Identification of an Evolutionarily Conserved Heterotrimeric Protein Complex Involved in Protein Targeting*

In Caenorhabditis elegans,lin-2, lin-7, and lin-10genetically interact to control the trafficking of the Let-23 growth factor receptor to the basolateral surface of body epithelia. The human homologue of the lin-10 gene has recently been identified as a member of the X11 gene family. The X11 proteins contain one phosphotyrosine binding (PTB) and two PSD-95·Dlg·ZO-1 (PDZ) domains as well as an extended amino terminus. We have previously shown that the PTB domain of X11α (also known as Mint1) can bind to the amyloid precursor protein (APP) in a phosphotyrosine-independent fashion and can markedly inhibit the processing of APP to the amyloid β (Aβ) peptide. Here, we report that X11α directly binds to the mammalian homologue of Lin-2 (mLin-2), also known as CASK. This binding is mediated by direct interaction between the Calmodulin Kinase II (CKII)-like domain of mLin-2 and the amino terminus of X11α. Furthermore, we can detect direct interactions between mLin-2 and mammalian Lin-7 (mLin-7). In mouse brain, we have identified a heterotrimeric complex that contains mLin-2, mLin-7, and X11α and that is likely important for the localization of proteins in polarized cells. This complex may play an important role in the trafficking and processing of APP in neurons.

fied in Shc and IRS-1 (2)(3)(4)(5) and subsequently found in a large number of proteins (6,7). In these proteins, this domain binds to Asn-Pro-X-pTyr, a ␤ turn motif found on activated growth factor receptors and other signaling molecules. After binding to growth factor receptors, both Shc and IRS-1 are themselves tyrosine-phosphorylated and coupled to downstream signaling molecules. Despite the name, there is strong evidence that some PTB domains bind to their target proteins in a phosphotyrosine-independent fashion (7). Perhaps this binding is best understood for the X11 protein where biochemical studies have demonstrated an interaction between the X11 PTB domain and APP (8,9). Structural studies indicate that the X11 PTB domain binds a nonphosphorylated beta turn motif on APP (10). Evidence has also been provided for phosphotyrosine-independent interactions of the Numb PTB domain (11)(12)(13). The finding that PTB domains can bind to their target peptides in a phosphotyrosine-independent fashion indicates that these domains can be involved in diverse cellular functions, not just signaling downstream of tyrosine kinases.
PTB domains are often found in combination with other protein-protein interaction domains. For example, Shc has both an SH2 domain and a PTB domain, whereas X11 proteins have two PDZ domains in addition to a PTB domain. PDZ domains have been shown to bind to the carboxyl terminus of proteins by wrapping around the extreme carboxyl-terminal residues (14,15). PDZ domain proteins such as PSD-95 have been demonstrated to play a role in receptor and channel clustering at synaptic junctions (16 -19). PDZ domains have also been implicated as being important in protein targeting to specific membrane surfaces. In Caenorhabditis elegans, the lin-2, lin-7, and lin-10 genes are important for the proper localization of the Let-23 growth factor receptor to the basolateral side of the body wall epithelium (20 -22). Let-23 is related to mammalian EGF receptors and binds to the lin-3 gene product, a protein related to TGF-␣ (23,24). Lin-3 is released by the anchor cell of the gonad and induces the epithelium to form a vulva by activating Let-23. Mutations in lin-2, lin-7, or lin-10 impair vulval formation likely because of mislocalization of the Let-23 receptor and its inability to efficiently bind Lin-3 (22). Originally Lin-10 was felt to be a protein unrelated to previously identified proteins (25). However recent work has reassigned the product of the lin-10 gene as a homologue of the X11 family of proteins. 2 In this work, we define a protein complex in mammalian brain that contains X11␣, mLin-2, and mLin-7. Thus, these protein interact both biochemically and genetically and likely control protein targeting in an evolutionarily conserved fashion.

EXPERIMENTAL PROCEDURES
Antibodies-Anti-Myc 9E10 (Oncogene Research Products) and anti-HA 12CA5 (Boehringer Mannheim) monoclonal antibodies were used for immunoprecipitation and immunoblotting. Anti-PSD-95 monoclonal antibody is from Upstate Biotechnology. Anti-syntaxin-1 monoclonal antibody is from Sigma. Polyclonal anti-mLin-2, anti-X11, and anti-mLin7 antibodies were prepared by injecting rabbits with the following * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM   Cell Culture-Human embryonic kidney 293 cells and A-172, a human neuroblastoma cell line, were grown in Dulbecco's modified Eagle's medium containing 100 units/ml Ϫ1 penicillin and 100 g/ml Ϫ1 streptomycin sulfate, supplemented with 10% fetal calf serum. NT2 cells were maintained in Dulbecco's modified Eagle's medium/F-12 medium containing 100 units/ml Ϫ1 penicillin and 100 g/ml Ϫ1 streptomycin sulfate, supplemented with 10% fetal calf serum.
