EAT-2, a SAP-like adaptor, controls NK cell activation through phospholipase Cγ, Ca⁺⁺, and Erk, leading to granule polarization

To help understand the relative contributions of SAP family members in NK cells, the expression patterns of the three members in the mouse were analyzed in various NK cell populations (Fig. 1). Real-time PCR analyses of RNA obtained from highly purified NK cell subsets showed that SAP (encoded by Sh2d1a) and EAT-2 (encoded by Sh2d1b1) were expressed at all stages of NK cell maturation, including NK cell precursors (NKPs), immature NK cells (iNK cells), and mature NK cells (mNK cells; Fig. 1, A and B). Lower levels of both adaptors were noted in NKPs, whereas higher amounts were observed in mNK cells. SAP and EAT-2 RNAs were also found in splenic NK cells from poly I:C–primed mice and lymphokine-activated killer (LAK) cells. In contrast, ERT (encoded by Sh2d1b2) was solely observed in LAK cells.

Complementary information was obtained from the ImmGen database (Fig. 1 C). Once again, SAP and EAT-2 RNAs were co-expressed in all NK cell subsets tested. The latter included Ly49C/I⁺ and Ly49C/I⁻ cells, which are educated or not by class I major histocompatibility complex molecules, respectively. They also included NK cells isolated at different times after infection with mouse cytomegalovirus. In contrast, little or no ERT RNA was found in all NK cell populations. Thus, SAP and EAT-2 were co-expressed at all stages of NK cell maturation and in all NK cell subsets. ERT was exclusively found in LAK cells.

SAP mediates its activating signals via an arginine at position 78 (R78) in the SH2 domain, which binds and activates the Fyn kinase (Latour et al., 2001, 2003; Chan et al., 2003). This arginine is not present in EAT-2. Rather, EAT-2 possesses tyrosines in the region C-terminal to the SH2 domain, the so-called tail, which can undergo phosphorylation (Roncagalli et al., 2005). In mice and most other nonprimate species, the tail bears two tyrosines, tyrosine 120 (Y120) and tyrosine 127 (Y127; Fig. 2 A and not depicted). In contrast, in humans and other primates, it contains a single tyrosine, Y127.

To determine if the single C-terminal tyrosine of human EAT-2 was critical for its activating function, we generated polyclonal derivatives of the human NK cell line YT-S that expressed equivalent amounts of either WT or tyrosine 127-to-phenylalanine 127 (Y127F) human EAT-2 (Fig. 2 B). Parental YT-S cells express endogenous SAP and SLAM family receptor 2B4 but do not contain EAT-2 (Fig. 2, B and C). Cells were then tested for their capacity to mediate cytotoxicity toward HeLa or K562 target cells, expressing or not the ligand of 2B4, CD48 (Fig. 2, D and E). The ability of YT-S to kill targets expressing CD48 was greatly enhanced by expression of WT EAT-2. This effect was not seen when targets lacked CD48, implying that it was mediated by 2B4. It was also eliminated when Y127 of EAT-2 was mutated. Hence, the capacity of human EAT-2 to promote the activating function of 2B4 required the C-terminal tyrosine Y127.

To ascertain the role of the two C-terminal tyrosines found in mouse EAT-2, we created a knock-in (KI) mouse in which Y120 and Y127 were mutated to phenylalanines (Sh2d1b1^Y120,127F mouse, hereafter called EAT-2 KI mouse). The vector depicted in Fig. 3 A was used. Immunoblot analyses of lysates from bone marrow–derived DCs of these mice showed that the mutations did not destabilize the EAT-2 protein (Fig. 3 B). DCs were used for this study, as the amounts of protein needed for immunoblotting could only be obtained with LAK cells and these cells also express ERT. The available anti–EAT-2 antibodies recognize ERT. Mutation of the C-terminal tyrosines also did not affect the ability of EAT-2 to bind 2B4 (Fig. 3 C). Furthermore, it had no impact on NK cell development, as assessed by expression of CD27, CD11b, and various Ly49 receptors, or on the expression levels of SLAM family receptors CD48 and SAP (unpublished data), as reported for EAT-2 deficiency (Roncagalli et al., 2005; Dong et al., 2009, 2012).

The impact of the EAT-2 KI mutation on NK cell activation was examined (Fig. 4). First, purified ex vivo NK cells were tested for their ability to kill hematopoietic target cells (Fig. 4 A). In comparison to WT NK cells, EAT-2 KI NK cells displayed a reduced ability to kill the hematopoietic cell lines RMA-S (lymphoma) and YAC-1 (thymoma). This defect was analogous to that seen in NK cells from EAT-2–deficient (EAT-2 KO) mice (Dong et al., 2009, 2012). Second, NK cells were assessed for their aptitude to produce IFN-γ in response to target cells, using intracellular staining with anti–IFN-γ antibodies (Fig. 4 B). Once again, in comparison to WT NK cells, EAT-2 KI NK cells exhibited reduced production of IFN-γ in response to RMA-S and YAC-1. This defect was similar to the one observed in EAT-2 KO NK cells (Dong et al., 2009, 2012). However, there was no defect when cells were activated with PMA and ionomycin, which bypass NK cell receptors.

To address the function of SLAM family receptors, similar experiments were conducted using as targets the nonhematopoietic cell line B16 (melanoma), expressing or not ligands for individual SLAM family receptors (Fig. 4 C). When compared with WT NK cells, NK cells expressing the EAT-2 KI mutation had dramatically compromised NK cell activation when SLAM family ligands were expressed on B16. This was true for cells expressing the ligands of 2B4 (that is, CD48), CRACC, CD84, and Ly-9. The defects were similar to those observed in EAT-2–deficient NK cells. However, neither the KI mutation nor EAT-2 deficiency had any impact on the capacity of NK cells to kill B16 lacking SLAM family ligands (B16-GFP).

