Research ArticleImmunology

NK cells integrate signals over large areas when building immune synapses but require local stimuli for degranulation

See allHide authors and affiliations

Science Signaling  25 May 2021:
Vol. 14, Issue 684, eabe2740
DOI: 10.1126/scisignal.abe2740

Ligand-patterned immune synapses

Natural killer (NK) cells execute virus-infected and tumor cells by forming immunological synapses with the target cells and releasing cytotoxic granules. The NK cell receptors LFA-1 and CD16 drive target cell recognition, synapse formation, and degranulation. Using artificial immune synapses (AIS) consisting of LFA-1 and CD16 ligands, Verron et al. found that the spatial distribution of receptor activation influenced the stability of synapses and subsequent degranulation. Although human NK cells formed synapses and properly positioned the lytic machinery on both dot- and donut-shaped AIS, degranulation did not occur in the central regions of donut-shaped AIS. The dependence of degranulation on local signaling suggests the existence of a late checkpoint for degranulation and suggests that target cells might be able to evade execution by controlling the spatial distribution of ligands (see also the Focus by Veillette).


Immune synapses are large-scale, transient molecular assemblies that serve as platforms for antigen presentation to B and T cells and for target recognition by cytotoxic T cells and natural killer (NK) cells. The formation of an immune synapse is a tightly regulated, stepwise process in which the cytoskeleton, cell surface receptors, and intracellular signaling proteins rearrange into supramolecular activation clusters (SMACs). We generated artificial immune synapses (AIS) consisting of synthetic and natural ligands for the NK cell–activating receptors LFA-1 and CD16 by microcontact printing the ligands into circular-shaped SMAC structures. Live-cell imaging and analysis of fixed human NK cells in this reductionist system showed that the spatial distribution of activating ligands influenced the formation, stability, and outcome of NK cell synapses. Whereas engagement of LFA-1 alone promoted synapse initiation, combined engagement of LFA-1 and CD16 was required for the formation of mature synapses and degranulation. Organizing LFA-1 and CD16 ligands into donut-shaped AIS resulted in fewer long-lasting, symmetrical synapses compared to dot-shaped AIS. NK cells spreading evenly over either AIS shape exhibited similar arrangements of the lytic machinery. However, degranulation only occurred in regions containing ligands that therefore induced local signaling, suggesting the existence of a late checkpoint for degranulation. Our results demonstrate that the spatial organization of ligands in the synapse can affect its outcome, which could be exploited by target cells as an escape mechanism.


Natural killer (NK) cells are innate cytolytic immune cells that are crucial for host defense against viral pathogens and cancer. NK cell surveillance relies on the recognition of potential target cells by engagement of activating and inhibitory receptors on the NK cell surface. This process is facilitated by the formation of a tight intercellular contact, or immune synapse, between the NK cell and its target. At this interface, NK cell receptors and target cell ligands segregate into distinct spatial domains called supramolecular activation clusters (SMACs) (1). Initial studies of immune synapse formation in T cells and NK cells describe a central SMAC (cSMAC) in which activating signals are concentrated, encircled by a peripheral SMAC (pSMAC) that promotes adhesion, which, in turn, is surrounded by a distal SMAC (dSMAC), containing proteins excluded from the central parts of the immune synapse (25). Although subsequent work has shown that the picture is complex (68), NK cell receptors involved in the immune synapse are commonly observed to distribute across the central part of the contact area (corresponding to the cSMAC) or in a peripheral ring structure (corresponding to the pSMAC) (1, 8). Thus, these two distribution patterns are hallmarks for the NK cell immune synapse.

After the NK cell has engaged in an immune synapse with its potential target, the balance between activating and inhibitory signals determines the outcome of the interaction [reviewed in (9)]. If activating signals dominate, then an enclosed cleft is formed in the central part of the synapse where cytolytic molecules including perforin and granzymes can be released from specialized intracellular vesicles, causing target cell death [reviewed in (10)]. The stepwise formation of this cytotoxic immune synapse involves recognition and adhesion to the target cell, minus-ended movement of lytic granules along microtubules until convergence at the microtubule-organizing center (MTOC), reorganization of the actin cytoskeleton at the synapse, receptor- and ion channel–mediated signaling, polarization of the MTOC and lytic granules toward the intercellular contact, and lastly fusion of the lytic granules with the NK cell membrane, resulting in the directed secretion of their cytotoxic contents into the synaptic cleft (1012). As for cytotoxic T cells, calcium (Ca2+) signaling mediated by the entry of extracellular Ca2+ is crucial for NK cell degranulation (1315).

These multiple steps toward cytotoxicity are differently driven by individual receptor engagement (16, 17). Engagement of the integrin lymphocyte function–associated antigen 1 (LFA-1), a heterodimer composed of the α-subunit CD11a and the β-subunit CD18, on the NK cell facilitates adhesion to the target cell and initiates the first steps of immune synapse formation, including actin polymerization (11, 12, 18). Engagement of LFA-1 alone or in combination with inhibitory receptors results in asymmetric spreading of the NK cell and resumption of migration, whereas its association with an activating signal leads to the formation of a stable, symmetrical synapse (19). Furthermore, engagement of LFA-1 alone stimulates polarization but not degranulation of lytic granules, and engagement of only the NK cell activating Fc receptor CD16 induces degranulation but not polarization of lytic granules (16). Ligation of both LFA-1 and CD16 leads to polarization of the lytic machinery and degranulation in the direction of the target cell (16).

Signaling at the immune synapse is regulated not only by the nature of receptors involved but also by their spatial distribution. Similar to receptors engaged in the immune synapses formed by T cells (2024), NK cell receptors form microclusters at the synapse. These include both inhibitory receptors such as killer cell immunoglobulin-like (KIR) receptors (1) and NK group 2 member A (NKG2A) (19) and activating receptors such as NKG2D (19), the natural cytotoxicity receptor NKp46 (25), and the Fc receptor CD16 (26). Signaling from these microclusters can be processed locally or in a cumulative fashion wherein the NK cell integrates signals from spatially separated ligands (19). Modifying the local distribution of NK cell ligands by altering their segregation into separate domains (27) or forcing T cell receptor (TCR) arrangement into defined static structures (28) modulates lymphocyte activation and reorganization of molecular assemblies in the cell.

Here, we used microcontact printing to create arrays of artificial immune synapses (AIS) to investigate the consequence of ligand distribution on NK cell function. Combining NK cell adhesion mediated by LFA-1 and activation through CD16, receptors were stimulated with typical immune synapse structures in the shape of uniform disks (“dots”) or rings (“donuts”). Time-lapse imaging of resting NK cells interacting with AIS revealed that the spatial distribution of ligands influenced the formation of the synapse as shown by more frequent transient, partial contacts on donut-shaped AIS compared to dot-shaped AIS. These results show that the NK cell response is regulated not only by the type and abundance of stimuli but also by their spatial distribution. NK cells that showed a strong Ca2+ response often spread symmetrically to cover the whole AIS, regardless of the shape of the AIS. These complete contacts also exhibited similar organization of the plasma membrane at the AIS interface and axial and lateral positioning of the MTOC, indicating that NK cells integrate signaling from spatially separated stimuli when building the immune synapse. However, degranulation was only observed in regions where there were local stimuli from ligands. This suggests that NK cell cytotoxicity is regulated by signal integration over a large area for establishment of an immune synapse and by local stimuli for execution of degranulation.