DNA Constructs-Full-length human X11␣ and X11␤ cDNAs have been described elsewhere (26). We identified a third form of X11 in the data base (EST vh50 g07) and have termed it X11␥. Human lin-2 cDNA was constructed from two human ESTs encompassing the coding sequences from 1 to 612 amino acids (EST yt03b09) and from 578 to 898 amino acids (EST zl68e09) of human Lin-2 protein. These overlapping ESTs were fused by a two-step polymerase chain reaction procedure to produce the full-length human mlin-2 cDNA. A mouse EST representing mlin-7 cDNA (EST vg65d05) was used to generate the mLin-7 constructs. The RK5-myc vector was used to express X11␣, mLin-2, and mLin-7 fused to the myc epitope (8). The RK5-X11␣ construct produces an untagged X11␣ protein. The pcDNA-HA vector was used to express mLin-2 fused to the HA epitope. mLin-2 protein (CKII ϩ PDZ) is from residue 1 to 612, mLin-2 (CKII) is from residue 1 to 320, and mLin-2 (SH3 ϩ guanylate kinase (GK)) is from residue 578 to 898. GST-mLin-7 PDZ and amino terminus were created by a polymerase chain reaction procedure. All constructs were sequenced using Sequenase Version 2.0 (United States Biochemical Corp.).
Protein Procedures-Cells were washed twice with cold phosphatebuffered saline and lysed in lysis buffer (50 mM HEPES, pH 7.5, 10% glycerol, 150 mM NaCl, 1% Triton X-100, 1.5 mM MgCl 2 , 1 mM EGTA) supplemented with 1 mM phenylmethylsulfonyl fluoride, 10 g/ml Ϫ1 aprotinin, and 10 g/ml Ϫ1 leupeptin. After centrifugation at 16,000 ϫ g for 20 min, lysate protein content was normalized using the Bio-Rad protein assay kit. Mouse brain proteins were extracted following a similar procedure. For immunoprecipitation, lysates were incubated with antibodies overnight at 4°C. Protein A-agarose was added, and immune complexes bound to beads were recovered after 1 h, washed three times with HNTG buffer (50 mM HEPES, pH 7.5, 10% glycerol, 150 mM NaCl, 0.1% Triton X-100), boiled in 1ϫ sample buffer, and separated by SDS-PAGE. Transfer and immunoblotting on nitrocellulose using HRP-protein A or HRP-anti-mouse antibody/chemiluminescence method were performed as described (8). For overlay assays, the membrane was incubated for 2 h at room temperature with soluble GST or GST-mLin-2 (CKII) at 1 g/ml in TBS, 5% bovine serum albumin, 1 mM dithiothreitol. After rinsing with TBS, 0.1% Triton X-100, and TBS buffers, the membrane was incubated with polyclonal anti-GST antibody diluted in TBS, 5% bovine serum albumin for 2 h. The immune complexes were revealed using HRP-protein A/chemiluminescence method. Cell transfection, metabolic labeling, and GST binding assays  and RK5-myc-mLin-2 (mLin-2) or RK5-myc-mLin-2 (CKII ϩ PDZ) or RK5-myc-mLin-2 (SH3 ϩ GK) were incubated with anti-myc antibody bound to beads. After washing, immune complexes were resolved on a 10% SDS-PAGE, transferred to nitrocellulose, and revealed with polyclonal anti-X11 (top panel) and anti-myc antibodies (bottom panel). Total lysates were subjected to Western blot and probed with anti-X11 antibody (middle panel). B, overexpressed X11␣ containing lysate was incubated with GST-mLin-2 fusion proteins, and bound X11␣ was revealed with anti-myc antibody by Western blot. The GST-mLin-2-(275-612) protein encompasses the PDZ domain and the linker region between the CKII and PDZ domains (see Fig. 1A). C, the following mouse brain fractions were subjected to overlay assay: lysate before depletion (lysate), lysate depleted of X11␣ (lysate post-IP anti-X11), proteins bound on pre-immune serum (IP control), or anti-X11 (IP anti-X11) antibodies. The membrane was probed with soluble GST-mLin-2 (CKII) protein, and bound proteins were revealed with anti-GST antibody followed by HRP-protein A and chemiluminescence detection. X11␣ is indicated by an arrow. No signal was detected with soluble GST protein (not shown). D, the same procedure was performed to detect myc-tagged X11␣, X11␤, X11␥, and Lin-10 in 293 cell lysates with GST-mLin-2 (CKII) (top panel). The level of protein expression was detected by blotting with anti-myc antibody (bottom panel).
