IL-2–dependent tuning of NK cell sensitivity for target cells is controlled by regulatory T cells

A previous study revealed an increase in numbers of NK cells upon diphtheria toxin (DT)–mediated ablation of T reg cells in Foxp3^DTR mice, in which T reg cells express the human DT receptor (DTR) under the control of the Foxp3 locus (Kim et al., 2007). To test whether NK cell reactivity was restrained by T reg cells, we isolated NK cells at different times after T reg cell depletion and stimulated with IL-12 and IL-18, PMA and ionomycin, or plate-bound agonistic antibody specific for the activating NK cell receptor Ly49D. We were unable to detect significant differences in the reactivity of NK cells to stimulation via Ly49D. In four out of six experiments, IFN-γ production and degranulation were comparable in DT-treated versus control mice at days 2 and 6 after DT administration (Fig. 1 A). In two experiments, we observed a moderately increased NK cell response on day 2 of DT treatment (DTX). These occasionally observed increased responses disappeared when CD4⁺ T cells were depleted in addition to T reg cells; similar results were obtained when NK cells were stimulated with NKp46 antibody (not depicted). Responsiveness of NK cells to IL-12 and IL-18 or PMA and ionomycin measured by IFN-γ production was only slightly increased on day 2 of DTX, whereas by day 6, NK cell reactivity was comparable with controls and even reduced at later time points (day 10; Fig. 1 A). This was in contrast to the progressively increasing activation of CD4⁺ and CD8⁺ T cells upon T reg cell depletion (not depicted).

Next, we examined the cytotoxic activity of NK cells in vitro in a ⁵¹Cr-release assay using YAC-1 target cells, which are missing-self targets whose killing is dependent on NKG2D signaling (Jamieson et al., 2002). We found that the cytotoxicity of NK cells isolated on days 2 and 6 after initial DT administration was comparable with that of NK cells from mock-treated mice (Fig. 1 B). To test whether the ability of NK cells to kill target cells in vivo was limited by T reg cells, we transferred splenocytes from mice expressing the transgene-encoded MCMV glycoprotein m157 (m157tg), an activating ligand for Ly49H, and found that NK cell–mediated target cell rejection was comparable in T reg cell–depleted and control mice (Fig. 1 C). We then analyzed the responsiveness of inhibitory receptor-negative NK cells, which have been proposed to achieve tolerance by dampening stimulatory signaling (Fernandez et al., 2005). These Ly49C/I⁻ NKG2A⁻ cells exhibited comparable IFN-γ production and even slightly reduced degranulation when stimulated with NKp46 and Ly49D antibodies on day 6 of T reg cell depletion (not depicted).

Together, these results suggest that under physiological conditions, the NK cell reactivity to strong activating ligands was independent of T reg cells. NK cells can have both proinflammatory and immunoregulatory functions, but the ablation of NK cells did neither alleviate nor facilitate the systemic myelo- and lymphoproliferative syndrome resulting from T reg cell elimination in Foxp3^DTR mice (Fig. 1, D–F). Although these experiments do not formally exclude potential T reg cell–dependent modulation of NK cell self-reactivity under physiological conditions, they are consistent with the observation that acute T reg cell ablation did not lead to the excessive generalized activation of NK cells (Fig. 1, A–C).