Combined engagement of CD16 and LFA-1 induces NK cell stopping and spreading

To mimic human NK cell immune synapses suitable for high-resolution, live-cell fluorescence microscopy, we generated AIS by microcontact printing (29). Antibodies or natural ligands that stimulate the NK cell receptors LFA-1 or CD16 were printed in arrays of dot- or donut-shaped, cell-sized patches onto glass substrates (Fig. 1, A to C). To stimulate these receptors, we used antibodies specific for LFA-1 or CD16 (αLFA-1 or αCD16, respectively), as well as the natural LFA-1 ligand recombinant human intercellular adhesion molecule-1 (rhICAM-1) and rituximab (RTX), which contains the human Fc region that is the natural ligand of CD16. The dynamic responses of primary human NK cells interacting with dot-shaped AIS stimulating LFA-1 and/or CD16 were studied by live-cell imaging.

Fig. 1 Antibodies patterned into AIS using microcontact printing.

(A) Schematic representation of the microcontact printing process. Poly(dimethylsiloxane) (PDMS) was molded on a microstructured silicon master (casting). The PDMS was peeled off from the master, loaded with a solution of the desired combination of αLFA-1, αCD16, rhICAM-1, and RTX (inking) and briefly dried. The stamp was applied to a poly-lysine–coated glass surface (printing), leaving an array of patches representing AIS that was used as substrate for NK cell binding and subsequent live or fixed imaging. (B and C) Example fluorescence images of AIS in dot (B) and donut (C) shapes. The loading antibody solution was supplemented with fluorescently labeled BSA (here shown in blue) for visualization. Scale bars, 10 μm.

Consistent with previous observations (19, 30), we found that engagement of LFA-1 alone by ligation to αLFA-1 triggered the initiation of an immune synapse, but in the absence of further activating signals, the cells remained elongated and asymmetrical (Fig. 2A). Ligation to both αLFA-1 and αCD16 caused the cells to stop and spread out symmetrically across the AIS, a morphology consistent with natural mature activating immune synapses (Fig. 2B). This behavior manifested as larger spreading area (Fig. 2C) and increased roundness (Fig. 2D) on AIS with both αLFA-1 and αCD16 (αLFA-1 + αCD16) compared to αLFA-1 alone. In addition, NK cells migrated faster on arrays of AIS with only αLFA-1 compared to AIS with αLFA-1 + αCD16 (Fig. 2E). Using a previously developed method based on detecting transient changes in migration behavior for individual cells, we divided NK cell migration tracks on AIS into periods of arrest, random migration, or directed movement (31, 32). NK cells on arrays of αLFA-1 + αCD16 AIS spent significantly more time in transient migration arrest periods (TMAPs; Fig. 2, F to H). Directed movement was only rarely observed on AIS of either composition. These differences in cell shape, spreading area, and migration behavior were not unique to arrays of AIS, because similar differences were observed also on surfaces evenly coated with αLFA-1 alone or a mix of αLFA-1 and αCD16 (fig. S1, A to H). Slightly higher migration speeds, associated with longer periods of directed migration, were measured on evenly coated surfaces, which could be due to the cells having access to a continuous layer of protein that is not present on arrays of AIS.

Fig. 2 Interaction of NK cells with LFA-1 alone promotes an exploratory phenotype, and engagement of LFA-1 and CD16 induces cell arrest and spreading.

(A and B) Representative fluorescence images of NK cells interacting with AIS (blue) composed of αLFA-1 alone (A) or αLFA-1 + αCD16 (B). NK cells were stained for F-actin (red), and maximum intensity projection images were generated from confocal z-stacks. (C and D) Spreading area and roundness of NK cells on each type of AIS. Spreading area (C) was measured as the footprint of the NK cell F-actin network in the imaging plane of the AIS. Roundness (D) was measured as 4π × cell area/perimeter2. n = 3 independent experiments, with 50 cells per condition per experiment. (E to H) NK cells were stained with the live cell marker Calcein Green and followed by time-lapse microscopy during their interaction with each type of AIS to measure NK cell migration speed (E) and directionality (F to H). Representative NK migration tracks that have been divided into three different modes of migration: Arrest (TMAP, magenta), random (black), and directed migration (orange) are shown (F and G) as well as the average percent of time spent in the different modes of migration on AIS (H). The P value in (H) was calculated by comparing the TMAP fraction between groups. n = 5 independent experiments, with 150 cells per condition per experiment. (I) Percentage of NK cells making contacts with one, two, three, or more AIS over the course of the assay (240 min). (J) Representative fluorescence image of a Calcein Green–stained NK cell (green) in simultaneous contact with three AIS through membrane tethers (white arrowhead). (K) Maximum number of AIS contacted simultaneously by individual NK cells. Data in (I) and (K) represent n = 4 independent experiments, with 60 cells per condition per experiment. Colored squares indicate the median measurement value for each independent experiment, with matching colors between paired experiments. Cluster plots in the background represent individual measurements. The P values in (C) to (E) and (H) were calculated using paired Student’s t test between the median of measurements for each experiment. The P values in (I) and (K) were calculated using paired Student’s t test between the mean count of contacts for each experiment. Scale bars, 10 μm.

The increased migration on AIS composed of αLFA-1 alone resulted in individual NK cells contacting several AIS on these arrays. On αLFA-1 AIS arrays, about 35% of the NK cells contacted multiple AIS during the 240-min assay, whereas the corresponding number for cells on αLFA-1 + αCD16 AIS was 12% (Fig. 2I). On αLFA-1 AIS, NK cells often appeared to have one end anchored to an AIS print while the other end of the cell scanned the surroundings, giving them an elongated shape (Fig. 2D). During this process, membrane protrusions resembling nanotubes that have been observed between NK and target cells (33, 34) were often formed, facilitating adhesion to one AIS while extending to bind to additional AIS (Fig. 2J). These membrane extensions could connect one NK cell with up to three different AIS simultaneously. Similar behavior was not observed on αLFA-1 + αCD16 AIS, where the maximum number of simultaneous contacts was 2 (Fig. 2K). When AIS were printed with a short center-to-center distance (≤30 μm), we observed that some cells assumed an elongated shape, forming stable contacts with two AIS prints at the same time, with the main part of the cell body oscillating between the prints (movie S1). These results show that signaling through LFA-1 triggers synapse initiation but allows for continued NK cell motility, which can be balanced by spreading and stopping induced by CD16.

Uniform AIS promote the formation of long-lasting, symmetrical synapses

To investigate whether NK cell responses were affected by the spatial distribution of ligands, we imaged NK cells on either dot- or donut-shaped AIS containing equal amounts of αLFA-1 and αCD16. NK cells formed significantly longer contacts on dot-shaped AIS, with 90% of contacts lasting 2 hours or longer, whereas the corresponding fraction was 57% for NK cells adhering to donut-shaped AIS. The mean contact times measured were 199 min for dot-shaped AIS and 129 min for donut-shaped AIS (Fig. 3A). These results could be explained by differences in how the NK cell interacted with the AIS. On dots, NK cells often formed contacts with one side of the printed AIS or directly at the center and then spread out symmetrically over the printed area to eventually cover the entire AIS (Fig. 3B). On donut-shaped AIS, contacts were initiated on the antibody ring, but, because of the absence of activating ligands in the central region, gradual symmetrical spreading across the AIS was hindered. Instead, NK cells often moved along one side of the printed AIS (movie S2) or even both sides without spreading across the center (Fig. 3C). NK interactions with AIS were thus classified as either complete contacts, if the NK cell spread out to eventually cover the entire print (movie S3), or as partial contacts if not. Complete contacts were significantly more frequent on dot-shaped AIS (75%) compared to donut-shaped AIS (46%) (Fig. 3D). NK cells forming complete contacts on either dot- or donut-shaped AIS showed a similar behavior, with a mean duration of the interaction of 221 or 220 min, respectively. This was markedly longer than for partial contacts, which lasted 162 min for dots and 38 min for donuts (Fig. 3E). The spreading time of NK cells forming complete synapses, which is the duration between the initial contact with the AIS and complete coverage of the print, was significantly longer on donut- compared to dot-shaped AIS (Fig. 3F). Thus, the spatial distribution of αCD16 and αLFA-1 influenced the fraction of NK cells forming mature immune synapses by altering the process of NK cell spreading across the AIS. However, for NK cells reaching complete spreading, the contact stability was similar for dot- and donut-shaped AIS.

