Formation and function of the lytic NK-cell immunological synapse

Preface

The natural killer (NK)-cell immunological synapse is the dynamic interface formed between an NK cell and its target cell. Formation of the NK-cell immunological synapse involves several distinct stages, beginning with the initiation of contact with a target cell and culminating in the directed delivery of lytic granule contents to lyse the target cell. Progression through the individual stages is methodical and underlies the precision with which NK cells select and kill susceptible target cells (including virally infected cells and cancerous cells) that they encounter during their routine surveillance of the body.

Introduction

Natural killer (NK) cells are lymphocytes of the innate immune system that provide effector functions after the ligation of germline-encoded receptors. In humans, NK cells are poised to deliver a response immediately after recognizing specific signals of stress, ‘danger’ or foreign origin. They were originally described as having rapid cell-mediated cytotoxicity. However, they are now also known to promote slower receptor-mediated apoptosis, provide contact-dependent costimulation and efficiently produce soluble mediators including cytokines. NK cells therefore participate in the defence against infections, the regulation of immune responses and the surveillance of stressed or cancerous cells (reviewed in¹). Although these functions are not unique to NK cells, they use NK-cell specific receptor systems, and the ability to mediate effector functions rapidly without the need to develop further is a key distinguishing feature of mature NK cells and cytotoxic T lymphocytes (CTLs), both of which are efficient at mediating cytotoxicity.

NK-cell cytotoxicity involves the secretion of cytolytic effector molecules from specialized organelles known as lytic granules. Much of what is known about this process is derived from CTLs due to an abundance of studies and derivative concepts that have often served as starting points for investigations of NK cells. NK cells, however, have important distinctions and thus the process in the two cell types should not be assumed identical. The lytic granules are preformed in resting human NK cells but not in CTLs. As a result, the cytolytic process in NK cells needs to be well regulated, and this may involve additional or enhanced mechanisms for controlling secretion of lytic granule contents in NK cells compared with CTLs. Indeed, direct comparisons between NK cells and CTLs have identified important kinetic and mechanistic differences².

The induction of many NK-cell effector functions, including cytotoxicity, requires that the NK cell contact its target cell. This ensures precise targeting of the cytolytic process to a single diseased cell within a tissue without affecting its neighbouring cells. Events occurring on interaction between a cytolytic cell and its target cell have been well studied. They include the delivery of cytolytic effector molecules to and the secretion of these molecules at the interface formed between the cytotoxic cell and its target cell through a process known as directed secretion.

Our understanding of directed secretion for cytotoxicity has been advanced by the concept of the immunological synapse. The immunological synapse was originally defined in the late 1990s^{3, 4} as the crucial junction between a T cell and an antigen-presenting cell (APC) at which T-cell receptors (TCRs) interact with MHC molecules. Subsequent studies extended these observations and identified relevant immunological synapses between different types of immune cells, as well as between immune cells and non-immune cells. Thus, an immunological synapse can be defined as the intentional arrangement of molecules in an immune cell at the interface with another cell. Numerous molecules have been identified as participating in the immunological synapse, including receptors, signalling molecules, cytoskeletal elements and cellular organelles. Some studies suggest that these molecules accumulate in distinct regions within an activating immunological synapse to form a supramolecular activation cluster (SMAC), which may be segregated into peripheral (pSMAC) and central (cSMAC) zones. However, the purpose of this molecular patterning is debated as function typically attributed to the SMAC can be induced in its absence under some circumstances⁵. Furthermore, single molecule and kinetic studies of the immunological synapse have shown that, in some settings, the engagement of individual receptors⁶ or the involvement of microclusters of cell-surface and signalling molecules can support cell activation without the need for a mature synapse^{7, 8}. Therefore, the importance of larger scale accumulations and distinct segregation of molecules at the synapse is currently being re-evaluated.

Although debate regarding the role of the immunological synapse in enabling immune responses continues, several of its potential functions seem relevant and worth considering (BOX 1). Many of these were originally assigned to the immunological synapse between a T helper cell and APC, but they also apply to the NK-cell synapse. Certain aspects, however, are more relevant to NK-cells, such as directed secretion of lytic granules for cytotoxicity.

Lytic granules are hybrid organelles that are specialized for secretion of the lytic effector molecules that reside within them (BOX 2). For an NK cell to mediate cytotoxicity, lytic granules are emptied onto a target cell at a prototypical mature lytic synapse (Figure 1). In the mature synapse, filamentous (F)-actin and adhesion receptors accumulate and are thought to form a ring in the pSMAC through which perforin and other lytic granule contents are secreted. Domains within the lytic synapse containing specific signalling molecules or secretory machinery have been described in CTLs⁹, but may not always be required for CTL-mediated cytotoxicity⁶. Similar molecule distribution patterns have been observed in NK cells, but can be inconsistent^10–12 and many aspects of NK cell synapse organization have not been elucidated. Nevertheless, LGs are large organelles and must traverse dense F-actin networks at the synapse, and actin reorganization is required for their release. Synapse formation, therefore, has a specific function in NK cells to enable cytotoxic function.

Figure 1. The prototypical mature NK-cell lytic synapse.

The mature natural killer (NK)-cell lytic synapse is defined by the formation of a supramolecular activation cluster (SMAC) at the interface between the NK-cell and target cell to which lytic granules polarize. The prototypical version of this synapse contains a central SMAC (cSMAC) that includes a secretory domain through which lytic granules may traverse. The series of images show a YTS human NK cell expressing a CD2-GFP (green fluorescent protein) fusion protein making contact with an EBV-transfomed B-cell line. The cell-cell conjugates were fixed, permeabilized and evaluated for the presence of perforin (red) using a monoclonal antibody. Perforin is contained within lytic granues and is presented as a surrogate for them, while CD2 patterning under normal conditions parallels F-actin at the mature synapse.¹² Serial images were obtained using a confocal microscope through the z-axis and the three-dimensional cell volume reconstructed in silico. The individual images on the left from top to bottom demonstrate a 180-degree rotation around the z-axis (Scale bar=5μM) The schematic on the right displays the position of the NK cell (yellow) and B cell (blue), CD2 at the SMAC (green) and perforin at the SMAC (red) in each image. Also see an interactive version in online Supplementary Figure 1.