SAP family adaptors enhance NK cell activation by a dual mechanism (Dong et al., 2009, 2012; Kageyama et al., 2012; Zhao et al., 2012). They promote the activating function of SLAM family receptors and prevent their inhibitory function. In the case of SAP, activation, but not prevention of inhibition, is dependent on the signaling motif, R78 (Dong et al., 2012). To evaluate if this was also the case for EAT-2, we presented the data of Fig. 4 C in a different manner for a side-by-side comparison of the impact of engaging or not engaging 2B4 in the different mice (Fig. 4 D). To show reproducibility, data from multiple independent experiments are represented in Fig. 4 E. Engagement of 2B4 by CD48 enhanced killing of B16 by WT NK cells. In contrast, it suppressed cytotoxicity by EAT-2 KO NK cells. Surprisingly, it was also the case for EAT-2 KI NK cells.

Thus, the C-terminal tyrosines were needed for the ability of mouse EAT-2 to promote NK cell activation by SLAM family receptors. Unlike the situation of SAP, this putative signaling motif of EAT-2 was also required to prevent the inhibitory function of SLAM family receptors.

Previous reports suggested that EAT-2 may couple SLAM family receptors to a wide range of signals in NK cells involving Fyn, Lck, phosphatidyl inositol 3′ kinase (PI3′K), Akt, PLCγ, c-Cbl, and others (Morra et al., 2001; Tassi and Colonna, 2005; Calpe et al., 2006; Clarkson et al., 2007; Clarkson and Brown, 2009). However, whether EAT-2 or other effectors were responsible for these signals was not addressed genetically.

To identify the signals mediated by EAT-2 in human NK cells, YT-S cells expressing WT EAT-2 or not were stimulated with anti-2B4 antibodies. Then, a variety of biochemical signals was analyzed (Fig. 5). Immunoblotting of total cell lysates with anti-phosphotyrosine antibodies showed that expression of WT EAT-2 had no impact on the overall pattern of protein tyrosine phosphorylation induced by 2B4 stimulation (Fig. 5 A). WT EAT-2 also had no influence on tyrosine phosphorylation of 2B4 (Fig. 5 B). Likewise, it had no effect on tyrosine phosphorylation of Vav-1, a key target of SAP in NK cells (Fig. 5 C; Dong et al., 2012). Tyrosine phosphorylation of SHIP-1 was also not affected. These observations suggested that, unlike SAP (Dong et al., 2012), EAT-2 was not triggering activation of kinases such as Fyn or Lck. If this were the case, EAT-2 would likely cause a global enhancement of 2B4-evoked protein tyrosine phosphorylation.

Nevertheless, WT EAT-2 caused a prominent augmentation (approximately fivefold) of 2B4-triggered PLCγ-1 tyrosine phosphorylation. PLCγ-1 is the predominant PLCγ isoform expressed in YT-S cells and is also highly expressed in normal human NK cells (Ting et al., 1992). It also provoked a small increase (less than twofold) in c-Cbl tyrosine phosphorylation. In keeping with the effect on PLCγ-1, cells expressing WT EAT-2 also displayed augmented 2B4-triggered Erk activation and Ca²⁺ fluxes, the two downstream effectors of PLCγ (Fig. 5, D and E). No effect was seen on activation of Akt, suggesting that PI3′K was not activated (Fig. 5 D). Importantly, these various effects of EAT-2 were completely eliminated by the Y127F mutation, implying that they were mediated by the C-terminal tyrosine (Fig. 5, C–E).

Next, we ascertained whether EAT-2 was mediating similar signals in mouse NK cells (Fig. 6). To this end, WT, EAT-2 KI, and EAT-2 KO NK cells were stimulated with anti-2B4 antibodies and tyrosine phosphorylation of Vav-1 and SHIP-1 was first examined (Fig. 6 A). In keeping with the data with YT-S cells, EAT-2 KI or EAT-2 KO NK cells displayed no defect in the ability of 2B4 to enhance tyrosine phosphorylation of these targets. This was unlike SAP deficiency, which abolished 2B4-triggered tyrosine phosphorylation of Vav-1 (Dong et al., 2012). Surprisingly, no detectable tyrosine phosphorylation of PLCγ-2, the major isoform of PLCγ expressed in mouse NK cells (Tassi et al., 2005; Caraux et al., 2006; Regunathan et al., 2006), was seen upon engagement of 2B4. Moreover, triggering of 2B4 did not enhance tyrosine phosphorylation of PLCγ-2 in response to stimulation of CD16 (Fig. 6 B). Likewise, stimulation of 2B4 alone failed to induce Ca²⁺ fluxes (Fig. 6 C). These findings indicated that the full spectrum of 2B4-triggered signals was not evoked in normal mouse NK cells, in comparison with the human NK cell line YT-S. This may not be surprising, as the functional impact of engagement of activating receptors in primary immune cells is frequently weaker than in cell lines. This presumably reflects constraints on the signals emanating from activating receptors in primary cells. Nonetheless, stimulation of 2B4 on WT NK cells potentiated Ca²⁺ fluxes induced by the activating receptor CD16. Importantly, this enhancement was severely attenuated in EAT-KO and EAT-2 KI NK cells. A quantitation of data from three independent experiments is depicted in Fig. 6 D. All cells responded equally to ionomycin (Fig. 6 E).

