In lymphocytes, stimulation of cell surface activating receptors induces the formation of protein microclusters at the plasma membrane that contain the receptor itself, along with other signaling molecules. Although these microclusters are generally thought to be crucial for promoting downstream cellular responses, evidence that specifically links clustering potential to signaling output is lacking. We found that protein kinase C-θ (PKCθ), a key signaling molecule in multiple lymphocyte subsets, formed microclusters in activated NK cells. These microclusters coalesced within the immunological synapse between the NK cell and its target cell. Clustering was mediated by the regulatory region of PKCθ and specifically required a putative phosphotyrosine-binding site within its N-terminal C2 domain. Whereas expression of wild-type PKCθ rescued the cytokine production defect displayed by PKCθ-deficient NK cells, expression of a PKCθ point-mutant incapable of forming microclusters had little to no effect. Hence, PKCθ clustering was necessary for optimal effector function. Notably, only receptors containing ITAMs induced PKCθ microclusters on their own, explaining previous observations that ITAM-coupled receptors promote stronger activating signals and effector responses than do receptors lacking these motifs. Taken together, our results provide a cell biological basis for the role of PKCθ clustering during NK cell activation, and highlight the importance of subcellular compartmentalization for lymphocyte signal transduction.
Natural killer lymphocytes play an important role in immune responses against viruses and tumors by directly killing infected or transformed target cells. NK cells express multiple germline-encoded activating receptors that recognize surface proteins characteristic of stress or infection. Ligand engagement by these receptors leads, within minutes, to the formation of a tight cell–cell interaction with the target cell, known as an immunological synapse (IS) (1–3). This is followed by the selective killing of the target cell and also by the secretion of inflammatory cytokines such as IFN-γ, which coordinate other branches of the immune system.
Activating NK receptors signal through a number of distinct cytoplasmic tyrosine-based motifs that are located within either the receptor itself or an adaptor chain that is associated with the receptor (4). Ligand binding induces the phosphorylation of these sequences by Src-family kinases, leading to the recruitment and activation of other signaling proteins. Many activating NK receptors contain ITAMs [consensus sequence Y-x-x-(L/I)-x6-8-Y-x-x-(L/I)] that recruit Syk-family kinases and promote the assembly of the LAT-Slp76 adaptor complex, which in turn activates effector enzymes like phospholipase C-γ. Other receptors, however, contain immunotyrosine-based switch motifs [consensus sequence T-x-Y-x-x-(V/I)], which signal via the SAP family of SH2 domain-containing adaptors, or YINM motifs, which directly recruit PI3K and the guanine nucleotide exchange factor Vav.
Although much is now known about the membrane proximal signaling events associated with each class of activating receptor, how these events are coupled to downstream effector responses, such as degranulation and cytokine production, is less clear. In that regard, it is intriguing that NK cells lacking the serine/threonine-directed protein kinase C-θ (PKCθ) display profound cytokine secretion defects, despite normal cytotoxicity (5). PKCθ is highly expressed in both T cells and NK cells, and is crucial for cellular responses ranging from cytokine production and differentiation to the induction of cytoskeletal polarity (6–8). Precisely how PKCθ couples receptor stimulation to these various outcomes is not fully understood, but it is likely that subcellular compartmentalization plays a crucial role. PKCθ localization is dictated by its N-terminal regulatory region, which, like other members of the PKC family, contains a C2 domain, together with tandem C1 domains (9). The C1 domains recognize 1,2-diacylglycerol (DAG) and phorbol esters, and mediate translocation of PKCθ to the plasma membrane in response to DAG production by phospholipase C-γ. Conventional C2 domains recognize negatively charged phospholipids in a calcium (Ca2+)-dependent manner. The C2 domain of PKCθ, however, lacks amino acid residues required for Ca2+ binding. Its precise function has remained unclear, although studies of related C2 domains have suggested that it might mediate protein–protein interactions instead of membrane binding (10–12).
In T cells, stimulation of the TCR induces DAG-dependent accumulation of PKCθ at the IS (8, 13). Recent imaging studies have revealed that the costimulatory receptor CD28 recruits this synaptically localized PKCθ into plasma membrane “microclusters” (14, 15). A number of distinct lymphocyte activating receptors are known to form microclusters in response to ligand binding (16–19). It is generally thought that this clustering facilitates signal transduction by providing activating receptors and receptor proximal enzymes privileged access to their substrates and binding partners while at the same time excluding negative regulators. It has been difficult, however, to assess the specific effects of microcluster formation on signaling output.
The subcellular distribution of PKCθ in the NK cell IS has not been characterized. In this study, we demonstrated that PKCθ forms transient microclusters at the NK cell IS in response to activating receptor stimulation. These microclusters, which were specific for PKCθ and not other PKC isozymes, were enriched for phosphotyrosine (pTyr) and phosphorylated signaling molecules, indicative of a role in activating signaling. Intriguingly, we found that only ITAM-coupled receptors, but not non–ITAM-coupled receptors, could drive PKCθ clustering on their own. Because these data correlated with previous studies indicating that ITAM-coupled receptors induce stronger downstream responses than their non–ITAM-coupled counterparts (20–23), we investigated whether PKCθ clustering might be important for promoting effector function. Indeed, using a point mutation in PKCθ that disrupted cluster formation without altering other PKCθ functionality, we demonstrated that PKCθ clustering is required for optimal effector responses. These results provide new insight into the importance of subcellular compartmentalization for effective signal transduction in lymphocytes.
