|Home | About | Journals | Submit | Contact Us | Français|
The NK cell activating receptor NKG2D plays a critical role in the destruction of malignant cells, but many of the cell-signaling mechanisms governing NKG2D-mediated cellular cytotoxicity are unknown. We have identified an NKG2D-mediated signaling pathway that governs both conjugate formation and cytotoxic granule polarization. We demonstrate that an interaction between the regulatory subunit of PI3K, p85, and the adaptor protein CrkL is required for efficient NKG2D-mediated cellular cytotoxicity. We show decreased NK cell - target cell conjugate formation in NK cells treated with PI3K inhibitors or depleted of CrkL. Independent of adhesion, we find that microtubule organization center polarization toward target cells expressing the NKG2D ligand MICA or toward anti-NKG2D coated beads is impaired in the absence of CrkL. Antibody-stimulated granule release is also impaired in NK cells depleted of CrkL. Furthermore, our data indicate that the small Ras-family GTPase Rap1 is activated downstream of NKG2D engagement in a PI3K- and CrkL-dependent manner and is required for conjugate formation, MTOC polarization and NKG2D-dependent cellular cytotoxicity. Taken together, our data identify an NKG2D-activated signaling pathway that collectively orchestrates NK cell adhesion, cell polarization and granule release.
Cytotoxic lymphocytes, such as NK cells and CD8+ T-lymphocytes, play a critical role in our body’s immune response to malignant and viral infected cells. A principle effector mechanism of cytotoxic lymphocytes is the process of cell-mediated cytotoxicity during which a target cell is destroyed by a cytotoxic lymphocyte. Cell-mediated cytotoxicity occurs in discrete steps: formation of a cytotoxic lymphocyte - target cell conjugate, polarization of the cytotoxic machinery toward the effector - target cell interface known as the cytotoxic synapse (CS)3, and finally lytic granule secretion (1, 2). Although the steps of cytotoxicity have been established, many questions remain regarding the regulation of these steps.
NKG2D is a non-ITAM containing NK- and T-cell activating receptor that recognizes “stress” ligands, such as the tumor-associated MICA (in human) and Rae-1 (in mouse), which are upregulated on the surface of cells undergoing genotoxic stress (3–5). Upon NKG2D ligation, the transmembrane protein DAP10 is phosphorylated by the Src kinase Lck (6). The signal then propagates down two distinct arms of the NKG2D signaling pathway – phosphatidylinositol-3 kinase (PI3K) and Grb2/Vav1 (4, 7, 8). Although a number of proteins have been shown to play a role downstream of these proximal signaling molecules, questions remain regarding the signaling pathways regulating NKG2D-mediated conjugate formation, granule polarization, and secretion.
Integrins have been established as regulators of cell adhesion and are required for proper NK cell - target conjugate formation. In fact, blocking β1 or β2 integrins inhibits NK cell-mediated cytotoxicity (9). Moreover, patients with leukocyte adhesion deficiencies display impaired NK cell function (10–12). ITAM-containing receptors have been shown to regulate integrins in T-lymphocytes, neutrophils, macrophages, and NK cells (13–15). Nolz and colleagues established a role for WAVE2 in integrin activation through Abl-and CrkL-C3G-mediated activation of Rap1 in T-cell signaling (16). Vinculin and talin have also been implicated in integrin affinity maturation downstream of the T-cell receptor (17). However, whether signaling pathways activated by NKG2D signaling affect integrin activation or NK cell adhesion is unknown. In fact, two recent reports did not observe any increase in conjugate formation upon stimulation of NKG2D suggesting that signaling through NKG2D/DAP10 may not activate integrins (8, 18).
The Crk family of adaptor proteins are a widely expressed family of signal transduction mediators involved in normal and malignant cell process (19). This family contains Crk, expressed as two isoforms - CrkI and CrkII, and CrkL (Crk-like) (19). CrkI has an N-terminal SH2 domain followed by a SH3 domain, whereas both CrkII and CrkL contain an additional C-terminal SH3 domain. These domains serve as binding sites for a number of proteins allowing Crk proteins to participate in multiple signaling pathways (19). Recent work has identified CrkL as an important mediator of integrin regulation in T-lymphocytes (16). Although no functional role for CrkL has been demonstrated in NK cells, Crk has been identified as a target following NK inhibitory receptor signaling (20). In this study, we demonstrate that NKG2D-mediated conjugate formation is PI3K-dependent, and involves the adaptor protein CrkL and the small Ras family-GTPase Rap1. Significantly, CrkL suppression results in diminished cytotoxicity, decreased conjugate formation, impaired MTOC polarization, and decreased granule secretion by NK cells. We also show that Rap1 is activated by NKG2D stimulation in a PI3K- and CrkL-dependent manner, and is required for NKG2D-mediated conjugate formation, MTOC polarization and cytotoxicity.
