Natural killer cells: integrating diversity with function

Abstract

The key role of natural killer cells in many aspects of the immune response is now being recognized. The last decade has seen an exponential increase in our understanding of the workings of these cells. Receptor diversity is crucial in allowing natural killer cells to respond effectively to a variety of different pathogens. This article reviews aspects of natural killer cell diversity that combine to generate populations of functional natural killer cells that exist within both the individual and throughout the population at large.

Introduction

Natural killer (NK) cells were first described in 1975 as a lymphocyte subset capable of cytotoxicity against leukaemia cells in vitro without previous sensitization.¹ Mature CD3⁻ CD56⁺ NK cells represent 5–15% of circulating lymphocytes, but make up a greater proportion of lymphocytes in some tissues such as the liver and pregnant uterus, where they may have specialized functions. They are distinct from T and B lymphocytes in that they do not show germ-line receptor rearrangement and their effector functions are governed by combinations of activating and inhibitory receptors. These allow the recognition of altered-self at a cellular, rather than the molecular level of T-cell receptor and immunoglobulin. Upon activation, NK cells can mediate direct cytotoxicity via perforin, or secrete cytokines predominantly of the T helper type 1 (Th1) type. Hence, they can have both a direct antiviral effect and can shape downstream immune responses. They have no requirement for priming so they confer early protection against tumour transformation and intracellular pathogens. However, such activity must be strictly regulated to prevent autoreactivity against the host. The last two decades have seen significant advances in the understanding of NK cell regulation. A multitude of activating and inhibitory receptors have been discovered and it is believed that the net balance of signals from these receptors enables NK cells to provide effective killing of target cells while maintaining self tolerance. Receptors that structurally belong to the same family may have activating or inhibitory effects. While some key receptors are shared between man and mice, others are distinct indicating that this is a system that has been subject to recent evolution. This review will concentrate on the structural and functional diversity of human receptors that regulate NK cell activity, while cross-referencing key lessons in NK cell biology from the murine system.

Diversity in NK cell receptors

Inhibitory

The critical role of inhibition as a means of regulating NK cell function was determined following a series of cytotoxicity experiments using major histocompatibility complex (MHC) class I-deficient targets. The results of these were synthesized in the landmark ‘missing-self’ hypothesis.² MHC class I molecules are key regulators of NK cell activity that are expressed ubiquitously on healthy cells and provide NK cells with a means of identifying ‘self’. Down-regulation of MHC class I or loss of its expression during viral infection or carcinogenesis releases the inhibitory signal to NK cells and permits their activation. Analysis of human NK cell clones originally demonstrated that inhibitory receptors specific for autologous MHC class I molecules are expressed by almost all mature NK cells.^3,4 This model has been refined as recent work has indicated that approximately 9% of circulating NK cells do not express an inhibitory receptor for MHC class I.⁵ In addition to controlling NK cell activity, these receptors have also been important for the maturation of NK cells to a fully functional state.⁶

Several inhibitory receptors for MHC class I have now been identified. The killer cell immunoglobulin-like receptor (KIR) family and the lectin-like CD94/NKG2A receptor play the most prominent role in NK regulation by MHC class I. In addition to these key receptors, NK cells can also be inhibited by members of the leucocyte immunoglobulin-like receptor/immunoglobulin-like transcript (LIR/ILT) family, specifically, LILRB1 (LIR-1/ILT2), which binds a broad range of MHC class I allotypes,⁷ LAIR-1, Siglec 7 and KLRG1/MAFA.^8–10 In the unusual situations of X-linked lymphoproliferative disease and TAP deficiency, NK cells can be inhibited by CD244 (2B4) and the homotypic interaction of carcinoembryonic antigen-related cell adhesion molecule (CEACAM), respectively (Table 1).^11,12

Table1.

