Direct and Indirect Interactions of the Cytoplasmic Region of CD244 (2B4) in Mice and Humans with FYN Kinase*

Engagement of the receptor CD244 (2B4) by its ligand CD48 has inhibitory and activating potential, and this differs depending on experimental systems in mouse and human. We show that, in both mouse and human upon engagement of its ligand CD48, CD244 can give a negative signal to natural killer cells, implying conservation of function between the two species. The signaling mechanisms used by CD244 in both human and mouse are conserved as shown by quantitative analyses of the direct molecular interactions of the SH2 domains of the adaptors SLAM-associated protein (SAP) and EAT-2 and of FYN kinase with CD244 together with the indirect interactions of the FYN SH2 domain with EAT-2. Functional experiments support the biochemical hierarchy of interactions and show that EAT-2 is not inhibitory per se. The data are consistent with a model in which the mechanism of signal transduction by CD244 is to regulate FYN kinase recruitment and/or activity and the outcome of CD48/CD244 interactions is determined by which other receptors are engaged.

Signaling by receptors that bind the intracellular adaptor SAP 2 is important for immune regulation in combating infection, and it is dysregulated in cancer and autoimmunity (1,4); thus, an understanding of the mechanisms involved has wide implications. CD244 (also known as 2B4) is one of the group of leukocyte surface receptors, including CD244, NTB-A (Ly-108), CD319 (CRACC, CS-1), CD150 (SLAM), CD84, and CD229 (Ly-9), that are characterized by related extracellular immunoglobulin superfamily domains and intracellular immunoreceptor tyrosine-based switch motifs (ITSMs) that can bind the SH2 adaptor protein SAP (see Fig. 1a) (3,5). These receptors are differentially distributed on immune cells and are involved in cell/cell interactions through heterophilic or homophilic interactions (2,3,6,8). The CD48/CD244 interaction is the only well characterized heterophilic interaction among the receptors that contain SAP-binding motifs in their cytoplasmic regions (3). CD244, primarily characterized as a receptor on natural killer (NK) and activated CD8 ϩ cells (6,9,10), binds the glycophosphoinositol-anchored protein CD48 with similar affinity in mouse and human (11). Rodent CD48 also retains a low but functionally significant affinity for CD2 (see Fig. 1a) (5,11). The ligand for human CD2 (CD58) is missing in rodents.
There has been confusion in the literature over the inhibitory and activating potential of CD244 (2B4) and over differences in behavior between mouse and human CD244. A study in a CD244-deficient mouse showed clearly that, when CD48 on a target cell is not able to bind to CD244 on an NK cell, killing is enhanced (12). These results assisted with interpreting data with monoclonal antibodies (mAbs) and distinguishing between blocking and cross-linking effects (12) and also suggested that the phenotype of the CD244-deficient mouse reflects loss of ligand engagement. Thus, binding of CD48 on the target cells to CD244 on NK cells has an inhibitory effect on NK cell killing. However, engagement of CD244 can also have positive effects in mice (10,13). Likewise, in human, the functional outcome of CD48/CD244 interactions can be positive and negative (14)(15)(16)(17). A model of CD244 signal transduction has to account for differences in functional outcome.
The altered behavior of NK cells deficient in CD244 (12), FYN (18), SAP (16,18,19), or EAT-2 and/or ERT (19) has implicated CD244 in being a major player in regulation among the SAP-binding receptors. The CD244 cytoplasmic tail contains the highest number (four) of the particular tyrosine motif among the SAP-binding receptors that, when phosphorylated upon cell activation, can bind SAP and/or EAT-2 (see Fig. 1) (3,19,21). These four motifs are present in mouse and human CD244 but are not identical in sequence. SAP can recruit FYN through an unusual interaction with the SH3 domain of FYN (22,23). EAT-2 and ERT are structurally related to SAP, but their tail regions contain one or two tyrosine motifs in human and mouse, respectively (19,24). ERT is not expressed in human (19,24). Unlike the SH2 domain of SAP, that of EAT-2 does not bind the SH3 domain of FYN (22,23). In mice, SAP and EAT-2 or ERT deficiencies have opposite phenotypes, with lack of SAP being associated with decreased responses and vice versa for EAT-2 or ERT (19). In mice, the inhibitory effects of EAT-2 and ERT have been shown to be dependent on phosphorylation of the tail tyrosine motifs (19). EAT-2 and ERT are expressed in NK cells (19,21,24), and CD244 is a major binding partner for EAT-2 and ERT (19), suggesting that the inhibitory function of CD244 is dependent on these interactions. Models for the mechanism of the inhibitory effects of SAP-binding receptors, particularly CD244, include direct binding of phos-phatases competing with SAP and recruitment of negative regulators through EAT-2 (3,25,26).
