Erbin Is a Protein Concentrated at Postsynaptic Membranes That Interacts with PSD-95*

Neuregulin is a factor essential for synapse-specific transcription of acetylcholine receptor genes at the neuromuscular junction. Its receptors, ErbB receptor tyrosine kinases, are localized at the postjunctional membrane presumably to ensure localized signaling. However, the molecular mechanisms underlying synaptic localization of ErbBs are unknown. Our recent studies indicate that ErbB4 interacts with postsynaptic density (PSD)-95 (SAP90), a PDZ domain-containing protein that does not interact with ErbB2 or ErbB3. Using as bait the ErbB2 C terminus, we identified Erbin, another PDZ domain-containing protein that interacts specifically with ErbB2. Erbin is concentrated in postsynaptic membranes at the neuromuscular junction and in the central nervous system, where ErbB2 is concentrated. Expression of Erbin increases the amount of ErbB2 labeled by biotin in transfected cells, suggesting that Erbin is able to increase ErbB2 surface expression. Furthermore, we provide evidence that Erbin interacts with PSD-95 in both transfected cells and synaptosomes. Thus ErbB proteins can interact with a network of PDZ domain-containing proteins. This interaction may play an important role in regulation of neuregulin signaling and/or subcellular localization of ErbB proteins. The neuromuscular junction is a cholinergic synapse that conveys signals rapidly from motoneurons to muscle cells. The fast and accurate neurotransmission at this synapse is guar-anteed by the high

The neuromuscular junction is a cholinergic synapse that conveys signals rapidly from motoneurons to muscle cells. The fast and accurate neurotransmission at this synapse is guaranteed by the high concentration of acetylcholine receptors in the postsynaptic membrane, which accounts for only 0.1% of total muscle surface (1,2). Muscle fibers are multinucleated cells. Remarkably, it is only the synaptic nuclei that actively transcribe genes encoding acetylcholine receptor subunits. Such synapse-specific transcription is mediated by neuregulin, a factor used by motoneurons to stimulate acetylcholine receptor synthesis at the neuromuscular junction. Neuregulin receptors are transmembrane tyrosine kinases of the ErbB family: ErbB2, ErbB3, and ErbB4. Stimulation by neuregulin of ErbB proteins leads to their tyrosine phosphorylation (3)(4)(5)(6) and subsequent activation of multiple intracellular signaling cascades (6 -9), essential for compartmental synthesis of acetylcholine receptors. In the central nervous system, neuregulin regulates expression of neuronal nicotinic acetylcholine receptor (10), N-methyl-D-aspartate receptor (11), and ␥-aminobutyric acid receptor (12). Recent studies suggest that in addition to an essential role during development, neuregulin appears to regulate synaptic plasticity in the adult brain (13).
ErbB proteins are not expressed evenly on the surface of cells. On the contrary, they are localized in subcellular compartments. In the nervous system, ErbB proteins are concentrated in postsynaptic membranes both at the neuromuscular junction (3, 14 -16) and in the central nervous system (13,17). In epithelial cells, ErbB2 appears to be enriched in basolateral membranes (18). The mechanism by which ErbB proteins are localized in the subcellular compartments remains largely unknown. The intracellular portions of ErbB receptor tyrosine kinases contain large C termini in addition to kinase domains. Thus, it is conceivable that ErbBs may interact with proteins that regulate their localization, surface expression, or kinase activity. Indeed, recent studies demonstrated that ErbB4, via its C terminus, interacts with postsynaptic density (PSD) 1 -95 (or SAP90), a PDZ domain-containing protein (13,17). PDZ domains are motifs of 80 -90 amino acids which often bind to specific sequences at the extreme C termini of target proteins (19 -22). They were originally identified in PSD-95, the Drosophila septate junction protein discs large, and the epithelial tight-junction protein zona occludens 1 (23)(24)(25)(26). PDZ domaincontaining proteins appear to coordinate the assembly of functional subcellular domains. PSD-95 uses multiple PDZ domains to cluster ion channels, receptors, and cytosolic signaling proteins in subcellular domains including synapses and cellular junctions (27). The interaction of PSD-95 with ErbB4 potentially may allow for a localized signaling complex at synapses while minimizing unwanted cross-talk. Moreover, PSD-95 could enhance neuregulin signaling probably by promoting dimerization of ErbB4 receptor tyrosine kinases (13).
However, PSD-95 interacts with ErbB2 poorly and does not interact with ErbB3 (13,17), which raises the possibility that other PDZ domain-containing protein may exist. Using a yeast two-hybrid strategy, we identified a novel PDZ domain-containing protein that interacts specifically with ErbB2, but not ErbB3 or ErbB4. This protein was named B2BP for ErbB2binding protein. B2BP was a polypeptide of 180 kDa. It had 16 leucine-rich repeats (LRRs) in the N terminus and a PDZ domain in the C terminus. While the study was in progress, Borg et al. reported the cloning of Erbin as an ErbB2-interact-ing protein (18,28). Sequence analysis suggests that B2BP is the mouse homolog of Erbin. We will refer B2BP as Erbin in the manuscript. We demonstrate that Erbin is concentrated at the neuromuscular junction and a component of the PSD in the central nervous system. Erbin interacts with ErbB2 in synaptosomes. Moreover, Erbin increases surface expression of ErbB2 in transfected cells. Furthermore, Erbin interacts with PSD-95 in synaptosomes and in mammalian cells. These results suggest that ErbB receptor tyrosine kinases interact with a network of PDZ domain-containing proteins that may regulate neuregulin signaling and localization.

