Research ArticleImmunology

Diacylglycerol kinase α establishes T cell polarity by shaping diacylglycerol accumulation at the immunological synapse

See allHide authors and affiliations

Science Signaling  26 Aug 2014:
Vol. 7, Issue 340, pp. ra82
DOI: 10.1126/scisignal.2005287

Abstract

Polarization of the T cell microtubule–organizing center (MTOC) to the immunological synapse between the T cell and an antigen-presenting cell (APC) maintains the specificity of T cell effector responses by enabling directional secretion toward the APC. The reorientation of the MTOC is guided by a sharp gradient of the second messenger diacylglycerol (DAG), which is centered at the immunological synapse. We used a single-cell photoactivation approach to demonstrate that diacylglycerol kinase α (DGK-α), which catalyzes the conversion of DAG to phosphatidic acid, determined T cell polarity by limiting the diffusion of DAG. DGK-α–deficient T cells exhibited enlarged accumulations of DAG at the immunological synapse, as well as impaired reorientation of the MTOC. In contrast, T cells lacking the related isoform DGK-ζ did not display polarization defects. We also found that DGK-α localized preferentially to the periphery of the immunological synapse, suggesting that it constrained the area over which DAG accumulated. Phosphoinositide 3-kinase activity was required for the peripheral localization pattern of DGK-α, which suggests a link between DAG and phosphatidylinositol signaling during T cell activation. These results reveal a previously unappreciated function of DGK-α and provide insight into the mechanisms that determine lymphocyte polarity.

INTRODUCTION

Cell polarity plays a central role in migration, asymmetric division, and intercellular communication. As such, it is essential for both the development and the homeostasis of complex tissues. In many cell types, polarized cellular architecture is directed by the movement of the centrosome [also called the microtubule-organizing center (MTOC)] to one side of the cell. This event realigns the microtubule cytoskeleton, positions key organelles, and is required for the elaboration of axons, primary cilia, and other specialized signaling structures (1). In lymphocytes, such as T cells, B cells, and natural killer cells, the MTOC reorients toward the immunological synapse that forms between the lymphocyte and its stimulatory target cell (2). This event brings the Golgi apparatus, secretory lysosomes, and other vesicular compartments associated with the MTOC into close apposition with the synaptic membrane, thereby enabling the directional secretion of soluble factors toward the target cell. In this manner, MTOC polarization maintains both the specificity and precision of cytokine-mediated communication and cytotoxic killing.

Reorientation of the MTOC is stimulated by engagement of the T cell antigen receptor (TCR) with cognate peptide–major histocompatibility complex (pMHC) molecules on the surface of the antigen-presenting cell (APC) (2). This event stimulates a membrane-proximal tyrosine kinase cascade that leads to the activation of several key signaling enzymes, among them phospholipase-Cγ (PLC-γ), which hydrolyzes phosphatidylinositol 4,5-bisphosphate (PIP2) to yield two second messengers: inositol trisphosphate (IP3) and diacylglycerol (DAG). Whereas IP3 diffuses into the cytoplasm to promote calcium (Ca2+) signaling, DAG accumulates in the plasma membrane, where it activates both Ras- and protein kinase C (PKC)–dependent signaling by recruiting proteins that contain typical C1 domains. DAG also forms a marked intracellular gradient that is centered at the immunological synapse (3). We previously showed that this gradient serves as the polarizing signal to stimulate MTOC reorientation (4), and that it does so, at least in part, by recruiting three distinct PKC isozymes to the immunological synapse in an ordered cascade (5). Perturbations that disrupt the shape or the stability of the DAG gradient compromise MTOC polarization (4), which suggests that controlling the size of the region in which DAG signaling occurs is crucial for this response.

The mechanisms that shape DAG accumulation at the immunological synapse are largely unknown, but there are indications that they involve not only the regulated production of DAG but also its regulated destruction (4). In that regard, enzymes that metabolize DAG or convert it to another species represent candidates for the regulation of T cell polarity. In multiple cell types, DAG signaling is opposed by DAG kinases (DGKs), a family of enzymes that phosphorylate DAG to yield phosphatidic acid (PA) (6, 7). The predominant members of this family in T cells are DGK-α and DGK-ζ. To date, studies of DGK-α−/− or DGK-ζ−/− mice have suggested that the two enzymes play partially redundant roles in the attenuation of DAG signaling during T cell activation (811); however, DGK-α and DGK-ζ display marked differences in structure in regions both N-terminal and C-terminal to their respective catalytic domains, which implies that they might also have isoform-specific functions. Whether DGK-α and DGK-ζ influence T cell polarity by shaping the accumulation of DAG at the immunological synapse is not known. Here, we investigated this possibility with a combination of single-cell imaging and targeted loss-of-function experiments. Our results reveal a role for DGK-α, but not DGK-ζ, in DAG gradient formation at the immunological synapse and MTOC reorientation.

RESULTS

DGK-α, but not DGK-ζ, is required for polarization of the MTOC

We first adopted a loss-of-function approach to assess the importance of DGKs for MTOC polarization to the immunological synapse. CD4+ helper T cells were prepared with lymphocytes from DGK-α−/− or DGK-ζ−/− mice expressing the 5C.C7 TCR, which recognizes the moth cytochrome C88–103 (MCC) peptide bound to the class II MHC molecule I-Ek. These T cells were allowed to form conjugates with MCC peptide–loaded CH12 B cells (which served as APCs) and then were fixed and stained for CD4 and pericentrin, a marker of the MTOC (Fig. 1A). Conjugates containing wild-type T cells derived from littermate control mice were prepared in parallel for comparison. MTOC polarization toward the APC was quantified by calculating a “polarization index,” which is equal to the distance between the immunological synapse and the MTOC divided by the distance between the immunological synapse and the back of the T cell. This analysis revealed a substantial defect in DGK-α−/− T cells, with average polarization indices increasing ~65% relative to those of wild-type control T cells (Fig. 1, A and B). Most of this change could be attributed to T cells displaying a partially polarized phenotype in which the MTOC reoriented toward the APC, but did not become tightly apposed to the immunological synapse. By contrast, polarization responses of DGK-ζ−/− T cells were indistinguishable from those of wild-type controls (Fig. 1, A and C), suggesting that the two DGK isoforms operate in a nonredundant manner in this context.

Fig. 1 DGK-α is required for MTOC polarization in T cell–APC conjugates.

