Cytosolic PLA₂ is required for CTL-mediated immunopathology of celiac disease via NKG2D and IL-15

NKG2D is licensed to mediate cytolysis independently of TCR activation in CTLs when they are in an effector stage and in the presence of IL-15 or high doses of IL-2 (6). Importantly, under these conditions it is possible to assess how cPLA₂ affects NKG2D effector functions independently from other receptors. To determine that our findings were not restricted to a particular subset of effector CTLs, we analyzed the effect of cPLA₂ inhibition in a variety of effector CTLs. Specifically, we studied freshly isolated effector intestinal intraepithelial CTLs (IE-CTLs) that were prestimulated in vitro with IL-15, normal IE-CTL clones, IE-CTL clones derived from celiac patients, peripheral blood effector CTL (PB-CTL) clones, and the leukemia TALL-104 CD8 T cell line. This latter cell line was previously used as a model to study the NKG2D cytolytic signaling pathway (6). All clones and cell lines were grown in the presence of a high concentration of IL-2, which is known to substitute for IL-15.

The cPLA₂ inhibitor AACOCF3 (CF3) impaired NKG2D-mediated cytolysis in antibody-redirected cytolytic assays (Fig. 1 A, left). Importantly, this finding could be extended to cytolytic assays using MIC-transfected C1R cells as targets (Fig. 1 A, right). Furthermore, arachidonic acid (AA) significantly restored cytolysis, strongly arguing against a nonspecific effect of the cPLA₂ inhibitor AACOCF3 (Fig. 1 A, right). Finally, AACOCF3 and AA had a dose-dependent effect on NKG2D-mediated cytolysis in freshly isolated IE-CTLs (Fig. S1, available).

To further assess the critical role of cPLA₂ in NKG2D-mediated cytolysis and granule release, cPLA₂ was “knocked down” in TALL-104 (Fig. 1, B and C) and PB-CTLs (Fig. 1 B, right) using cPLA₂ small interfering (si)RNA. Importantly, cPLA₂ siRNA specifically targeted cPLA₂ expression, as it affected neither ERK nor JNK expression (not depicted) and activation (Fig. 1 B, left), nor 5-lipoxygenase (5-LO) expression (Fig. 1 C, left). The effect of cPLA₂ knock down on NKG2D-mediated cytolysis was assessed in two ways. First, in redirected cytolytic assays, as expected, NKG2D-mediated cytolysis was inhibited by up to 50% in the presence of cPLA₂ siRNA, but not control siRNA. In contrast, TCR-mediated cytolysis remained essentially intact for this level of specific cytolysis (Fig. 1 B, middle). The level of inhibition observed for NKG2D-mediated cytolysis was highly significant considering that some residual cPLA₂ could be detected (Fig. 1 B, left). Second, similar results were obtained when using MIC⁺C1R target cells (Fig. 1 D). Importantly, addition of AA (Fig. 1 D), but not a control polyunsaturated fatty acid, linoleic acid (LA), rescued NKG2D-mediated cytolysis (Fig. 1 D, right). The assays were performed either with global unsorted siRNA-transfected CTLs (Fig. 1, B and D, right), or with CTLs sorted into GFP⁺-transfected and GFP⁻-untransfected CTLs (Fig. 1 D, middle).

Having shown that cPLA₂ played a role in NKG2D-mediated cytolysis, we wanted to more specifically demonstrate its role in the release of cytolytic granules. We previously reported, that ligation of NKG2D receptors expressed on CTLs with monoclonal anti-NKG2D mAb triggered granule release using a BLT esterase assay (6). As anticipated, knock down of cPLA₂ by siRNA (Fig. 1 C) blocked degranulation. Similar results were obtained in the presence of the pharmacological inhibitor CF3 (unpublished data). Importantly, cPLA₂ inhibition did not affect the level of granzyme and perforin expression in CTLs (Fig. S2, available), further suggesting that cPLA₂ plays a role in NKG2D-mediated degranulation.

Interestingly, cPLA₂ inhibition by AACOCF3 or knock down by siRNA affected only marginally TCR-mediated cytolysis (Figs. 1 and 2).

Altogether, these results demonstrate that cPLA₂ and AA play a critical role in NKG2D-mediated, but not TCR-mediated, cytolytic granule exocytosis in effector CTLs.

