Fine Specificity and Molecular Competition in SLAM Family Receptor Signalling

SLAM family receptors regulate activation and inhibition in immunity through recruitment of activating and inhibitory SH2 domain containing proteins to immunoreceptor tyrosine based switch motifs (ITSMs). Binding of the adaptors, SAP and EAT-2 to ITSMs in the cytoplasmic regions of SLAM family receptors is important for activation. We analysed the fine specificity of SLAM family receptor phosphorylated ITSMs and the conserved tyrosine motif in EAT-2 for SH2 domain containing signalling proteins. Consistent with the literature describing dependence of CRACC (SLAMF7) on EAT-2, CRACC bound EAT-2 (KD = 0.003 μM) with approximately 2 orders of magnitude greater affinity than SAP (KD = 0.44 μM). RNA interference in cytotoxicity assays in NK92 cells showed dependence of CRACC on SAP in addition to EAT-2, indicating selectivity of SAP and EAT-2 may depend on the relative concentrations of the two adaptors. The concentration of SAP was four fold higher than EAT-2 in NK92 cells. Compared with SAP, the significance of EAT-2 recruitment and its downstream effectors are not well characterised. We identified PLCγ1 and PLCγ2 as principal binding partners for the EAT-2 tail. Both PLCγ1 and PLCγ2 are functionally important for cytotoxicity in NK92 cells through CD244 (SLAMF4), NTB-A (SLAMF6) and CRACC. Comparison of the specificity of SH2 domains from activating and inhibitory signalling mediators revealed a hierarchy of affinities for CD244 (SLAMF4) ITSMs. While binding of phosphatase SH2 domains to individual ITSMs of CD244 was weak compared with SAP or EAT-2, binding of tandem SH2 domains of SHP-2 to longer peptides containing tandem phosphorylated ITSMs in human CD244 increased the affinity ten fold. The concentration of the tyrosine phosphatase, SHP-2 was in the order of a magnitude higher than the adaptors, SAP and EAT-2. These data demonstrate a mechanism for direct recruitment of phosphatases in inhibitory signalling by ITSMs, while explaining competitive dominance of SAP and EAT-2.


