CD22 Forms a Quaternary Complex with SHIP, Grb2, and Shc

CD22 is a cell surface molecule that regulates signal transduction in B lymphocytes. Tyrosine-phosphorylated CD22 recruits numerous cytoplasmic effector molecules including SHP-1, a potent phosphotyrosine phosphatase that down-regulates B cell antigen receptor (BCR)- and CD19-generated signals. Paradoxically, B cells from CD22-deficient mice generate augmented intracellular calcium responses following BCR ligation, yet proliferation is decreased. To understand further the mechanisms through which CD22 regulates BCR-dependent calcium flux and proliferation, interactions between CD22 and effector molecules involved in these processes were assessed. The adapter proteins Grb2 and Shc were found to interact with distinct and specific regions of the CD22 cytoplasmic domain. Src homology-2 domain-containing inositol polyphosphate-5′-phosphatase (SHIP) also bound phosphorylated CD22, but binding required an intact CD22 cytoplasmic domain. All three molecules were bound to CD22 when isolated from BCR-stimulated splenic B cells, indicating the formation of a CD22·Grb2·Shc·SHIP quaternary complex. Therefore, SHIP associating with CD22 may be important for SHIP recruitment to the cell surface where it negatively regulates calcium influx. Although augmented calcium responses in CD22-deficient mice should facilitate enhanced c-Jun N-terminal kinase (JNK) activation, BCR ligation did not induce JNK activation in CD22-deficient B cells. These data demonstrate that CD22 functions as a molecular “scaffold” that specifically coordinates the docking of multiple effector molecules, in addition to SHP-1, in a context necessary for BCR-dependent SHIP activity and JNK stimulation.

the cytoplasm, and the activation of mitogen-activated protein kinases (MAPKs). Upon cell activation, MAPKs translocate to the nucleus and activate specific transcription factors (6 -9). Subfamilies of MAPKs include the extracellular signal-regulated kinases (ERKs), c-Jun N-terminal kinases (JNKs), and p38, which become fully activated as a result of dual phosphorylation on single threonine and tyrosine residues (reviewed in Ref. 10). BCR-induced calcium mobilization is essential for JNK, but not ERK or p38, activation (10 -12). Also critical to each of these events in B cells are cell surface receptors such as CD22 that modify and provide a context for BCR signal transduction (13).
Critical interactions between CD22 and other effector molecules may provide a molecular explanation for augmented calcium responses yet decreased proliferation by CD22-deficient B cells in response to BCR ligation. Although decreased recruitment of SHP-1 to the plasma membrane due to the absence of CD22 could explain augmented calcium responses, evidence suggests that other molecules are also involved. SHP-1 and SHIP differentially down-regulate calcium responses by preventing release from intracellular stores and from the influx of extracellular free calcium, respectively (33). Like the killer cell inhibitory receptor class of molecules, which block the release of calcium from intracellular stores (34), it has been suggested that CD22 inhibits calcium release solely through SHP-1 activity (33). However, CD22 modulates the mobilization of both intracellular and extracellular calcium (27), suggesting that CD22 up-regulates both SHP-1 and SHIP activity. A SHIP⅐Grb2⅐Shc ternary complex forms following BCR ligation in primary B cells, which may be required for recruitment of SHIP to the plasma membrane where it down-regulates calcium mobilization (35). To evaluate further how CD22 regulates calcium mobilization, we have assessed the ability of SHIP to associate with CD22 in the current study. Since calcium mobilization is dramatically elevated in CD22-deficient B cells and calcium mobilization is essential for JNK activation (10 -12), we have also assessed whether CD22 loss enhances JNK activation following BCR ligation. These studies reveal that phosphorylated CD22 specifically associates with SHIP, Grb2, and Shc to form a quaternary complex. Intriguingly, despite augmented calcium mobilization in CD22-deficient B cells, JNK activation was selectively impaired, which may explain their decreased proliferative responses. Thus, adapter protein-mediated interactions between CD22 and other effector molecules are likely to contribute to the regulatory activities of CD22.

EXPERIMENTAL PROCEDURES
Mice-CD22-deficient (B6 ϫ 129) and wild type control mice (B6 ϫ 129) were generated by breeding mice heterozygous for the CD22 deficiency as described (24, 36). All mice were 2 months of age when used and were housed in a specific pathogen-free barrier facility. All studies and procedures were approved by the Animal Care and Use Committee of Duke University.
Biotin-conjugated peptides corresponding to regions of the mouse CD22 cytoplasmic domain were synthesized by the University of North Carolina Peptide Synthesis Facility (Chapel Hill, NC) in both tyrosinephosphorylated and non-phosphorylated forms.