NKG2D was shown to function as a co-stimulatory receptor that enhances TCR-dependent cytolytic responses (13, 29). We therefore investigated whether cPLA₂ was also implicated in NKG2D cytolytic co-stimulatory functions. This study had to be performed in CTLs grown in medium with low IL-2, in which NKG2D was not licensed to mediate direct killing. As previously reported (29), NKG2D significantly increased TCR-mediated cytolysis in two distinct PB-CTL clones (Fig. 2). Importantly, cPLA₂ inhibition prevented NKG2D enhancement of TCR-mediated cytolysis, and AA significantly rescued NKG2D co-stimulation (Fig. 2). In contrast, cPLA₂ inhibition had little or no significant effect on TCR-mediated cytolysis (Fig. 2).

In summary, these observations also support a role for cPLA₂ in NKG2D cytolytic co-stimulatory functions.

To establish a biochemical basis for cPLA₂ functions in CTLs, we determined NKG2D-induced cPLA₂ phosphorylation and AA release in several different CTL subsets, including in IE-CTL clones derived from celiac patients. Expression of cPLA₂ in mature T cells has been controversial (19, 30). However, using Western blot analysis, we demonstrated cPLA₂ expression in different effector CTLs (Fig. 3, left). Furthermore, as shown in Fig. 3, NKG2D cross-linking induced cPLA₂ phosphorylation in a dose-dependent manner (Fig. 3, left).

To further assess the ability of NKG2D to activate cPLA₂, we studied perinuclear translocation of cPLA₂ and cPLA₂ activity in IE- and PB-CTLs. As anticipated, NKG2D engagement induced perinuclear translocation of cPLA₂ (Fig. S3 A, available) and up-regulated cPLA₂ activity (Fig. S3 B), which were blocked in the presence of the pharmacological inhibitor CF3 (Fig. S3).

To determine whether the capacity of NKG2D to induce cPLA₂ phosphorylation in CTLs was dependent on their activation status, we sorted freshly purified resting CD8β⁺ peripheral blood T cells that were used either immediately for biochemical studies or to establish a short-term effector CTL line by allogenic stimulation, as previously described (31). Interestingly, NKG2D and IL-15 failed to induce cPLA₂ phosphorylation in resting CD8⁺TCRαβ⁺ peripheral blood lymphocytes, but not in the short-term derived effector CTL line (Fig. S4, available). These findings are in accordance with previously reported data (6) showing that licensing of NKG2D by IL-15 to mediate cytolysis was restricted to CTLs in the effector stage.

Finally, we investigated whether induction of cPLA₂ phosphorylation was associated with AA release (Fig. 3, right). This was investigated under conditions of stimulation resulting in maximal levels of cPLA₂ phosphorylation. TCR stimulation induced AA release. However, importantly, NKG2D could induce AA release independently of TCR engagement.

Collectively, these results demonstrate that NKG2D can induce cPLA₂ activation and mediate AA release independently of TCR stimulation. This unexpected function of NKG2D suggests that in tissues expressing stress-induced ligands, CTLs could participate in the activation of granulocytes sensitive to the presence of AA.

Previous studies had shown that NKG2D activated ERK and JNK through distinct pathways and that both MAP kinases regulated NKG2D-mediated cytolysis in CTLs in an additive manner (6). However, how ERK and JNK played a role in cytolysis remained elusive. Interestingly, cPLA₂ activation by surface receptors was shown to be dependent on MAP kinase activation (for review see reference [28]). This prompted us to investigate whether there was a link between ERK, JNK, cPLA₂ activation, and NKG2D-mediated cytolysis.

We first determined the kinetic of ERK, JNK, and cPLA₂ phosphorylation upon NKG2D stimulation. The peak of cPLA₂ phosphorylation occurred at 10–15 min after ERK and concomitant with JNKK1 and JNK phosphorylation (Fig. 4 A, top left). These results suggested that cPLA₂ phosphorylation was downstream of ERK and JNK.

To further evaluate the function of JNK and ERK in cPLA₂ activation and NKG2D-mediated cytolysis, we performed studies with pharmacological inhibitors targeting specific MAP kinases. As anticipated, cPLA₂ phosphorylation (Fig. 4 A, bottom left) and AA release (not depicted) were blocked in an additive manner upon ERK (PD98059) and JNK (SP600125) inhibition. In contrast, the p38 inhibitor SB 203580 had no effect (unpublished data), a result that is in agreement with its lack of function in NKG2D-mediated cytolysis (6). Finally, addition of AA reversed JNK inhibition, more significantly than ERK inhibition of NKG2D-mediated cytolysis (Fig. 4 A, right).