(A) Wild-type (WT, top), DGK-α−/− (αKO, middle), and DGK-ζ−/− (ζKO, bottom) 5C.C7 T cell blasts were incubated with MCC-loaded CH12 target cells (APCs), fixed, and stained with anti-CD4 and anti-pericentrin antibodies. Representative images are shown, with the position of the APC indicated by white text. Scale bars, 10 μm. (B and C) The polarization index was measured (see Materials and Methods for details) for (B) DGK-α−/− T cells (n = 48 conjugates) and (C) DGK-ζ−/− T cells (n = 40 conjugates) together with WT controls (n ≥ 40 conjugates). Data are presented in scatter plot format (left) as well as in histogram format (right) after binning the data into four cellular regions proceeding from the immunological synapse (region 1) to the distal pole (region 4), which is represented in the schematic inset in (B). Red lines and error bars in the scatter plots denote means and SEM, respectively. ***P < 0.001. All data are representative of at least two independent experiments.

To investigate the effects of DGK deficiency on MTOC dynamics more closely, we used a single-cell TCR photoactivation approach that enabled both the stimulation and the imaging of polarization responses with high spatiotemporal resolution (12). Briefly, T cells were attached to glass surfaces coated with immobilized I-Ek containing a photoactivatable version of the MCC peptide [NPE (1-ortho-nitrophenylethyl urethane)–MCC], which is nonstimulatory until exposed to ultraviolet (UV) light (fig. S1A). Then, UV irradiation of a micrometer-scale region of the surface beneath an individual T cell was used to stimulate localized TCR activation. This approach typically results in recruitment of the MTOC to the stimulatory region within 2 min of irradiation (4), where it remains for the duration of the imaging experiment (usually a total of 8 min).

DGK-ζ−/− T cells behaved normally in this assay, exhibiting polarization of the MTOC to a similar extent to that seen in T cells derived from wild-type littermate control mice (Fig. 2, A and B, and movies S1 and S2). By contrast, T cells lacking DGK-α often failed to reorient their MTOC properly, despite making apparent attempts to do so (Fig. 2, A and C, and movies S1 and S3). Indeed, the average distance between the irradiated region and the MTOC was substantially greater in DGK-α−/− T cells than in wild-type control T cells (Fig. 2C). This was particularly obvious during the second half of the time-lapse experiments, when the MTOC had typically settled at its new position. Analysis of individual MTOC trajectories revealed that the MTOC in DGK-α−/− T cells often reoriented to positions either short of or beyond the irradiated region (fig. S2). This marked imprecision was consistent with a role for DGK-α in focusing the polarization response.

Fig. 2 DGK-α is required for MTOC polarization in TCR photoactivation experiments.

(A to C) WT, DGK-α–/– (αKO), and DGK-ζ–/– (ζKO) 5C.C7 T cell blasts expressing GFP-tubulin were imaged and stimulated by localized UV irradiation (red circles) on surfaces coated with I-Ek-NPE-MCC. (A) Representative time-lapse montages of WT (top), αKO (middle), and ζKO (bottom) T cells are shown, with the irradiation time indicated by the white “UV” text. Scale bars, 10 μm. (B and C) Quantification of MTOC polarization in (B) ζKO and (C) αKO 5C.C7 T cell blasts (n ≥ 16 cells for each condition) compared to that in WT cells. Left: Average distance between the MTOC and the center of the irradiated region plotted against time, with the vertical purple line indicating UV irradiation. Right: Distance measurements from the second half of all time-lapse experiments (4 to 8 min) plotted in histogram format. (D) WT or αKO OT-1 CTLs (n = 15 cells each) were subjected to photoactivation experiments as described in (A), and the data were analyzed as described in (B) and (C). Error bars in graphs denote SEM. P values were computed using distance measurements from the second half of each time-lapse. All data are representative of at least four independent experiments.

Next, we investigated whether DGK-α played a similar role in CD8+ T cells, using as our model system cytotoxic T lymphocytes (CTLs) expressing the OT-1 TCR. To perform polarization experiments with these cells, we developed a photocaged form of their cognate ligand, the ovalbumin257–264 peptide SIINFEKL (hereinafter referred to as OVA peptide) bound to the class I MHC protein H-2Kb (fig. S1B). With this reagent, we could stimulate robust polarization responses in CTLs from wild-type mice (Fig. 2D). These responses were markedly impaired, however, in DGK-α−/− CTLs, which failed to properly reorient their MTOC toward the irradiated region (Fig. 2D). This phenotype was similar to what we observed in CD4+ T cells, which suggested a conserved role for DGK-α in both cell types.

The polarization defect observed in DGK-α−/− T cells was reversed by expression of full-length DGK-α, which demonstrated that the loss-of-function phenotype was specific (Fig. 3A). Furthermore, expression of a kinase-defective point mutant of DGK-α (KD DGK-α) failed to rescue MTOC reorientation in DGK-α−/− T cells (Fig. 3A), suggesting that the catalytic activity of the enzyme was required for regulation of MTOC polarity. Increasing the abundance of DGK-ζ through transduction was also ineffective at reversing the phenotype of DGK-α−/− T cells (Fig. 3B), lending further support to the interpretation that the two isoforms operate nonredundantly in this context. Together, these data suggest that DGK-α, but not DGK-ζ, is required for MTOC polarization in both CD4+ and CD8+ T cells.

Fig. 3 The defective polarization phenotype is specific to DGK-α.

(A and B) DGK-α–/– (αKO) 5C.C7 T cells expressing (A and B) GFP-labeled WT DGK-α, (A) KD DGK-α, or (B) DGK-ζ together with centrin-2–RFP (an MTOC marker) were imaged and stimulated by localized UV irradiation on surfaces coated with I-Ek-NPE-MCC. Polarization was analyzed by plotting the distance between the MTOC and the irradiated region as a function of time (left), as well as by plotting distance measurements from the second half of all time-lapse experiments in histogram format (right). Purple vertical lines indicate UV irradiation, and error bars denote SEM. Each curve represents n ≥ 10 cells. P values were computed using distance measurements from the second half of each time-lapse. Data are representative of at least two independent experiments.