Although we did not observe a marked increase in the ability of NK cells to produce IFN-γ or to kill target cells upon T reg cell ablation, the assessment of in vivo NK cell reactivity at a population level could conceal T reg cell–mediated control of a small number of self-reactive NK cells. Therefore, we tested a role of T reg cells in the maintenance of tolerance of NK cells with a known specificity for self-ligands. Analogous to the T cell anergy that results from continuous exposure to a cognate self-antigen, NK cells that encounter activating ligands during differentiation similarly exhibit peripheral tolerance (Oppenheim et al., 2005; Sun and Lanier, 2008; Tripathy et al., 2008). In the case of T cells, T reg cells are thought to be essential to maintain this state of tolerance keeping peripheral autoreactive T cells at bay (Kim et al., 2007). To assess a role for T reg cells in the restraint of NK cells that exhibit self-reactive activating receptors, we generated Foxp3^DTR mice that transgenically express m157 (m157tg; Tripathy et al., 2008). NK cells continuously exposed to m157 as self-ligand in m157tg mice exhibit decreased amounts of Ly49H and a Ly49H receptor–specific unresponsiveness but can be activated through other activating receptors (Tripathy et al., 2008). As expected, NK cells in m157tg Foxp3^DTR mice similarly down-regulated Ly49H and were hyporesponsive to its stimulation. Depletion of T reg cells did not trigger expansion of Ly49H⁺ NK cells nor did it rescue Ly49H expression or IFN-γ production in response to Ly49H engagement (Fig. 2, A and B). NK cell responses to Ly49D stimulation were comparable in Foxp3^DTR mice expressing or lacking m157tg in the presence of T reg cells and upon their depletion. NK cells isolated from T reg cell–depleted mice expressing or lacking m157tg killed YAC-1 target cells as efficiently as NK cells from mock-treated animals. However, NK cells from m157tg mice were unable to lyse m157-expressing Ba/F3 (Ba/F3-m157) cells, and this state of unresponsiveness was not reversed upon ablation of T reg cells (Fig. 2 C). NK cells from m157tg mice maintained tolerance in the absence of T reg cells not only upon short-term (2 d) but also upon long-term (6–10 d) ablation of T reg cells when systemic T cell autoimmunity developed (Fig. 2 A and not depicted). Thus, T reg cells are dispensable for the maintenance of NK cell tolerance to activating self.

On the basis of the observation that NK cell reactivity and tolerance to activating ligands were regulated independently of T reg cells, we speculated that NK cell–intrinsic inhibitory mechanisms obviated the requirement for T reg cell–mediated suppression. Therefore, we sought to explore whether T reg cells influenced NK cell function in the absence of cell-intrinsic inhibitory signaling. Interactions of NK cell inhibitory receptors with self–MHC class I molecules prevent licensed functionally mature NK cells from attacking cells that provide suboptimal stimulation too weak to overcome the activation thresholds imposed by inhibitory receptors (Elliott and Yokoyama, 2011). In the absence of this continuous inhibition, lack of self-recognition enables the killing of cells that down-regulate MHC class I molecules as the result of oncogenic transformation or viral infection. To determine whether T reg cells safeguard NK cell cytotoxicity against missing-self in vivo, we cotransferred splenocytes from β2-microglobulin (β2M)–deficient and m157tg mice as targets for missing-self (loss of inhibitory signals because of MHC class I deficiency)– and non-self–recognition (engagement of an activating viral ligand in the presence of MHC class I), respectively (Fig. 3 A). To exclude a potential effect of different composition of experimental and control target cell populations (e.g., lack of CD8⁺ T cells in β2M-deficient mice), the specific rejection was assessed for CD19⁺ B cells. We observed a modest increase in β2M-deficient target cell rejection on day 2 after T reg cell depletion followed by a significant increase on day 6, whereas the rejection of m157tg targets was comparable in mock-treated or T reg cell–depleted mice (Fig. 3, B and C). The target cell rejection in T reg cell–depleted mice was largely mediated by NK cells because NK1.1 antibody–mediated depletion of NK cells in DT-treated control mice reduced the observed rejection of β2M-deficient and m157tg target cells to background levels (7.7 ± 5.3% and 6.4 ± 5.0%, respectively; not depicted). Increased NK cell cytotoxicity was unlikely the consequence of NK cell expansion because absolute NK cell numbers were only slightly increased, with differences reaching statistical significance when the data from several experiments were pooled (Fig. 3 D). Importantly, enlarged spleens in T reg cell–depleted mice also harbored proportionally more target cells, leading to comparable ratios of NK effector to control target cells in mock- and DT-treated mice (Fig. 3, E and G). Moreover, the enhanced rejection was specific to β2M-deficient target cells because NK cell activity against m157tg targets coinjected into the same animals was unaffected. NK cell–dependent rejection of target cells lacking classical MHC class I molecules H-2K^b and -D^b was similarly increased in the absence of T reg cells (Fig. 3 F). These data suggested that T reg cells control the activation of NK cells in the absence of inhibitory receptor engagement.

