Research ArticleImmunology

Superresolution Microscopy Reveals Nanometer-Scale Reorganization of Inhibitory Natural Killer Cell Receptors upon Activation of NKG2D

See allHide authors and affiliations

Science Signaling  23 Jul 2013:
Vol. 6, Issue 285, pp. ra62
DOI: 10.1126/scisignal.2003947

Abstract

Natural killer (NK) cell responses are regulated by a dynamic equilibrium between activating and inhibitory receptor signals at the immune synapse (or interface) with target cells. Although the organization of receptors at the immune synapse is important for appropriate integration of these signals, there is little understanding of this in detail, because research has been hampered by the limited resolution of light microscopy. Through the use of superresolution single-molecule fluorescence microscopy to reveal the organization of the NK cell surface at the single-protein level, we report that the inhibitory receptor KIR2DL1 is organized in nanometer-scale clusters at the surface of human resting NK cells. Nanoclusters of KIR2DL1 became smaller and denser upon engagement of the activating receptor NKG2D, establishing an unexpected crosstalk between activating receptor signals and the positioning of inhibitory receptors. These rearrangements in the nanoscale organization of surface NK cell receptors were dependent on the actin cytoskeleton. Together, these data establish that NK cell activation involves a nanometer-scale reorganization of surface receptors, which in turn affects models for signal integration and thresholds that control NK cell effector functions and NK cell development.

Introduction

Natural killer (NK) cells participate in immune defense against viral infections and tumor transformations through cellular cytotoxicity and the secretion of cytokines and chemokines. NK cell activation is regulated by cytokines, the local environment, and by cell-cell contacts. At the interface (the immune synapse) between an NK cell and a target cell, NK cells are regulated by the balance between activating and inhibitory signals induced by the ligation of germ line–encoded cell-surface receptors (1). Depending on the repertoire of receptors expressed on the NK cell surface and the quality and quantity of ligands found on the target cell, the interaction can lead to the formation of a cytolytic synapse, resulting in target cell death, or an inhibitory synapse, in which inhibitory signals dominate the outcome of the interaction and the target cell is spared (2, 3).

An important class of inhibitory NK cell receptors is the family of killer cell immunoglobulin-like receptors (KIRs), which recognize classical major histocompatibility complex (MHC) class I proteins [human leukocyte antigen A (HLA-A), HLA-B, and HLA-C] and induce inhibitory signals through immunoreceptor tyrosine-based inhibition motifs (ITIMs) within their cytoplasmic region. Target cells that are deficient in MHC class I are susceptible to killing by KIR-positive NK cells because of the lack of inhibitory signaling triggered at the immune synapse, consistent with the “missing self” hypothesis (4).

Over the past decade, fluorescence imaging of immune synapses has demonstrated that immunological recognition by T, B, and NK cells involves the segregation and organization of proteins into micrometer- and submicrometer-scale domains (1, 59). At the NK cell synapse, activating receptors such as NKG2D (10) and CD16 (11), as well as the inhibitory receptor KIR2DL1 (12), accumulate in small aggregates termed microclusters. High-resolution imaging of the interface between NK cells expressing KIR2DL1 and target cells expressing one of its ligands HLA-Cw6 demonstrated that, over time, small clusters of KIR2DL1 move from the periphery of the synapse toward the center (12).

Early evidence for the importance of the organization of KIR2DL1 at the immune synapse was the observation that its micrometer-scale organization varied with the abundance of MHC on the target cell (13). Moreover, phosphorylation of KIR2DL1, as visualized by fluorescence lifetime imaging microscopy (FLIM) to detect fluorescence resonance energy transfer (FRET), was spatially restricted to discrete microdomains within the immune synapse, rather than being homogeneously distributed across the synapse (14). Perhaps the best evidence that the organization of NK cell synapse proteins is important in signaling is that when the juxtaposition of activating and inhibitory receptor-ligand interactions is disrupted, by changing the sizes of the relevant proteins, the outcome of NK cell interactions is altered (15), consistent with data using micropatterned surfaces of activating and inhibitory ligands (16). Specifically, the inhibitory NK cell receptor KIR2DL1 could only inhibit NKG2D-mediated activating signals when these receptors and ligands were colocalized (15). Together, these data indicate that the synaptic organization of proteins is central to signal integration.

A major difficulty in understanding signal integration at this level, however, is that imaging of individual activating and inhibitory proteins, adaptors, or kinases has been hampered by the limit in resolution—at best about 250 nm—of conventional diffraction-limited microscopy techniques. Fluorescence imaging techniques that are capable of nanometer-scale resolution are now available (1720). Here, we used two methods for superresolution microscopy: photoactivated localization microscopy (PALM) and ground-state depletion microscopy followed by individual molecule return (GSDIM). Both techniques localize single molecules by exploiting the properties of stochastic switching of fluorophores (18, 20). When used with total internal reflection fluorescence (TIRF), PALM and GSDIM can be used to visualize the distribution of single proteins at the cell-surface membrane.

Some groups have applied superresolution microscopy techniques to the study of T cell recognition (2125), generating new insights into the pathway of T cell activation. There is evidence that the T cell receptor (TCR) and the adaptor molecule linker of activated T cells (LAT), for example, cluster in small aggregates, which concatenate to form microclusters upon TCR triggering (21). Two-color PALM revealed that clusters of TCR and the kinase ζ chain–associated protein kinase of 70 kD (ZAP-70) are partially segregated from clusters of LAT, with only small regions of overlap that likely represent “hot spots” for LAT phosphorylation (24). Additionally, some TCR signaling, perhaps at later times, may depend on a pool of LAT associated with subsynaptic intracellular vesicles rather than on LAT clusters that are already present in the plasma membrane (23, 26). Together, these studies have uncovered details of the TCR signaling pathway that were previously hidden, and helped establish that immune receptor signaling is facilitated by dynamic interactions between nanometer-scale protein clusters and vesicles rather than by a linear cascade of single protein-protein interactions.