Reagents were purchased from Sigma unless otherwise stated. Recombinant ICAM-1 and VCAM-1, and mouse monoclonal IgG1 anti-NKG2D were purchased from R&D Systems. Human NK cells were cloned and passed as previously described (21). Both the wild type BaF3 myeloid cell line and the MICA expressing variant were a gift from Dr. L.L. Lanier (UCSF, San Francisco, CA). The anti-LFA-1 antibody, mHm23, was a gift from T. A. Springer (CBR Institute for Biomedical Research, Boston, MA). Mouse monoclonal IgG1 anti-FcR (CD16) was a gift from Dr. D. Segal (Immunology Branch, NCI, NIH, Bethesda, MD). Mouse monoclonal anti-(31 integrin (Chemicon), mouse monoclonal CrkL (ABNOVA), mouse monoclonal p85, clone AB6 (Upstate) were also used. Rabbit polyclonal anti-Rap1 was purchased from Cell Signaling Technology. Rabbit polyclonal anti-SLP76 was generated as previously described (22).
Adhesions assays were performed as previously described (16). Briefly, a 96-well flat bottom plate was coated overnight with ICAM-1 or VCAM-1, then blocked with 2.5% BSA in PBS. Human NK cells were stained with 5 µM calcein AM (Invitrogen) for 30 min in serum-free media and washed with PBS. Cells were stained with 10 µg/mL IgG, anti-FcR, or anti-NKG2D at 20 × 106 cells/mL then added to wells containing goat anti-mouse F(ab’)2. Plates were incubated on ice for 1 hr, then in a 37°C water bath for 15 min. After incubation, cellular input was determined using a fluorescent plate reader. Plates were then washed with PBS and the remaining fluorescence was measured to determined cell adherence.
Cell stimulation, immunoprecipitation, and immunoblot analysis were conducted as previously described (23).
CrkL, C3G, and Rap1 were suppressed using the nucleofection technique as previously described (7). Briefly, 300 pMol of siRNA per 5 × 106 human NK clones were introduced in human NK clones by Amaxa nucleofection using Soln V and Program O-17. Cells were then placed in RPMI containing 10% fetal bovine serum, 1% sodium pyruvate, 1% L-Glutamine, and 20 U/mL IL-2 and allowed to recover for 72 hours. The following target sequences used to generate siRNA duplexes against proteins in human NK cells: CrkL: GAGTTCTTTTGGATCATAA (targeted to the 3’ UTR, nucleotides 1634-1652 of NM_005207.3), C3G: GGGTTGTGTGAACTGAAAT (targeted to the 3’ UTR, nucleotides 5290–5308 of NM_005207.3), Rap1A: siGENOME SMART pool M-003623-02-005, targeted against NM_002884 (Dharmacon), Rap1B: siGENOME SMART pool M-010364-03-005, targeted against NM_001010942 (Dharmacon), Negative Control: TTCTCCGAACGTGTCACGT.
Wild type CrkL and Rap1A cDNA were cloned from human NK cell mRNA by RT-PCR. W160Y and W161Y mutations were introduced into the wild type CrkL cDNA to create the N-terminal SH3 CrkL mutant. The S17A mutation was introduced into wild type Rap1A cDNA to create a dominant negative (24). Wild type and mutant CrkL and Rap1A cDNA were introduced into the pSP11.FLAG vector as previously described (25). The sequence of wild type and mutant CrkL and Rap1A vaccinia were confirmed. Human NK cells were infected at a multiplicity of infection of 20:1 for CrkL, or 5:1 and 30:1 with F.Rap1A and F.Rap1 S17A respectively with appropriate WR compensation. Cells were infected for 5 hrs at 37°C in serum-free media.