Interestingly, the increased cytotoxicity against β2M-deficient targets was abolished upon concomitant depletion of CD4⁺ and CD8⁺ T cells (not depicted), suggesting that the presence of activated T cells rather than the lack of T reg cells was responsible for increased NK cell cytotoxicity in DT-treated mice. Based on these findings, it seemed likely that T cell–dependent factors helped the NK cells to gain function. Multiple studies suggested that lymphokines, particularly IL-2, can activate NK cells in vitro, or when therapeutically administered in vivo (Henney et al., 1981; Caligiuri et al., 1993). IL-2–dependent help has been proposed to play a role for IFN-γ production by NK cells during infection (Bihl et al., 2010; Horowitz et al., 2010; Lee et al., 2012). It was not known, however, whether in the absence of infection T cells help NK cells through provision of IL-2 and whether T reg cells have the capacity to interfere with this help. To test whether increased amounts of IL-2 could account for the T cell–dependent increase of NK cell cytotoxicity in T reg cell–depleted mice, we treated these and control mice with IL-2 blocking antibody JES6-1A12 or isotype control IgG. IL-2 neutralization prevented an increase in NK cell cytotoxicity against missing-self targets in the absence of T reg cells (Fig. 4 A). Importantly, rejection of m157tg MHC class I–sufficient targets was not affected by IL-2 blockade (not depicted). Therefore, increased levels of IL-2 selectively boosted cytotoxicity against missing-self. Neutralization of IL-2 did not prevent the activation of T cells in T reg cell–depleted mice as we found similar frequencies of CD44^hi and IFN-γ–producing CD4⁺ T cells and only a modest reduction of activated CD8⁺ T cells in JES6-1A12– or isotype control–treated mice (Fig. 4 B). Interestingly, depletion of CD4⁺ T cells, which readily produce IL-2 in T reg cell–depleted mice (Fig. 4 C), fully recapitulated the effect of IL-2 blocking antibodies (Fig. 4 A). Together, these data suggested that CD4⁺ T cell–derived IL-2 activated NK cells in the absence of T reg cells. The neutralization of IL-2 in mock-treated mice harboring T reg cells did not alter NK cell cytotoxicity. Therefore, T reg cells appeared to limit amounts of IL-2 under steady-state conditions and, thereby, dampen NK cell activation. Thus, T reg cells restrained the IL-2–dependent CD4⁺ T cell help for NK cell cytotoxicity in the absence of inhibitory receptor signaling. In contrast, killing of MHC-sufficient m157tg targets, which can engage both activating and inhibitory receptors, was not affected by T reg cell depletion with or without the concomitant depletion of effector T cells or the neutralization of IL-2.

These observations raised the question of how IL-2 could affect the NK-dependent rejection of missing-self targets while not affecting non-self targets. Previously, effects of IL-2 on the NK cell expansion and on granzyme B and perforin mRNA translation have been suggested to increase cytotoxicity against a broad range of targets (Fehniger et al., 2007). We hypothesized that the observed selective impact of the relief from T reg cell–mediated control on the in vivo responsiveness of NK cells against MHC class I–deficient targets was a consequence of a short-term effect of IL-2 on proximal events in target cell recognition or NK cell activation rather than a long-term generalized potentiation of NK cell cytotoxicity. Interestingly, a recent study suggested that IL-2 rescues the function of human WASP-deficient NK cells through actin reorganization (Orange et al., 2011). This observation in conjunction with our data raised the possibility that IL-2 can act early on target cell recognition by wild-type NK cells. To test this hypothesis, we performed live cell imaging of the interactions of FACS-purified NK cells with RMA-S target cells. In contrast to their unstimulated, spherically symmetrical counterparts, IL-2–pretreated NK cells appeared elongated and readily adhered to target cells (Fig. 5 A). To visualize productive encounters with their targets, NK cells were preloaded with the calcium-sensitive dye Fura-2 to monitor activation-induced Ca²⁺ flux. We found that IL-2–treated NK cells had a significantly increased contact efficiency as determined by the number of target cell contacts that cells underwent before initiating Ca²⁺ flux for the first time (Fig. 5 B). Furthermore, whereas only ∼20% of mock-stimulated NK cells eventually engaged with target cells in a manner leading to the initiation of Ca²⁺ flux during the 45-min observation period, Ca²⁺ flux was observed in ∼70% of NK cells that had been preincubated with 250 U/ml IL-2 (for 60 min; Fig. 5 C). Consistent with these results, a brief exposure to IL-2 amplified up to approximately fourfold the NK cell conjugate formation with MHC class I–deficient RMA-S target cells in a dose-dependent manner but had a much reduced effect (∼1.2 increase) on the conjugate formation with Ba/F3-m157 cells (Fig. 5, D and E). In addition, punctual IL-2 stimulation was sufficient to increase cytotoxicity of NK cells (Fig. 5 F). It is noteworthy that IL-2 differentially affected NK cell activity against the two types of targets, boosting RMA-S lysis by >5-fold, whereas lysis of Ba/F-m157 was increased only marginally (∼1.4-fold). Therefore, the ability of IL-2 to modulate NK cell cytotoxicity directly correlated with its ability to increase target–cell conjugation. This observation raised the possibility that the preferential IL-2–dependent rejection of MHC class I–deficient versus m157tg target cells in mice subjected to T reg cell depletion (Fig. 3 B) could be explained by its differential impact on the ability of NK cells to engage target cells. To test this possibility, we FACS-purified NK cells from T reg cell–depleted and control mice and tested their ability to form conjugates with target cells. In agreement with the aforementioned in vivo and in vitro experiments, we found that in the absence of T reg cells, NK cells had an increased ability to engage RMA-S cells, whereas NK cell conjugate formation with Ba/F3-m157 cells was unaffected (Fig. 5 G). Furthermore, targeting of IL-2 bound to S4B6 antibody to NK cells was sufficient to rapidly augment the rejection of MHC class I–deficient but not m157tg target cells (Fig. 5 H).

