Integrins Influence the Size and Dynamics of Signaling Microclusters in a Pyk2-dependent Manner*

Background: The mechanism by which integrin signaling regulates lymphocyte cytolytic activity is poorly understood. Results: The extent of integrin ligation influences the size and dynamics of signaling microclusters. Conclusion: Integrins regulate the magnitude and duration of proximal signaling that affect lymphocyte degranulation kinetics and the rate of target cell lysis. Significance: Integrin engagement ensures more rapid cytolytic granule release and faster target cell destruction. Integrin engagement on lymphocytes initiates “outside-in” signaling that is required for cytoskeleton remodeling and the formation of the synaptic interface. However, the mechanism by which the “outside-in” signal contributes to receptor-mediated intracellular signaling that regulates the kinetics of granule delivery and efficiency of cytolytic activity is not well understood. We have found that variations in ICAM-1 expression on tumor cells influence killing kinetics of these cells by CD16.NK-92 cytolytic effectors suggesting that changes in integrin ligation on the effector cells regulate the kinetics of cytolytic activity by the effector cells. To understand how variations of the integrin receptor ligation may alter cytolytic activity of CD16.NK-92 cells, we analyzed molecular events at the contact area of these cells exposed to planar lipid bilayers that display integrin ligands at different densities and activating CD16-specific antibodies. Changes in the extent of integrin ligation on CD16.NK-92 cells at the cell/bilayer interface revealed that the integrin signal influences the size and the dynamics of activating receptor microclusters in a Pyk2-dependent manner. Integrin-mediated changes of the intracellular signaling significantly affected the kinetics of degranulation of CD16.NK-92 cells providing evidence that integrins regulate the rate of target cell destruction in antibody-dependent cell cytotoxicity (ADCC).


Integrin engagement on lymphocytes initiates "outside-in"
signaling that is required for cytoskeleton remodeling and the formation of the synaptic interface. However, the mechanism by which the "outside-in" signal contributes to receptor-mediated intracellular signaling that regulates the kinetics of granule delivery and efficiency of cytolytic activity is not well understood. We have found that variations in ICAM-1 expression on tumor cells influence killing kinetics of these cells by CD16.NK-92 cytolytic effectors suggesting that changes in integrin ligation on the effector cells regulate the kinetics of cytolytic activity by the effector cells. To understand how variations of the integrin receptor ligation may alter cytolytic activity of CD16.NK-92 cells, we analyzed molecular events at the contact area of these cells exposed to planar lipid bilayers that display integrin ligands at different densities and activating CD16specific antibodies. Changes in the extent of integrin ligation on CD16.NK-92 cells at the cell/bilayer interface revealed that the integrin signal influences the size and the dynamics of activating receptor microclusters in a Pyk2-dependent manner. Integrinmediated changes of the intracellular signaling significantly affected the kinetics of degranulation of CD16.NK-92 cells providing evidence that integrins regulate the rate of target cell destruction in antibody-dependent cell cytotoxicity (ADCC).
A great deal of evidence demonstrates that adhesion molecules function as a molecular glue to mediate contact between immune cells and target cells. However, they also mediate "outside-in" signaling regulating functions of immune cells. Particularly, ligation of lymphocyte function-associated antigen-1 (LFA-1 2 or ␣ L ␤ 2 integrin) by intercellular adhesion molecule 1 (ICAM-1) is necessary to induce cytoskeleton remodeling and formation of a well-organized immune cell/target cell interface (1)(2)(3). The formation of such interface is especially important for cytotoxic lymphocytes to exercise efficient cytolytic activity against target cells via delivery of cytolytic granules (4 -6). However, how specific engagement of the integrin receptors regulates kinetics of granule delivery and efficiency of cytolytic activity is not well understood.

EXPERIMENTAL PROCEDURES
Cells, Antibodies, and Chemicals-The human NK-92 cell line transduced to express Fc␥RIIIa receptor (176V allele of CD16 that binds to the antibody Fc fragment with high affinity) has been described in detail elsewhere (7,8). The human breast cancer cell line SKBR3 was supplied by ATCC. In some experiments, SKBR3 cells were treated with 100 units/ml IFN-␥ for 48 h prior to the assay.
Herceptin, humanized antibody against human HER2/neu used in this study have been previously described (7,9). Hybridomas producing mouse blocking antibody TS1/18 specific for human CD18 (␤ 2 integrin chain), mouse non-blocking TS2/4 antibody against human CD11a (␣ L integrin chain), mouse blocking antibody R6.5 specific for human ICAM-1, and mouse 3G8 antibody specific for human CD16 were purchased from ATCC. The hybridoma producing mouse H4A3 antibody specific for human CD107a was a kind gift from Dr. J. Thomas August. Purified H4A3 and 3G8 IgG were labeled with Alexa Fluor 568 and Alexa Fluor 488 (Molecular Probes) according to the manufacturer's instruction. Monovalent Fab fragments of anti-CD107a antibody were produced and labeled with Alexa Fluor 568 as described (10,11). Rabbit polyclonal antibody specific for phospho-Src family kinases (Tyr-416) was supplied by Cell Signaling Technology. Alexa Fluor 546 goat anti-rabbit IgG F(ab)Ј 2 were obtained from Invitrogen.