Fig. 3 NK cells more commonly build short-lived, partial contacts on donut-shaped αLFA-1 + αCD16 AIS.

(A) Duration of contacts made by NK cells on dot- or donut-shaped αLFA-1 + αCD16 AIS. Only contacts formed during the first 60 min were included, with a total assay time of 240 min. (B and C) Representative images of NK cells in complete (B) or partial (C) contact on donut-shaped αLFA-1 + αCD16 AIS (blue). NK cells were labeled for F-actin (red). Scale bars, 10 μm. (D) Proportion of contacts in which the NK cell reached complete, symmetric coverage of the AIS. (E) Duration of partial and complete contacts on dot- and donut-shaped AIS. (F) Spreading time of NK cells forming complete contacts on AIS, measured as the time between initial contact and reaching complete, symmetrical spreading over the entire AIS. Data in (A) and (E) represent n = 3 independent experiments, with 60 to 90 cells per condition per experiment. Data in (D) and (F) represent n = 6 independent experiments, with 60 to 90 cells per condition per experiment. Colored squares indicate the median measurement value for each independent experiment, with matching colors between paired experiments. Cluster plots in the background represent individual measurements. The P values in (A) and (D) to (F) were calculated using paired Student’s t test between the median of measurements for each experiment.

To confirm that this result was not restricted to the stimulation of NK cells by antibodies against CD16 and LFA-1, we also generated AIS containing the natural ligands of these receptors: rhICAM-1, which is recognized by LFA-1, and RTX, which contains the human Fc portion recognized by CD16 on NK cells (3537). Similar to what we observed on αLFA-1 + αCD16 AIS, NK cells formed stable synapses on rhICAM-1 + RTX AIS; however, fewer cells interacted with the prints and we observed larger variation between NK cells from different donors. In addition, when stimulated with the natural ligands, NK cells more often spread out to form a complete contact on dot-shaped AIS compared to donut-shaped AIS (fig. S2A), and this process took a longer time on donut-shaped compared to dot-shaped AIS (fig. S2B).

NK cell spreading on AIS correlates with Ca2+ response

Migration arrest of thymocytes upon antigen recognition is associated with transient increases in intracellular Ca2+ (38). To investigate how NK cell recognition and spreading are related to Ca2+ signaling, we loaded NK cells with the Ca2+-sensitive dyes fluo-4 and Fura Red and imaged the cells while they were interacting with dot- and donut-shaped AIS of αCD16 and αLFA-1. We observed strong Ca2+ fluxes upon synapse formation with AIS and throughout the spreading process, confirming NK cell activation by the AIS (Fig. 4A and movies S4 and S5). Most Ca2+ fluxes took the shape of a steep main Ca2+ peak, sometimes followed by a brief drop in fluorescence intensity and one or several secondary peaks, and a slower decay down to a plateau, commonly involving Ca2+ oscillations (Fig. 4B). In accordance with previous reports (15), the timing of the onset of the Ca2+ flux was most often associated with morphological changes indicating NK cell commitment to forming a synapse: cell body polarization and membrane spreading at the contact (Fig. 4C). A small fraction of the NK cells did not show a Ca2+ signal upon contact with AIS, possibly due to a lack or low amount of CD16. These cells almost exclusively formed short-lived partial contacts lacking NK cell commitment, whereas the vast majority of cells forming committed partial or complete contacts on dot- or donut-shaped AIS showed a Ca2+ signal (Fig. 4D).

Fig. 4 NK cell Ca2+ responses on αLFA-1 + αCD16 AIS coincide with commitment to the synapse and spreading.

(A) Time-lapse sequence of an NK cell building a complete contact on a dot-shaped αLFA-1 + αCD16 AIS (blue). The schematic illustrates the main stages of interaction as defined by NK morphological changes. Bright-field and fluorescence images show morphological changes and Ca2+activation. Scale bar, 10 μm. (B) Ca2+activity curve of the NK cell during the interaction depicted in (A), corresponding to the normalized ratio between the fluo-4 and Fura Red fluorescent signals. The different stages of the interaction are color coded as indicated. (C) Relative time between the different interaction stages and the onset of Ca2+ activity. Negative values indicate that the interaction stage happened before the onset of Ca2+ signaling. (D) Proportion of contacts exhibiting Ca2+ activity above the activation threshold, sorted depending on the interaction stage reached: “Complete” if symmetric spreading was achieved, “committed partial” if the NK cell committed but did not spread symmetrically across the AIS, or “noncommitted” otherwise. (E) Peak Ca2+amplitude in NK cells reaching complete or partial contact with either AIS shape, defined as the fold change between the Ca2+ activity at peak and at baseline (before contact initiation). Data in (C) to (E) represent n = 3 independent experiments, with 70 to 100 cells per condition per experiment. Colored squares indicate the median measurement value for each independent experiment, with matching colors between paired experiments. Cluster plots in the background represent individual measurements. The P values in (E) were calculated using paired Student’s t test between the median of measurements for each experiment.

Comparing NK cells that formed partial or complete contacts on either type of AIS (dot or donut), we observed higher Ca2+ amplitudes for NK cells that went on to spread out over the entire AIS regardless of AIS shape (Fig. 4E). Also, the spreading time of NK cells forming complete contacts inversely correlated with the amplitude of the initial Ca2+ peak (fig. S3C). We observed similar results for NK cells interacting with rhICAM-1 + RTX AIS (fig. S2, C and D). For NK cells interacting with αLFA-1 + αCD16 AIS, we further characterized the decay profile of the Ca2+ fluxes in individual cells by measuring the “sustained fraction,” defined as the fraction of time points where the Ca2+ signaling intensity was higher than half of the peak amplitude, over 20 min (60 imaging frames) after the initial Ca2+ peak (fig. S3D). Although sustained fractions were slightly higher in cells forming complete contacts on either AIS shape, no correlation was observed between sustained fraction and spreading time across the AIS (fig. S3, E and F). These results show that NK cell engagement is accompanied by Ca2+ signaling, with the amplitude of the initial Ca2+ peak, rather than sustained signaling, correlating with the spreading response and formation of a mature immune synapse.

NK cells form tight synapses with AIS despite local depletion of ligands in the center

Previous studies have described the formation of a tight immune synapse with a narrow synaptic cleft between the cells spanning up to 35 nm (39, 40). This has also been observed in artificial synapses established on coated surfaces or lipid bilayers (41). We used total internal reflection fluorescence (TIRF) microscopy to assess the distance of the synaptic interface formed between the NK cell and the AIS, analogous to the distance between the NK and target cell in a natural immune synapse. This technique excites fluorescence only within a very narrow section (≈130 nm) closest to the glass substrate and the AIS. We loaded NK cells that had interacted with αLFA-1 + αCD16 AIS for 30 min with the membrane dye CellMask Deep Red and performed TIRF imaging (Fig. 5, A and B). Selecting NK cells that had formed complete synapses on dot or donut AIS, we observed sustained membrane fluorescence intensity across both types of AIS, showing that the NK plasma membrane spread flat along the entire AIS surface.

Fig. 5 NK cells form a tight contact surrounded by a region enriched in F-actin on both dot- and donut-shaped αLFA-1 + αCD16 AIS.