The observation that IL-2 rapidly increased the adhesiveness of NK cells to RMA-S cells lacking MHC class I raised the question of whether this effect was specific for missing-self targets or whether it could also be relevant for a broad range of target cells. We found that a brief preincubation with IL-2 dramatically boosted cytotoxicity against RMA, RMA-S, and Ba/F3 cells, which in the absence of IL-2 were inefficiently killed. In contrast, IL-2 had little effect on the killing of target cells that were readily lysed by NK cells without IL-2 stimulation (Fig. 6 A). Consistently, IL-2 also boosted NK cell conjugation to these weak targets but had little impact on NK cell adhesiveness to strong targets (Fig. 6 B). Importantly, the ability of IL-2 to increase adhesiveness to weak suboptimal targets correlated with its effect on cytotoxicity (Fig. 6 C), suggesting that this adaptive cytokine is important for the regulation of NK cell functional affinity for weak target cells.

In conclusion, our results suggest that IL-2 can rapidly enhance NK cell cytotoxicity by boosting their ability to engage weak target cells. T reg cells control this CD4⁺ T cell–dependent tuning of NK cell responsiveness by limiting the amounts of available IL-2 and therefore regulate NK cell function by interfering with this previously unappreciated adaptive-innate lymphocyte cross talk.

An essential role for T reg cells in the immune homeostasis is highlighted by the finding that their depletion leads to the activation and expansion of CD4⁺ and CD8⁺ T cells and B cells (Kim et al., 2007). Here, we asked whether T reg cells restrain NK cells, a third lymphocyte lineage with the capacity to recognize self- and foreign antigens. Given the parallels between T and NK cell biology, we anticipated a prominent role for T reg cell–mediated suppression of NK cells. We found that T reg cells were dispensable for NK cell tolerance to an activating self-ligand and seemed to only modestly influence the responsiveness of NK cells to a non-self–ligand. However, we observed increased cytotoxicity of NK cells against missing-self targets in the absence of T reg cells. Because any gain of NK cell function in T reg cell–depleted mice was dependent on the presence of activated CD4⁺ T cells and IL-2, these observations suggested that T reg cells restricted IL-2–dependent T cell help for NK cell function rather than directly suppressing NK cells. These results provide experimental support for the idea that the therapeutic manipulation of T reg cells could influence NK cells mainly by indirect means (Barao et al., 2006; Terme et al., 2008; Galani et al., 2010). Our findings are consistent with the idea that CD4⁺ T cells capable of activating other immune cell types are the intended targets of T reg cell–mediated regulation (Kim et al., 2007).

The notion that T reg cells restrain NK cells by limiting T cell help does not exclude the possibility that a particular local inflammatory environment might enable direct T reg cell–mediated control of NK cell responsiveness. For example, T reg cells may directly suppress NK cell responses in settings where they serve as major producers or exhibitors of TGF-β, an important negative regulator of both NK and effector T cell activation (Laouar et al., 2005; Li et al., 2006). Indeed, T reg cells isolated from tumors suppressed NKG2D expression and tumor cell lysis by NK cells in a TGF-β–dependent manner (Ghiringhelli et al., 2005; Smyth et al., 2006). However, inhibition of NK cells by these tumor-derived T reg cells was neutralized in the presence of IL-2 (Ghiringhelli et al., 2005).

