Molecular basis of the recruitment of the SH2 domain-containing inositol 5-phosphatases SHIP1 and SHIP2 by fcgamma RIIB.

FcgammaRIIB are single-chain low affinity receptors for IgG that negatively regulate immunoreceptor tyrosine-based activation motif-dependent cell activation. They bear one immunoreceptor tyrosine-based inhibition motif (ITIM) that becomes tyrosyl-phosphorylated upon coaggregation of FcgammaRIIB with immunoreceptor tyrosine-based activation motif-bearing receptors and that recruits SH2 domain-containing inositol 5-phosphatases (SHIPs) in vivo. Synthetic FcgammaRIIB ITIM phosphopeptides, however, also bind SH2 domain-containing protein-tyrosine phosphatases (SHPs) in vitro. To identify SHIP-binding sites, we exchanged residues between the FcgammaRIIB ITIM and the N-terminal ITIM of a killer cell Ig-like receptor that does not bind SHIPs. Loss of function and gain of function substitutions identified the Y+2 leucine, in the FcgammaRIIB ITIM, as determining the binding of both SHIP1 and SHIP2, but not the binding of SHP-1 or SHP-2. Conversely, the Y-2 isoleucine that determines the in vitro binding of SHP-1 and SHP-2 affected neither the binding nor the recruitment of SHIP1 or SHIP2. One hydrophobic residue, in the ITIM of FcgammaRIIB therefore determines the affinity for SHIPs. This residue is symmetrical to the hydrophobic residue that determines the affinity of all ITIMs for SHPs. It defines a SHIP-binding site, distinct from a SHP-binding site, that enables FcgammaRIIB to recruit SHIP1 and SHIP2 and that is preferentially used in vivo.

Fc␥RIIB are single-chain low affinity receptors for IgG that negatively regulate immunoreceptor tyrosinebased activation motif-dependent cell activation. They bear one immunoreceptor tyrosine-based inhibition motif (ITIM) that becomes tyrosyl-phosphorylated upon coaggregation of Fc␥RIIB with immunoreceptor tyrosinebased activation motif-bearing receptors and that recruits SH2 domain-containing inositol 5-phosphatases (SHIPs) in vivo. Synthetic Fc␥RIIB ITIM phosphopeptides, however, also bind SH2 domain-containing protein-tyrosine phosphatases (SHPs) in vitro. To identify SHIP-binding sites, we exchanged residues between the Fc␥RIIB ITIM and the N-terminal ITIM of a killer cell Ig-like receptor that does not bind SHIPs. Loss of function and gain of function substitutions identified the Y؉2 leucine, in the Fc␥RIIB ITIM, as determining the binding of both SHIP1 and SHIP2, but not the binding of SHP-1 or SHP-2. Conversely, the Y؊2 isoleucine that determines the in vitro binding of SHP-1 and SHP-2 affected neither the binding nor the recruitment of SHIP1 or SHIP2. One hydrophobic residue, in the ITIM of Fc␥RIIB therefore determines the affinity for SHIPs. This residue is symmetrical to the hydrophobic residue that determines the affinity of all ITIMs for SHPs. It defines a SHIP-binding site, distinct from a SHP-binding site, that enables Fc␥RIIB to recruit SHIP1 and SHIP2 and that is preferentially used in vivo.
Immunoreceptor tyrosine-based inhibition motifs (ITIMs) 1 are present in the intracytoplasmic (IC) domains of a large group of transmembrane molecules that negatively regulate cell activation induced by receptors bearing immunoreceptor tyrosine-based activation motifs (ITAMs) (1). Fc␥RIIB, a subgroup of Fc receptors that bind IgG complexes (2), and killer cell Ig-like receptors with a long IC domain (KIRLs), which bind major histocompatibility complex class I molecules (3), are two prototypes of ITIM-bearing receptors. Fc␥RIIB exist as two (Fc␥RIIB1 and B2 in humans) or three (Fc␥RIIB1, B1Ј, and B2 in mice) alternatively spliced products of a single gene. Fc␥RIIB were shown to negatively regulate cell activation induced by all ITAM-bearing immunoreceptors (4) and to control the magnitude of both antibody responses and anaphylactic reactions (5). Fc␥RIIB were also recently found to negatively regulate cell proliferation induced by growth factor receptors with an intrinsic protein-tyrosine kinase activity (6). KIRLs are polymorphic molecules with two (KIR2DLs) or three (KIR3DLs) Ig-like extracellular (EC) domains encoded by related but distinct genes. KIRLs inhibit NK and T cell-mediated cytotoxicity (7,8). Negative regulation exerted by Fc␥RIIB (9) and KIR2DL3 (10) was shown to require their coaggregation with ITAM-bearing receptors by extracellular ligands. Fc␥RIIB possess one ITIM, while KIR2DL3 possess two ITIMs.