In this Review, the specific steps leading to the formation and function of the NK-cell lytic synapse are considered. Recent evidence has defined a sequence of events that has some surprising linearity and a series of important checkpoints. Appreciation of these will help to better define and control NK-cell cytotoxic ability.

Stages of the NK-cell lytic synapse

The formation of a mature and functional NK-cell lytic synapse can be divided into a series of discrete stages — the recognition and initiation stage, the effector stage and the termination stage (Figure 2). Each of the stages can be further subdivided into multiple steps. Some of these have been defined in a linear sequence with clear prerequisites and thus occur in series and not in parallel. Together, these processes enable the delivery of lytic granules to the synapse followed by their close association with the NK-cell membrane to which they can fuse and release their contents onto the target cell. Because lytic granules exist in resting NK cells before activation, each stage must be controlled to prevent accidental release of cytotoxic mediators and to enable rapid directed secretion at the appropriate moment.

Figure 2. Model for sequential stages in the formation and function of the NK-cell lytic synapse.

It is proposed that the formation of a functional natural killer (NK)-cell lytic synapse can be divided into three main stages — recognition, effector and termination stages — that are each subdivided into multiple steps. Important steps proposed to occur in the recognition stage include adhesion and initial activation signaling. In the effector stage key steps include actin reorganization, receptor clustering, MTOC and lytic granule polarization, and lytic granule fusion with the plasma membrane. In the termination stage crucial steps are proposed to include a period of inactivity and detachment. The specific time required to progress through the various stages varies and is likely to be a feature of the given target cell as well as the activation state of the NK cell. The linearity of connections between certain steps is supported by experimental data as outlined in the text, such as a requirement for actin reorganization for MTOC polarization. In others, however, linearity is inferred and is presented as a proposed model. The inhibitory synapse (Box 3) has been shown to halt progression of the lytic synapse by interfering with the late recognition stage and early effector stage – specifically activation signaling, actin reorganization and receptor clustering steps.

Recognition and initiation stages

The initial stage in the formation of a lytic synapse includes establishing a close association between the NK cell and target cell, initial signaling and adherence of the NK cell to its target cell (Figure 2). These steps are in large part theoretical, but are suggested by certain experimental evidence. In the first step, the NK cell either intentionally or accidentally encounters its potential target cell. This can be in response to chemotactic signals, enabling their localization within the organism at sites where NK cell effector functions are required (reviewed elsewhere¹³). The first true contact between an NK cell and its target is a cellular association akin to tethering. For this, the NK cell localizes to the site of, and would establish interactions with the target cell. Although studies of the NK-cell synapse have not firmly established the exact molecules, they may include the selectin family members, as has been shown in in vivo studies of NK cell localization¹⁴. They may also include the recognition of sialyl Lewis X by CD2 expressed by NK cells¹⁵, as CD2 has been shown to accumulate rapidly at the developing NK-cell synapse¹². Identification of the molecules involved in initial contact between an NK cell and its target cell is challenging, as in vitro approaches used to study the immunological synapse may reduce the requirements for this particular step. The early interactions, however, could then result in longer lasting association leading to an initial adhesion. These events are likely to contribute to NK cell activation as receptors potentially engaged at this point may participate in activation signaling, such as CD2¹⁶.

A next step in the formation of the lytic synapse is firm adhesion, which is facilitated by receptor–ligand interactions of higher affinity. The integrin family of adhesion molecules is important, and in particular the integrins lymphocyte function-associated antigen 1 (LFA1; CD11a/CD18) and MAC1 (CD11b/CD18) expressed by NK cells have been well studied. These integrins cluster rapidly at the NK-cell synapse following initiation^{12, 17, 18}, but they probably function in adhesion prior to their rearrangement to the synapse and participate in signaling¹⁹, even in resting NK cells²⁰). Integrin signalling can fully activate some NK cells²⁰ and partially activate others²¹. As early events, however, integrin signalsmay be less likely to commit a NK cell to cytotoxicity although still physiologically relevant in promoting maturation of the synapse. Owing to the diffuse cell-surface distribution of more-potent NK-cell activating receptors such as the natural cytotoxicity receptor family, under physiological conditions, subsets of these receptors are also likely to be engaged at this stage by ligands on the target cell. These interactions provide additional signals that complement early NK-cell activation. The function of these activating receptors however is probably distinct from that of TCRs at the CTL immunological synapse. In favour of this, TCRs couple with ten immunoreceptor tyrosine-based activation motifs (ITAMs), whereas NK-cell activating receptors couple with less. So, the ligation of individual TCRs could theoretically provide more potent signals for synapse formation and function compared with NK-cell activating receptors. This may contribute to the observation of rapid minimal triggering of TCRs leading to CTL activation⁶, and may not apply to NK cells.

Importantly, these initial steps in the formation of the NK-cell lytic synapse probably occur before molecular patterning or polarization is evident and are completed quickly. Whether the NK cell progresses to molecular reorganization at the synapse seems to depend on the level of signals through inhibitory receptors, such as KIRs (killer-cell immunoglobulin-like receptors), which can establish a so-called inhibitory synapse (Box 3). Such regulation ensures that NK cells effectively carry out their surveillance function, by leaving most cells undisturbed while being poised to destroy those that are diseased. The NK-cell inhibitory synapse is especially elegant in that it directly interferes with the ability of the lytic synapse to progress past the initiation stage^22–24.

Effector stages

After target-cell recognition and synapse initiation has occurred, and in the absence of overriding inhibitory signals, reorganization of the immunological synapse can proceed. The key steps of the effector stage can allow: formation of a stable NK-cell–target-cell interface that has a ‘cleft’ into which cytolytic molecules are secreted; recruitment of lytic granules from throughout the cell to the synapse; clearance of a conduit in the NK-cell cortex through which lytic granules could be directed to the cell membrane; and fusion of the lytic-granule membrane with plasma membrane for release of lytic-granule contents. The exact sequence of events is debatable, but several observations support some linearity.