Granule Release Assay-SKBR3 cells were plated in 96-well flat-bottom plates at sub-confluent density and cultured in the presence or absence of IFN-␥ for 48h prior to the assay. CD16.NK-92 cells were then added to the monolayers of SKBR3 cells at E:T ratio 1:1 in the presence of 1 g/ml Herceptin antibody. Degranulation of CD16.NK-92 cells was detected at various time points by staining the cells with Alexa 568conjugated CD107a-specific antibody followed by flow cytometry analysis on LSRII instrument (BD Biosciences).
Conjugate Formation Assay-SKBR3 cells at density 5 ϫ 10 5 cells/ml were labeled with 0.1 M Calcein-AM in the presence of 10 g/ml Herceptin for 1 h in the dark, and washed free of unreacted reagents. CD16.NK92 cells at 1 ϫ 10 7 cells/ml were labeled with PKH26 according to manufacturer's instruction, for 5 min, washed three times, and incubated in serum-containing medium for 1 h at 37°C & 5% CO 2 to remove excess of the dye. Stained cells were then mixed at E:T ratio of 2:1, and after a short centrifugation at 300 rpm the conjugates were allowed to form for different periods of time at 37°C and 5% CO 2 . The samples were fixed with PFA, and the conjugates were identified by double staining using LSRII flow cytometer (BD Biosciences).
Streptavidin (2 g/ml) and monobiotinylated anti-CD16 (3G8) monoclonal antibody labeled with Alexa Fluor 488 (2 g/ml) were reacted sequentially with the biotinylated bilayers (14) to produce the antibody density of 50 molecules/m 2 , unless indicated otherwise. In some experiments the density of anti-CD16 on bilayer was varied from 5 to 500 molecules/m 2 by changing the concentration of biotinyl-CAP lipids in the bilayer from 0.001 mol% to 0.3 mol%, respectively. Where indicated, Cy5-ICAM-1-His 6 molecules were incorporated into the bilayers at the density 300 molecules/m 2 , a physiological density of this molecule on the surface of APC (4,10,11,13). Densities of Cy5-ICAM-1 and anti-CD16 antibody on the bilayers were determined as described elsewhere (4). In one set of experiments, the density of Cy5-ICAM-1 was varied while keeping the density of the anti-CD16 unchanged.
Imaging-CD16.NK-92 cells were injected into pre-warmed to 37°C flow cell. In some experiments the cells were treated with Pyk2 inhibitor. The inhibitor was present in extracellular medium at concentrations ranging from 1 to 10 M. IRM images of the cells that formed tight contact with the bilayers were obtained using a confocal fluorescence microscope (Zeiss LSM510 Meta confocal Laser Scanning Microscope). Images of CD16 clusters bound to fluorescent anti-CD16 antibodies were acquired with a TIRF microscope (Andor Revolution XD system equipped with Nikon TIRF-E illuminator and 100ϫ/1.49 NA objective, Yokogawa CSU-X1 spinning disk confocal, and Andor iXon X3 EM-CCD camera). To image activated Src kinases, CD16.NK-92 cells attached to the bilayers were fixed with 2% paraformaldehyde, permeabilized with 0.1% Triton X-100 and stained for phospho-Src (pSrc) kinases with anti-Tyr416 rabbit polyclonal antibody followed by Alexa Fluor 568labeled secondary goat anti-rabbit antibody. Fixed and stained cells were then imaged by TIRF microscopy.
To ensure accurate comparison, the measurements were performed in a series of parallel experiments on the same day using the same reagents and freshly prepared planar bilayers. The images were acquired using the identical microscope settings. The same batches of the cells were utilized in all experiments.
Events Timing and Degranulation Kinetics at CD16.NK92/ Bilayers Interface-The timing of ICAM-1 accumulation, signaling microcluster formation and the onset of granule release have been measured in individual CD16.NK-92 cells interacting with the bilayers. Mean timing for each of the events was calculated by averaging the observed timings for more than 20 cells. For measuring the degranulation kinetics, percentage of degranulating cells at various time points in several imaging fields was determined.
Evaluation of CD16 Microcluster Size-To characterize moving microclusters we analyzed multiple images in each field during the period of 90 min. The images were taken every 10 or 30 s. We defined microcluster as a coherent group of bright pixels. Moving microclusters were defined as those observed on several consequent imaging planes with little or no significant change in shape, and the position of its center at least once over the period of observation is located outside of the original position.
