The Down Syndrome Cell Adhesion Molecule (DSCAM) Interacts with and Activates Pak*

The Down syndrome cell adhesion molecule (DSCAM) is a member of the immunoglobulin superfamily that maps to a Down syndrome region of chromosome 21q22.2–22.3. In Drosophila , Dscam functions as an axon guidance receptor regulating targeting and branching. Genetic and biochemical studies have shown that in Drosophila , Dscam activates Pak1 via the Dock adaptor molecule. The extracellular domain of human DSCAM is highly homologous to the Drosophila protein; however, the intracellular domains of both human and Drosophila DSCAM share no obvious sequence identity. To study the signaling mechanisms of human DSCAM, we investigated the interaction between DSCAM and potential downstream molecules. We found that DSCAM directly binds to Pak1 and stimulates Pak1 phosphorylation and activity, unlike Drosophila where an adaptor protein Dock mediates the interaction between Dscam and Pak1. We also observed that DSCAM activates both JNK and p38 MAP kinases. Furthermore, expression of the cytoplasmic domain of DSCAM induces a morphological change in cultured cells that is JNK-dependent. These observations suggest that human DSCAM also signals through Pak1 and may function in axon guidance similar to the Drosophila Dscam. Down syndrome cell adhesion molecule (DSCAM), 1 a novel member of the immunoglobulin superfamily, maps to the chromosome at 21q22.2–22.3. Mutations in this region lead to men-tal retardation (1–3). DSCAM is mainly expressed in the brain, and its particular expression pattern suggests that it may play used are: DSCAM, Down syndrome cell adhesion molecule; JNK, c-Jun NH 2 -terminal kinase; MAP, mitogen-activated protein; PBS, phosphate-buffered saline; GST, glutathione S -transfer- ase; MOPS, 4-morpholinepropanesulfonic acid; IP, immunoprecipitation; MBP, maltose-binding protein; HA, hemagglutinin.

Down syndrome cell adhesion molecule (DSCAM), 1 a novel member of the immunoglobulin superfamily, maps to the chromosome at 21q22. 2-22.3. Mutations in this region lead to mental retardation (1)(2)(3). DSCAM is mainly expressed in the brain, and its particular expression pattern suggests that it may play an important role in the formation of neural networks (4,5). DSCAM is a type I transmembrane protein that likely functions as a cell surface receptor mediating axon growth and pathfinding (3)(4)(5)(6). Besides these important implications, little is know about the physiological function or the molecular mechanism of DSCAM signal transduction in mammalian systems.
In Drosophila, Dscam was identified by both biochemical purification and genetic analysis (7). Mutation in Dscam causes lethality in the early larval stage in Drosophila. Further studies demonstrated that Dscam is required for the formation of axon pathways in the central nervous system. In the embryonic nervous system in Drosophila, Dscam promotes Bolwig's nerve growth cones to an intermediate target (7). In the postembryonic nervous system, it has been shown that the primary axons of mushroom body neurons frequently migrate along inappropriate pathways and exhibit excessive branching in terminal regions in Dscam mutants (8). Dscam also plays a crucial role in the targeting of a subset of Drosophila olfactory receptor neurons (ORNs) (9). It has been found that ORNs frequently terminate in ectopic sites both within and outside the antennal lobe in Dscam mutants. These observations indicate that Dscam plays an important role in neuronal axon guidance and axon bifurcation (7,8,10,11).
Dock, a Drosophila adaptor protein, is highly similar to the mammalian Nck and consists of three SH3 domains followed by one SH2 domain (12). Dock binds a proline-rich motif in the intracellular domain of Dscam through its SH3 domain and may bind tyrosine-phosphorylated Dscam through its SH2 domain (7). Dock has been shown to play a role in axon guidance (13), therefore supporting a possible function of Dscam in neuronal guidance.