An initial stage in the commitment to lytic-synapse formation is actin reorganization. This involves the formation of F-actin networks from the cellular pool of monomeric G-actin — a step that was first observed in polarized NK cells in 1983 using fluorescently labelled phalloidin (which binds junctions between approximated actin subunits)²⁵. F-actin reorganization at the NK-cell synapse occurs downstream of activating-receptor-induced VAV1 activity^{19, 26} and depends on Wiskott-Aldrich syndrome protein (WASP)²⁷ a known part of the protein unit that promotes F-actin branching. Accordingly, in the absence of WASP, or in the presence of actin inhibitors, F-actin accumulation at the synapse and NK-cell cytotoxicity are reduced^{12, 27–29}. The WASP-dependent reorganization of F-actin at this stage is also important for characteristic changes in NK-cell shape that occur soon after synapse formation^{2,12, 27, 29} (J.S.O., unpublished observations).

The events occurring at the same time as actin reorganization have not yet been fully elucidated, but they include receptor clustering, lipid-raft aggregation, further activation signalling and lytic-granule redistribution. Although many receptors might be present within the synapse at the initial stages, owing to their diffuse distribution in the plasma membrane, some accumulate rapidly after cell conjugation, including those that are important in both adhesion and triggering of cytotoxicity. Of these, CD11a, CD11b and CD2 do not seem to cluster in the absence of actin reorganization¹² thus providing some evidence for linearity in synapse formation. The mechanisms underlying actin-dependent receptor clustering in NK cells are unknown, but the ezrin, radixin and moesin (ERM) family of proteins are present in the NK-cell lytic synapse^{23, 30, 31} and are known to function in lateral receptor motility in other cell types^32–35. A related issue is that of lipid rafts. Whether these membrane domains represent discrete entities or more fluid collections of molecules, they have been shown to aggregate at the NK-cell lytic synapse (reviewed in³⁶) and their integrity is likely required for cytotoxicity as demonstrated using the somewhat nonspecific cholesterol-sequestering drug β-methylcyclcodextrin³⁷. The existence and function of lipid rafts, unfortunately, has been defined using indirect methods such as cell membrane cholera toxin binding, detergent resistant biochemical fractionation and cholesterol sequestration and thus remains controversial. Similar to receptor clustering, however, lipid-raft aggregation at the NK-cell synapse depends on actin polymerization²³ additionally supporting a linear progression through the early effector steps of synapse formation.

It has been shown that receptor clustering at the NK-cell lytic synapse is important for the generation of robust signalling in live NK cells³⁸. In T cells, TCR signalling stems from microclusters of TCR molecules that move from the pSMAC into the cSMAC as activation proceeds³⁹. Although functional microclusters have been identified in NK cells, these have only been investigated at the supramolecular inhibitory cluster (SMIC)⁴⁰. It is unclear whether this level of signaling organization of the pSMAC and cSMAC identified in T cells can be similarly applied to NK cells.

Another requirement for effector function of the NK-cell lytic synapse is the polarization of lytic granules to the synapse. This begins with movement of the granules along the microtubules to the microtubule-organizing centre (MTOC). As the minus ends of microtubules are present at the MTOC, lytic granules require a minus-ended directed motor such as dynein to move along microtubules. The exact motor used, however, has not been defined. While lytic granules aggregate around the MTOC, the MTOC also begins to polarize towards the immunological synapse. Signals required for MTOC polarization include ERK (extracellular-signal-regulated kinase) phosphorylation, VAV1 activation and PYK2 (protein tyrosine kinase 2) activity in NK cells^{26, 41, 42}. In T cells MTOC polarization requires CDC42 (cell-division cycle 42)⁴³ and activation of a signalling platform comprising ZAP70 (ζ-chain-associated protein kinase of 70 kDa), SLP76 (SH2-domain-containing leukocyte protein of 76 kDa) and LAT (linker for activation of T cells)⁴⁴, as well as the function of the formins Diaphanous-1 and Formin-like-1⁴⁵. The molecular motors used for propulsion or force-generating steps required to pull the MTOC to the NK-cell synapse, however, have not been defined. The CDC42 interacting protein 4 (CIP4), which interacts with tubulin, CDC42 and WASP, localizes to the MTOC after immunological synapse formation and is required for complete MTOC polarization⁴⁶. Unlike CDC42⁴⁷ in T cells or VAV1 in NK cells²⁶, however, reducing CIP4 function does not impair actin reorganization in NK cells⁴⁶.

Although the mechanism underlying CIP4 function is not known, it may participate actively in MTOC motility or may help to anchor the MTOC at the NK-cell synapse by interacting with WASP or CDC42. Polarization of the MTOC in NK cells is also likely to depend on the appropriate insertion of the plus ends of microtubules into the accumulated F-actin at the synapse, which in T cells has been shown to be mediated through ADAP (adhesion- and degranulation-promoting adaptor protein)⁴⁸ and could function to apply force to the MTOC as the synapse reorganizes. In NK cells, the integrity and reorganization of the F-actin network is required for MTOC and lytic-granule polarization to the synapse, as blockade of actin function prevents such polarization^{2, 12, 26}, thus further defining a linear series of events in synapse formation. Requirements for cytokine secretion at the synapse, are less well established but have been shown in T cells to require WASP⁴⁹ and can occur as both directed and multi-directional⁵⁰. The cellular mechanisms underlying cytokine secretion, however, have important distinctions from secretion of lytic granule contents as the separation of the two is likely critical for enabling specificity in immune responses.