(A and B) Representative TIRF images of the NK cell membrane during complete contacts on dot-shaped (A) or donut-shaped (B) AIS (blue). (C and D) Cell membrane radial distribution profile of NK cells on either dot-shaped (C) or donut-shaped (D) AIS. Individual curves represent four independent experiments, with 16 cells per condition per experiment. (E and F) Representative confocal images of F-actin in NK cells forming complete contacts on dot-shaped (E) or donut-shaped (F) AIS. (G and H) F-actin radial distribution profile of NK cells on either dot-shaped (G) or donut-shaped (H) AIS. Individual curves represent three independent experiments, with 50 to 125 cells per condition per experiment. Scale bars, 10 μm.

For a more detailed assessment, we analyzed the radial distribution of fluorescence across the AIS (Fig. 5, C and D). For donut-shaped AIS, the fluorescence intensity was slightly higher on the ring of antibodies but relatively even in the central region, confirming that the NK cell membrane was close to the entire AIS, despite the lack of specific ligand engagement in the inner region (Fig. 5D). Because the TIRF evanescent wave intensity decreases exponentially from the glass surface, intensity variations in TIRF images can be read as differences in distances from the surface for samples with evenly distributed fluorophores (42). The intensity profile on donut-shaped AIS is thus in good accordance with other descriptions of the synaptic cleft. However, exact distance measurements are difficult in our assay because membrane ruffling can also result in locally higher TIRF intensities (42). These results suggest that the complete spreading and formation of a tight and relatively uniform synaptic cleft of NK cells over dot- and donut-shaped AIS containing αLFA-1 and αCD16 is not only a mechanical consequence of local ligand engagement. Rather, NK cells integrate the signals from spatially separated ligands and respond by establishing a synapse, where tight contact can be maintained despite a local void of ligands.

Stable NK synapse formation is coupled to symmetric actin polymerization and PKC-Θ distribution

The correlation we observed between symmetric spreading and longer contact duration is in line with previous suggestions that radial symmetry and actin enrichment in the periphery are important for the stability of the immune synapse (43, 44). Imaging of NK cells fixed and stained for F-actin after 40-min incubation on αLFA-1 + αCD16 AIS showed that cells that had spread symmetrically had increased actin polymerization at the periphery of the contact (Fig. 5, E to H).

Protein kinase C Θ (PKC-Θ) is a principal kinase in T cell signaling and redistributes to the immune synapse upon activation of T and NK cells (45, 46). Because PKC-Θ has been shown to break synapse symmetry and promote migration in T cells (28, 43), we investigated the localization of PKC-Θ in NK cells that had formed complete contacts on αLFA-1 + αCD16 AIS of either shape. In accordance with previous observations in NK cells (45), we found that PKC-Θ organized into microclusters close to the synaptic interface (fig. S4, A and B). A radial analysis of the fluorescence intensity revealed that PKC-Θ was principally distributed in an annular structure at the interface between the cell body and lamellipodia on both dot- and donut-shaped AIS (fig. S4, C and D). This corresponds to the junction between the pSMAC and the dSMAC, which also correlates with the distribution of PKC-Θ in naïve T cells interacting with activating planar bilayers (43). The intensity of PKC-Θ was generally also higher around the MTOC as shown by directional analysis of the center of mass of intensity for MTOC and PKC-Θ (fig. S4, E and F), which is consistent with a role in positioning of the MTOC at the synapse (47). For NK cells forming asymmetric partial contacts with αLFA-1 + αCD16 AIS, PKC-Θ was often distributed around the MTOC but could also be found at other locations in the cells (fig. S4, G and H). Overall, the distribution appeared less regular in elongated NK cells compared to NK cells symmetrically spread on AIS prints. Thus, PKC-Θ assumed a circular distribution with particular weight around the MTOC in NK cells evenly spread across AIS of αCD16 and αLFA-1.

NK cells arrange the MTOC and lytic granules independently of central ligand engagement

We set out to investigate how the spatial distribution of ligands influenced the polarization of the lytic machinery toward activating AIS. We fixed NK cells interacting with dot- and donut-shaped αCD16 + αLFA-1 AIS and stained them for microtubules and the cytolytic granule marker perforin (Fig. 6A). We selected NK cells that had spread symmetrically across the AIS for analysis and determined the positions of the MTOC and lytic granules relative to the center of the AIS (Fig. 6B). In the vast majority of NK cells, the MTOC was polarized (z direction) toward the AIS (Fig. 6C) and slightly laterally displaced (xy plane) from the center of the AIS (Fig. 6D). These observations are consistent with previous reports that the MTOC is subject to mechanical forces resulting in an off-center position in the cytotoxic immune synapse (4851). We observed no difference in lateral positioning of the MTOC between dot- and donut-shaped AIS (Fig. 6D). The MTOC was located within the central, empty region in 74% of complete contacts on donut-shaped AIS (fig. S5, A and B). Thus, positioning of the MTOC seems to be regulated by integrated signals from the entire AIS rather than local signaling from engaged receptors.

Fig. 6 The distribution of αLFA-1 and αCD16 in dot- or donut-shaped AIS has little influence on the organization of the lytic machinery.

(A) Representative fluorescence images of cytotoxic granules (labeled for perforin, magenta) and microtubules (green) in NK cells building complete contacts on either dot- or donut-shaped αLFA-1 + αCD16 AIS (blue). Scale bars, 5 μm. (B) Schematic representation of the parameters describing the spatial organization of granules and MTOC. L1 represents the xy distance between the MTOC and the center of the AIS; L2 represents the distance between the MTOC and the centroid of the granule cloud. (C) Axial MTOC polarization on dot- and donut-shaped AIS, defined as the distance between the MTOC and the plane of the AIS. (D) MTOC lateral position, L1, on dot- and donut-shaped AIS. (E) Granule cloud offset from MTOC, L2, on dot- and donut-shaped AIS. (F) Granule cloud spread, defined as the average distance between individual granules and the centroid of their cloud, on dot- and donut-shaped AIS. All data represent n = 6 independent experiments, with 25 cells per condition per experiment. Colored squares indicate the median measurement value for each independent experiment. Cluster plots in the background represent individual measurements. The P values in (C) to (F) were calculated using unpaired Student’s t test between the median of measurements for each experiment.

We next determined the position of the lytic granules with respect to the MTOC. For both AIS shapes, the lytic granules were most often found in a tight cluster with a few satellites. For both dot- and donut-shaped AIS, lytic granules localized near the MTOC, with the cloud centroid less than 2 μm away from the MTOC (Fig. 6E). The average distance between individual lytic granules and the centroid of the cloud was about 2 to 3 μm (Fig. 6F), corresponding well with results found for NK cells activated through interactions with target cells (52). We noticed that the granule cloud was slightly more spread out on donut- compared to dot-shaped AIS (Fig. 6F), which could be attributed to a higher number of granules in the cells (fig. S5, C and D). The MTOC and granule cloud were, however, positioned similarly on both AIS shapes, with the granule cloud found most commonly on the outer side of the MTOC relative to the center of the AIS (fig. S5, E to G). Similar results were also obtained for NK cells forming contacts with AIS of natural ligands (fig. S2, E to H). Together, these results confirm that NK cells forming complete contacts on AIS organized their lytic machinery in a manner consistent with mature cytotoxic synapses, with the granules tightly packed around the MTOC and close to the synaptic interface. The similarity between dot- and donut-shaped AIS shows that this spatial organization is independent of receptor engagement in the center of the contact.