were performed as described previously (8). For fractionation, brains were homogenized in a 10 mM Tris, pH 7.4, 0.2 mM MgCl 2 , 5 mM KCl buffer containing protease inhibitors. Lysates were adjusted to 0.25 M sucrose and 1 mM EDTA, and debris were removed after a 1000 ϫ g centrifugation during 10 min at 4°C. Supernatants were centrifuged at 100,000 ϫ g during 1 h at 4°C. Supernatants represent the cytosolic fraction. Pellets were lysed in lysis buffer to extract membrane bound proteins.

RESULTS AND DISCUSSION
We have identified three members of the X11 protein family. X11␣/Mint1 and X11␤/Mint2 are primarily expressed in the central nervous system, whereas X11␥/Mint3 is more widely expressed (26). 3 All X11 family members have conserved PTB and PDZ domains but divergent amino termini. Our previous work has detected an interaction between X11 family members and APP via the X11 PTB domain (8). This binding prolongs the half-life of APP and slows its processing to the pro-amyloidogenic A␤ peptide (26,27). As X11␣ has multiple proteinprotein interaction domains (Fig. 1A), we looked for proteins other than APP that could bind to X11␣. In [ 35 S]methioninelabeled A-172 cells, we were able to show that X11␣ coimmunoprecipitates with a 110-kDa protein. X11␣ binds to this 110-kDa protein via residues 163-436 in its amino terminus (Fig.  1B).
Recent evidence indicates that the X11 proteins have a close homologue in C. elegans encoded by the lin-10 gene. 2 In worms, the lin-2 gene is linked to the lin-10 gene and encodes a 110-kDa protein (Fig. 1A), a size compatible with the protein binding to the amino terminus of X11␣. A rat homologue of lin-2 encoding a neurexin-binding protein known as CASK has been identified (28). We identified the human mlin-2/CASK gene from the EST data base and constructed an mLin-2 protein fused to a myc epitope. Antibodies raised against mLin-2 recognized the 110-kDa protein bound to X11␣ (not shown). When 293 cells were transfected with X11␣ and myc-tagged mLin-2, we were able to coimmunoprecipitate the proteins. Deletions of mLin-2 show that the SH3 and GK domains are dispensable for this interaction (Fig. 2A). The binding site for the amino terminus of X11␣ mapped to the CKII-like domain. This CKII domain, when expressed as a GST fusion protein, could bind X11␣ in precipitation reactions and in Far Western blotting (Fig. 2, B and C). This domain not only recognized X11␣ but also the C. elegans X11 homologue, Lin-10. In contrast the mLin-2 CKII domain did not recognize two other isoforms of X11, X11␤ and X11␥ (Fig. 2D). This is consistent with the finding that, although X11␣, X11␤, and X11␥ have conserved PTB and PDZ domains, they have divergent amino termini. Similarly, in 293 cells we could show that mLin-2 coimmunoprecipitates with X11␣ but not X11␤ or X11␥ (results not shown). The kinase and calmodulin binding site of calmodulin kinase II ␣ is 45% identical to the CKII domain of mLin-2 but

FIG. 3. In vivo interaction of X11␣ and mLin-2 in mouse brain.
A, proteins from mouse brain lysate extracted in lysis buffer were immunoprecipitated with pre-immune serum (IP control) or anti-X11 antibody (IP anti-X11), and bound proteins were separated on an 8% SDS-PAGE. After Western blot, proteins were revealed with anti-X11 (top panel), anti-mLin-2 (middle panel), and anti-PSD95 (bottom panel) antibodies. As a control, one-tenth the amount of lysate used for immunoprecipitation was run on the gel as lysate. B, total lysate, cytosolic, and membrane fractions of mouse brain were prepared as discussed under "Experimental Procedures." The protein fractions were then separated by SDS-PAGE. After transfer to nitrocellulose, proteins were revealed with the respective antibodies from top to bottom: anti-mLin-2, anti-X11, and anti-syntaxin. C, same as panel A but cytosolic and membrane fractions were used for immunoprecipitation. D, the cytosolic fraction was immunodepleted of X11␣ by immunoprecipitation with anti-X11 antibody (not shown). Lysate before (lysate) or after (X11␣ depleted lysate) depletion was run on SDS-PAGE, and proteins were transferred on nitrocellulose. The membrane was probed with anti-X11 (top panel) and anti-mLin-2 (bottom panel) antibodies.