Although actin function is required for MTOC and lytic granule polarization, a discrete region of the F-actin network needs to be disassembled to create a conduit through which the granules gain access to the plasma membrane. Although the function of these conduits remains hypothetical, large channels of 1–4μ in diameter, are often observed in the cSMAC of a prototypical NK-cell lytic synapse^{12, 18, 51, 52}. Theoretically, these conduits, could also be smaller channels just large enough to allow the ~500nm diameter lytic granules through. The mechanism responsible for this targeted F-actin disassembly is unknown, but it seems to be independent of MTOC polarization, as actin clearance occurs normally in NK cells treated with microtubule depolymerizing agents¹². Thus this step occurs before MTOC polarization in the sequence leading to synapse maturation. Once the MTOC has become polarized to the synapse, there are several ways in which the lytic granules might proceed through the conduit in the cortical F-actin. Originally, lytic granules, once they had been delivered near to the synapse by the MTOC, were thought to traffic to the plasma membrane using plus-ended microtubule motors. Consistent with this, lytic granules have been shown in vitro to undergo plus-ended microtubule motility using kinesin⁵³. Recently however, the MTOC in CTLs was shown to associate closely with the synapse and directly deliver the lytic granules, thereby without the need for additional plus-ended motility of the granules along individual microtubules⁵⁴. The transit of lytic granules or even the MTOC together with the granules through clearances in the F-actin network in NK cells, however, may require additional motor functions. Myosin-II is a candidate protein for this function. NK cells in which myosin-II function is inhibited biochemically or its expression is downregulated by small interfering RNA form a mature lytic synapse with polarized MTOC and lytic granules, but do not degranulate⁵². Lytic granule approximation to the membrane after polarization is thus another important linear step in synapse maturation.

Once the lytic granules have traversed the F-actin and can interact with the plasma membrane, there are most likely several final steps in the effector stage of NK-cell lytic synapse formation. These include concepts defined in neurological synapses, and identified in CTLs, but in NK cells are predicted to inlcude the docking of lytic granules to the synapse and their priming for membrane fusion. After priming, the membrane of the lytic granules could fuse with the plasma membrane, causing the release of granule contents into the cleft formed between the two interacting cells. Although all of these events occur within the confined space of the synapse, individual steps can be visualized using electron microscopy⁵⁵ and fluorescent imaging techniques such as total-internal reflection fluorescence microscopy⁵⁶.

Although most of these steps remain hypothetical in NK cells insights from CTLs are instructive. Herelytic granule docking to the synapse requires members of the RAB family of small GTPases, which are important regulators of vesicle trafficking and compartmentalization (reviewed in⁵⁷). RAB proteins undergo post-translational prenylation to gain hydrophobicity, which allows them to link to a cargo to and facilitate interaction with a target membrane. RAB27a perfoms this function in docking lytic granules in CTLs⁵⁸. Furthermore, active RAB27a can interact with key effector molecules, which enable vesicle priming for fusion with the plasma membrane. Priming refers to the process of preparing the lytic granule for a functional ability to fuse with the inner leaflet of the NK cell membrane at the synapse. In NK cells, MUNC13-4 has been shown to interact with RAB27a⁵⁹ and is probably responsible for priming. Interestingly, MUNC13-4 can be derived from distinct vesicles, through a process of regulated fusion with developing lytic granules, thereby suggesting the existence of an independent step in and introducing an opportunity to regulate this stage of synapse formation. Probable targets of lytic-granule-associated MUNC13-4 are members of the SNARE (soluble N-ethylmaleimide-sensitive-factor accessory-protein receptor) family. SNARE proteins act in a coordinated manner to facilitate the fusion of two distinct membranes (reviewed in⁶⁰). SNARE proteins present on vesicles are termed v-SNAREs and those present on target membranes are termed t-SNAREs; an interaction between v-SNAREs and t-SNAREs is requisite for membrane fusion. Although direct interactions between Munc13-4 and SNAREs have not been shown in NK cells, homologous proteins in other cell types have been shown interact to enable fusion^{61, 62}. The only SNARE components required for lytic function in NK cells identified so far are the t-SNAREs VAMP7 (vesicle-associated membrane protein 7)⁶³ and syntaxin-11^64,65. The v-SNARE syntaxin-7 is present on lytic-granule membranes in NK cells and is likely to participate⁶⁶. The regulation of SNAREs and other proteins of these late effector stages will probably prove to be crucial in the fine-tuning of the NK-cell lytic synapse.

Termination stages

Termination stages of the NK-cell lytic synapse refer to those that occur after the lytic-granule contents have been secreted. Their sequence is presently hypothetical, but include a period of inactivity and down modulation of the accumulated activating receptors followed by NK-cell detachment from the target cell and recycling of cytolytic capacity (Figure 2). The cleft formed at the lytic synapse between the NK and target cell creates a protected pocket of up to 55 nm deep that is present 45 minutes after conjugation³¹. Although it is unclear how long this cleft is stable for, it probably remains intact after the granules have been released during a period of relative inactivity. The cleft and this delay in progression to subsequent stages may serve to increase the concentration of the lytic effector molecules exposed to the target cell while protecting neighbouring cells from exposure to these damaging molecules, which is likely to give the lytic synapse a specific purpose in vivo After this time, the synapse function is downmodulated, which in T cells is achieved by TCR internalization via the cSMAC^{39, 67} and includes localized TCR ubiquitylation, suggesting that the receptors are targeted for degradation⁶⁸. In NK cells, downregulation of the activating receptor NKG2D from the synapse has been observed under some circumstances⁶⁹ and may also occur for other receptors, such as 2B4 and NKp46^{70, 71}, although the mechanisms responsible for and purpose of downregulation of NK-cell activating receptors may be different to those underlying TCR downmodulation. Receptor downregulation at the NK-cell synapse could potentially happen at any point after signal generation is complete, but the persistence of certain activating receptors at the SMAC at late times after conjugation^{12, 51} suggests that it may be a relatively late event. Once the NK cell has carried out its cytolytic function, it can detach from the target cell and restore its ability to kill another susceptible cell. The physical process of detachment has been evaluated in the context of the NK-cell inhibitory synapse⁷² but not the lytic synapse. At the inhibitory synapse, detachment may result from reduced integrity of interactions between the F-actin cortex and the plasma membrane through dephosphorylation of ERM protein targets²³. Although this may also participate in the lytic synapse it has not yet been defined in NK cells. There may, however, be additional active detachment processes with similarity to those defined at the T-cell synapse, such as an involvement of protein kinase Cθ (PKCθ) in promoting synapse destabilization⁷³. Indeed, PKCθ has been found to localize to the lytic synapse in NK cells¹⁸. After detachment, the NK cell can restore its cytotoxic potential by generating new lytic granules and re-expressing any downmodulated activating receptors. Early functional studies⁷⁴ suggested that some NK cells retain the capacity to form a second synapse immediately after dissipation of the first, and more recent work defines the serial killing ability of activated NK cells⁷⁵. Other NK cells, as well as those completing serial kills will need to actively renew their functional capacity; this might occur rapidly (within a few hours) given the known kinetics of lytic-granule refilling as defined in CTLs⁷⁶. The signals initiating the process of recycling NK-cell cytolytic capacity are unknown but might be derived from those involved in cytotoxic function at the NK-cell lytic synapse. Interestingly, ligation of the activating receptor NKp30 induces rapid activation of nuclear factor-κB (NF-κB)⁷⁷, which has been shown to serve as a transcription factor for expression of the lytic granule component perforin⁷⁸. The physical process of detachment itself might also provide a signal for recycling in NK cells, although this remains hypothetical.