These new imaging technologies have yet to be applied to visualizing recognition by immune cell types other than T cells. Here, with a view to understanding how positive and negative signals are integrated, we used PALM and GSDIM in combination with quantitative cluster analysis to study the nanometer-scale organization of the NK cell immune synapse. We found that a large fraction of the inhibitory receptor KIR2DL1 is constitutively aggregated in small nanometer-scale clusters at the NK cell surface. Surprisingly, activation of NK cells through ligation of the activating receptor NKG2D caused a rearrangement in the nanoscale organization of the inhibitory receptor KIR2DL1. These data establish a previously unappreciated aspect of the crosstalk between activating and inhibitory immune receptors, which we suggest should be taken into account in models of signal integration.

Results

The inhibitory receptor KIR2DL1 is organized in nanoscale clusters in NK cell–surface membranes

To achieve nanometer-scale resolution, PALM exploits the properties of photoactivatable or photoswitchable fluorescent proteins to reconstruct a superresolution image containing single-molecule localizations (18). The basic principle behind PALM is the stochastic photoactivation of small numbers of fluorescent proteins and the localization of each singly activated molecule with mathematical fitting. Here, we stably transduced the human NK cell lines NKL and YTS to express KIR2DL1 fused to tdEosFP, a photoswitchable fluorescent protein that is capable of an irreversible photoconversion from green fluorescence (with a peak emission wavelength of 516 nm) to red fluorescence (peak, 581 nm) upon near-ultraviolet irradiation at ~390 nm (27). With the correct laser settings, tdEosFP can be converted from the green to the red form such that sparse subsets of the red form can be imaged iteratively. The cell-surface abundance of the receptor was assessed by flow cytometry, and the photoswitching of KIR2DL1-tdEosFP was confirmed by confocal microscopy (figs. S1 and S2). Furthermore, cytotoxicity assays confirmed that KIR2DL1-tdEosFP was functionally active because expression of this receptor was sufficient to inhibit the lysis of target cells expressing the cognate ligands HLA-Cw4 or HLA-Cw6 (figs. S1 and S2). The transfected cell lines used in this study expressed numbers of KIR2DL1 molecules comparable to those of KIR2DL1+ primary NK cells isolated from human blood, as assessed by flow cytometry and quantification with beads (table S1).

First, we used PALM to characterize the spatial patterning of KIR2DL1 at the surface of resting NKL/KIR2DL1-tdEosFP and YTS/KIR2DL1-tdEosFP cells on poly-l-lysine–coated slides (Fig. 1A). We used Ripley’s K function analysis to determine the extent of clustering of a population of molecules compared to a randomly distributed set of molecules. Notably, quantifications of this type provide information only on the molecules detected and not on all molecules. Hence, this approach is most effective in revealing relative differences between conditions, and precise parameters obtained from the analysis are indicative but are not definitive. This analysis revealed that KIR2DL1 was evidently more clustered than would be expected for a spatially random set of molecules, indicated by an increased extent of peak clustering (Fig. 1, B and C). Depiction of the cluster profiles as a binary map facilitated a quantitative analysis of cluster morphology (Fig. 1, D and E). KIR2DL1 was organized into small, nanoscale clusters in both NKL/KIR2DL1-tdEosFP and YTS/KIR2DL1-tdEosFP cells, with a density of 3.9 ± 0.2 clusters/μm2 in NKL transfectants (Fig. 1D) and 3.3 ± 0.2 clusters/μm2 in YTS transfectants (Fig. 1E). Clusters were relatively circular. One measure of circularity, the ratio of their perimeter to their area, ranged between 0.75 and 0.79 (table S2); a perfect circle would have this parameter as 1. The clusters had similar diameters: 110 ± 4 nm in NKL/KIR2DL1-tdEosFP cells and 120 ± 8 nm in YTS/KIR2DL1-tdEosFP cells. The density of proteins detected within the clusters was 1920 ± 180 molecules/μm2 in NKL transfectants and 1140 ± 160 molecules/μm2 in YTS transfectants. No obvious difference in cluster morphology was observed between those at the center and those at the periphery of the synapse. Moreover, these nanoclusters were not an artifact resulting from fixation with paraformaldehyde (PFA) because we also detected clusters of KIR2DL1 in nonfixed, live NKL/KIR2DL1-tdEosFP cells placed on non-activating surfaces (Fig. 1, F and G).

Fig. 1 KIR2DL1 is organized in nanoscale clusters in NK cell–surface membranes.

(A) Representative brightfield and PALM images of NKL/KIR2DL1-tdEosFP and YTS/KIR2DL1-tdEosFP cells on slides coated with poly-l-lysine (PLL). Scale bars, 10 μm. The 3 μm × 3 μm regions outlined in red are enlarged and shown with corresponding cluster maps and binary maps (scale bars, 500 nm). Colors correspond to the extent of clustering according to the pseudocolor scale shown below the cluster map. (B and C) Ripley’s K analysis of the molecules in the selected regions [red boxes from (A)] in (B) NKL/KIR2DL1-tdEosFP cells and (C) YTS/KIR2DL1-tdEosFP cells. L(r) − r represents the degree of clustering relative to simulated random distributions [indicated by the 99% confidence intervals (CIs)]; r is the radial scale. (D and E) Quantitative analysis of KIR2DL1 clustering in (D) NKL/KIR2DL1-tdEosFP cells and (E) YTS/KIR2DL1-tdEosFP cells on PLL-coated slides. Cluster density, average cluster diameter, and density of protein within clusters were obtained from PALM images by spatial point-pattern analysis and cluster thresholding. Each symbol represents the mean from several regions analyzed in an individual cell. Horizontal bars and errors represent the means and SEM, respectively. Data are from 13 cells from two to four independent experiments. (F) Representative PALM image of live (unfixed) NKL/KIR2DL1-tdEosFP cells on slides coated with an isotype control antibody (left panel). Scale bar, 5 μm. The 3 μm × 3 μm region outlined in red is enlarged (right panel; scale bar, 500 nm). (G) Ripley’s K function analysis of live NKL/KIR2DL1-tdEosFP cells in the same region.