FIG. 4.
A heterotrimeric complex of mLin-7⅐mLin-2⅐X11␣ exists in the brain. A, proteins from mouse brain lysate were immunoprecipitated with pre-immune serum (IP control) or anti-mLin-7 antibody (IP anti-mLin-7), and bound proteins were separated on an 8% SDS-PAGE. After Western blot, proteins were revealed with anti-X11 (top panel) and anti-mLin-2 (bottom panel) antibodies. B, full-length mLin-7 was expressed as an myc-tagged protein and precipitated with GST proteins. After Western blot, bound mLin-7 was revealed with anti-myc antibody. C, HA-tagged mLin-2 produced in 293 cells was precipitated with GST-mLin-7 proteins and revealed with anti-HA antibody.
does not bind to X11␣ (result not shown). The CKII domain of mLin-2 does not appear to encode an active kinase (21,28) but rather appears to function as a protein-protein interaction domain.
We examined the interaction of X11␣ and mLin-2 in mouse brain. We were able to show that antibodies to X11␣ were able to coimmunoprecipitate mLin-2 from mouse brain (Fig. 3A). We found both proteins in cytosolic and membrane fractions (Fig.  3B) where they coimmunoprecipitate (Fig. 3C). When we immunodepleted brain cytosolic fraction of X11␣, we also removed a large amount of mLin-2 from the lysates (Fig. 3D), indicating that a significant fraction of X11␣ and mLin-2 are bound together. In worms, lin-7 mutants yield a similar phenotype as is seen with lin-2 and lin-10 mutants (22). In mouse brain, we were able to show that immunoprecipitation with anti-mLin-7 antibodies coimmunoprecipitated X11␣ and mLin-2 proteins (Fig. 4A). GST-mLin-7 can bind mLin-2 but not X11␣ (results not shown), demonstrating that mLin-7 is bound to mLin-2, whereas mLin-2 is bound to X11␣. We have mapped the site of interaction for mLin-7 to the region between the CKII and PDZ domain of mLin-2 (Fig. 4B). Conversely, we have mapped the binding site for mLin-2 to the amino-terminal half of Lin-7 (Fig.  4C).
In summary, we have identified a heterotrimeric complex in the brain that consists of X11␣, mLin-2, and mLin-7. In worms, genetic analysis has shown that homologous proteins play a central role in targeting of the Let-23 receptor to the basolateral surface of epithelial cells (22). This process is mediated by the binding of the Lin-7 PDZ domain to Let-23 (20). However, we have found that none of the mammalian EGF-receptor family members (HER1 through HER4) binds to mLin-7. 4 In neurons, the X11␣⅐mLin-2⅐mLin-7 protein complexes may bind to specific target membranes sites via Munc-18 -1, a protein that can bind syntaxins and possibly modulate trafficking and secretion (29). Once bound to membranes, X11␣, mLin-2, and mLin-7 proteins either individually or in a complex may serve to localize or retain proteins at specific membrane sites. Recent studies have detected a fraction of mLin-2/CASK at neuronal synapses (30) as well as at the basolateral surface of epithelial cells 4 (31). Because of the brain-specific expression of X11␣, we can exclude a role for X11␣ in the basolateral localization of mLin-2 in mammalian epithelia. X11␥, the only X11 species that we have detected in epithelium, does not interact with mLin-2. This suggests a possible divergence between worm and mammalian epithelia because, in worm epithelia, the X11␣ homologue, Lin-10, is crucial for basolateral targeting. The heterotrimeric complex contains several protein-protein interaction domains that would be useful to contact a large number of different proteins (30,31). In addition, the presence of the PTB domain, a domain that can bind to beta turn motifs, might make X11 proteins particularly suitable for detecting traffick-ing signals with tyrosine-based motifs (7). An alteration in localization of APP or its retention in a subcellular compartment induced by X11␣ may explain the effects of X11␣ on the processing of APP (26). Its is presently unknown if mLin-2 and mLin-7 also participate to this effect. Identification of additional proteins that can bind to X11, mLin-2, and mLin-7 family members will be important to further elucidate the role of this protein complex in receptor localization and function.