Genetic studies have implicated the serine/threonine kinase Pak1 functions downstream of Dock in axon guidance. In Drosophila, Pak1 is recruited to the Dscam signaling complex through Dock. Furthermore, Dock has been shown to interact with DSH3PX1 and the Wiskott-Aldrich syndrome protein (Wasp) to regulate actin polymerization and receptor trafficking (14). Thus, the Drosophila Dscam protein appears to interact with multiple downstream proteins to control neuronal axon guidance and axon bifurcation. An interesting feature of the Drosophila Dscam is that it can generate as many as 38,016 theoretical splicing forms. In fact, many different splicing forms of Dscam mRNA have been experimentally verified (7,15). The potential large numbers of Dscam splicing forms may contribute to diversity and specificity of neuronal connectivity (8,9). Dscam may form a dimer with other isoforms or act as co-receptor to modulate the activities of other guidance receptors.
Surprisingly, the intracellular domain of human DSCAM shares no obvious sequence homology with the Drosophila Dscam (7). The lack of sequence homology between the intracellular domains of the Drosophila and human proteins suggests that the signal transduction mechanisms could be different. Though unlikely, it is possible that the human DSCAM and Drosophila Dscam proteins may have very different biological functions. Despite strong functional data in Drosophila, little is known about the signaling mechanisms and physiological functions of DSCAM in higher eukaryotes. In this report, we show that human DSCAM directly interacts with Pak1. DSCAM binding activates Pak1 and also stimulates the JNK and p38 MAP kinases. Expression of the cytoplasmic domain of DSCAM induces a morphological change in cultured cells. Our studies suggest that the human DSCAM may have biological functions similar to the Drosophila protein and also provide a possible mechanism for DSCAM signaling.

DNA Constructs and Cell
Culture-Full-length DSCAM and JNK were PCR-amplified from a human brain cDNA library based on the sequence in the GenBank TM (DSCAM, gi:6740012; JNK, gi:20986493) and subcloned into a pcDNA3 vector. A FLAG or HA tag was added at the C terminus of DSCAM. FLAG-JNK or HA-JNK was tagged by FLAG or HA at the N terminus. DSCAM intracellular domain and its deletion mutants were created by PCR and subcloned into pcDNA3, pEBG3X, pGEX-KG, and pMAL-c2 vectors for expression in either mammalian cells or Escherichia coli, respectively. Myc-Pak1 and Myc-Rac1(L61) were kind gifts from Drs. J. Chernoff and A. Hall. HA-Pak1, Myc-Pak1K299A, Nck1 SH2 domain and SH3 domain, and His-Rac(L61) were constructed by PCR amplification and subcloned into pcDNA3HA, pEBG, and pSJ4 vectors, respectively. GST⅐c-Jun/pGEX-KG, HA-p38/pcDNA3, HA-ERK/pcDNA, GST⅐MKK4(K/R)/pEBG, and HA-Nck1/pcDNA3 are from laboratory stocks. HEK293, COS-7, and HeLa cells were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum at 5% CO 2 .
Antibodies and Chemicals-DSCAM antibody was raised by immu-nizing a rabbit with the intracellular domain of DSCAM purified from E. coli. Antibody was purified from serum by immunoaffinity binding following the manufacturer's procedures (Cat. No. 20422; Pierce). HA, Myc (Convence), FLAG (Sigma), GST (Novagen), Pak, and phospho-Pak1 (Cell Signaling Technology), phospho-JNK (Cell Signaling and Upstate Biotechnology), anti-mouse IgG fluorescein isothiocyanate (Molecular Probe), and phospho-c-Jun (Cell Signaling) antibodies were purchased from commercial sources. JNK inhibitor II (SP600125) and PP2 inhibitor were from Calbiochem. Immunoprecipitation, Immunoblotting, and in Vitro Binding-HEK293 cells were transiently transfected using LipofectAMINE (Invitrogen). To test protein-protein interaction, 40 h after transfection the cells were washed once with cold PBS and then lysed in Nonidet P-40 buffer (10 mM Tris-HCl, pH 7.5, 2 mM EDTA, 150 mM NaCl, 1% Nonidet P-40, 50 mM NaF, 1 mM sodium vanadate, 1 mM phenylmethylsulfonyl fluoride, 2 g/ml aprotinin, 2 g/ml leupeptin). The lysates were cleared by centrifugation at 13,000 rpm. The supernatant was incubated with specific antibodies for 2 h and followed by further incubation with protein A-or protein G-agarose for another hour. Co-precipitated proteins were detected by Western blot. To assay kinase activity, transfected cells were starved for 8 h in serum-free Dulbecco's modified Eagle's medium and then lysed in lysis buffer (1% Nonidet P-40, 0.5% sodium deoxycholate, 0.1% SDS, 1 mM EDTA, 1 mM EGTA, 1 mM dithiothreitol in phosphatebuffered saline, pH 7.4). The immunoprecipitation (IP) procedure was similar to that used by co-immunoprecipitation above.