Human NK clones were stained with PKH26 per included protocol and resuspended at 0.5 × 106 cells/mL in RPMI. Target cells were stained with either 50 µg/mL 5-sulfofluorescein diacetate (SFDA) (Invitrogen) in PBS for 1 hr in a 37°C water bath or 50 µM 7-amino-4-chloromethylcoumarin (CMAC) (Invitrogen) for 30 min in serum-free media then quenched for 15 min in RPMI+10%CS. Target cells were washed twice in PBS and resuspended at 0.5 × 106 cells/mL in RPMI. 0.25 × 106 NK cells were mixed with 0.25 × 106 target cells and spun at 500 rpm for 5 min. The resulting pellet was placed in a 37°C water bath for the desired time. 0.5 mL of media was removed from each sample after all samples had been heated for the desired amount of time. Samples were then vortexed at maximum speed for 5 sec to disrupt non-specific conjugates and then fixed with 0.5 mL of 4% paraformaldehyde. Conjugate formation was analyzed using a LSRII flow cytotometer (Becton Dickenson Immunocytometry Systems). The SFDA was excited with a 488 nm laser and the emission was measured through a 530/30 nm BP filter. CMAC was excited with a 355 nm laser and the emission was measured through a 450/65 nm BP filter. The PKH26 was excited with a 532 nm laser and the emission was measured through a 575/26 nm BP filer. Data was analyzed with the FlowJo software package (Treestar).
51Cr release assays were performed as previously described (21). NK cell cytotoxicity was triggered either by coating P815 tumor targets with a “triggering” antibody, such as anti-NKG2D, or the BaF3 myeloid cell line expressing the NKG2D receptor ligand MICA.
GST-CrkL SH2, GST-CrkL SH3-N, and GST-CrkL SH3-C were constructed as previously described (26). Briefly, 10 × 106 per sample human NK clones were stimulated with anti-NKG2D and cross-linked, then lysed on ice for 10 min with buffer containing 50 mM Tris Base, 20 mM EDTA, and 1% Igepal CA-630. Lysates were clarified by centrifugation and transferred to GSH-agarose coated with 40 µg GST fusion protein. Lysates and beads were rotated at 4°C for 1 hr. Protein complexes were eluted with 40 µL SDS sample buffer, resolved by SDS-Page, and transferred to Immobilon-P membranes (Millipore). The presence of protein complexes was detected with primary antibodies followed by goat α-mouse-HRP (Santa-Cruz) or goat α-rabbit HRP (Santa-Cruz) and visualized with the Super Signal detection system (Pierce Biotechnology).
Immunoflouresence staining of NK cell - target cell and NK cell - bead conjugates was performed similarly to T-cell - B-cell conjugate immfluoresence described previously (16, 27). Briefly, BaF3 and MICA tumor targets were stained with 50 µM 7-amino-4-chloromethylcoumarin (CMAC) for 30min in serum-free media then quenched for 15 min in RPMI+10%CS. For NK - bead conjugates sulfate latex beads (Invitrogen) were coated with 2 µg anti-NKG2D or IgG for 1.5 hrs while rocking at 37°C. Beads were washed 3 times in PBS + 2.5% BSA. NK - target cell conjugates were formed by conjugating human NK clones with either MICA expressing or control BaF3 target cells at an E:T ratio of 2:1. NK - bead conjugates were formed by conjugating human NK clones with antibody coated sulfate latex beads at an E:T ratio of 1:1. Cells were conjugated by spinning NK - targets or NK - beads at 500 rpm for 5 min at room temperature, incubated at 37°C for 10 min, transferred to poly-L-lysine coated coverslips and allowed to settle on coverslips for 4 min at 37°C. Conjugates were washed, fixed with 4% paraformaldehyde and then permeablized with 0.1% Triton X-100. Conjugates were blocked with PBS containing 1% glycerol, 0.1% Gelatin from Cold Water Fish, 5% normal goat serum, 0.1% BSA, and 0.4% azide, then stained with primary antibody followed by either a AlexaFlour568 goat anti-mouse (Invitrogen), Cy5-goat anti-mouse (Zymed). Anti-α-tubulin-FITC was also used for microtubule organizing center (MTOC) staining. Actin was stained with AF488- or Rhodamine-Phalloidin (Invitrogen). Coverslips were mounted in ProLong Gold (Invitrogen). Fixed conjugates were examined at room temperature on a Zeiss Axiovert 200 with an Axiocam HRC digital camera using a 100x/1.4 NA oil-immersion DIC objective or Zeiss LSM 510 scanning confocal microscope using a 100x/1.4 NA oil-immersion objective. Image acquisition was done with either the Zeiss Axiovision or LSM software packages. Figure construction of images was performed in Adobe Photoshop and Illustrator CS3.