DGK-α shapes the gradient of DAG at the immunological synapse

Given the importance of localized DAG signaling for reorientation of the MTOC, as well as the capacity of DGKs to control the abundance of DAG in the cell, we reasoned that DGK-α might influence MTOC polarization responses by controlling the range over which DAG accumulates at the immunological synapse. To test this hypothesis, we analyzed wild-type, DGK-α−/−, and DGK-ζ−/− 5C.C7 T cells expressing a fluorescent biosensor for DAG, which consists of the tandem C1 domains of PKCθ fused to green fluorescent protein (C1θ-GFP). These cells were stimulated on supported lipid bilayers containing I-Ek-MCC together with B7.1 and intercellular adhesion molecule-1 (ICAM-1), which are ligands for the costimulatory receptor CD28 and the integrin lymphocyte function–associated antigen-1 (LFA-1), respectively, on T cells. After 10 min, the T cells were fixed, stained for filamentous actin (F-actin), and imaged by total internal reflection fluorescence (TIRF) microscopy. Under these conditions, T cells normally form radially symmetric immunological synapses with the bilayer, which are bounded by a peripheral ring of F-actin (13). In wild-type cells, C1θ-GFP localized within this F-actin ring (Fig. 4A), indicating that the DAG gradient was constrained to the center of the immunological synapse. We quantified this result by dividing the diameter of the area over which C1θ-GFP was distributed by the diameter of the immunological synapse. This calculation yielded a ratio of less than 1 (Fig. 4B), consistent with a centralized distribution of DAG. A similar pattern of C1θ-GFP recruitment was observed in T cells deficient in DGK-ζ (Fig. 4, A and B), suggesting that this isoform does not influence the compartmentalization of DAG at the immunological synapse. In DGK-α−/− T cells, however, the distribution of DAG was substantially broader, with C1θ-GFP invading the peripheral space occupied by F-actin (Fig. 4, A and B). These results suggest that DGK-α, but not DGK-ζ, is required for shaping DAG accumulation at the immunological synapse, and further support the interpretation that DGK-α plays a unique role, among DGK isoforms, in establishing T cell polarity.

Fig. 4 DGK-α determines the range of the DAG gradient.

(A) WT, DGK-α−/− (αKO), and DGK-ζ−/− (ζKO) 5C.C7 T cells were transduced with retrovirus expressing C1θ-GFP; incubated on lipid bilayers containing ICAM-1, B7.1, and I-Ek-MCC; fixed; and stained for F-actin. Left: Representative TIRF images. Right: Linescans (derived from the white lines shown in the TIRF images) showing the fluorescence intensities of F-actin and C1θ-GFP. Scale bars, 10 μm. (B) Quantification of the relative size of the area of DAG accumulation (see Materials and Methods for details), which was calculated by normalizing the diameter of the DAG fluorescence signal to the diameter of the F-actin ring. Data are derived from n > 30 cells per condition. Red lines and error bars in the scatter plots denote means and SEM, respectively. ****P < 0.0001, *P < 0.05. (C) DGK-α+/− (Het, left) and DGK-α−/− (KO, right) 5C.C7 T cells expressing C1θ-GFP were imaged by TIRF microscopy and stimulated by localized UV irradiation on surfaces coated with I-Ek-NPE-MCC. Representative time-lapse montages are shown, with the irradiation time indicated by the white “UV” text. Scale bars, 10 μm. (D) The average normalized autocorrelation width (see Materials and Methods for details) of the accumulation pattern of C1θ-GFP was plotted against time for both 5C.C7 T cell blasts (left) and OT-1 CTLs (right). Curves represent n ≥ 10 cells each. Purple vertical lines indicate UV irradiation, and error bars denote SEM. P values were computed using averaged autocorrelation measurements from all time points after UV irradiation. All data are representative of at least two independent experiments.

We also analyzed the range of DAG accumulation in TCR photoactivation experiments. Recruitment of C1θ-GFP at the plasma membrane was typically observed within ~90 s of stimulation in both wild-type and DGK-α−/− T cells (Fig. 4C and movies S4 and S5), consistent with previous findings (4). The degree to which this response was compartmentalized to the irradiated region was measured by calculating the normalized autocorrelation width of the distribution of C1θ-GFP (see Materials and Methods for details). In wild-type T cells, photoactivation induced a decrease in normalized autocorrelation width (Fig. 4D), indicating that the DAG distribution polarized toward the site of TCR stimulation. We did not observe this focusing behavior in DGK-α−/− T cells, however, consistent with the idea that DGK-α is required for controlling the area over which DAG accumulates (Fig. 4D). We observed similar results in experiments with either 5C.C7 or OT-1 T cells (Fig. 4D), although the effects of DGK-α deficiency were less pronounced in the OT-1 system. In general, photoactivation experiments with OT-1 cells tended to be noisier than were the analogous experiments with 5C.C7 cells, which was possibly because of the low-level, UV-independent stimulation of the OT-1 TCR by photocaged H-2Kb-OVA. Nevertheless, together with the lipid bilayer studies described earlier, these data suggest that DGK-α limits the size of the area at the immunological synapse in which DAG accumulates in both CD4+ and CD8+ T cells.

Loss of DGK-α augments signaling downstream of the TCR, but not cytotoxicity

DGK-α deficiency enhances DAG-dependent signaling and cytokine secretion in both naïve and effector T cells (8, 9, 11, 14, 15). Consistent with these studies, we found that DGK-α−/− 5C.C7 helper T cells released substantially more interleukin-2 (IL-2) than did their wild-type counterparts after stimulation with antibodies against CD3 and CD28 (Fig. 5A); however, loss of DGK-α had no substantial effect on the killing of target cells by OT-1 CTLs (Fig. 5B). This finding was unexpected given that DGK-α−/− CTLs exhibited enhanced TCR-dependent phosphorylation of extracellular signal–regulated kinases 1 and 2 (ERK1/2) at early time points (Fig. 5, C and D). Previous studies established the importance of reorientation of the MTOC for optimal cytotoxic responses (4, 1619). Hence, the potential gains in cytotoxicity that could result from enhanced DAG signaling would likely be offset by suboptimal polarization of the MTOC toward the immunological synapse.

Fig. 5 DGK-α deficiency enhances signaling but not cytotoxicity.

(A) WT and αKO T cell blasts were activated on plastic surfaces containing ICAM-1, B7.1, and either stimulatory (MCC) or nonstimulatory (HB) pMHC, as indicated. IL-2 secretion was assessed by enzyme-linked immunosorbent assay. **P < 0.01. (B) WT and αKO OT-1 CTLs were mixed with OVA-loaded RMA-s target cells at an effector/target (E/T) ratio of 2:1. Specific lysis of RMA-s cells is graphed as a function of the concentration of OVA peptide. Functional assays in (A) and (B) were performed in triplicate, and data are representative of at least three independent experiments. (C) DGK-α+/+ (WT) and DGK-α–/– (αKO) OT-1 T cell blasts were stimulated by cross-linking of anti-CD3 and anti-CD28 antibodies for the indicated times. Cell lysates were then analyzed by Western blotting with antibodies against pERK and total ERK, which was a loading control. Blots are representative of three independent experiments. (D) Quantification of the relative ratios of pERK1/2 abundance normalized to that of total ERK for the indicated cells over time. Data were pooled from three independent experiments. P values refer to pairwise comparisons between WT and αKO cells. All error bars denote SEM.