For co-immunoprecipitation of endogenous proteins, mouse brains were homogenized in PBS buffer with proteinase inhibitors. The lysate was centrifuged at 7000 rpm for 5 min, and the supernatant was further centrifuged at 55,000 rpm for 1 h to pellet the membrane fraction. The pellets were solubilized with Nonidet P-40 buffer (for composition, see above) and immunoprecipitated with affinity-purified DSCAM antibody. The immunoprecipitates were analyzed by Western blot.
For in vitro binding, recombinant proteins were expressed and purified from E. coli. In vitro binding was performed in binding buffer (50 mM Tris-HCl at pH 7.5, 150 mM NaCl, 1.0 mM dithiothreitol, 1% Triton X-100, 5 mM MgCl 2 , and 25 M ZnCl2). Purified GST or MBP fusion proteins (ϳ5 g/ml) were incubated with the indicated proteins together for 2 h, and then glutathione-agarose beads (GST pulldown) or amylose resin (MBP pulldown) were added for another hour. The beads were washed four times with binding buffer, eluted with 5 mM glutathione or maltose, and analyzed by Western blot.
Kinase Activity Assay-Proteins (JNK, Pak1, ERK, and p38) were immunoprecipitated from lysates of transfected cells (16). The IP was washed three times with IP buffer and once with kinase assay buffer (20 mM MOPS, pH 7.2, 10 mM MgCl 2 , 2 mM EGTA, 0.1% Triton 100, and 1 mM dithiothreitol). Kinase reactions were initiated by incubating the precipitated kinase with substrates (JNK with 2 g of GST⅐c-Jun, Pak1 with 2 g of histone 2B, ERK and p38 with 2 g of GST⅐Elk, respectively) in the kinase assay buffer in the presence of 10 Ci of [␥-32 P]ATP at 30°C for 20 min. The kinase reactions were analyzed by SDS-PAGE and quantified by phosphorimaging. Pak1 and JNK activities were also analyzed by phospho-Pak1, phospho-JNK, and phospho-c-Jun Western analysis.
Cell Morphology and Immunofluorescence Staining-COS-7 and HeLa cells were transfected with DSCAM and various mutants. 40 h after transfection, cells were treated with the inhibitors or solvent controls for 2 h. The cells were fixed with 4% paraformaldehyde in PBS  ). B, DSCAM is specifically expressed in brain. Tissues from mice were analyzed for DSCAM expression using the anti-DSCAM antibody. DSCAM was mainly detected from mouse brain lysate. Low level expression was also observed in the heart and lung. The transfected DSCAM was included as a positive control (lane 9). C, co-immunoprecipitation of DSCAM and Pak1 from brain. The membrane fraction was prepared from mouse brain and lysed in Nonidet P-40 buffer (see "Materials and Methods"). DSCAM was immunoprecipitated with DSCAM antibody, and Pak1 was detected in DSCAM immunoprecipitates. Preimmune serum and anti-actin rabbit antibody were used as negative controls. fixed cells were incubated with rhodamine phalloidin for 45 min. The samples were washed twice with PBS. The cells were visualized by fluorescence microscopy. Images were quantified using NIH Image 1.62 (National Institutes of Health), and statistical analysis was performed using StatView 5.01 (SAS Institute Inc.). Student's t test was used to determine the significant differences between the mean relative sizes of the various experimental groups, represented as p values.