Inhibitory receptors and their ligands

	Receptor	Ligand
KIR family	2DL1	Group 2 HLA-C
	2DL2/3	Group 1 HLA-C
	2DL5	Unknown
	3DL1	Bw4+ HLA-B
	3DL2	HLA-A3/A11
C-type lectin-like receptors	CD94:NKG2A	HLA-E
	NKR-P1A	LLT1
LIR/ILT family	LILRB1/ILT2/LIR1	HLA-A, -B, -C
Others	LAIR1	Collagen
	Siglec-7	Sialic acid
	KLRG-1/MAFA	Cadherins
	CEACAM1	CEACAM1

HLA, human leucocyte antigen; KIR, killer cell immunoglobulin-like receptor; LIR/ILT, leucocyte immunoglobulin-like receptor/immunoglobulin-like transcript.

Both KIR and CD94/NKG2A synergize to generate NK cells that are responsive to changes in MHC class I expression; however, they are derived from distinct gene families. The KIR are type 1 transmembrane glycoproteins belonging to the immunoglobulin superfamily. These highly polymorphic genes are encoded within the leucocyte receptor complex on chromosome 19q13.4, and segregate independently from genes of their ligand, classical MHC class Ia molecules (human leucocyte antigens (HLA)-A, -B and -C) located on chromosome 6. The KIR show extensive genetic, expression and functional diversity that can impact NK cells at many different levels (Fig. 1).

Figure 1.

Killer cell immunoglobulin-like receptor (KIR) diversity can impact natural killer (NK) cells at different levels. Schematic diagram of the impact of the areas in which KIR diversity may influence NK cell function. This can occur as the result of differences in gene content and allelic diversity at the KIR locus. This may result in different individuals having different numbers of activating and inhibitory KIR. These KIR are expressed stochastically on NK cells to generate a ‘KIR’ repertoire, which differs among different individuals. The presence or absence of human leucocyte antigen (HLA) class I ligands for these KIR may further impact on this repertoire, and on the functionality of these KIR-expressing subpopulations in different individuals.

There are 14 different expressed KIR genes encoding both inhibitory and activating receptors. There is substantial diversity of KIR at both locus and allelic levels.^13,14 Furthermore, the KIR are clonally expressed in a stochastic fashion to generate a repertoire of NK cells expressing different combinations of KIR in the same individual.^3,15 The KIR can be subcategorized by the number of extracellular immunoglobulin domains, either two or three (‘2D’ or ‘3D’), and by the length of their cytoplasmic tails, short or long (‘S’ or ‘L’). In general, long cytoplasmic tails are associated with specific inhibitory motifs (immunoreceptor tyrosine-based inhibitory motif; ITIM). Receptor nomenclature would therefore imply that KIR2DL is composed of two immunoglobulin domains (2D), with a long (L) intracytoplasmic tail and an inhibitory function. The inhibitory KIR and their ligands are listed in Table 1. Disease-association studies have shown that KIR receptor–ligand interactions can influence the outcome of viral infections such as human immunodeficiency virus (HIV) and hepatitis C virus,^16–18 pregnancy and autoimmune diseases.^19–24

There are six members of the type 2 transmembrane glycoprotein C-type lectin-like receptor family. Their genes are located on the NK cell receptor complex (chromosome 12p13). Each heterodimeric receptor consists of an invariant CD94 molecule covalently associated with a molecule of the NKG2 group. NKG2A and NKG2B (a splice variant of NKG2A) are inhibitory whereas NKG2C, -D, -E and -F are activating. Consequently, CD94/NKG2A is a heterodimeric inhibitory receptor consisting of two C-type lectin subunits, CD94 (30 000 MW) and NKG2A (43 000 MW), linked by disulphide bonds. The ligand for CD94/NKG2A is the non-classical MHC class Ib molecule HLA-E (Qa1-b in mice). It has a widespread tissue distribution but requires specific nonamer peptides derived from the leader sequence (aa 3–11) of classical MHC class Ia molecules (HLA-A, -B and -C) and HLA-G to stabilize expression.²⁵ Unlike HLA-A, -B and -C, HLA-E is relatively non-polymorphic, as is its receptor CD94/NKG2A. Overall the binding of CD94/NKG2A to HLA-E is dominated by the CD94 moiety.²⁶