Conservation of extracellular and intracellular interactions with CD48 and the adaptors SAP and EAT-2 and expression in cytotoxic cells in both mouse and human (6,9,10) suggest that the function of CD244 is conserved between the two species. We sought to identify the features of CD244 function that are conserved between mouse and human with the aim of establishing the fundamental mechanism by which CD244 regulates immune responses. An understanding of the differences between mouse and human CD244 will allow interpretation of the relevance of mouse model data to human. There is evidence that manipulating CD48/CD244 interactions may be useful therapeutically, as CD244-deficient mice are more effective in tumor clearance (12,27). We show that, in both mice and humans, engagement of mouse and human CD244 on normal NK cells by CD48 on target cells can be inhibitory. By determining the hierarchy of interactions likely to compete for direct binding to CD244 at physiological temperature, we show that intracellular interactions of CD244 are generally conserved between mouse and human CD244. We identify FYN kinase as a potential binding partner for the tail of EAT-2 and provide functional data to show that EAT-2 is not inhibitory per se. This suggests a model in which the functional outcome of CD48/ CD244 engagement is a reflection of the potential to recruit FYN kinase and the balance between co-engaged inhibitory and activating receptors. The general mechanism is applicable to the other SAP-binding receptors.

EXPERIMENTAL PROCEDURES
Recombinant Proteins-The rat basophil leukemia cell line was stably transfected with the pBabe vector encoding a chimeric receptor containing the extracellular region of the C57/ BL6 allele of mouse CD244 (11) and the transmembrane and cytoplasmic regions of the mouse -chain, CD244-(28), using FuGENE TM transfection reagent. Full-length mouse CD244 (11) in pBabe was expressed in 2B4 and 171 mouse T cell hybridoma cells by transfection of ecoPhoenix cells and transduction. Cells expressing 2B4 were selected at 48 h by Dynal bead selection using anti-2B4 mAb (Pharmingen), and selection was maintained with puromycin (1 g/ml) in RPMI 1640 medium containing 5-10% fetal calf serum. Constructs of full-length mouse SAP and the SH2 domain (amino acids 2-104; provided by Andre Veillette), full-length human SAP and the SH2 domain (amino acids 1-108; provided by Cox Terhorst), fulllength mouse EAT-2 and the SH2 domain (amino acids 1-103; provided by Andre Veillette), full-length human EAT-2 and the SH2 domain (amino acids 1-108; provided by Marco Colonna), and the human FYN SH2 (amino acids 144 -248) and SH3 (amino acids 82-149) domains (both provided by Rose Zamoyska, National Institute for Medical Research, London, UK) for production of soluble recombinant proteins were produced by PCR from templates. Fragments were cut with BamHI and SalI and cloned into BamHI and XhoI sites in the pTrcHISA vector (Invitrogen). The vector stop codon was used for the isolated domain constructs except for the mouse SAP SH2 domain, in which it was engineered directly after the final SAP amino acid. Constructs for His-tagged mouse FYN SH3-SH2 (amino acids 82-248) and human FYN SH3-SH2 (amino acids 82-248) proteins were provided by Louise Bird (Oxford Module Consortium). Proteins were expressed as N-terminally His-tagged fusion proteins and purified using nickel-agarose affinity chromatography. Proteins were subjected to gel filtration on Superdex 75 (GE Healthcare) prior to Biacore analysis. Concentration was determined using absorbance at 280 nm and the following theoretical extinction coefficients (Vector NTI): mouse SAP and SH2, 23 For mammalian expression as enhanced green fluorescence protein (EGFP) fusion proteins, BamHI-HindIII fragments from pTrcHISA constructs were cloned into the retroviral vector pLEGFP-C1 (Invitrogen) and expressed in 2B4 hybridoma cells as described for the pBabe vector, and selection was maintained with G418 (0.4 mg/ml).
Biacore TM Analyses-Biacore analyses using a Biacore TM 2000 were carried out essentially as described previously (11). Analyses of CD244 mAb blocking were carried out at 25°C (11), and all other experiments were carried out at 37°C. Streptavidin (ϳ3000 response units (RU)) and mAbs (ϳ15,000 RU) were directly coupled to CM5 chips by amine coupling for immobilization of biotinylated phosphorylated peptides (Peptide Protein Research Ltd. (Fareham, UK) and Sigma) (see Tables 1 and  2) and native receptors, respectively. The FYN SH3 domain (ϳ1250 -2000 RU) was immobilized by amine coupling. In measurements of binding to the four ITSM peptides, one flow cell was used as a negative control and then coated with the fourth peptide, and equilibrium binding was repeated. At the level of immobilization of peptide (ϳ25 RU), there was no difference between negative controls of streptavidin or an irrelevant phosphorylated peptide. Native receptor was captured from cell lysate (10 8 cell eq/ml) by passing the lysate at 1 l/min over a mAb-coated flow cell. Cell lysates were prepared from lymphokine-activated killer (LAK) cells that had been treated with pervanadate or not at 10 8 cells/ml for 7-10 min at 37°C; washed with phosphate-buffered saline; and lysed at 10 8 cells/ml in 10 mM Tris-HCl (pH 7.4) supplemented with 140 mM NaCl, 1 mM EDTA, 10% (v/v) glycerol, 0.02% (w/v) NaN 3 , 1% Brij 96, 1 mM sodium pervanadate, mammalian protease inhibitor mixture (Sigma), and 1 mM NaF for ϳ10 min at 4°C. The lysate was centrifuged at 20,800 ϫ g in a microcentrifuge for 10 min at 4°C. Lysates were used immediately in Biacore experiments.