Degranulation is targeted to areas of ligand engagement

Having established that both granule convergence and polarization of the lytic machinery occur in complete synapses formed on either AIS shape, we sought to determine whether these synapses could support degranulation. For this purpose, glass substrates were coated with capture antibodies against perforin (capture αPrf) before microcontact printing with either dot- or donut-shaped AIS containing αLFA-1 and αCD16 and supplemented with capture αPrf to ensure capture across the entire AIS (fig. S6, A to H). NK cells were fluorescently labeled with Calcein Green and continuously monitored by time-lapse imaging during their interaction with the AIS. After the cells were enzymatically detached from the surface, captured perforin was detected by immunofluorescence, allowing us to correlate NK cell dynamics and contact formation with degranulation resulting from the formation of an activating synapse (Fig. 7A and movies S6 and S7). We measured substantial degranulation from NK cells having formed complete contacts on either AIS shape, confirming that the synapses were indeed mature and cytotoxic (Fig. 7B). NK cells making complete contacts on donut-shaped AIS degranulated less often than cells making complete contacts on dot-shaped AIS, but the amount of perforin captured for individual degranulation events was comparable between the two AIS shapes (Fig. 7, B and C). Looking at the spatial distribution of degranulation, we found that released perforin formed a tight cluster on both dot- and donut-shaped AIS (Fig. 7D). Dividing AIS into concentric regions (Fig. 7E), the perforin cluster on donut-shaped AIS most often localized on the ring containing αLFA-1 and αCD16, with some granules found in the central ligand–devoid region, whereas it localized closer to the center on dot-shaped AIS (Fig. 7F). This shows that degranulation was targeted toward areas with local signaling. These results confirm the maturity of complete synapses on either type of AIS, but they also indicate a role for the spatial distribution of ligands in regulating the cytotoxic outcome of the synapse.

Fig. 7 AIS shape influences the frequency and position of NK cell degranulation.

(A) Time-lapse sequences for NK cells (green) building complete contacts on dot- or donut-shaped αLFA-1 + αCD16 AIS (blue). Perforin (magenta) captured by antibodies attached to the glass substrate was imaged after the time-lapse series. Scale bar, 10 μm. (B) Proportion of contacts resulting in degranulation for NK cells having formed complete contacts on αLFA-1 + αCD16 dot- or donut-shaped AIS. (C) Amount of perforin captured from NK cells forming complete contacts on dot- or donut-shaped AIS, defined as the normalized integrated perforin fluorescent intensity over the AIS. (D) Spread of the cloud of captured perforin, defined as the average distance between individual perforin clusters and the centroid of the perforin cloud. (E) Schematic representation of the concentric regions used to describe the spatial distribution of captured perforin: 1, a 6-μm-wide central region of the AIS, corresponding to the void of ligands for donut-shaped AIS; 2, a 1-μm-wide ring corresponding to the inner border of donut-shaped AIS; 3, a 6- to 8-μm-wide ring corresponding to the main antibody-covered part of the donut-shaped print; 4, a 1-μm-wide band at the outer border of the AIS; 5, a rim surrounding the AIS, stretching from the edge of region 4 up to 10 μm from the AIS center. (F) Spatial distribution of individual perforin clusters into the regions defined in (E). Data represents n = 3 independent experiments, with 80 cells per group per experiment. (G) To investigate our hypothesis of hindered degranulation in the central void of ligands on donut-shaped AIS, we simulated the effect of degranulation hindrance in region 1 of dot-shaped AIS by excluding cells that degranulated to an area centered in region 1 from the data on dot-shaped AIS and reanalyzing the data. (H) Proportion of NK cells that degranulated after forming complete contacts on centrally hindered dot- and donut-shaped AIS. Cells on dot-shaped AIS that degranulated in region 1 were excluded from this analysis to simulate central hindrance. (I) Spread of the cloud of captured perforin on centrally hindered dot- or donut-shaped AIS. (J) Spatial distribution of individual perforin clusters on centrally hindered dot- and donut-shaped AIS, according to the regions defined in (E). Data represent n = 3 independent experiments, with 80 cells per condition per experiment. Colored squares indicate the median measurement value for each independent experiment, with matching colors between paired experiments. Cluster plots in the background represent individual measurements. The P values in (B) to (D), (H), and (I) were calculated using paired Student’s t test between the median of measurements for each experiment. The P values in (F) and (J) were calculated using paired Student’s t test between the median measured distance between the perforin centroid and the center of the AIS, for each experiment.

Because the lateral organization of granules and MTOC was almost identical between cells on both AIS shapes, the difference in the resulting degranulation profile at the surface must result from a difference in granule fusion (exocytosis) at the cell surface. We hypothesized that the position of the MTOC and granules is determined according to the integration of signaling from the entire synapse but that the actual release of granules is conditioned by local signaling. In donut-shaped AIS, this would result in granules targeted to the central region not being able to fuse and release their contents, with only granules docking close to ligand interactions being exocytosed and their content captured at the AIS surface. Therefore, practically all granules would be targeted to regions presenting signaling and could thus be released on dot-shaped AIS. To test our hypothesis, we sorted degranulating cells on dot-shaped AIS depending on the location of the centroid of their degranulation cloud, here used as a measure for the targeting of granules. We then excluded cells with perforin clouds aimed at the central region of the print (Fig. 7G). This hypothetical situation mimics what would happen on donut-shaped AIS if the granules were targeted exactly as on dot-shaped AIS but were only able to be released in the presence of local signaling near the site of docking. In this situation, the number of degranulating cells was strongly reduced, to levels comparable to that on donut-shaped AIS (Fig. 7H). The radial distribution profile of degranulation on the remaining cells was also nearly identical to that observed on donut-shaped AIS, as was the spread of the degranulation cloud (Fig. 7, I and J). These results support a model wherein NK cells use the overall shape of the AIS to position the lytic machinery but require local stimuli at the site of degranulation, suggesting the existence of a late signaling-dependent spatial checkpoint to complete the final steps of degranulation.


Here, we have used microcontact printing to pattern antibodies against NK activating receptors in layouts designed to mimic ligand presentation in the immune synapse and used time-lapse imaging to study the dynamic responses of NK cells making contacts with these AIS. In accordance with previous results involving NK cells interacting with antibodies presented on evenly coated surfaces (19) or in lipid bilayers (3, 53), we observed that ligation of LFA-1 induced a migratory response, with NK cells assuming elongated shapes. In this context of spatially separated synapses, this translated into a fraction of NK cells initiating contacts with several AIS simultaneously, sometimes facilitated by the formation of thin membrane tethers stretching between the distal AIS and the NK cell body. NK cells engaging LFA-1 alone commonly stayed in contact with a single AIS while actively seeking for subsequent contacts with the rest of the cell body, revealing a seemingly contradictory consequence of LFA-1 signaling, wherein both tethering and motility are supported in parallel. In contrast, combined ligation of LFA-1 and CD16 led to reduced NK cell migration often followed by spreading across a single AIS into symmetrical, stable contacts. This confirms the role of CD16 engagement in inducing NK cell commitment to the synapse, providing a stop signal that balances the motility signals from LFA-1.

We further investigated the importance of local signaling in the immune synapse formation and outcome by comparing the effect of CD16 + LFA-1 signaling in synapses with central ligand engagement (dot-shaped AIS) and in synapses with a central depletion of ligands (donut-shaped AIS). Altering the distribution of ligands affected the stability of contacts, because NK cells generally formed shorter-lived contacts on donut-shaped AIS. This could be explained by a higher proportion of partial contacts on donut-shaped AIS, in which the NK cells were unable to continue to spread across the AIS and thus to cover the central void of ligands. The fraction of NK cells spreading across AIS to establish complete contacts showed a higher Ca2+ activation response compared to cells forming only partial contacts. The difficulty for NK cells to reach a radially symmetrical configuration on donut-shaped AIS without spreading over the entire AIS often resulted in continued motion and eventually detachment, whereas spreading over dot-shaped AIS led to maintenance of symmetry and high stability (19, 44). Here, we used a static model system of immobilized antibodies or natural ligands, which is obviously very different from NK cell spreading across a target cell with a fluid membrane. Nevertheless, our results suggest that not only the nature of ligands but also their distribution across the target cell membrane could influence the outcome of NK cell surveillance by promoting or inhibiting NK cell spreading, possibly by altering the symmetry of the synapse, which could result in resumed migration (44, 54).