For analysis of the size of individual CD16 microclusters, we examined the series of images taken every 30 s. To single out individual microclusters, intensity profiles of each of the moving microclusters were examined. Microclusters that demonstrated unimodal intensity distribution (see supplemental Information) were considered as individual microclusters. We then calculated average background fluorescent level for each of the analyzed cells. Each coherent pixel group with all pixels' intensities equal or exceeding 1.3 times the fluorescent background level was considered having a positive signal and included in the analysis. Regions were drawn around each selected individual moving microcluster at image planes corresponding to 5, 10, or 30 min after the initial cell-bilayer contact. MetaMorph Region Measurements function was utilized to measure sizes and intensities of these microclusters. Average values for each of these parameters were calculated and expressed in m 2 or arbitrary units, respectively.
Analysis of CD16 Microclusters Movement-Parameters of microcluster movement were studied using the series of sequential images collected every 10 s. For each cell we determined the time difference between initiation and termination of centripetal microcluster movement. This time difference was termed mobility. We then examined the two planes corresponding to the beginning and the end of the microcluster movement and measured the distance that each microcluster traveled. The distance was termed microcluster displacement. We calculated average mobility and displacements for at least 20 cells at each experimental condition.
Quantification of Activated Src Kinases Microclusters-The average size of activated Src kinases (pSrc) clusters and average integrated fluorescence intensities of the microclusters in each cell were calculated and expressed in m 2 and arbitrary units, respectively. We then evaluated total amount of the signaling molecules and their average density at the contact area of CD16.NK-92 cells exposed to the bilayers. For this purpose, the smallest possible circular regions were drawn to include all the pSrc clusters observed at the contact area of each cell. Integrated fluorescence intensity (the parameter proportional to total amount of the fluorescent molecules recruited to the cell/ bilayer interface) and the area were determined for these regions using the Region measurement function of MetaMorph Software suite. The integrated fluorescence of pSrc molecules and pSrc density at the contact area of each cell were calculated, and average values are expressed in arbitrary units and arbitrary units per m 2 , respectively.

Ligation of ␤ 2 Integrins Accelerates the Kinetics of Target Cell
Killing by CD16.NK-92 Cells-We examined the kinetics of Herceptin-induced killing of SKBR3 breast cancer cells mediated by CD16.NK-92 effectors. SKBR3 cells have a low level of ICAM-1 that serves as a ligand for ␤ 2 integrins on CD16.NK-92 cells (7). We have shown that treatment of SKBR3 with IFN-␥ for 48 h results in significant up-regulation of cell-surface ICAM-1 molecules but not Her2/neu receptor (7). We compared the extent of Herceptin-induced specific lysis of IFN-␥ treated or untreated SKBR3 cells by CD16.NK-92 effectors at various time-points. The killing rate was determined as the initial slope of the plots (Fig. 1A). The up-regulation of ICAM-1 resulted in ϳ2-fold increase of the rate of target cell killing by CD16.NK-92 (Fig. 1B). Because IFN-␥ treatment did not change Her2/neu receptor expression on SKBR3 cells, the observed increase in ADCC rate cannot be attributed to an elevated level of the epitope recognizable by the antibody. However, IFN-␥ treatment could also up-regulate expression of HLA-E, which is a ligand for CD94/NKG2A receptor on NK-92 cells, and modulates expression of other activating receptor ligands (16). To test the role of ICAM-1 in the observed increase of the ADCC, we utilized TS1/18 antibody that blocks specific interactions of all ␤ 2 integrins with their ligands (17), TS2/4 antibody that also bind to the integrins but are not inhibitory (18), and R6.5 blocking antibody against ICAM-1 (19). The presence of TS1/18 antibodies in the extracellular medium completely abrogated the lysis of IFN-␥-treated SKBR3 cells (Fig. 1A), whereas TS2/4 antibodies had no effect on the rate of the killing, and anti-ICAM-1 antibody significantly reduced the kinetics and magnitude of target cell lysis (supplemental Fig.  S1). The killing of IFN-␥-treated SKBR3 cells was also diminished after lowering expression of ICAM-1 with ICAM-1 siRNA (supplemental Fig. S2). These data provide evidence that increasing the extent of ␤ 2 integrin ligation by ICAM-1 results in significant acceleration of the killing kinetics of tumor cell destruction by CD16.NK-92 cytolytic effectors in ADCC responses.
The Level of ICAM-1 on Target Cells Influences Conjugate Formation and the Kinetics of Cytolytic Granule Release by CD16.NK-92 Cells-Variations in the ICAM-1 level on target cells could affect the killing kinetics in two principle ways. First, a higher extent of ␤ 2 integrin engagement by ICAM-1 could merely enhance effector/target cell conjugate formation resulting in more efficient killing. Second, increasing the ␤ 2 integrin ligation could potentiate the integrin-mediated signaling, accelerating recruitment and release of cytolytic granules. The latter is consistent with the increase of the killing rate of SKBR3 cells after ICAM-1 up-regulation (Fig. 1B).