PLCγ-2 is critical for the activating function of multiple NK cell receptors, including SLAM family receptors (Tassi et al., 2005; Caraux et al., 2006; Regunathan et al., 2006; Dong et al., 2012). To determine which downstream effectors of PLCγ were critical for the function of EAT-2, we analyzed the impact of prevention of Ca²⁺ fluxes, Erk activation, or both (Fig. 7 A). YT-S cells were incubated with the chelator of Ca²⁺ BAPTA-AM (1,2-bis(2-aminophenoxy)ethane-N,N,N′,N′-tetraacetic acid tetrakis(acetoxymethyl ester)), the MEK1 inhibitor PD98059, or both. BAPTA-AM or PD98059 alone resulted in a partial inhibition of cytotoxicity toward K562 cells expressing CD48. A more pronounced effect was seen when BAPTA-AM and PD98059 were combined. These effects were seen whether YT-S expressed EAT-2 or not, in keeping with the idea that multiple activating receptors are using these effectors.

As expected, BAPTA-AM also fully prevented 2B4-evoked Ca²⁺ fluxes (Fig. 7 B). Surprisingly, it also abolished 2B4-triggered Erk activation (Fig. 7 C). In contrast, PD98059 had only a small inhibitory impact on 2B4-induced Erk activation. The latter two findings suggested that Ca²⁺ fluxes, rather than MEK1, were primarily responsible for Erk activation in response to 2B4.

Therefore, unlike earlier proposals that EAT-2 triggers a broad range of signals in NK cells, EAT-2 induced a specific subset of signals in response to 2B4 engagement. In YT-S cells, these signals included tyrosine phosphorylation of PLCγ, Ca²⁺ fluxes, and Erk activation. In mouse NK cells, they were limited to an enhancement of Ca²⁺ fluxes triggered by CD16. All these responses were strictly dependent on the C-terminal tyrosines of EAT-2.

Next, the molecular mechanism by which EAT-2 coupled to PLCγ was evaluated. First, we ascertained which PTK was mediating tyrosine phosphorylation of EAT-2 (Fig. 8 A). Co-transfection experiments in Cos-1 cells showed that only the Src family kinases Fyn and, to a lesser extent, Lck, Src, and Lyn were able to induce tyrosine phosphorylation of EAT-2. Then, we tested whether EAT-2 was able to associate with full-length PLCγ in intact cells (Fig. 8, B and C). cDNAs encoding mouse EAT-2, PLCγ-1, and Fyn were co-transfected in cells (Fig. 8 B). When Fyn was expressed, there was prominent co-immunoprecipitation of EAT-2 with PLCγ-1. A much less extensive association was observed when Fyn was not expressed, implying that the interaction was mediated by Fyn-dependent tyrosine phosphorylation. To ensure that the interaction was due to binding of phosphorylated EAT-2 to PLCγ-1, rather than binding of EAT-2 to phosphorylated PLCγ-1, similar experiments were performed using EAT-2 polypeptides mutated at a residue critical for phosphotyrosine binding by the SH2 domain (arginine 54-to-leucine 54; R54L mutation; Morra et al., 2001) in the absence or the presence of additional mutations of the C-terminal tyrosines (Fig. 8 C). Y127 of EAT-2, but not Y120 or an intact SH2 domain, was required for the EAT-2–PLCγ-1 interaction.

To show directly that binding was dependent on phosphorylation of Y127, synthetic peptides corresponding to the C terminus of mouse EAT-2, phosphorylated or not at Y120, Y127, or both, were incubated with lysates from mouse LAK cells (Fig. 8 D). EAT-2 peptides phosphorylated at Y127, alone or in combination with Y120, bound to PLCγ-2. However, peptides phosphorylated at Y120 alone or not phosphorylated did not bind. A peptide encompassing the C-terminal tail of human EAT-2 phosphorylated at Y127, but not a nonphosphorylated version, also bound to PLCγ-1, as reported previously (Clarkson and Brown, 2009).

We also analyzed the ability of immobilized fusion proteins bearing the N-terminal or C-terminal SH2 domain of PLCγ-1 to bind mouse EAT-2, which was expressed in cells in combination with Fyn to enable phosphorylation of EAT-2 (Fig. 8 E). Unfortunately, the N-terminal domain from PLCγ-2 was not soluble when produced in bacteria (unpublished data). As controls, SH2 domains of Fyn, Lck, Hck, Csk, Vav-1, SHIP-1, and Grb-2, as well as combined SH3 and SH2 domains of Fyn and Hck, were tested. Only the N-terminal SH2 domain of PLCγ-1 interacted with EAT-2. Unlike previously reported (Clarkson et al., 2007), there was no binding of the Fyn SH2 domain to EAT-2.

Lastly, we tested the capacity of immobilized human or mouse EAT-2 peptides, phosphorylated or not at the C-terminal tyrosines, to bind the SH2 domain of PLCγ-1 (Fig. 8 F). Phosphorylated EAT-2 peptides from human or mouse, but not unphosphorylated peptides, interacted with the SH2 domain of PLCγ-1. No interaction was seen with GST alone.

These data implied that, upon phosphorylation by Src family kinases, EAT-2 directly associated with PLCγ. This association was mediated by Y127 of EAT-2 and the N-terminal SH2 domain of PLCγ.

NK cell–mediated cytotoxicity involves several steps (Orange, 2008). The initiation phase enables formation of stable conjugates between NK cells and target cells. SAP plays a key role during this phase (Dong et al., 2012). This is followed by the effector phase, which sequentially involves: (1) actin polymerization at the area of contact between NK cells and target cells, at the so-called NK cell synapse, (2) polarization of microtubule-organizing center (MTOC) and cytotoxic granules toward the NK cell synapse, (3) release of cytotoxic granules toward target cells, and (4) target cell killing.