When we focused on the fraction of NK cells that went on to form complete contacts over the AIS, we found that their morphology, contact stability, Ca2+ signaling profile, plasma membrane organization at the interface, and positioning of the MTOC and perforin-containing granules were largely similar between dot- and donut-shaped AIS. NK cells forming complete contacts showed robust Ca2+ signals upon commitment to the contact with the AIS and adopted a symmetric, round morphology with actin accumulation at the cell periphery after spreading. The synapses proved to be very stable with contacts often lasting for several hours. TIRF imaging confirmed that the NK cell plasma membrane was in close proximity with the whole AIS surface with some local variation, suggesting either tighter contact or increased accumulation of membrane in the center for NK cells on dots and on top of the antibody ring for NK cells on donut-shaped AIS. In NK cells that had formed complete contacts with AIS, the MTOC, surrounded by a tight cluster of lytic granules, was commonly found within 2 μm from the AIS surface on both dot- and donut-shaped AIS, confirming the cytotoxic commitment of these cells (16, 55, 56). Also, the granule cloud and MTOC showed similar lateral organization on both AIS shapes, with the MTOC placed off-center, about 2.5 μm away from the center of the AIS. This off-center localization of the MTOC is consistent with the results from numerical simulations and polarization microscopy experiments on cytotoxic T cells, which indicated that pulling forces on microtubules cause the MTOC to be positioned about 2 μm from the center (50, 51). Thus, on donut-shaped AIS, the MTOC was most often found over a region devoid of ligands.

Together, the similarities observed between dot- and donut-shaped AIS indicate that once a complete contact is formed over the AIS, the assembly of the immune synapse is independent of the difference between dot- and donut-shaped prints. NK cells thus sense the overall circular shape of the print rather than the central part. We believe that the circular shapes of the AIS promoted symmetric spreading by facilitating a directional balance between the centripetal contractile forces [(19, 43, 44, 57); reviewed in (58)]. In NK cells forming complete contacts with either dot- or donut-shaped AIS, PKC-Θ, a principal kinase with a suggested role in synapse symmetry in T cells [(43, 46, 59); reviewed in (60)], formed symmetric and radial microclusters at the junction between the pSMAC and dSMAC, a distribution that was not found in NK cells that only partially covered an AIS or migrated on the glass between AIS. This observation agrees with the general view that the immune synapse, where the effector cell spreads symmetrically across the target cell and activating receptors and ligands accumulate into circular clusters, is a structure that promotes stability, sustained signaling, and preparation for effector functions.

Using a combination of live-cell time-lapse imaging and antibody capture of exocytosed perforin to associate NK cell–AIS interaction with degranulation, we measured substantial degranulation from NK cells forming complete contacts on both dot- and donut-shaped AIS, confirming that NK cell interaction with AIS could mimic NK-target cytotoxic synapses. Fewer cells degranulated in complete contacts on donut-shaped AIS compared to dots, which suggests a link between ligand distribution and the cytotoxic outcome of the synapse. This concept has been proposed for breast cancer cells wherein the induction of an “actin response” in reaction to NK cell recognition resulted in reduced cytotoxicity, possibly through modulation of the local concentration of NK cell ligands (61).

The radial distribution of captured perforin on the AIS showed that a population of cells degranulating toward the center of dot-shaped AIS was absent on donut-shaped AIS. In fixed NK cells, the lateral distribution of granules and the positioning of the MTOC were nearly identical between cells on both AIS, suggesting that the lytic machinery was being targeted to the same region independently of the central composition of the AIS. We hypothesized that the difference in the resulting degranulation profile at the surface resulted from differences in the subsequent steps leading to exocytosis and that this depended on local signaling. We tested this by excluding from the analysis NK cells in which the centroid of the degranulated perforin cloud was found in the center of the dot-shaped AIS. This brought down the hypothetical fraction of degranulating NK cells down to a similar level as that observed on donuts and the radial distribution of perforin to a profile similar to that measured on donuts. This supports the idea that local signaling is required for the final steps leading to CD16-dependent degranulation, in line with the observation that sites of granule release are commonly found in proximity with CD16 microclusters (62). A lack of local signaling in the predetermined area primed for degranulation would then result in impaired exocytosis, as observed in the central region on donut-shaped AIS.

The molecular mechanism for such a late signaling checkpoint for degranulation has not previously been described, but it is possible that it may be connected to the formation of hypodensities that are necessary for granules to pass through the actin mesh to reach the plasma membrane (63, 64) or to the recruitment of proteins important for the formation of SNAP receptor complexes, such as Munc18-2 and Syntaxin-11, which are required for granule fusion with the plasma membrane (6569). Syntaxin-11 is recruited to the immune synapse before lytic granule docking and delivered by vesicle-associated membrane protein-8 (VAMP8)-positive recycling endosomes (70, 71). Upon initial receptor engagement in cytotoxic T cells, essential signaling molecules are delivered by a similar endosomal pathway, including additional TCR, CD3, Lck, and linker of activated T cells (LAT) (7275). LAT-bearing VAMP7+ vesicles have recently been shown to dock in the vicinity of and interact with TCR-ZAP70 microclusters (76), suggesting that other recycling endosomes could follow a similar targeting mechanism toward sites of receptor engagement. The observed proximity of Syntaxin-11 and CD3 at the immune synapse could indeed be a clue that the docking machinery is at least partially recruited toward local signaling in T cells (77). Thus, the dependence of CD16 signaling on CD3ζ, ZAP70, and LAT activation in NK cells (78) motivates further investigation into exocytic mechanisms involved at the immune synapse and whether local recruitment of the docking and fusion machinery represents a final checkpoint for degranulation.

A limitation of this study is that we have not directly measured the dynamics of the MTOC and surrounding granules in NK cells forming contacts with AIS. It has been shown that the MTOC in T cells moves around with transient stops at off-center positions, likely representing unstable potential energy minima formed by forces generated by continuously rearranging microtubules (50, 51). Furthermore, in T cells, it has been debated whether granules can be transported by plus-ended motors toward the intercellular contact for degranulation (79), and in NK cells, synaptic vesicles can show high motility apparently searching for sites allowing degranulation (63, 80). It is probable that similar dynamics occur also in our system, which, in principle, could give NK cells on donut AIS the opportunity to rearrange their lytic machinery or transport individual granules until reaching areas permissible for degranulation. However, on the basis of the overall lower amount of degranulation found on donuts (Fig. 7, B and H) and the similar amount of perforin captured for NK cells on dots and donuts (Fig. 7C), we believe that the contributions from such mechanisms are limited in the current system.

In this study, we investigated how NK cells interact with static protein ligand prints sized as typical NK cell immune synapses and shaped as dots or donuts. We found that ligand distribution affected the NK cells’ ability to spread across the AIS, indicating that target cells could play an active role in inhibiting or promoting the establishment of an immune synapse by controlling the spatial distribution of the ligands. For cells establishing a complete contact, stability and assembly of the lytic machinery were regulated by the overall shape of the print, showing that NK cells can sense large-scale patterns and integrate spatially separated signals. Furthermore, local activating signals were important for the NK cells to proceed with degranulation, representing a spatial checkpoint in cytotoxicity. Together, these findings indicate that mechanisms that disrupt ligand assemblies in designated areas of the synapse could be a way for target cells to avoid attack. Therefore, the regulation of ligand dynamics in target cells during immune synapse formation deserves further studies.