Ligation by monoclonal antibody alters the nanometer-scale organization of KIR2DL1

To assess whether ligation of KIR2DL1 affected its nanoscale organization in a manner that could be detected by PALM, we imaged NKL/KIR2DL1-tdEosFP cells after a 15-min incubation on slides coated with the anti-KIR2DL1 monoclonal antibody EB6 (widely used as a reagent to block KIR ligation) or with an isotype-matched control antibody [immunoglobulin G1 (IgG1)] (Fig. 2A). Indeed, ligation of KIR2DL1 substantially altered its nanometer-scale organization (Fig. 2, B and C). Ripley’s K analysis of representative regions indicated that KIR2DL1 was more clustered on EB6-coated slides than on IgG1-coated slides, as was demonstrated by the higher degree of peak clustering and by a shift to a lower radial scale at which peak clustering occurred (Fig. 2B). From quantitative analysis of the binary maps, we found that the density of clusters remained unchanged at 4.0 ± 0.1 clusters/μm2 on IgG1-coated slides and 3.8 ± 0.2 clusters/μm2 on EB6-coated slides (P = 0.50; Fig. 2C). However, the diameters of the clusters in NKL/KIR2DL1-tdEosFP cells decreased by 20% upon ligation of the receptor with specific antibody: from 122 ± 3 nm on IgG1-coated slides to 97 ± 2 nm on anti-KIR2DL1–coated slides (P < 0.0001). Furthermore, the density of protein within the clusters of EB6-treated cells increased by more than 60%, from 1420 ± 160 to 2330 ± 200 molecules/μm2 (P = 0.001).

Fig. 2 Nanoclusters of KIR2DL1 get smaller and denser upon ligation with monoclonal antibody.

(A and D) Representative brightfield and PALM images of (A) NKL/KIR2DL1-tdEosFP cells and (D) YTS/KIR2DL1-tdEosFP cells on slides coated with either isotype control mAb (monoclonal antibody) (IgG1) or anti-KIR2DL1 mAb (EB6). Scale bars, 10 μm. The 3 μm × 3 μm regions outlined in red are enlarged and shown with the corresponding cluster maps and binary maps (scale bars, 500 nm). (B and E) Ripley’s K function analysis of the molecules in selected regions in (B) NKL/KIR2DL1-tdEosFP cells and (E) YTS/KIR2DL1-tdEosFP cells on IgG1- or EB6-coated slides. (C and F) Quantitative analysis of KIR2DL1 clustering in (C) NKL/KIR2DL1-tdEosFP cells and (F) YTS/KIR2DL1-tdEosFP cells on IgG1- or EB6-coated slides. In all panels, data are from 19 to 22 cells from two to five independent experiments. n.s., not significant.

We detected similar changes in experiments with YTS/KIR2DL1-tdEosFP cells (Fig. 2, D to F), demonstrating that changes in the nanoscale clustering of KIR were not specific to NKL transfectants. Specifically, ligation of KIR2DL1 in YTS transfectants with specific monoclonal antibody decreased the cluster diameter by 16% (P = 0.011) and increased the density of KIR2DL1 within the clusters by 76% (P = 0.001) (Fig. 2F and table S2). Hence, these data suggest that the inhibitory receptor KIR2DL1 forms nanoclusters in resting NK cell membranes and that these clusters become smaller and denser upon ligation with monoclonal antibody. These experiments confirm that small changes in clustering can be detected with our methodology.

Ligation of the activating receptor NKG2D induces changes in the nanoscale organization of KIR2DL1

Consistent with previous findings (16), we found that ligation of the activating receptor NKG2D resulted in a cell-spreading response in NKL/KIR2DL1-tdEosFP cells as detected by confocal microscopy (fig. S3). We next set out to use superresolution microscopy to test whether ligation of NKG2D could cause any changes to the nanometer-scale organization of KIR2DL1 (Fig. 3). Upon ligation of NKG2D, we observed changes in the nanoscale clustering of KIR2DL1 (Fig. 3, A and B). An enlarged view of representative nanoclusters helps to visualize these differences (Fig. 3C). For visualization purposes, we denote the position of molecules by black dots of arbitrary size, which do not reflect the actual precision of localization. The density of protein within the clusters increased by 58%, from 1422 ± 157 molecules/μm2 in IgG1-treated cells to 2241 ± 155 molecules/μm2 in anti-NKG2D–treated cells (P = 0.0007; Fig. 3D). In addition, the density of KIR2DL1 clusters increased slightly from 4.0 ± 0.1 clusters/μm2 on IgG1-coated slides to 4.7 ± 0.1 clusters/μm2 on slides coated with anti-NKG2D antibody (P = 0.0002; Fig. 3D). Furthermore, the diameters of the clusters statistically significantly decreased by 19% in the presence of activating antibody, from 122 ± 3 nm in cells exposed to IgG1 control to 99 ± 3 nm in cells exposed to anti-NKG2D antibody (P < 0.0001).

Fig. 3 Ligation of the activating receptor NKG2D changes the nanoscale organization of KIR2DL1.