ITIMs are constituted by a tyrosine, preceded by an isoleucine, valine, or leucine at position YϪ2, and followed by a valine or leucine at position Yϩ3 (1,11). Upon coaggregation of ITIMbearing receptors with ITAM-bearing receptors, ITIMs are tyrosyl-phosphorylated by Src family protein-tyrosine kinases (12,13). Phosphorylated ITIMs (pITIMs) then recruit cytoplasmic phosphatases containing Src homology 2 (SH2) domains that interfere with signals transduced by ITAM-bearing receptors (14 -16). These include the two-SH2 domain-containing protein-tyrosine phosphatases SHP-1 (14,17) and SHP-2 (18) and the single-SH2 domain-containing inositol 5-phosphatases SHIP1 (19) and SHIP2 (20). SHP-1 is thought to dephosphorylate tyrosines in ITAMs, protein-tyrosine kinases, and/or adapter proteins whose phosphorylation is critical for activation signals, thereby stopping the initial steps of transduction. SHP-1 was recently found to inhibit the redistribution of cholesterol/sphyngolipid-rich membrane lipid microdomains following the engagement of a KIR3DL, on NK cells, by major histocompatibility complex class I molecules on target cells (21). The possible role of SHP-2 is not clear, because both positive and negative effects have been assigned to this phosphatase (22)(23)(24). SHIP1 and SHIP2 remove 5-phosphate groups in inositol phosphates and phosphatidylinositol phosphates (25). The preferred substrate of SHIP1 is phosphatidylinositol 3,4,5-trisphosphate, which enables the membrane recruitment of the Bruton's tyrosine kinase via its pleckstrin homology domain (15,16). Bruton's tyrosine kinase is manda-tory for phospholipase C ␥ to be activated and to hydrolyze phosphatidylinositol (4, 5)-bis-phosphate into inositol 1,4,5trisphosphate, which induces a Ca 2ϩ response, and diacylglycerol, which activates protein kinase C. In addition, SHIP1 was recently found to function as a linker molecule that recruits the RasGAP-binding protein p62 dok , leading to an inhibition of the Ras pathway (26). The possible role of SHIP2 is unknown.
The affinity of pITIMs for SH2 domain-containing phosphatases requires the conservation of both the Y and the Yϩ3 residues (11). Synthetic peptides corresponding to pITIMs of all ITIM-bearing molecules were found to bind SHP-1 and SHP-2 in vitro (14,17,27). Phosphorylated peptides corresponding to the Fc␥RIIB ITIM (19), but not phosphorylated peptides corresponding to the KIR2DL3 ITIMs (27), also bound SHIP1. The in vitro binding of SHP-1 and SHP-2 to the pITIMs of KIR2DL3 and Fc␥RIIB depends on the YϪ2 residue (11,27). The molecular basis for the binding of SHIPs to the pITIM of Fc␥RIIB is unknown. Noticeably, ITIM-bearing molecules recruit fewer phosphatases in vivo than bind in vitro to corresponding pIT-IMs. Thus, when expressed in mast cells and coaggregated with ITAM-bearing high affinity IgE receptors (Fc⑀RI), Fc␥RIIB were found to recruit SHIP1 but not SHP-1 or SHP-2 in vivo (28).
We investigated here the molecular basis of the recruitment of SHIP1 by Fc␥RIIB1. This question was addressed by a mutational analysis of the YϪ2 residue, which determines the recruitment of SHPs, and of the Yϩ1 and Yϩ2 residues, which contribute to the binding of SH2 domains (29). These residues of the Fc␥RIIB ITIM were substituted for corresponding residues of the N-terminal KIR2DL3 ITIM (KIR N-ITIM), which is unable to bind SHIP1. Conversely, the same residues of the KIR N-ITIM were substituted for corresponding residues of the Fc␥RIIB ITIM. The properties of wild type (w.t.) and modified ITIMs were analyzed by examining the in vitro binding of SH2 domain-bearing phosphatases to pITIM peptides and the in vivo recruitment of these phosphatases by Fc␥RIIB1 bearing corresponding ITIMs, when expressed in mast cells or in B cells. We found that, like a YϪ2 hydrophobic amino acid determines the affinity of ITIMs for SHP-1 and SHP-2, a symmetrical Yϩ2 hydrophobic amino acid determines the affinity of the Fc␥RIIB ITIM for SHIP1 and SHIP2. The presence of this residue is correlated with the ability of pITIMs to bind SHIP1/2. The Fc␥RIIB ITIM therefore contains two distinct though overlapping binding sites, for SHPs and for SHIPs, respectively, the SHIP-binding site being preferentially used in vivo.

EXPERIMENTAL PROCEDURES
Cells-RBL-2H3 were cultured in Dulbecco's modified Eagle's medium or RPMI supplemented with 10% fetal calf serum, 100 IU/ml penicillin, 100 g/ml streptomycin, and 2 mM L-glutamine. IIA1.6 cells were cultured in the same RPMI-based culture medium supplemented with 0.5 M 2-mercaptoethanol and 2 mM sodium pyruvate. Culture reagents were from Life Technologies, Inc.
cDNA Constructs-Mutant Fc␥RIIB were constructed in two steps. Sequences encoding the N-terminal part of the IC domain were first amplified using a 5Ј primer that hybridized with sequences encoding the nine N-terminal amino acids (Primer N): AAG AAA AAG CAG GTA CCA GCT CTC CCA, containing a KpnI site (underlined) and a 3Ј primer encoding the mutation to introduce. Sequences encoding the C-terminal part of the IC domain were amplified using a 5Ј primer encoding the mutation and a 3Ј primer that hybridized with sequences encoding the five C-terminal amino acids and the first 16 nucleotides of the 3Ј-untranslated sequences (Primer C): GAG ACA CTA GAG CTC GGC CTT TCT GGC TTG C, containing a SacI site (underlined). The two overlapping PCR fragments were used as templates to amplify the whole mutated sequence using Primers N and C. The resulting PCR product was either used as such or used as a template to introduce additional mutations. cDNA templates and corresponding primers used are listed below. When sense and corresponding antisense primers are complementary, only the sense primer is indicated.