Materials and Methods
DNA constructs and recombinant proteins
cDNAs encoding full-length KIR2DS2, mouse Bcl10, and human DAP12 were subcloned into a murine stem cell virus retroviral plasmid upstream of either GFP or mCherry. Retroviral expression vectors for full-length mouse PKC constructs fused to GFP or Tag-RFP-T have been described (8). Constructs encoding the following fragments were obtained by PCR of the full-length PKCθ gene: C2 (amino acids 1–151), tandem C1 (160–281), C2–C1 (1–282). Chimeric PKC proteins were constructed using a three-step PCR approach in which different domains from PKCε and PKCθ were amplified from either PKCε or PKCθ cDNA and stitched together in two consecutive PCR steps. The chimeric PCR products were then cloned into the murine stem cell virus-GFP vector. In all cases, PKC proteins or fragments were positioned upstream (i.e., N-terminal) of the fluorescent protein. Point mutations of KIR2DS2 and PKCθ were inserted using the QuikChange protocol (Stratagene). Retroviral constructs expressing full-length PKCθ or PKCθ-H63D linked by an internal ribosome entry site to the extracellular domain of human CD2 were prepared by subcloning the coding sequence for PKCθ into the pMigR2 plasmid. Biotinylated HLA-Cw3, ULBP3, and ICAM-1 were prepared as described (16). HLA-Cw3 used for these studies contained a mutation that blocks binding of the inhibitory receptor ILT2 (16). The extracellular region of CD48, fused to both a histidine tag and a BirA recognition sequence, was expressed in Hi-5 cells by baculoviral transduction and purified by Ni2+ affinity. The protein was then biotinylated using the BirA enzyme followed by size exclusion chromatography.
Primary NK cells and NKL cells
Primary mouse NK cells were purified from splenocytes by negative selection using Abs against erythroid cells (TER-119), CD4 (GK1.5), CD8α (YTS169.4), CD5 (53-7.3.4), CD19 (1D3), and GR-1 (RB6-8C5; all Abs from the University of California, San Francisco, Monoclonal Antibody Core). Enriched NK cells were cultured in RPMI 1640 supplemented with 10% FCS and 1000 IU/ml of human IL-2. NKL cells were maintained in RPMI 1640 with 10% FCS and 200 IU/ml IL-2. Retroviral production and transduction of NKL cells was performed as described (16). Expression of transduced proteins was quantified 48 h after transduction by flow cytometry (BD LSR II), using either the transduced fluorescent protein label or an Ab against KIR2 (DX27; BD Biosciences) for detection. NKL cells expressing the transduced protein (typically representing 2–10% of the total population) were isolated by FACS 1–2 wk after transduction and maintained as stable cell lines.
Plastic surfaces for stimulation of primary murine NK cells were prepared using 48-well plates coated first with DOTAP (Sigma-Aldrich) solution and then with Abs (5 μg/ml) against mouse 2B4 (m2B4; BioLegend), Ly49H (3D10; eBioscience), NK1.1 (PK136; eBioscience), or NKG2D (A10; eBioscience). After coating, 2–3 × 105 NK cells were stimulated in each well for 8 h at 37°C. Cells were then fixed and stained for intracellular IFN-γ (XMG1.2; eBioscience), CD49b (DX-5; BioLegend), and CD3 (17A2; eBioscience). Flow cytometric analysis was performed on the CD49b+CD3− population. To assess degranulation, fluorescently conjugated Abs against CD107a (1D4B; BD Biosciences) were added at the start of the experiment, and the staining was evaluated after fixation and labeling with the other Abs. In certain experiments, cells were stimulated in the presence of 10 μM PP2 (Sigma-Aldrich) or 0.5 μM BAY61-3606 (Calbiochem) to inhibit Src- or Syk-family kinases, respectively. NKL cells were stimulated in plastic wells coated with streptavidin followed by biotinylated ICAM-1 (1 μg/ml), HLA-Cw3 (1 μg/ml), CD48 (1 μg/ml), ULBP3 (1 μg/ml), or Abs against NKG2D (1 μg/ml). For comparative studies in which one or more constituents were left out, a nonstimulatory biotinylated mouse MHC molecule (either I-Ek or H2-Db) was added to keep the total protein concentration for that experiment constant. To assess degranulation, fluorescently conjugated Abs against CD107a were included at the start of the experiment. After washing, CD107a staining was quantified by flow cytometry. All flow cytometric data were analyzed using FlowJo software.
IL-2–cultured NK cells were incubated with 51Cr-labeled target cells for 4 h at 37°C, after which supernatants were analyzed for 51Cr release. Maximum (MAX) lysis was determined by repeated freeze–thaw lysis of labeled target cells. Spontaneous (MIN) lysis was determined from wells containing target cells in the absence of effectors. Individual samples were run in triplicate. Percent specific lysis = (Lysissample − LysisMIN)/(LysisMAX − LysisMIN) × 100%.