We next dissected the signaling pathway leading to JNK activation and further assessed its critical role in cPLA₂ activation using a dominant-negative and overexpression approach. Vav1, which regulates JNK activation through a Rho/Rac1→MAPK kinase kinase 1 pathway (32), was also reported to play an essential role in the NKG2D cytolytic pathway in NK cells (33, 34). These observations suggested that Vav1 could control JNK activation. In agreement, our kinetic analysis indicated that Vav1 activation preceded JNKK1 and JNK phosphorylation (compare Fig. 4, A and B, top left). In addition, a dominant-negative of Vav1, Vav1-C (33), blocked both JNK and cPLA₂ activation after NKG2D cross-linking in CTLs (Fig. 4 B, top right). To further assess the role of JNK in cPLA₂ phosphorylation, TALL-104 CTLs were transfected with a dominant-negative JNKK2, which was shown to prevent JNK activation. As anticipated, blocking JNK activation resulted in inhibition of cPLA₂ phosphorylation upon NKG2D engagement (Fig. 4 B, bottom left). Conversely, simple overexpression of constitutive active JNK (JNKK2-JNK1), but not inactive JNK (JNKK2[KM]-JNK1) (35), induced similar levels of cPLA₂ phosphorylation than NKG2D stimulation in CTLs (Fig. 4 B, bottom right).

Finally, because the upstream events leading to ERK activation were defined by a previous study (6) (see also Fig. 7), we focused on assessing the role of ERK in cPLA₂ activation. As predicted by the pharmacological studies, the dominant-negative of MEK-1 (MEK-2A) (36, 37), which is located upstream of ERK, blocked ERK and cPLA₂ phosphorylation (Fig. 4 C, left), whereas overexpression of constitutive active MEK-1 (MEK-2E) (36, 37), induced cPLA₂ phosphorylation (Fig. 4 C, right).

Altogether, these observations suggest that Vav is upstream of JNK, and that JNK and ERK control NKG2D-mediated cPLA₂ activation in CTLs in an independent and additive manner.

We had previously demonstrated that IL-15 or high levels of IL-2 (which can substitute for IL-15) licensed NKG2D to mediate cytolysis independently of TCR activation in effector CTLs (6, 13). Furthermore, we showed that IE-CTLs exposed in vivo to IL-15 in active celiac disease killed MIC⁺ target cells (6). Here, we investigated whether the licensing of NKG2D to mediated direct cytolysis in freshly purified IE-CTLs involved cPLA₂ activation by IL-15. We first determined whether IL-15 could activate cPLA₂. Interestingly, IL-15 induced not only ERK and JNK phosphorylation, as previously shown (6, 38), but also cPLA₂ phosphorylation (Fig. 5 A, left). The time course was similar to that of NKG2D (compare Fig. 5 A, left, and Fig. 4 A, left). In addition, cPLA₂ expression was increased in primary IE-CTLs stimulated for 48 h with IL-15, but not with IL-7 (Fig. 5 A, right). Furthermore, IL-15 synergized with NKG2D to induce cPLA₂ phosphorylation in a dose- (Fig. 5 B, left) and time-dependent manner (Fig. 5 B, right). cPLA₂ activity was significantly up-regulated only in freshly isolated IE-CTLs that were prestimulated with IL-15 for 48 h. Importantly, induction of cPLA₂ activity paralleled the ability of NKG2D to mediate degranulation (compare Fig. 5, C and D). Conversely, cPLA₂ inhibition blocked cPLA₂ activity (Fig. 5 C) and degranulation (Fig. 5 D), which was restored in the presence of AA (Fig. 5 D).

We next investigated whether NKG2D could induce AA release in primary effector IE-CTLs. As shown in Fig. 5 E, IL-15 and NKG2D independently induced AA release. The release was maximal when IE-CTLs were stimulated simultaneously by IL-15 and NKG2D. Importantly, AA release was also observed upon engagement of NKG2D by plate-coated monoclonal anti-NKG2D mAb (Fig. 5 E, left) and MIC-expressing EL4 fibroblasts (Fig. 5 E, right).

Altogether, these observations suggest that priming of the NKG2D cytolytic pathway by IL-15 in primary effector CTLs entails up-regulation of cPLA₂ expression and activation. Furthermore, IL-15 and NKG2D can induce AA release.

NKG2D and IL-15 were shown to play a critical role in celiac disease by licensing IE-CTLs to kill enterocytes (6). Furthermore, previous studies in active celiac patients reported that ERK (6) and JNK (38) phosphorylation were induced in IE-CTLs of active celiac patients. Together, these observations suggested that cPLA₂ might be activated in IE-CTLs of celiac patients.

