Termination of the Activating NK Cell Immunological Synapse Is an Active and Regulated Process

Natural killer cells are cytotoxic innate lymphocytes that are able to detect, bind, and kill malignant and infected cells (1, 2). In addition, they influence the immune response through the release of cytokines and chemokines. During the initial contact with a target cell, NK cells are activated by binding specific ligands on the target cell membrane, which leads to the formation of the immunological synapse (IS) and a tight cellular contact. The target cell death is then induced by the release of cytotoxic granules, filled with pore-forming perforin and apoptosis-inducing granzymes, or by stimulation of death receptors via TRAIL or FasL (reviewed in Ref. 3).

The IS is initiated by binding of integrins and selectins on the NK cell to their respective ligands on the target cell. The binding of LFA-1 to its ligands ICAMs 1, 2, or 3 is especially important for IS maturation (4–6). First, the actin cytoskeleton is reorganized and the NK cell receptors are recruited to the IS. Through inside-out signaling the affinity and avidity of LFA-1 is increased, which increases the binding to ICAMs, and thereby strengthens the IS (7, 8). The tighter binding allows the clustered activating and inhibitory receptors to bind their respective ligands and signal via their intracellular motifs. Activating NK cell receptors include NKG2D; the NCRs NKp30, NKp44, and NKp46; DNAM-1; and members of the SLAM family of receptors, such as 2B4 and NTB-A (9). These receptors transmit their activating signal via phosphorylation of Src and Syk family kinases, phospholipase C γ (PLCγ), as well as PI3K (10–12). The cytotoxic granules move along microtubules via dynein-dynactin motor proteins and through the actin mesh via myosin IIA in an ATP-dependent manner, and they are released into the IS in a process called degranulation (13, 14). This ultimately results in the death of the locally attached target cell.

Although the initiation and formation of the IS are well characterized, much less is known about signaling processes that are important for the maintenance of the IS and the detachment after the death of the target cell. However, these processes may be important for cytokine secretion (15) and serial killing activity of NK cells. Already in 1975 it was shown that cytotoxic T cells and NK cells can kill multiple cells in a process named serial killing (16, 17). The number of targets cells killed by an individual NK cell differs from study to study and seems to be influenced by several factors, such as the activation status of the NK cell and the choice of target cell (18–20). In addition, only a fraction of NK cells seems to be able to kill more than one cell, and those cells can quickly form new lytic granules or recycle existing ones to prevent exhaustion (20–22). Interestingly, the first kill of a serial-killing NK cell seems to be slower than the following kills, which could be explained by an addition and integration of activating signals from new and old target cells (19). Also, the use of therapeutic Abs or CAR-mediated retargeting of NK cells increases the number of serial-killing NK cells, which makes the processes leading to serial killing interesting for therapeutic applications involving NK cell cytotoxicity (18, 23).

In this study, we investigated the process of NK cell detachment from a target cell. Most studies that analyze NK cell serial killing use microscopic methods to determine the dissociation of an NK cell from its target. In this study, we use a modified FACS-based conjugate assay to calculate the half-life of the contact between human NK cells and K652 target cells. Our data show that ongoing activating signaling is required to maintain the contact in an energy-dependent fashion, even after the IS was formed successfully. NK cells that detached from target cells show clear signs of activation (CD69 upregulation, downregulation of activating receptors) and degranulation. More importantly, the initiation of a contact to new target cell results in accelerated detachment from an initially bound target cell. This demonstrates that not only the attachment, but also the detachment of NK cells from target cells is a highly regulated process.

The human cell line K562 was maintained in IMDM (Thermo Fisher) and the human HeLa cell line in DMEM (Thermo Fisher). All media were supplemented with 10% heat-inactivated FCS (Thermo Fisher) + 100 U/ml penicillin and 100 μg/ml streptomycin (P/S; Thermo Fisher). HeLa cells were stably transfected with CD48 and selected with 1 μg/ml puromycin (Merck Millipore). Primary NK cells were isolated from blood of healthy donors as described earlier (24). In brief, PBMCs were obtained via density centrifugation, and NK cells were isolated with Dynabeads Untouched Human NK Cells Kit (Thermo Fisher). Primary NK cells were cultivated in IMDM + 10% FCS + 1% P/S and supplemented with rIL-15 (5 ng/ml) and rIL-2 (200 IU/ml). At days 0 and 7 after isolation, irradiated JY feeder cells were added. All cells were incubated in a humidified incubator with 37°C and 5% CO₂.