Supported lipid bilayers
Bilayers were generated as described from small unilamellar vesicles containing a 10:1 mixture of 1,2-dioleoyl-sn-glycero-3-phosphocholine and biotinyl cap phosphoethanolamine (Avanti Polar Lipids) (16). After formation, bilayers were incubated with streptavidin, washed with PBS, and then incubated with biotinylated NK receptor ligands or biotinylated Abs. ULBP3, HLA-Cw3, and CD48 were used at 2–6 μg/ml, and ICAM was used at 2 μg/ml. Biotinylated Abs against NK1.1 (PK136; eBioscience), Ly49H (3D10; eBioscience), NKG2D (A10; eBioscience), and 2B4 (m2B4 BioLegend) were used at 10 μg/ml. Stimulatory bilayers for 5C.C7 T cells were prepared using biotinylated I-Ek containing the stimulatory moth cytochrome c peptide, B7-1, and ICAM-1 (all proteins at 0.5 μg/ml). For comparative studies in which one or more constituents were left out of the bilayer, a nonstimulatory biotinylated mouse MHC molecule (either I-Ek containing the null Hb peptide or H2-Db) was added to keep the total protein concentration for that experiment constant. After protein loading, bilayers were stored at room temperature for up to 4 h prior to use.
NK cells were incubated on stimulatory bilayers for 15 min at 37°C and then fixed with 2% paraformaldehyde in PBS. After extensive washing in PBS, cells were permeabilized with Triton X-100, blocked with BSA, and then incubated with primary Abs against pSrc (Cell Signaling), pZap70 (Cell Signaling), pTyr (4G10; Millipore), PKCθ (C-18; Santa Cruz Biotechnology), or GFP (Invitrogen), followed by staining with Alexa 594-labeled secondary Abs (Jackson Immunoresearch) and phalloidin (labeled with Alexa Fluor 488 or 594; Invitrogen) for 1 h. After a final PBS wash, cells were imaged by total internal reflection fluorescence (TIRF) microscopy. In certain experiments, stimulation was performed in the presence of 10 μM PP2 or 0.5 μM BAY61-3606 to inhibit Src- or Syk-family kinases, respectively.
Phoenix E cells were transfected with plasmids containing GFP-labeled PKCθ-H63D, PKCθ-Y90F, PKCθ, or GFP only. Cells were collected after 2 d and lysed in ice-cold 10 mM Tris/Cl, pH 7.5; 150 mM NaCl; 0.5 mM EDTA; 0.5% Nonidet P-40, with 1× protease inhibitor mixture (Roche) freshly added. Lysates were incubated with GFP-Trap beads (ChromoTek) for 2 h at 4°C, washed 4 times with ice-cold wash buffer (10 mM Tris/Cl, pH 7.5; 150 mM NaCl; 0.5 mM EDTA, with 1× protease inhibitor mixture freshly added), and resuspended in wash buffer. To 39.5 μl lysate was added 2 μg purified Marcksl1 (MLP, a PKC substrate) or mutant Marcksl1 lacking the Ser93 and Ser104 consensus PKC phosphorylation sites (MLP*), together with 5 μl 10× kinase buffer (final concentration, 25 mM Tris/Cl, pH 7.5; 10 mM MgCl2; 1 mM NaF). The reaction was started by adding 0.5 μl ATP (final concentration, 1 mM) and incubated at 37°C for 30 min. Then 6× SDS loading buffer was added to stop the reaction, and samples were analyzed by immunoblot using α-phosphoserine PKC substrate Ab (Cell Signaling). Total PKCθ levels were assessed using α-GFP Ab (Invitrogen).
All imaging experiments used an inverted fluorescence video microscope (Olympus IX-81) attached to an electron-multiplying charge-coupled device camera (Hamamatsu). Lasers of 488 nm and 561 nm (Melles Griot) were used for TIRF imaging of GFP and mCherry/RFP/Alexa 594, respectively, and a DG-4 Xe lamp (Sutter) was used for epifluorescence imaging. A 60×, 1.45 numerical aperture objective lens was used for TIRF imaging, and a 20× epifluorescence objective (0.75 numerical aperture, all objectives from Olympus) was used for Ca2+ imaging. For live imaging experiments, cells were transferred into RPMI 1640 supplemented with 5% FCS and lacking phenol red, and then imaged at 37°C. Prior to Ca2+ imaging, cells were loaded with 5 μg/ml Fura-2AM. Fixed samples were imaged in PBS. Time-lapse recordings were made using Slidebook software (Intelligent Imaging Innovations). In general, images were acquired every 10 s, 30 s, or 60 s for 20–30 min after addition of cells to the stimulatory bilayer. Experiments involving photoactivation of the 5C.C7 TCR were performed as previously described, using a Mosaic digital diaphragm apparatus (Photonic Instruments) attached to a mercury lamp (Olympus) (24). We used the following criteria to verify TIRF illumination: 1) detecting a decrease in the intensity of excitation light emerging from the objective lens, 2) finding enhanced spatial resolution where the cell made contact with the supported lipid bilayer, and 3) confirming that it was impossible to focus on features above the plane of the bilayer.