In situ analysis of immunohistochemical staining with anti-CD3 antibody in combination with anti–phospho-cPLA₂ antibody or control antibody revealed that the frequency of IE-CTL–expressing phospho-cPLA₂ was highly increased in celiac patients (n = 6; 79 ± 8%) compared with controls (n = 4; 13 ± 4.6%; P = 0.001) (Fig. 6 B). In addition, to control for evaluation bias inherent to subjective morphological inspection and background staining, automatic image analysis was performed using the ACIS software (see Materials and methods) as previously described (39, 40). Specific cPLA₂ staining appears in red when staining is above background threshold values after overlay (Fig. 6 A). Importantly, the automatic count confirmed the manual count by showing a similar significant increase in the percentage of cPLA₂-positive IE-CTLs (Fig. 6 B). This result is even more striking in regard to the augmentation in the number of IE-CTLs in celiac disease, reaching classically up to 10 times the value of that observed in normal controls (41). The total number of IE-CTL–expressing phospho-cPLA₂ in celiac patients can therefore be estimated to be 40–60-fold higher than in controls. This finding is particularly relevant in the context of our observations that cPLA₂ plays a critical role in NKG2D cytolytic functions of celiac and IL-15–stimulated IE-CTLs (Figs. 1 and 5).

Furthermore, phospho-cPLA₂, similar to phospho-ERK and JNK (6, 38), was also highly up-regulated in enterocytes of celiac patients (Fig. 6 A), especially in the nuclear compartment (not depicted) where accumulation correlates with activation (for review see reference [28]).

Altogether, these results illustrate that cPLA₂ is activated in IE-CTLs of active celiac patients and support, in association with other present (Figs. 1 and 5) and past (6) observations, an in vivo role for cPLA₂ in celiac disease pathogenesis.

Our study establishes an unexpected connection between Il-15– and CTL-mediated immunopathology in celiac disease and inflammatory lipid biology. Our observations also suggest that NKG2D effector functions, particularly cytolysis, are critically regulated by cPLA₂ in CTLs. Furthermore, we identify an unexpected role for Vav, JNK, and ERK in NKG2D-mediated cytolysis by showing that they mediate cPLA₂ activation. Finally, our study suggests that NKG2D and IL-15 could contribute to tissue inflammation and granulocyte activation by promoting the release of AA in CTLs (Fig. 7).

The molecular basis for exocytosis is a field of high investigation. It involves trafficking of granules to the plasma membrane and membrane fusion events. There is increasing evidence that exocytosis can be triggered, either under conditions of high-level calcium mobilization sufficient to activate snare-mediated membrane fusion (for review see references [42]), or under conditions of low levels of calcium mobilization (43) involving AA and its eicosanoids metabolites, which are known as fusogens (44). This latter mechanism is thought to play a role in neutrophil and neuroendocrine cell degranulation (45, 46). It may be particularly relevant for surface receptors that, similar to NKG2D, signal through an adaptor molecule lacking an ITAM motif (10), and induce relatively low levels of calcium (34, 47) (Fig. S5, available). Pharmacological studies performed in the 1980s (48–52) suggested that inflammatory lipid mediators may play a role in NK cell–mediated lysis in human and rodents without defining the receptors involved (48–52). More recently, cPLA₂ was proposed to contribute to NKR-P1A–induced cytolysis in rat NK cells (53). However, recent findings in human indicate that engagement of NKR-P1A by its ligand LLT1 (54, 55) inhibited rather than induced NK cell–mediated lysis (54). Furthermore, the nature of the adaptor molecule associated with NKR-P1A remains undefined. Thus, the role of cPLA₂ could not be ascribed to particular cytolytic signaling pathways and to individual NK receptors in human. Our study unambiguously identifies the NKG2D–DAP10 in human CTLs as an immunoreceptor complex that requires cPLA₂ activation and AA release as integral components of its cytolytic machinery. In addition, our preliminary data indicate that this finding is not limited to CTLs, because the NKG2D cytolytic pathway in NK cells also involves cPLA₂ activation (unpublished data). Interestingly, our data suggest that cPLA₂ is not significantly involved in TCR-mediated cytolysis. Future studies will determine whether, more generally, ITAM-signaling receptors mobilizing high levels of calcium mediate granule exocytosis independently of cPLA₂. In agreement, NK cell receptors associated with ITAM-bearing adaptor molecules induce high levels of calcium flux (47) and cPLA₂-independent cytolysis (56). Importantly, however, even though cPLA₂ is not directly involved in TCR-mediated cytolysis, it plays a critical role in the co-stimulation of cytolytic T cell responses. Notably, NKG2D was unsuccessful to induce cPLA₂ phosphorylation in resting CD8⁺TCRαβ⁺ peripheral blood lymphocytes. This finding may explain why NKG2D fails to exert co-stimulatory function in resting noneffector CD8 T cells (57). Together, these observations are in agreement with the general concept, reported by Bryceson et al. (58, 59) that receptors using different adapters and signaling pathways contribute distinctively and synergistically to cytolysis.