To test if EAT-2 promoted the initiation phase, conjugate formation was evaluated using mouse NK cells lacking EAT-2 (Fig. 9 A). NK cells from WT, EAT-2 KO, or SAP KO mice were incubated with RMA-S cells, and conjugate formation was monitored. Whereas SAP deficiency resulted in compromised conjugate formation, loss of EAT-2 had no impact on this response. The impact of EAT-2 on conjugate formation was also ascertained in the YT-S system (Fig. 9 B). Compared with YT-S cells lacking EAT-2, cells expressing EAT-2 did not exhibit enhanced conjugate formation with K562 cells expressing CD48. In fact, even though EAT-2 promoted NK cell cytotoxicity (Fig. 2 E), it caused a reduction of conjugate formation with these targets (Fig. 9 B). This was seen at multiple time points (unpublished data). This effect was not seen in NK cells expressing EAT-2 Y127F. The latter finding implied that this effect was not simply a result of displacement of SAP from 2B4 due to EAT-2 overexpression but perhaps related to an inhibitory effect of EAT-2 signaling on adhesion.

We also examined the impact of CRACC on conjugate formation (Fig. 9 C). CRACC is a member of the SLAM family that interacts with EAT-2 but not SAP (Tassi and Colonna, 2005; Cruz-Munoz et al., 2009). Being a self-ligand, it is triggered when CRACC is expressed on target cells. When CRACC was expressed with EAT-2 in YT-S cells, it did not augment conjugate formation with K562 cells expressing CRACC. This was despite the fact that CRACC augmented killing of K562 cells expressing CRACC (Cruz-Munoz et al., 2009). We concluded that, unlike SAP, EAT-2 did not enhance conjugate formation with target cells.

To ascertain whether EAT-2 was promoting the effector phase, events occurring in conjugates of YT-S cells with K562 cells expressing CD48 were visualized by confocal microscopy (Fig. 10). YT-S, rather than primary NK cells, were chosen for these studies, as they are more homogeneous and easier to analyze by confocal microscopy. Phalloidin staining highlighted polymerized actin, whereas staining with anti-tubulin or anti-perforin antibodies identified MTOCs and cytotoxic granules, respectively (Fig. 10 A). Using this approach, various stages in the effector phase were identified (Fig. 10 B). Stage 0 corresponds to conjugates in which actin is not polymerized, and MTOC and granules are not polarized toward the synapse. Stage 1 represents conjugates in which polymerized actin, but not MTOC and granules, accumulates at the synapse. Stage 2 indicates conjugates where MTOC and granules have partially migrated toward the synapse. And stage 3 shows conjugates in which the MTOC and granules have fully migrated toward the synapse.

Comparison of YT-S cells expressing WT EAT-2 or not showed that EAT-2 augmented the proportions of conjugates exhibiting fully polarized (stage 3) or partially polarized (stage 2) MTOC and granules (Fig. 10 C). Conversely, it reduced the proportions of conjugates lacking MTOC or granule polarization (stage 1), or devoid of actin polymerization (stage 0). Such effects were not seen in YT-S cells expressing Y127F EAT-2. These data were statistically validated by correlation coefficient (r) analysis. Importantly, the enhancement of polarization by WT EAT-2 was seen even after a correction was made for the reduction of conjugate formation (Fig. 10 D). This implied that the increase in stages 2 and 3 conjugates was not simply due to a reduction of the stability of stages 0 and 1 conjugates but rather reflected a true acceleration of polarization.

To determine which downstream effectors of EAT-2 were mediating these effects, the impact of BAPTA-AM and PD98059 was examined (Fig. 10 E). BAPTA-AM alone, but not PD98059 alone, caused a marked increase in stage 0 conjugates, coupled to a decrease in stage 1, 2, and 3 conjugates. The combination of BAPTA-AM and PD98059 had effects similar to those of BAPTA-AM alone. Of note, these effects were seen not only in YT-S cells expressing WT EAT-2 but also in those expressing GFP alone. This finding indicated that Ca²⁺ fluxes were generally required for granule polarization, even in the absence of EAT-2.

Although microscopic analyses of primary mouse NK cells are challenging, we wanted to obtain support for the idea that EAT-2 signaling also promoted granule polarization in the mouse system (Fig. 10 F). To this end, we analyzed CD107a externalization, a consequence of degranulation, in NK cells from WT, EAT-2 KO, or EAT-2 KI mice. Compared with WT NK cells, NK cells from EAT-2 KO or EAT-2 KI mice displayed reduced exposure of CD107a in response to RMA-S targets. However, CD107a exposure in response to PMA and ionomycin was not affected. Therefore, EAT-2 promoted MTOC and granule polarization toward the NK cell synapse. This effect required Y127 and was mediated by the ability of EAT-2 to enhance Ca²⁺ fluxes.

To gain insight into the respective roles of EAT-2 and SAP in NK cells, their expression patterns in various NK cell populations were examined in the mouse. These studies showed that EAT-2 and SAP were co-expressed at all stages of NK cell differentiation and in all mNK cells. They also implied that EAT-2 and SAP might need to complement each other in all subsets of NK cells. In contrast, the third SAP family adaptor in the mouse, ERT, was detected only in LAK cells. It was not found in any freshly isolated NK cell population. The latter finding was in agreement with earlier studies showing that ex vivo NK cells from mice lacking ERT had no appreciable defect (Wang et al., 2010).

To elucidate the mechanism by which EAT-2 cooperates with SAP to promote NK cell activation, we focused on its C-terminal tyrosines. These tyrosines are not present in SAP and can undergo phosphorylation in NK cells (Roncagalli et al., 2005). Comparison of the EAT-2 sequence from several species showed that, in humans and other primates, the C-terminal tail contains only one tyrosine, Y127. In contrast, in most other species including mice, the tail encompasses two tyrosines, Y120 and Y127.