PKCθ clustering was quantified by calculating the fraction of synapse area containing PKCθ fluorescence greater than an established “bright pixel” threshold. For images of primary mouse NK cells stained with anti-PKCθ Abs, the bright pixel threshold was set at 2× the average, background-corrected PKCθ fluorescence intensity for that synapse. For NKL cells expressing GFP-labeled PKCθ, the bright pixel threshold was set at 3× the average, background-corrected GFP fluorescence intensity for that synapse. Quantification of enrichment within PKCθ microclusters (Fig. 4D) was performed by calculating the ratio of the average, background-corrected pTyr, pZap70, or Bcl10 fluorescence intensity within PKCθ microclusters to the average, background-corrected pTyr, pZap70, or Bcl10 fluorescence intensity over the entire IS. The pSrc staining in Fig. 1D was quantified by determining the average, background-corrected synaptic pSrc fluorescence for each condition. Analysis of single-cell Ca2+ flux in Fura-2AM–loaded cells was performed by normalizing the Fura ratio of each cell, using the last image prior to the initial rise in Ca2+. All data sets used for quantification comprised images recorded on the same day. Analysis was performed using Slidebook, Microsoft Excel, and GraphPad Prism.
PKCθ−/− donor mice were injected i.p. with 0.3 ml 10 mg/ml 5-fluorouracil (Sigma-Aldrich) in PBS. After 72 h, bone marrow was isolated from femurs and tibias of donor mice and cultured at 1 × 106 cells/ml in stem cell media [DMEM + 15% FBS supplemented with 100 ng/ml recombinant murine stem cell factor (R&D), 20 ng/ml recombinant human IL-6 (R&D), and 10 ng/ml recombinant murine IL-3 (R&D)]. After 48 h culturing, bone marrow cells were retrovirally transduced with constructs coding for the human CD2 extracellular domain and either wild-type PKCθ or PKCθ−H63D. Retroviruses were mixed with bone marrow (105 cells/ml) in stem cell media and centrifuged at 1400 g for 2 h at 37°C in the presence of 4 μg/ml polybrene and cultured 24 h at 37°C. After 24 h, the transduction procedure was repeated. Bone marrow cells were harvested and 1–2 × 105 cells were resuspended in 200 μl PBS and injected retro-orbitally into CD45.1+ recipients that had been exposed to sublethal irradiation (800 rads). After an 8-wk reconstitution period, NK cells were isolated, cultured in IL-2 for 7–10 d, and then stimulated with Abs against NK1.1, as described above. Prior to flow cytometric analysis, cells were stained with Abs against CD45.1, CD45.2, CD49b, CD3, IFN-γ, and CD107a.
PKCθ forms membrane microclusters at the NK cell IS
To explore the cellular basis for PKCθ function in NK cells, we analyzed the subcellular localization of PKCθ at the NK cell IS. IL-2–cultured mouse NK cells were stimulated on supported lipid bilayers containing the adhesion molecule ICAM-1 and crosslinking Abs against the activating receptors NK1.1, Ly49H, and 2B4. After contact formation, cells were fixed and stained for PKCθ and F-actin, and then imaged using TIRF microscopy, which enables high-resolution analysis of the plasma membrane attached to the bilayer. Activated NK cells formed radially symmetric synapses under all conditions, often characterized by a ring of dense F-actin in the periphery surrounding a region of hypodense F-actin at the center, consistent with previous observations (17, 25). In cells stimulated through NK1.1 and Ly49H, both of which are ITAM-coupled receptors, PKCθ formed distinct microclusters in the plasma membrane that localized inside the peripheral actin ring (Fig. 1A). In contrast, PKCθ clusters were not observed in cells stimulated through 2B4, which does not contain ITAMs (Fig. 1A). The extent of PKCθ clustering was quantified by calculating the fraction of synapse area taken up by high-intensity PKCθ staining. This analysis demonstrated that the apparent differences in clustering observed between 2B4-activated cells and cells stimulated through ITAM-coupled receptors were highly significant (Fig. 1B). Importantly, all receptor crosslinking Abs used for these experiments induced robust Ca2+ flux in addition to Src kinase phosphorylation at the NK cell IS (Fig. 1C, 1D). These early signaling responses were much stronger than those observed in cells plated on ICAM-1 alone, indicating that all of the crosslinking Abs were effective at stimulating their cognate NK receptors.
Consistent with previous reports (20–23), we found that stimulation of NK1.1 and Ly49H induced stronger degranulation and cytokine production than did stimulation of 2B4 in both resting and IL-2–cultured NK cells (Fig. 2). Hence, the ability of each activating receptor to induce PKCθ microclusters correlated with both the presence of ITAMs and the potential to generate robust effector responses.
Clustering is specific for PKCθ but does not require ITAM signaling
To characterize the dynamics and molecular composition of PKCθ microclusters, we developed a system for routine live imaging of cluster formation in the human NK cell line NKL. As model non–ITAM-coupled receptors, we used 2B4 and NKG2D, both of which are expressed in NKL cells. The receptor 2B4 recognizes the Ig domain protein CD48, whereas NKG2D binds to a set of proteins (the MIC and ULBP families in humans) that are upregulated in response to infection and cellular transformation (26). As a model ITAM-coupled receptor, we used KIR2DS2, which associates with the ITAM-containing signaling adaptor DAP12 (26). The affinity of KIR2DS2 for its ligand, the class I MHC protein HLA-Cw3, is too low to mediate effective signaling. Therefore, we introduced a point mutation (Y45F) known to enhance ligand binding (Supplemental Fig. 1A, 1B) (27). NKL cells expressing KIR2DS2-Y45F (called simply KIR2DS2 hereafter) degranulated in response to immobilized HLA-Cw3, CD48, and ULBP3 (Supplemental Fig. 1B, 1C), validating KIR2DS2, 2B4, and NKG2D, respectively, as activating receptors in this system.