The amount of GTP bound Rap1was determined by a RalGDS RBD pulldown assay as previously described (28). Briefly, human NK cells were stimulated with either anti-NKG2D or anti-FcR and cross-linked, then resuspended in buffer containing 50 mM Tris (pH 7.5), 50 0mM NaCl, 5 mM MgCl2, 10% Glycerol, 0.5% NP-40 (Igepal CA-630). Cells were immediately spun at 25,000 × g for 5 min at 4°C. Lysates were transferred to GSH-Agarose coated with GST-Ral-Ras-binding domain (RBD). Lysates and beads were rotated at 4°C for 15 min. A portion of the lysates was reserved to determine the total amount of Rap1 in the lysate (input). Proteins were eluted and visualized as previously described (23). Densitometry was performed using ImageJ (NIH), corrected for total Rap1A loading, and then plotted as fold change relative to control baseline.
Granule secretion was determined in human NK cells as previously described (29). Briefly, human NK clones were stimulated by in a 96-well plate with wells that had been previously coated with anti-FcR, anti-NKG2D, anti-CD56, or TPA and Ionomycin. After a 4 hr incubation at 37°C, supernatants were harvested. The granzyme A activity of the supernatants was then assessed by measuring N-α-carbobenzoxy-L-lysine-thiobenzyl (BLT)-esterase activity.
The importance of NKG2D as an activating receptor of NK cells has been clearly established (4, 26), but the interplay between integrins and NKG2D mediated cytotoxicity is unclear, despite the fact that integrins are know to play a critical role in natural cytotoxicity (30). Two recent reports did not observe any increase in conjugate formation upon stimulation of NKG2D suggesting that signaling through NKG2D/DAP10 may not activate integrins (8, 18). To more directly examine whether NKG2D signaling might regulate integrin activation, we determined whether stimulation through NKG2D activates integrins in clonally derived, non-transformed, IL-2-activated human NK cells. Unstimulated NK clones displayed limited adhesion to VCAM-1 or ICAM-1 coated surfaces. In contrast, 30.1 ± 2.5% of NK clones stimulated with anti-FcR and 27.2 ± 3.9% of clones stimulated with anti-NKG2D adhered to 0.1 µg/well of VCAM-1 compared to 15.8 ± 3.2% of clones treated with an IgG control (Figure 1A). Similarly, 21.4 ± 1.5% of NK clones stimulated with anti-FcR and 18.3 ± 2.2% of clones stimulated with anti-NKG2D adhered to 0.1 µg/well of ICAM-1 compared to 8.8 ± 0.4% of clones treated with an IgG control (Figure 1B). To confirm that these effects were specific to the receptor, and not to non-specific antibody binding to the FcR expressed on the surface of human NK clones, we examined granule secretion by cross-linking the NK cell surface receptor CD56, which would not be expect to trigger granule release. As expected, ligation of FcR or NKG2D enhanced secretion, whereas ligation of CD56 did not increase secretion demonstrating that NK cell activation in response to anti-FcR or anti-NKG2D is not a result of non-specific binding to the NK cell FcR (Supplemental Figure 1.) These results demonstrate that stimulation through either FcR or NKG2D increases β1- and β2-integrin mediated adhesion of human NK cells.
Conjugate and CS formation have been established as critical steps in NK cell-mediated cytotoxicity (31). Integrin recruitment to the CS has been demonstrated in a K562 tumor target model of natural cytotoxicity, but has not been examined in primary human NK cells activated by ligation of NKG2D with its natural ligand, MICA (1). BaF3 cells stably expressing MICA were conjugated with NK cells to determine if integrins are recruited to the CS by NKG2D ligation (Figure 1C). β1- and β2-integrins localized diffusely within the plasma membrane of isolated NK clones. In NK cells conjugated with wild type BaF3 cells, β1- or β2-integrins did not localize to the CS. However, β1 integrins were recruited to the NK CS in 48.7 ± 5.5% of NK/MICA conjugates versus 7.7 ± 1.5% in NK/BaF3 conjugates. Similarly, β2 integrins were recruited to the NK CS in 51 ± 6% of NK/MICA conjugates versus 11 ± 1% in NK/BaF3 conjugates. These data demonstrate that β1- and β2-integrins are recruited to the NKG2D-mediated CS.
The role of LFA-1 in NKG2D-mediated cytotoxicity in NK cells has been suggested in previous studies but has not been established in primary human lymphocytes (32, 33). In this study, we used human NK clones and two models of cytotoxicity, reverse-ADCC and MICA expressing BaF3 cells, to establish a role for the β2 integrin, LFA-1, in NKG2D-triggered NK cell-mediated cytotoxicity. Treatment of NK clones with 1 µg/mL of the anti-LFA-1 blocking antibody, mHm23, decreased NK cell conjugate formation greater than 50% compared with IgG control antibody (Figure 1D). mHm23 treatment affected cytotoxicity in our reverse-ADCC model and caused a significant decrease in the killing of MICA expressing BaF3 cells (Figure 1E). To show that lysis of MICA expressing BaF3 cells is specific to the ligation of NKG2D, we inhibited NKG2D-mediated NK cell activation with an anti-NKG2D blocking antibody, which confirmed that MICA engagement by NKG2D was required for killing target cells (Supplemental Figure 2). These results establish a role for LFA-1 in NKG2D-mediated cytotoxicity in primary human NK cells.