RESULTS
Human DSCAM Interacts with Pak-We investigated whether the mammalian DSCAM regulates Pak function. The human DSCAM cDNA was obtained by PCR from a human fetal brain cDNA library. We first conducted experiments to test whether human DSCAM interacts with Pak1 by co-immunoprecipitation. Our results show that the intracellular domain of DSCAM interacts with Pak1 (Fig. 1, A and B). The intracellular domain of DSCAM was expressed as a GST fusion in mammalian expression vector pEBG and was co-expressed in HEK293 cells with Pak1. Then, GST⅐DSCAM was precipitated using glutathione-agarose beads. GST⅐DSCAM specifically coprecipitated HA-Pak1, whereas the negative control GST did not (Fig. 1B). To determine the region in DSCAM responsible for Pak1 interaction, numerous deletion constructs were created and analyzed. Our data clearly showed that the N-terminal half (amino acid residues 1637-1791) of the intracellular domain of DSCAM is necessary and sufficient for interaction with Pak1 (Fig. 1, A and B). Therefore, this domain in DSCAM is defined as the Pak-interacting domain.
Pak1 is a serine/threonine kinase composed of regulatory domains at the N terminus and a kinase domain at the C terminus ( Fig. 2A). Pak1 has been reported to be involved in many signaling cascades (17)(18)(19). We performed deletion experiments to map the domain in Pak1 responsible for interaction with DSCAM. Deletion of residues before 351 or residues after 451 does not abolish the interaction with DSCAM, whereas any deletion within residues 351-451 abolishes DSCAM binding (Fig. 2, A and B). The deletion construct (400 -545) does not interact with DSCAM. These results show that amino acid residues between 351 and 400, the N terminus region of the Pak1 kinase domain, are required for interaction with DSCAM and thus define a DSCAM-interacting domain ( Fig. 2A, DID).
The Drosophila Dscam also interacts with Pak. However, this interaction is mediated by the Dock adaptor protein. To test whether the human DSCAM directly interacts with Pak, GST⅐DSCAM and MBP⅐Pak (251-500) were expressed and purified from E. coli. The purified proteins were tested in an in vitro binding assay. Our results show that GST⅐DSCAM was pulled down by MBP⅐Pak1 but not by the negative control, MBP⅐plexin B1 (Fig. 3A). Reciprocal pulldown experiments also showed that Pak1 was pulled down by GST⅐DSCAM but not the negative control GST (Fig. 3B). These data indicate that the intracellular domain of DSCAM directly interacts with Pak1.
DSCAM Interacts with Pak in Vivo-To test whether DSCAM interacts with Pak1 under physiological conditions, we first prepared DSCAM antibody using purified GST⅐DSCAM as an antigen. The anti-DSCAM antibody was affinity-purified from immunized rabbit antiserum. The purified DSCAM antibody specifically recognized DSCAM and showed a sensitivity compatible with the anti-FLAG antibody when transfected DSCAM was tested (Fig. 4A). We performed Western blot of mouse tissues to determine the DSCAM expression pattern. Our data showed that DSCAM is mainly expressed in mouse brain (Fig. 4B), consistent with previous reports with the mRNA expression patterns (3,20). To test the interaction between DSCAM and Pak1, membrane preparations from mouse brain were used to immunoprecipitate DSCAM, followed by immunoblot to detect Pak1 with anti-Pak1 antibody. Our data showed that Pak1 is co-immunoprecipitated by anti-DSCAM antibody but not by the preimmune serum or the control antiactin antibody (Fig. 4C). These data demonstrate that DSCAM and Pak1 form a complex under physiological conditions.