To examine the recruitment of PKCθ under various stimulus regimens, we prepared NKL cells expressing KIR2DS2, together with GFP-labeled PKCθ, and then imaged them using TIRF microscopy on bilayers presenting ICAM-1 and either HLA-Cw3, ULBP3, or CD48. Consistent with our observations in primary cells, PKCθ formed distinct membrane microclusters in NKL cells stimulated with HLA-Cw3, but not in cells stimulated with either CD48 or ULBP3 (Fig. 3A). These microclusters appeared within seconds of contact formation near the center of the synapse, and tended to fade over time, becoming imperceptible after 10-20 min (Supplemental Video 1). To assess whether microcluster formation was a general property of PKC proteins, we also examined the recruitment behavior of PKCδ, PKCε, PKCη, PKCα, and PKCβ. Only PKCθ formed KIR2DS2-induced microclusters (Fig. 3B), indicating that clustering is a specific property of this isoform.
Previous studies have shown that NKG2D and 2B4, when stimulated together, induce stronger activating signals and effector responses than when either receptor is triggered alone (21, 28). To assess whether synergy between NKG2D and 2B4 could promote PKCθ clustering, we imaged NKL cells expressing GFP-labeled PKCθ on bilayers containing CD48 and ULBP3. Remarkably, PKCθ formed microclusters under these conditions (Fig. 3A). These data imply that PKCθ clustering does not result from ITAM signaling per se, but rather that it is correlated with the overall strength of activating signals and their potential to induce robust effector responses.
PKCθ colocalizes with activating receptors and signaling molecules
In B cells, T cells, and NK cells, activating receptors form plasma membrane microclusters at the IS in response to ligand binding (16, 18, 19). To explore the relationship between these activating receptor microclusters and PKCθ, we labeled DAP12 and transduced it into NKL cells expressing KIR2DS2, together with GFP-labeled PKCθ. Using TIRF microscopy, we observed that stimulatory bilayers containing HLA-Cw3 specifically induced the formation of DAP12 microclusters in these cells. Microclusters formed initially during cell spreading and began to traffic centripetally toward the center of the synapse once the size of the contact had stabilized (Fig. 4A, Supplemental Video 2). This behavior is quite similar to what has been observed for Ag receptors in T cells and B cells (18, 19). Clusters of DAP12 eventually coalesced into a stable aggregate at the center of the synapse that persisted for the duration of our experiments (typically 20–40 min). Because this central accumulation behaved analogously to the central supramolecular activation cluster (cSMAC) that has been observed in T cell synapses (19), we refer to it as a cSMAC.
Analysis of cells expressing labeled DAP12 and PKCθ revealed extensive overlap between the two proteins. This overlap, however, was incomplete, and it changed as the NK cell IS matured (Fig. 4A, Supplemental Video 2). PKCθ was not recruited to activating receptor microclusters as they formed in the periphery (e.g., yellow arrowheads in Fig. 4A), but rather accumulated in these microclusters when they approached the center of the contact. Subsequently, as the clusters coalesced into a mature cSMAC, PKCθ accumulation faded. Hence, PKCθ association with activating microclusters is transient, beginning as the clusters approach the center of the synapse and diminishing as a stable cSMAC is formed.
It is generally thought that activating signals propagate from receptor microclusters at the periphery of IS and that these signals are downregulated as microclusters accumulate in the cSMAC (29–32). Consistent with this idea, we observed robust staining for pTyr in peripheral DAP12 microclusters, but proportionately less staining in the cSMAC that formed at later time points (Fig. 4B). It remained unclear, however, whether microclusters containing PKCθ, which are centrally located, still engage in activating signaling. To address this issue, we stimulated NKL cells expressing GFP-labeled PKCθ on bilayers containing HLA-Cw3 and then stained them with Abs against pTyr and the phosphorylated form of the Syk-family kinase Zap70 (pZap70). Unambiguous colocalization was observed between these markers and PKCθ (Fig. 4C). Indeed, both pTyr and pZap70 were enriched 4- to 5-fold in PKCθ clusters relative to the rest of the IS (Fig. 4D). This finding suggests that activating receptor microclusters are still signaling while PKCθ is associated. Taken together, our data indicate that PKCθ is transiently recruited to microclusters containing activating receptors, where it participates in activating signaling before dissociating as the microclusters are downregulated in the cSMAC.