Antibodies-Hybidomas producing CD22-specific monoclonal antibodies were generated by the fusion of NS-1 myeloma cells with spleen cells from a CD22-deficient mouse that was immunized three times with the GST-CD22cyt fusion protein in adjuvant. Supernatant fluid from four hybridomas reacted with GST-CD22cyt in enzyme-linked immunosorbent assays but not GST alone. Each anti-CD22 antibodyproducing hybridoma was subcloned twice by limiting dilution and used to generate tissue culture supernatant fluid for these studies. Antibody isotypes were determined using a mouse monoclonal antibody isotyping kit (Amersham Pharmacia Biotech).
Protein Precipitation and Immunoblot Analysis-B cells were purified from single cell splenocyte suspensions by removing T cells with anti-Thy1.2 antibody-coated magnetic beads (Dynal Inc., Lake Success, NY). B cell suspensions were analyzed by immunofluorescence staining with flow cytometric analysis following isolation and were always Ͼ94% B220 ϩ . The B cells were resuspended (2 ϫ 10 7 /ml) in RPMI 1640 medium containing 5% fetal calf serum. Following incubation for 5 min at 37°C, B cells were cultured with media alone or containing F(abЈ) 2 fragments of goat anti-mouse IgM antibodies (40 g/ml) at 37°C and then lysed in buffer containing 1% Nonidet P-40, 150 mM NaCl, 50 mM Tris-HCl (pH 8.0), 1 mM sodium orthovanadate, 2 mM EDTA, 50 mM NaF, and protease inhibitors as described (37). For the detection of MAPK subfamily kinase activation, B cell lysates were subjected to SDS-polyacrylamide gel electrophoresis (PAGE) followed by electrophoretic transfer to nitrocellulose with indirect immunoblotting using antibodies against active forms of MAPK subfamily members.
For immunoprecipitation studies using antibodies reactive with proteins of interest, B cell lysates were pre-cleared using protein G-Sepharose beads (50 l; Amersham Pharmacia Biotech) and then incubated with 5 g/ml antibody for 2 h at 4°C, followed by the addition of protein G-Sepharose beads (50 l) for an additional 2 h. The beads were washed five times with 1 ml of lysis buffer and resuspended in 100 l of SDS-PAGE sample buffer followed by boiling for 5 min. For coprecipitation studies using CD22 fusion proteins, B cell lysates were precleared using glutathione-Sepharose beads (50 l; Amersham Pharmacia Biotech) and then incubated with 25 g/ml GST-CD22cyt or GST-CD22cyt(Tyr(P)) for 2 h at 4°C, followed by the addition of glutathione-Sepharose beads (50 l) for an additional 2 h. The beads were washed five times with 1 ml of lysis buffer and resuspended in 100 l of SDS-PAGE sample buffer followed by boiling for 5 min.
For coprecipitation studies using CD22 peptides, B cell lysates were pre-cleared with avidin-agarose beads (50 l; Pierce) and then incubated with each peptide (20 M) for 2 h at 4°C, followed by the addition of avidin-agarose beads (50 l) for an additional 2 h. The beads were washed five times with 1 ml of lysis buffer and resuspended in 100 l of SDS-PAGE sample buffer followed by boiling for 5 min.
Precipitated proteins were subjected to SDS-PAGE with subsequent electrophoretic transfer to nitrocellulose membranes. After blocking the membranes with buffer containing 2% (w/v) bovine serum albumin, protein presence was detected either by direct immunoblotting with horseradish peroxidase-conjugated anti-phosphotyrosine antibody or by indirect immunoblotting using antibodies reactive with the proteins of interest. Incubation periods for primary and secondary antibodies were 1 h at 25°C with extensive washing of the membranes following each step. Immunoblots were developed using enhanced chemiluminescence kits from Pierce. To verify equivalent amounts of protein in each lane, the same blots were stripped and reprobed with antibodies against proteins of interest. Relative band intensities of MAPK immunoblots were determined using NIH Image software (version 1.60).
In Vitro Lyn Kinase Assays-Lyn phosphorylation of CD22 peptides was assessed using in vitro kinase reactions. Each reaction mixture (30 l) contained CD22 non-phosphorylated peptides or control (phosphorylated) peptides at the indicated concentrations, Src kinase reaction buffer (10 l), a 1:10 dilution of Src manganese/ATP mixture, 2 units of purified Lyn kinase (all from Upstate Biotechnology, Inc.), and 10 Ci of [␥-32 P]ATP (ICN Biomedicals, Inc., Costa Mesa, CA). The reactions were carried out for 20 min at 25°C before termination using 30 l of 4.75 M guanidine HCl. 20 l of each reaction mixture was then spotted onto SAM TM Biotin Capture Membranes (Promega). The membranes were washed three times with 2 M NaCl, once with 1 M NaCl ϩ 0.75% phosphoric acid, and once with 2 M NaCl containing 1% Tween 20. Radioactivity was quantified by scintillation counting.