PI3K has been established as an important signaling mediator in NKG2D-stimulated NK cell cytotoxicity (4). Although Giurisato et al. reported that PI3K did not affect basal conjugate formation in the NK cell tumor line NK92 (18), this is surprisingly not the case in primary human NK cells. In fact, NK clones treated with the classic PI3K inhibitors wortmanin or LY294002 decreased conjugate formation by >50% (Figure 2A). Cytotoxicity was also substantially decreased in NK clones treated with either PI3K inhibitor. As expected Akt phosphorylation was decreased with wortmannin treatment (Figure 2B). Therefore, in primary human NK clones, PI3K activity regulates NKG2D-mediated conjugate formation and subsequent cytolysis.
CrkL has been shown to regulate integrins following TCR ligation in T-lymphocytes (16). We therefore examined whether CrkL regulates integrin function in NK cells. CrkL was clearly recruited to the cytotoxic synapse in NK clones conjugated with MICA targets but not in control BaF3 cells (Figure 3A). In fact, CrkL was recruited to NK CS in 33.3 ± 5.0% of NK/MICA conjugates versus 3.3 ± 1.2% in NK/BaF3 conjugates (Figure 6A). The staining of CrkL at the CS was highly similar to the polarized F-actin polymerization seen in NK – MICA cell conjugates. Next, we explored phosphorylation of CrkL in NK cells upon stimulation of NKG2D. Crk family members are initially recruited by activating receptors, but then phosphorylated by Abl tyrosine kinase at and inhibitory residue in order to terminate signaling. In fact, phosphorylation of CrkL occurs in TCR-mediated signaling regulating integrin activation (16). Interestingly, Crk was recently shown to be phosphorylated in an Abl-dependent manner upon ligation of inhibitory KIRs in order to terminate signaling from activating receptors (20). Significantly, crosslinking NKG2D on NK clones results in a time-dependent phosphorylation of CrkL, further confirming that CrkL is recruited to NKG2D-dependent signaling pathways (Figure 3B). Additionally, to determine the importance of CrkL in NK cell-mediated cytotoxicity, we suppressed CrkL in human NK clones with siRNA and used these clones in a cytotoxicity assay against various NK-sensitive targets. CrkL suppression resulted in a significant decrease in cytotoxicity against P815 tumor targets coated with either anti-NKG2D or anti-FcR, and against MICA-expressing BaF3 cells (Figure 3C). These results indicate that CrkL mediates NK cell-mediated cytotoxicity triggered by multiple activating receptors, including NKG2D.
Given that integrins are critical to conjugate formation in NK cell-mediated cytotoxicity and the known requirement of CrkL for integrin function in T-cells, we suppressed CrkL in human NK cells to explore the role of CrkL in NK – target cell conjugate formation. CrkL suppression in NK clones results in a >90% decrease in conjugate formation between NK cells and MICA-expressing BaF3 cells compared to NK clones treated with a negative control siRNA (siNeg) (Figure 4A), implicating CrkL in NKG2D-mediated cell – cell adhesion.
Our data demonstrate that NKG2D-mediated conjugate formation is both CrkL- and PI3K-dependent. Yet it remained possible that suppression of CrkL was affecting global signaling from NKG2D through an unknown mechanism. In order to examine this we measured the phosphorylation kinetics of Akt and Erk, two PI3K-regulated pathways (34), in NK cells suppressed for CrkL. Importantly, treatment of NK cells with siCrkL did not did not affect Akt or Erk phosphorylation compared to siNeg treated controls (Figure 4B). Therefore, CrkL suppression does not affect either of these PI3K-dependent pathways. Moreover, NKG2D-dependent tyrosine phosphorylation of Vav1, SLP76 and PLCγ2 are unaffected by suppression of CrkL (C.M. Segovis and D.D. Billadeau, unpublished observation).