In T cells, it has been shown that PKCθ induces activation of the transcription factor NF-κB by phosphorylating a scaffolding complex containing the adaptor molecules CARMA1, Bcl10, and Malt1 (CBM complex) (33). A recent imaging study suggested that PKCθ associates with the CBM complex in microclusters at the T cell IS, consistent with the importance of PKCθ for NF-κB signaling (15). To investigate whether analogous coclustering occurs in NK cells, we imaged GFP-labeled Bcl10, together with RFP-labeled PKCθ. Although Bcl10 did form microclusters upon stimulation of KIR2DS2, we observed little to no overlap between these clusters and clusters of PKCθ (Fig. 4D, 4E), suggesting that PKCθ does not associate with the CBM complex at the NK cell IS. This result is consistent with previous observations that NF-κB signaling is unaffected in NK cells lacking PKCθ (5).
PKCθ clustering requires Src-family, but not Syk-family, kinases
To address the intracellular signaling requirements for PKCθ microcluster formation, we monitored PKCθ recruitment dynamics in the presence of PP2 and BAY61-3606, which are small-molecule inhibitors of Src- and Syk-family kinases, respectively. Treatment with PP2 completely disrupted microcluster formation in NKL cells, whereas BAY61-3606 had no effect on the process (Fig. 5A). Both compounds completely blocked IFN-γ secretion from murine NK cells after stimulation of NK1.1, confirming their efficacy as inhibitors (Fig. 5B). Together, these results suggest that PKCθ clustering operates downstream of Src kinases but that it is independent of Syk kinase activity. This idea is consistent with our observation that simultaneous stimulation of NKG2D and 2B4, which are not known to signal though Syk kinases, nevertheless induced PKCθ microcluster formation (Fig. 3A).
PKCθ clustering requires the putative pTyr binding site in its C2 domain
Subcellular localization of PKCθ is mediated by its N-terminal regulatory region, which contains both tandem C1 domains and a C2 domain (9, 34). To explore the requirements for PKCθ clustering, we assessed the recruitment behavior of various GFP-labeled PKCθ fragments in NKL cells. We found that the C2 domain and the tandem C1 domains were insufficient on their own to form synaptic clusters (Fig. 6A). However, a fragment containing both the C2 and the tandem C1 domains completely reconstituted clustering behavior, indicating that the entire N-terminal regulatory region participates in the process.
To confirm and extend these results, we prepared a set of chimeric constructs in which portions of the PKCθ regulatory region were replaced by the corresponding part of PKCε, a related isoform that does not form microclusters at the NK cell IS (Fig. 3B). We also generated reciprocal chimeras in which domains of PKCθ were introduced into PKCε. The clustering behavior of each protein was then analyzed by TIRF in NKL cells (Fig. 6B). Only constructs containing the PKCθ C2 domain formed activation-induced microclusters, indicating that this specific domain is crucial for the response. Conversely, the tandem C1 domains of both PKCθ and PKCε were permissive for clustering, provided the PKCθ C2 domain was present.
Taken together with the results obtained from isolated PKCθ fragments, these data indicate that both the C2 and the tandem C1 domains contribute to clustering behavior, but that the C1 domains merely serve as a generic DAG binder, possibly to localize the protein to the DAG-rich synaptic membrane (see 21Discussion). By contrast, the C2 domain of PKCθ plays a much more specialized and isoform-specific role in this process. To further explore this role, we analyzed two different PKCθ mutants bearing single amino acid substitutions in the C2 domain. It has been shown that the Src kinase Lck phosphorylates PKCθ at Tyr 90 within the C2 domain, and that this residue is required for optimal transcriptional and proliferative responses to TCR stimulation (35). To assess the importance of Tyr 90 for PKCθ clustering in NK cells, we examined the recruitment behavior of a PKCθ mutant in which Tyr90 was changed to Phe (PKCθ-Y90F). In NKL cells, this mutant formed clusters that were indistinguishable from those of wild-type PKCθ (Supplemental Fig. 2A; compare with Fig. 4C), indicating that Lck-mediated phosphorylation of the C2 domain is not involved in synaptic microcluster formation.
It has also been shown that the C2 domain of PKCδ, the PKC isoform most closely related to PKCθ, can specifically recognize pTyr-containing peptides (10). pTyr recognition requires a histidine residue in the PKCδ C2 domain that is conserved in PKCθ (His63 in PKCθ). Accordingly, we mutated His63 to Asp and examined the recruitment behavior of the resulting protein at the NK cell IS. Experiments using NKL cells expressing either wild-type or mutant PKCθ labeled with GFP indicated that the H63D mutation completely disrupted microcluster formation (Fig. 6C). Both proteins exhibited robust accumulation at the NK cell–target cell interface (Supplemental Fig. 2B), indicating that the distinct clustering behavior we observed did not result from differences in IS recruitment. To guard against artifacts related to the comparison of independently derived stable cell lines, we expressed both mCherry-labeled wild-type PKCθ and GFP-labeled PKCθ-H63D together in the same NKL cells. Strikingly, stimulation of KIR2DS2 induced no observable clustering of PKCθ-H63D, despite the fact that wild-type PKCθ clustered robustly in the same synapses (Fig. 6D, Supplemental Video 3). Hence, the putative pTyr binding site within the C2 domain is required for PKCθ microcluster formation in NK cells.