Human genetic diseases affecting the NK-cell immunological synapse

Primary immunodeficiency diseases are characterized by genetic aberrations that impair immunological function, defence or both. Several of these diseases affect NK cells (reviewed in^{79, 80}) and an informative subset are characterized by a specific blockade in the stages leading to the formation of a functional lytic synapse (Table 1). To date, none of the diseases in this subset has been defined as impairing NK cells in isolation and are expected to affect all cytotoxic lymphocytes. So, insight into how the lytic synapse is formed in cells from patients with these diseases has been gained mostly from T-cell studies. For those considered here, however, the functional defect has at least been established in patient NK cells. Although cytotoxic lymphocytes have crucial roles in host defence and immune regulation, there are likely to be specific contributions of NK-cell deficiency to the clinical phenotypes.

Table I.

Defects of the lytic synapse in natural killer cells

Disease	Gene	Protein	HLH phenotype	Effect on the lytic synapse	Step affected†	References§
LAD-I*	ITGB2	CD18	No	Decreased conjugation with target cells	3–5	87, 88
WAS	WASP	WASP	Some	Decreased F-actin reorganization and integrin redistribution	6–10	12, 27, 91
CHS	LYST	Lysosomal trafficking regulator	Yes	Inability to generate normal lytic granules for trafficking to the synapse.	6–10	98–103
HPS2	ADTB3A	AP3	Yes	Inappropriate formation of lytic granules and movement along microtubules	6–10	106, 107
GS2	RAB27A	RAB27a	Yes	Lytic granules move to the synapse but remain with microtubules	14–15	106, 107
FHL3	UNC13D	Munc13-4	Yes	Lytic granules move to the synapse but fail to dock and thus do not obtain an intimate association with the NK-cell plasma membrane.	16–17	85
FHL4	STX11	Syntaxin-11	Yes	Lytic granules polarize to and dock at the NK-cell plasma membrane, but fail to fuse.	18	64, 65, 114

^*Abbreviations: LAD-I leukocyte adhesion deficiency type I, WAS – Wiskott-Aldrich syndrome, CHS – Chediak-Higashi syndrome, HPS2 - Hermansky-Pudlak syndrome type-II, GS2 – Griscelli syndrome type-II, FHL - Familial erythrophagocytic lymphohistiocytosis

^†Step affected refers to the progress in formation of the NK-cell lytic synapse as depicted in Figure 2.

^§References directly relevant to NK cells are provided, others indirectly related to NK cells are provided in the text.

Most of these diseases can result in haemophagocytic lymphohistiocytosis (HLH)⁸¹. HLH represents an inappropriately robust immune response to infection (typically with herpesviruses), which results in a persistent systemic inflammatory syndrome. This leads to the physiological symptoms of septic shock, but is also associated with the pathological finding of haematophagocytosis (the ingestion of red blood cells by phagocytes). The defect in cytotoxic lymphocytes is believed to contribute to this phenotype, as the infected and other activated cells that promote inflammation cannot be eliminated (Figure 3). NK cells may be most relevant to the HLH phenotype, given their localization to marginal zones in lymphoid organs after viral infection, their innate function early in the course of infection⁸² and their inherent ability to eliminate hyperactivated macrophages⁸³. The defective NK cells in these diseases, although unable to eliminate infected cells by direct cytotoxicity can still carry out other functions, including the production of cytokines and stimulation of inflammatory responses. This further underscores distinct requirements for secretion of cytokines and lytic granule contents at the synapse. Nevertheless, the inability of NK cells to eliminate infected cells early in the infection (before T cells would be expanded) may be an important cause of HLH.

Figure 3. Proposed mechanism of HLH immunopathogenesis due to defective NK-cell lytic synpase function.

Pathogenic viral infection in a normal individual leads to the production of pro-inflammatory factors, such as tumour-necrosis factor (TNF), interferon-α (IFNα) and interleukin-12 (IL-12) by the infected cell (light green arrows). This induces relevant responses from uninfected cells including dendritic cells, monocytes, natural killer (NK) cells and other lymphocytes. These cells produce additional factors (dark green arrows) to further activate the induced cells and elicit the responses of others. Once induced, NK-cell cytotoxicity can help to rapidly eliminate the infected cell (thick red arrow) and serves to prevent further immune-cell activation induced by the infected cell. NK cell cytotoxicity can also eliminate other non-infected, activated monocytes and dendritic cells to provide additional and critical immunoregulatory function (thin red arrows). These processes have the potential to remain localized and focused to sites of infection. In an individual with impaired NK-cell cytotoxicity but a normal capacity for NK-cell activation, the NK-cell lytic synapse directed against the infected cell does not eliminate it (x). So, the inflammatory response continues unabated, leads to further activation of uninfected cells, and further pro-inflammatory activity from the induced cells. Although the infection may be localized, and the initial responding cells localized to the infection, the response amplifies and leads to an uncontrolled systemic inflammatory syndrome (purple arrows). The inflammatory response may control the viral replication, but without eliminating the source the inflammation itself may not be containable.