Introduction
The balance between activation and inhibition in an immune response is critically dependent on the specificity of phosphorylated tyrosine motifs in the cytoplasmic regions of leukocyte surface receptors and associated adaptors for recruitment of intracellular enzymes. The division between activation and inhibition is commonly delineated by specificity of Immunoreceptor Tyrosine based Activation Motifs (ITAMs) and Immunoreceptor Tyrosine based Inhibitory motifs (ITIMs) for an activating kinase or an inhibitory phosphatase respectively. For example, the T cell receptor complex containing multiple ITAMs is an activating receptor and SIRPa with multiple ITIMs is inhibitory. The balance between activation and inhibition is then dependent on recruitment of the receptors at the cell surface.
An alternative strategy to regulating the balance between activation and inhibition through tyrosine phosphorylation is in the specificity of receptor motifs for the intracellular binding partners. The SLAM family receptors within the CD2 family of leukocyte surface receptors contain multiple Immunoreceptor Tyrosine-based ''Switch'' Motifs (ITSMs) (Figure 1). These motifs carry the consensus sequence TxYxxV/I/L and biochemical and functional analysis indicate they have overlapping specificity for activating and inhibitory binding partners. CD244 has the greatest number of ITSMs, four ( Figure 1). The primary binding partners for these ITSMs are the small SH2 domain containing adapter proteins SAP [1][2][3][4][5][6][7][8] and EAT-2 [9][10][11]. In vivo and in vitro studies in mouse and human have shown that activation through SLAM family receptors is dependent on recruitment of SAP and EAT-2. In the absence of SAP and EAT-2, there is loss of activating function through SLAM family receptors, most notably in humans with mutations in the SH2D1A (SAP) gene who develop X-linked lymphoproliferative disease (XLP), a condition characterized by the failure to clear persistent viral infections such as EBV, lymphoid hyperplasia, and lymphoma development [8,12]. Among the defects in lymphocyte function are impairment of cytotoxic activity in both NK cells and CD8+ T cells [13][14][15][16][17][18].
NK cells express both SAP and EAT-2 proteins [5,9]. The best characterized molecular interaction of SAP family adapters with an intracellular enzyme is the recruitment of the Src-family kinase Fyn ( Figure 1) [19][20][21][22][23]. There is evidence that SAP can also couple SLAM family receptors to other cytosolic effector enzymes [24][25][26] through the conserved interaction of SH3 domains with Arg78 on SAP [20]. EAT-2 lacks a similar structural motif and so cannot interact with SH3 domains. Instead, it contains a conserved YxxV motif which binds SH2 domain containing proteins [22,27]. In mice, NK cell-mediated cytotoxicity against SLAM-family receptor-expressing target cells is impaired in both SAP 2/2 and EAT-2/ERT 2/2 mice, ERT being an additional EAT-2 homologue in mice [21,28,29]. Activation through EAT-2 is dependent on tyrosine phosphorylation of the tail of EAT-2 [27,30]. Recruitment of Fyn by EAT-2 [11,21,27] may contribute to early phosphorylation events, however, an interaction between PLCc1 and the tyrosine phosphorylated tail of EAT-2 [11,27] suggests a mechanism of recruitment of PLCc through SLAM family receptors ( Figure 1) [9,11,27,31,32]. Furthermore, mouse NK cells lacking PLCc2 fail to mobilize calcium and have impaired cytotoxicity through all activating receptors examined including CD244 [33]. PLCc1 deficient mouse embryos are not viable [34] but complementation with PLCc1 indicated specific niches for each of the two isoforms [35].
In addition to loss of activating function in the absence of SAP, EAT-2 and ERT, SLAM family receptors exhibited effects which suggested the receptors were actively attenuating signalling [28,36]. Recently signal attenuation by Ly108/NTB-A has been dissected providing evidence for biochemical loops of activation and inhibition [23,37], and analysis of functional dependence on the tyrosine phosphatases, SHP-1 and SHP-2 versus an inositol phosphatase, SHIP suggested a dominant role for the latter [23]. Colocalisation of SHP-1 with a SLAM family receptor in the absence of SAP indicates a role for the tyrosine phosphatases in signal attenuation [37,38]. Physical associations with inhibitory binding partners of SLAM family receptor ITSMs have been proposed based on indirect assays such as coprecipitation [5,18], pulldowns [39] and colocalization, [37,38] and on direct binding studies on protein arrays [40] and surface plasmon resonance (SPR) [22,40].
We sought to more clearly define the hierarchy of binding partners for ITSMs and the EAT-2 tail. We distinguished between specificity of ITSMs for the adaptors, SAP and EAT-2 and for inhibitory binding partners [22,27,39,40]. PLCc1 and PLCc2 were identified as binding partners for the EAT-2 tail, and functional experiments confirmed that SLAM family receptor mediated killing in human is critically dependent on SAP, PLCc1 and PLCc2 and to a lesser extent on the presence of EAT-2.

Protein expression and purification
Soluble recombinant proteins were produced in BL21(DE3)-pLysS E. coli (Calbiochem) at 37uC in 2TY medium and induced at OD 600 = 0.6 with 0.1 mM IPTG for 4 h or in 2XYT autoinduction (Formedium) medium overnight. Bacterial pellets were resuspended in ,0.1 volume 10 mM HEPES pH7.4, 150 mM NaCl and used immediately or stored at 220uC in 10 ml aliquots. His-tag mediated protein purification was performed on a Co 2+ or Ni 2+ column by standard protocols. Protein purification of SAP, EAT-2, PLCc1 and PLCc2 was optimized by being performed on the day of SPR experiment in order to use protein at the highest activity with minimum aggregation. Where applicable, proteins were stored frozen in 10-25 mM HEPES pH 7.4, 200-300 mM NaCl (pH 7.4).All proteins were purified by size exclusion chromatography immediately prior to SPR analysis.