The following Abs were used in this study: anti-2B4 (C1.7; BioLegend), anti-CD56 (MEM-188; BioLegend), anti-CD69 (FN50; BioLegend), anti–DNAM-1 (DX11; BD Biosciences), anti-CD107a (H4A3; BioLegend), anti-NKG2D (149810; R&D), anti-NKG2A (Z199; Beckmann Coulter), and IgG-control (MOPC-21; BioLegend).

The following inhibitors were used in this study: PD98059, piceatannol, okadaic acid, and U0126 obtained from Cayman Chemical; PP1, wortmannin, U73122, and SB202190 obtained from Biomol; Syk Inhibitor IV, calyculin A, and nocodazole obtained from Merck Millipore; cytochalasin D obtained from US Biological; sodium azide obtained from Carl Roth, and paclitaxel obtained from Cell Signaling. The inhibitors did not influence the viability of NK cells at the concentrations used in this study.

Other reagents used in this study included PKH26 Red and PKH67 Green Fluorescent Cell Linker Kit for general cell membrane labeling (Sigma-Aldrich), CellTracker Violet BMQC Dye (Thermo Fisher), and ATPlite–Luminescence ATP Detection assay system (PerkinElmer).

A total of 5 × 10⁶ NK cells and target cells was stained with 1 μM PKH26 or PKH67 membrane dyes for 5 min. Cells were washed thoroughly and resuspended in NK cell medium (IMDM, 10% FCS, 1% P/S). A total of 5 × 10⁴ NK cells was added to 1 × 10⁵ target cells in 100 μl volume, centrifuged (20 × g, 1 min), and coincubated for 0–90 min. The reaction was stopped by brief vortexing and addition of 100 μl 4% paraformaldehyde (PFA). Cells were analyzed immediately using a BD FACSCalibur, and conjugates were determined as double-positive events. Data were analyzed using FlowJo (Tree Star) software, and statistics were calculated using GraphPad PRISM software.

Cells were stained using PKH dyes as described earlier. A total of 5 × 10⁵ NK cells was mixed with 1 × 10⁶ target cells in a small volume (650 μl) and centrifuged (20 × g, 1 min). If not indicated otherwise, cells were pre-coincubated for 30 min at 37°C to form initial conjugates. Subsequently, cells were diluted 1:25 in NK cell medium, divided into six equal samples, and incubated while rotating at 37°C, allowing the NK cells to detach but preventing the formation of new conjugates. The reaction was stopped at different time points by brief vortexing and addition of an equal volume 4% PFA. Cells were analyzed immediately using a BD FACSCalibur. Conjugates were determined as double-positive events within all NK cells using FlowJo (Tree Star) software. The half-life of conjugates was calculated using the formula for one-phase exponential decay.

The plate-bound detachment assay was performed using adherent HeLa cells, stably transfected with CD48, as target cells. This assay required the use of adherent target cells to separate unbound NK cells from NK cells in conjugates. HeLa cells were grown in a six-well plate overnight to obtain a confluent layer. A total of 3–5 × 10⁶ NK cells was added, and the plate was centrifuged (20 × g, 1 min) and incubated for 30 min (37°C, 5% CO₂) to form initial conjugates. Subsequently, the culture medium was removed and the plate was washed thoroughly to remove unbound NK cells, by centrifugation upside down, covered with another six-well plate. Culture and washing medium was collected, stored on ice, and fresh medium was added to the plate. After 30-, 60-, and 90-min incubation on a shaker (37°C, 5% CO₂), the newly detached NK cells were collected as described earlier. The collected NK cells were stained with the indicated Abs and analyzed using a BD FACSCalibur.