Our study provides experimental evidence of IL-2–mediated T cell help for NK cells. Early studies indicated that NK cell function can be boosted upon provision of IL-2 in experimental and therapeutic settings (Henney et al., 1981; Caligiuri et al., 1993). IL-2–mediated T cell help has been suggested to occur in vivo in the context of infection (Bihl et al., 2010; Lee et al., 2012), but its role under physiological conditions or in autoimmunity remained unclear. Our observation that T reg cells restrict access of NK cells to IL-2 might explain why T cell help for NK cells has been difficult to observe in the presence of competent T reg cells. Neutralization of IL-2 had no effect on NK cell cytotoxicity in control mice, suggesting that basal IL-2 amounts produced by T cells are not accessible to NK cells in the presence of T reg cells. This might be explained by the lack of CD25 expression on resting mature NK cells in mice. Therefore, adaptive help for NK cells in the presence of T reg cells might only occur in situations of markedly increased IL-2 production or CD25 expression on NK cells. Interestingly, during systemic viral infection the induction of IL-12 and IL-18 can drive high-level CD25 expression on NK cells (Lee et al., 2012), which might allow NK cells to compete with T reg cells for IL-2. Thus, in addition to direct activation of NK cells, CD25-inducing therapeutic (e.g., intratumoral) delivery of IL-12 and IL-18 might enable NK cells to receive T cell help even in the presence of T reg cells (Ni et al., 2012).

In contrast to previously described effects of IL-2 (Henney et al., 1981; Caligiuri et al., 1993; Fehniger et al., 2007; Lee et al., 2012), our experiments revealed a context-dependent, immediate role of IL-2 for NK cells in vivo. The presence of IL-2 did not result in the generalized activation of NK cells because killing of m157 targets was not affected by the presence or absence of T reg cells. However, IL-2 increased the in vivo rejection of MHC class I–deficient target cells. Thus, IL-2–induced potentiation of NK cell effector function seemed to occur in the absence of inhibitory receptor signaling. However, IL-2 did not further boost NK cell responses to a strong activating signal in the presence of inhibitory ligands. Therefore, it seems likely that the balance between activating and inhibitory signals defines the impact of IL-2–dependent help on NK cell function and its control by T reg cells. IL-2 rapidly lowered the activation threshold of NK cells by tuning their ability to adhere to and engage with their targets, suggesting potential interactions of IL-2R with integrin or activating-receptor signaling pathways. Integration of activating and inhibitory signals occurs at the interface with the target cell. A study in T cells suggested that the IL-2R might be recruited to the immunological synapse (Maldonado et al., 2004). Furthermore, a cross talk between cytokine and activating receptor signaling in NK cells has been described for IL-15 and NKG2D (French et al., 2006; Horng et al., 2007). We found no evidence for the IL-2–dependent enhancement of Ly49H- and NKG2D-dependent interactions, but these results do not exclude IL-2–mediated potentiation of NK cell cytotoxicity through the priming of other activating receptor pathways (e.g., downstream of SLAM receptors). MHC class I–reactive inhibitory receptors could potentially block such a receptor cross talk and therefore explain the IL-2–dependent increase of NK cell responses against missing-self targets. Of particular interest in this regard are SHP tyrosine phosphatases, which associate with inhibitory receptors and can be recruited to the IL-2R complex and inhibit IL-2 signaling (Migone et al., 1998; Xiao et al., 2009). Interestingly, we found that IL-2 rapidly boosted the ability of NK cells to engage with their target cells. However, the IL-2–dependent modulation of conjugate formation and cytotoxicity was also observed when using MHC class I–sufficient RMA or Ba/F3 target cells. These results suggested that IL-2 drastically amplifies NK cell responsiveness against weak targets, but it had little effect on the conjugate formation with and the killing of strong targets such as YAC, Ba/F3-Rae1, and Ba/F3-m157 expressing high amounts of activating ligands. MHC class I–deficient and m157tg cells used in our in vivo experiments likely represent weak and strong targets, respectively. MHC class I–deficient cells express endogenous activating ligands that are normally insufficient to overcome the threshold imposed by inhibitory receptors, whereas m157tg cells express high-affinity viral ligands that activate NK cells even in the presence of inhibitory receptor signaling. Thus, IL-2–mediated help might be dispensable for NK cell responses to strong activating signals but is likely a critical determinant of NK cell–mediated cytotoxicity against weak targets.