We next examined secretion in NK cells treated with CrkL siRNA (siCrkL). Surprisingly, siCrkL-treated NK clones displayed decreased secretion when stimulated by either anti-FcR or anti-NKG2D (Figure 5A). This effect was observed at multiple antibody concentrations: 30% and 46% decrease at 1 µg/mL and 0.3 µg/mL anti-FcR respectively, and 84% and 87% decrease at 1 µg/mL and 0.3 µg/mL anti-NKG2D respectively. TPA/Ionomycin induced secretion was not affected by CrkL suppression indicating that granule/membrane fusion was not affected, thus suggesting a role for CrkL in establishing polarized granule secretion in NK cells.
Microtubule polarization to the CS is required for localized release of granules (35). We therefore examined MTOC polarization in primary human NK cells treated with CrkL siRNA. CrkL suppression significantly impaired MTOC polarization to the CS of NK – MICA targets compared to NK clones treated with negative control siRNA (Figure 5B). Since the loss of MTOC polarization in the siCrkL NK clones could be due to diminished conjugate formation (Figure 4), we used anti-NKG2D beads to explore the role of CrkL in integrin-independent MTOC polarization. A 52% decrease in MTOC polarization was observed in NK clones treated with siCrkL in NK – bead conjugates compared to siNeg-treated NK clones (Figure 5C). These data demonstrate that CrkL regulates not only conjugate formation, but also MTOC polarization and subsequent granule secretion, all of which are required for NK cell-mediated cytotoxicity.
PI3K is a critical proximal signaling regulator of NKG2D-mediated NK cell cytotoxicity (4, 36). Our results indicate that both PI3K activity and CrkL regulate NKG2D-mediated adhesion, suggesting that CrkL may signal downstream of PI3K. To test this, we treated NK clones with wortmannin and determined if CrkL recruitment to the CS was affected.
Treatment of human NK clones with wortmannin decreased recruitment of CrkL to the CS of NK – MICA target cell conjugates by 78% (Figure 6A). However, we cannot rule out the possibility that this defect in CrkL recruitment is a result of defective conjugate formation in the wortmannin-treated NK cells. Since CrkL has been observed in complex with p85 in NK cells (37), we used immunoprecipitation (IP) to establish an association between CrkL and p85 in our system and to determine if the interaction was NKG2D-induced. Indeed, CrkL co-immunoprecipitatred p85 in both unstimulated and NKG2D– stimulated human NK clones indicating a constitutive interaction between these two molecules (Figure 6B). Similarly, an IP of p85 indicated constitutive p85-CrkL association (Figure 6B). The N-terminal SH3 domain of CrkII has been shown to interact with the proline-rich region of p85, the regulatory subunit of PI3K, upon TCR-activation, and in vitro an interaction has been demonstrated between the SH3 domains of CrkL and p85 (38, 39). However, which of the two CrkL SH3 domains interacts with p85 is unknown in NK cells. Using GST fusion proteins of individual SH2 and SH3 domains of CrkL, we found that the N-terminal SH3 domain of CrkL bound p85 (Figure 6C). We next sought to determine the functional significance of this interaction in NKG2D-mediated cellular cytotoxicity by analyzing whether disruption of this p85-CrkL complex would lead to impaired killing. To test this, we performed siRNA – rescue experiments in which wild type CrkL or an N-terminal SH3 mutant (SH3-Nmt) were overexpressed in human NK clones pretreated with siCrkL. Significantly, overexpression of wild type CrkL partially rescued the decrease in NKG2D-dependent cytotoxicity seen with CrkL suppression, whereas the CrkL SH3-1 was unable to rescue the cytotoxic defect in the CrkL suppressed cells (Figure 6D). These results demonstrate a regulatory role for the N-terminal SH3 domain of CrkL in NKG2D-mediated NK-cell cytotoxicity. Taken together, our results demonstrate that p85 and CrkL are in a complex, and that this interaction regulates the development of NKG2D-mediated cytotoxicity in human NK cells.
Rap1 has been demonstrated to be critical regulator of integrin activation in T-lymphocytes through a CrkL/C3G complex (16), but no role for Rap1 has been established in human NK cells (40, 41). Ligation of either the FcR or NKG2D results in activation of Rap1 (Figure 7A). To confirm these results, we studied the activation of MST1 upon FcR and NKG2D receptor ligation. The Rap1-RAPL-MST1 signaling cascade has been shown to be required for adhesion and cell polarity in lymphocytes in response to chemokine stimulation, and activation of Rap1 results in phosphorylation of MST1 (42). Ligation of either FcR or NKG2D increased phosphorylation of MST1 supporting the idea that Rap1 is coupled to these activating receptors in NK cells (Figure S3). Previous work has demonstrated that CrkL regulates the GTP-loading and activation of Rap1 downstream of several cell surface receptors (43–45). Since Rap1 is activated by NKG2D ligation we asked if either PI3K activity or CrkL regulates Rap1-GTP loading. In fact, treatment with wortmannin (Figure 7B) or siRNA depletion of CrkL (Figure 7C) abrogated NKG2D-stimulated Rap1 activation. Taken together, these data indicate that Rap1 is activated following FcR and NKG2D ligation in a PI3K- and CrkL-dependent manner.