Other primary immunodeficiency diseases that impair cytolytic-cell function but do not disrupt the formation of the NK-cell synapse can also result in HLH. For example, HLH occurs in patients with a mutation in the gene encoding perforin⁸⁴ and in this setting is characterized by the normal secretion of lytic granules that are unable to mediate cytotoxicity due to the absence of perforin⁸⁵. The diseases that provide specific insight into the NK-cell lytic synapse are considered in two groups. Diseases in the first group affect steps involved with the recognition stage, or activation steps of the effector stage of the synapse. Diseases in the second group affect steps in lytic granule traffic to the synapse within the effector stage.

Diseases affecting recognition or activation steps of the NK-cell lytic synapse

Leukocyte adhesion deficiency type I (LAD-I) results from a defect in the CD18 (β-integrin) component of leukocyte integrin heterodimers⁸⁶. Thus, LAD-I leukocytes do not appropriately adhere to inflamed or activated cells and cannot localize effectively to tissues and sites of inflammation. This leads to increased numbers of leukocytes in the blood and susceptibility to infectious diseases. Because early steps in NK-cell synapse formation — adhesion and activation signalling — depend on integrins, NK cells from patients with LAD-I do not adhere to their target cells, resulting in defective cytotoxicity^{20, 87–89}. LAD-I is also distinguished from other diseases discussed here because it does not lead to HLH. This is presumably because NK cells from LAD-I patients do not form and are not activated through the synapse to produce cytokines.

Wiskott-Aldrich syndrome (WAS) results from a haematopoietic-cell-specific defect in actin reorganization and cell signalling due to WASP deficiency (reviewed in⁹⁰). Patients lacking WASP expression or expressing abnormal WASP have NK cells with decreased cytolytic capacity^{27, 91}. Clinically, patients with WAS are susceptible to herpesviruses⁹² and can develop HLH^93–95, thereby demonstrating the functional relevance of WASP deficiency for the NK-cell lytic synapse. Formation of the lytic synapse is abnormal in NK cells from WAS patients and includes decreased F-actin accumulation and adhesion-receptor clustering at the synapse^{12, 27, 91}. This underscores a requirement for WASP in early effector stages of the NK-cell lytic synapse. Importantly, exposure to interleukin-2 in vitro reverses the defect in WASP-deficient NK cells by restoring lytic function and F-actin accumulation at the synapse^{91, 96}. This demonstrates the existence of potential alternative pathways of actin polymerization in NK cells. At baseline, however, WASP is crucial for lytic synapse function but not early steps in synapse formation, thereby potentially enabling some pro-inflammatory functions to result in an occasional HLH phenotype.

Diseases affecting lytic granule traffic to the NK-cell lytic synapse

Chediak-Higashi syndrome (CHS) and Hermansky-Pudlak syndrome type II (HPS2) both affect the normal formation of lytic granules and lead to the presence of “giant” lytic granules. Both are also associated with albinism, which is caused by aberrant functioning of melanocytes, which pigment skin via secretion of melanosomes (an equivalent of lytic granules). CHS and HPS2 are similar in that they represent a failure in generation of the NK-cell lytic synapse at the end of the effector stages, as the abnormal lytic granules will not migrate along the microtubules to the MTOC. CHS results from a mutation in the LYST gene, which encodes the lysosomal trafficking regulator (reviewed in⁹⁷). Most patients with CHS ultimately experience an “accelerated phase” of the disease, which is an infection-induced HLH. The mechanics of NK-cell cytotoxicity and the NK-cell lytic synapse in CHS have not been studied in depth in recent years. Nevertheless, older studies identified defective cytolytic activity in patient NK cells despite normal adhesion to target cells^98–102. NK cells from CHS patients contain abnormal giant lytic granules¹⁰³, and studies in CTLs indicate that these giant granules arise from the fusion of individual lytic granules ¹⁰⁴. HPS2 is caused by a mutation in the ATD3BA gene (reviewed in¹⁰⁵), which encodes adaptor protein 3 (AP3) and, unlike CHS, HPS2 is associated with excessive bleeding due to an absence of the platelet storage pool and ensuing abnormal platelet aggregation. HPS2 patients exhibit defective NK-cell cytotoxicity^{106, 107} and the HLH phenotype¹⁰⁷. AP3 is required for the appropriate sorting of molecules from the Golgi into lytic granules. Similar to patients with CHS, CTLs from patients with HPS2 have enlarged granules that fail to move along microtubules¹⁰⁸. Griscelli syndrome type II (GS2) is a third syndrome that combines albinism and immunodeficiency and is also associated with an accelerated phase HLH phenotype. It is distinct from CHS and HPS2, in that patients have a broader range of hypopigmentation. GS2 is caused by mutation in the RAB27A gene, which encodes the RAB27A small GTPase¹⁰⁹. NK-cell cytotoxicity is decreased in patients with GS2, but not necessarily absent^{110, 111}, and can be reversed by culturing the NK cells in the presence of IL-2¹¹¹. So, the mechanism underlying the defect of the NK-cell lytic synapse in GS2 patients is also distinct from that in CHS and HPS2 patients. Studies performed in mouse RAB27A-deficient CTLs show that lytic granules migrate and polarize towards but fail to dock at the synapse⁵⁸. Specifically, they were separate from the membrane at the level of resolution in confocal microscopy. The reversal of the defect by IL-2, however, suggests there may be some redundancy in RAB27A function. Interestingly, RAB27B is capable of substituting for deficient RAB27A function under certain circumstances¹¹² and may provide a means for activation-induced augmented lytic granule traffic.

Familial erythrophagocytic lymphohistiocytosis (FHL) types 3 and 4 are similar to GS2, but are not associated with albinism demonstrating that the affected genes are not essential in melanocytes. FHL3 is caused by mutation in the UNC13D gene, which encodes MUNC13-4. Initially defined in CTLs, the lytic granules in MUNC13-4-deficient cells polarize towards the synapse and dock at the plasma membrane, but do not fuse¹¹³. Defective cytolytic activity and decreased granule fusion with the plasma membrane have also been observed in MUNC13-4-deficient NK cells⁸⁵. As explained above, MUNC13-4 interacts with RAB27A and primes the lytic granules for SNARE-mediated fusion with the NK-cell plasma membrane at the lytic synapse.