Studies had identified PLC-γ2, PI3-K, and Grb2/Vav1 as key upstream signaling molecules of the NKG2D cytolytic pathway (Fig. 7) (33, 60, 61). Our previous analysis of the human NKG2D–DAP10 cytolytic signaling pathway suggested that ERK and JNK, activated through independent pathways, played critical roles in NKG2D- but not TCR-mediated cytolysis (Fig. 7) (6). The role of PI3K in cytolysis is complex and remains incompletely understood; among its potential roles, PI3K is thought to be involved in ERK activation (6, 62, 63). A critical role for Vav in natural killer signaling and cytotoxicity has been reported in mouse and human by several groups (33, 64–68). Vav is thought to play a role similar to that of ERK (62, 63) in cytolysis by promoting granule polarization (67). This study reveals that Vav1 is upstream of JNK in the NKG2D cytolytic pathway (Fig. 7). More importantly, it sheds a new light on the role of the PI3K–ERK and Vav–JNK pathways in NKG2D-mediated cytolysis by showing that ERK and JNK critically regulate cPLA₂ activation, which in turn critically regulates NKG2D-mediated degranulation and cytolysis (Fig. 7). Whether AA, which is released upon cPLA₂ activation, plays a role in granule exocytosis as fusogen (22, 44, 45) or indirectly through its role in the biosynthesis of eicosanoids remains to be determined. AA serves as a substrate to 5-LO and cyclooxygenase (COX), which drive the leukotriene and prostaglandin synthesis pathways, respectively (28). Interestingly, although NKG2D engagement induces 5-LO and COX activation, only 5-LO is involved in NKG2D-mediated cytolysis (unpublished data). These observations are in agreement with reports suggesting that leukotrienes, but not prostaglandins, play a role in lymphokine-activated killer activity (51, 52) and NK-mediated cytolysis (69, 70).

Convergent observations suggest that IL-15 and NKG2D may coordinately regulate CTL effector functions and mediate organ-specific immunopathology (for review see reference [15]). For instance, the role of NKG2D as a direct mediator of cytolysis was shown in celiac disease (6). Celiac disease is an inflammatory intestinal disorder with an autoimmune component triggered by dietary gluten in genetically susceptible individuals (71, 72). Although the role of gluten-specific, DQ2- or DQ8-restricted CD4 T cells in the lamina propria was accepted early on (73), the role of IE-CTLs was questioned because no gluten-specific IE-CTLs could be identified (for review see reference [71]). Our finding that IL-15 primed NKG2D in CTLs to kill MIC-expressing enterocytes in celiac disease (6, 13) provided an explanation as to how IE-CTLs could cause intestinal damage and malabsorption, despite their inability to recognize gluten. Licensing by IL-15 was shown to involve up-regulation of NKG2D and DAP-10 expression, and activation of the PI3-K→ERK signaling pathway (6). We now show that IL-15 is also involved in the activation of JNK. More importantly, we demonstrate that IL-15–mediated licensing of the NKG2D cytolytic pathway requires cPLA₂ activation, and that cPLA₂ is highly activated in IE-CTLs of active celiac patients. Interestingly, cPLA₂ is concomitantly activated in the intestinal epithelial cells. This activation of cPLA₂ may be secondary to the activation of enterocytes by IL-15 (74) and/or AA released by IE-CTLs.

Finally, our study reveals that NKG2D and IL-15 signaling in effector CTLs causes CTLs to release AA and potentially eicosanoids, such as leukotrienes and prostaglandins. Eicosanoid synthesis by CTLs is supported by our findings that NKG2D induces AA production and activates 5-LO and COX (unpublished data). Interestingly, AA and eicosanoids favor granulocyte recruitment and activation (75–79). These findings establish an unrecognized link between CTLs, NKG2D, IL-15, and inflammation, which may have as of yet unrecognized pathological implications. In particular, it may be relevant in immune-mediated diseases, such as rheumatoid arthritis and celiac disease, where NKG2D and IL-15 play a pathogenic role (for review see reference [15]) and where granulocytes are activated (80–84).

Collectively, our findings invite us to reexamine the role of IL-15 and NKG2D in inflammation and autoimmunity, and develop therapeutic strategies aimed at blocking cPLA₂ in diseases associated with dysregulated IL-15 expression and NKG2D activation. Future studies will determine whether 5-LO and COX should also be targeted. Our preliminary observations suggest that 5-LO, but not COX, is involved in NKG2D-mediated immunopathology. In addition, in view of observations suggesting that prostaglandins inhibit NK cell cytolytic responses (85, 86), blocking COX activation may even be deleterious.