DGK-α localizes to the periphery of the immunological synapse

Activation of the TCR stimulates the recruitment of both DGK-α and DGK-ζ to the plasma membrane (2023); however, the compartmentalization of each protein within the immunological synapse has not been examined. To address this, we prepared OT-1 CTLs that expressed GFP-labeled forms of DGK-α or DGK-ζ. These cells were stimulated on bilayers containing ICAM-1, H-2Kb-OVA, and B7.1; stained for F-actin; and imaged by TIRF microscopy. We found that DGK-α accumulated preferentially in the periphery of the immunological synapse, where it displayed substantial overlap with the F-actin ring (Fig. 6A). By contrast, DGK-ζ was distributed in a uniform manner over the entire immunological synapse (Fig. 6A). We quantified the annularity of these localization patterns by calculating the ratio between the average fluorescence intensity in the center of the immunological synapse and the average fluorescence intensity in the periphery. This analysis revealed a statistically significant difference in the localization of DGK-α and DGK-ζ (Fig. 6B). The annular configuration adopted by DGK-α was essentially the reciprocal of the centralized pattern that we observed earlier for C1θ-GFP (Fig. 4A), which suggests that DGK-α might surround the region of DAG accumulation to constrain its size.

Fig. 6 DGK-α localizes to the periphery of the immunological synapse in a PI3K-dependent manner.

(A) OT-1 CTLs were transduced with retroviruses expressing GFP-labeled DGK-α or DGK-ζ as indicated; stimulated on bilayers containing H-2Kb-OVA, B7.1, and ICAM-1; fixed; and stained for F-actin. Left: Representative TIRF images. Right: Linescans (derived from the white lines in the TIRF images) showing the fluorescence intensities of F-actin and GFP-DGK. Scale bars, 10 μm. (B) Quantification of the localization patterns of the indicated GFP-labeled DGK constructs at the immunological synapse was performed by comparing the mean fluorescence intensity (MFI) of GFP at the center of the immunological synapse with its MFI at the periphery (see Materials and Methods for details). Data are from n > 40 cells per condition. Red lines and error bars in the scatter plot denote means and SEM, respectively. ****P < 0.0001, **P < 0.01, *P < 0.05. (C and D) DGK-α−/− 5C.C7 T cell blasts were transduced with retroviruses expressing GFP-labeled (C and D) DGK-α, (C) ΔEF DGK-α, or (D) ΔEF2C1 DGK-α together with centrin-2–RFP. The resulting cells were used in TCR photoactivation experiments to assess MTOC polarization. Results were quantified by plotting distance measurements from the second half of all time-lapse experiments in histogram format. Data are from n ≥ 10 cells per condition. P values were computed using distance measurements from the second half of each time-lapse. (E and F) DGK-α−/− OT-1 CTLs expressing GFP–DGK-α were treated with wortmannin (wort) or dimethyl sulfoxide (DMSO; vehicle control); activated on lipid bilayers containing H-2Kb-OVA, ICAM-1, and B7.1; fixed; and stained for F-actin. (E) Left: Representative TIRF images. Right: Linescans (derived from the white lines in the TIRF images) showing the fluorescence intensities of F-actin and GFP–DGK-α. Scale bars, 10 μm. (F) Synaptic localization of GFP–DGK-α was quantified ratiometrically as described in (B). Data are from n = 14 cells per condition. Red lines and error bars in the scatter plot denote means and SEM, respectively. ***P < 0.001. All data are representative of at least three independent experiments.

The recruitment of DGK-α to the plasma membrane is dependent on its N-terminal regulatory region, which contains a recoverin homology domain, two Ca2+-binding EF hands, and two atypical C1 domains, which do not recognize DAG (21, 22, 24). The EF hands are thought to endow DGK-α with Ca2+ responsiveness by inhibiting the membrane localization and enzymatic activity of DGK-α in the absence of Ca2+ (21), whereas the C1 domains are required for membrane binding (24). To assess the relative importance of these domains for the compartmentalization of DGK-α within the immunological synapse, we analyzed two deletion mutants, one lacking the recoverin homology domain and EF hands (ΔEF DGK-α) and the other lacking the entire N-terminal regulatory region (ΔEF2C1 DGK-α) (fig. S3A). The localization pattern of the ΔEF DGK-α variant was similar to that of full-length DGK-α (Fig. 6B and fig. S3B), suggesting that the EF and recoverin homology domains are dispensable for enrichment at the periphery of the immunological synapse. By contrast, the ΔEF2C1 DGK-α variant exhibited weak membrane recruitment and no enrichment at the periphery (Fig. 6B and fig. S3B). Together, these results suggest that the compartmentalization of DGK-α within the immunological synapse requires the tandem C1 domains. In certain experimental systems, membrane recruitment of DGK-α is dependent on its own enzymatic activity (21, 24); however, we found that the localization of the KD DGK-α variant at the immunological synapse was indistinguishable from that of the wild-type protein (Fig. 6B and fig. S3B), arguing against a role for DAG turnover in this context.

To explore the relationship between the localization of DGK and the regulation of MTOC reorientation, we performed TCR photoactivation experiments with DGK-α−/− T cells that had been transduced with retroviruses expressing either the ΔEF DGK-α or the ΔEF2C1 DGK-α variant. Expression of ΔEF DGK-α in DGK-α−/− T cells rescued MTOC polarization to a similar extent as did expression of full-length DGK-α (Fig. 6C), suggesting that regulation by the EF hands is not required for DGK-α–dependent reorientation of the MTOC. By contrast, the ΔEF2C1 DGK-α variant failed to reverse the DGK-α−/− phenotype (Fig. 6D). Hence, DGK-α constructs that exhibited peripheral localization at the immunological synapse were, in general, capable of potentiating MTOC polarization. The exception was the KD DGK-α variant, which was localized to the periphery but lacked catalytic activity (Figs. 3A and 6B). Together, these results suggest that the annular accumulation of DGK-α in the membrane at the immunological synapse is required for its effects on T cell polarity.