Cytotoxic capacity was determined by a standard 4-h chromium release assay as described previously (25). In brief, 5 × 10⁵ target cells were labeled in 100 μl medium with 100 μCi [⁵¹Cr] (Hartmann Analytic) for 1 h at 37°C, 5% CO₂. A total of 5 × 10⁴ target cells was added in a 96–u-bottom–well plate and mixed with NK cells at ratios from 20:1 to 1.25:1 (depending on the experiment). NK cells were either pretreated with inhibitors or inhibitors were added at the point of coincubation. After 4 h, supernatant was harvested and analyzed in a WIZARD² (PerkinElmer) gamma counter. Specific lysis was calculated as (experimental release − spontaneous release)/(maximum release − spontaneous release) × 100.

To analyze the detachment of NK cells, we used a modified flow cytometry–based conjugate assay (Fig. 1A) (26). Primary human IL-2–activated NK cells and K562 target cells were stained in different colors using membrane dyes and coincubated in a small volume to allow conjugate formation. After 30 min, before conjugate formation reached a plateau phase (Supplemental Fig. 1A), the preformed conjugates were diluted with medium and samples were agitated by rotation to allow the NK cells to detach while preventing the formation of new conjugates (Supplemental Fig. 1B). Samples were collected at different time points after dilution, and the number of conjugates was determined by flow cytometry (Fig. 1B). By plotting the amount of conjugates over time, we determined the conjugate half-life. For IL-2–activated human NK cells and K562 target cells, the conjugate half-life was 47.0 ± 11.8 min (Fig. 1C). Changing the ratio of NK to target cells or shortening the pre-coincubation time did not significantly change this value (Supplemental Fig. 1C, 1D). We also did not find any correlation between the amounts of conjugates that formed within the 30 min of pre-coincubation and the conjugate half-life (Supplemental Fig. 1E). As we used membrane dyes to label the NK and target cells, we observed a dye transfer between the cells, resulting in a population of NK cells with an intermediate level of target cell dye (Fig. 1B). This population was excluded from our analysis. When using intracellular dyes (cell tracker) to stain NK and K562 cells, we did not observe any dye transfer and we obtained essentially the same half-life of conjugates, demonstrating that the membrane dyes do not influence the detachment assay.

To analyze the expression of surface proteins after detachment, we established a plate-bound detachment assay using adherent target cells (see 2Materials and Methods). Using this method, we investigated the surface expression of several proteins from early, intermediate, and late detached NK cells and compared them with NK cells, which were either incubated without target cells or which did not bind to a target cell during the pre-coincubation phase. Compared with these controls, NK cells showed increased expression of the activation marker CD69 (Fig. 2A) and the degranulation marker CD107a (Fig. 2B) after detachment. This demonstrates that the target cell contacts were productive, leading to the activation and degranulation of the NK cells. The engagement of activating NK cell receptors by their ligands on target cells results in receptor internalization (27–29). Supporting these findings, we observed a reduction in the surface expression of the activating receptors 2B4, DNAM-1, and NKG2D (Fig. 2C–E), because the ligands for these receptors were present on the HeLa-CD48 target cells used in the assay. This downregulation was specific because we did not see significant changes in the expression of NKp30, NKp44, or NKp46, the NK cell marker CD56, or the α-integrin CD11a after NK cell detachment (data not shown). For the inhibitory receptor NKG2A, we observed a slight upregulation in late-detaching NK cells, although these changes were not significant (Fig. 2F). This demonstrates that freshly detached NK cells internalize activating receptors and upregulate proteins that are associated with NK activation and degranulation.