The The resulting mutated fragments were cloned at KpnI and SacI sites into a NT vector containing sequences encoding the Fc␥RIIB EC and TM domains under the control of the SR␣ promoter as described (30). All amplified fragments were sequenced on the two strains.
cDNA sequence encoding the TM and IC domains of human KIR2DL3 were amplified from the p58.183 KIR2DL3 cDNA (31) by PCR with the following primers: sense, CCC AGA CAG GTA CCT GTT CTG ATT GGG ACC, and antisense, CTG ACT GTG GAG CTC ATG GGC AGG, for KIR2DL3. KpnI (GGTACC) and SacI (GAGCTC) sites are underlined. PCR products were inserted into an expression cassette under the control of the SR␣ promoter in pBR322, containing a resistance gene to neomycin (NT-neo) and the EC domain of Fc␥RIIB.
Transfectants-cDNAs were stably transfected in RBL-2H3 and IIA1.6 cells by electroporation. Transfectants were selected with neomycin (Life Technologies, Inc.) and cloned as described (30,32). The expression of recombinant receptors on clones remained stable over the duration of experiments. Several clones of each transfectant were used and gave similar results.
GST Fusion Proteins-cDNA encoding the SH2 domain of SHIP1 was amplified by PCR, using as a template cDNA generated from RNA extracted from RBL-2H3 cells. The following primers were used: sense, 5Ј-CTG ACC CAG TCT AGA GGA TCC ATG CCT GCC ATG GTC CCT G-3Ј; antisense, 5ЈGAC ACC TCG AGC TCT CAG GGA GGC AGC TCA-3Ј. The sequence was checked on the two strands by dideoxynucleotide sequencing. The SHIP1 SH2 domain cDNA and the SHP-1 SH 2 domain cDNA were inserted into the pGEX-4T-2 vector (Amersham Pharmacia Biotech) and transfected into DH5-␣ Escherichia coli. All GST fusion proteins were produced in DH5-␣ E. coli following isopropyl-1-thio-␤-D-galactopyranoside induction, purified on glutathione-agarose (Sigma), and analyzed by SDS-PAGE. Soluble SH2 domain-containing GST fusion proteins were eluted from glutathione-agarose beads with a solution of 50 mM Tris, 25 mM glutathione, pH 8.0.
ITIM Peptides and in Vitro Binding of Phosphatases-Biotinylated ITIM peptides were purchased from Neosystems (MPS, San Diego, CA) or from Sigma-Genosys (The Woodlands, TX). They were coupled to streptavidin-agarose beads. Beads were saturated with 1 mg/ml Dbiotin, washed in lysis buffer, and incubated for 2 h in lysates from 1 ϫ 10 7 cells or with SH2 domain-containing GST fusion proteins. Beads were washed and boiled in sample buffer. Eluted material was fractionated by SDS-PAGE, transferred onto an Immobilon-P membrane, and Western blotted with anti-phosphatase antibodies or anti-GST antibodies.
Surface Plasmon Resonance Analysis-Surface plasmon resonance measurements were performed on a BIAcore apparatus (BIAcore, Uppsala, Sweden). Before use, GST-SHIP1 SH2 fusion proteins were dialyzed in HBS buffer, pH 7.4 (10 mM HEPES, 150 mM NaCl, 3.4 mM EDTA). Protein purity was assessed by 12.5% SDS-PAGE and Coomassie blue staining. Running buffer consisted of HBS buffer supplemented with 0.05% surfactant P20. Biotinylated phosphopeptides were immobilized on streptavidin microchips (Sensorchip SA, BIAcore). The measurement of phosphorylated peptide binding to GST-SHIP1 SH2 fusion protein was performed at a constant 30 l/min flow rate. The regeneration was performed using HBS buffer supplemented with 0.02% SDS. k off and k on determination were performed using the BIAevaluation 3.0 software. The equilibrium dissociation constants K D were calculated as the k off /k on ratio.