(A) Representative brightfield and PALM images of NKL/KIR2DL1-tdEosFP cells incubated on slides coated with IgG1 or anti-NKG2D (α-NKG2D) antibodies. Scale bars, 10 μm. The 3 μm × 3 μm regions outlined in red are enlarged and shown with corresponding cluster maps and binary maps (scale bars, 500 nm). (B) Ripley’s K function analysis of the molecules in the selected regions in NKL/KIR2DL1-tdEosFP cells on slides coated with control IgG1 or anti-NKG2D mAb. (C) Enlarged images of representative nanoclusters of KIR2DL1 at the surface of NKL/KIR2DL1-tdEosFP cells on slides coated with control IgG1 antibody or with anti-NKG2D antibody. Black dots represent the location of individual molecules; however, these black dots are of arbitrary size and do not reflect the precision with which molecules are located. Scale bars, 50 nm. (D) Quantitative analysis of KIR2DL1 clustering in NKL/KIR2DL1-tdEosFP cells on IgG1- or anti-NKG2D–coated slides. Data are from 19 cells from two to three independent experiments. (E) Distribution in the size of KIR2DL1 clusters. Cluster sizes were analyzed by nonlinear curve fitting (dashed lines). Vertical lines represent the mean cluster diameter. (F) Quantitative analysis of the same data as in (D) with the O-ring cluster analysis method.

The entire distribution of sizes appeared to shift to smaller clusters upon ligation of NKG2D (Fig. 3E). We confirmed the same changes in KIR2DL1 clustering in GSDIM experiments (fig. S4). Moreover, there are different ways in which protein clusters can be analyzed. To test the importance of this, we compared the quantitative results obtained when regions were analyzed as O-rings rather than as circles; the O-ring analysis was used, for example, in a superresolution microscopic study of T cells (24). We found that small differences in protein cluster parameters were indeed dependent on the method of analysis used; however, the trends and conclusions of our study were not affected (Fig. 3F and fig. S5). These results establish that ligation of NKG2D on NKL transfectants changed the organization of KIR2DL1 such that nanoclusters of KIR2DL1 were more numerous, smaller, and denser.

One possible reason for this observation is that NKG2D ligation might cause a generic change to the cell surface that influences the nanoscale organization of all surface proteins. Thus, we next set out to test whether ligation of NKG2D also influenced the organization of the class I MHC protein in GSDIM experiments with a directly labeled monoclonal antibody to detect HLA protein. GSDIM performs stochastic, single-molecule switching by using high laser power to promote the transition of standard fluorophores to a dark state while recording the fluorescence of those that have returned to the ground state (20). This analysis revealed that HLA was also organized in nanoclusters (Fig. 4A). The density of clusters of HLA on IgG1-coated slides was higher than that of clusters of KIR2DL1, with 8.7 ± 0.2 clusters of HLA per μm2 compared to 4.0 ± 0.1 clusters of KIR2DL1 per μm2. However, the diameters of HLA nanoclusters (116 ± 2 nm) were very similar to those of KIR2DL1 clusters (122 ± 3 nm) (Fig. 4B). We detected that there were more proteins within clusters of HLA than there were in KIR2DL1 clusters (8810 ± 930 HLA molecules/μm2 compared to 1420 ± 160 KIR2DL1 molecules/μm2). However, ligation of NKG2D caused no statistically significant change in the density of HLA clusters at the cell surface, the diameters of the clusters, or the density of HLA proteins within the clusters (Fig. 4B). Thus, ligation of the activating receptor NKG2D affected the cluster morphology of KIR2DL1 without having a general effect on all cell-surface proteins.

Fig. 4 The ligation of NKG2D has no effect on the nanoscale organization of HLA.

(A) Representative GSDIM images of HLA protein in NKL cells on IgG1- or anti-NKG2D–coated slides. Scale bars, 10 μm. The 2 μm × 2 μm regions outlined in red are enlarged and shown with the corresponding cluster maps and binary maps (scale bars, 500 nm). (B) Quantitative analysis of HLA protein clustering in NKL cells on IgG1- or anti-NKG2D–coated slides. Data are from 16 to 23 cells over two independent experiments.

Ligation of the activating receptor CD28 on YTS cells does not change the nanoscale organization of KIR2DL1

We next set out to establish whether the nanoscale reorganization of KIR2DL1 was directly caused by ligation of NKG2D or was an indirect effect, such as a consequence of the cell-spreading response caused by NK cell activation (16). To address this, we compared whether ligation of another activating receptor influenced the distribution of KIR2DL1. Because NKL cells are best activated through NKG2D, and less so by other activating receptors, we used the NK cell line YTS because it is specifically activated through CD28 (28). We confirmed that ligation of CD28 caused a spreading response in YTS cells, as detected by the appearance of a dense ring of F-actin at the synapse periphery, which was stained with phalloidin; 78% of cells spread on anti-CD28–coated slides compared to 6% on IgG1-coated slides (Fig. 5A).

Fig. 5 Ligation of the activating receptor CD28 has no effect on the nanoscale organization of KIR2DL1.

(A) Representative confocal microscopy images of F-actin in YTS/KIR2DL1-tdEosFP cells incubated on slides coated with isotype-matched control mAb (IgG1) or anti-CD28 mAb. Scale bars, 2 μm. (B) Representative brightfield and PALM images of YTS/KIR2DL1-tdEosFP cells incubated on IgG1- or anti-NKG2D–coated slides. Scale bars, 10 μm. The 3 μm × 3 μm regions outlined in red are enlarged and shown with the corresponding cluster maps and binary maps (scale bars, 500 nm). (C) Quantitative analysis of KIR2DL1 clustering in YTS/KIR2DL1-tdEosFP cells incubated on IgG1- or anti-CD28–coated slides. Data are from 20 to 23 cells from five independent experiments. (D) Representative confocal microscopy images of F-actin in YTS/KIR2DL1-tdEosFP cells incubated on slides coated with isotype control mAb (IgG2a) or with both anti-CD28 and anti-CD11a mAbs. Scale bars, 2 μm. (E) Quantitative analysis of KIR2DL1 clustering in YTS/KIR2DL1-tdEosFP cells incubated on slides coated with IgG2a mAb or with both anti-CD28 and CD11a mAbs. Data are from 16 cells from three independent experiments.