Activating

In comparison to the inhibitory receptors there are many more activating receptors expressed by each NK cell. It is now thought that when a critical threshold of activating signalling exceeds the counterbalancing influence of the inhibitory receptors, NK cells will mount an effector response.²⁷ In contrast to inhibitory receptors, most activating receptors are expressed on most NK cells. Unlike their inhibitory counterparts it appears that the activating KIRs (KIR2DS and KIR3DS) represent a relatively minor component of the activating NK cell repertoire. Instead there are a number of different receptors that transduce activating signals through adaptor molecules. These receptors are listed in Table 2, and include members of both the immunoglobulin superfamily and the C-type lectin-like family of receptors. Key activating receptors include the natural cytotoxicity receptors, NKG2D, the low-affinity immunoglobulin G receptor CD16 and CD244, which may have both activating and inhibitory functions.¹²

Table 2.

Activating receptors and their ligands

	Receptor	Ligand
Natural cytotoxicity receptors	NKp30	BAT-3
	NKp44	Viral haemagglutinin
	NKp46	Viral haemagglutinin
C-type lectin-like receptors	CD94:NKG2C	HLA-E
	CD94:NKG2E	HLA-E
	NKG2D	MIC-A/B, ULBPs
KIR family	2DS1	Group 2 HLA-C
	2DS2	Group 1 HLA-C
	3DS1	Bw4+ HLA-B?
	2DS3	Unknown
	2DS4	HLA-Cw4
	2DS5	Unknown
	2DL4	HLA-G
Others	CD244 (2B4)	CD48
	CD16	IgG
	CD226 (DNAM-1)	CD112, CD155
	CRACC	CRACC
	NTB-A	NTB-A

HLA, human leucocyte antigen; KIR, killer cell immunoglobulin-like receptor.

Natural cytotoxicity receptors are specific to NK cells. There are three members that are named according to their molecular weight NKp46 (46 000 MW), NKp30 (30 000 MW) and NKp44 (44 000 MW). The cellular ligands for NCRs have yet to be defined but there is evidence that they recognize viral haemagglutinins.²⁸ The ligands of natural cytotoxicity receptors are also thought to be expressed on healthy and diseased cells. NKp46 may be expressed on almost all NK cells.²⁹ Unlike NKp46 and NKp30, which are expressed on resting NK cells, NKp44 is only expressed on activated NK cells.³⁰ Recently, it has been shown that NKp44 marks a unique subset of interleukin-22 (IL-22)-secreting NK cells in the gut-associated lymphoid tissue.³¹

NKG2D belongs to the C-type lectin-like family of receptors and is expressed on both NK cells and T cells.³² Unlike other NKG2 family members it does not associate with CD94 but instead forms a homodimeric receptor.³³ In addition it recognizes the MHC class I-like stress-induced self ligands (MICA/B),³⁴ UL16 binding proteins (ULBP) 1–5 in humans³⁵ and H60/RAE-1/MULTI-1 family members in mice.³⁶ These ligands have restricted expression on healthy cells but are up-regulated by cellular stress, tumour transformation and viral infection. MICA/B and ULBP molecules share structural similarities to MHC class I molecules. In particular, MICA/B alleles are highly polymorphic and may vary in their affinity for NKG2D.

NKG2C/E form heterodimers with CD94 that also bind HLA-E.^25,37 Interestingly CD94/NKG2C has a sixfold lower affinity for HLA-E than both CD94/NKG2A and CD94/NKG2E.^37,38 The functional significance of this is not clear. However, HLA-E loaded with an HLA-G-derived nonamer triggers NKG2C activation very efficiently. Furthermore, the low affinity of CD94/NKG2C for its ligands parallels that of the activating KIR, which also appear to bind their HLA class I ligands with low affinity.