FHL4 is caused by mutations in the STX11 gene, which encodes syntaxin-11¹¹⁴. Although NK cells from patients harboring a STX11 mutation are defective in cytotoxic activity, it was initially thought that syntaxin-11 conferred a costimulatory role in promoting susceptibility of monocytes to NK-cell cytotoxicity. Subsequent studies however indicated a direct role for syntaxin-11 in NK-cell degranulation^{64, 65}. As described above, syntaxin-11 is a t-SNARE in NK cells and presumably interacts with respective v-SNAREs to enable lytic granule fusion. Importantly, the polarization of lytic granules to the synapse in NK cells from FHL4 patients, without subsequent degranulation, has been directly observed⁶⁵. FHL4 patients have diverse clinical phenotypes, with some showing late onset of disease¹¹⁵, which implies that syntaxin-11 mutations are hypomorphic or that there is some redundancy of the protein for enabling lytic granule fusion. In support of some level of redundancy, stimulation of NK cells from FHL4 patients with IL-2 can restore degranulation⁶⁵, suggesting an activation-induced synthesis of proteins that complement deficient syntaxin-11 function.

Concluding remarks

The NK-cell lytic synapse is formed in a series of stages that are required for cytolytic function. These include recognition/initiation, effector and termination stages that together enable the precise delivery of lytic-granule contents onto a susceptible target cell. Experimental investigations have defined some sequence to the individual steps, but have also raised many new and unanswered questions. First, what governs the access of lytic granules from the NK cell cytoplasm to the plasma membrane? A prerequisite conduit through the actin cortex at the synapse is likely a crucial element of cytotoxic cells, which may be generated by a mechanism as elegant as that of cytotoxicity itself. A second question relates to the regulation of the individual steps in synapse formation. Each step may have its own specific signal requirements, or may more simply represent a gradient emanating from strong initial signals. The former would enable the ideal fine-tuning of cytotoxic host defence.

There are important similarities of the NK-cell lytic synapse to the immunological synapse in T cells, however, there are also several notable actual and potential distinctions. These may relate to the fact that NK cells are armed for cytotoxicity in the resting state and rely on a balance of signals from inhibitory and activating receptors to function in immunosurveillance. Thus it is likely that studies of synapse regulation in NK cells will highlight mechanisms that control progression through the steps of lytic synapse formation. Some of these may be augmented, or even specific to NK cells to provide exacting control of the innate immune response in the face of less antigen specificity compared to the adaptive immune response. Identifying these mechanisms spatially and in real time as they relate to the synapse will provide valuable insight into the cellular control of the NK-cell cytolytic process.

Our understanding of the mechanisms of NK-cell lytic synapse formation and function has also been obtained from the study of rare diseases that affect these processes. Dissection of the HLH phenotype, in particular, has enhanced understanding of the discrete stages in the formation of the NK-cell lytic synapse. Although a limited number of genes have been ascribed to the HLH phenotypes, there are many patients with defective NK-cell function and HLH for which the genetic defects are unknown. It is likely that further study of these individuals will uncover additional proteins involved in the progression of NK-cell lytic-synapse formation.

Box 1: Proposed functions of the immunological synapse

There are numerous functions ascribed to the immunological synapse, many of which are relevant to natural killer (NK) cells.

Ligand recognition

The immunological synapse creates a critical zone at which ligands may be recognized with extraordinary precision amidst complex surroundings of non-relevant interactions.

Signal amplification and integration

Whether in large clusters at the central supramolecular activation cluster (cSMAC) or in discrete microclusters located in the periphery, important signalling molecules localize with receptors at the synapse. Innovative studies in NK cells have demonstrated function at these sites⁴⁰ as well as functional integration among signals¹¹.

Costimulation

NK-cell expression of critical costimulatory ligands such as CD40L¹¹⁶ and OX40L¹¹⁷ can enable their direct presentation to adaptive immune cells for facilitating responses. NK cells can form dual synapses in vivo with other innate cells and T cells¹¹⁸. The innate cell presumably induces NK-cell costimulatory ligands, which are then recognized by adjoined T cells.

Cytotoxicity

Precise targeting of cytotoxicity achieved through the immunological synapse seals the interaction between NK cell and target cell³¹ and can protect neighbouring cells from damage.

Directed secretion

Arrangement of molecules at the synapse can create a conduit through which cellular components can be secreted⁹. The synapse can facilitate linkage to intracellular transport machinery to allow focused secretion of small and large components including lytic granules²¹ and cytokines¹¹⁹.

Multi-driectional secretion

Activation signals from the synapse can induce secretion from within an NK cell, but at sites away from the synapse^{21, 50}. This might be useful after recognizing opsonizing IgG through NK cell FcγRIII, combating extracellular infections, and promoting local inflammation.

Protein transfer

Cell-surface proteins can be transferred from the NK cell to target cell membrane through the immunological synapse^{51, 120–124} to induce target-cell signal generation and protect NK cells from fratricide^{69, 125}.

Cell-fate determination

The immunological synapse polarizes cellular components in CTLs to enable unequal cell division leading to distinct cell fates¹²⁶. This has not yet been defined in NK cells, but will probably be important.

Inhibition of activation

The immunological synapse affords the opportunity to specifically ligate inhibitory receptors and prevent activation (see Box 3).

Signal termination

Molecular patterns at the immunological synapse may facilitate NK cell receptor internalization and/or end their productive signalling¹²⁷. This may be important in recycling or creating a refractory period and has been defined in T cells^{39, 67}.

Box 2: NK-cell lytic granules

The lytic granules found in natural killer (NK) cells are secretory lysosomes that have characteristics of both the cellular lysosomal compartment and specialized secretory machinery (reviewed in¹²⁸). They arise from the fusion of endosomes with specific secretory components from the trans-Golgi network. This results in a dual function organelle, that is, one specialized for destructive activities due to its lysosomal properties and secretory function. Important regulators of granule maturation are adaptor protein 3 (AP-3) and lysosomal trafficking regulator (LYST) protein, which regulate the addition of lysosomal proteins from the Golgi and lysosomal fusion, respectively. Components that are sorted into the lytic granules include granzymes A and B, which are transported from the trans-Golgi network by the mannose-6-phosphate receptor, as well as FAS ligand, granulysin and perforin. Mannose-6-phosphate receptor-independent sorting mechanisms exist and are likely also relevant to NK cell lytic granule development¹²⁹.