RESULTS
The Mutation of the Y(ϩ1ϩ2) Serine and Leucine into Alanine and Glutamine in Fc␥RIIB ITIM Abrogates the Recruitment of SHIP1 by Phoshorylated Fc␥RIIB1-To identify ITIM amino acids that determine the in vivo recruitment of SH2 domain-containing phosphatases by Fc␥RIIB, mutations were introduced in cDNA encoding murine Fc␥RIIB1 by exchanging nucleotides encoding the serine and leucine residues, at positions Yϩ1 and Yϩ2, for nucleotides encoding alanine and glutamine that are present at the same positions in the N-terminal ITIM of KIR2DL3 (SL (ϩ1ϩ2) AQ Fc␥RIIB1). Nucleotides encoding the isoleucine, at position YϪ2 were or not also exchanged for nucleotides encoding alanine (I (Ϫ2) A Fc␥RIIB1). w.t. and mutated cDNAs were stably transfected in the rat mastocytoma cells RBL-2H3 (Fig. 1A) and in the Fc␥RIIBnegative mouse lymphoma B cells IIA1.6 ( Fig. 1B), which constitutively express Fc⑀RI and B cell receptors for antigen (BCR), respectively. Recombinant Fc␥RIIB1 were coaggregated with Fc⑀RI in RBL transfectants sensitized with mouse IgE anti-dinitrophenyl, incubated with F(abЈ) 2 fragments of the rat anti-mouse Fc␥RIIB mAb 2.4G2, and stimulated with TNP-MAR F(abЈ) 2 as described (13). Recombinant Fc␥RIIB1 were coaggregated with BCR in IIA1.6 transfectants stimulated with intact RAM IgG.
These results indicate 1) that the conservation of SL (ϩ1ϩ2) is critical for Fc␥RIIB1 to recruit SHIP1 in vivo, 2) that, although it prevents the recruitment of SHIP1, the SL (ϩ1ϩ2) AQ mutation does not confer Fc␥RIIB1 the ability to recruit SHP-1 or SHP-2, and 3) that I (Ϫ2) plays no role in the recruitment of SHIP1 by Fc␥RIIB1.
The Substitution of the Y(ϩ1ϩ2) Serine and Leucine for Alanine and Glutamine in Fc␥RIIB pITIM Peptides Abrogates the Binding of SHIP1-The specificity of tyrosyl-phosphorylated w.t. or modified peptides corresponding to the Fc␥RIIB ITIM was analyzed by examining their ability to bind SH2 domain-containing phosphatases when coated onto agarose beads and incubated with a lysate of RBL-2H3 cells. Neither w.t. (Fig. 2A) nor modified peptides (not shown) bound detectable amounts of SHP-1, SHP-2, or SHIP1 when nonphosphorylated. As observed previously (17,19), tyrosyl-phosphorylated peptides corresponding to the Fc␥RIIB ITIM bound SHP-1, SHP-2, and SHIP1. The substitution of the serine and leucine residues, at positions Yϩ1 and Yϩ2, for alanine and glutamine, respectively (SL (ϩ1ϩ2) AQ), abrogated the ability of phosphopeptides to bind SHIP1 but had no effect on their ability to bind SHP-1 and SHP-2. As previously reported (27), the substitution of the isoleucine, at position YϪ2, for an alanine (I (Ϫ2) A), abrogated the ability of phosphopeptides to bind SHP-1 and SHP-2 but had no effect on their ability to bind SHIP1. The substitution of I (Ϫ2) and SL (ϩ1ϩ2) for alanine, alanine, and glutamine, respectively (I (Ϫ2) A ϩ SL (ϩ1ϩ2) AQ), abrogated the ability of phosphopeptides to bind all three phosphatases ( Fig. 2A).
The binding of phosphopeptides to phosphatases present in a cell lysate may be direct or indirect via intracellular intermediates that could possibly bridge the two molecules. To exclude the latter possibility, we examined the in vitro binding of soluble GST fusion proteins containing either the single SH2 domain of SHIP1 (GST-SHIP1 SH2) or the two SH2 domains of SHP-1 (GST-SHP-1 SH2s) to agarose beads coated with w.t. or SL (ϩ1ϩ2) AQ Fc␥RIIB pITIM. As revealed by Western blotting with anti-GST antibodies, w.t. Fc␥RIIB pITIM bound both GST fusion proteins, whereas SL (ϩ1ϩ2) AQ Fc␥RIIB pITIM bound GST-SHP-1 SH2s but not GST-SHIP1 SH2 (Fig. 2B). Biacore analysis of the interactions between w.t. and SL (ϩ1ϩ2) AQ Fc␥RIIB pITIMs with GST-SHIP1 SH2 confirmed the requirement of SL (ϩ1ϩ2) for Fc␥RIIB pITIM to bind the SH2 domain of SHIP1 (Fig. 2C) and permitted to quantify the affinity of the interaction. The SL (ϩ1ϩ2) AQ substitution resulted in a 118-fold reduction of the k on , a 34-fold increase of the k off and, as a consequence, a 35-fold increase of the K D (Table I). SL (ϩ1ϩ2) , but not I (Ϫ2) , therefore appear critical for the in vitro binding of SHIP1 to the Fc␥RIIB pITIM. Conversely, I (Ϫ2) , but not SL (ϩ1ϩ2) determines the in vitro binding of SHP-1.

The Mutation of Y(ϩ1ϩ2) Residues into Serine and Leucine Confers Fc␥RIIB1
Bearing a KIR N-ITIM the Ability to Recruit SHIP1-To complement the results obtained using loss of function mutations made in Fc␥RIIB1 with gain of function mutations, we replaced the Fc␥RIIB1 ITIM by w.t. or mutated KIR N-ITIM. The KIR N-ITIM was mutated by exchanging nucleotides encoding the alanine and glutamine residues at positions Yϩ1 and Yϩ2 for nucleotides encoding serine and leucine (AQ (ϩ1ϩ2) SL) that are present at the same positions in the Fc␥RIIB ITIM. Nucleotides encoding the valine, at position YϪ2, were or not also exchanged for nucleotides encoding alanine (V (Ϫ2) A) (Fig. 3A). w.t. and mutated cDNAs were stably transfected in RBL-2H3 cells (Fig. 3B) and in IIA1.6 cells (Fig.  3C), and they were coaggregated with Fc⑀RI or with BCR under the same conditions as in Fig. 1.