In order to determine the role of Rap1 in NKG2D-mediated cytotoxicity in human NK cells, we initially expressed a wild type or dominant negative version of Rap1 (S17A) in NK cells and examined NKG2D-mediated killing (24). In contrast to expression of wild type Rap1, which did not affect killing through NKG2D, expression of Rap1S17A inhibited killing of either MICA expressing BaF3 cells or P815 cells coated with anti-NKG2D (Figure 7D). Since overexpression of S17A Rap1A may have off-target effects associated with its dominant-negative function, we utilized siRNA to suppress Rap1A and Rap1 B in NK clones as both Rap1A and Rap1B are expressed in human NK cells (Figure 7E, immunoblot). Importantly, suppression of either Rap1A or Rap1B decreased NKG2D-mediated cytotoxicity against multiple tumor targets, which was further diminished by suppression of both Rap1 isoforms (Figure 7E).
Given that both PI3K and CrkL regulate NKG2D-dependent Rap1 activation and conjugate formation, we used dominant negative Rap1A to assess the role of Rap1 in NK – target conjugate formation. Indeed, overexpression of Rap1A had no effect on conjugate formation between NK cells and MICA expressing BaF3 cells compared to WR infected cells (Figure 8A). In contrast, overexpression of S17A Rap1A significantly decreased conjugate formation between NK cells and MICA expressing targets equal to conjugate formation between WR expressing NK clones and BaF3 control targets (Figure 8A). We also examined the effect of Rap1A or dominant negative Rap1 on integrin-independent MTOC polarization in primary human NK cells. Rap1A overexpression did not affect MTOC polarization to the CS of NK – NKG2D-coated bead conjugates compared to WR infected cells (Figure 8B). However, overexpression of dominant negative Rap1A decreased MTOC polarization by 53% in NK – bead conjugates compared to WR infected cells (Figure 8B). Overall, these results demonstrate that Rap1 activation by NKG2D is PI3K and CrkL-dependent, and that Rap1 regulates NKG2D-mediated adhesion and MTOC polarization leading to NK cell cytotoxicity.
In this paper we describe an NKG2D-mediated signaling pathway for the regulation of conjugate formation, MTOC polarization and granule secretion. Integrins have been shown to play a critical role in lymphocyte function regulated by ITAM-containing receptors such as the TCR, but little is know about the regulation of integrin by non-ITAM containing receptor such as NKG2D. We establish that ligation of FcR and NKG2D increase adhesion to β1 and β2 integrin ligands, VCAM-1 and ICAM-1. Using blocking antibodies we have established a role for the β2 integrin in NKG2D-mediated NK – tumor target formation and cytotoxicity. This is the consistent with the observation that patients with leukocyte adhesion deficiencies display impaired NK cell function (10–12)
The results presented herein indicate that NKG2D signaling regulates NK cell adhesion and that LFA1 is required for optimal killing through NKG2D. Surprisingly, previous studies aimed at dissecting NKG2D-mediated signaling in both human and mouse NK cells suggest that neither the PI3K nor Grb2/Vav1 pathways regulate NKG2D-mediated cellular adhesion (8, 18). Taken at face value, these studies would seem to indicate that neither of the two known signaling pathways engaged by NKG2D are important in stimulating adhesion. However, the amount of conjugate formation between NK cells and NKG2D-sensitive targets in both studies is only slightly, if at all, increased over control targets (8, 18), suggesting the possibility that NKG2D-engagement does not trigger integrin activation, similar to known ITAM-containing receptors. However, our data provide evidence that NKG2D ligation does in fact increase integrin-mediated adhesion to plate-bound substrates, as well as conjugate formation to an NKG2D-ligand-bearing target cells. Moreover, our data identify PI3K signaling as an important mediator of NKG2D-stimulated adhesion. Thus, our work adds to the understanding of the NKG2D-mediated signaling pathways by establishing that NKG2D-mediated conjugate formation is PI3K-dependent. Future studies will be important to determine if the Grb2/Vav1 pathway also participates in NKG2D-stimulated adhesion.