We did indeed find that a lower level of ICAM-1 expression on SKBR3 cells impaired the ability of CD16.NK-92 effectors to form conjugates with these cells (supplemental Fig. S3). But we also examined the kinetics of granule release by CD16.NK-92 cells exposed to monolayers of SKBR-3 cells in the presence of Herceptin. The target cells were either untreated or treated with IFN-␥ as above to vary the level of cell surface ICAM-1. Degranulation was measured by the appearance of CD107 (LAMP-1) molecules on the surface of the effector cells exposed to monolayers of the target cells at different time points. Fig. 1C shows that the percentage of degranulating CD16.NK-92 cells and average amount of granules released by individual effector cells responding to SKBR3 with elevated levels of ICAM-1 was substantially higher at every time point. The observed difference suggested that ␤ 2 integrin mediated signaling enhances the kinetics of granule release (Fig. 1D). Consistent with this, inhibition of Pyk2, a proximal signaling kinase in the integrin-mediated signaling pathway (20), resulted in a decrease of the specific lysis of SKBR3 targets by CD16.NK-92 cells (supplemental Fig. S4A). The inhibitory effect was dose-dependent and was more profound for killing of IFN-␥-treated SKBR3 cells (supplemental Fig. S4B).
␤ 2 -Integrins Facilitate Cellular Adhesion and the Kinetics of Cytolytic Granule Release at the CD16.NK-92/Bilayer Interface-To dissect the influence of ␤ 2 integrins on adhesion and kinetics of granule delivery at the single cell level, we analyzed interactions of CD16.NK-92 cells to planar lipid bilayers containing ligand for the CD16 receptor in the presence or absence of ICAM-1 and visualized degranulation of individual cells, essentially as we have previously described in cytotoxic T cells (4,10,11).
Although CD16.NK-92 cells formed contacts with the bilayers regardless of the presence of ICAM-1, both the percentage of cells contacting the bilayers ( Fig. 2A) and the adhesion area (Fig. 2, B and C) were significantly larger on ICAM-1-containing bilayers. Importantly, 90% of the cells that established contact with ICAM-1-containing bilayers degranulated, while only 60% of the cells contacting bilayers without ICAM-1 did so ( Fig.  2A). These data suggest that in addition to promoting cellular adhesion, ␤ 2 integrins induce a signal facilitating granule delivery to the contact surface.
We then examined the kinetics of granule release at the CD16.NK-92/bilayer interface by TIRF microscopy (Fig. 3A). Granule release was observed at distinct locations that were preserved during the time of observation ( Fig. 3A and supplemental Fig. S5). These locations were adjacent to, but did not overlap with the clusters of CD16 receptors (Fig. 3A). On average, the granule release by CD16.NK-92 cells on ICAM-1-containing bilayers was observed 5 min after the cells were brought in contact with the bilayers, while granule release by CD16.NK-92 cells interacting with bilayers that do not present ICAM-1 was delayed until 10 min ( Fig. 3A and supplemental Fig. S6). The kinetics of granule release was assessed by measuring the fraction of degranulating cells as a function of time followed by the appearance of the CD16 microclusters. The amount of time between formation of CD16 microclusters and the release of the granules in the presence of ICAM-1 was 3.3 times shorter (Fig. 3B). These data reiterate findings produced with live target cells (see Fig. 1, C and D) and provide evidence that integrin-induced signaling regulates the kinetics of cytolytic granule release by CD16.NK-92 effectors.
Analysis of the Dynamics of Activating Microclusters-It is well established that proximal signaling mediated by antigenspecific receptors in T and B lymphocytes is compartmentalized and occurs in signaling microclusters containing activating receptors (21)(22)(23)(24)(25)(26). To understand mechanism by which ␤ 2 integrins influence intracellular signaling from activating receptors that regulates the kinetics of granule delivery and release, we analyzed the dynamics of CD16-containing microclusters at the CD16.NK-92/lipid bilayer interface in the presence and absence of ICAM-1.
Upon initial contact of CD16.NK-92 cells with the bilayers, several undersized CD16-containing activating microclusters were formed in the center of a very small contact area. The contact area containing the microclusters was subsequently enlarged during the first 1.5-2 min after the initial contact. Within this period, the newly formed microclusters were small and remained stationary over the entire area of cell/bilayer interface. Then the microclusters began to grow in size and started to move centripetally (supplemental Movie S1 and Movie S2). Once centripetal movement of a microcluster had begun, new microclusters were observed to be formed in its place. Distances that the microclusters traveled were different for each microcluster and depended on the location of their initial formation. The majority of the microclusters was formed on the periphery and traveled longer distances, while those formed in the center remained almost completely stationary.
The movement of microclusters continued for about 10 -15 min (supplemental Movies S1 and S2). The moving microclusters then coalesced into distinct, well-structured donut-shaped structures ( Fig. 3A and supplemental Fig. S7). As pointed out above, the microclusters were adjacent to but did not overlap with the sites of granule release, and granules were released within the donut-shaped aggregates (Fig. 3A). The observed pattern of microclusters and adjacent granule release sites at the interface resembled multifocal synapses (6,27).