The C2 domain pTyr binding site is not required for PKCθ recruitment in T cells
To further explore the role of the putative pTyr binding site within the C2 domain, we examined the effects of the H63D mutation on PKCθ dynamics during T cell activation. TCR signaling induces the DAG-dependent accumulation of PKCθ in regions of the plasma membrane containing activated receptors (8). The kinetics and strength of this accumulation can be quantified using a photochemical approach in which TCRs in a defined region of membrane are activated with a pulse of UV light (36). Using this assay, we compared the recruitment dynamics of wild-type PKCθ and PKCθ-H63D, and found no difference between the two proteins (Supplemental Fig. 2C). These results are consistent with our previous data suggesting that the DAG-binding C1 domains primarily mediate plasma membrane association of PKCθ in this context (8).
As mentioned above, stimulation of CD28 together with the TCR induces the formation of PKCθ microclusters that coalesce into a ring around the cSMAC (14). This localization pattern can be visualized readily by TIRF imaging of T cells on lipid bilayers containing agonist peptide-MHC along with the CD28 ligand B7-1. Under these conditions both wild-type PKCθ and PKCθ-H63D formed microclusters (Supplemental Fig. 2D), suggesting that determinants other than the putative pTyr binding site in the C2 domain are required for CD28-dependent clustering in T cells.
The absence of an observable recruitment defect in T cells implies that the H63D mutation does not substantially alter the structure of full-length PKCθ. Consistent with this idea, the kinase activity of PKCθ-H63D is comparable to that of both wild-type PKCθ and PKCθ-Y90F (Supplemental Fig. 2E). We conclude that the H63D mutation selectively disrupts PKCθ microcluster formation in NK cells without substantially altering other properties of the protein.
PKCθ clustering is required for robust NK cell effector responses
It has been shown previously that PKCθ is required for optimal NK cell cytokine secretion (5). Consistent with these results, we found that IFN-γ production in response to ITAM receptor stimulation was suboptimal in PKCθ-deficient NK cells (Supplemental Fig. 3A). We also observed decreased levels of degranulation in the absence of PKCθ, implying that a defect in cytotoxicity might be present (Supplemental Fig. 3A). To further explore this possibility, we performed in vitro killing assays using three different types of target cell. We observed no differences between wild-type and PKCθ-deficient NK cells in these assays (Supplemental Fig. 3B), in agreement with previous work (5). This finding indicates that the degranulation defect resulting from loss of PKCθ does not lead to reduced cytotoxicity.
The ability of the H63D mutation to disrupt the synaptic clustering of PKCθ in NK cells without altering its other recruitment properties enabled us to assess whether microcluster formation is required for PKCθ to promote effector responses. If clustering is required, one would predict that expression of wild-type PKCθ in PKCθ-deficient NK cells would restore normal cytokine production and degranulation, and that expression of PKCθ-H63D would not. Accordingly, we transduced hematopoietic stem cells (HSCs) from PKCθ knockout mice, with retroviruses expressing either wild-type PKCθ or PKCθ-H63D together with human CD2, which served to identify transduced cells (Supplemental Fig. 4). The HSCs were then transferred into sublethally irradiated hosts expressing the congenic marker CD45.1. After 8 wk, the splenic NK cell compartment in these chimeric mice generally contained three distinct populations of cells: 1) a CD45.1+CD45.2− population derived from the host (Gate 1 in Fig. 7A, called “host” cells hereafter); 2) a CD45.1−CD45.2+human CD2− population derived from the PKCθ−/− HSCs (Gate 2 in Fig. 7A, called “untransduced” cells); and 3) a CD45.1−CD45.2+human CD2+ population derived from PKCθ−/− HSCs that expressed exogenous PKCθ (Gate 3 in Fig. 7A, called “transduced” cells). Importantly, the host population expressed endogenous PKCθ and thereby served as an internal positive control for fully functional NK cells.
After reconstitution and isolation, total splenic NK cells from chimeric mice were stimulated using plate-bound Abs against NK1.1 and then analyzed for IFN-γ production and degranulation (Fig. 7B, 7C). To quantify PKCθ-dependent rescue over multiple independent experiments, we determined a “rescue ratio” for both the untransduced NK cells and the transduced NK cells in each mouse (Fig. 7D). This ratio was calculated by dividing the percentage of responding cells in the transferred population by the percentage of responding cells in the host population. If transferred cells responded as well as host cells, the rescue ratio would be ∼1, and if transferred cells responded poorly relative to host cells, the ratio would be <1. As expected, untransduced cells mounted substantially weaker effector responses than did host cells. This finding presumably reflected the fact that untransduced cells were PKCθ−, whereas host cells were PKCθ+. By contrast, IFN-γ production and degranulation in transduced cells expressing wild-type PKCθ were comparable to that of host cells, consistent with the importance of PKCθ for effector function. Remarkably, little to no rescue was observed in transduced cells expressing PKCθ-H63D. These results strongly suggest that PKCθ microcluster formation is required for optimal effector responses in NK cells.
Although it is generally thought that microcluster formation at the IS plays a crucial role in lymphocyte activation, little evidence to date directly links the clustering potential of a specific signaling molecule with the strength of downstream responses. In this article, we showed that strong activating stimulation, through either ITAM-coupled receptors or a synergistic combination of non–ITAM-coupled receptors, induces the recruitment of PKCθ into activating receptor microclusters at the NK cell IS. Using a mutation that selectively disrupts clustering behavior, we also showed that PKCθ microcluster formation is required for optimal effector responses. Indeed, the similarity in phenotype displayed by PKCθ-deficient NK cells and NK cells expressing PKCθ-H63D implies that clustering is necessary for a substantial fraction of PKCθ-dependent IFN-γ production and degranulation in this cell type. These results lay the groundwork for future analysis of the functional relevance of signaling microclusters during NK cell activation.