Next, we compared PALM images of YTS/KIR2DL1-tdEosFP cells on slides coated with anti-CD28 monoclonal antibody to those of cells on slides coated with a control IgG1 monoclonal antibody (Fig. 5B). We found that there were no statistically significant differences in the density of KIR2DL1 clusters at the cell surface (P = 0.65), the diameters of the clusters (P = 0.88), or the density of protein within the clusters (P = 0.1) (Fig. 5C), even though ligation of CD28 was sufficient to induce a spreading response in these cells. Signaling from CD28 synergizes with ligation of the adhesion molecule CD11a [also known as leukocyte function-associated antigen–1 (LFA-1)] to activate YTS cells (29); however, co-ligation of these two proteins, although it caused a spreading response in YTS/KIR2DL1-tdEosFP cells (Fig. 5D), did not affect the organization of KIR2DL1 (Fig. 5E). This establishes that ligation of an activating receptor capable of inducing a spreading response does not generally alter the nanoscale organization of KIR2DL1 in NK cells. This argues for a specific interplay between the ligation of NKG2D (and perhaps other activating NK cell receptors not tested here) and the organization of KIR2DL1.

Changes in the nanoscale organization of KIR2DL1 induced by ligation of NKG2D are dependent on actin

To test whether the organization of actin influenced the nanoscale architecture of NK cell–surface proteins, we pretreated NK cells with latrunculin A, which inhibits F-actin polymerization and promotes filament disassembly by binding to and sequestering G-actin monomers (30). Confocal imaging of latrunculin A–treated cells demonstrated a complete disruption of the actin cytoskeleton (fig. S6A). However, latrunculin A had no substantial effect on the nanoclusters of KIR2DL1 in cells incubated on anti-IgG1–coated slides (Fig. 6A). Thus, we conclude that the cortical actin mesh is not essential for the organization of KIR2DL1 into nanoscale clusters.

Fig. 6 The change in KIR2DL1 organization induced by NKG2D ligation is dependent on actin.

(A to C) Quantitative analysis of KIR2DL1 clustering in NKL/KIR2DL1-tdEosFP cells incubated on slides coated with (A) control IgG1, (B) anti-NKG2D mAb, or (C) anti-KIR2DL1 mAb after treatment with latrunculin A (Lat A). Graphs compare data from untreated and Lat A–treated cells. Data are from 14 to 19 cells for each condition from two to three independent experiments.

Next, we tested whether an intact actin cytoskeleton was required for the reorganization of KIR2DL1 proteins in response to ligation of NKG2D. An intact cytoskeleton plays an important role in NK cell activation and cytotoxicity (16, 31). However, at the concentration used here, latrunculin A did not entirely abrogate the ability of NKL/KIR2DL1-tdEosFP cells to be activated by slides coated with anti-NKG2D monoclonal antibody because a spreading response was still detected (fig. S6B). However, in the presence of latrunculin A, the density of clusters at the cell surface, the diameter of the clusters, and the density of protein within the clusters were unaffected by ligation of NKG2D in NKL/KIR2DL1-tdEosFP cells. That is, the organization of KIR2DL1 remained similar to that in cells incubated on control IgG1-coated slides (Fig. 6B). Thus, disruption of the actin cytoskeleton completely inhibited the effects of NKG2D ligation on the organization of KIR2DL1. In contrast, latrunculin A had no effect on the changes in the nanometer-scale organization of KIR2DL1 caused by anti-KIR2DL1 monoclonal antibody in NKL/KIR2DL1-tdEosFP cells (Fig. 6C). Similarly, the organization of KIR2DL1 in latrunculin A–treated YTS/KIR2DL1-tdEosFP cells was still affected by ligation with an anti-KIR2DL1 monoclonal antibody (fig. S7). Thus, monoclonal antibody–mediated ligation of KIR2DL1 directly changed the organization of KIR2DL1 independently of the cytoskeleton; however, the actin cytoskeleton was critical to the mechanism by which NKG2D mediated a change in the nanoscale clustering of KIR2DL1.

Ligation of NKG2D also affects the nanoscale distribution of KIR2DL1 in primary NK cells

The use of NK cell transfectants throughout these experiments raised the question of whether KIRs endogenously expressed on primary human NK cells were similarly clustered. To test this, we isolated primary NK cells from the peripheral blood of healthy donors and cultured the cells with interleukin-2 (IL-2) for 6 to 10 days. Primary KIR2DL1+ NK cells were then imaged by GSDIM on slides coated with IgG1 or anti-NKG2D antibody, with anti-KIR2DL1 antibody used to detect KIR2DL1 (Fig. 7A). In primary NK cells, the organization of KIR2DL1 proteins also exhibited a high degree of clustering (Fig. 7A and table S2), and ligation of NKG2D substantially altered the nanoscale organization of KIR2DL1 (Fig. 7B). In particular, cluster diameter was significantly decreased by 17%, from 146 ± 7 nm on IgG1-coated slides to 121 ± 6 nm on anti-NKG2D–coated slides (P = 0.017), and the density of proteins within clusters was statistically significantly increased from 12,720 ± 1390 molecules/μm2 on IgG1-coated slides to 16,850 ± 1280 molecules/μm2 on anti-NKG2D–coated slides (P = 0.038) (Fig. 7B). Thus, as also established by PALM analysis of NK cell transfectants, ligation of NKG2D influenced the nanoscale organization of KIR2DL1 in primary NK cells, resulting in smaller and denser nanoclusters of KIR2DL1.

Fig. 7 Ligation of NKG2D affects the nanoscale organization of KIR2DL1 in primary human NK cells from peripheral blood.

(A) Representative single-molecule GSDIM images of primary human NK cells incubated on slides coated with isotype-matched (IgG1) or anti-NKG2D mAb. Scale bars, 5 μm. The 2 μm × 2 μm regions outlined in red are enlarged and shown with the corresponding cluster maps and binary maps (scale bars, 500 nm). (B) Quantitative analysis of KIR2DL1 clustering in primary NK cells from one donor incubated on IgG1- or anti-NKG2D–coated slides. Data are from 13 to 14 cells and are representative of cells isolated from three healthy donors assessed in independent experiments.