Protein pull-down and LC-mass spectrometry
For pull-down of associated proteins, streptavidin Sepharose (GE Healthcare) was pre-coated with saturating amounts of peptide or EAT-2. NK92 cells (4610 8 ) were treated with 500 mM pervanadate for 15 minutes, then lysed in PBS, 1% Triton X-100, 1 mM sodium orthovanadate in the presence of mammalian protease inhibitor solution (Sigma). Lysates were pre-cleared by centrifugation, followed by incubation with streptavidin Sepharose. Peptide or EAT-2-coated streptavidin beads were incubated with NK92 lysates for 2 hours at 4uC with rotation. Beads were washed once with 15 ml ice cold PBS/1% (v/v) Triton X-100. Bound protein was eluted using 100 mM glycine-HCl pH 2.5.
Protein samples were prepared for mass spectrometry as described previously [42]. Briefly, eluted proteins were spun onto the membranes of Vivacon 500 filters (10,000 MWCO -Sartorius). Membranes were washed once with PBS then proteins were denatured with 8 M urea, reduced with 10 mM DTT in 25 mM ammonium bicarbonate buffer, followed by alkylation with 50 mM iodoacetamide. Proteins were then digested overnight with trypsin. Tryptic peptides were then eluted from the membrane and analyzed on an Ultimate 3000 HPLC-coupled Orbitrap XL mass spectrometer. Proteins were identified by liquid chromatography-coupled tandem mass-spectrometry (LC-MS/ MS). Raw data files were converted to mzXML format using ReADW (v. 4.2.1) and submitted to the Central Proteomics Facilities Pipeline [43] for database comparison and peptide identification. Searching against the International Protein Index (IPI) human protein sequence database for protein identification [44], protein and peptide hits were statistically validated by applying a 1% false discovery rate (FDR) by searching against a concatenated target/decoy database. Datasets were searched with variable peptide modifications for phosphotyrosine, phosphoserine, phosphothreonine, oxidation of methionine and deamination.
Protein identification data were exported into ProteinCenter (v3.8.3013 -Thermo Fisher) for filtering and annotation. Results were filtered to remove any proteins with less than 3 unique peptides, less than 10 percent sequence coverage, or median protein probability scores less than 0.5. To establish sample specificity, any protein with at least 10-fold greater sequence coverage in one sample compared with the other was considered to be specific for that sample.

Surface plasmon resonance (SPR)
Measurements of protein-protein affinity were performed on BIAcore 3000 and T200 systems using CM5 or biotin capture chips (GE Healthcare Ltd, UK). Streptavidin was coupled to CM5 sensor chips using standard NHS/EDC chemical linkage, followed by immobilization of biotinylated peptides. The levels of peptide immobilized were approximately 25 and 50 response units respectively of single and tandem tyrosine motifs to facilitate saturating binding and limitation of ''bridging'' [22,45]. Purified recombinant proteins were passed over the immobilized peptides at a range of concentrations in HBS-EP buffer, 10 mM HEPES pH 7.4, 150 mM NaCl, 3.4 mM EDTA, 0.005% surfactant, P20. Equilibrium binding coefficients and dissociation and association rates were calculated using BIAevaluation and/or GraphPad Prism 5.4 software.

RNA interference
Lentiviral constructs expressing gene-specific shRNAs in vector pLKO.1 were obtained from Open Biosystems. Panels of 5 shRNAs were tested for each protein to select the most effective construct. The specific clones used were as follows: TRCN0000006979 for PLCc1, TRCN0000002322 for PLCc2, TRCN0000082710 for SAP, and TRCN0000082657 for EAT-2. As a non-targeting control, the hairpin sequence 59-GCGCGCTTTGTAGGATTCGTTCAAGAGACGAATCC-TACAAAGCGCGCTTTTTT was expressed in pLKO.1. Lentiviral particles were produced through polyethylenimine (PEI-Sigma) co-transfection of 293 T cells with pLKO.1 along with packaging vectors pMD2.G and pSPAX2 (Addgene; Didier Trono Laboratory). Virus was produced in RPMI 1640/10% heatinactivated FCS for 48-72 hours. Viral supernatants were collected and NK92 cells were transduced twice overnight in the presence of 1000 U/ml IL-2. Cells with stable shRNA expression were selected using 3 mg/ml puromycin (Invitrogen) for 72 hours prior to use in functional assays.