NK cell signalling pathways

Activating

Although there is a plethora of activating receptors, several of them share signalling pathways through their association with common transmembrane adaptor proteins such as CD3ξ, FCRγ and DAP12 (Fig. 2). These adaptor proteins contain cytoplasmic tails with immunoreceptor tyrosine-based activation motifs (ITAMs). This motif contains a tyrosine separated from a leucine by any two other amino acids, giving the signature YxxL. Two of these signatures are typically separated by between seven and twelve amino acids in the tail of the molecule [YxxLx(7–12)YxxL]. Receptor–ligand binding leads to phosphorylation of tyrosine residues within these motifs. This in turn leads to recruitment of protein tyrosine kinases of the Syk family, such as Syk and ZAP70, and subsequent initiation of downstream signalling pathways through Vav, Rac, PAK1, MEK and ERK.³⁹

Figure 2.

Natural killer (NK) cells are diverse at the level of receptors, ligands and signalling adaptors. Examples of the diverse array of molecules that generate diversity in the control of NK cell function are shown. This can occur at the level of the receptor, the ligand or the adaptor molecule. The signalling motifs that control these functions are illustrated: ITIM, immunoreceptor tyrosine-based inhibitory motif; ITAM, immunoreceptor tyrosine-based activating motif; ITSM, immunoreceptor tyrosine-based switch motif and the YINM motif of DAP10.

In contrast, NKG2D associates with a different transmembrane adaptor molecule, DAP10, which lacks ITAMs. Instead its intracellular signalling domain contains a YINM motif, which may lead to the associations of DAP10 with phosphatidylinositol 3 kinase or the adaptor molecule grb2 as opposed to Syk or ZAP-70.⁴⁰ Despite these differences, ultimately similar processes of cytoskeletal reorganization to form an activating synapse and cytotoxicity may be initiated. However, it has yet to be shown that DAP10 signalling initiates cytokine secretion.²⁷

Inhibitory

Although the extracellular domains of NK cell inhibitory receptors are diverse, the intracytoplasmic signalling motifs of these transmembrane receptors are remarkably similar. The cytoplasmic tails have a conserved sequence of amino acids (S/I/V/LxYxxI/V/L), known as ITIMs. Binding of ligands to inhibitory receptors activates Src family kinases that phosphorylate tyrosine residues, leading to the recruitment of other enzymes such as the tyrosine phosphatases SHP-1 and SHP-2, or the inositol-phosphatase known as SHIP.⁴¹ These tyrosine phosphatases are able to dephosphorylate protein substrates of tyrosine kinase linked to activating NK cell receptors.⁴² This would imply that there is coaggregation of activating and inhibitory receptors.^43,44 Events are transient and spatially localized such that if the same NK cell encounters a target cell lacking inhibitory ligands, negative signalling pathways can be overcome leading to NK activation.²⁷

Pathways to activation

Activation and coactivation

Signals derived from activating and inhibitory receptors are integrated by the NK cell to ultimately determine their functional outcome. Critically, NK cells are under a predominant inhibitory control and this maintains self-tolerance. However, this self-tolerant state can be overcome by receptor-mediated or cytokine stimulation.

For cytotoxicity of a target to occur a number of co-ordinated processes need to happen. These include cell adhesion, synapse formation, granule polarization and granule exocytosis. Recent work has highlighted the complexity of this process and identified some of the key molecular players. Furthermore, it has become clear that the broad term ‘activating receptor’ is not sufficient to describe the cell surface molecules that initiate these processes. In experiments in which individual receptor–ligand pairings are isolated, only the low-affinity immunoglobulin G receptor CD16 is sufficient to induce granule exocytosis and tumour necrosis factor-α secretion in unstimulated NK cells.^45,46 In reverse antibody-dependent cell-mediated cytotoxicity experiments using unstimulated NK cells and the P815 mastocytoma cell line, cross-linking NKp46, NKG2D, 2B4, CD2 or DNAM in isolation was insufficient to mediate cytotoxicity or cytokine secretion. However, specific combinations of receptors were able to synergize to induce NK cell effector functions. Conversely, IL-2-activated NK cells can be triggered by the isolated ligation of NKp46, NKG2D, 2B4 or DNAM.⁴⁶ Consequently, in vivo signals from these receptors synergize to induce NK cell effector function and they have therefore been termed ‘coactivating’ receptors. The relative contributions of these signals to activate an NK cell may vary according to the target cell involved, and it is clear that there is interplay between cytokine- and receptor-mediated signalling. For instance the adaptor molecule DAP10, which transduces activating signals from NKG2D, is associated with the IL-15 receptor β- and γ-chains⁴⁷ and this adaptor molecule can be phosphorylated by the IL-15-receptor-associated Jak-3 kinase. This system therefore provides a proximal molecular link between cytokine- and receptor-mediated NK cell activation, which may extend to other cell surface activating receptors.