Discussion

Over the past decade, imaging of immune cell interactions has established the presence of a supramolecular organization of proteins at micrometer and submicrometer scales within immune synapses. Through improvements in imaging technology, it has emerged that kinases, adaptors, and antigen receptors accumulate in nanometer-scale clusters within the broader micrometer-scale organization of immune synapses that were detected initially. Changes in the nanoscale organization of cell-surface proteins are important because they provide the foundation for most models of immunoreceptor triggering and are central in coordinating transient protein-protein interactions that facilitate downstream signals (23, 24).

NK cells are less reliant than are T cells or B cells on the triggering of one dominant activating immune receptor; instead, they are controlled by the balance between activating and inhibitory signals from multiple cell-surface receptors. Thus, NK cells are an important prototypical cell type for studying the critical interplay between positive and negative signaling surface receptors. The integration of these signals is important for the development of NK cells, as well as for controlling NK cell effector functions (32). One contemporary model for NK cell development requires that individual cells acquire an ability to kill target cells according to the specific repertoire of activating and inhibitory receptors they express, implying that signal integration is a critical determinant of NK cell development. However, despite the importance of NK cells in immune responses, very little is known about how NK cell receptors are organized at the nanometer scale.

To address this, we set out to investigate the organization of an inhibitory NK cell receptor, KIR2DL1, at the single-protein level. The diffraction limit for the spatial resolution of light microscopy has been circumvented through various complementary techniques, collectively known as superresolution microscopy, and it is this technological advance that here enabled the direct imaging of the nanometer-scale architecture of an inhibitory NK cell receptor for the first time. This analysis revealed that KIR2DL1 was constitutively organized in nanometer-scale clusters in resting NK cells, in both NK cell lines and primary human NK cells from peripheral blood.

Quantitative analysis of our PALM and GSDIM data provided estimates of several parameters that describe these clusters. Notably, however, each single-molecule superresolution technique does have caveats, and in particular, PALM underestimates the total number of molecules (because of photobleaching of fluorescent proteins), and GSDIM is subject to oversampling (because of multiple rounds of photoactivation). Thus, in our studies and in others with superresolution microscopy, quantitative analysis gives an indication of specific parameters, but it is most reliable in comparative studies in which the differences detected must reflect real changes in protein organization. We also only compared samples prepared and imaged with the same method. The nanoclusters of KIR2DL1 observed in this study were relatively circular, with an average diameter of 95 to 120 nm, which is comparable to the size of clusters determined in resting T cells for TCR and LAT (70 to 140 nm in diameter) (21) and for Lck (92 to 102 nm in diameter) (25). The density of KIR2DL1 clusters at the cell surface was between 3 and 5 clusters/μm2 in NKL and YTS cell lines, which is relatively similar to the cluster density of the TCR (10 to 20 clusters/μm2) (21), LAT (4.2 clusters/μm2) (23), and Lck (8 clusters/μm2) (25). Similarly, the number of KIR2DL1 proteins that we detected in each cluster (9 to 20) was in the same range as the number of TCR and LAT molecules per cluster (7 to 20) (21). These similarities in the size and density of protein nanoclusters likely reflect general properties of membrane protein compartmentalization (33, 34). The mechanism by which KIR2DL1 is confined to nanometer-scale clusters is unclear, as it is for all nanoscale clusters of proteins at immune cell surfaces. However, several nonmutually exclusive models have been proposed, as recently reviewed (34).

Particularly surprising was our discovery that when NK cells were activated through ligation of NKG2D, there were substantial changes in the nanoscale architecture of KIR2DL1 clusters, such that the clusters became smaller and denser. The effect of NKG2D ligation on the organization of KIR2DL1 was specific because there was no effect on the distribution of surface HLA proteins. Also, the ligation of a different activating receptor (CD28) had no effect on the distribution of KIR2DL1. The existence of an interplay between activating and inhibitory NK cell receptors has previously been indicated through fluorescence correlation spectroscopy of murine NK cells (35). In that study, the authors suggested that such a crosstalk between activating and inhibitory receptors is key to NK cell education, whereby engagement of MHC class I–specific inhibitory receptors controls the confinement of activating receptors in the plasma membrane (35). Expanding this concept here, our data demonstrate an alternative interplay (in human NK cells), whereby the engagement of activating receptors affects the nanoscale organization of inhibitory receptors.

The use of an inhibitor of actin polymerization demonstrated that although the actin mesh was not needed for KIR2DL1 clusters per se, the changes in KIR2DL1 clustering caused by NKG2D ligation were dependent on an intact actin cytoskeleton. Several mechanisms have been proposed to account for proteins not being randomly distributed in the plasma membrane, as was first proposed in the fluid mosaic model (36). Two such models that involve actin are the “protein island” model and the “picket fence” model. The protein island model describes the plasma membrane as containing cholesterol-low, protein-free domains that are distinct from cholesterol-enriched, protein-high domains within which an actin network is responsible for protein island stability (21, 37). The picket fence model uses actin to suppress the free diffusion of transmembrane proteins through the corralling, tethering, or both of receptors to the cytoskeleton (38, 39).

When actin polymerization is disrupted in T cells, the density of protein clusters is reduced; however, raft and nonraft markers are still clustered and colocalized with regions containing high protein density (37). Thus, in resting NK cells, KIR2DL1 may be localized within analogous “protein islands.” When NK cells are activated, there is a remodeling of the actin cytoskeletal mesh at the NK cell synapse (4042). It is possible that remodeling of the actin mesh directly affects the organization of KIR2DL1, or there may be an indirect effect because of a requirement of the actin mesh for signals downstream of NKG2D ligation. Although speculative, a transient increase in the density of KIR2DL1 might make NK cells especially sensitive to the recognition of inhibitory ligands on target cells that express activating proteins, thereby playing an important role in establishing appropriate self-tolerance. This could work through increasing the probability that serial triggering will occur (43); if the ligand is present on the target cell, the aggregation of its corresponding receptor will increase the chances that when the ligand dissociates, it will be able to associate with another receptor in that cluster. As a consequence, a lower number of ligands would be needed to transduce a productive signal (43). Specifically, we can estimate that the average distance between KIR2DL1-tdEoSFP proteins within clusters decreased from 29 to 22 nm, based on an estimation calculated by d = √(1/D), where d is the average distance between neighboring molecules, and D is the molecular density.