Phosphatidylinositol 3-kinase promotes synaptic DGK-α localization and MTOC reorientation

The apparent importance of the peripheral accumulation of DGK-α prompted us to investigate the signaling pathways that controlled this behavior. Much attention has focused on the role of intracellular Ca2+ (21, 22); however, we previously demonstrated that Ca2+ signaling is dispensable for the TCR-stimulated reorientation of the MTOC (4), which implies the existence of other mechanisms controlling the localization of DGK-α at the immunological synapse. For example, phosphatidylinositol 3-kinase (PI3K) signaling stimulates the recruitment of DGK-α to the plasma membrane in a Ca2+-independent manner, and phosphatidylinositol 3,4,5-trisphosphate (PIP3), the product of PI3K, directly stimulates DGK-α activation (25). These observations are particularly intriguing in light of a study that showed that PIP3 also accumulates in an annular pattern at the periphery of the immunological synapse (26).

To assess the importance of PI3K signaling for DGK-α localization in our system, we treated OT-1 CTLs that expressed GFP-labeled DGK-α with wortmannin, a small-molecule PI3K inhibitor. The cells were then applied to stimulatory bilayers and imaged by TIRF microscopy. Wortmannin completely disrupted the peripheral accumulation of DGK-α (Fig. 6, E and F), yielding a uniform fluorescence distribution across the immunological synapse. These results suggest that PI3K signaling is required for the compartmentalization of DGK-α at the immunological synapse.

One would expect that disrupting the localization of DGK-α in this manner would also alter the distribution of DAG. To test this hypothesis, we inhibited PI3K in OT-1 CTLs expressing C1θ-GFP and then imaged the cells on stimulatory bilayers. Wortmannin resulted in a substantial broadening of the distribution of C1θ-GFP at the immunological synapse (Fig. 7, A and B). This phenotype was similar to that seen in DGK-α−/− T cells, which suggests that the peripheral localization of DGK-α is crucial for constraining DAG in this context. Consistent with this interpretation, wortmannin also impaired MTOC polarization in TCR photoactivation experiments (Fig. 7C and fig. S4A). As we had observed for DGK-α−/− T cells, this defect was associated with imprecise targeting of the MTOC to the irradiated region after reorientation (fig. S4B). Together, these data suggest that by compartmentalizing DGK-α within the immunological synapse, PI3K signaling is required for optimal MTOC polarization.

Fig. 7 PI3K activity promotes MTOC polarization and DAG gradient formation.

(A) WT and DGK-α−/− (αKO) OT-1 CTLs were transduced with retrovirus expressing C1θ-GFP and incubated on lipid bilayers containing ICAM-1, B7.1, and H-2Kb-OVA in the presence or absence of wortmannin (wort), as indicated. Cells were then fixed and stained for F-actin. Left: Representative TIRF images. Right: Linescans (derived from the white lines in the TIRF images) showing the fluorescence intensities of F-actin and C1θ-GFP. Scale bars, 10 μm. (B) Quantification of the relative sizes of the areas of accumulated DAG, which were calculated ratiometrically as described for Fig. 4. Data are from n > 30 cells per condition. Red lines and error bars in the scatter plot denote means and SEM, respectively. ****P < 0.0001, **P < 0.01. (C) WT and DGK-α−/− (αKO) OT-1 CTLs were transduced with retrovirus expressing centrin-2–RFP and subjected to TCR photoactivation experiments in the presence or absence of wortmannin, as indicated. Polarization was analyzed by plotting distance measurements from the second half of all time-lapse experiments in histogram format. P values were computed using distance measurements from the second half of each time-lapse. All data are representative of three independent experiments.

Studies suggest that class IA PI3K family members, primarily PI3Kδ, are responsible for the production of PIP3 at the immunological synapse (26, 27). To assess the role of PI3Kδ in the DAG-dependent polarization of the MTOC, we used IC87114, a specific inhibitor of PI3Kδ. IC87114-treated OT-1 T cells exhibited broadening of the DAG gradient and impaired reorientation of the MTOC compared to that in vehicle-treated cells (fig. S4, C and D), similar to the effects of wortmannin. We conclude that PI3Kδ-dependent PIP3 production stimulates DAG accumulation and cytoskeletal polarization at the immunological synapse.

DISCUSSION

Although it has been known for some time that DGK activity attenuates DAG-dependent signaling in T cells, the function of specific DGK isoforms remains poorly understood. Here, we demonstrated that DGK-α in T cells was required for optimal MTOC polarization toward the APC by constraining the area in which DAG accumulated at the immunological synapse. DGKs are generally viewed as being inhibitors of lymphocyte responses; however, our data suggest that by promoting cytoskeletal polarity, DGK-α exerts a positive effect on T cell activation and subsequent effector function. In that regard, it is probably worth reevaluating the role of this enzyme within the TCR signaling cascade.

Previous work suggested that there was substantial functional overlap between DGK-α and DGK-ζ. DGK-α−/− and DGK-ζ−/− T cells display phenotypic similarities, which include enhanced mitogen-activated protein kinase (MAPK) signaling and resistance to anergy (9, 14, 28). Furthermore, mice lacking both of the DGK isoforms exhibit a block in thymocyte maturation that is not observed in mice deficient in either isoform alone, which implies that they have redundant roles in T cell development (8). There are indications, however, that DGK-α and DGK-ζ might regulate DAG in different ways. DGK-ζ−/− T cells display more pronounced dysregulation of MAPK signaling than do their DGK-α−/− counterparts (8, 9, 14). Furthermore, overexpression of DGK-ζ in Jurkat cells attenuates the TCR-dependent production of DAG to a greater extent than does overexpression of DGK-α (20). Hence, DGK-ζ appears to be more important for determining the cellular abundance of DAG. In contrast, our data suggest that DGK-α plays a modulatory role in shaping the signaling cascade.

This functional distinction between the two DGK isoforms is mirrored by differences in their patterns of accumulation at the immunological synapse. Although initial studies of Jurkat cells suggested that DGK-ζ, but not DGK-α, is recruited to the immunological synapse (20), another study of primary T cells demonstrated that both isoforms accumulate there (14). We used TIRF microscopy to extend these observations, and found that DGK-α localizes preferentially to the periphery of the immunological synapse, whereas DGK-ζ displays a uniform distribution. Together with previous work, our results suggest a model in which DGK-ζ controls the overall magnitude of DAG signals, whereas DGK-α controls their subcellular organization. It will be interesting to see if this conceptual distinction is borne out in future studies of isoform-specific DGK function in T cells and other cell types.