␤ 2 -Integrins Ligation Controls the Size of the Microclusters-To characterize CD16-containing signaling microclusters in CD16.NK-92 cytolytic effectors, we evaluated the size of the microclusters by either directly measuring their linear dimensions (supplemental Fig. S8) or from the integrated fluorescence intensity of the observed microclusters (supplemental Fig. S9). By either measure, we found that the area of microclusters on ICAM-1-containing bilayers was almost twice as large as the area of microclusters on bilayers without ICAM-1 ( Fig.  4A and supplemental Fig. S9). The observed difference remained the same for up to 30 min (Fig. 4B). Importantly, the Dark areas correspond to tight contact between the cells and bilayers. C, average areas of tight contact formed by CD16.NK-92 cells exposed to the bilayers with (slashed bars) or without (empty bars) ICAM-1 are shown. Regions were drawn around tight adhesion areas on the IRM images, and areas of those regions were determined. Representative results of three independent experiments with at least 20 cells analyzed in each group; the data are shown as mean Ϯ S.D., p Ͻ 0.0001 by two-tailed Student's t test. size of individual microclusters did not depend on the changes of the density of anti-CD16 antibodies incorporated into the bilayers over a 100-fold range, but was linked to the presence or absence of ICAM-1 (Fig. 4, C and D). These data suggest that the ligation of ␤ 2 -integrins triggers outside-in signaling that influence the size of CD16-activating microclusters. To provide evidence for the role of integrin-mediated signaling in controlling the size of microclusters, we first varied the extent of integrin ligation on CD16.NK-92 cells by changing the density of ICAM-1 on planar bilayers and found that lowering the density below 60 mol/m 2 led to a statistically significant decrease in size of the activating microclusters (Fig. 4E). It has to be noted that regardless of ICAM-1 density, we observed accumulation of ICAM-1 molecules at the cell/bilayer interface indicative of the productive engagement of integrins on the surface of the effector cells. We then examined the effect of the Pyk2 inhibitor that diminishes proximal integrin-mediated signaling. Fig. 4F shows that treatment of CD16-NK-92 cells with the inhibitor caused a decrease of the size of signaling microclusters at the CD16-NK-92/bilayers interface. Because the size of CD16 signaling microclusters correlates with the number of activating receptors recruited to each microcluster, the data provide evidence that ␤ 2 -integrin-mediated signaling could effectively modulate the proximal signaling from activating receptor, which is linked to the kinetics of cytolytic granule release and the efficiency of NK cell cytolytic activity (6,11,28).
␤ 2 -Integrins Influence the Microcluster Displacement and Mobility-Because microclusters signal when they are on the move (23), those that move long distances are expected to contribute more to the magnitude and kinetics of proximal signaling. This prompted us to investigate parameters that are associated with the observed movement of signaling microclusters ( Fig. 5A and supplemental Movie S3). The first parameter analyzed is the shortest distance from the initial to the end point of microcluster travel at the synaptic interface, which was termed microcluster displacement. Thus, microcluster displacement is the length of an imaginary straight path, which is typically distinct from the path that microclusters actually travel ( Fig. 5A  and supplemental Fig. S10). The second parameter analyzed is the average time period within which individual microclusters are moving. This parameter was called microcluster mobility (supplemental Fig. S10).
We found that the microcluster displacement correlated with the size of the interface and the location of initial microcluster formation within the synaptic interface. CD16.NK-92 exposed to ICAM-1 containing bilayers formed a more sizable adhesion area that increases the potential for a microcluster to travel a longer distance. Indeed, an average displacement of microclusters at the contact area of CD16.NK-92 cells exposed to bilayers containing ICAM-1 at 200 -300 mol/m 2 was 1.5 times longer than the microcluster displacement at the contact area of CD16.NK-92 exposed to bilayers without ICAM-1, i.e. 7 m and 5 m, respectively (Fig. 5B). The mobility of microclusters on ICAM-1-containing bilayers was 1.7-times greater than for the cells exposed to bilayers free of ICAM-1, namely, 17 and 10 min, respectively (Fig. 5D). ICAM-1 at a density significantly below 300 mol/m 2 did not enhanced microcluster displacement and mobility (Fig. 5, B and D) despite productive integrin engagement as was evident from ICAM-1 accumulation at the cell/bilayer interface. Importantly, inhibition of Pyk2 signaling kinase had more profound effect on both displacement and mobility of microclusters (Fig. 5, C and E) then on their size (Fig. 4F).