Our results suggest that an IL-2–dependent adaptive-innate lymphocyte cross talk regulates NK cell function during early stages of target cell recognition. By restricting the availability of IL-2, T reg cells provide an additional checkpoint for NK cell activation. The ability of T reg cells to deprive NK cells from access to IL-2 might be of special relevance in autoimmune lesions, where autoreactive T cells produce IL-2 and could thereby unleash NK cell activity against weak tissue targets. Consistent with this idea, T reg cell depletion in prediabetic BDC2.5/NOD mice rapidly activated NK cells (Feuerer et al., 2009), which appears to depend on T cell–derived IL-2 (in this issue, see Sitrin et al.). The different kinetics of NK cell activation in the BDC2.5/NOD and Foxp3^DTR mice used in our study might therefore be explained by the different kinetics of T cell activation.

An important clinical implication of our findings is that IL-2 doses that override the IL-2 buffering capacity of T reg cells will likely boost NK cell cytotoxicity despite the presence of T reg cells. Indeed, treatment of type I diabetes patients with a combination of rapamycin and IL-2 increased NK cell numbers and β cell dysfunction despite markedly increased T reg cell numbers (Long et al., 2012). Our data also suggest that therapies that can increase IL-2 availability might have immediate effects on the activity of NK cells.

The indirect regulation of NK cell activity by T reg cells is likely relevant in the context of tumor growth. A recent report suggests increased control of MHC class I–deficient lymphomas by NK cells that expand in response to IL-2 from homeostatically activated T cells (Ni et al., 2012). Intriguingly, several tumors down-regulate MHC class I expression. Recently, frequent mutations in the β2M gene and the resulting loss of MHC class I expression have been reported in human B lymphoma patients (Challa-Malladi et al., 2011). One possible explanation of why these tumors are not sufficiently cleared by NK cells was provided by the observation that a large fraction of those tumors that lose MHC class I expression show concomitant mutations in the cell adhesion molecule CD58. These results together with our findings imply that NK target cell adhesion is a critical factor of tumor cell rejection likely controlled by T reg cells. Our data suggest that the provision of IL-2 either directly by NK cell–targeted IL-2–antibody complexes or indirectly might prove therapeutically beneficial in these patients. In summary, our study revealed that T reg cell control of IL-2 availability serves as an important checkpoint in the dynamic modulation of NK cell activation.

Foxp3^DTR mice (Kim et al., 2007) used in this study have been backcrossed to C57BL/6 for >16 generations. m157tg mice (Tripathy et al., 2008) were provided by S.K. Tripathy (Washington University in St. Louis, St. Louis, MO). β2M^−/− and K^bD^b−/− mice were purchased from Taconic. All of the mice were bred and housed in the specific pathogen–free animal facility at the Memorial Sloan-Kettering Cancer Center and used in accordance with the guidelines approved by the Institutional Animal Care and Use Committee.

To deplete T reg cells, Foxp3^DTR mice were injected i.p. with 50 µg/kg DT (Sigma-Aldrich) every other day (q2d). NK cells and CD4⁺ and CD8⁺ T cells were depleted by i.p. injections of 300 µg PK136 or 400 µg GK1.5 or 2.43 antibody, respectively (BioXcell). To allow for complex formation, 1 µg of recombinant mouse IL-2 (PeproTech) was incubated for 10 min at room temperature with 5 µg anti–IL-2 antibody (S4B6-1; BioXcell) before i.p. injection. To block IL-2 in vivo, mice received 200 µg of the indicated antibodies q2d i.p.