The genesis of the secretory lysosome in NK cells is a multistep process and is similar to that in cytotoxic T cells. There are, however, likely to be important differences between the processes in the two cell types as the generation of the lytic granules in T cells is only induced after cell activation. By contrast, secretory lysosomes are preformed and maintained in resting human NK cells¹³⁰. In addition, NK cells contain various related organelles with distinct morphology, including mature lytic granules, lytic granule precursors, multivesicular bodies and late endosomes (see figure – red arrowheads denote lytic granules and their precursory forms). These can polarize to the immunological synapse (as shown) where their contents can be secreted into a cleft formed between NK and target cell.

Box 3: The inhibitory immunological synapse in NK cells

An important structure besides the lytic synapse that can form between an NK cell and another cell is the inhibitory synapse¹⁷. Because resting NK cells contain lytic granules and express germline-encoded activating receptors, a robust mechanism for restraining the formation of a lytic synapse is essential to avoid inadvertent cytolysis. NK cells express a family of inhibitory receptors that recognize determinants of self and prevent the lysis of healthy cells. This inhibitory activity is mediated through the formation of an inhibitory synapse. Here inhibitory receptors such as the killer-cell immunoglobulin-like receptors (KIRs), which contain long cytoplasmic tails with immunoreceptor tyrosine-based inhibitory motifs (ITIMs), engage their ligands and induce the activity of phosphatases such as SHP1 (SH2-domain-containing protein tyrosine phosphatase 1)^{40, 131}. The clustering of KIRs and associated components at the inhibitory synapse has been termed the supramolecular inhibitory cluster (SMIC)¹⁷. The SMIC is distinguished from the NK-cell supramolecular activation cluster (SMAC) in that it excludes lipid raft membrane domains^{37, 132–134}, does not accumulate substantial F-actin³¹, and includes inhibitory signalling molecules such as SHP1^{10, 18, 131, 134}. The function of the SMIC in preventing SMAC formation is achieved by preventing actin reorganization^{22, 23}, blocking the recruitment of activating receptors^{24, 135} and promoting detachment from a target cell⁷². Mechanistically, this is likely due to the dephosphorylation of key molecules involved in F-actin dynamics including VAV1^{23, 136} and ezrin-radixin-moesin proteins ²³. As a result the NK cell synapse is likely a key site for signal integration between inhibitory and activating receptors¹¹.

The three-dimensional organization of the SMIC has been described to include an early recruitment of KIRs to the centre of the SMIC, which then migrate to the periphery, although several molecular patterns for the SMIC have been reported^{10, 17, 18}. The figure provides aschematic of the inhibitory synapse in which NK cells expressing KIR2DL1 are paired with target cells expressing the cognate ligand HLA-Cw4 (Left). The SMIC is characterized by clustered KIR2DL1 on the NK cell and HLA-Cw4 on the target cell. F-actin does not accumulate at the NK cell SMIC and the MTOC is not polarized. KIR2DL1 expressing NK cells do not form a SMIC with non-cognate HLA-Cw3-expressing target cells and neither accumulate at the synapse (Right). Here a SMAC is formed as F-actin accumulates in the NK cell at the synapse and the MTOC is polarized.

Supplementary Material

Figure S1

Online Supplementary information S1 (Figure). Interactive version of a prototypical mature NK cell lytic synapse. The mature NK cell lytic synapse in Figure 1 is provided in an interactive format. The image shows a human NK cell (YTS) expressing CD2-GFP (green fluorescent protein) conjugated to an EBV-transformed B cell (721.221). NK cells were conjugated to 721.221 cells, fixed, permeabilized and stained with a perforin-specific monoclonal antibody (red) to visualize the lytic granules. Serial images were obtained using a confocal microscope through the z-axis and the three-dimensional cell volume reconstructed in silico and QTVR output generated with complete rotation around the z-axis generated using Improvision Volocity visualization software. The accumulated GFP fluorescence indicates the supramolecular activation cluster (SMAC) formed at the synpase between the NK cell and the target cell. Notably, a region in the centre of the SMAC that is devoid of GFP is consistent with the presence of a cSMAC and delineates a channel through which the secretion of lytic granule contents is likely to occur. Although there is an accumulation of perforin internal to the SMAC (consistent with the lytic granules surrounding the microtubule-organizing centre (MTOC)), perforin can be seen coming from this region to the cSMAC. This is consistent with the secretory domain of the cytolytic immunological synapse.

Glossary

Danger signal

Agents that alert the immune system to danger and thereby promote the generation of immune responses. Danger signals can be associated with microbial invaders (exogenous danger signals), but can also be produced or expressed by damaged cells (endogenous danger signals)

Microclusters

Discrete collection of molecules at the immunological synapse capable of moving and generating signals. They are smaller than and exclusive of the central supramolecular activation cluster (cSMAC), which can represent a coalescence of multiple microclusters

Lipid rafts

Liquid-ordered membrane domains that are enriched in sphingolipidsand cholesterol, and also contain glyophosphatidylinositol-anchored receptors. They can provide ordered structure to the lipid bilayer and have the ability to include or exclude specific signalling molecules and complexes. There is some degree of controversy regarding lipid rafts given that most techniques for their investigation are indirect and include cell-membrane binding of cholera toxin, cholesterol sequestration using drugs, and cell fractionation based upon detergent sensitivity

microtubule-organizing centre

(MTOC). A structure that is found in all plant and animal cells, from which microtubules radiate. The two most important types of MTOCs are the basal bodies that are associated with cilia and the centrosome, which is composed of γ-tubulin-ring complexes for microtubule nucleation

NKG2D

(Natural-killer group 2, member D). A primary activating receptor encoded by the NK-cell gene complex and expressed by all mature NK cells. It recognizes distinct families of ligands that are generally expressed only by infected, stressed or transformed cells

Perforin

A component of cytolytic granules that participates in the permeabilization of plasma membranes, allowing granzymes and other cytotoxic components to enter, or be taken up by, target cells