Isolation and culture of NK cells

Human NK cells were isolated from blood from anonymous healthy donors according to local ethics regulations, following either of the following protocols. For experiments involving fixed NK cells on AIS, peripheral blood mononuclear cells (PBMCs) were separated from buffy coats by density gradient centrifugation (Ficoll-Paque, GE Healthcare). NK cells were then isolated from PBMCs by negative selection according to the manufacturer’s instructions (Miltenyi Biotec). For control of NK cell purity, cells were stained with monoclonal antibodies for CD56-PE (BioLegend, clone MEM-188) and CD3-FITC (BioLegend, clone OKT3) and analyzed with a FACSCalibur cytometer. For all other experiments, NK cells were directly isolated from buffy coats using negative magnetic selection according to the manufacturer’s instructions (STEMCELL Technologies). NK cells were then stained using the following antibodies: CD56-PE (BioLegend, clone HCD56), CD3-BV421 (BioLegend, clone UCHT1, or BD Horizon, clone SK7), CD16-APC (BD Pharmingen, clone 3G8), NKp46-PE-Cy7 (BD Pharmingen, clone 9E2/NKp46), and the viability dye BV510 (BD Biosciences), before purity analysis on a FACSCanto II flow cytometer (Becton Dickinson).

NK cells isolated using either protocol were maintained in RPMI cell culture medium containing 10% human AB+ serum, 1% penicillin-streptomycin, 2 mM l-glutamine, 1 mM sodium pyruvate, and 1× nonessential amino acids (all from Sigma-Aldrich). NK populations contained more than 95% CD56+ CD3 cells, of which 79 to 95% were found to be CD16+. NK cells were used within 48 hours after isolation.

Microcontact printing

Stamp masters for printing were produced by etching patterns in silicon using a previously described process (81). Stamps of poly(dimethylsiloxane) (PDMS; Sylgard 184, Dow Corning) were produced by casting prepolymer solution in silicon masters and curing at 65°C for at least 6 hours. To improve the uptake of the loading solution by the hydrophobic PDMS surface, stamps were washed with ethanol and washed and degassed twice in phosphate-buffered saline (PBS). Stamps were then incubated for 1 hour with (i) antibodies against LFA-1 (10 μg/ml) (αLFA-1; BioLegend, clone HI111), (ii) αLFA-1 (10 μg/ml) and antibodies against CD16 (10 μg/ml) (αCD16; BioLegend, clone 3G8), or (iii) rhICAM-1 (20 μg/ml) (R&D Systems) + RTX (20 μg/ml) (InvivoGen). All loading solutions were mixed in PBS and supplemented with bovine serum albumin (BSA) (10 μg/ml) conjugated to Alexa Fluor 555 (BSA-AF555; Thermo Fisher Scientific) for visualization. The stamps were then briefly washed with PBS and Milli-Q, dried, and placed on the substrate to print, with the structured surface facing the glass. Either a 35-mm glass-bottom dish (MatTek) or eight-chamber glass slide (ibidi) was coated with poly-lysine (precoated from factory or using reagent from Thermo Fisher Scientific). To reduce convection and cellular drift, a closed chamber with small volume was designed and used in experiments using natural ligands. For this, a 300-μm-thick double-sided adhesive tape was cut into an open 1-mm-wide ring with 16-mm diameter. This ring was placed on the 18-mm-diameter glass bottom of a 35-mm dish, and AIS were printed in the open central region of the ring. Two types of prints were used, either dot or donut shaped. The outer diameter of dot-shaped stamps was between 8 and 15 μm, whereas the inner and outer diameters of donut stamps were 6 to 7 and 14 to 15 μm, respectively. The center-to-center distance between prints was 18 to 32 μm for dot-shaped AIS and 22 to 32 μm for donut-shaped AIS. For experiments involving AIS of both shapes, outer AIS diameters and center-to-center distances were matched.

Live-cell migration and Ca2+ imaging

For migration analysis, NK cells were washed twice in PBS then incubated in 1 ml of RPMI with 1 μM Calcein Green (Thermo Fisher Scientific) for 20 min at 37°C, 5% CO2. For Ca2+ imaging, cells were washed twice with Hanks’ balanced salt solution (HBSS) (Invitrogen) before staining with 3 μM fluo-4 (Invitrogen) and 4 μM Fura Red (Thermo Fisher Scientific) for 30 min in RPMI. After washing in either PBS or HBSS, NK cells were left to rest for another 30 min before imaging. For experiments on αLFA-1 + αCD16 AIS, the cells were seeded at 2 × 105 to 3 × 105 cells/ml onto micropatterned glass in complete RPMI, supplemented with 10 mM Hepes (Sigma-Aldrich) for migration studies. Imaging was conducted using a 20× Plan-Apochromat objective on a LSM 880 confocal microscope (Carl Zeiss AG) with an incubation chamber set to 37°C, 5% CO2, capturing one frame every 45 to 60 s for 4 hours (migration) or every 20 s for 90 min (Ca2+ signaling). For experiments on rhICAM-1 + RTX AIS, NK cells stained with Ca2+ reporters were seeded in 20 μl of complete RPMI at 0.5 × 107 to 1.5 × 107 cells/ml, and the small chamber made by adhesive tape was immediately closed with an 18-mm-diameter coverslip before directly proceeding to imaging.

NK cell segmentation and single-cell tracking were performed using the “Pixel classification” and “Object tracking with learning” modules in Ilastik (82). For all duration measurements (contact duration and spreading time), only contacts formed within the first hour of the 4-hour assay were included to avoid measurement artifacts due to cells landing at the surface later in the assay. Different modes of NK cell migration were defined according to the mean square displacement (MSD) along the migration track as previously described (31). Briefly, the MSD was evaluated using a sliding window of 20 min. TMAPs were identified by comparing the mean diffusion coefficient along the curve, with the random diffusion coefficient estimated for a spherical particle of size comparable to a cell. Directed migration was detected by fitting the MSD to tα. α = 1 corresponds to a situation of random Brownian motion, and directed migration was characterized as α > 1.5 for at least 10 consecutive frames.

Ca2+ signaling curves were obtained by dividing the mean fluo-4 (fluorescent in its Ca2+-bound form) fluorescence intensity by the mean Fura Red (fluorescent in its Ca2+-free form) fluorescence intensity in individual tracked NK cells, at each time point. The resulting curves, here denoted as Ca2+ intensity curves, were then normalized to their value at the first time point, before synapse formation. To define an activation threshold for all cells, potential Ca2+ intensity peaks were identified by dividing the Ca2+ intensity at each time point to baseline measured before interaction (fold change). Receiver operating characteristic curves were then calculated using NK cells that did not interact with any AIS (true negatives) and a representative sample of NK cells exhibiting a clear Ca2+ intensity peak (true positives). The threshold was chosen to maximize the probability of detection (true positive rate) while minimizing the probability of false alarm (false positive rate) (83), leading to a threshold value of 2.63 (fig. S3, A and B). Contact times were manually measured in ImageJ [National Institutes of Health (NIH)], whereas individual migration tracks, MSD, and Ca2+ signaling curves were analyzed in MATLAB (MathWorks).

High-resolution confocal imaging of fixed NK cells on AIS

NK cells were seeded on the micropatterned glass bottom of 35-mm dishes (MatTek) at 5 × 105 cells/ml and were incubated for 40 min at 37°C and 5% CO2 in RPMI medium containing 1% human serum. Cells were fixed and permeabilized for 20 min in Fix/Perm solution (BD Biosciences) and washed using Perm/Wash buffer (BD Biosciences), followed by a blocking step with PBS supplemented with 5% goat serum for 60 min. For respective experiments, the cells were then stained with phalloidin conjugated to Abberior STAR 635 (Abberior), antibodies for α-tubulin conjugated to Alexa Fluor 488 (Millipore, clone DM1A), and for perforin conjugated to Pacific Blue (BioLegend, clone dG9). For PKC-Θ labeling, the cells were first stained with a primary rabbit antibody (Santa Cruz Biotechnology, polyclonal) followed by a goat anti-rabbit secondary antibody conjugated to Alexa Fluor 405 (Thermo Fisher Scientific). Cells were washed, and images were acquired using a 63×/1.40 Plan-Apochromat oil immersion objective on an LSM 880 confocal microscope (Carl Zeiss AG).