Overexpression studies of EAT-2 previously indicated that the tyrosines of EAT-2 were required for EAT-2 to suppress NK cell activation (Roncagalli et al., 2005). However, subsequent studies demonstrated that EAT-2 exhibits distinct functions whether target cells were hematopoietic or nonhematopoietic in origin. In the presence of hematopoietic target cells, EAT-2 is a positive regulator of NK cell activation (Dong et al., 2009, 2012). This correlates with the fact that only hematopoietic cells express ligands for SLAM family receptors, and is the predominant function of EAT-2 in NK cells.

To address the roles of the C-terminal tyrosines in this function, human and mouse EAT-2 polypeptides carrying or lacking these tyrosines were expressed in NK cells. Expression of human EAT-2 in the NK cell line YT-S showed that Y127 was absolutely needed for the capacity of human EAT-2 to promote NK cell–mediated cytotoxicity in response to target cells expressing the ligand of 2B4. Likewise, creation of a KI mouse mutant expressing Y120,127F EAT-2 revealed that the tyrosines were required for the ability of mouse EAT-2 to augment NK cell cytotoxicity and IFN-γ production in response to targets expressing the ligand of 2B4, Ly-9, CD84, or CRACC. The loss of function caused by the tyrosine mutations was not due to reduced expression of EAT-2 or diminished association with SLAM family receptors. Rather, it was related to a compromised ability to mediate activating signals.

Biochemical studies in YT-S cells showed that human EAT-2 coupled 2B4 to specific signals involving PLCγ and its downstream effectors, Ca²⁺ fluxes and Erk. These signals were strictly reliant on Y127. There was no impact on other characterized 2B4-triggered tyrosine phosphorylation substrates, including 2B4 itself, Vav-1, and SHIP-1 (Chen et al., 2004). Moreover, there was no influence on Akt activation. A small increase in c-Cbl tyrosine phosphorylation was seen, although its mechanism and functional impact remain to be clarified. As c-Cbl is a ubiquitin ligase involved in signal termination (Thien and Langdon, 2005), c-Cbl tyrosine phosphorylation might be part of a negative feedback mechanism aimed at terminating the function of EAT-2.

Analyses of NK cells from EAT-2–deficient mice showed that, unlike SAP (Dong et al., 2012), EAT-2 did not promote tyrosine phosphorylation of Vav-1. However, in keeping with the effects of human EAT-2 in YT-S cells, mouse EAT-2 augmented the ability of 2B4 to enhance CD16-triggered Ca²⁺ fluxes. This effect was diminished when the C-terminal tyrosines were mutated. Although we were unable to detect tyrosine phosphorylation of PLCγ in mouse NK cells, these observations supported the idea that, by way of the C-terminal tyrosines, mouse EAT-2 was also coupling SLAM family receptors to the PLCγ pathway. It is possible that PLCγ was activated upon 2B4 triggering in mouse NK cells, but was not undergoing tyrosine phosphorylation. Alternatively, it is plausible that the extent of tyrosine phosphorylation of PLCγ in response to 2B4 stimulation in mouse NK cells was not sufficiently marked for detection in our assays.

Transfection experiments provided evidence that Src family kinases, in particular Fyn, were responsible for phosphorylation of the C-terminal tyrosines of EAT-2. This notion was also consistent with the fact that the tyrosines are located in typical consensus sequences for phosphorylation by Src kinases (Songyang et al., 1994a, 1995). Co-immunoprecipitation experiments and in vitro binding studies also supported the idea that the interaction between the C-terminal tyrosines of EAT-2 and PLCγ was direct. Moreover, it was dependent on phosphorylation of Y127, which bound to the N-terminal SH2 domain of PLCγ. The sequence surrounding phosphorylated Y127, pYVDV (where pY is phosphotyrosine, V is valine, and D is aspartic acid), is a typical binding site for the PLCγ SH2 domain (consensus sequence pY[V/I/L][E/D][L/I/V], where I is isoleucine, L is leucine, and E is glutamic acid; Songyang et al., 1993).

Although Y127 played a pivotal role in the function of EAT-2, it is noteworthy that, in mice and other nonprimate species, EAT-2 contains an additional tyrosine in the tail, Y120. This tyrosine is located in a conserved sequence (pYENT in the mouse, where N is asparagine and T is threonine), which represents a typical binding site for signaling molecules like Grb-2 and 3BP2 (Songyang et al., 1994b). Hence, in these species, EAT-2 may mediate additional signals via Y120. It is also possible that, in species lacking Y120, EAT-2 couples to these effectors via another sequence. Additional studies are needed to examine these possibilities.

One surprising observation was that treatment of NK cells with the Ca²⁺ chelator, BAPTA-AM, resulted in prominent decrease in 2B4-evoked Erk activation. In contrast, the MEK1 inhibitor PD98059 caused a minimal compromise in Erk activation. Combined, these data suggested that Ca²⁺, rather than MEK1, was primarily responsible for activation of Erk in response to PLCγ activation. These findings were reminiscent of results recently obtained in B cells (Limnander et al., 2011). In immature B cells, PLCγ-dependent Ca²⁺ fluxes led to Erk activation via a PKC (protein kinase C)-δ and Ras-GRP (guanyl nucleotide-releasing protein)-1–dependent mechanism. An analogous mechanism may underlie the role of Ca²⁺ in 2B4-triggered Erk activation in NK cells.

Another unexpected finding was that, in mouse NK cells, the C-terminal tyrosines were needed not only to promote the activating effect of 2B4 but also to prevent the inhibitory impact of 2B4. Because the C-terminal tyrosines were not needed for EAT-2 to associate with 2B4, these findings implied that the signals mediated by the C-terminal tyrosines were responsible for prevention of inhibition. This is unlike the situation with SAP, in which the signaling motif (R78) was not needed for prevention of inhibition (Dong et al., 2012). Such a disparity suggests that SAP and EAT-2 prevent the inhibitory function of SLAM family receptors by distinct mechanisms. SAP may physically interfere with coupling of SLAM family receptors to inhibitory effectors through a natural blocking effect, as previously suggested (Sayos et al., 1998; Dong et al., 2012). In contrast, EAT-2 may mediate signals that inactivate or antagonize the inhibitory effectors of SLAM family receptors.