Characterization of GST-CD22cyt Fusion Proteins and
Anti-CD22 Antibody-To evaluate effector molecule interactions with the CD22 cytoplasmic domain, a fusion protein consisting of GST and the entire 140-amino acid cytoplasmic domain of mouse CD22 was generated (Fig. 1). The fusion protein was produced in bacteria in non-phosphorylated (GST-CD22cyt) and tyrosine-phosphorylated (GST-CD22cyt(Tyr(P))) forms as described under "Experimental Procedures." Migration of the GST-CD22cyt(Tyr(P)) protein in SDS-PAGE gels was significantly slower than for GST-CD22cyt protein due to tyrosine phosphorylation (Fig. 2C). In addition, peptides containing either single tyrosine motif or tandem tyrosine motif sequences corresponding to specific regions of the CD22 cytoplasmic domain were used in these studies in non-phosphorylated and tyrosine-phosphorylated forms ( Fig. 1).
Since the two currently reported anti-mouse CD22 monoclonal antibodies react with the extracellular domain of mouse CD22 and do not react with CD22 in Western blots, new antibodies were generated that react with CD22 under reducing conditions. Monoclonal antibody secreting hybridomas were generated by immunizing CD22-deficient mice with the GST-CD22cyt fusion protein. Of 546 hybridomas screened, four secreted IgG1 antibodies that reacted with the GST-CD22cyt fusion protein in enzyme-linked immunosorbent assays but not GST alone (31 antibodies): MB22-1, MB22-2, MB22-3, and MB22-4. One of the four antibodies, MB22-1 reacted predominantly with a band of M r 145,000 and to a lesser extent with a band of M r 130,000 in immunoblots of mouse splenocytes from wild type but not CD22-deficient mice ( Fig. 2A). These bands appear to represent CD22 with differences in post-translational processing, and the M r 145,000 protein was further resolved into two similar-sized bands in some experiments. The MB22-1 antibody effectively immunoprecipitated the M r 145,000 isoform of CD22 from B cell lysates as detected by immunoblotting with an anti-phosphotyrosine antibody (Fig.  2B). The MB22-1 antibody did not immunoprecipitate phosphoproteins from CD22-deficient B cells (Fig. 2B). Also, MB22-1 did not react with CD22 in lysates of human B cell lines (data not shown). The MB22-1 antibody reacted equally with the GST-CD22cyt and GST-CD22cyt(Tyr(P)) fusion proteins bound to nitrocellulose but did not react with a CD19-GST fusion protein (Fig. 2C).
CD22 Forms a Complex with SHIP, Grb2, and Shc Following BCR Stimulation-Since CD22 interacts with Grb2 following BCR stimulation in K46 cells (29), and primary B cells generate a SHIP⅐Grb2⅐Shc ternary complex following BCR ligation (35), we assessed whether CD22 could serve as a membrane-proximal docking site for the SHIP⅐Grb2⅐Shc complex. The ability of the SHIP⅐Grb2⅐Shc complex to associate with CD22 in primary cells was evaluated by immunoprecipitation of SHIP, Grb2, and Shc from cell lysates of resting and anti-IgM activated splenic B cells, followed by detection of CD22 in the complexes by immunoblotting. SHIP, Grb2, and Shc each coprecipitated CD22 at low levels from resting B cells but at significantly higher levels following BCR stimulation for 5 min (Fig. 3, A-C). Equivalent levels of SHIP and Grb2 were precipitated from resting and activated B cells as revealed by stripping the immunoblots and reprobing with anti-SHIP or anti-Grb2 antibodies (Fig. 3, A and B). The anti-Shc antibody used in these studies was inadequate for reprobing Western blots, so the cell lysates were reprecipitated with an anti-Vav antibody and FIG. 2. Characterization of the CD22-specific MB22-1 monoclonal antibody. A, purified splenic B cells (1 ϫ 10 7 cells/sample) from wild type or CD22-deficient mice were incubated in media alone or with anti-IgM antibodies before detergent lysis. Whole cell lysates (1 ϫ 10 6 cell equivalents) were subjected to SDS-PAGE, electrophoretically transferred to nitrocellulose membranes, and immunoblotted with a 1:50 dilution of MB22-1 hybridoma culture supernatant fluid. Antibody reactivity was revealed using a horseradish peroxidase-conjugated secondary antibody with detection by enhanced chemiluminescence. B, splenic B cell lysates were prepared as in A with immunoprecipitation (IP) using a 1:10 dilution of MB22-1 hybridoma supernatant fluid followed by capture with protein G-Sepharose beads. Precipitated proteins were subjected to SDS-PAGE, transferred to nitrocellulose, and subsequently immunoblotted with an anti-phosphotyrosine monoclonal antibody. C, GST-CD22 cytoplasmic domain fusion proteins or a control GST-CD19 cytoplasmic domain fusion protein were subjected to SDS-PAGE (5 g per lane) and transferred to nitrocellulose membranes. The membrane was first immunoblotted using MB22-1 hybridoma supernatant fluid. The membrane was subsequently stripped and reprobed with anti-GST antibody as indicated. Heavy bars indicate peptides with single tyrosine residues, and thin lines indicate dual tyrosine-containing peptides. CD22 sequence and numbering is as described (16,59), with tyrosines indicated in bold type.