Quantitation of protein expression
For quantitation of protein expression, NK92 cell lysates were prepared using buffer (150 mM NaCl, 20 mM Tris-HCl pH 7.4, 10% glycerol, 2 mM EDTA, 10 mM NaF, 1 mM PMSF) containing 0.5% Triton X-100. Lysates were resolved by SDS-PAGE under reducing conditions using Novex gradient gels and transferred to nitrocellulose membranes for western blotting by standard protocols. Following RNA interference, blots containing protein lysates from 10 5 cells were developed with: anti-EAT-2 antibody (rabbit polyclonal; a kind gift from Marco Colonna and Ilaria Tassi), anti-PLCc1 (2822) from Cell Signalling Technology or anti-human PLCc2 (Q-20) from Santa Cruz. For quantitation of SAP and EAT-2 protein levels in NK92 cells, the equivalent of 0.522610 6 cells from lysates prepared at 5610 7 cells/ml was analysed alongside known amounts of recombinant protein standards, recombinant full-length human SAP (50, 25, 12.5 and 6.25 ng) or EAT-2 (20, 10, 5 and 2.5 ng), generated as previously described and developed with anti-SAP antibody (rabbit polyclonal clone FL128, Santa Cruz Biotechnology) or anti-EAT-2. For chemiluminescent detection, HRP conjugated anti-mouse and anti-rabbit IgG antibodies were obtained from Cell Signaling Technology. SHP-2 levels in 10 5 cell equivalents were analysed in a similar manner with recombinant SHP-2 N+C as a standard (25, 50 and 100 ng) and anti-SHP-2 (Cell Signaling #3752). Fluorescent detection of protein bands was performed on the LI-COR Odyssey Sa system. IRDye 680LT-conjugated anti-rabbit IgG and IRDye 800 CW anti-mouse IgG were obtained from LI-COR Biosciences. Quantitation of protein expression was performed using LI-COR Odyssey Sa software version 1.0. In initial experiments, estimations of % knockdown in comparison with another protein blotted with a specific antibody or in comparison with a nonspecific ''reference'' band gave the same result. Subsequently, a ''reference'' band was routinely used. After RNA interference, SAP expression was also quantified by intracellular staining using anti-human SAP (XLP-1D12) from Cell Signalling Technology and median fluorescence intensity measured by flow cytometry.

Cytotoxicity assays
Cytotoxicity through individual receptors was measured in a redirected lysis assay. P815 target cells were labeled for 3 hours with EXPRE 35 S 35 S Protein Labeling Mix (Perkin Elmer, Inc.) in cysteine/methionine-free RPMI (Sigma). Target cells were washed 3 times in complete RPMI and dispensed into 96-well round bottom plates at 5000 cells/well. Lentiviral-transduced NK92 cells were harvested and mixed with target cells at the indicated effector:target ratio in duplicate unless otherwise stated in the presence of IL-2. The indicated antibodies were added to the cellular mixture for the duration of the 4 hour cytotoxicity assay. The antibodies used to induce cytotoxicity were as follows: anti-NKp30 (P30.15) and anti-NTB-A (NT-7) were obtained from Biolegend; anti-CRACC (162) and anti-CD58 (MEM-63) were obtained from Abd Serotec; anti-CD2 (X53) and anti-CD244 (C1.7) were purified in our laboratory. The numbers of independent experiments were: SAP -2, EAT2 -3, PLCc1 -4, and PLCc2 -2.

SAP and EAT-2 show differential binding to motifs in SLAM family receptors
The most distinct differences between SAP and EAT-2 are the mechanisms of recruitment of downstream effectors and their distribution. The specificities of their SH2 domains overlap, introducing the concept of competition between them for ITSMs in cells where they are co-expressed, including NK cells. In our previous studies we focused on comparison between human and mouse CD244 ITSMs showing that SAP and EAT-2 had similar specificity for phosphorylated ITSMs in both species [22]. Signalling through the SLAM family receptor, CRACC has been reported to be independent of SAP [9,30], suggesting a difference in specificity of the CRACC ITSMs for SAP and EAT-2. We measured the affinity of SAP and EAT-2 for CRACC peptides by SPR ( Figure 2). Binding of SAP to CRACC ITSM2 was observed, however, EAT-2 bound about a hundred fold more strongly to CRACC ITSM2 with a K D = 0.003 mM ( Figure 2 and Table 1). The preference of CD244 ITSM1 for EAT-2 was less striking, with the K D for binding of EAT-2 to CD244 ITSM1 being one order of magnitude stronger than SAP ( Figure 2, Table 1). This selectivity of the CRACC ITSM may explain the failure to detect SAP binding to CRACC by immunoprecipitation [9] and the reported dominance of EAT2/ERT in mouse CRACC signalling [30]. Neither SAP nor EAT-2 bound phosphorylated human CRACC ITSM1 (data not shown). In mouse the tyrosine motif (ADYDTI) in the equivalent position to human CRACC ITSM1 deviates from the specificity and consensus for an ITSM (TxYxxV/I/L), did not bind EAT-2 and differed from ITSM2 in its functional effects [30]. These measurements were carried out with full length SAP and EAT-2 purified on the same day (see Methods) and affinities were comparable to those obtained with the isolated SH2 domains ( [22] and data not shown). Lower affinities obtained with full length SAP and EAT-2 in our earlier study [22] were most likely related to protein activity.