DSCAM Forms a Complex with Pak and Rac and Activates
Pak-Rho small GTPase family members, Rac1, Cdc42, and RhoA, play important roles in axon growth and pathfinding (11,(21)(22)(23). We tested the possibility of interaction between DSCAM with Rho GTPases by co-expressing them in HEK293 cells. We failed to see any obvious direct interaction among them (data not shown). However, when DSCAM was incubated with Rac in the presence of Pak1 using purified proteins from E. coli, Rac was pulled down in the DSCAM precipitation (Fig.  5A). Interestingly, the Pak and DSCAM interaction was enhanced by the presence of Rac protein (Fig. 5A). Our data suggest that the Rac1-bound Pak1, hence active, has higher affinity for the interaction with DSCAM, which may further regulate Pak activation. The formation of a DSCAM⅐Pak1⅐Rac complex is consistent with the observation that Rac interacts with the CRIB motif in the N-terminal domain of Pak1 and DSCAM interacts with the kinase domain of Pak1 ( Fig. 2A).
We tested whether DSCAM modulates Pak1 activity. Pak1 was co-transfected with GST⅐DSCAM in HEK293 cells and immunoprecipitated. Pak1 kinase activity positively correlates with phosphorylation in the activation loop of the kinase domain (24,25). Phosphorylation of Pak1 was determined with phospho-Pak1 antibodies. Our data showed that phosphorylation of Pak1 Ser-199/204 was enhanced by DSCAM (Fig. 5B), whereas Pak1 phosphorylation induced by RacL61 was observed as a positive control. To further confirm Pak1 activation by DSCAM, we performed a Pak1 kinase assay by co-expressing Pak1 with DSCAM in 293 cells in the presence of RacL61 (Fig. 5C). Our data showed that Rac activates Pak1 as expected. Expression of DSCAM further increased Pak1 kinase activity, consistent with the Pak1 phosphorylation data in Fig.  5B. We also tested various DSCAM deletion mutants. Our data indicate that Pak1 interaction appears to be important for DSCAM to stimulate Pak1 activity (Fig. 5C). However, the ability of DSCAM to bind Pak1 does not linearly correlate with its ability to activate Pak1; DSCAM(1637-1791) is not effective in activating Pak1, whereas this deletion mutation can potently bind Pak1 (Figs. 1B and 5C). These results indicate that besides the Pak1 binding domain, other regions of DSCAM also contribute to Pak activation.
DSCAM Activates JNK and p38 but Not ERK-One of the downstream events induced by Pak1 is activation of JNK, a member of the MAP kinase superfamily (26). We tested whether DSCAM could activate JNK by co-expression in HEK293 cells. Fig. 6A shows that JNK activity was strongly elevated by wild type GST⅐DSCAM. The deletion mutants that did not bind Pak1 dramatically reduced the ability of DSCAM to activate JNK. We also examined the effect of DSCAM on p38 and ERK kinases. DSCAM also potently activated p38 kinase activity (Fig. 6B) and had no effect on ERK activation, whereas the positive control, epidermal growth factor, stimulated ERK (Fig. 6C). The above observations demonstrate that the intracellular domain of DSCAM can activate JNK and p38, but not the ERK pathway.
JNK is phosphorylated and activated by MKK4, a MAP kinase kinase, upstream of JNK (27). We tested the role of MKK4 in DSCAM-stimulated JNK activation. A dominant negative MKK4(K/R) was co-transfected with JNK and GST⅐DSCAM. JNK was immunoprecipitated and assayed using recombinant GST⅐c-Jun as a substrate. Phosphorylation of GST⅐c-Jun was determined by the phospho-c-Jun antibody. Our data show that the MKK4(K/R) mutant completely blocked JNK activation (Fig. 6D), suggesting a role of MKK4 in DSCAM-induced JNK activation. To test the function of Pak1 in regulation of DSCAM-induced JNK activation, the kinase dead mutant Pak1(K299A) was tested (28). We found that the kinase dead mutant Pak1(K299A) significantly inhibited DSCAM-induced JNK phosphorylation (Fig. 6E). DSCAM deletion mutants 1637-1884 and 1910 -2012 lost their ability to activate JNK partially and completely, respectively (Fig. 6, A  and E).