When coaggregated with Fc⑀RI in RBL cells (Fig. 3B) or with BCR in IIA1.6 cells (Fig. 3C), all chimeras became tyrosylphosphorylated. As expected, the substitution of the Fc␥RIIB ITIM for the KIR N-ITIM abrogated the coprecipitation of SHIP1. Unexpectedly, however, it did not confer the chimera an ability to coprecipitate SHP-1 or SHP-2. Under the same conditions, a chimera made of the EC domain of Fc␥RIIB and of the TM and IC domains of KIR2DL3, that was inducibly tyrosyl-phosphorylated when coaggregated with Fc⑀RI, in RBL cells (Fig. 3B) or with BCR in IIA1.6 cells (Fig. 3C), recruited SHP-1 and SHP-2 but not SHIP1. Remarkably, the AQ (ϩ1ϩ2) SL mutation enabled the Fc␥RIIB chimera bearing a KIR N-ITIM to coprecipitate SHIP1, and SHIP1 coprecipitation was not affected by a V (Ϫ2) A mutation (Fig. 3, B and C). The above results indicate that the substitution of the Fc␥RIIB ITIM for  that of the KIR N-ITIM, in Fc␥RIIB1, rendered the chimera unable to detectably recruit not only SHIP1 but also SHP-1 and SHP-2 and that an AQ (ϩ1ϩ2) SL mutation enabled the chimera to recruit SHIP1, whether the V (Ϫ2) residue was conserved or not.

The Substitution of the Y(ϩ1ϩ2) Alanine and Glutamine for Serine and Leucine Confers KIR N-pITIM
Peptides the Ability to Bind SHIP1-As observed previously (11,27), when incubated in cell lysates, tyrosyl-phosphorylated peptides corresponding to the KIR N-ITIM bound SHP-1 and SHP-2 but not SHIP1 in vitro, and the substitution of the valine, at position YϪ2, for an alanine (V (Ϫ2) A) abrogated the ability of KIR N-pITIM peptides to bind SHP-1 and SHP-2. The substitution of the alanine and glutamine, at positions Yϩ1 and Yϩ2 in the KIR N-ITIM, for corresponding residues of the Fc␥RIIB ITIM, i.e. serine and leucine (AQ (ϩ1ϩ2) SL), had no effect on the ability of phosphopeptides to bind SHP-1 and SHP-2 but conferred KIR N-pITIM peptides the ability to bind SHIP1 (Fig. 4A).
KIR N-pITIM-coated beads bound GST-SHP-1 SH2s but not GST-SHIP1 SH2 in vitro. The AQ (ϩ1ϩ2) SL substitution conferred these peptides the ability to bind also to GST-SHIP1 SH2 (Fig. 4B). When assessed by Biacore analysis, w.t. KIR N-pITIM had no measurable affinity for SHIP1 SH2. The AQ (ϩ1ϩ2) SL substitution conferred KIR N-pITIM an affinity for the SH2 domain of SHIP1 that was of the same order of magnitude as that of Fc␥RIIB pITIM ( Fig. 4C and Table I). An AQ (ϩ1ϩ2) SL substitution therefore conferred KIR N-pITIM a measurable affinity for the SH2 domain of SHIP1 in vitro.  Fig. 1A. RBL transfectants were incubated and challenged as in Fig. 1A. Cells were lysed, and protein G-Sepharose was used to precipitate 2.4G2-bound Fc␥RIIB chimeras. Immunoprecipitates were analyzed as in Fig. 1. WCL from RBL transfectants expressing Fc␥RIIB1 bearing a w.t. KIR N-ITIM is shown. RBL cells expressing Fc␥RIIB whose TM and IC domains were replaced by those of KIR2DL3 were used as positive controls for the recruitment of SHP-1 and SHP-2 (right panel). C, histograms show the expression of Fc␥RIIB chimeras in IIA1.6 cells, assessed by indirect immunofluorescence as in Fig. 1B. IIA1.6 transfectants were stimulated as in Fig. 1B. Cells were lysed, and 2.4G2coated Sepharose beads were used to precipitate Fc␥RIIB chimeras. Immunoprecipitates were analyzed as in Fig. 1. WCL from IIA1.6 transfectants expressing Fc␥RIIB1 bearing a w.t. KIR N-ITIM is shown. IIA1.6 cells expressing Fc␥RIIB whose TM and IC domains were replaced by those of KIR2DL3 were used as positive controls for the recruitment of SHP-1 and SHP-2 (right panel). SHP-2 is shown by an arrow .   