Antibodies-AO rats (haplotype RT1 u ; Harlan UK) were initially immunized intravenously with 4 ϫ 10 6 irradiated rat basophil leukemia cells (haplotype RT1 u ) expressing the CD244chimera, as it was more highly expressed than native CD244. The rats were given three additional immunizations, once with complete and twice with incomplete Freund's adjuvant of 20 g of CD244-CD4d3ϩ4 soluble fusion protein purified from Chinese hamster ovary (CHO) cells stably transfected with a CD244-CD4d3ϩ4-PEE14 construct (11) subcutaneously. Rats were boosted three times intravenously with 4 ϫ 10 6 rat basophil leukemia cells expressing CD244-and 20 g of CD244-CD4d3ϩ4 soluble fusion protein. 3 days after the final boost, the spleen was removed, and a single cell suspension was prepared and fused with the Lewis rat myeloma cell line Y3 AG8.153. Three hybridomas producing mAb that stained CD244/171 cell lines but not the 171 untransfected cell line were isolated and subcloned twice by limiting dilution. The mAbs were of the IgG1-(OX120) and IgG2a-(OX121 and OX122) isotypes as determined by enzyme-linked immunosorbent assay (Pharmingen). mAbs OX121 and OX122 were purified on protein G-Sepharose (GE Healthcare). Fab fragments were prepared by papain digestion and gel filtration on a Sephadex HR 200 column (GE Healthcare) in phosphate-buffered saline. C57/BL6 and BALB/c mice obtained from Sir William Dunn School of Pathology were used for flow cytometric analysis of normal CD244.
Cellular Assays-Murine NK cells were enriched by passing a splenic cell suspension from C57/BL6 mice through a nylon wool column for negative depletion of adherent cells. LAK cells were generated by culturing the passaged cells in RPMI 1640 medium with L-glutamine supplemented with 10% fetal calf serum, 100 units/ml penicillin, 100 units/ml streptomycin, and 50 M 2-mercaptoethanol (complete RPMI medium) and containing 0.5% spent tissue culture supernatant from a rat interleukin-2 (IL-2)-producing CHO line at 37°C and 5% CO 2. On day 3, half of the medium was replaced with fresh medium. Murine LAK cells were used on day 6. Human NK cells were purified from non-adherent peripheral blood cells (National Blood Service, Bristol, UK) by depletion with anti-human CD3, CD4, and CD8 mAbs and cultured for ϳ12 days with an irradiated Epstein-Barr virus-positive B cell line (221.721) and irra-diated allogeneic peripheral blood cells. 2 ϫ 10 4 RMA-S or 221.721 cells, target and effector NK cells (LAK cells), or human NK cells at the indicated effector/target ratios were incubated in complete RPMI medium in 96-well plates. mAb or Fab fragments were added to the wells at a final concentration of 10 g/ml. After 4 -5 h of incubation at 37°C, the supernatants were removed, and the amount of lysis was measured using a nonradioactive cytotoxicity assay (CytoTox 96, Promega Corp.). Antigen-specific IL-2 production of hybridoma cells (10 5 cells/well) using mouse CD48 ϩ and CD48 Ϫ , major histocompatibility complex Class II ϩ CHO cells (30) (10 5 cells/well) with a final concentration of 0.1 M moth cytochrome c peptide was performed as described (29). In all cellular experiments, triplicate samples were assayed, and the mean is displayed with S.E. Each experiment was performed at least four times.

CD244 Can Be Inhibitory in Killing by Murine NK Cells-To
distinguish between the effects of CD244 and CD2 engagement by CD48 in rodents, we produced an anti-CD244 mAb that was more effective at blocking CD48/CD244 interactions than that characterized previously (11). Of three mAbs that were specific for mouse CD244 and bound to a mouse T cell hybridoma transduced with mouse CD244 (Fig. 1b) (data not shown), one (mAb OX122) completely blocked binding of CD48-CD4d3ϩ4-coated fluorescent beads to cells (30) expressing CD244 (Fig. 1c). mAb OX121 partially blocked binding of CD48-CD4d3ϩ4-coated beads, similar to a mAb used in our previous study (11). Biacore analysis with purified proteins confirmed the specificity of the mAbs for CD244 as described (data not shown) (11). mAbs OX122 and OX121 blocked binding of soluble CD48-His by 90 and 20%, respectively. mAb OX122 was noncompetitive with OX121. mAbs OX121 and OX120 were competitive with each other and with a mAb also shown to partially block in a similar experiment (11). Fab fragments from mAb OX122 were as effective as whole IgG in blocking CD48/ CD244 interactions.
mAbs OX122 and OX121 were tested for binding to CD244 on normal mouse cells. Both mAbs gave identical staining with a distribution consistent with preferential expression on NK cells and a low percentage of T cells as expected for an anti-CD244 mAb (data not shown) (10). Similar to the allelic mAb mentioned above (31), mAb OX122 and OX121 staining was restricted to the C57/BL6 allele used for immunization and did not stain cells from BALB/c mice (data not shown).