Although our data suggest that DGK-α controls MTOC polarization by shaping the area in which DAG accumulates at the immunological synapse, we cannot exclude a role for DGK-generated PA in this process. Indeed, PA activates PLC-γ (29), which implies that DGK activity at the immunological synapse might promote the local production of DAG. This type of positive feedback loop could conceivably sharpen the gradient of DAG at the immunological synapse and enhance the precision of polarization responses. PA can also activate PKCζ (30, 31), an integral component of the PAR (for partitioning-defective) polarity complex, which includes the adaptor proteins PAR-3 and PAR-6 (32). PA binds directly to the Drosophila PAR-3 ortholog Bazooka (33), further suggesting that this complex is regulated directly by PA. Both PAR-3 and PKCζ are implicated in stimulating T cell polarity (34, 35), but precisely how they are coupled to DAG signaling, if at all, has remained unclear. It is tempting to speculate that PA might function as a molecular link between the two systems.

It initially seemed unlikely that DGK-α could influence MTOC polarization, because the TCR-dependent recruitment of DGK-α to the plasma membrane was thought to require Ca2+ signaling (21, 22), and Ca2+ is dispensable for polarization responses (4). We have potentially resolved this paradox by demonstrating that PI3K activity stimulates the localization of DGK-α to the periphery of the immunological synapse. We showed previously that PIP3 is enriched in this peripheral domain, where it promotes actin polymerization through the exchange factor Dock2 (26). A direct interaction with this pool of PIP3 would explain the pattern of accumulation of DGK-α at the immunological synapse, and would be consistent with a previous study that showed that PIP3 stimulates DGK-α activity in vitro (25). It is also possible, however, that PIP3 controls DGK-α localization through an as yet undefined adaptor complex. Regardless, our results demonstrate that crosstalk between PIP3-dependent and DAG-dependent signaling plays an important role in coordinating cytoskeletal responses downstream of the TCR.

It was somewhat surprising that DGK-α−/− CTLs did not exhibit enhanced cytotoxicity in our hands despite the fact that their DAG signaling responses were amplified relative to those of CTLs from wild-type mice. Optimal target cell killing, however, requires not only T cell activation but also proper reorientation of the MTOC. In the absence of DGK-α, it is likely that the effects of stronger signaling are antagonized by inefficient cytoskeletal polarization, which might yield little net change in cytotoxicity. This interpretation is consistent with data showing that pharmacological inhibition of DGK-α does not enhance CTL functionality unless the protein is overexpressed (36); however, our results are inconsistent with a study that showed that loss of DGK-α boosts killing responses in CTLs expressing chimeric antigen receptors (CARs) (15). The experimental system used in this previous study, however, is not strictly comparable to ours. CARs bind to their cognate ligands with much higher affinity than do TCRs, which presumably enhances both adhesion to the target cell and the strength of activating signals. These parameters could influence the relationship between polarity and cytotoxicity. Over time, the concept of blocking inhibitory regulators, such as DGK-α, to augment T cell function during adoptive immunotherapy has received considerable attention (15, 26, 37, 38). It may be useful to keep in mind, however, that suppression of presumed inhibitory proteins in complex signaling networks can have unanticipated results.

The idea that DAG gradient formation at the immunological synapse requires the coordinated production and destruction of DAG by PLCγ and DGK-α, respectively, is quite reminiscent of work that demonstrated that the accumulation of PIP3 at the leading edge of migrating cells is maintained by the opposing activities of PI3K and the lipid phosphatase PTEN (39, 40). In both systems, negative regulators act to focus a second messenger gradient and make it contingent on continuous activating signals. The polarized signaling that results is highly responsive to minute changes in the orientation and the intensity of receptor stimulation. This conserved strategy is likely to be a common theme in biological systems designed to follow moving targets in complex intercellular environments.

MATERIALS AND METHODS

Mice

The Institutional Animal Care and Use Committee of Memorial Sloan Kettering Cancer Center (MSKCC) approved the animal protocols used in this study. DGK-α−/− 5C.C7 TCR-transgenic mice and DGK-ζ−/− 5C.C7 TCR-transgenic mice were generated by crossing 5C.C7 TCR-transgenic RAG2−/− mice (Taconic) with mice deficient in DGK-α or DGK-ζ, respectively, which was followed by crossing of the heterozygous progeny. DGK-α−/− OT-1 TCR-transgenic mice were generated by crossing OT-1 TCR-transgenic RAG2−/− mice (Taconic) with DGK-α−/− mice, which was followed by crossing of the heterozygous progeny.

Cells

CD4+ T cells were isolated from the lymph nodes of 5C.C7 TCR-transgenic mice and stimulated with irradiated splenocytes from B10A mice at a ratio of 1:5 in the presence of 5 μM MCC peptide. CD8+ T cells were isolated from the lymph nodes of OT-1 TCR-transgenic mice and were stimulated at a 1:4 ratio with a mixture of irradiated splenocytes from B6 mice preincubated with 100 nM OVA peptide. Cells were maintained in RPMI medium containing 10% (v/v) fetal calf serum (FCS), 2 mM l-glutamine, 1 mM sodium pyruvate, nonessential amino acids, penicillin (50 U/ml), and streptomycin (50 μg/ml, Invitrogen/Gibco). IL-2 (30 IU/ml) was added 24 hours after the lymphocytes were isolated. CD8+ T cell cultures were split everyday, whereas CD4+ T cell cultures were split every 2 days with complete RPMI containing IL-2 (30 IU/ml). CH12 cells and RMA-s cells (a lymphoma cell line used as APCs for OT-1 CTLs) were cultured in RPMI medium fully supplemented as described earlier. Phoenix-ECO cells (which were used for retroviral transduction) were maintained in Dulbecco’s modified Eagle’s medium supplemented with 10% (v/v) FCS, 2 mM l-glutamine, 1 mM sodium pyruvate, nonessential amino acids, penicillin (50 U/ml), and streptomycin (50 μg/ml, Invitrogen/Gibco).