FIG. 4. In vitro binding of phosphatases by w.t. and modified  KIR N-pITIM peptides. A and B, nonphosphorylated and phosphorylated peptides corresponding to the w.t. and modified KIR N-ITIM, bound to agarose beads, were incubated in RBL-2H3 cell lysates (A) or with GST-SHIP1 SH2 or GST-SHP-1 SH2s (B). Precipitated material was fractionated by SDS-PAGE, transferred onto Immobilon-P, and Western blotted with anti-SHIP1, anti-SHP-1, and anti-SHP-2 antibodies (A) or anti-GST antibodies (B). C, the measurement of phosphorylated peptide binding to GST-SHIP1 SH2 fusion protein was performed at a constant 30 l/min flow rate, using 75 resonance units of phosphorylated peptides immobilized on streptavidin microchips, as in Fig. 2C. Curves correspond to concentrations of GST-SHIP1 SH2 of 80, 40, 20, and 10 nM. Y(ϩ1ϩ2) Serine and/or Leucine Determines the Ability of pITIMs to Bind and to Recruit SHIP2-SHIP2, a new SH2 domain-containing inositol 5-phosphatase having been described while our work was in progress, we wondered whether SL (ϩ1ϩ2) might also determine the recruitment of SHIP2. Indeed, SHIP2 was recently found to coprecipitate with the Fc␥RIIB1Ј isoform (20). Fc␥RIIB1 mutants that were used in Fig. 1 were examined for their ability to recruit SHIP2 in vivo when coaggregated with BCR in IIA1.6 cells. w.t. Fc␥RIIB1 and I (-2) A Fc␥RIIB1 recruited SHIP2, but SL (ϩ1ϩ2) AQ Fc␥RIIB1 and I (-2) A ϩ SL (ϩ1ϩ2) AQ Fc␥RIIB1 did not (Fig. 5A). The same serine and/or leucine residues that determine the recruitment of SHIP1 are therefore also critical for Fc␥RIIB1 to recruit SHIP2 in vivo.

FIG. 3. In vivo recruitment of phosphatases by Fc␥RIIB bearing a w.t., AQ (؉1؉2) SL, V (؊2) A, or V (؊2) A ؉ AQ (؉1؉2) SL KIR N-ITIM in mast cells and B cells, and structure of corresponding
Agarose-bound w.t. and modified Fc␥RIIB pITIM or KIR N-pITIM peptides that were used in Figs. 2 and 4, respectively, were incubated in cell lysates, and the binding of SHIP2 was examined by Western blotting. w.t. Fc␥RIIB pITIM bound SHIP2 in a lysate of IIA1.6 cells (Fig. 5B) and in a lysate of RBL-2H3 cells (data not shown). SHIP2 binding was abrogated by a SL (ϩ1ϩ2) AQ substitution, but not by an I (Ϫ2) A substitution. Conversely, the w.t. or V (Ϫ2) A KIR N-pITIM did not bind SHIP2, and the AQ (ϩ1ϩ2) SL substitution conferred the KIR N-pITIM peptide the ability to bind SHIP2 (Fig. 5C). The in vitro binding of SHIP2 to the two pITIM peptides therefore depends on the presence of a serine and/or a leucine, between the tyrosine and the Yϩ3 leucine.
The Y(ϩ2) Leucine Is Sufficient to Determine the Ability of Fc␥RIIB and KIR N-ITIM pITIMs to Bind SHIP1 and SHIP2-To determine the respective roles of S (ϩ1) and L (ϩ2) in the ability of pITIMs to bind SHIP1 and SHIP2, they were individually substituted for A and Q respectively, in the Fc␥RIIB pITIM. Conversely, the A (ϩ1) and Q (ϩ2) , were individually substituted for S and L, respectively, in the KIR N-pITIM. The ability of SHIP1 and SHIP2 to bind to these peptides and, as controls, to pITIMs with the double substitution, was examined and compared with the ability of the same peptides to bind SHP-1 and SHP-2 in the same cell lysates. As expected, no mutation affected the ability of pITIMs to bind SHP-1 and SHP-2. Remarkably, the single L (ϩ2) Q substitution, but not the single S (ϩ1) A substitution, abrogated the binding of both SHIP1 and SHIP2 to Fc␥RIIB pITIM, as efficiently as the double SL (ϩ1ϩ2) AQ substitution. Conversely, the single Q (ϩ2) L substitution, but not the single A (ϩ1) S substitution, conferred KIR N-pITIM the ability to bind both SHIP1 and SHIP2, as efficiently as the double AQ (ϩ1ϩ2) SL substitution (Fig. 6). The same results were obtained with the same peptides and GST fusion proteins containing the SH2 domain of SHIP1 (Table I and data not shown). A single amino acid, L (ϩ2) , therefore determines the binding of SHIP1 and SHIP2 to Fc␥RIIB pITIM.