SAP can regulate signalling through CRACC
Overlapping specificity of SAP and EAT-2 for ITSMs suggested both SAP and EAT-2 had the potential to regulate signalling through SLAM family receptors, including CRACC. We tested for functional dependence of cytotoxicity on SAP and EAT-2 in a human NK cell line, NK92, targeting expression by lentiviral transduction of a specific shRNA. Receptor-specific effects on cytotoxicity were determined by redirected lysis assays using the Fc receptor-expressing, mouse mastocytoma cell line P815 as targets. Antibodies to CD244, NTB-A, and CRACC, the SLAM family receptors expressed on NK92 cells were used to cross-link these receptors and anti-CD58 was used as a negative control. To examine the specificity of the relative contribution of proteins important for SLAM family activating signal transduction in additional receptor pathways, the activation of cytotoxicity through CD2 (which contains multiple PxxP motifs for SH3 domain binding) and NKp30 (which associates with CD3f) were tested. The cytotoxic activity of the SLAM family receptors, including CRACC, was reproducibly reduced (n = 2) following SAP knockdown ( Figure 3A, B). Interestingly, the reduction in SAP expression consistently resulted in higher CD2 mediated cytotoxicity, indicating possible interference by SAP in unknown pathways downstream of CD2.
Since there is competition between EAT-2 and SAP for ITSMs, the concentrations of SAP and EAT-2 are factors in determining dependence of SLAM family receptors on these adaptors. We measured the concentration of SAP and EAT-2 in NK92 cells by western blotting (Figure 4, Table 2) and determined a SAP to EAT-2 ratio of 4:1. Using an approximation of the diameter of NK92 cells as 10 mm, which gives a cellular volume of 523 fL, we estimated the total cellular concentration of SAP as 1.9 mM and EAT-2 as 0.44 mM.
RNA interference in the redirected lysis assay in NK92 cells reduced the concentration of EAT-2 by up to ,50% ( Figure 5). In the majority of experiments (n = 3), no effect of EAT-2 knockdown on SLAM family mediated cytotoxicity was observed. This may be due to an insufficient reduction in intracellular EAT-2 concentration. We determined the intracellular EAT-2 concentration in NK92 cells to be 0.44 mM, and so a 50% reduction in EAT-2 through shRNA mediated knockdown will yield an intracellular concentration of approximately 0.2 mM. Based on the dissociation constants for EAT-2 binding to CD244 and CRACC (Figure 2), an EAT-2 concentration of 0.2 mM would approach saturation of the phosphorylated ITSMs. This is in contrast to SAP where the endogenous concentration, 1.9 mM was knocked down to 0.6 mM. According to the equilibrium binding data (Figure 2), this concentration of SAP is well below saturation for both CD244 and CRACC, and so we observe a functional impact of the knockdown. In each experiment CD2 and NKp30 mediated cytotoxicity were consistently enhanced by EAT-2 knockdown ( Figure 5 and data not shown). Analysis of spontaneous lysis of the NK-susceptible target B cell line 721.221 indicated that engagement of multiple receptors facilitates the activating function of EAT-2 in NK92 cells (( Figure 5 [39]).

ITSMs show a hierarchy in binding SH2 domains from activating and inhibitory signalling proteins
Analysis of binding of SAP and EAT-2 to ITSMs revealed differences in specificity which can explain different modes of activation. Similarly, differences in specificity of inhibitory SH2 domain containing proteins for ITSMs are relevant to understanding mechanisms of inhibition by ITSMs. We focused on CD244 ITSMs to establish general paradigms. As CD244 binding to the adaptors, SAP and EAT-2 is dependent on phosphorylation [11,22,47] only binding to phosphorylated peptides was analysed. We first confirmed the potential of CD244 ITSM1 to recruit activating and inhibitory binding partners by performing pulldown experiments with human NK cell lysates and mass spectrometry analysis of proteins bound specifically to CD244 ITSM1 when compared with a control phosphopeptide (SHIP-1 Y 865 ). In addition to identifying SAP and EAT-2, the SH2 domain containing phosphatases SHIP-1, SHIP-2, SHP-1, SHP-2 and CSK, as well as GRB-2 and PLCc1 were also pulled down by the ITSM peptide (Table 3).
We examined whether hierarchies of specificity of CD244 ITSMs for activating and inhibitory binding partners differ. SHIP-1 was studied as an example of a single SH2 domain containing phosphatase and SHP-2 an example of tandem SH2 domains. We compared the relative specificity of the single SH2 domain from SHIP-1 for the 4 ITSMs from CD244 ( Figure 6A, C and Table 4). A representative sensorgram trace for binding of SHIP-1 SH2 domain to CD244 ITSM1 and ITSM2 and a positive control, a well characterized SHIP binding peptide from FCRcIIb, are shown ( Figure 6A). Equilibrium binding analysis ( Figure 6C, Table 4) revealed a preference of SHIP SH2 domain for CD244 ITSM2 and ITSM3. The same pattern was observed in binding studies with mouse SHIP SH2 domain and mouse CD244 ITSMs (data not shown). While the pattern of binding to the different ITSMs was similar for the inhibitory SHIP and activating adaptor domains (Table 4), SAP and EAT-2 consistently bound ITSMs ten to a hundred fold more strongly than SHIP-1. FCRcIIb was included as a monitor of SHIP SH2 domain activity and the K D was comparable with previous measurements [41].