Constructs and retroviral transduction

Full-length coding sequences for mouse DGK-α and mouse DGK-ζ isoform 2 were amplified from murine T cell complementary DNA (cDNA) by polymerase chain reaction (PCR) and inserted downstream of and in frame with GFP using a TOPO-GFP cloning vector (4). The fusion construct was then transferred into a murine stem cell virus (MSCV) retroviral expression vector. Fragments encoding the ΔEF and ΔEF2C1 constructs were generated by PCR with full-length DGK-α cDNA as a template and subcloned into pMSCV as GFP fusions as described earlier. KD DGK-α, bearing the G432A point mutation, was obtained by two-step PCR mutagenesis as described previously (41). Centrin-2–red fluorescent protein (RFP), GFP-tubulin, and C1θ-GFP constructs were described previously (4, 41). Constructs were retrovirally transduced into T cells 48 or 72 hours after initiation of culture for CD8+ and CD4+ T cells, respectively, as previously described (4).

Peptides and proteins

To photocage the OVA peptide, an NPE protecting group was installed at the ε-amino position of the p7 Lys. OVA was prepared on resin by standard solid-phase peptide synthesis, with Lys7 incorporated as an ivDde-protected derivative. After chain assembly, the ivDde moiety was selectively removed by hydrazine treatment, and the exposed amino group was reacted with NPE–N-hydroxysuccinimide ester, which was prepared as previously described (12). The peptide was cleaved from the resin and globally deprotected with trifluoroacetic acid, and then purified by reversed-phase high-performance liquid chromatography. H-2Kb heavy chain bearing a C-terminal BirA tag was expressed into inclusion bodies in Escherichia coli and purified under denaturing conditions. It was then refolded in the presence of excess β2-microglobulin and either NPE-OVA or OVA peptide. After concentration in Millipore ultracentrifugation devices, the samples were biotinylated with the BirA enzyme and further purified by gel filtration chromatography (Superdex 200, GE Pharmacia). The preparation of biotinylated I-Ek-NPE-MCC, I-Ek-MCC, ICAM-1, and B7.1 was described previously (12, 42, 43).

Enzyme-linked immunosorbent assays

Ninety-six–well Immuno plates (Nunc) were coated with streptavidin (10 μg/ml) in 0.1 M NaHCO3 for 2 hours, followed by blocking for 2 hours with 2% (w/v) bovine serum albumin (BSA) in Hepes-buffered saline (20 mM Hepes, 150 mM NaCl). Plates were coated overnight with biotinylated B7.1 (1 μg/ml), ICAM-1 (1 μg/ml), and either I-Ek-MCC (1 μg/ml) or I-Ek-HB (1 μg/ml, nonstimulatory control pMHC). After coating, 1 × 105 5C.C7 T cell blasts were added to the plates. After 16 hours of incubation at 37°C, diluted supernatants were added into Immuno plates previously coated overnight with anti-mouse IL-2 antibody (1 μg/ml, clone JES6-1A12, BD Biosciences). After 2 hours of incubation of supernatants, the plates were washed and then incubated for an extra 2 hours with biotinylated anti-mouse IL-2 antibody (clone JES6-5H4, BD Biosciences). Plates were then incubated with extra-avidin alkaline phosphatase (at a 1:1000 dilution, Sigma) for 1 hour, and the substrate, p-nitrophenyl phosphate (Pierce), was added. Plates were read before saturation, a few minutes after addition of the substrate, at a wavelength of 405 nm.

Biochemical signaling assays

OT-1 CTLs (1 × 106 to 2 × 106) were preincubated on ice with biotinylated anti-CD3 (5 μg/ml, clone 2C11, eBioscience) and biotinylated anti-CD28 (1 μg/ml, clone 37.51, BioLegend) antibodies. The antibodies were then cross-linked by the addition of streptavidin (20 μg/ml, PeproTech) and incubated at 37°C. At various time points, 2 × 105 to 4 × 105 cells were removed and lysed by incubation in cold lysis buffer (10 mM tris-HCl, 5 mM EDTA, 1% NP-40, 0.5% sodium deoxycholate, 0.15 M NaCl, 1 mM sodium fluoride, 0.1 mM orthovanadate, and protease inhibitors). Soluble lysates were analyzed by Western blotting with anti–phospho-ERK1/2 (pThr202/Tyr204, clone 20G11, Cell Signaling) and anti-ERK1/2 (clone 137F5, Cell Signaling) antibodies.

Killing assays

RMA-s target cells were incubated with different concentrations of peptides for 45 min. Both RMA-s cells and CTLs were washed three times with RPMI supplemented with 1% serum without phenol red. Then, 15,000 RMA-s cells were incubated with 30,000 T cell blasts per well in 200 μl of medium, 1% serum for 5 hours. The release of lactate dehydrogenase (LDH) was measured with the LDH cytotoxicity detection kit (Clontech) according to the manufacturer’s instructions. All assays were performed in triplicate.

Photoactivation experiments

An eight-well Lab-Tek chambered cover glass (Nunc) was coated as previously described (12). For 5C.C7 T cells, surfaces were incubated with photocaged I-Ek-NPE-MCC at 0.5 μg/ml, the nonstimulatory pMHC I-Ek-HB at 3 μg/ml, and anti–MHC class I H-2Kk at 0.5 μg/ml (clone 36-7-5, BD Biosciences). For CD8+ T cells, surfaces were coated with H-2Kb-NPE-OVA (0.1 μg/ml), H-2Db-KAVY (1 μg/ml), and ICAM-1 (1 μg/ml). T cells (2 × 105) were added to each well in imaging medium (RPMI without phenol red medium, containing 5% FCS and Hepes). In certain experiments, cells were pretreated for 20 min with 0.5 μM wortmannin, 1 μM IC87114, or DMSO (as a vehicle control), and the drugs were maintained in the medium for the duration of the imaging experiment. Three-second time-lapse series were recorded for 8 min with an inverted fluorescence videomicroscope (Olympus IX-81) equipped with a 150× [numerical aperture (NA) 1.45] objective lens. Photoactivation was performed with a Mosaic digital diaphragm apparatus (Photonic Instrument) attached to a mercury (HBO) lamp (Olympus). One UV pulse was performed for 500 ms after 30 s of time-lapse recording to decage the peptide and deliver the TCR signal. To image GFP- or RFP-tagged constructs, 488- and 564-nm lasers (Melles Griot) were used, respectively. MTOC probes (tubulin and centrin) were imaged by epifluorescence illumination, whereas C1θ was imaged by TIRF microscopy.