immunoblotted with the anti-Vav antibody to verify equal protein concentrations in the cell lysates (Fig. 3C). The GTP exchange factor SOS, an activator of the ERK pathway, has also been reported to interact with CD22 following BCR stimulation in K46 cells (29). However, SOS did not coprecipitate CD22 at detectable levels in primary B cells, so the anti-SOS antibody was used as a negative control in these experiments (Fig. 3). Although it is possible that SOS may interact with CD22 in primary splenic B cells at later time points of activation, these results suggest that a fundamental difference in signaling events exists between primary splenic B cells and the K46 cell line.
The ability of CD22 to coprecipitate SHIP, Grb2, and Shc from lysates of resting and anti-IgM activated splenic B cells was also assessed by immunoblotting. CD22 immunoprecipitated from activated B cell lysates coprecipitated SHIP, but SHIP was not coprecipitated from lysates of unstimulated B cells (data not shown). Shc and Grb2 interactions with CD22 could not be effectively evaluated in this manner since the MB22-1 antibody heavy and light chains overlapped with these molecules in the immunoblots. Nonetheless, CD22 phosphorylation resulted in the formation of a CD22⅐Grb2⅐Shc⅐SHIP complex that may enhance SHIP, Grb2, and Shc recruitment to the plasma membrane.
Effector Molecule Interactions with the CD22 Cytoplasmic Domain-To verify further that SHIP, Grb2, and Shc interact with the CD22 cytoplasmic domain, the ability of purified CD22 fusion proteins to precipitate these proteins from lysates of resting and activated B cells was assessed. In addition, the ability of the fusion proteins to bind SHP-1 was evaluated. Proteins that coprecipitated with the non-phosphorylated (GST-CD22cyt) and tyrosine-phosphorylated (GST-CD22cyt-(Tyr(P))) fusion proteins were subjected to SDS-PAGE with subsequent immunoblotting using SHIP-, Grb2-, Shc-, or SHP-1-specific antibodies. As expected, SHP-1 interacted strongly with the GST-CD22cyt(Tyr(P)) fusion protein but not with GST-CD22cyt (Fig. 4A). Also, the p145, p135, and p110 isoforms of SHIP were each precipitated by GST-CD22cyt(Tyr(P)) ( Fig. 4A) at levels proportional to their intracellular concentrations in B cells (Fig. 5A) but not by GST-CD22cyt (Fig. 4A). Similar levels of SHIP were precipitated from resting and anti-IgM-stimulated primary B cells using the GST-CD22cyt(Tyr(P)) protein. Grb2 and Shc each bound GST-CD22cyt(Tyr(P)), although Shc also bound GST-CD22cyt at low levels (Fig. 4A).
Only the p66 isoform of Shc, but not the p52 and p46 isoforms, could be definitively evaluated because the antibodies used for Shc immunoblotting reacted weakly with the 45-55-kDa GST-CD22 fusion proteins (data not shown). As with SHIP, the levels of Grb2 and Shc binding by GST-CD22cyt(Tyr(P)) proteins were similar in lysates from both resting and activated cells. None of these proteins was coprecipitated with free GST (data not shown). These data indicate that bacterial phosphorylation of the fusion protein occurs at the appropriate tyrosine residues to recapitulate effector molecule interactions with CD22 found in primary B cells.
As was the case for CD22 precipitated from primary B cells, SOS did not coprecipitate with GST-CD22cyt(Tyr(P)) protein, although SOS was readily detectable in whole cell lysates (Fig.  4B). BLNK, an adapter protein involved in BCR-mediated activation of the JNK pathway (38), was not coprecipitated at detectable levels in these experiments, although BLNK was only detectable at low levels in whole cell lysates (Fig. 4B). Equivalent levels of fusion protein were used for each precipitation as demonstrated by reprobing the membranes with anti-GST antibodies (Fig. 4, A and B). Therefore, SHIP, Grb2, and Shc appear to bind CD22 specifically through interactions facilitated by phosphorylated tyrosine residues.