Recruitment of Activated Src Kinases to the Contact Surface-To provide additional evidence that ICAM-1 interactions with ␤ 2 -integrins influence the magnitude of proximal activating signaling in CD16.NK-92 cells, we visualized accumulation of activated (phosphorylated) Src kinases (pSrc) at the synaptic . The dependence of CD16 microcluster size upon the extent of integrin ligation and integrin-mediated signaling. CD16.NK-92 cells were exposed to the bilayers containing anti-CD16 antibody and ICAM-1 molecules at indicated concentrations. Individual CD16 microclusters were selected and analyzed as described under "Experimental Procedures." A, distribution of CD16 microcluster size at the contact surface of CD16.NK-92 cells exposed to bilayers containing anti-CD16 antibody (50 mol/m 2 ) in the presence (blue) or absence (red) of ICAM-1. B, average areas of CD16 microclusters observed at indicated time points after initial cell/bilayer contact. CD16.NK-92 cells exposed to the bilayers that do (filled bars) or do not (empty bars) display ICAM-1. C and D, average areas of CD16 microclusters that was determined 10 min after the initial cell/bilayer contact are shown. CD16.NK-92 cells were exposed to bilayers containing anti-CD16 antibody at indicated concentrations in the absence (C) or presence (D) of ICAM-1. E, average area of CD16 microclusters was determined 10 min after the initial contact of CD16.NK-92 with bilayers containing anti-CD16 antibody (50 mol/ m 2 ) and ICAM-1 at indicated densities. F, CD16.NK cells were treated with Pyk2 inhibitor at indicated concentrations and were exposed to bilayers containing anti-CD16 antibody (50 mol/m 2 ) and ICAM-1 (300 mol/m 2 ) and average areas of CD16 microclusters was determined 10 min after the initial cell/bilayer contact. Results are representative of 2 to 3 independent experiments. At least 100 microclusters (nՆ100) from at least 10 cells in each group were analyzed in every experiment. Results are shown as mean Ϯ S.D. p values were calculated by Student's t test: *, p Ͻ 0.05; ***, p Ͻ 0.001; ns, not significant.
interface in response to CD16 receptor ligation in the presence or absence of ICAM-1 on the bilayers. Regardless of the presence of ICAM-1, we observed the formation of pSrc clusters (Fig. 6A). Although the number of pSrc clusters per cell was similar on the bilayers with or without ICAM-1, the presence of ICAM-1 resulted in the formation of larger pSrc clusters having a higher density of pSrc (Fig. 6, B and C). Thus, ␤ 2 integrin signaling mediated more robust recruitment of pSrc to the contact surface. In the presence of ICAM-1, the observed pSrc microclusters were also more persistent because their size and the density did not change significantly from 5 to 10 min after initial contact of the cells with the bilayers (Fig. 6, B and C). In contrast, without ICAM-1 on the bilayers, the average size and density of pSrc microclusters decreased approximately by 20% from 5 to 10 min after initial cell exposure to the bilayers (Fig. 6,  B and C). Thus, the proximal activating signaling at the synaptic interface of CD16.NK92 cells exposed to the bilayers containing ICAM-1 appeared to be more robust and was maintained for a longer time as opposed to that in CD16.NK-92 cells on ICAM-1-free bilayers.
We then measured the kinetics of pSrc recruitment to the interface and the average density of the recruited pSrc (Fig. 6, D  and E). There was no difference in the total amount or the density of pSrc recruited to the contact area in the presence or absence of ICAM-1 within 5 min after the initial contact of FIGURE 5. Effect of integrin ligation and integrin-mediated signaling on parameters of CD16 microcluster movement. CD16.NK-92 cells were exposed to bilayers containing anti-CD16 antibody (50 mol/m 2 ) and ICAM-1 at indicated density, and CD16 microcluster mobility and displacement were determined 10 min after the initial cell/bilayer contact. A, representative image of CD16 microclusters and the microcluster tracks are shown. White lines indicate the microcluster displacement. See supplemental Fig. S10 and "Experimental Procedures" for details. B, average displacements of CD16 microclusters at the CD16.NK-92/bilayer interface in the presence of ICAM-1 at indicated density. C, average displacements of CD16 microclusters at the CD16.NK-92/bilayer interface in the presence of indicated concentrations of Pyk2 inhibitor. D, average mobility of CD16 microclusters at the contact area of CD16.NK-92 cells exposed to the bilayers containing anti-CD16 antibody and various densities of ICAM-1 molecules. E, average mobility of CD16 microclusters at the contact area of CD16.NK-92 cells exposed to the bilayers containing anti-CD16 antibody (50 mol/m 2 ) and ICAM-1 (300 mol/m 2 ). CD16.NK-92 cells were treated with Pyk2 inhibitor at indicated concentrations. At least 100 microclusters (nՆ100) from at least 20 cells were analyzed at each condition in three independent experiments. Results are shown as mean Ϯ S.D. p values were calculated by Student's t test: **, p Ͻ 0.01; ***, p Ͻ 0.001; ns, not significant. FIGURE 6. Influence of integrin engagement on recruitment of pSrc to the CD16.NK-92 cell/bilayer interface. CD16.NK-92 cells were exposed to bilayers with or without ICAM-1 for 5 or 10 min. The bilayers were then fixed with paraformaldehyde, permeabilized, stained with antibodies against pSrc followed by staining with fluorescent secondary antibody. A, representative TIRFM images of contact area of CD16.NK-92 cells exposed to bilayers in the presence (left panel) or absence (right panel) of ICAM-1. Average areas (B) and intensities (C) of pSrc-containing microclusters recruited to the contact surface of CD16.NK-92 in the presence (filled bars) or absence (empty bars) of ICAM-1, 5 or 10 min after the initial contact formation. At least 100 microclusters from 3 independent experiments were analyzed for each group, and the results are shown as mean Ϯ S.D., *, p Ͻ 0.05 by Student's t test. Total average intensity (D) and density (E) of pSrc recruited to the cell/bilayer interface in individual cells exposed to bilayers that do (filled bars) or do not (empty bars) display ICAM-1, 5, or 10 min after the initial contact formation. At least 20 cells were analyzed in each group. Results are representative of three independent experiments and are shown as mean Ϯ S.D., *, p Ͻ 0.05 by Student's t test.