It is also possible that the increased density of KIR2DL1 in clusters reflects the fact that some KIR2DL1 proteins became oligomerized. KIR2DL1 forms dimers in the presence of metal ions, such as cobalt and zinc (44). Furthermore, the removal of these ions results in monomeric KIR2DL1 and abrogates the inhibitory signals from this receptor (45, 46). This would provide an alternative mechanism by which the observed increase in aggregation of nonligated KIR2DL1 reported here may increase the efficiency of KIR2DL1 signaling, mediated through the downstream binding of Src homology 2 (SH2) domain–containing protein tyrosine phosphatase 1 (SHP-1) to the receptor (47). SHP-1 physically associates with the phosphorylated membrane-proximal ITIM in KIR2DL1 through its tandem SH2 domains (48). However, structural analyses demonstrate that full activation of SHP-1 is only achieved when both of its SH2 domains are simultaneously associated with phosphorylated tyrosines (49, 50). This suggests that active SHP-1 would involve the cross-linking of two KIR2DL1 receptors, each contributing one phosphorylated ITIM. This is further supported through studies that demonstrate the need for an optimal distance and orientation between two phosphorylated tyrosines to achieve full enzyme activity (49, 50). Thus, tighter clustering of KIR2DL1 proteins may lower the threshold of the inhibitory signal by placing the receptors in a more optimal spacing or orientation for the binding of SHP-1.

With superresolution imaging technologies ever improving, a future goal is to use superresolution microscopy to observe the spatial and temporal locations of both KIR2DL1 and NKG2D at the interface between cells; however, for now, this is technically not possible. Yet already here, by studying NK cells interacting with coated slides, superresolution single-molecule imaging has revealed nanometer-scale clusters of inhibitory receptors whose structure is unexpectedly altered by activating receptor signals, a previously unrecognized concept that has important functional consequences for signal integration in NK cells.

Materials and Methods

Cell lines and primary NK cells

The NK cell lines YTS and NKL were maintained in RPMI 1640 (Gibco) supplemented with 10% fetal calf serum (FCS), 2 mM l-glutamine, and 1 mM penicillin and streptomycin (all from Invitrogen; complete medium). NKL cells were cultured in the presence of IL-2 (100 U/ml; Roche). All transfectants were generated by retroviral transduction as previously described (51). In brief, the packaging cell line Phoenix amphotropic (Nolan Lab, Stanford) was transfected with Lipofectamine LTX (Invitrogen) with the PINCO retroviral vector encoding KIR2DL1 fused to tdEosFP. Viral supernatant collected 24 and 48 hours after transfection was used for three sequential centrifugations (at 300g for 45 min at 30°C) for infection of 106 YTS or NKL cells. Two weeks after infection, cells expressing KIR2DL1 were enriched by flow cytometry. Primary human NK cells were isolated from the peripheral blood of healthy donors by negative selection with magnetic bead–based NK cell isolation kit (Miltenyi Biotec) and were cultured as described previously (52). Experiments were performed with freshly isolated NK cells that were stimulated with human recombinant IL-2 (150 U/ml) for 6 days.

Plasmid generation

To express KIR2DL1*002 tagged at the C terminus with tdEosFP, the coding sequence of tdEosFP (MoBiTec) was substituted for the coding sequence of green fluorescent protein (GFP) in the retroviral vector PINCO encoding KIR2DL1-GFP (12) at the Bam HI and Not I sites. This resulted in an eight–amino acid residue linker, GVPSSLE.

Preparation of coated slides

Glass coverslips (Lab-Tek, Nunc) were prepared as described previously (16). Briefly, slides were cleaned with 1 M HCl and then 70% ethanol, coated with 0.01% poly-l-lysine, dried, and then coated with monoclonal antibody (at 3 μg/ml) in phosphate-buffered saline (PBS) overnight. Slides were then washed and blocked with complete medium. Monoclonal antibodies specific for KIR2DL1 (EB6; Beckman Coulter), NKG2D (clone 149810; R&D Systems), CD28 (BD Biosciences), or CD11a (BD Biosciences) and murine IgG1 (BD Biosciences) or IgG2a (R&D Systems) isotype controls were used.

Sample preparation

Cells were allowed to settle on the surface of precoated slides for 15 min at 37°C. For experiments with latrunculin A, cells were pretreated with 2 μM latrunculin A (Tocris Bioscience) for 1 hour before being added to the slides. Cells were then fixed with 6% PFA for 20 min at room temperature and washed three times in PBS. To visualize F-actin for confocal microscopy, cells were additionally permeabilized with 0.1% Triton X-100 (Sigma) in PBS for 4 min, blocked in 5% bovine serum albumin (BSA) in PBS for 15 min, and then stained with Alexa Fluor 647–conjugated phalloidin (2 U/ml; Molecular Probes). For GSDIM, fixed cells were washed and blocked with 0.1% saponin, 4% BSA, 4% FCS, 0.2% NaN3 in PBS for 20 min at room temperature. The sample was then stained with Alexa Fluor 488–conjugated anti–MHC class I antibody (5 μg/ml, W6/32; Serotec) or with anti-CD158a antibody (5 μg/ml; BD Pharmingen) in blocking solution overnight at 4°C. For samples stained with an unconjugated primary antibody, the sample was washed and incubated with Alexa Fluor 488–conjugated anti-mouse IgM antibody (5 μg/ml; Invitrogen) for 1 hour at room temperature. All samples were then washed and imaged in 50 mM cysteamine (Sigma) in PBS.