DISCUSSION
This work aimed at understanding the molecular bases that permit Fc␥RIIB to recruit SHIP1, which has been proposed to account for the negative regulatory properties of this receptor (19,26). Fc␥RIIB is the only described ITIM-bearing molecule that was formally demonstrated to recruit SHIP1 in vivo. Having noticed that Fc␥RIIB is also the only such receptor whose ITIM contains a serine and a leucine at positions Yϩ1 and Yϩ2, respectively (33), we replaced these two residues by corresponding residues of the KIR N-ITIM, which has no affinity for SHIP1 (27). We then determined which amino acid substitutions rendered Fc␥RIIB unable to bind SHIP1 and/or SHIP2. SHIP2 was indeed recently found to bind to the Fc␥RIIB ITIM (20). Conversely, we replaced the Fc␥RIIB ITIM by the KIR N-ITIM, and Fc␥RIIB ITIM residues whose mutation prevented Fc␥RIIB1 from binding SHIPs were reintroduced in the KIR N-ITIM. The ability of receptors bearing w.t. and mutated ITIMs to recruit phosphatases in vivo were studied, when coaggregated with Fc⑀RI in RBL cells or with BCR in IIA1.6 cells, as well as the ability of corresponding pITIMs to bind in vitro phosphatases or SH2 domains of phosphatases.
As observed previously by us (27) and by others (11), both the Fc␥RIIB pITIM and the KIR N-pITIM bound SHP-1 and SHP-2 when incubated in cell lysates. Likewise, both phosphorylated peptides bound GST fusion proteins containing the two SH2 domains of SHP-1. Among all amino acid substitutions studied here, only that of I (Ϫ2) , in the Fc␥RIIB ITIM, or of V (Ϫ2) , in the KIR N-ITIM, abrogated the binding of SHPs to corresponding phosphopeptides. To bind to pITIMs, SHPs therefore require a hydrophobic residue at position YϪ2. Despite their in vitro affinity for the Fc␥RIIB pITIM, SHPs were not detectably recruited by tyrosyl-phosphorylated Fc␥RIIB1 in vivo. This discrepancy could have several explanations. Even though we readily observed the in vivo recruitment of SHP-1 and SHP-2 by a Fc␥RIIB chimera bearing the IC domain of KIR2DL3 in both B cells and mast cells, we cannot exclude the possibilities that Fc␥RIIB1 did recruit SHPs below detection levels or that recruited SHPs were lost during coprecipitation. Two groups previously reported that Fc␥RIIB1 recruited low levels of SHP-1 following coaggregation with BCR in B cells (17,34). Alternatively, the in vivo recruitment of SHPs to the phosphorylated Fc␥RIIB ITIM might be prevented by non-ITIM sequences of Fc␥RIIB1. Supporting this possibility, the phosphorylated KIR N-ITIM recruited SHP-2 when kept in the KIR2DL3 IC domain (30) but not when transposed into the Fc␥RIIB1 IC domain. Another possibility could be that SHIPs and SHPs might compete for being recruited in vivo and that, for some reason, SHIP recruitment might be dominant over SHP recruitment. The absence of recruitment of SHPs by SL (ϩ1ϩ2) AQ Fc␥RIIB1 and by Fc␥RIIB1 bearing the KIR N-ITIM that did not bind SHIPs does not favor this interpretation. Finally, although both SHIPs and SHPs bound in vitro to 12-amino acid phosphopeptide-coated beads, SHIPs, which have one SH2 domain, might be recruited by Fc␥RIIB that have one ITIM, whereas SHPs, which have two SH2 domains, might need receptors that have two ITIMs, such as KIR2DL3, to be recruited. Supporting this possibility, all molecules bearing more than one ITIM that have been examined were found to recruit SHP-1 and/or SHP-2 in vivo.
As already known, the Fc␥RIIB pITIM, but not the KIR N-ITIM, bound SHIP1 in vitro. The Fc␥RIIB pITIM, but not the KIR N-ITIM, bound also SHIP2. The substitution of SL (ϩ1ϩ2) , in the Fc␥RIIB ITIM, for the AQ residues present in the KIR N-ITIM at the same positions, abrogated the binding of SHIP1 and SHIP2 to phosphopeptides. Conversely, the substitution of AQ (ϩ1ϩ2) for SL in the KIR N-ITIM, conferred the phosphopeptides with an ability to bind SHIP1 and SHIP2. The same results were obtained when the binding of GST-SHIP1 SH2 to the same phosphopeptides was examined. A single L (ϩ2) Q substitution in the Fc␥RIIB ITIM, was sufficient to abrogate the binding of SHIP1 and SHIP2, and a single Q (ϩ2) L substitution, in the KIR N-ITIM, was sufficient to enable the binding of these two phosphatases. These results demonstrate that the L (ϩ2) residue determines the ability of pITIMs to bind SHIPs. That the same residue determines the binding of the two phosphatases is surprising because the SH2 domains of SHIP1 and SHIP2 have a relatively low (54%) homology. Supporting our conclusion that L (ϩ2) , rather than S (ϩ1) , is critical for binding SHIPs, S (ϩ1) is conserved in ITIM-bearing molecules that are not known to bind SHIPs, whereas L (ϩ2) is present in the Fc␥RIIB ITIM only. Whether this conclusion can be extended to all ITIMs may be challenged by previous works reporting that phosphopeptides corresponding to the second ITIM of gp49B1 (35) or to the third ITIM of p91 (36) bound SHIP1 in vitro. These two ITIMs contain the VTYAQL sequence. We (this work and Ref. 27) and others (11), however, detected neither SHIP1 nor SHIP2 binding to the KIR N-pITIM, which contains the same VTYAQL sequence. Another molecule known to recruit SHIP1 in vivo is the erythropoietin receptor that becomes tyrosyl-phosphorylated upon binding of erythropoietin. A mutational analysis of the intracytoplasmic domain of this receptor, which contains eight potential tyrosyl-phosphorylation sites, recently showed that the truncation of a segment containing the FEYTIL sequence abrogated the binding of the receptor to GST-SHIP1 SH2 (37). Two sequences that bind the SH2 domain of SHIP1 therefore contain either a leucine (Fc␥RIIB) or an isoleucine (erythropoietin receptor) at position Yϩ2. These closely related residues are, with valine, the three most hydrophobic amino acids. To bind to pITIMs, SHIPs may therefore require a hydrophobic residue at position Yϩ2. Supporting this conclusion, all ITIMs present in KIRLs, PIR-B, gp49B1, SIRP␣, NKG2A, CD72, and ILTs/LIRs that do not recruit SHIP1 possess a hydrophilic Yϩ2 residue.