The requirement for the synergy of activating signals to generate efficient NK cell responses indicates that discrete activating receptor–ligand stimuli may be required for full lysis of target cells. This has been shown by the co-operation of natural cytotoxicity receptors to maximally lyse tumour targets.⁴⁸ Additionally, disruption of adhesion is sufficient to abrogate the lysis of virus-infected targets.⁴⁹ The NK cells require the up-regulation of multiple ligands on potential target cells to achieve recognition and activation in vivo.

Ligand up-regulation

Understanding the interactions of NK cells with target cells has been hampered by a lack of understanding of the ligands for the activating receptors. In particular, self-ligands expressed on the cell surface for the natural cytotoxicity receptors have yet to be isolated. Haemagglutinin is a viral ligand for both NKp46 and NKp44, and BAT3 is a potential soluble ligand for NKp30, but these molecules do account for the participation of these receptors in tumour targets.^28,50 In contrast to these receptors, the ligands for NKGD are inducible by cellular stress, infection and DNA damage.⁵¹ Up-regulation of these activating ligands may tip the balance of the NK cell from inhibition to activation and so result in cytotoxicity or cytokine secretion. This has been termed ‘induced-self’ recognition, to distinguish it from ‘missing-self’ recognition, in which the primary change is a down-regulation of ligands for inhibitory receptors.⁵² Additionally, the balance of the cell surface complement of the NK cell may be altered by up-regulation of an inducible activating receptor such as that of NKp44 by IL-2.

Diversity in inhibitory signals

Tuning of NK cell responses by multiple activating receptor–ligand interaction is one mechanism that prevents inappropriate self-recognition. However, inhibitory receptor–ligand pairings are also important for NK cell self-tolerance. Indeed, NK cells appear to be finely tuned by this balance of activating and inhibitory receptor expression. Stochastic expression of inhibitory receptors for MHC class I on NK cells within a single individual generates a repertoire of NK cells that express discrete inhibitory receptor combinations. To release an NK cell from inhibition, down-regulation of its HLA ligand is required. This can be achieved globally by viral infection or during the course of tumorigenesis. However, as these inhibitory receptors have discrete HLA ligands they may also be subject to discrete down-regulatory processes. For instance the HIV protein nef selectively targets HLA-A and HLA-B, while leaving HLA-C relatively intact.⁵³ As a result, the NK cells expressing KIR3DL1 are more likely to be released from inhibition by HIV-infected targets and hence become activated, than are NK cells that express other inhibitory receptors, or even NK cells that express KIR3DL1 in combination with an inhibitory receptor for HLA-C. Additionally, in vaccinia virus infection HLA-E appears to be preferentially down-regulated compared with classical class HLA-A, -B or -C molecules,⁵⁴ rendering these infected cells susceptible to lysis by NKG2A-positive NK cell clones. These data suggest that biochemical differences between the different HLA molecules may define patterns of down-regulation and hence potential for NK cell recognition.