The new anti-CD244 mAb was used to distinguish between the roles of CD48/CD244 and CD48/CD2 interactions in an NK cell killing assay. LAK cells were prepared as effector cells from C57/BL6 mice. Approximately 10 -20% of the LAK cell preparation expressed CD244, and all the cells expressed CD2 and CD48 (Fig. 1d) (data not shown). The RMA-S cell line, which was used as a source of target cells, expressed CD48 and low levels of CD2 and was negative for CD244 (Fig. 1e).
Incubation of LAK cells with RMA-S cells in the presence of excess soluble anti-CD244 mAb or mAb OX122 or OX121 resulted in enhanced killing (Fig. 1f). OX122 was consistently more effective than OX121 at blocking CD48/CD244 interactions. LAK cell killing was enhanced by non-cross-linking Fab fragments from OX122 (Fig. 1g), supporting the conclusion that the mechanism of action was blocking CD48/CD244 interactions. In contrast to the effects of the anti-CD244 mAb, an anti-CD2 mAb previously shown to be effective in blocking CD48/CD2 interactions (30) reduced killing (Fig. 1f), and the opposing effects of blocking both interactions by the anti-CD48 mAb appeared to cancel each other out (Fig. 1f).
CD244 Can Be Inhibitory in Killing by Human NK Cells-We tested the effects of blocking CD48/CD244 interactions on killing CD48 ϩ human target cells by activated CD244 ϩ NK cells derived from human peripheral blood (Fig. 2). The majority of the NK cell preparation expressed CD244, CD2 (Fig. 2a), and CD56 and were CD3-negative (data not shown). Disruption of CD48/CD244 interactions with a blocking anti-CD244 or anti-CD48 mAb resulted in increased killing, consistent with human CD244 having an inhibitory effect (Fig. 2b). There was no evidence that blocking CD48 had an effect other than on an interaction with CD244 in this assay (Fig. 2b), confirming the conclusion that a barely detectable interaction between human CD48 and CD2 is not physiologically significant (32).
The Adaptors SAP and EAT-2 and the Protein-tyrosine Kinase FYN Bind Directly to Mouse and Human CD244-To understand the conserved features of the signaling mechanism used by CD244 and to establish a hierarchy of which interactions are likely to occur in the cell, we measured binding of intracellular proteins reported to associate with the cytoplasmic region of CD244 to phosphorylated peptides representing the four ITSMs in mouse and human CD244. The affinity of SAP for each of the ITSMs in CD244 is an important factor for evaluating a model of competitive binding with phosphatases or other SH2 domain-containing proteins, including EAT-2, as an explanation for inhibitory signaling by CD244. Phosphorylated peptides representing each of the four SAP-binding motifs in mouse and human CD244 were immobilized on a Biacore chip, and a series of concentrations of monomeric recombinant SH2 and SH3-SH2 domains were injected over them at 37°C. Examples of data are shown for the mouse SAP SH2 domain to demonstrate that binding reached equilibrium rapidly and dissociated completely without evidence of a complicating contribution from aggregates that bind and dissociate more slowly (Fig. 3a). Equilibrium binding data were obtained by subtracting responses in control flow cells and plotted to determine the K D (Fig. 3b) . Affinity data are summarized in Table 1. In both species, the affinity of binding of SAP and EAT-2 SH2 domains (K D ϭ 0.1-1.75 M) suggests that interactions with all four motifs could be physiologically relevant, as they are in the range of affinities for other functional SH2 domain interactions (Fig. 3, a and b, and Table 1) (29). The mouse EAT-2 SH2 domain tended to aggregate, and binding data are shown for full-length mouse EAT-2. Comparisons between full-length SAP and EAT-2 and the isolated SH2 domains for binding to CD244 peptides revealed differences in affinities, generally within the range of 1-5-fold weaker for the full-length proteins, which could be due to differences in the activity of the proteins or could have functional significance ( Table 1). There was a significantly lower affinity of human SAP compared with the SAP SH2 domain or EAT-2 for the third motif. Binding to the C-terminal fourth motif of CD244 tended to have a lower affinity than the other three, consistent with this motif having another function (33). However, the affinity of human EAT-2 for the distal motif (K D ϭ 0.5 M) was comparable with a functionally relevant SH2 domain interaction (Table 1) (29).