Mapping of SHIP, Grb2, and Shc Interactions with the CD22 Cytoplasmic Domain-The SHIP-and Shc-binding sites in the CD22 cytoplasmic domain were mapped using biotinylated peptides containing the six CD22 tyrosine residues either in a tandem array or as single tyrosine motifs as shown in Fig. 1. SHIP did not interact at detectable levels with phosphorylated or nonphosphorylated CD22 in either the tandem tyrosine (Fig.  5A) or single tyrosine motif formats (data not shown). Grb2 bound to a tandem tyrosine phosphopeptide containing Tyr-807 (Fig. 5B), the previously described Grb2-binding site (29). Grb2 did not interact at detectable levels with the Tyr-807/Tyr-822 phosphopeptide, indicating that additional sequences present in the Tyr-796/Tyr-807 phosphopeptide may facilitate Grb2 binding, or Grb2 binding to the Tyr-807/Tyr-822 phosphopeptide may be displaced by another molecule, possibly SHP-1 that interacts strongly with Tyr-822 (Fig. 5C). Shc bound tandem tyrosine phosphopeptides Tyr-752/Tyr-762 and Tyr-807/Tyr- FIG. 3. Association of CD22 with SHIP, Grb2, and Shc. Purified splenic B cells (5 ϫ 10 7 cells/sample) from wild type mice were incubated in media alone or with anti-IgM antibodies and lysed. Lysates were then immunoprecipitated (IP) with 5 g of anti-SHIP, anti-Grb2, anti-Shc, or anti-SOS antibodies followed by capture with protein G-Sepharose beads. Immunoprecipitated proteins were then subjected to SDS-PAGE and electrophoretic transfer to nitrocellulose membranes. Whole cell lysates (1 ϫ 10 6 cell equivalents) were also subjected to SDS-PAGE and transfer to nitrocellulose as controls. The membranes were subsequently immunoblotted using a 1:50 dilution of MB22-1 hybridoma culture supernatant fluid. The membranes were also directly probed or stripped and reprobed with antibodies against immunoprecipitated proteins as indicated. Data are representative of three independent experiments that produced similar results. 822 exclusively and at high levels (Fig. 5B). Shc bound single phosphopeptides of CD22 corresponding to Tyr-762 and Tyr-822 (Fig. 5C). Grb2 and Shc did not interact with non-phosphorylated CD22 peptides at significant levels. That Shc binding to the Tyr-822 phosphopeptide was significantly weaker than binding to the Tyr-762 phosphopeptide suggests that additional sequence requirements present in the Tyr-807/Tyr-822 tandem tyrosine phosphopeptide may be necessary for optimal Shc interactions with this region of CD22. These results demonstrate that as for Grb2, the interaction of Shc with CD22 was specific for particular phosphotyrosine domains of CD22. Moreover, SHIP binding to CD22 either requires a larger region of the CD22 cytoplasmic domain than was represented by the CD22 phosphopeptides used, or SHIP interactions with CD22 are indirect and mediated through bridging by Grb2 and Shc.
Previous studies have shown that SHP-1 interacts with phosphorylated Tyr-842 of CD22 (19,23). However, SHP-1 did not bind the Tyr-822/Tyr-842 phosphopeptide with tandem tyrosines in our experiments (Fig. 5A), although SHP-1 was precipitated by the Tyr-842 single tyrosine motif peptide (Fig.  5C). Since the Tyr-822/Tyr-842 phosphopeptide lacks the Cterminal histidine of CD22 that was present in the Tyr-842 single tyrosine motif peptide, this histidine may be essential for stable SHP-1 binding to CD22. SHP-1 also bound strongly to two other regions of CD22 containing phosphorylated Tyr-796 and Tyr-822, as previously reported (19,23). (39,40). Therefore, whether Lyn could phosphorylate the CD22 tyrosine residues bound by Grb2 and Shc was assessed. In vitro kinase assays were performed using purified Lyn kinase and single and tandem tyrosine motif CD22 peptides. Each kinase reaction contained equimolar concentrations of tyrosines, so that the degree of phosphorylation between CD22 peptides containing single or tandem tyrosines could be compared. CD22 peptides containing single tyrosines were phosphorylated by Lyn with the exception of peptide Tyr-752 that was not phosphorylated above background levels (Fig. 6). Accordingly, CD22 Tyr-752 has not been shown to mediate intermolecular interactions. Thus, the Grb2-and Shc-binding sites on CD22 are potential substrates for Lyn phosphorylation in vivo.

Sites of CD22 Phosphorylation by Lyn Kinase-Lyn expression is required for a majority of CD22 phosphorylation in vivo
Each CD22 tandem tyrosine motif peptide was phosphorylated by Lyn since each contained potential Lyn phosphoryla-  Fig. 1. Whole cell lysates from purified splenic B cells (2 ϫ 10 7 cells/sample) were incubated with the indicated non-phosphorylated (NP) or tyrosine-phosphorylated (P) CD22 peptides prior to capture with avidin-agarose beads. Whole cell lysates (1 ϫ 10 6 cell eq) were also subjected to SDS-PAGE and transferred to nitrocellulose as controls where indicated. Coprecipitated proteins were subjected to SDS-PAGE and immunoblotting with antibodies against Grb2, Shc, SHIP, SHP-1, or Lyn. Data are representative of three independent experiments that produced similar results.