SAP stimulates NK cell activation by promoting conjugate formation between NK cells and target cells (Dong et al., 2012). We observed herein that EAT-2 did not augment conjugate formation. This was seen both in mouse NK cells and in YT-S cells. The lack of a stimulating effect of EAT-2 on conjugate formation was in agreement with the absence of effect on Vav-1, which likely mediates the effect of SAP on conjugate formation (Dong et al., 2012). Confocal microscopy studies provided indication that EAT-2 accelerated the effector phase of NK cell activation. Indeed, in YT-S cells, EAT-2 enhanced the proportions of conjugates having MTOCs and granules polarized toward the NK cell synapse. Fewer conjugates exhibited nonpolarized MTOCs or granules, or had a lack of polymerized actin at the NK cell synapse. This effect was absolutely dependent on Y127. It was also consistent with the impact of EAT-2 in mouse NK cells, in which WT EAT-2, but not EAT-2 Y120,127F, enhanced CD107a exposure. Moreover, experiments with pharmacological inhibitors suggested that it was mediated by the ability of EAT-2 to promote Ca²⁺ fluxes. However, an effect of the Erk pathway was possible, given that inhibition of Ca²⁺ fluxes also prevented Erk activation.

In summary, our data showed that EAT-2 promotes NK cell activation by molecular and cellular mechanisms distinct from those used by SAP. EAT-2 functions through Y127-dependent recruitment of PLCγ, which evokes Ca²⁺ fluxes and Erk activation. These signals enable faster polarization of MTOCs and cytotoxic granules toward the NK cell synapse. This ability of EAT-2 and SAP to couple SLAM family receptors to different signaling pathways and cellular effects likely explains their complementary roles in NK cell activation. It permits SLAM family receptors to mediate a broader range of signals and effects, and to be more efficient at activating NK cells.

Given that EAT-2 strengthens the activating function of SLAM family receptors in NK cells, it may be intriguing that other lymphocytes expressing SAP, in particular T cells, do not express EAT-2 (Roncagalli et al., 2005). Perhaps this distinction reflects the fact that, in T cells, activation is primarily driven by the antigen receptor and the co-receptors CD4 and CD8. In this context, SLAM family receptors may not need to mediate strong signals. In fact, the presence of EAT-2 could have deleterious consequences in T cells, perhaps by enabling antigen receptor-independent signals that were too strong. Nonetheless, it is worth pointing out that some T cells, including activated Vγ5⁺ intraepithelial lymphocytes, may express high levels of EAT-2 (www.immgen.org). The possibility that EAT-2 contributes to the innate-like characteristics of these T cells deserves consideration.

To generate a mouse strain in which Y120 and Y127 of EAT-2 were replaced by phenylalanines (Sh2d1b1^Y120,127F mouse), the targeting vector depicted in Fig. 3 A was created. In brief, this construct contained a 4.5-kb 5′ arm that replaced the codon for Y120 (TAT) by TTT, which codes for phenylalanine. This codon is located in exon 3. The construct also possessed a 3-kb 3′ arm that changed the codon for Y127 (TAT) to TTC, which also codes for phenylalanine. The codon for Y127 is in exon 4. In addition, a silent mutation creating an EcoRI site (GAATTC) for screening was introduced at this site. The construct also encompassed the neomycin resistance (neo) marker, which was flanked by frt sites. After linearization with NotI, the construct was transfected in C57BL/6-derived Bruce 4 embryonic stem cells. Cells were selected in the presence of G418 and clones showing homologous recombination were identified by Southern blotting. Clones containing both the Y120F and the Y127F mutations were identified by sequencing of PCR-generated fragments containing exons 3 and 4. They were then injected into blastocysts and chimeric mice were used for germ line transmission. The neo marker was eliminated by breeding mice with a transgenic mouse expressing the Flpe recombinase (B6.SJL-Tg(ACTFLPe)9205Dym/J; The Jackson Laboratory; Rodríguez et al., 2000). Mice were then screened by PCR, using oligonucleotide primers at the positions depicted in Fig. 3 A. C57BL/6 mice lacking EAT-2 or SAP were described previously (Al-Alem et al., 2005; Dong et al., 2012). SAP-deficient mice were provided by L. Yin (International Agency for Research on Cancer, Lyon, France). In all experiments, littermates were used as WT controls. WT C57BL/6 mice were obtained from The Jackson Laboratory. Animal experimentation was approved by the Animal Care Committee of IRCM and performed in accordance with the guidelines of the Canadian Council of Animal Care.

cDNAs coding for human EAT-2 (WT or Y127F), mouse EAT-2 (WT; Y120F; Y127F; Y120,127F; R54L; R54L,Y120F; and R54L,Y127F), FLAG-tagged mouse EAT-2, human CD48, and PLC-γ1 were generated by PCR and verified by sequencing. Those encoding mouse 2B4, mouse CRACC, various PTKs, and an SH2 domain–deleted variant of Fyn were reported previously (Cao et al., 1998; Chen et al., 2006; Cruz-Munoz et al., 2009). For expression in YT-S, K562, and HeLa, cDNAs were usually cloned in the retroviral vector pFB-GFP, which also encodes GFP. For expression in Cos-1 cells, cDNAs were cloned in the vector pXM319.