The PKCθ microclusters we have characterized bear a superficial similarity to the clusters of PKCθ that form in response to CD28 stimulation in T cells (14). Close examination of the requirements and dynamic properties of both types of cluster, however, suggests that they are fundamentally distinct structures. PKCθ microclusters in NK cells fade as the cSMAC matures, whereas the T cell clusters appear to persist for the lifetime of the IS. In T cells, PKCθ clustering requires the YINM receptor CD28, whereas in NK cells the ITAM-coupled receptors more effectively promote microcluster formation. Indeed, YINM signaling from NKG2D is insufficient, on its own, to induce PKCθ microclusters in NK cells and must be supplemented by coactivation through 2B4. Finally, a recent study indicated that the V3 linker between the tandem C1 domains and the kinase domain is necessary and sufficient for recruitment to the cSMAC in T cells (37). In contrast, we have shown that clustering in NK cells requires the putative pTyr binding site in the C2 domain.
The idea that PKCθ microclusters in NK cells are not directly analogous to CD28-dependent clusters in T cells is consistent with results from our laboratory and others suggesting that PKCθ induces distinct downstream signaling responses in NK cells relative to T cells. Phosphorylation of the CBM complex by PKCθ in T cells is crucial for NF-κB activation, and components of the CBM complex appear to colocalize with CD28 and PKCθ at the T cell IS (15, 33). In contrast, NK cells lacking PKCθ do not display any defects in NF-κB signaling (5), and we did not observe coclustering between PKCθ and Bcl10 at the NK cell IS. Given the apparent functional differences between PKCθ clusters in T cells and NK cells, it will be interesting in future studies to explore the biological consequences of disrupting each type of clustering behavior. Access to the H63D and V3 mutations should facilitate this line of investigation.
The C2 domains of PKCδ and PKCθ are highly conserved. Indeed, all of the amino acid residues directly involved in the recognition of pTyr-containing peptides by PKCδ are identical in PKCθ (10). Hence, it is surprising that PKCδ does not form detectable microclusters in our hands. A likely explanation for this discrepancy is that compartmentalization within the NK cell synapse is a combinatorial process requiring multiple regulatory domains. We have found that both the tandem C1 domains and the C2 domain are required for PKCθ microcluster formation. However, the role of the C1 region is likely to be quite generic, as the corresponding portion of PKCε is permissive for clustering. In T cells, the C1 domains of PKCθ and PKCε mediate DAG-dependent recruitment of the protein to the IS (8, 38). In contrast, PKCδ does not accumulate synaptically, but instead associates with lysosomal compartments (39). Hence, it is tempting to speculate that PKCθ localization at the NK cell IS cells is a two-step process involving initial, C1-dependent recruitment to the synaptic membrane, followed by C2-dependent recruitment into activating receptor microclusters.
The reduced degranulation we observed in NK cells lacking PKCθ or PKCθ clustering is seemingly incongruous with the observation that target cell killing is unaffected by loss of PKCθ (5). However, it is likely that suboptimal degranulation responses are nevertheless sufficient for cytotoxicity in many contexts, particularly those in which FasL-Fas–mediated killing is still intact. Overall, our data are consistent with the previously articulated hypothesis that different signaling thresholds exist for cytotoxicity and cytokine secretion (5, 23, 40). A lower threshold for the former relative to the latter would presumably be useful for enabling target cell killing in cases in which concomitant inflammatory responses might be unnecessary or even harmful.
Taken together, our results suggest that PKCθ clustering is an early and necessary step toward optimal NK cell activation. It is remarkable that the subcompartmentalization of a protein that is already localized to the NK cell IS can have such strong implications for downstream signaling. Within microclusters, PKCθ presumably has access to substrates and other interacting proteins that would not otherwise be available. The identification and characterization of these signaling partners will be crucial for determining the mechanism of PKCθ-dependent signal amplification and remains an area for future investigation.
We thank N. Bantilan for technical assistance, S.S. Yi and the Memorial Sloan-Kettering Cancer Center Microchemistry Core Facility for peptide synthesis, D. Littman for PKCθ knockout mice, S. Rudensky and laboratory members for reagents and advice, A. Hall for critical reading of the manuscript, and members of the Huse and M.O. Li laboratories for helpful discussions.
This work was supported by the Spanish Ministry of Science and Innovation (to E.M.); the Lucille Castori Center for Microbes, Inflammation, and Cancer (to M.A.F.); the Metastasis Research Center, the Searle Scholars Program, and the Cancer Research Institute (to J.C.S. and M.H.).
The online version of this article contains supplemental material.
Abbreviations used in this article:
- CBM complex
CARMA1, Bcl10, and Malt1
central supramolecular activation cluster
hematopoietic stem cell
protein kinase C-θ
total internal reflection fluorescence.
The authors have no financial conflicts of interest.