As FYN kinase is recruited to CD244 via SAP, it is implicated in regulation by CD244 and other SAP-binding receptors (18). We tested whether FYN can bind directly to CD244, as this has been suggested (34). We tested the FYN SH2 domain for species-specific binding to both mouse and human CD244. As the construct for the mouse FYN SH2 domain did not readily produce monomeric protein, a protein containing mouse FYN SH3 and SH2 domains was used. These recombinant human and mouse FYN proteins bound the phosphorylated membraneproximal motif of CD244 with an affinity that could be functionally significant (Table 1 and see Fig. 4b). No detectable binding of the SH2 domain of another kinase, human CSK (C-terminal SRC kinase), to human phosphorylated CD244 peptides was observed (data not shown).
Binding of SAP to Synthetic Peptides Is Comparable with Binding to Native Phosphorylated CD244-We have shown previously for CD6 that binding of signaling proteins to short peptides and native cytoplasmic regions is comparable (29). To confirm this for CD244, we analyzed direct binding of the purified mouse SAP SH2 domain to native mouse CD244 (Fig. 3, c-f). Building on techniques pioneered with CD6 (29), CD244 was purified on a Biacore chip using mAb OX121 directly from lysates from LAK cells treated or not with pervanadate. The SAP SH2 domain was injected over phosphorylated and unphosphorylated CD244 and a control flow cell at 37°C (Fig.  3c). Analysis of the binding data revealed saturable binding to  . Surface plasmon resonance analysis shows that the mouse SAP SH2 domain binds directly to mouse CD244 (Table 1). a and b, representative equilibrium binding of the mouse SAP SH2 domain to immobilized mouse CD244 peptides (Table 1) AUGUST 31, 2007 • VOLUME 282 • NUMBER 35 phosphorylated CD244, with K D ϭ 0.3 M (Fig. 3d) and, in another independent experiment, K D ϭ 0.6 M (data not shown). The SAP SH2 domain did not bind in a concentrationdependent manner to unphosphorylated CD244 (Fig. 3d). The levels of immobilized unphosphorylated CD244 were higher than those of phosphorylated CD244 (Fig. 3e), which emphasizes the specificity of SAP for phosphorylated CD244 (Fig. 3f). Western blot analysis of CD244 immunoprecipitates from pervanadate-treated LAK cell lysate with antiphosphotyrosine mAb followed by mAb OX122 revealed only the phosphorylated long major form of mouse CD244 (data not shown) (35). (25,34,36) and peptide binding (37) experiments have indicated a direct interaction of phosphatases with human CD244 and other SAP-binding receptors, suggesting that the inhibitory effects of these receptors are due to a switch from SAP or EAT-2 to a phosphatase (3,20,25). To assess whether the SH2 domains of phosphatases would compete with SAP and EAT-2 effectively, we measured binding of the mouse and human SHIP (SH2 domain-containing inositol 5-phosphatase) SH2 domains, which are most similar to the SAP and EAT-2 SH2 domains (38), and of human SHP-2 to the phosphorylated CD244 peptides by surface plasmon resonance analysis as described for Fig. 3. Species-specific binding was weaker by about an order of magnitude or greater than binding to a phos-phorylated peptide from Fc receptor-␥IIb, which we used as a control for the activity of the recombinant SHIP, SHP-2, and SHP-1 SH2 domains, as the binding specificities of their SH2 domains overlap (39). It has been suggested that phosphatases bind the third motif in human CD244 (34), but although the human SHIP SH2 domain had a marginal preference for the third motif in human CD244, this was still 6-fold weaker (K D ϭ 3 M at 37°C) than binding to an Fc receptor-␥IIb peptide (K D ϭ 0.5 M at 37°C) (29) in the same experiment. Binding of the mouse SHIP SH2 domain to the third motif of mouse CD244 was similar (K D ϭ 7 M at 37°C). Binding of human SHP-2 to the CD244 peptides was at least 10-fold weaker compared with binding to the same Fc receptor-␥IIb peptide (K D ϭ 1 M at 37°C) in the same experiment. In additional experiments, human SHP-1 bound even more weakly than human SHP-2 to human CD244 peptides. We detected no binding of the fulllength human tyrosine phosphatases to native phosphorylated mouse CD244 (data not shown); thus, there was no evidence that cooperation between two SH2 domains would effect a functional interaction with CD244. In affinity measurements, phosphotyrosine was still present on peptides (data not shown) and native CD244 (Fig. 3e) after injection of full-length tyrosine phosphatases. Repeated injection of high concentrations of the phosphotyrosine phosphatases eventually dephosphorylated the peptides as revealed with anti-phosphotyrosine mAb.