The Crk family of adaptor proteins, Crk and CrkL, are important regulators of cell motility and adhesion in multiple cell types, and in the regulation of the oncogenic BCR-Abl mutation (19). The lymphocyte Crk family member, CrkL, has been clearly established as an important adapter protein in T-lymphocyte biology (16) and it has been suggested that CrkL plays a role in NK cell activation, but it has not been established if CrkL plays a role in the activation of NK cells by non-ITAM containing receptors such as NKG2D (46, 47). In this study we establish that CrkL regulates NKG2D-mediated NK cell-mediated cytotoxicity, in part through its effects on cell adhesion and establishing cell polarity. These results are significant since a recent study has shown that engagement of an inhibitory KIR on NK cells results in the inactivation of Crk through Abl-mediated phosphorylation at a known inhibitory site (20). Interestingly, the KIR-generated inhibitory signal could be overcome by expression of a Crk protein lacking the inhibitory phosphorylation site. Our data implicate CrkL in both ITAM-dependent and ITAM-independent signaling pathways leading to granule release and cellular cytotoxicity. Whether CrkL is phosphorylated by KIR engagement and serves a functionally redundant role with Crk in the regulation of NK cell cytotoxicity remains to be determined.
PI3K has been shown to regulate cell polarity and adhesion in multiple cell types (48, 49). Additionally, the regulatory subunit of PI3K, p85, has also been shown to interact with CrkL in transformed cell lines (38). These observations are consistent with our study using normal human NK cells, and we add the observation that PI3K interacts with the N-terminal SH3 domain of CrkL in primary human NK clones and this interaction is important for NKG2D-stimulated killing. Moreover, our data indicate that CrkL is involved in establishing cell polarity downstream of NKG2D, affecting both MTOC polarization and granule release. Bryceson et. al. has indicated that in resting NK cells, granule polarization is integrin-regulated whereas degranulation is controlled by FcR ligation (50). In our study, beads that are coated with only anti-NKG2D stimulate recruitment of the MTOC to the NK – bead interface in a CrkL-dependent manner, suggesting that in IL-2 expanded human NK clones, integrins are not required for MTOC polarization.
How an activation signal passes from NKG2D-PI3K-CrkL to Rap1 remains to be determined. Indeed, PI3K has been shown to regulate Rap1-dependent adhesion (48). Rap1 is an integral protein involved in the activation of integrins in T cells, where it has also been shown to regulate cell motility and polarity (51, 52). Recently, Rap1 activation following TCR ligation has been shown to occur via a complex of proteins including CrkL and C3G (16, 53). Interestingly, although CrkL is required for NKG2D-mediated conjugate formation and cytotoxicity, C3G does not appear to participate in NKG2D-mediated cellular cytotoxicity (Supplemental Figure 4). Still, our data suggest that Rap1 is activated in a PI3K- and CrkL-dependent manner following NKG2D ligation and mediates NK cell cytotoxicity through regulation of both NKG2D-mediated adhesion and MTOC polarization in human NK cells. Given that C3G suppression does not effect NKG2D-mediated cytotoxicity, it is likely that the NKG2D-PI3K-CrkL signaling cascade signals through another Rap1 activator such as CalDAG-GEF, PDZ-GEF, or DOCK4 (53). Additionally, the Rap1 effector pathways required for NKG2D-initiated cell killing remain to be determined.
Taken together, we have identified that PI3K activity regulates cell adhesion stimulated through engagement of NKG2D. Moreover, we have shown that the adaptor protein CrkL and the small GTPase Rap1 play important roles downstream of PI3K controlling not only integrin activation, but also establishing NKG2D-dependent cell polarity leading to granule exocytosis and optimal cellular cytotoxicity. A further understanding of the NKG2D-PI3K-CrkL-Rap1 signaling pathway, as well as other signaling pathways engaged by this activating receptor, will likely provide important insight into the causes of leukocyte adhesion deficiencies as well as other disorders of NK cell function.
1Supported by the NIH (CA47752 and AI65474 to D.D.B.). D.D.B. is a Leukemia and Lymphoma Society Scholar.
3Abbreviations used in this paper: CS, cytotoxic synapse; IP, immunoprecipitation; ITAM, immunoreceptor tyrosine-based activation motif; ITIM, immunoreceptor tyrosinebased inhibition motif; MTOC, microtubule organizing center; MICA, MICA expressing BaF3 cells; siCrkL, siRNA directed against CrkL; siRNA, small interfering RNA; siNeg, negative control siRNA