The contact area of NK cells on AIS was obtained using ImageJ (NIH) by thresholding the phalloidin fluorescence intensity in the z plane of the print, followed by topologically closing and filling holes in the obtained mask. Similarly, NK cell roundness was measured on thresholded maximum intensity projections of the phalloidin channel. To determine the position of the MTOC and lytic granules relative to the AIS center on complete contacts, single NK cells with roundness over 0.8 covering at least 90% of the AIS were selected. The position of the MTOC was manually set as the convergence point of microtubules, often corresponding to the α-tubulin cluster of highest intensity in the cell. Lytic granules were detected in Volocity (PerkinElmer) by segmenting clusters of perforin, and the position of each was defined as the fluorescence intensity–weighted center of mass. The granule cluster centroid was defined as the cumulated intensity-weighted center of mass of all individual granules in the cell. The average distances between individual granules and other objects were always weighted to granule cumulated intensity, thus compensating for segmenting effects (12). The relative position of the MTOC and the granule cluster was analyzed by measuring the angle between two vectors, both starting from the AIS center, but one drawn to the MTOC and the other to the granule cluster centroid. Radial fluorescence intensity profiles of F-actin and PKC-Θ were generated from single-slice images (F-actin) or maximum intensity projection images (PKC-Θ) using the Radial Profile plugin in ImageJ (NIH).

TIRF imaging of fixed NK cells on AIS

NK cells were seeded on the micropatterned glass bottom of 35-mm dishes (MatTek) at 5 × 105 cells/ml and left to interact for 30 min. NK cell outer membranes were then stained with CellMask Deep Red (Thermo Fisher Scientific) at 10 μg/ml for 10 min and washed with PBS before 10-min fixation with Cytofix (BD Biosciences) containing 4% paraformaldehyde. The samples were then blocked using PBS with 5% (w/v) BSA and lastly mounted for immediate imaging. Images were acquired using an Elyra S1 microscope equipped with an α Plan-Apochromat 100×/1.46 oil immersion objective (Carl Zeiss AG). NK cells having spread symmetrically over the entire AIS were selected by wide-field imaging before switching to TIRF mode where the imaging focus plane was set according to the AIS.

To assess whether NK cells’ plasma membrane were in close proximity to the surface across the AIS, radial fluorescence intensity profiles of the NK cell membrane over the prints were generated using the Radial Profile plugin in ImageJ (NIH). The resulting curves were normalized in radius (0 to 1) and in intensity (integrated fluorescence intensity set to 1) to account for individual staining differences.

Perforin capture on AIS

Perforin capture was performed using a protocol inspired by Srpan et al. (84), adapted for use with AIS. Capture surfaces were prepared by successively coating eight-chamber glass slides (ibidi) with poly-l-lysine (Thermo Fisher Scientific) and anti-perforin capture antibody mix (5 μg/ml) (capture αPrf; Mabtech, combination of clones Pf-80/164). The resulting surfaces were micropatterned as described before with αLFA-1 (10 μg/ml), αCD16 (10 μg/ml), capture αPrf (5 μg/ml), and BSA-AF555 (10 μg/ml). NK cells labeled with 1 μM Calcein Green were added to the chambers at 2 × 105 to 3 × 105 cells/ml. During their interaction with AIS, cells were imaged every 2 to 5 min using a 20× Plan-Apochromat objective on a Cell Observer 7 wide-field microscope (Carl Zeiss AG). After 60 min, the cells were detached using the Accumax detachment solution (Merck Millipore) at 37°C, 5% CO2 for 15 min. Cells were washed away with PBS, and complete cell removal was confirmed by visual inspection using a phase contrast microscope (Carl Zeiss AG). The slides were then washed twice with enzyme-linked immunosorbent assay (ELISA) buffer [PBS, 0.05% Tween-20, and 0.1% (w/v) BSA]. The detection antibody solution composed of PBS and anti-perforin conjugated to biotin (5 μg/ml) (clone Pf-344, Mabtech) was added to the slide chambers. After washing with buffer, captured perforin was revealed by labeling with streptavidin conjugated with Alexa Fluor 647 (BioLegend) at 1 μg/ml in PBS, and the slides were washed again with ELISA buffer solution. The surfaces were then taken back to the microscope, and the slides were realigned so that the same regions of interest were used for degranulation measurements as for migration.

The images were analyzed using ImageJ (NIH). Briefly, time-lapse and perforin detection images were combined and spatially aligned to account for stage drift and imprecision when taking the slide off and on the microscope. On AIS where a single NK cell had established a complete contact, degranulation distribution was measured in a 20-μm-wide circular region centered on the print.

Statistical analysis and data plotting

Following recently published suggestions (85), we chose to only compare summary measures (here set to the median of all individual measurements) between independent experiments as opposed to comparing pooled data from the independent experiments. The plots reflect this, as the large colored squares indicate the median measurement value for each independent experiment, and the cluster plots in the background (where applicable) show the corresponding individual measurements. Paired experiments are color coded. The thicker black line (or bar height where applicable) indicates the mean of summary measures from independent experiments, and the error bars represent the SEM of summary measures. All P values were calculated using (paired where applicable, unpaired otherwise) Student’s t test between the summary measures of indicated groups. P values were calculated and plots were produced using Prism (GraphPad) or MATLAB (MathWorks).


Fig. S1. Interaction of NK cells with surfaces evenly coated with αLFA-1 or αLFA-1 + αCD16.

Fig. S2. Responses of NK cells to dot- and donut-shaped rhICAM-1 + RTX AIS.

Fig. S3. Correlation between NK cell spreading dynamics and Ca2+ activation.

Fig. S4. Intracellular organization of PKC-Θ in NK cells on αLFA-1 + αCD16 AIS.

Fig. S5. Relative position of lytic granules and MTOC on αLFA-1 + αCD16 AIS.

Fig. S6. Perforin capture on donut-shaped αLFA-1 + αCD16 AIS.

Movie S1. Oscillations between two αLFA-1 AIS.

Movie S2. Complete contact on donut-shaped AIS.

Movie S3. Partial contact on donut-shaped AIS.

Movie S4. Ca2+ activation on dot-shaped AIS.

Movie S5. Ca2+ activation on donut-shaped AIS.

Movie S6. Perforin capture on dot-shaped AIS.

Movie S7. Perforin capture on donut-shaped AIS.


Acknowledgments: We thank the Önfelt lab and D. Davis for discussions and feedback on the manuscript and D. Jans for introduction to TIRF experiments. Funding: We thank the Swedish Foundation for Strategic Research (SBE13-0092), the Knut and Alice Wallenberg Foundation (KAW 2018.0106), the Swedish Research Council (2019-04925), the Swedish Cancer Foundation (CAN 2016/730, 19 0540 Pj), and the Swedish Childhood Cancer Foundation (MT2019-0022) for financial support. Author contributions: Q.V. and E.F. designed, performed, and analyzed experiments and wrote the manuscript. M.L. performed and analyzed experiments. L.B. and P.E.O. analyzed experiments. T.W.F. manufactured the silicon masters. B.Ö. conceptualized the study and wrote the manuscript. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper or the Supplementary Materials.

Stay Connected to Science Signaling

Navigate This Article