The Affinities of the Phosphatase SHIP, SHP-1, and SHP-2 SH2 Domains for Mouse or Human CD244 Are Not in the Range Expected for Direct Competitive Binding with SAP and EAT-2-Coprecipitation
The FYN SH2 Domain Binds to Phosphorylated Tyrosine Motifs in the Tail of EAT-2-Phosphorylation of the tyrosines in the tail of mouse EAT-2 has been shown to be critical for the

EAT-2, and FYN bind phosphorylated CD244 peptides with affinities consistent with a direct functional interaction
A summary of K D values (M, 37°C, S.E. 5-20%) is provided for binding of monomeric recombinant proteins measured as described in the legend to Fig. 3. Interactions are species-specific, i.e. human protein on human peptides and murine protein on murine peptides. ITSM1 is membrane-proximal, and ITSM4 is distal. The sequence of each peptide is given with the phosphorylated Tyr in boldface. m, mouse; h, human. Peptide % purity is given in parentheses. For each protein, K D values were measured for all four peptides in the same experiment. Variability between experiments was Ͻ20%.  inhibitory effects of EAT-2 (19). The C-terminal tyrosine motif is conserved between human and mouse, and there is a second more proximal tyrosine motif in mouse (Table 2). An obvious question is whether these motifs recruit negative regulators directly or indirectly. The surface plasmon resonance experiments described above showed that direct interactions of the phosphatases were too weak to compete for EAT-2. Thus, further experiments tested whether indirect interactions occurred, i.e. did the phosphatases bind EAT-2? The affinity of the mouse and human SHIP SH2 domains (K D Ͼ 50 M at 37°C) and of human SHP-2 (SH2 domains are 100% conserved between human and mouse; K D ϭ 40 M at 37°C) for the mouse EAT-2 phosphorylated Tyr 128 peptide containing the tyrosine motif that is conserved between human and mouse ( Table 2) was very low. In addition, none of these proteins bound the mouse EAT-2 phosphorylated Tyr 119 peptide (K D Ͼ 50 M at 37°C). Regulation of FYN kinase activity has already been implicated in CD244 function through SAP and direct binding of the FYN SH2 domain (Table 1). A positive effect of EAT-2 on phosphorylation of CD244 has been observed (20,21), consistent with EAT-2 also recruiting a SRC kinase (20,21). We tested the FYN SH2 domain for binding to the EAT-2 tail peptides. The FYN SH3-SH2 domain bound to the phosphorylated tyrosine motifs of EAT-2 in both mouse and human (K D ϳ1 M at 37°C) (Fig. 4, a and b, and Table 2), indicative of a direct functional interaction (Table 1) (29). The human FYN SH2 domain did not bind the proximal motif in human EAT-2 containing Phe instead of Tyr compared with binding the conserved phosphorylated tyrosine motif, Tyr 128 ( Table 2). The lower affinity of the human FYN SH2 domain compared with the SH3-SH2 protein for the EAT-2 tail (K D ϭ 3 and 1 M, respectively) may be due to protein activity, as there was a comparable difference in binding human CD244 ITSM1 (K D ϭ 0.6 and 0.1 M, respectively) ( Fig.  4a and Tables 1 and 2).
The FYN SH2 Domain Binds to EAT-2 with Higher Affinity than SAP Binds the FYN SH3 Domain-To compare a potential interaction between the FYN SH2 domain and EAT-2 with that between the FYN SH3 domain and SAP, we measured binding of the human SAP SH2 domain to the immobilized FYN SH3 domain at 37°C ( Table 2). The affinity of the human SAP SH2 domain for the FYN SH3 domain ( Table 2) was ϳ2-4-fold weaker than that measured for the human SAP SH2 domain (22) or full-length human SAP (37) for the FYN SH3 domain at 25°C. As discussed above, the weaker binding by full-length SAP may be due to protein activity or may be of functional significance. The mouse SAP SH2 domain also bound the immobilized FYN SH3 domain (100% conserved between human and mouse; K D ϭ 3 M at 25°C) (data not shown). Thus, the interaction between SAP and the FYN SH3 domain is conserved between human and mouse.  . c-f, complexes containing EAT-2 and FYN were isolated from transduced murine 2B4 hybridoma cells (c and d) and human NK cells derived from normal peripheral blood (e and f). c, mouse EAT-2-EGFP was detected with anti-phospho-EAT-2 (pEAT-2) antibody in lysates (1.6 ϫ 10 6 cell eq) from cells treated with or without pervanadate (pV) (lanes 1 and 2) and nized a band the correct size for EAT-2-EGFP (Fig. 4c, lanes 1  and 2) that was not present in cells expressing SAP-EGFP (lane 3). In cells treated with pervanadate, immunoprecipitated EAT-2-EGFP reacted with anti-phosphotyrosine mAb, showing that it could be phosphorylated (Fig. 4c, lane 4). Probing an equivalent blot of immunoprecipitated EAT-2-EGFP with anti-FYN antibody showed that a complex containing phosphorylated EAT-2 and FYN could be formed in cells (Fig. 4d). Detection of phosphorylated EAT-2 in pervanadate-treated human NK cells (Fig. 4e) showed that human EAT-2 has the potential to be phosphorylated (40). Probing a blot containing immunoprecipitates of EAT-2 with anti-FYN antibody revealed the presence of complexes containing FYN and phosphorylated EAT-2 in pervanadate-treated human NK cells. Comparison of detection of complexes between EAT-2 and FYN showed that antibody raised against a phosphorylated EAT-2 peptide preferentially precipitated phosphorylated EAT-2 (Fig. 4, e and f, lane 1). Immunoprecipitation of FYN as a complex with EAT-2 (Fig. 4f, lanes 1, 5, and 6) was more efficient than with an antibody against the N-terminal region of FYN (lanes 3 and 4). The presence of FYN in the immunoprecipitate with antibody preferentially recognizing the phosphorylated form of EAT-2 (Fig. 4f, lane 1) implies a higher order complex, as the EAT-2 antibody would block the specific association of the tail with FYN. In agreement with this interpretation, anti-EAT-2 antibody raised against the unphosphorylated EAT-2 tail peptide was equally efficient in precipitating phosphorylated and unphosphorylated EAT-2 (Fig. 4f, lanes 5 and 6).