For immunohistochemical studies, six adult patients with active celiac disease and four control aged-matched individuals undergoing biopsies for functional intestinal disorders of nonceliac origin were studied. Diagnosis of celiac disease was based on detection of antitransglutaminase antibodies, the presence of HLA DQ2 or DQ8, villous atrophy, and clinical and histological response to a gluten-free diet. In addition, IE-CTLs were isolated from surgical specimens of individuals undergoing gastric bypass for morbid obesity, as previously described (6). All subjects gave written informed consent and research was approved by the University of Chicago Institutional Review board.

IE-CTLs were purified from jejunal biopsies and surgical specimens, as previously described (31). Intraepithelial and peripheral blood NKG2D⁺TCR⁺CD8⁺ CTL lines and clones were obtained and cultured as previously described (31). PB-CTL clone #348 and #414 (cultured with 5 U/ml of IL-2) in which NKG2D exerts co-stimulatory cytolytic functions but cannot induce direct cytolysis were obtained and cultured, as previously described (1, 8). Resting peripheral blood CD8⁺TCRαβ⁺ T lymphocytes were purified and an effector CTL line was generated according to a previously described protocol (6, 31). In brief, peripheral blood lymphocytes were isolated from whole blood of healthy volunteers after Ficoll density gradient centrifugation (GE Healthcare). Cells were stained with anti-CD8β APC-conjugated mAb (BD) and purified by AutoMACS (Miltenyi Biotec) using anti-APC magnetic beads (Miltenyi Biotec). Purity (>95%) of isolation was confirmed by flow cytometry. Purified CD8⁺TCRαβ⁺ T cells were divided into two parts. One part was directly used for cPLA₂ signal transduction experiments. The other part was used to generate a short-term effector CTL line by allogenic stimulation, as previously described (31), to investigate whether the effector status impacted on the ability of NKG2D to induce cPLA₂ phosphorylation.

MICA-transfected (C1R-MIC) or control vector–transfected (C1R-Neo) C1R cells and CD8⁺TCRαβ⁺ cytotoxic leukemia TALL-104 cell line (American Type Culture Collection) were cultured, as previously described (6). EL4 (American Type Culture Collection TIB-39) are a murine T lymphoma cell line. MICA-transfected EL4 (EL4-MICA) and control EL4 cells were grown in RPMI 1640 supplemented with 10% FCS, glutamine, antibiotics, and G 418.

cPLA₂ inhibitor AACOCF3, MEK1/2 inhibitor PD98059, and JNK inhibitor SP600125 were obtained from Calbiochem. AA and LA were obtained from Sigma-Aldrich. Anti-CD3 (clone UCHT1, IgG1) and anti-NKG2D mAbs (clone 1D11, IgG1) and isotype-matched control IgG1 were purchased from BD; PE-conjugated anti-NKG2D 1D11 was purchased from eBioscience; and anti-Vav1, -cPLA₂, -ERK, -JNK-phospho-cPLA₂, -phospho-JNKK1, -phospho-PLCγ2, -phospho-ERK, and -phospho-JNK antibodies were purchased from Cell Signaling Technology. Antiphosphotyrosine mAb 4G10 was obtained from Millipore. Antiactin mAb was purchased from Sigma-Aldrich. Anti–5-LO mAb was obtained from Fitzgerald Industries Int. F(ab')₂ goat anti–mouse antibodies were obtained from Jackson ImmunoResearch Laboratories. Human IL-15 and IL-2 were purchased from BD.

Dominant-negative mutant Vav1 plasmid (pcDNA3.flag.h.Vav1-C) that contains only the C-terminal portion of Vav1 was a gift from P. Leibson and D. Billadeau (Mayo Clinic College of Medicine, Rochester, MN,) (33). Dominant-negative MEK1(S218a, S222a, MEK[2A]), and constitutively activated MEK1(MEK[2E]) expression plasmid were gifts from M.R. Rosner (University of Chicago, Chicago, IL) (36, 37). Expression vectors encoding JNKK2-JNK1 (which encodes a constitutively active form of JNK), JNKK2(K149M)-JNK1 (a kinase-deficient mutant), and JNKK2(KM) have been previously described (35). cPLA₂, 5-LO, and control synthesized siRNA oligonucleotides were purchased from Santa Cruz Biotechnology, Inc. TALL-104 CTL cells were electroporated with the Amaxa Nucleofector (Amaxa, Inc.), using Amaxa Cell Line nucleofection solution V and program T-20. Cells were cultured for 24–72 h before being used for experiments. For overexpression experiments involving HA-tagged JNKK2-JNK1, HA-tagged JNKK2(KM)-JNK1, or MEK(2E), TALL-104 CTLs were cotransfected with the indicated plasmids and GFP plasmid at a 5:1 ratio. 40 h after transfection, cells were sorted by flow cytometry into GFP⁺ and GFP⁻ cells with FACSAria (BD) before being analyzed for p-cPLA₂, p-JNK, and p-ERK expressions by Western blot. For siRNA experiments, 4 × 10⁶ cells were transfected with 120 pmol siRNA and 30 pmol GFP plasmid. On average, 40% of cells were transfected based on GFP expression. Efficiency and specificity of protein knock-down was assessed by Western blot with the appropriate antibodies.