Tandem SH2 domains bind tandem ITSMs and ITIMs
Binding of inhibitory tyrosine phosphatases through their SH2 domains to ITIMs is enhanced by phosphorylation of two tyrosine motifs which are appropriately spaced [45]. We identified pairs of ITSMs (based on spacings of 23-30 amino acids between tyrosines) in the cytoplasmic tails of SLAM family receptors including human CD244 ITSMs 1 and 2 and human CD244 ITSMs 3 and 4 with the potential to simultaneously bind to tandem SH2 domains of the tyrosine phosphatases. We measured binding of SHP-2 tandem SH2 domains (recombinant SHP-1 SH2 domains were unstable in solution) to tandem ITSMs in CD244 ( Figure 6B and Table 4). A representative sensorgram (Fig. 4B) shows that equilibrium binding was achieved with minimal interference with aggregated material only apparent at the higher concentrations of soluble SHP-2. We measured an order of magnitude increase in affinity of binding of tandem SH2 domains to the tandem ITSM peptides compared with individual ITSMs (Fig. 4D, Table 4). No increase in affinity was observed when both individual ITSM peptides were immobilized. Binding to the tandem peptide representing CD244 ITSM3+ITSM4 was 3-5 fold weaker compared with CD244 ITSM1+ITSM2 at 37uC and 25uC and   we observed the same phenomena for the SIRPa tandem ITIM peptides (Table 4 and data not shown). Binding of SHP-2 tandem SH2 domains to CD244 was twothree fold weaker, but comparable with binding to tandem ITIMs from a well characterized inhibitory receptor SIRPa [48] ( Figure 6D, Table 4). We obtained similar K D values for at least three different preparations of SHP-2 tandem SH2 domains and observed similar affinities with full length SHP-2 (not shown). Quantitation of the levels of SHP-2 in NK92 cells, as was done for SAP and EAT-2, indicated there is an order of magnitude higher concentration compared with SAP, and is approximately an order of magnitude higher than the K D , indicating that SHP-2 will be competitive for binding highly phosphorylated receptor.

The tail of EAT-2 binds phospholipase C (PLC) c1 and PLCc2
In contrast to the characterization of the specificity of ITSMs, the underlying molecular basis of activation and inhibition by EAT-2 has not been clarified. Activating effects of EAT-2 downstream of SLAM family receptors are dependent on phosphorylation of the tyrosine motif in the tail [27,29]. In our previous studies, PLCc1 was identified as a potential binding partner for the EAT-2 tail in human T cell blasts [27]. Since EAT-2 is predominantly expressed by NK cells and human CD8 + T cells [9], it was important to determine whether there are additional binding partners for the phosphorylated tail of EAT-2 in human NK cells. We carried out an EAT-2 Y 127 phosphopeptide pull-down of proteins using NK92 cell lysates. Proteins identified by at least 3 unique peptides associating with EAT-2 Y 127 phosphopeptide but not with EAT-2 SH2 domain are listed in Table 5. Of these, only PLCc1 and PLCc2 are known to contain SH2 domains capable of phosphotyrosine-dependent interactions, and none of the proteins contain a Phospho-Tyrosine Binding (PTB) domain. In addition, PLCc1 and PLCc2 were identified with substantially more unique peptides and greater sequence coverage than any other proteins, suggesting lesser ''hits'' are likely to be non-specific. Notably absent from this list is Fyn, which we previously found to have a measurable affinity (K D ,1 mM) for the EAT-2 tail [22].
We compared direct binding of PLCc1 and PLCc2 to EAT-2 Y 127 and LAT Y 132 by SPR. Direct binding of recombinant PLCc1 to EAT-2 Y 127 was confirmed (Figure 7 and Table 6 [27]). Recombinant PLCc2 SH2 domains also bound immobilized EAT-2 Y 127 and LAT Y 132 peptides with comparable affinity (Figure 7, Table 6) demonstrating the ability of EAT-2 to recruit both PLCc isoforms. PLCc1 and PLCc2 are encoded on human chromosomes 20 and 16 respectively and share only 63% sequence identity in their SH2 domains, so their overlapping specificity is for EAT-2 is remarkable. As for their relative affinity for EAT-2, we cannot rule out that the two fold lower affinities observed for PLCc2 may be due to protein activity as recombinant PLCc2 was not as stable as PLCc1. Binding of CD244 ITSM1 to PLCc was quantifiable but an order of magnitude weaker than the interaction between EAT-2 and PLCc indicating a potential role for the adapter in linking surface receptors with effector enzymes (Figure 7, Table 6).