For real-time PCR analyses, cells were purified by cell sorting. In brief, NKPs (Lin⁻, CD122⁺, NK1.1⁻, and CD49b⁻), iNK cells (Lin⁻, CD122⁺, NK1.1⁺, and CD49b⁻), and mNK cells (Lin⁻, CD122⁺, NK1.1⁺, and CD49b⁺) were obtained from bone marrow of C57BL/6 mice using antibodies directed against CD19, Ter-119, B220, CD122, NK1.1, and CD49b, as described elsewhere (Ramirez et al., 2012). Splenic mNK cells were isolated by staining with antibodies against CD122, NK.1.1, and CD49b, whereas splenic B cells were obtained by staining with antibodies directed against B220 and CD19. In all cases, cell purity was >90%. Freshly isolated splenic NK cells from poly I:C–primed mice and spleen-derived LAK cells were generated as outlined elsewhere (Dong et al., 2009). RMA-S (lymphoma), YAC-1 (thymoma), and B16 (melanoma) expressing or not SLAM family ligands were described elsewhere (Dong et al., 2009, 2012). Control B16 cells expressed GFP alone. YT-S, K562, HeLa, and Cos-1 cells were reported elsewhere (Chen et al., 2006; Cruz-Munoz et al., 2009). YT-S cells expressing GFP alone or in combination with WT or Y127F human EAT-2 were generated by retroviral infection, using the retroviral vector pFB-GFP. Transduced cells were sorted for expression of GFP. K562 or HeLa derivatives expressing human CD48 or mouse CRACC were generated in a similar way. Cos-1 cell transfections were performed as outlined previously (Latour et al., 1997).

Antibodies recognizing CD19 (mAb 1D3), Ter-119 (mAb Ter119), B220 (mAb RA3-6B2), CD122 (mAb TM-b1), NK1.1 (mAb PK136), CD49b (mAb DX5), human 2B4 (mAb C1.7), human CD48 (mAb TÜ145), mouse 2B4 (mAb 2B4), mouse CD48 (mAb HM48-1), CD3 (mAb 145-2C11), CD11b (mAb M1/70), CD27 (LG.7F9), IFN-γ (mAb XMG1.2), CD107a (mAb eBio 1D4B), CD16 (mAb 2.4G2), Ly49D (mAb 4E5), Ly49H (mAb 3D10), and Ly49C/I/F/H (mAb 14B11), as well as isotype controls, were purchased from eBioscience or BD. Anti–phospho-Erk (pT202pY204; mAb E10), anti–phospho-Akt (pS473; mAb 193H12), and anti-Akt were obtained from Cell Signaling Technology, and antibodies directed against phosphotyrosine (mAb 4G10) or Erk were from Millipore. Anti-FLAG mAb M2 was obtained from Sigma-Aldrich. Antibodies against mouse EAT-2 (mAb 8F12), SAP (mAb 1A9), mouse Ly-9, mouse Ly108, mouse CD84, mouse SLAM, and mouse CRACC were produced in our laboratories (Roncagalli et al., 2005; Veillette et al., 2008; Zhong and Veillette, 2008; Cruz-Munoz et al., 2009). Monoclonal antibodies against human EAT-2 (mAb 10F7) were generated by immunizing mice with the full-length human EAT-2 protein (COVALAB). Polyclonal rabbit antibodies against human 2B4 were provided by E. Long (National Institutes of Health, Rockville, MD). Those against Fyn, Vav-1, SHIP-1, and c-Cbl were described elsewhere (Chen et al., 2006). Antibodies against PLCγ-1 and PLCγ-2 were purchased from Santa Cruz Biotechnology, Inc.

RNA was purified with the RNeasy micro kit (QIAGEN), and reverse transcribed using the SensiScript kit (QIAGEN) and oligo (dT) primers (Ambion). PCR reactions were performed using single-stranded cDNAs, gene-specific primers, and the PerfeCTa SYBR Green Supermix (Quanta), according to the manufacturer’s instructions. Real-time PCR was performed in duplicates, and fluorometric data were collected at the annealing step of each cycle, using the ViiA Real-Time PCR system (Life Technologies). A dissociation curve was obtained at the end of 40 cycles to confirm specificity of amplification. Expression is reported as Δ cross threshold (ΔCT), which is relative to Gapdh mRNA and normalized to values for LAK cells. The sense and antisense PCR primers were: Sh2d1a (SAP), 5′-ACAGAAACAGGTTCTTGGAGTG-3′ and 5′-GCATTCAGGCAGATATCAGAATC-3′; Sh2d1b1 (EAT-2), 5′-CAGGATAGAGACTAATGCTCATAC-3′ and 5′-CAGGATAGAGACTAATGCTCATAC-3′; Sh2d1b2 (ERT), 5′-AGGATAGAGACTGAGCCCAG-3′ and 5′-CTGGACAGAAGCCGCTTCCTC-3′; and Gapdh, 5′-AAATGGTGAAGGTCGGTGTG-3′ and 5′-GCTCTGGAAGATGGTGATG-3′.

Amino acid sequences of EAT-2 from different species were obtained by searching the public database PROTEIN at the NCBI website. Alignments were done using ClustalW2 online software from the EBI website and results were edited with BioEdit software v7.2.1, Ibis BioScience. Microarray data assembled by the Immunological Genome Project Consortium (ImmGen) were downloaded from their website (www.immgen.org). Data are shown as reported.

Poly I:C–activated ex vivo NK cells or IL-2–expanded NK cells were generated as outlined previously (Dong et al., 2009, 2012). NK cell–mediated cytotoxicity and IFN-γ production were assayed as detailed elsewhere (Dong et al., 2009, 2012).