FIG. 4. Coprecipitation of effector molecules by CD22 cytoplasmic domain fusion proteins.
Purified splenic B cells (2 ϫ 10 7 cells/sample) incubated in media or with anti-IgM antibodies were lysed and incubated with GST-CD22cyt or GST-CD22cyt(Tyr(P)) fusion proteins prior to capture with glutathione-Sepharose beads. Coprecipitated proteins were subjected to SDS-PAGE and electrophoretic transfer to nitrocellulose membranes. Immunoblotting was carried out using antibodies reactive with the indicated effector proteins. Data are representative of at least three independent experiments that generated similar results.
tion sites (Fig. 6). Strikingly, peptides Tyr-796/Tyr-807 and Tyr-807/Tyr-822 were phosphorylated at significantly higher levels than single tyrosine peptides corresponding to the same regions. This suggests that phosphorylation of specific tyrosine residues within these peptides may allow Lyn binding and the processive phosphorylation of the other tyrosine residue within the peptide. Therefore, the ability of Lyn to interact with CD22 tandem tyrosine phosphopeptides was examined. Lyn bound phosphopeptide Tyr-796/Tyr-807 at the highest level, followed by Tyr-807/Tyr-822, and then peptides Tyr-752/Tyr-762 and Tyr-822/Tyr-842 (Fig. 5B). Lyn also bound each of the nonphosphorylated peptides but at much lower levels than the phosphopeptides. Since these results correlate closely with the observed pattern of Lyn phosphorylation of each peptide (Fig.  6), these peptides may be phosphorylated at low levels by Lyn during coprecipitation experiments such that the observed associations of Lyn and Shc with non-phosphorylated CD22 peptides might actually represent phosphotyrosine-mediated interactions. Nonetheless, these results demonstrate that each CD22 peptide contains sites appropriate for Lyn phosphorylation and binding.
Impaired Activation of the JNK Signaling Pathway in CD22deficient Mice-Since CD22⅐Grb2⅐Shc⅐SHIP complex formation may contribute to the augmented calcium responses of CD22deficient B cells, and JNK activation is dependent upon calcium mobilization, JNK activation was assessed in CD22-deficient B cells. Activation of the ERK and JNK subfamilies of MAPKs was assessed in purified splenic B cells from wild type and CD22-deficient mice following BCR stimulation. The B cells were treated at predetermined time points with an optimal dose of F(abЈ) 2 anti-mouse IgM antibodies, and MAPK activation was then evaluated by immunoblotting using antibodies specific for the dually phosphorylated (fully active) forms of ERK and JNK MAPKs as described (41). In contrast to ERK1/2 activation, which exhibited normal kinetics and intensity of phosphorylation, JNK2 activation was severely impaired in CD22-deficient B cells (Fig. 7A). In fact, despite a moderately higher base-line level of dually phosphorylated JNK in CD22deficient B cells, JNK activation did not increase significantly above resting levels following BCR ligation (Fig. 7C), despite equivalent levels of cytoplasmic JNK2 and ERK2 in CD22deficient and wild type B cells (Fig. 7B). The JNK1 isoform of JNK was not significantly activated in either wild type or CD22-deficient mice (data not shown). Also, p38 was not significantly activated in primary B cells following BCR ligation (data not shown) as recently reported (42). Therefore, JNK activation was not up-regulated following BCR ligation despite augmented calcium responses in CD22-deficient B cells. DISCUSSION The current study demonstrates novel interactions between CD22 and effector molecules involved in the modulation of calcium flux and MAPK activation. Specifically, the Grb2 and Shc adapter proteins bound CD22 in primary B cells (Fig. 3), and their binding sites mapped to distinct regions of the CD22 FIG. 6. Lyn phosphorylation of CD22 cytoplasmic domain peptides. In vitro kinase reactions were carried out using the nonphosphorylated CD22 peptides described in Fig. 1, purified Lyn kinase, and [␥-32 P]ATP. Phosphorylated CD22 peptide Tyr-807/Tyr-822 was used as a negative control. Following a 20-min kinase reaction, the radioactivity incorporated into each peptide was quantified. Values represent the mean (ϮS.E.) of results from four independent experiments. * indicates values that were significantly different from control values, p Ͻ 0.01.

FIG. 7. BCR-induced MAPK activation in CD22-deficient B cells.