Lipid bilayer experiments

Lipid bilayers were formed in eight-well Lab-Tek chambers as previously described (43). After incubation for 45 min with streptavidin (20 μg/ml), the bilayers were washed and incubated with biotinylated pMHC (H-2Kb-OVA or I-Ek-MCC), ICAM-1, and B7.1 proteins (1 μg/ml each). After bilayer preparation, 2 × 105 T cells transduced with retroviruses expressing the appropriate constructs were added to each well and incubated for 10 min at 37°C before cells were fixed for 5 min with 2% paraformaldehyde (PFA). Cells were then permeabilized with 0.5% Triton X-100 for 5 min, blocked with 3% BSA in phosphate-buffered saline (PBS), and stained for 45 min with Alexa Fluor 594–coupled phalloidin (Invitrogen) to visualize F-actin. When necessary, cells were pretreated with 0.5 μM wortmannin or DMSO, and drugs were maintained during cell activation. Fixed samples were imaged on an inverted fluorescence videomicroscope (Olympus IX-81) with a 63× (NA 1.45) or 150× (NA 1.45) TIRF objective lens.

Conjugate experiments

5C.C7 T cell blasts (1 × 105) were mixed with CH12 target cells (1 × 105), centrifuged at 1500g for 1 min to force conjugate formation, and then incubated for 10 min at 37°C. Cell conjugates were then resuspended in warm PBS, immobilized on polylysine-coated coverslips, and fixed with 2% PFA for 5 min at 37°C. Cells were stained for CD4 with a primary rat anti-mouse CD4 antibody (5 μg/ml, clone GK1.5, eBioscience) and then were permeabilized with 0.5% Triton X-100. After blocking with 2% BSA, samples were further stained with polyclonal rabbit anti-mouse pericentrin antibody (5 μg/ml, Abcam), followed by Cy3-labeled donkey anti-rat secondary antibody (1 μg/ml, Jackson ImmunoResearch) and fluorescein isothiocyanate–labeled donkey anti-rabbit secondary antibody (1 μg/ml, Jackson ImmunoResearch). Fixed samples were imaged with an SP-5 upright confocal microscope (Leica) equipped with a 63× objective lens.

Image analysis

SlideBook (Intelligent Imaging Innovations Inc.), ImageJ, and MatLab (MathWorks) software were used for image processing and analysis. MTOC reorientation after photoactivation was quantified by calculating the distance between the center of the irradiated region and the MTOC as a function of time. Polarization histograms were generated with Prism software (GraphPad) by sorting distance measurements from the second half of time-lapse experiments (4 to 8 min) into 0.2-μm bins (4). Statistical tests of MTOC reorientation data were computed by calculating the mean distance between the MTOC and the irradiated region for each cell during the second half of the time-lapse experiments. Conjugate experiments were analyzed by calculating a polarization index equal to the distance from the MTOC to the immunological synapse divided by the distance from the immunological synapse to the distal pole of the T cell. These data were then binned into one of four equally spaced regions starting at the immunological synapse and moving toward the distal pole of the T cell (44). To quantify the size of the DAG gradient in lipid bilayer experiments, the fluorescence intensity of C1θ-GFP and F-actin along the line bisecting the immunological synapse was computed with ImageJ software. The diameters of the C1θ-GFP and F-actin distributions were determined from this linescan by calculating the distance between the point at which the fluorescence signal reached 50% of its peak intensity above background and the point at which it fell below this 50% threshold. The relative size of the C1θ-GFP gradient relative to the F-actin ring was then expressed as ratio of these diameters. The size of the DAG gradient in the photoactivation experiments was determined by calculating the two-dimensional autocorrelation function for each image in the time-lapse (5). Gaussian fitting was then used to determine the characteristic width of the fluorescent signal for each image. Width measurements were normalized to values from the first 10 time points in the experiment (that is, before UV irradiation). Statistical tests of DAG gradient autocorrelation data were computed by calculating the average autocorrelation width for each cell for all time points after UV irradiation. The localization of DGK in lipid bilayer experiments was analyzed with linescans across the immunological synapse. For each linescan, the background-corrected MFI at the edges (positions F1 and F2) of the immunological synapse was compared with the background-corrected MFI of three equally spaced central positions (F3, F4, and F5) as follows: mean(F3 + F4 + F5)/mean(F1 + F2). We expect this “clearance ratio” to be ~1 for uniform distributions and <1 for annular patterns.

Statistical analysis

Statistics were calculated with Prism software. Student’s t test was used to assess the statistical significance of the data presented in Fig. 5, A and D. One-way analysis of variance followed by a Tukey’s multiple comparison test was used to analyze the data present in Fig. 4B, whereas a Kruskal-Wallis test followed by a Dunn’s multiple comparison test was used to analyze the data presented in Figs. 6B and 7B. All other P values were calculated with the Mann-Whitney test. Data conformity to a Gaussian model was assessed with a D’Agostino and Pearson omnibus normality test.

SUPPLEMENTARY MATERIALS

www.sciencesignaling.org/cgi/content/full/7/340/ra82/DC1

Fig. S1. Photoactivatable ligands for the 5C.C7 and OT-1 TCRs.

Fig. S2. DGK-α is required for MTOC polarization in TCR photoactivation experiments.

Fig. S3. Localization patterns of DGK-α variants at the immunological synapse.

Fig. S4. PI3K activity promotes MTOC polarization and synaptic DAG.

Movie S1. Localized TCR photoactivation induces MTOC reorientation.

Movie S2. DGK-ζ is dispensable for MTOC reorientation.

Movie S3. DGK-α is required for MTOC reorientation.

Movie S4. Localized TCR photoactivation stimulates focal DAG accumulation.

Movie S5. DGK-α is required for proper DAG gradient formation.

REFERENCES AND NOTES

Acknowledgments: We thank X. Zhong (Duke University) for the mice; the Molecular Cytology Core Facility at MSKCC for confocal microscopy; the Microchemistry Core Facility at MSKCC for peptide synthesis; and R. Joshi (University of Pennsylvania), E. Merino, and members of the Huse and M. O. Li laboratories for advice. Funding: Supported in part by the NIH (R01-AI087644) (to M.H.) and the Cancer Research Institute (to A.C.). Author contributions: A.C. and M.H. designed the experiments; A.C. collected and analyzed the data; N.S.B. provided technical assistance; A.L.F. developed the photoactivatable H-2Kb reagent; G.A.K. provided DGK-α−/− and DGK-ζ−/− mice and gave critical conceptual advice; and A.C. and M.H. wrote the paper. Competing interests: The authors declare that they have no competing interests.
View Abstract

Navigate This Article