EAT-2 Induces Increased IL-2 Production in T Hybridoma Cells-The biochemical data suggest that signaling through EAT-2 in mouse and human is regulated through the level of recruitment of FYN kinase and that this will be amplified in the presence of phosphorylated EAT-2, which has a higher affinity for FYN compared with SAP. We compared the effects of EAT-2 and SAP in an antigen-specific T cell hybridoma model. Mouse EAT-2 and SAP fusion proteins or EGFP alone were expressed together with mouse CD244 in murine T hybridoma cells (Fig. 5a) and stimulated with antigen and CHO cells expressing major histocompatibility complex Class II and mouse CD48. Antigen-specific IL-2 production was increased with cells expressing EAT-2 relative to cells expressing SAP or EGFP. The positive effect of EAT-2 in these cells was enhanced by CD48 on the antigen-presenting cells, but was observed in the absence of CD48 expression on the antigen-presenting cells (Fig. 5b). In hybridoma cells expressing transduced mouse CD2 and EAT-2-EGFP, IL-2 production was also increased (data not shown). 2B4 hybridoma cells express at least one other SAPbinding receptor, CD150 (SLAM) (data not shown), that would contribute to recruitment of EAT-2. The blocking experiments do not allow the contribution of CD244 and CD2 in terms of adhesion and/or signaling to the increase in IL-2 production in the presence of EAT-2 to be elucidated. The main conclusion from these experiments is that EAT-2 is not necessarily intrinsically inhibitory, and the results are consistent with EAT-2 recruiting FYN kinase.

Inhibition by CD48/CD244 Interactions Does Not Correlate with EAT-2 Expression in T Hybridoma
Cells-We compared the dependence of an inhibitory outcome of CD48/CD244 interactions on expression of EAT-2 in the hybridoma model. In the presence of the blocking anti-CD244 mAb and antigenpresenting cells expressing CD48, IL-2 production was increased in cells expressing SAP or EGFP, but not significantly in cells expressing high levels of EAT-2 (Fig. 5b). Thus, inhibition by CD48/CD244 engagement does not strictly correlate with levels of EAT-2. Inhibition by the anti-CD244 mAb was specific as shown by the neutral effect of anti-CD48 and anti-CD2 mAbs. IL-2 production by cells expressing EAT-2, SAP, or EGFP in the presence of CHO cells lacking CD48 was insensitive to blocking mAbs (Fig. 5b). CD48 expression on the antigen-presenting cells may be important to achieve a level of activation where CD244 is functional.

DISCUSSION
An understanding of engagement at the cell surface and the mechanism of signal transduction by CD244 and its relationship to CD2 in mice and humans is necessary to allow interpretation of the relevance of mouse model data to human. An inhibitory outcome of engagement of human and mouse CD244 by CD48 in normal NK cell killing implies that the function of CD244 is conserved. Blocking CD48 and CD244 in humans appears to be equivalent, whereas dissection of the effects of blocking CD48 in rodents is complex and depends on whether the effects are due to blocking CD244 and CD2 on the same or different cells.
Dissection of the pattern of binding by intracellular proteins to the tyrosine motifs in the cytoplasmic regions of mouse and human CD244 showed that intracellular interactions are conserved. The hierarchy of affinities showed that SAP, EAT-2, and the FYN kinase SH2 domain could compete for direct binding to CD244. The largest differential was between full-length EAT-2 and SAP binding to the third ITSM of human CD244, which may be of functional significance. Weak binding of SAP may be influenced by phosphorylation of an adjacent N-terminal tyrosine residue. The affinities measured at 37°C of the SAP and EAT-2 SH2 domains for phosphorylated CD244 are compatible with values determined for binding to a human phosphorylated CD150 peptide (K D ϭ 0.1-0.3 M at 25°C) (20,37,41,42). The similarity between affinity measurements for peptides and native CD244 purified directly from LAK cells shows that affinity values obtained with synthetic peptides provide a good representation of the hierarchy of interactions in vivo.