A, splenic B cells purified from CD22-deficient or wild type control littermates (1 ϫ 10 7 cells/sample) were incubated with either medium alone (time 0) or with anti-IgM antibodies for the indicated times and lysed. Subsequently, lysates of the B cells were subjected to SDS-PAGE and transferred to nitrocellulose membranes for subsequent immunoblotting with phospho-specific antibodies for the activated forms of the MAPKs. Membranes were stripped and reprobed with anti-ERK2 antibody to verify equivalent loading of proteins between samples. B, cytoplasmic ERK2 and JNK2 protein levels in purified B cells from spleens of wild type and CD22-deficient mice as determined by immunoblotting with the indicated antibodies. C, ERK2 and JNK2 phosphorylation following anti-IgM antibody treatment of B cells from CD22-deficient (CD22Ϫ/Ϫ, filled bars) and wild type (wt, shaded bars) control littermates as show in A. Values represent the relative mean optical density (Ϯ S.E.) of band intensities determined by scanning densitometry from three independent immunoblotting experiments. Values obtained for wild type control B cells incubated with medium alone (time 0) were adjusted to 1.0, with all other density values shown relative to this. * indicates that differences in mean values between CD22-deficient and wild type control B cells were significantly different, p Ͻ 0.01. cytoplasmic domain (Figs. 4A and 5, B and C). CD22 also bound SHIP (Figs. 3A and 4A), which may provide a molecular explanation for augmented calcium responses following BCR ligation in CD22-deficient B cells in addition to SHP-1 recruitment. Since SHIP did not bind to individual CD22 phosphopeptides at detectable levels (Fig. 5A) and required an intact CD22 cytoplasmic domain (Figs. 3A and 4A), SHIP probably binds CD22 indirectly through the formation of a newly described ternary complex that includes Shc, Grb2, and SHIP (35). SHIP binding by both Grb2 and Shc is required for the generation of a stable Grb2⅐Shc⅐SHIP ternary complex (35). In support of this model for CD22/SHIP interactions, the GST-CD22cyt(Tyr(P)) fusion protein precipitated all three SHIP isoforms, p145, p135, and p110 (Fig. 4A), although p110 lacks an SH2 domain (43). SHIP has been previously shown to bind a CD22 phosphopeptide containing Tyr-796 at low levels in a study using the A20 B cell line (44). However, a similar phosphopeptide was unable to precipitate SHIP from primary B cells in the current study ( Fig. 5A and data not shown). Thus, SHIP (p145 and p135) may alternatively bind phosphorylated CD22 through a weak SH2 domain-mediated interaction. Nonetheless, CD22 may be an important component in the formation of a stable Grb2, Shc, and SHIP complex.
CD22 may be utilized for Shc/SHIP mobilization to the plasma membrane since a Shc-associated 145-kDa phosphoprotein, presumably SHIP, is recruited from the cytosol to a membrane-bound fraction after BCR stimulation (7,45). SHIP recruitment to the plasma membrane is considered critical for its functional activity. At the membrane, activated SHIP downregulates calcium responses by reducing the availability of phospholipids such as phosphatidylinositol 3,4,5-trisphosphate that activate effector molecules such as Btk, an activator of PLC-␥2 (1,46,47). SHIP binding to CD22 may also be required for optimal SHIP phosphorylation since SHIP phosphorylation is decreased in CD22-deficient mice following BCR ligation (32). Given that SHIP tyrosine phosphorylation correlates positively with its phosphatase activity (48 -50), a lack of CD22mediated membrane localization and decreased SHIP phosphorylation may be molecular explanations for augmented calcium responses in CD22-deficient B cells.
Like CD22, Fc␥RIIB contains ITIM motifs, and co-crosslinking of the BCR and Fc␥RIIB results in the recruitment of SHIP to Fc␥RIIB and suppression of BCR-generated signals that include calcium mobilization (33). It has been postulated that CD22 down-regulation of calcium mobilization is primarily mediated by SHP-1 (34). However, SHIP also plays a critical role in suppressing calcium responses following BCR ligation independent of Fc␥RIIB co-ligation (1,52). SHIP suppresses calcium responses following BCR stimulation by decreasing inositol 1,4,5-trisphosphate production (53). Therefore, SHIP prevents the hyperactivation of resting B cells stimulated by free antigen in addition to its role in Fc␥RIIB signaling. Thus, SHIP may primarily limit BCR signal strength by hydrolysis of phospholipids after localization to the plasma membrane through cell surface molecules including CD22.
Grb2 and Shc interactions with CD22 required tyrosine phosphorylation of CD22. It may be functionally significant that the two Shc-binding regions flanked the known Grb2binding site on CD22 (Fig. 5, B and C). This may provide an optimal target for SHIP binding to a CD22⅐Grb2⅐Shc complex assembled following BCR ligation and CD22 phosphorylation. Grb2 has a single SH2 domain flanked by two Src homology 3 (SH3) domains, which allows it to interact with tyrosine-phosphorylated and proline-rich sequences, respectively. Shc contains one SH2 domain and one phosphotyrosine-binding domain that specifically recognizes phosphotyrosines within NPXY motifs (51). Shc likely binds CD22 phosphotyrosines via its SH2 domain since CD22 does not contain a NPXY motif. In addition, the phosphotyrosine-binding domain of Shc interacts with tyrosine-phosphorylated SHIP NPXY motifs in vitro (49). That similar levels of Grb2, Shc, and SHIP bound GST-CD22cyt(Tyr(P)) in lysates from both resting and activated B cells (Fig. 4A) suggests that adequate levels of these phosphoproteins are present in resting B cells to enable high level interactions with tyrosine-phosphorylated CD22 following BCR stimulation. Analogous to the constitutive association of CD22 with SHP-1 in resting primary B cells (32), low level CD22⅐Grb2⅐Shc⅐SHIP complex formation may also occur spontaneously in vivo to maintain the homeostasis of signaling thresholds in resting B cells.