In addition to HLA diversity, there is also diversity in the interactions of KIR with its HLA ligands. In hepatitis C virus infection KIR2DL3 and its cognate ligand group 1 HLA-C is protective against chronic infection, but its allele KIR2DL2 is not protective. In vitro KIR2DL2 has a greater avidity for group 1 HLA-C than KIR2DL3.⁵⁵

There is also greater complexity at the HLA-B locus for the Bw4-positive allotypes and KIR3DL1. Different KIR3DL1 alleles are expressed on the cell surface at different levels and individual KIR3DL1 allotypes bind to different HLA-B allotypes with different avidities. The clinical correlate of this observation is that in HIV infection KIR3DL1 alleles that are expressed at high levels are the most protective.¹⁸

The effect of different inhibitory KIR–HLA combinations on NK cells could be related directly to release from inhibition of an NK cell on encountering its target. However recent work has shown that NK cells acquire functional maturity or are ‘licensed’ on the basis of expression of inhibitory receptors for cognate MHC class I.⁶ This is a complex and poorly understood process but implies that the interaction of inhibitory KIR with its MHC class I ligand could further define the ability of an NK cell to respond to an activating signal irrespective of the presence or absence of MHC class I on the target cell. This effect was originally demonstrated in vitro using antibody cross-linking of the NK1.1 activating receptor on murine NK cells and has subsequently been shown to be valid in humans.^6,56,57 Further subtleties to this system exist in that NKG2A provides a relatively constant level of functional activity in different donors, but because of the KIR system, can vary depending both on the KIR gene and the HLA class I type of the individual studied.^5,57 For instance, KIR2DL3-positive NK cells from donors with an HLA-Cw*07 allele are more responsive to a class I negative cell line than are KIR2DL3-positive NK cells from individuals expressing other KIR2DL3 binding allotypes. Additionally, KIR3DL2 and LILRB1 are seen to confer no discernable positive functional reactivity on NK cells as determined by these assays.

Linking inhibition and activation

In addition to the diversity generated by the cell surface receptors of NK cells, further diversity is present at the level of signalling (Fig. 2). The contribution to signalling of the binding of KIR to its ligand is important because the clustering of KIR at the inhibitory synapse is independent of ATP and cytoskeletal reorganization. Once a ligand is engaged then inhibitory signalling involves recruitment of the protein tyrosine phosphatase SHP1 to the ITIMs of the KIR. This process is dependent on the ITIMs being phosphorylated by an unidentified mechanism and is also dependent on the multifunctional signalling molecule β-arrestin 2.⁵⁸ The SHP1 dephosphorylates Vav1, which inhibits its guanine exchange factor activity.⁵⁹ This in turn prevents actin polymerization and thence activating receptor clustering and recruitment into lipid rafts. Therefore inhibition is a relatively proximal event and is ideally positioned to prevent NK cell activation. Alternatively it can be envisaged that multiple activating receptor interactions can overcome this inhibitory signalling by favouring recruitment and phosphorylation of Vav1, leading to actin polymerization and formation of an activating NK cell synapse. Additionally, 2B4 and CD2 accumulate within inhibitory synapses and so signalling by this receptor–ligand pairing may be prevented at a stage before cytoskeletal repolarization.⁶⁰

As NK cells integrate the signals derived from both activating and inhibitory receptors this can be modulated by the receptor complement expressed by NK cells. In particular, combinatorial expression of activating and inhibitory receptors can lead to subpopulations of NK cells that can be activated in different situations dependent on the up-regulation of ligands for specific NK cell receptors and the down-regulation of HLA class I receptors for inhibitory receptors. At a molecular level the KIR–HLA system provides a mechanism for fine-tuning NK cell reactivity based on allelic diversity. This may have an evolutionary advantage at a population level in that different individuals may be able to respond better to specific viral infections. That NK cell receptors can exert significant selection pressure on viruses has been elegantly demonstrated for murine cytomegalovirus infection.⁶¹ In the human system, pressures as diverse as viral infections and pregnancy may be acting on KIR to maintain a balance between KIR haplotypes with either few or many activating receptors.⁶² This population diversity can be followed down to the molecular level because of the combinatorial expression of activating and inhibitory receptors and the ability of KIR–HLA interactions to fine-tune NK cell reactivity (Fig. 2).

The challenge for NK cells is that of self: non-self discrimination in the absence of recombining germ-line genes. They have risen to this by generating diversity at the population, receptor, ligand, expression and signalling levels. The challenge for immunologists is to unravel this complexity and to harness these features to benefit human health.

Disclosures

None.