DSCAM Induces Cell Contraction-Expression of axon guidance receptors in COS-7 cells causes cell morphological alterations, such as cell shrinkage (29 -32). We observed that overex- pressing GST⅐DSCAM(1637-2012) results in a dramatic cell morphology change in COS-7 cells (Fig. 7, A-D). Deletion of either the N-or C-terminal region of the intracellular domain of DSCAM cannot induce cell shrinkage (data not shown). Therefore, the ability to reduce cell size requires the intact intracellular domain. The cell shrinkage effect was not because of cell death, as assayed by staining to check cell viability. We also tested the involvement of JNK. We observed that JNK inhibitor II decreased JNK activity (data not shown) (21,33), and treatment with JNK inhibitor significantly blocked the effects of DSCAM on cell morphology (Fig. 7, E and F). The quantification and analysis of cell size of transfected COS-7 cells is shown in Fig. 7G. We also tested whether DSCAM can induce cell shrinkage in HeLa cells. The results are summarized in Fig. 7H. Our data show that DSCAM moderately reduced HeLa cell size, and this reduction was partially blocked by treatment of JNK inhibitor. Together, our observations indicate that activation of JNK by DSCAM expression may be important in the DSCAM-induced morphological alteration.

DISCUSSION
Drosophila genetic and neurobiological studies have established that Dscam plays an important role in neuronal axon guidance and targeting (7)(8)(9). Genetic studies have already demonstrated a functional importance of Pak1 in Drosophila (34). Despite the divergence between the intracellular domains of Drosophila and human DSCAM, Pak1 is common in the signal transduction pathways. Our studies demonstrate that the specific mechanism linking DSCAM to Pak1 are different between Drosophila and human (Fig. 8). Drosophila Dscam interacts with Pak1 via the Dock adaptor protein. It is Dock that recruits Pak1 to the Dscam complex, and presumably Pak1 is activated. In contrast, human DSCAM directly interacts with Pak1. The domains in both human DSCAM and Pak1 responsible for their interaction are defined in this study.
Pak1 positively functions downstream of the Drosophila Dscam receptor. However, no biochemical data have been reported on whether DSCAM activates Pak1. Our data clearly demonstrate that human DSCAM activates Pak1 kinase. These results complement and extend the genetic results from Drosophila. Furthermore, we showed that interaction between DSCAM and Pak1 is required for Pak1 activation. Our experiments with DSCAM deletion constructs also indicate that mere binding is not sufficient for DSCAM to activate Pak1. These observations suggest that activation of Pak1 by DSCAM likely involves other parts of the DSCAM molecule. It is possible that DSCAM may also recruit a Pak activator to the complex. It is worth noting that Rac1, which is a potent Pak1 activator, is also present in the DSCAM complex and enhances the interaction between DSCAM and Pak1 (Fig. 8). These data indicate that Rac may participate in Pak1 activation in response to DSCAM.
We observed that DSCAM can stimulate the activation of JNK and p38 MAP kinase but does not increase ERK activity. Using various dominant negative mutants, our data suggest that Pak1 is important for DSCAM to stimulate JNK activation. Furthermore, we observed that expression of a kinase inactive MKK4 blocks JNK activation by DSCAM. These results suggest a model of JNK activation by DSCAM as outlined in Fig. 8. Despite the exciting findings in Drosophila regarding the biological functions of Dscam, little functional data are available for human DSCAM. The function of DSCAM in Down syndrome has not been unequivocally established; therefore, it will be important to reveal whether DSCAM indeed plays a role affecting neuronal wiring and patterning in humans. This report provides biochemical evidence that human DSCAM likely activates intracellular signaling events similar to those activated by Drosophila Dscam, suggesting that the human DSCAM molecule may have physiological functions similar to the Drosophila Dscam in neuronal axon guidance and bifurcation.