It is striking that JNK was not significantly activated above resting levels following BCR ligation in CD22-deficient B cells, whereas ERK activation progressed normally (Fig. 7). It has been shown previously that in chronically stimulated (tolerant) B cells, JNK activation is impaired, but ERK activation is within normal limits (52). This was explained by the fact that tolerant B cells exhibited dramatically impaired calcium mobilization, and JNK activation, but not ERK activation, is dependent upon the calcium response. However, CD22-deficient B cells do not exhibit such signs of chronic stimulation, since these cells mobilize calcium at increased levels following BCR ligation. Therefore, impaired JNK activation in CD22-deficient B cells is not explained by a chronically activated phenotype. Also, deficient JNK activation may explain the decreased proliferative responses of CD22-deficient B cells after BCR ligation. By contrast, BCR-dependent JNK activation is moderately enhanced in B cells from Lyn-deficient mice (40), which also have enhanced calcium responses to BCR ligation. Since CD22 loss had no significant effect on ERK activation via BCR signaling, it is unlikely that CD22 is a major regulator of this signaling pathway during early B cell activation. In agreement with this, CD22 ligation induces JNK, but not ERK, activation (53). This suggests that cellular components involved in JNK activation are intricately linked to CD22 expression and function.
Lyn undoubtedly plays a major role in regulating CD22 function (40). Basal and BCR-induced CD22 phosphorylation requires amplification of Lyn activation through a CD19-dependent regulatory loop (54,55). Although it remains to be formally proven that Lyn phosphorylates CD22 in vivo, in vitro experiments in the current study suggest that multiple CD22 tyrosine residues are subject to Lyn phosphorylation (Fig. 6). Following its activation, Lyn is likely to preferentially phosphorylate certain tyrosines of CD22, subsequently bind CD22 through its SH2 domain, and then phosphorylate other CD22 tyrosine residues through "processive phosphorylation" (56). This model of CD22 phosphorylation correlates with data presented here that certain CD22 peptides that contain tandem tyrosine motifs were phosphorylated at much higher levels than single tyrosine motif peptides representing the same regions (Fig. 6). This may result in the recruitment of SHP-1, Shc, Grb2, and SHIP, which may become phosphorylated and further activated by Lyn bound to CD22. Although the order of significance of this cascade of events is likely to be highly regulated, the finding that Lyn phosphorylates different regions of CD22 may allow CD22 to interact with numerous regulatory proteins through spatially distinct phosphotyrosine residues.
This study demonstrates that CD22 does not merely serve as a docking site for tyrosine-phosphorylated proteins that are in turn dephosphorylated by SHP-1, thus limiting BCR signal strength. Rather, it appears that CD22 carries out multiple positive and negative regulatory functions. We propose that CD22 facilitates the mobilization of SHIP, Grb2, and Shc into a tetrameric complex at the cell surface that prevents B cell hyperstimulation by modulating calcium mobilization following BCR ligation. Therefore, a CD22⅐Shc⅐Grb2⅐SHIP complex may be generated in a manner analogous to assembly of the Shc⅐Grb2⅐SOS complex following BCR ligation (7,57). That CD22 forms a multimolecular complex with Grb2, Shc, and SHIP in primary B cells complements a recent study demonstrating CD22⅐Grb2⅐SOS complex formation in K46 B lymphoma cells (29). CD22⅐Grb2⅐Shc⅐SHIP or CD22⅐Grb2⅐SOS complex formation may be regulated by extrinsic factors unrelated to the ability of CD22 to associate with Grb2 and Shc. Alternatively, complex formation with SHIP or SOS may reflect a competitive process whereby SHIP blocks the interaction of Shc with SOS and thereby interferes with Ras activation as proposed by others (44,49,58). Nonetheless, the binding of these different adapter/effector molecules to discrete cytoplasmic regions of CD22 suggests that membrane-proximal CD22 phosphotyrosines are necessary for the mobilization of signaling complexes required for the positive regulation of SHIP and JNK activation. Together these data provide a plausible explanation for the paradox of augmented calcium flux, yet hypoproliferative responses by B cells from CD22-deficient mice after BCR stimulation.