Antibody-mediated stimulation of 2B4, CD16, or both on YT-S cells or mouse NK cells was performed as explained elsewhere (Chen et al., 2006; Dong et al., 2009, 2012). Cells were then processed for immunoprecipitation, immunoblotting, or Ca²⁺ fluxes, which have been previously described (Chen et al., 2006; Dong et al., 2009, 2012).

Immunoprecipitations and immunoblots were performed as reported (Veillette et al., 1988).

The Ca²⁺ chelator BAPTA-AM and the MEK1 inhibitor PD98059 were purchased from Merck. They were dissolved in DMSO. To establish the optimal concentration of the compounds for our experiments, YT-S cells were incubated for 30 min with increasing amounts of the compounds in culture medium. 2B4-triggered intracellular Ca²⁺ fluxes or Erk activation was then examined. The lower concentration needed for maximally inhibiting these responses, without compromising cell viability, was chosen. These concentrations were: BAPTA-AM, 20 µM; and PD98059, 100 µM. PD98059 failed to cause complete inhibition of Erk1/2, even at a concentration of 100 µM. In all experiments, compounds were added 30 min before the functional or biochemical assays. Cell viability was monitored by trypan blue exclusion or flow cytometry, using forward side channel (FSC) and side scatter channel (SSC) analysis.

Biotinylated peptides encompassing the 29–amino acid C-terminal sequence of human or mouse EAT-2, phosphorylated or not at Y120, Y127, or both, were synthesized by the W.M. Keck Facility (Yale University, New Haven, CT). Peptides were dissolved in DMSO at a concentration of 10 µg/µl and coupled to Neutravidin agarose beads (Thermo Fisher Scientific). Beads were then incubated with the indicated cell lysates for 90 min at 4°C. They were extensively washed and association with PLCγ was detected by immunoblotting. GST fusion proteins encompassing various domains were produced in bacteria, as detailed elsewhere (Roncagalli et al., 2005). They were then incubated with lysates of HeLa cells expressing cDNAs encoding FLAG-tagged mouse EAT-2 in the presence of SH2 domain-deleted Fyn, or with immobilized EAT-2 peptides. After several washes, binding was detected by immunoblotting.

YT-S cells were first labeled on ice by incubation with Alexa Fluor 647–conjugated anti–human CD335 mAb 9E2 (BioLegend), which recognizes NKp46. Although NKp46 is expressed on YT-S, it is not functional (our unpublished results). K562 cells were labeled in a similar way, using PE-conjugated anti–human CD71 mAb CY1G4 (BioLegend). After washing the unbound antibodies, cells were resuspended at a concentration of 2 × 10⁶ cells per ml and conjugate formation was analyzed as previously described (Dong et al., 2012).

K562 cells were first stained with the Cell Tracker Violet dye (Life Technologies), according to the manufacturer’s instructions. After washing, cells were allowed to recover by incubation in culture medium for at least 10 min. YT-S cells and K562 cells were then mixed at a 2:1 ratio in serum-free culture medium and incubated for 10 min at 37°C to allow conjugate formation. The suspension was subsequently transferred to poly-l-lysine–coated coverslips and incubated an additional 5 min at 37°C for attachment. Cells were then fixed and permeabilized in PBS containing 4% formaldehyde and 0.1% saponin. After that, they were washed twice with PBS containing 0.1% saponin and blocked for 30 min in PBS supplemented with 5% mouse serum. They were washed again and incubated for 1 h with biotinylated anti-tubulin mAb 236–10501 (Life Technologies). After further washing, cells were incubated for 1 h with Alexa Fluor 647–coupled anti-perforin mAb dG9 (BioLegend), Brilliant violet 605–coupled streptavidin (BioLegend), and Alexa Fluor 546–coupled phalloidin (Invitrogen). After an additional wash, coverslips were mounted over glass slides using fluorescent mounting medium (Dako). Data were acquired using a laser-scanning microscope LSM-710 (Carl Zeiss). Only conjugates in which one YT-S cell was coupled to one K562 cell were analyzed. At least 100 different conjugates were analyzed for each cell type or condition. Conjugate stages were defined as: 0, conjugates lacking actin polymerization, and MTOC and granule polarization; 1, conjugates with polymerized actin, but devoid of MTOC and granule polarization; 2, conjugates where MTOC and granules have partially migrated toward the synapse; and 3, conjugates where MTOC and granules have fully migrated toward the synapse.

To measure induction of CD107a expression at the cell surface, splenocytes (2 × 10⁶) from poly I:C–primed mice were cultured with the same number of RMA-S cells, GolgiStop (BD), and Alexa Fluor 488–conjugated anti-CD107a antibody. After 6 h, cells were harvested and analyzed by flow cytometry. NK cells were identified by gating on NK1.1⁺CD3⁻ cells. As control, cells were stimulated with a combination of PMA plus ionomycin.

For statistical analyses of cytotoxicity and IFN-γ production, paired Student’s t tests (two-tailed) were performed. For statistical comparison of the relative distribution of YT-S:K562 conjugates in the various stages of cytotoxicity, data for individual cell populations were compared using the Pearson correlation coefficient (r). All statistics were calculated using Prism 6 (GraphPad Software).

We thank the members of our laboratory for discussions. We also acknowledge Eric Long for the anti–human 2B4 serum and Luo Yin for the SAP-deficient mouse. We thank Kevin Ramirez and Barbara Kee for advice on purification of NK cell subsets.

This work was supported by grants from the Canadian Institutes of Health Research and the Canadian Cancer Society Research Institute (CCSRI) to A. Veillette. L.-A. Pérez-Quintero and H. Guo were recipients of Studentships from the Clinical Research Institute of Montreal and McGill University, respectively. R. Roncagalli was recipient of a Studentship from CCSRI. A. Veillette holds the Canada Research Chair in Signaling in the Immune System.

The authors declare no competing financial interests.