The formation of the NK cell IS with a target cell depends on signals via activating NK cell receptors (8) and the integrin LFA-1 (30, 31), and involves distinct and regulated steps (3). We were interested to investigate whether any of these signals are also essential for the maintenance of the IS after it is established. To not interfere with the formation of the IS, we added chemical inhibitors to our detachment assay after the conjugates were formed and studied their effect on the decay of preformed conjugates (Fig. 1A). Artificially increasing the affinity of LFA-1 by treating the conjugates with Mg/EGTA prevented the detachment of NK cells, demonstrating the involvement of LFA-1 in conjugate stability (Supplemental Fig. 2A, 2B). The inhibition of early signaling events via the inhibition of Syk and Src family kinases by the inhibitors Syk IV, PP1, or piceatannol significantly accelerated the detachment of NK cells (Fig. 3A–C). PLCγ is recruited to the IS after activation by Syk and is important for NK cell effector functions. Inhibition of PLCγ with U73122 resulted in a strongly reduced half-life of conjugates (Fig. 3D). Syk also activates PI3K, which regulates several intracellular signaling pathways by the generation of secondary lipids (32, 33). Inhibition of PI3K by wortmannin accelerated the NK cell detachment (Fig. 3E). These data demonstrate that the activity of Src and Syk family kinases, PI3K and PLCγ, is not only important for the establishment, but also for the maintenance of the IS. Another key factor for the functional generation of the IS is the rearrangement of the actin cytoskeleton (3). We targeted actin with the depolymerizing reagent cytochalasin D or the stabilizing reagent jasplakinolide to interfere with normal F-actin dynamics. Treatment with either of the two reagents led to an instant decay of preformed conjugates, showing the importance of dynamic F-actin for the stability of the IS (Fig. 3F, 3G). The microtubule cytoskeleton facilitates the transport of cytotoxic granules toward the IS (34, 35). To investigate the influence of microtubules on the stability of the IS, we used the reagents nocodazole to depolymerize microtubules or paclitaxel to stabilize them. Nocodazole treatment accelerated the NK cell detachment slightly but significantly, whereas paclitaxel had no effect on the conjugate stability (Fig. 3H, 3I), suggesting that polymerized tubulin is necessary to prevent conjugate decay. These results show that not only the establishment, but also the maintenance of the IS depends on early activating signals and the integrity of the cytoskeleton.

MAPK acts further downstream of PI3K and is important for polarization and degranulation of cytotoxic granules (36). The inhibition of the MAPKs MEK1/2 with PD98059 or U0126 did not affect the half-life of NK conjugates (Fig. 4A, 4B), although the treatment reduced the cytotoxicity against K562 (data not shown). The serine-threonine kinase p38 is involved in a broad spectrum of NK cell effector functions and is also important for NK cell cytotoxicity (37). However, p38 inhibition with SB202190 did not influence NK cell detachment (Fig. 3C), although it did inhibit the killing of K562 (data not shown). Those results demonstrate that in contrast with the early signaling events, downstream signaling via MEK1/2 or p38 does not influence the stability of NK cell conjugates.

Signaling events and other dynamic processes such as the rearrangement of the cytoskeleton are dependent on energy in the form of ATP. However, the importance of ATP for stability of the IS and its influence on NK cell detachment has not been studied so far. Therefore, we investigated the role of energy in these processes by depleting ATP using azide. The pretreatment of NK cells with 50 mM sodium azide for 2 h resulted in a reduction of intracellular ATP levels by >60% compared with control-treated cells (Fig. 5A) and decreased NK cell–mediated killing of K562 (Fig. 5B). Next, we wanted to test the importance of ATP for NK cell conjugate stability. Although the pretreatment with azide reduced the half-life of conjugates compared with control cells, it also resulted in the formation of fewer conjugates (Supplemental Fig. 3), making the interpretation of the results difficult (Fig. 5C, 5D). In addition, we observed inhibition of target cell lysis when azide was added directly to a cytotoxicity assay without preincubation of the NK cells (Fig. 5B). To not interfere with conjugate formation, we therefore added azide only at the time point of dilution. This also resulted in a faster detachment of NK cells and a significant reduction of the conjugate half-life (Fig. 5C, 5D). These results show that the maintenance of NK cell IS is highly dependent on the availability of intracellular energy in the form of ATP.

Several signaling molecules are regulated by serine/threonine phosphorylation. Treatment of NK cells with the serine/threonine phosphatase inhibitor calyculin A was shown to reduce NK cell–mediated cytotoxicity (38), and the serine/threonine phosphatase PP2A was found to negatively regulate IL-2 signaling in NK cells (39). We were therefore interested to investigate whether serine/threonine phosphatases are also involved in the process of IS maintenance. Inhibition of the serine/threonine phosphatases PP1, PP2A, and PP3 by the broadband inhibitor okadaic acid had no effect on NK cell cytotoxicity (Fig. 6A) and only very slightly increased the half-life of conjugates in the detachment assay (Fig. 6B, 6C). In contrast, the treatment of NK cells with calyculin A resulted in a complete inhibition of NK cell cytotoxicity against K562 (Fig. 6D). The detachment assay showed a decreased conjugate half-life after treatment of preformed conjugates with calyculin A compared with control cells (Fig 6E, 6F). Therefore, the reduced cytotoxicity caused by calyculin A may be a result of early NK cell detachment caused by an unstable IS.