SLAM family receptor mediated cytotoxicity is dependent on PLCc1 and PLCc2
PLCc1 and PLCc2 have distinct functions in NK cells [35]. PLCc2 is expressed by NK cells and has been previously shown in mice to be critical for calcium flux and granule exocytosis during mouse NK cell activation [33]. Functional dependence of cytotoxicity on PLCc1 has not been studied because mice lacking PLCc1 are not viable. We compared the effects of knocking down PLCc1 and PLCc2 on SLAM family mediated cytotoxicity using lentiviral shRNA (Figure 8). Targeting of either PLCc1 or PLCc2 was isoform specific ( Figure 8B). In the case of PLCc1, protein expression was routinely reduced by 70% or more ( Figure 8B). When PLCc1 expression was suppressed, specific lysis of P815 cells through CD244, NTB-A and CRACC was reduced ( Figure 8A). Consistent with a previous report that receptors that signal through ITAMs such as NKp30 can activate either PLCc1 or PLCc2 [49], a partial reduction in NKp30-mediated cytotoxicity in PLCc1 knockdown cells was observed ( Figure 8A). Consistent with data from PLCc2 2/2 mice [33,35,50], knockdown of PLCc2 in human NK92 cells had a negative impact on cytotoxicity through all of the activating receptors analyzed. The  Table 2. Quantitation of SAP, EAT-2 and SHP-2 in NK92 cells. data show clearly that SLAM family receptor mediated cytotoxicity depends on both isoforms PLCc1 and PLCc2.