To look at cPLA₂, JNK, and ERK phosphorylation, cells were serum-starved for 30 h. To test the effects of kinase inhibitors, cells were preincubated for 30 min at 37°C with the indicated inhibitors before stimulation. To cross-link immunoreceptors, cells were incubated for 4 min at 4°C with the indicated monoclonal antibody before adding F(ab')₂ GAM for the indicated duration at 37°C. Cells were lysed for 20 min in ice-cold lysis buffer containing fresh protease and phosphatase inhibitors (50 mM Tris-Hcl [pH 7.5]; 150 mM NaCl; 1% Triton-X100; 1 mM EDTA; 1 mM Na₃VO₄; 1 mM NaF; and protease inhibitor cocktail tablets). Cellular debris was removed by centrifugation at 15,000 rpm for 20 min at 4°C. Total lysates were subjected to SDS-PAGE electrophoresis and transferred to PVDF membranes (Bio-Rad Laboratories). Proteins were then detected by using the indicated antibodies, followed by HRP-conjugated goat anti–mouse (HRP-GAM) or donkey anti–rabbit (HRP-DAR) antibodies (Jackson Immunoresearch Laboratories) using the enhanced chemiluminescence (ECL) kit from GE Healthcare. Tyrosine-phosphorylated Vav1 was revealed by antiphosphotyrosine (4G10) antibody after immunoprecipitation of the cell lysates with rabbit polyclonal anti-Vav1 antibody.

Chromium release assays were performed as previously described using P815 cells (a Fcγ⁺ mouse mastocytoma; American Type Culture Collection), C1R-MICA1, EL4-MICA1, control C1R transfectants, and control EL4 cells at the indicated E:T ratio in triplicate wells (6). For Fc-dependent redirected cytotoxicity, effectors and targets were incubated in the presence of soluble anti-NKG2D or anti-CD3 mAbs at the indicated concentration. Chromium release was measured using a scintillation counter (Packard). Maximum release was determined by the addition of detergent (10% SDS) and spontaneous release ranged from 5–10% of the maximum. The percentage of specific cytotoxicity was calculated using the formula 100 × (cpm experimental − cpm spontaneous)/(cpm maximum − cpm spontaneous). When indicated, effector cells were treated for 30 min before and during the cytotoxic assay with different inhibitors, lipid mediators, or equivalent concentrations of the DMSO or ethanol vehicles.

10⁶ cells/ml IE-CTLs, PB-CTLs, and TALL-104 CTLs were labeled with 0.2 µCi ³H AA (specific activity 62.5 Ci/mmol; PerkinElmer) in RPMI 1640 or IMDM with 0.2% fetal bovine serum at 37°C 5% CO₂ for 1 or 8 h. After labeling, cells were washed at least three times to remove free ³H AA and, when indicated, pretreated with inhibitors or vehicles for 30 min in medium with 0.2% bovine serum albumin (BSA). Cells were then incubated with anti-NKG2D, anti-CD3 mAbs, or mouse IgG1 control at 37°C for 4 min followed by F(ab')₂ goat anti–mouse IgG for 1 h in medium with 0.2% BSA. Supernatants and cell pellets were separately collected by centrifugation. ³H AA was measured with a scintillation counter. Percentage of ³H AA release was calculated as supernatant ³H/(supernatant ³H + pellet ³H) × 100.

For surface staining, cells were incubated with fluorochrome-conjugated antibodies according to standard protocols. Fluorescence was analyzed on a six-color FACSCanto (Becton Dickinson) with quadrants set to score as negative >99% of control Ig-stained cells.