Choi and Mitchison (19) suggested the model of kinetic priming. The model states that NK cells use signaling events from former contacts to lower the activation threshold for subsequent kills, resulting in a more effective serial killing. They suggested that an NK cell might stay in contact with a target cell until a new target induces the detachment. To test the model of kinetic priming in our assay system, we added new K562 target cells, which were labeled in a different color, to the preformed conjugates. To allow the formation of new contacts, we did not rotate the cells during the experiment. The presence of new target cells accelerated the detachment of NK cells from the initially bound target cells (Fig. 7A). At the same time, we observed that the NK cells formed conjugates with the new target cells (Fig. 7B). As a control, we used formaldehyde-fixed K562 cells. NK cells did not form conjugates with fixed target cells (Fig. 7B), and the presence of fixed target cells also did not change the stability of existing conjugates (Fig. 7A). We saw essentially the same result when using methanol-fixed targets or Daudi cells as nonsusceptible targets for resting NK cells (Supplemental Fig. 2C, 2D). This demonstrates that the possibility to engage into a contact with a new, susceptible target cell is another factor that determines the stability of existing NK cell conjugates.

The steps of NK cell synapse formation up to the release of lytic granules and the induction of apoptosis in target cells are well understood (3). Because it was described that NK cells can kill not only one target, but in some cases consecutively up to 16 different cells (18–20), the question of what leads to the termination of the synapse and the detachment from the target cell became more intriguing. In this study, we used a flow cytometry–based approach to investigate the factors influencing NK cell detachment. The determined half-life for conjugates between human IL-2–activated NK cells and the target cell line K562 was ∼47 min. It is important to acknowledge that this is an average value based on a large and diverse pool of NK cells and that it was derived from an in vitro system. It is known that multiple factors such as the preactivation state of NK cells, the type of effector or target cells, or the assay system used can influence on NK:target cell contacts (19, 20, 40).

The analysis of detached NK cells clearly demonstrated that the conjugates in our assays were functional and resulted in activation of NK cells, as evident by the upregulation of the activation marker CD69 and the surface expression of CD107a as a marker for degranulation. Consistent with the fact that the HeLa-CD48 target cells used in the assay express ligands for 2B4, NKG2D, and DNAM-1 (41, 42), and in line with previous observations (27–29), we observed a downregulation of these receptors in detached NK cells as a sign of receptor engagement. However, HeLa cells also express B7-H6, the ligand for NKp30 (43, 44). The fact that we did not see a significant downregulation of NKp30 could be explained by insufficient expression of its ligand or indicate that NKp30 is not internalized after ligand engagement.

NK cell activity is strictly regulated by integration of activating and inhibitory signaling (9). Conjugate formation and induction of effector functions depend on certain key factors, for example, Src and Syk family kinases, PI3K or PLCγ (2, 11). Our data show that these signaling events are not only necessary for the establishment of the IS, but also for the maintenance of the contact. However, because we treated existing NK:K562 conjugates with inhibitors, we cannot exclude that this may also interfere with signaling events in the K562 cells. Src kinases or PLCγ inhibition had a very strong effect, leading to almost immediate collapse of most of the conjugates. Considering the importance for Src family kinase-mediated phosphorylation in the signaling pathways of all activating receptors, this was not surprising (45). Similarly, PLCγ activity was described to be crucial for NK cell function (46, 47). Interfering with PI3K or Syk activity also affected the stability of conjugates, although less effectively than PLCγ inhibition (Supplemental Fig. 4). This suggests that these signaling pathways may be less important to maintain the stability of the IS. The IS is a dynamic structure. Receptor–ligand interactions are constantly broken and re-established. Signals from activating NK cell receptors are necessary to induce a high ligand binding activity of the integrin LFA-1 by affecting its affinity and avidity through what is known as inside-out signaling (7, 8). Our data suggest that inside-out signaling is not only required to induce LFA-1 activity to establish a firm contact to a target cell, but that it is also necessary to maintain LFA-1 binding activity. Another process induced by early signaling events is the formation of F-actin, which facilitates actin cytoskeleton rearrangements (48). Our data show that actin dynamics are essential for the maintenance of the IS, because we observed rapid detachment upon interference with the actin cytoskeleton. Recruitment and clustering of receptors are dependent on F-actin. Hence, loss of the anchor provided by the actin cytoskeleton would enable the redistribution of receptors and integrins, and thereby weaken the contact between NK and target cells.