Discussion
Analysis of interactions of activating partners for ITSMs in the cytoplasmic regions of SLAM family receptors distinguished specificity of SAP and EAT-2 SH2 domains which correlates with published functional data [9,30]. Independence of the SLAM family receptor, CRACC on SAP has been observed in SAP deficient mouse and human primary cells [9,30]. However, we observed reliance of CRACC on SAP in redirected lysis assays in the human cell line, NK92. SPR analysis showed that SAP bound to human CRACC ITSM2, albeit two orders of magnitude more  weakly than EAT-2, supporting the concept that there is competition between SAP and EAT-2 for SLAM family receptor ITSMs [9] depending on adapter concentration and differences in binding kinetics. The concentration of EAT-2 in NK92 cells was approximately 4 fold lower than that of SAP. However, it was two orders of magnitude above the K D for binding ITSMs which may explain the marginal effects on cytotoxicity of halving the concentration of EAT-2 by RNA interference. Additional factors including phosphorylation of its tail are likely to be limiting in determining activating effects of EAT-2.
In our previous measurements of SH2 domain binding to phosphopeptides we observed a difference in specificity between tyrosine motifs which bound activating and inhibitory SH2 domains [41]. However, the potential of ITSMs to recruit inhibitory binding partners does not appear to involve an alteration in which motifs are likely to bind most strongly. There was a trend for the membrane proximal tandem ITSMs and ITIMs to bind more strongly, compared with the membrane distal ones, to the tandem SH2 domains of SHP-2 which may reflect some more general conservation of topology for multiple motifs [51]. Experiments with a recombinant cytoplasmic region of human CD244 have implicated CD244 ITSM3 as being dominant in SHP-1 binding [39] indicating other factors influence the relative importance of ITSMs in a given response.
Analysis of interactions of tandem SH2 domains of tyrosine phosphatases with tandem ITSMs indicated that there will be  productive recruitment if concentrations of the adaptors, SAP and EAT-2 are compromised. This is consistent with functional data from SAP-family deficient mice [28] and human XLP patients. In the case of XLP patients, the function of NTB-A, which was mainly activating in NK clones from healthy individuals, became largely inhibitory in the absence of SAP. [18] Without SAP present, SH2 domains from SHP1 and SHP2 could more easily compete for ITSMs, and the likelihood of forming the higher avidity tandem interaction will increase. There was a discernible difference between ITSMs and ITIMs in specificity for inhibitory binding partners. Subtle differences in sequence and consequent binding affinities may have a significant impact on signalling [52]. Increased affinity through tandem binding of SH2 domains to tandem phosphorylated tyrosine motifs over mono phosphorylated motifs has been observed previously and most likely contributes to the specificity of SH2 domain containing proteins [45,53], though in our case we cannot exclude the contribution of additional sequence within the peptides, rather than a tandem domain interaction. Inhibitory effects of SLAM family receptors do not necessarily depend on tandem ITSM phosphorylation implying that single SH2 domain recruitment of inhibitory binding partners is feasible [37]. Activation of the single SH2 domain containing protein, SHIP is a proximal event in CD244 signal transduction [47] and there are data supporting competition between SAP and SHIP for a direct interaction with an ITSM in SLAM [40,54]. A stronger dependency of mouse CD244 on SHIP compared with SHP-1 and SHP-2 [23] may reflect insufficient phosphorylation of tandem ITSMs. Alternatively, in mouse CD244, ITSMs 1 and 2 may be less favourably spaced for binding tandem SH2 domains, being further apart than in human CD244. Overlaid SPR sensorgrams (A) and equilibrium binding curves (B) for ten serial injections of two-fold dilutions of human PLCc1 SH2-SH2-SH3 or PLCc2 SH2-SH2-SH3 (top concentration 1 mM) over immobilised phosphorylated peptides representing EAT-2 pY127, LAT pY132 or CD244 ITSM1 at 37uC. Signal from a blank reference cell was subtracted as background. K D values were determined by fitting equilibrium binding data using a one-site specific binding model. Proteins were purified on the same day as SPR analysis. doi:10.1371/journal.pone.0092184.g007 In contrast to ITSMs, the specificity of the conserved phosphotyrosine motif in the tail of EAT-2 appears to be more limited. We have characterized interactions with FYN and PLCc, the latter binding up to two orders of magnitude more strongly [27]. In pulldown experiments by comparison with another phosphopeptide [27] or with the EAT-2 SH2 domain we identified both PLCc proteins as being specifically isolated with the EAT-2 tail phosphopeptide, and PLCc1 and PLCc2 displayed similar specificity for EAT-2 and for the positive control phosphopeptide from LAT. Actual binding affinities in cells depending on phosphorylation of PLCc may be an order of magnitude lower than measured by SPR with the recombinant PLCc produced in bacteria [32]. Activation of PLCc is a proximal event in SLAM family signal transduction [9,31,55] and we observed a dependence of cytotoxicity on both PLCc1 and PLCc2. We have observed a striking difference in the effects of PLCc1 and PLCc2 on IFNc production by NK92 cells, inhibitory and activating respectively (TJW, unpublished) but the molecular mechanism of differences in regulation by the two isoforms is unknown [35]. Mechanisms of recruiting and activating PLCc1 and PLCc2 downstream of SLAM family receptors include colocalization of CD244 with LAT [55,56]. A role for EAT-2 in activation of PLCc is suggested by compromised calcium influx through CRACC in mice deficient in both EAT-2 and the mouse specific EAT-2 related protein, ERT [30]. Coprecipitation of EAT-2 and PLCc1 in transduced hybridoma cells indicated this interaction could occur in cells [27]. Therefore recruitment of PLCc by EAT-2 may provide an auxiliary mechanism for activating PLCc.
In summary, we have identified differential and overlapping specificities of positive and negative regulators of cell signalling for the ITSMs of SLAM family receptors. The affinities measured indicate an intracellular environment where positive and negative regulators will compete for the phosphorylated ITSMs of SLAM family receptors. The relative affinities of the SH2 domains studied indicate that the adapters SAP and EAT-2 will play a dominant role in promoting cellular activation when present, however, signal attenuation by phosphatases is possible through direct recruitment to ITSM-containing receptors, especially in the absence of SAP and EAT-2.