Cell suspensions were stained with fluorophore-conjugated antibodies purchased from eBioscience, BioLegend, or BD. Intracellular staining of Ki-67, Foxp3, and Eomes was performed using the Foxp3 mouse T reg cell staining kit (eBioscience). To analyze cytokine production and degranulation, 10⁶ splenocytes/well were incubated on MaxiSorp 96-well plates (Thermo Fisher Scientific) coated with 10 µg/ml of agonist antibodies against Ly49D or Ly49H (provided by L. Lanier, University of California, San Francisco, San Francisco, CA) for 5 h in the presence of 1 µg/ml Brefeldin A (Sigma-Aldrich) and 2.5 µg/ml CD107a antibody. Alternatively, cells were stimulated as indicated with 20 ng/ml IL-12 and 10 ng/ml IL-18 (R&D Systems) or 50 ng/ml PMA and 500 ng/ml ionomycin. Intracellular cytokine staining was performed using the Perm/Wash kit (BD). For live/dead discrimination, propidium iodide or LIVE/DEAD fixable yellow dye (Molecular Probes) was used. Cells were analyzed on a LSRII cytometer (BD) and analyzed with FlowJo software (Tree Star).

FACS-sorted NK cells were incubated in the presence or absence of 250 U/ml human IL2 (PeproTech) for 40 min and then loaded with 5 µg/ml Fura-2AM for 20 min and transferred into colorless RPMI (no phenol red) containing 5% FCS ± IL-2. 10⁵ NK cells were mixed 1:1 with RMA-S cells and imaged in 8-well chamber slides (Thermo Fisher Scientific) using an inverted fluorescence video microscope (IX-81; Olympus) fitted with a 20×, 0.75 NA objective lens (Olympus). A Xe lamp (DG-4; Sutter Instrument) was used for fluorophore excitation, and data were collected on an electron-multiplying charge-coupled device camera (ImagEM; Hamamatsu Photonics). Time-lapse recordings were made using SlideBook software (Intelligent Imaging Innovations). In general, one brightfield image and one Fura-2 image were taken every 10 s for 45 min. Contact efficiency (Fig. 5 B) was defined as the reciprocal of the number of target cell contacts made by each NK cell before its initial Ca²⁺ flux. For NK cells that did not flux Ca²⁺ despite 12 contacts with target cells, the contact efficiency was set to 0.

NK cells were stained with anti-NKp46 and FACS-purified using an Aria2 sorter (BD). NK cells were then incubated for 20–60 min in the presence or absence of 10–250 U/ml IL-2 in RPMI supplemental 10% or used directly ex vivo. 10⁴ NK cells were pelleted with 3 × 10⁴ target cells in 96-well V-bottom plates and incubated for 20 min. The plate was immediately transferred on ice, and samples were carefully resuspended in ice-cold PBS and analyzed on an LSRII flow cytometer.

Splenic NK cells were negatively enriched and incubated in triplicates with ⁵¹Cr-labeled target cells. Supernatants were assayed for ⁵¹Cr release after 5 h of incubation at 37°C, and percentage of specific lysis was calculated.

Splenocytes from CD45.1 BL6, CD45.1/CD45.2 m157tg, or CD45.2 β2M^KO mice were labeled with CellTrace Violet (Invitrogen), and 3–5 × 10⁶ cells/each were transferred by i.v. injection. 12–14 h after transfer, splenocytes were isolated and analyzed by FACS. Specific lysis of CD19⁺ cells was calculated based on the relative numbers recovered from each donor population compared with the mean ratios of cells recovered from two to three NK-depleted mice and in the input mix.

All statistical analyses were performed using Prism5 software (GraphPad Software). Results are expressed as means ± SD. Differences between individual groups were analyzed for statistical significance using the nonparametrical Mann-Whitney test. Only when comparing groups of n = 3 (triplicate cultures in in vitro killing and conjugate formation assays in Figs. 1 B, 2 C, and 5, F and G) was the unpaired Student’s t test used: *, P < 0.05; **, P < 0.01; ***, P < 0.001; ns, not significant.

We thank Sandeep K. Tripathy and Lewis Lanier for providing mice and critical reagents. We would also like to thank our laboratory members for helpful discussions and Wolfgang Kastenmüller for critically reading the manuscript.

This work was supported by an Irvington Fellowship of the Cancer Research Institute (to G. Gasteiger) and National Institutes of Health (NIH) grant R37 AI034206 (to A.Y. Rudensky). J.C. Sun is supported by the Searle Scholars Program and NIH grants AI085034 and AI100874. A.Y. Rudensky is an investigator with the Howard Hughes Medical Institute.

The authors declare no competing financial interests.