Although our data show that later signaling events such as MAPK signaling were required for NK cell cytotoxicity, but not for the stability of conjugates, inhibition of the serine/threonine phosphatases PP1, PP2A, and PP3A by okadaic acid or calyculin A had differential effects on conjugate stability. Calyculin A clearly inhibited NK cell cytotoxicity, confirming a previous result (38). Okadaic acid inhibits PP1, PP2A, and PP3A, whereas calyculin A inhibits PP1 and PP2A. Although the IC₅₀ values for PP2A (0.5–1 nM) are comparable for both inhibitors, calyculin A has a higher inhibitory effect on PP1 (IC₅₀ value 2 nM for calyculin A versus 60–500 nM for okadaic acid) (49). Thus, because only calyculin A treatment affected conjugate stability, this effect may be caused by more effective inhibition of PP1. However, we cannot exclude the possibility that an off-target effect of calyculin A is responsible for our result. In T cells PP1 was described to dephosphorylate and thereby activate cofilin (50). Cofilin is required in T cell migration and activation by depolymerizing and/or severing actin filaments (51). Therefore, the observed effect in this study may be because of a yet unknown role of cofilin in regulating the actin dynamics in NK cells.

Already in 1975, Berke and Gabison (52) reported that the formation, but not the maintenance, of T:target cell conjugates is energy dependent. Our data demonstrate that in NK cells also the maintenance of NK:target cell conjugates is dependent on energy in the form of ATP, because its depletion upon treatment with azide resulted in a rapid loss of conjugate stability. Upon target cell contact, NK cells rapidly consume their metabolic energy and mitochondria reorganize toward the synapse, possibly to compensate for this local loss of energy (53). This may explain why inhibiting the function of mitochondria had such a drastic effect on conjugate stability. In addition, the kinases recruited by activating receptors require ATP as substrate. Therefore, reduction of the substrate would decrease the activating signaling and thereby destabilize the NK:target cell conjugates.

Serial killing is important for the efficiency of NK cell cytotoxicity. Some NK cells can consecutively kill multiple times (16, 18–20, 54). It was suggested that the availability of a new target in close proximity induces detachment from a previous target and that NK cells integrate signals from the previous and the current target to increase its killing speed. This process was called kinetic priming (19). In line with these observations, we found that addition of an exceeding amount of targets to preformed conjugates accelerated the NK cell detachment from the initial targets. This indicates that engagement of a new susceptible target induces processes leading to detachment. Formation of a new target cell contact requires the redistribution of adhesion molecules and activating receptors, which would reduce ongoing signaling at the previous IS. Our data show that ongoing signaling is essential for the maintenance of a conjugate, providing a functional explanation why a new contact can weaken an existing one. Interestingly, only susceptible targets affect NK cell detachment, because fixed target cells or nonsusceptible Daudi cells did not influence conjugate stability. Similarly, it was shown that an inhibitory synapse in the form of a contact with an MHC class I⁺ target cell does not interfere with an activating synapse with an MHC class I⁻ target cell (55). Interestingly, the formation of an inhibitory synapse is energy independent and is not affected by depletion of ATP (56). Therefore, an inhibitory synapse clearly differs from an activating one. Because it does not reduce the ongoing signaling at an existing contact, it does not interfere with it. This would suggest a model by which an NK cell can maintain an activating IS with a susceptible target cell even while surrounded by MHC class I⁺, nonsusceptible targets. However, in the presence of additional susceptible targets, the dynamics of the existing IS may change, enabling effective serial killing. This demonstrates that the regulation of NK cell detachment is an important factor for NK cell cytotoxic responses.

The authors have no financial conflicts of interest.