Granule release was evaluated as previously described (6). In brief, CTLs were suspended in RPMI medium containing 2% FCS, incubated in 96-well with 4 µg/100 µl of anti-NKG2D mAb or isotype control for 4 min, and then stimulated for 4 h with goat anti–mouse IgG F(ab')₂ at 37°C. Maximum granule release was determined using 1% Triton X-100. The supernatants were evaluated for esterase secretion using a standard N-benzyloxycarbonyl lysine thiobenzyl ester (BLT; Calbiochem). The percentage of BLT esterase activity was calculated using the following equation: (experimental BLT esterase release − spontaneous BLT esterase release)/(maximum BLT esterase release − spontaneous BLT esterase release) × 100.

cPLA₂ activity was assessed in lysates of primary IE-CTL, PB, and IE-CTL lines according to the manufacturer's procedure (Cayman Chemical), as reported (87), using arachidonoyl thio-PC as a synthetic substrate. In brief, 4 × 10⁶ cells were stimulated with anti-NKG2D or isotype control, and the pellets were lysed by three cycles of freeze thawing in lysis buffer (50 mM Hepes, pH 7.4, containing 1 mM EDTA, 1 µg/ml leupeptin, 5 µg/ml aprotinin, 1 mM PMSF, 2 mM Na₃VO₄, 2.5 mM NaF, and 1 mM DTT). 10 µl sample supernatant and 5 µl assay buffer were well mixed and the reaction was initiated by adding 200 µl substrate solution to the wells. After 1 h incubation at room temperature, 10 µl of DTNB/EGTA was added to each well to stop the reaction. Absorbance was measured at 414 nm. cPLA₂ activity was expressed as the percentage of nonstimulated control (absorbance of stimulated cells/absorbance of nonstimulated cells) × 100.

Immunohistochemical double staining for CD3 and phospho-cPLA₂ was performed on 4 µM 10% formalin-fixed paraffin sections using the double staining blocking kit (DAKO). Monoclonal anti-CD3 antibody was used at 1:200 dilution (BioGenex), and rabbit polyclonal anti–p-cPLA₂ and isotype antibody were used at 1:50 dilution.

To control for evaluation bias inherent to subjective morphological inspection and background staining, automatic image analysis with the pixel and cell/object-based image analysis was performed using ACIS software from Clarient, as previously described (39, 40). For each slide, 5–6 areas were scanned at 200× magnification to yield high-resolution digital images. The ACIS software applies the color-specific thresholds configured by an experienced user-pathologist (MT) to differentiate specific blue phospho-cPLA₂ staining within lymphocytes and epithelial cells from the surrounding blue background staining. The same threshold values for specific blue staining were applied to both control and active celiac biopsy sections. The overlay tool was used to highlight phospho-cPLA₂–positive areas (defined by blue staining above background) in red color within the CD3⁺ (brown) intraepithelial lymphocyte population (Fig. 6). From each digital image, at least 100 IE-CTLs were selected and numbered for analysis of phospho-cPLA₂ staining. All cells were classified as either double positive or only CD3 positive (phospho-cPLA₂ negative). Results obtained by software analysis were compared with cell counts obtained by manual morphological inspection.

Mixed effects models were constructed for all bar graphs in the figures. The treatment variable was included as a fixed predictor. Experimental plate was included as a random predictor in the model. The outcome variable (specific lysis) was log transformed after checking normality of the residuals from the mixed model. Tukey-adjusted pairwise comparisons are reported unless otherwise noted. The main effect of treatment was statistically significant for all conditions tested (P < 0.001).

For analysis of immunohistochemical data, celiac cases were compared with controls using exact Wilcoxon rank sum tests.

Fig. S1 shows dose-dependent inhibition by AACOCF3 (CF3) and rescue by AA of NKG2D-mediated cytolysis. Fig. S2 shows the effect of cPLA2 inhibition by AACOCF3 (CF3) on granzyme and perforin expression in IE- and PB-CTLs. Fig. S3 shows that NKG2D stimulation induces perinuclear cPLA2 translocation and cPLA2 activity in IE- and PB-CTLs in a time-dependent manner. Fig. S4 shows that NKG2D and IL-15 induce cPLA2 phosphorylation in effector, but not resting, peripheral blood CTL. Fig. S5 shows that NKG2D induces notably lower levels of calcium release than the TCR in CTLs. A supplemental materials and methods is also provided.

This paper is dedicated to the memory of Paul J. Leibson who pioneered the field of NK cell signaling. His work, in particular on NKG2D, has been a source of inspiration.

The authors would also like to thank Jean-Pierre Kinet, Mark Musch, William De Paolo, and Albert Bendelac for critical reading of the manuscript, Daniel D Billadeau for providing the Vav1 dominant-negative construct, Emily Kistner for help with statistical analysis, Ryan Duggan for cell sorting, and Terry Li for help in the execution of the immunohistochemistry studies. We would also like to thank the University of Chicago Celiac Disease Center and the Columbia Presbyterian Hospital Celiac Disease Center.

The work was supported by RO1 DK063158, RO1 DK58727, A130581 (TS), and University of Chicago DDRCC P30DK42086.

The authors have no conflicting financial interests.