The peptide binding data, together with no detectable binding by the full-length tyrosine phosphatases to native CD244, make it unlikely that direct phosphatase recruitment underlies the inhibitory effects of CD244. The affinities of phosphatase binding directly to CD244 were below SAP, EAT-2, and FYN in a hierarchy of interactions likely to occur in the cell. Even the strongest interaction measured between the SHIP SH2 domain and CD244, which is ϳ6-fold weaker than a functionally relevant interaction between the SHIP SH2 domain and Fc receptor-␥IIb (K D ϭ 0.5 M at 37°C), would not effectively compete with SAP and EAT-2. Competition would depend on the relative concentrations of competing components. The levels of SAP but not SHP-2 protein in human NK cells are increased upon activation (43). Moreover, CD244 maintains its inhibitory effects in SHIP-deficient cells (44).
Indirect binding of phosphatases to CD244 through the phosphorylated tyrosine motifs in the tail of EAT-2 does not seem a likely mechanism for the inhibitory effects of CD244 based on the low affinities (K D Ͼ 50 M at 37°C) for phosphorylated EAT-2 tail peptides. An interaction between the phosphorylated tyrosine residues in the tail of EAT-2 and the FYN SH2 domain and the increased IL-2 production in EAT-2-expressing hybridoma T cells suggest instead that EAT-2 is involved in FYN recruitment. We were able to detect complexes containing EAT-2 and FYN in artificially activated cells using immunoprecipitating reagents with a different specificity to those tested previously (21). We have also shown, as has been stated previously (40), that human EAT-2 has the potential to be phosphorylated. An interaction between EAT-2 and SRC family kinase domains, including FYN, has been demonstrated in a yeast two-hybrid assay (24). The failure to demonstrate an interaction between the human FYN SH2 domain and mouse EAT-2 in these studies may reflect inefficient phosphorylation of the EAT-2 tail in this assay (24) or a weaker cross-species affinity. Cross-species binding of the mouse FYN SH3-SH2 domain to the human EAT-2 tail phosphorylated Tyr 128 peptide was 5-fold weaker than binding to the equivalent mouse peptide. 3 The potential of EAT-2 to lead to increased IL-2 production may have relevance to X-linked proliferative disease. Phosphorylated CD244 has the potential to bind SAP and EAT-2 at the same time, and there are data that are compatible with this hypothesis (21). A correct balance between SAP and EAT-2 is important to develop effector function, as cytotoxicity in SAPdeficient patients is impaired, and CD244 is not correctly localized (45). Juxtaposition of SAP and EAT-2 might cooperatively facilitate and control the correct level of recruitment of FYN through its SH3 and SH2 domains. There is evidence that the FYN SH2 domain contributes to the SAP/FYN SH3 domain interaction (37). Defects in SAP expression may allow excessive recruitment of FYN by EAT-2, leading to the uncontrolled proliferation in cells expressing the Epstein-Barr virus proteins LMP1, which inhibits expression of SAP (46), and LMP2, which maintains a sustained level of ligand-independent activation (47).
CD244 can be phosphorylated in the absence of SAP (16), and EAT-2 has been shown to enhance phosphorylation of CD244 (20,21). In terms of initiating phosphorylation of CD244, the higher affinity of CD48/CD244 (11) over CD150/CD150 (7) 4 interactions in human and mouse will facilitate more effective recruitment of CD244 to cell contacts, where the concentration of active kinases may be sufficient to begin phosphorylation of the tail (28). The FYN SH2 domain can bind directly to the membrane-proximal motif of CD244 and in transfection experiments, expression of FYN induces phosphorylation of CD244 (20). Initial phosphorylation of the membrane-proximal motif fits in with a general mechanism for phosphorylation of a membrane-proximal motif preceding the more distal motif (48) and more specifically for receptors related to CD244 (49).
A model to explain inhibition by CD244 needs to account for differences in functional outcome. One in which the functional consequences of CD48/CD244 interactions depend on which other receptors are engaged at the cell surface would be compatible with recruitment of FYN by CD244 and account for differences in outcome depending on the interacting cells expressing CD48 (1). Such a mechanism is also applicable to the behavior of EAT-2. Experiments with genetically manipulated mice and overexpression in cell lines led to the designation of EAT-2 as a negative regulator (19). In contrast, we have shown that EAT-2 is not inhibitory per se. Seemingly paradoxically, overexpression of EAT-2 in mice has an inhibitory effect on overall phosphorylation levels (19), which, as pointed out pre-viously, could be due to loss of specificity (19). Alternatively, increased EAT-2 concentration could induce an altered balance in favor of phosphorylated inhibitory receptors and a change in balance between phosphorylation and dephosphorylation. A model in which the effects of CD48/CD244 are dependent on which other receptors are engaged at the same time is applicable to other SAP-binding receptors. The niche and specificity of each may be determined largely by distribution, ligand engagement, and the number and spacing of the intracellular ITSMs.