CD16.NK92 cells with the bilayers. The difference became apparent later, i.e. 10 min after the contact was established, and the total amount and the density of the recruited pSrc were significantly higher in the presence of ICAM-1.

DISCUSSION
It is thought that an efficient response by cytolytic effectors requires the formation of a highly specialized interface between target and effector cell called the cytolytic synapse (4 -6, 10, 11, 13). The formation of this structure usually depends on the ligation of both activating and integrin receptors. Signaling initiated by each receptor is required for cytoskeleton remodeling, formation of activating microclusters and the adhesion ring junction, polarization of the effector cell, and cytolytic granules release. While major signaling and adaptor proteins that are involved in the development of full blown signaling in cytotoxic lymphocytes are known, how each signaling pathway contributes to the initiation of the cytolytic response is not well understood. The principal finding of this study is that engagement of ␤ 2 -integrins but not CD16-activating receptors influences the size and the dynamics of signaling microclusters in CD16.NK-92 cells. This result is consistent with previous findings showing that variation of the extent of TCR ligation in T cells exposed to ICAM-1-containing lipid bilayers presenting stimulatory peptide MHC ligands at various densities did not significantly affect the size of TCR signaling microclusters (23). In addition, it has also been shown that in the absence of the integrin ligation, the contact interface is profoundly diminished between CTL and a lipid bilayer that displays only cognate pMHC ligands (4).
To reveal the critical role of ␤ 2 -integrins in regulating the proximal signaling and the formation of contact interface, we utilized CD16.NK-92 cytolytic effectors that could still kill target cells and establish contacts with planar lipid bilayers in the absence of ICAM-1 ligand for ␤ 2 -integrins. This may be due to the presence of other ␤ 2 -integrins, such as ␣ M ␤ 2 and ␣ X ␤ 2 on the surface of NK-92 cells (1), which could recognize other protein and non-protein ligands (29). For instance, ␣ M ␤ 2 integrin has been shown to interact with casein and serum albumin (30,31) presumably via recognition of RGD sites on denatured proteins (32). These proteins were used to block lipid bilayers and could serve as altered ␤ 2 -integrin ligands. In fact, blocking ␤ 2 -integrins with TS1/18 antibodies precludes CD16-mediated cytolytic activity of CD16.NK-92 cells (7) and prevents these cells from establishing contact with lipid bilayers (data are not shown). Similarly, the ability to augment the naturally low level of ICAM-1 on SKBR3 target cells by IFN-␥ treatment, allowing us to study how differential expression of ICAM-1 by live target cells influences CD16.NK-92-mediated ADCC.
Thus, variations of ICAM-1 ligand density on planar lipid bilayers and the surface of live target cells allowed us to vary the extent of ␤ 2 -integrin ligation on CD16.NK-92 effectors and to analyze systematically how variations in ␤ 2 -integrin ligation affect the killing kinetics of the target cells, the area of the contact surface, the size and dynamics of activating CD16 signaling microclusters, and the kinetics and topography of cytolytic granule release.