We (27) and others (11) reported previously that the YϪ2 isoleucine residue that is critical for SHP-1 and SHP-2 binding is not critical for SHIP-1 binding. We show here that this residue is irrelevant for SHIP2 binding too. We also show here that the Yϩ2 leucine that is critical for SHIP1 and SHIP2 binding is irrelevant for SHP-1 and SHP-2 binding, as well as S (ϩ1) . The YϪ1 threonine does not seem to be involved either (data not shown). The Fc␥RIIB core ITIM therefore contains two overlapping but distinct SH2 domain-binding sites, for SHIPs and for SHPs, respectively. We propose IxpYxxL as being the SHP-1/2-binding site and xxpYxLL as being the SHIP1/2-binding site in the Fc␥RIIB ITIM. Interestingly, two symmetrical hydrophobic residues, at positions YϪ2 and Yϩ2, therefore determine the binding of SHPs and SHIPs, respectively.
Fc␥RIIB1 recruited in vivo not only SHIP1, when coaggregated with Fc⑀RI in mast cells or with BCR in B cells, but also SHIP2, when coaggregated with BCR in B cells. Fc␥RIIB1 whose ITIM bore an SL (ϩ1ϩ2) AQ mutation lost their ability to recruit SHIP1 and SHIP2, when tyrosyl-phosphorylated following coaggregation with Fc⑀RI or with BCR. Conversely, a Fc␥RIIB chimera bearing a w.t. KIR N-ITIM failed to recruit SHIP1, and a chimera whose KIR N-ITIM bore a AQ (ϩ1ϩ2) SL mutation gained the ability to recruit SHIP1 when tyrosylphosphorylated following coaggregation with Fc⑀RI or with BCR. Our in vitro data may therefore account for the in vivo recruitment of SHIPs by Fc␥RIIB1, and they emphasize the mandatory role of ITIMs with an appropriate sequence. The ITIM may, however, not account alone for the in vivo recruitment of SHIPs by Fc␥RIIB1. The IC domain of Fc␥RIIB1 indeed contains three other tyrosines that are also likely to become phosphorylated during coaggregation with ITAM-bearing receptors. Recent works indicate that the tyrosine C-terminal to the ITIM, which is conserved in all three murine isoforms, is indeed phosphorylated in Fc␥RIIB1 and is required for this receptor to recruit SHIP1 (38): by binding the SH2 domain of Grb2 (39) whose C-terminal Src homology 3 (SH3) domain has an affinity for proline-rich sequences in SHIP1 (40), it stabilizes the recruitment of the phosphatase to the Fc␥RIIB pITIM. Interestingly, the SH3 domain of Grb2 has no affinity for SHIP2 (40), suggesting either that the Fc␥RIIB1 pITIM is sufficient for the recruitment of SHIP2 or that another molecule might play the same role for SHIP2 as Grb2 for SHIP1. That Fc␥RIIB can recruit also SHIP2 may diversify the negative regulatory properties of this receptor. SHIP1 and SHIP2 do not have identical substrates, and they may interact with different effector and adapter molecules in different cell types. Both hydrolyze phosphatidylinositol 3,4,5-trisphosphate, whereas only SHIP1 hydrolyzes inositol 1,3,4,5-tetraphosphate (40), which has been proposed to promote Ca 2ϩ influx (41). Both can bind the SH3 domains of Abl and Src via their proline-rich regions and the phosphotyrosine-binding domain of Shc via phosphorylated tyrosines, whereas only SHIP1 binds the SH3 domain of Grb2, and only SHIP2 binds the SH3 domain of Crk (40). SHIP1 is restricted to cells of the hematopoietic lineage, whereas SHIP2 is also expressed in nonhematopoietic cells (42). Fc␥RIIB might thus differentially use SHIP1 and/or SHIP2 in different cell types with possible different functional consequences.
In conclusion, we demonstrate here that the Fc␥RIIB ITIM contains two binding sites, for SHPs and for SHIPs whose binding depends on two hydrophobic residues, located at positions YϪ2 and Yϩ2, respectively, i.e. symmetrically N-and C-terminal to the phosphorylated tyrosine residue. Although both binding sites are functional in vitro, the SHIP-binding site is preferentially used in vivo in both mast cells and B cells.