Microscopy

For laser scanning confocal microscopy, cells were imaged with an inverted microscope (TCS SP5 RS, Leica Microsystems) with a 63×, 1.2 numerical aperture, water immersion objective using LAS AF (Leica Application Suite Advanced Fluorescence) software (Leica). For PALM, images of fixed samples were acquired on an inverted wide field fluorescence microscope (Zeiss) with TIRF illumination, with a 100×, 1.45 numerical aperture, oil immersion objective. Photoconversion of tdEosFP was achieved with a polychrome light source (Polychrome IV, TILL Photonics) set to 390 nm, and the red-converted form of tdEosFP was imaged with the 561-nm laser line. For each PALM acquisition, 2 × 104 raw images were acquired with a cooled electron-multiplying charge-coupled device camera (C9100-13, Hamamatsu) with an exposure time of 30 ms using SimplePCI Software (Hamamatsu). Brightfield images were acquired with an exposure time of 108.5 ms. For GSDIM, images of fixed samples were acquired on a Leica SR GSD microscope with TIRF illumination, with a 100×, 1.47 numerical aperture, oil immersion objective. For each GSDIM acquisition, samples were first pumped in epifluorescence mode with 100% of the 488-nm laser power (20.8 mW at the objective) until most molecules moved to the triplet/dark states. Then, 2.5 × 104 raw images were acquired in TIRF mode with an exposure time of 10 ms with 100% of the 488-nm laser power.

Data analysis

Superresolution PALM images were reconstructed from raw image sequences with QuickPALM software (53), which is available as a plug-in for ImageJ processing software (U.S. National Institutes of Health). Superresolution GSDIM images were reconstructed from raw image sequences using Leica GSD wizard software with a minimum photon cutoff of 400 for each point source and a filter that discards events that reappear within 10 frames and are within 50 nm of the first point source. Non-overlapping regions of 3 μm × 3 μm and 2 μm × 2 μm were selected for quantitative analysis for PALM and GSDIM, respectively. Several non-overlapping regions were selected for each cell (avoiding cell edges). The data obtained from several regions from the same cell were then averaged, giving a representative result for that cell. The number of regions analyzed per cell varied from 1 to 6, with a mean of 3. For primary NK cells, which are smaller than YTS and NKL cells, a smaller region area was selected (2 μm × 2 μm rather than 3 μm × 3 μm), and this size was kept for all GSDIM analysis. The two-dimensional coordinates of events within a region were subjected to Ripley’s K function analysis and quantitative cluster mapping, as has been previously described in detail (22, 23, 25). Ripley’s function was calculated with SpPack, an add-in for Microsoft Excel (54). For PALM and GSDIM, 50 and 80 nm were used as the expected scales of the clusters, respectively (55). Quantitative cluster mapping generated two-dimensional pseudocolored cluster maps and thresholded binary maps. Cluster maps show the extents of clustering [where values of L(r) of each point are interpolated on a surface plot according to a pseudocolor scale]. Binary maps were obtained by thresholding of cluster maps to define clusters. The fraction of molecules detected in clusters ranged from 40 to 60% of the total molecules detected. This fraction of total molecules detected was then used to generate the cluster densities, cluster diameters, and molecular densities. The circularity, size, and number of clusters and other parameters were extracted from the binary maps with ImageJ software.

Statistical analysis

Statistical significance was determined by performing two-tailed, unpaired Student’s t tests (GraphPad software, Prism). Multiple means were compared with one-way analysis of variance (ANOVA) and Bonferroni corrections. Graphs show mean values, and error bars represent the SEM. In statistical analysis, P > 0.05 is indicated as not significant, and statistically significant P values are indicated by asterisks as follows: *P < 0.01, **P < 0.001, and ***P ≤ 0.0001.

Supplementary Materials

www.sciencesignaling.org/cgi/content/full/6/285/ra62/DC1

Methods

Fig. S1. Functionality of KIR2DL1 and tdEosFP in NKL/KIR2DL1-tdEosFP cells.

Fig. S2. Functionality of KIR2DL1 and tdEosFP in YTS/KIR2DL1-tdEosFP cells.

Fig. S3. Ligation of NKG2D causes a spreading response in NKL/KIR2DL1-tdEosFP cells.

Fig. S4. GSDIM confirms the nanoscale reorganization of KIR2DL1 upon ligation of NKG2D.

Fig. S5. Comparison of cluster analysis by the circle and ring methods.

Fig. S6. Latrunculin A disrupts the actin cytoskeleton but does not inhibit activation in NKL/KIR2DL1-tdEosFP cells.

Fig. S7. Latrunculin A has no effect on the changes in organization of KIR2DL1 in YTS/KIR2DL1-tdEosFP cells induced by EB6.

Table S1. Quantification of the cell-surface abundance of KIR2DL1 molecules in NK cell line transfectants and primary human NK cells.

Table S2. Mean values for different parameters assessing KIR2DL1 distribution and cluster morphology.

References and Notes

Acknowledgments: We thank T. Magee and M. Spitaler from the FILM (Facility for Imaging by Light Microscopy) imaging facility (Imperial College London) and members of our laboratory for discussion. Funding: The work was funded by the Biotechnology and Biological Sciences Research Council (ref: BB/I013407/1), the Medical Research Council (ref: G1001044), and a Wolfson Royal Society Research Merit Award. Author contributions: S.V.P. and S.-P.C. performed the experiments and analyzed the data; D.M.O. helped with data analysis; S.M.R. helped establish novel instrumentation; A.O. performed the experiments; and S.V.P., S.-P.C., and D.M.D. conceived the project, designed the experiments, and wrote the manuscript. Competing interests: The authors declare that they have no competing interests.
View Abstract

Navigate This Article