We have found that in the presence of ICAM-1, the average size of moving microclusters was uniform and was characterized by a tight unimodal distribution (Fig. 4A). The average microcluster size remained steady for over 30 min (Fig. 4B). When CD16.NK-92 cells were stimulated via CD16 receptor in the absence of ICAM-1 on bilayers, the smaller moving microclusters still maintained their size and uniformity (Fig. 4, A and  B). We postulate that RGD-containing denatured proteins present on bilayers engaged the NK cell integrins, albeit to a lower extent, resulting in a smaller average size of the microclusters. Importantly, variations in the extent of CD16 ligation within a wide range did not affect either the size or temporal uniformity of the signaling microclusters regardless of the presence of ICAM-1 (Fig. 4, C and D). This suggested that the size of the microclusters in CD16.NK-92 cells is regulated by the integrin signaling, but not by the CD16-mediated signaling. Indeed, inhibiting Pyk2 kinase, a key signaling protein in the ␤ 2 -mediated signaling pathway (Fig. 7), led to a smaller size of signaling microclusters (Fig. 4F) and a significant decrease of microcluster mobility and displacement (Fig. 5, C and E). Importantly, the inhibition of microcluster dynamics (Fig. 5, C and E) was more profound than the decrease of microcluster size (Fig. 4F). This suggests that integrin-mediated signaling has a greater impact on microcluster dynamics than on their size. FIGURE 7. Schematic illustration of integrin-mediated signaling pathway that regulates actin polymerization, actin cytoskeleton remodeling, and microcluster formation. Binding of ICAM-1 to activated integrin receptors initiates outside-in signaling. The ligand binding leads to integrin aggregation and recruitment of scaffolding proteins such as integrin-linked kinase (ILK) to cytoplasmic tail of integrins (43,44). ILK interacts with Particularly interesting Cys-His-rich protein (PINCH) and Parvin to form an ILK-PINCH-Parvin (IPP) complex that functions as a signaling platform for integrins (43,45). This leads to activation of paxillin and proline-rich tyrosine kinase 2 (Pyk2) and the formation of Pyk2/Paxillin complexes (20). The formation of these complexes is facilitated by the interaction with vinculin, talin, and multiple F-actin binding and regulating proteins. The latter regulate the Cdc42 pathway, which directly controls actin polymerization, cytoskeleton remodeling (46 -50), and formation and growth of micrtoclusters in distinct, multiple loci. It is also thought that vinculin plays an important role in linking de novo actin polymerization to integrin activation (51).
Analysis of the recruitment of pSrc to the CD16.NK-92/bilayer interface showed that the development of proximal activation signaling in the NK cell cytolytic effectors takes a relatively long period of time, i.e. about 10 min (Fig. 6, D and E). This is significantly longer as compared with cytolytic T cells, which could reach maximal signaling within 2 min (11,33). This difference is in accord with the delay in cytolytic granule release by the NK cells ( Fig. 3A and supplemental Fig. S6) as opposed to that by cytotoxic T cells, which release granules within 1-2 min after the contact with the bilayer (11).
Our data are consistent with previous findings that "outsidein" integrin-mediated signaling contributes to activating signal longevity through the regulation of persistence and mobility of SLP-76 microclusters and their interaction with ZAP-70 (34,35). In accord with this, recent findings have shown that effective integrin signaling in T cells is linked to ZAP-70 phosphorylation, while the catalytic activity of ZAP-70 contributes to TCR signal amplification (36).
Our previous findings showing that the kinetics of proximal signaling regulates the kinetics of granule release and target cell lysis by CTL (6,11,28). Consistent with this, the size and dynamics of signaling microclusters appears to regulate magnitude and kinetics of proximal signaling, which influence the kinetics of granule release (Figs. 1C and 3B). The latter control the kinetics of target cell lysis (Fig. 1, A and B).
Cytolytic granules travel along the microtubules from the periphery toward the synaptic interface (37,38) and are released in multiple locations (Fig. 3A, supplemental Movies S4 and S5). These locations are likely determined by the points of microtubule attachment to the actin cytoskeleton mesh at the interface through Adhesion and Degranulation-Promoting Adapter Protein (ADAP), as has been previously described for T cells (37). The granule release sites had characteristically lower density of ICAM-1 molecules (supplemental Fig. S5), which is consistent with previous observations showing that cytolytic granules are released in the locations devoid of actin cytoskeleton (39). Here we show that the sites of granule release are surrounded by clusters of activating CD16 receptors (Fig.  3A, supplemental Movies S4 and S5). Such structures resemble multifocal synapses in T cells (6,27). Consistent with previous observations (40), we have shown here that the locations of granule release at the NK cell interface remain unchanged over a long period of time (supplemental Fig. S5), ensuring uninterrupted delivery of cytolytic granules to the same locations on the target cell membrane. This increases the probability of successful granule penetration into a target cell and the induction of target cell death. Thus, even though a mature cytolytic synapse is not required for cytolytic activity (41,42), the formation of a very well structured synaptic interface is essential for efficient cytolytic granule delivery and target cell destruction (6).
In conclusion, integrin signaling mediates the formation of a large synaptic interface and controls the size, dynamics, and stability of signaling microclusters, thereby ensuring the triggering of robust, rapid, and sustained signaling that is necessary to exercise effective cytolytic activity by cytolytic lymphocytes against target cells. CTL mostly rely on ␣ L ␤ 2 (LFA-1) integrin and, therefore, are more effective during inflammation, a condition resulting in up-regulation of the LFA-1 ligand, ICAM-1, on the target cell surface. Because NK cells express and could effectively utilize other integrin receptors such ␣ M ␤ 2 , ␣ X ␤ 2 , and ␣ D ␤ 2 , their functioning appeared to be less dependent on inflammation and they play an important role in the earlier detection and destruction of tumor cells, virus-infected cells, or other aberrant cells prior to the development of inflammation and full blown acquired immune response.