EXPERIMENTAL PROCEDURES
Two-hybrid Studies in Yeast-To identify ErbB2-interacting proteins, the ErbB2 carboxyl terminus (amino acids 1251-1260) was generated by two complimentary oligonucleotides and subcloned into the pGBT9 yeast vector containing the Gal4 DNA binding domain (CLON-TECH). The bait plasmid was then transformed into the yeast strain Y190 and used to screen mouse cDNA libraries in ACT2 or in pACT2 that contained the Gal4 transcription activation domain. Positive clones were selected on plates lacking leucine, tryptophan, and histidine and were confirmed further by a filter assay for ␤-galactosidase activity as described previously (13). Constructs containing PDZ domains of PSD-95, nNOS, ␣1-syntrophin, ␤1-syntrophin, or ␤2-syntrophin, or C termini of ErbB3 or ErbB4 have been described previously (13). C termini of ErbB2 mutants and NR2A (amino acids 1416 -1464) were generated by polymerase chain reaction and subcloned in pGBT9. Sequences of all constructs were confirmed by DNA sequencing. The yeast vectors are transformed into HF7c and Y190. Interactions were characterized by growth without leucine, tryptophan, and histidine and by a filter assay for ␤-galactosidase activity.
Antibody Production and Purification-The GST fusion protein containing the PDZ of Erbin (amino acid residues 1241-1371) was produced, affinity purified, and concentrated as described previously (29). Antiserum against the GST-Erbin/PDZ fusion protein was raised in a New Zealand White rabbit by standard procedures (30). To purify the antibodies, two affinity columns were prepared by coupling GST protein and GST-Erbin/PDZ fusion protein to Affi-Gel 15 (Bio-Rad), respectively, according to the manufacturer's instruction. The columns, 2 ml each, were washed sequentially with 10 ml of 100 mM glycine, pH 2.5, 10 ml of 10 mM Tris/HCl, pH 8.8, 10 ml of 100 mM triethylamine, pH 11.5, and equilibrated with 10 ml of 10 mM Tris/HCl, pH 7.5. 2 ml of antiserum was diluted in 18 ml of 10 mM Tris/HCl, pH 7.5, and passed through the GST-coupled Affi-Gel 15 column three times. The flowthrough was then loaded on the GST-Erbin/PDZ-coupled Affi-Gel 10 column three times. After washing the column with 20 ml of 10 mM Tris/HCl, pH 7.5, and 20 ml of 500 mM NaCl in 10 mM Tris/HCl, pH 7.5, antibodies were eluted with aliquots of 100 mM glycine, pH 2.5, and collected in vials containing 500 l of 1 M Tris/HCl, pH 8.0. The antibodies were dialyzed overnight against 10 mM Tris/HCl, pH 7.5, 0.9% saline and stored in 0.02% sodium azide. Affinity-purified antibodies were used in all experiments unless otherwise specified.
Primary rat muscle cell cultures were prepared as described previously (32) with minor modifications. Muscles were isolated from hind legs of day 19 rat embryos. Muscles were tweezed apart in PBS and incubated in PBS containing 0.25% trypsin at 37°C for 50 min with frequent trituration. Dissociated cells were filtered through a 20-mesh screen and pelleted twice. They were resuspended in the growth medium (Dulbecco's modified Eagle's medium containing 10% fetal bovine serum, 10% horse serum, 2% chick embryo extract, and 25 g/ml gentamicin). The cells were plated on regular culture dishes for 30 min to get rid of fibroblasts prior to being plated on collagen-coated culture dishes at a density of 7.5 ϫ 10 6 cells/10-cm dish. By the 4th day in vitro, myotubes were formed, and the cultures were treated with 3 g/ml cytosine arabinoside for 24 h to inhibit fibroblast proliferation.
Cells were transfected using the standard calcium phosphate technique. 2 days after transfection, cells were washed with PBS and lysed in the modified RIPA buffer (1 ml/100-mm plate), containing 20 mM sodium phosphate, pH 7.4, 50 mM sodium fluoride, 40 mM sodium pyrophosphate, 1% Triton X-100, 2 mM sodium vanadate, 50 M phenylarsine, 10 mM p-nitrophenyl phosphate, including protease inhibitors (6). Lysed cells were incubated on ice for 30 min and centrifuged at 13,000 ϫ g for 10 min at 4°C. The supernatant was designated as cell lysate.
Northern Blotting-Northern blot was done using a membrane containing mRNAs from multiple tissues (CLONTECH). An Erbin cDNA fragment (nucleotides 4044 -4651) was labeled with [␣-32 P]dCTP by a random prime method. The membrane was hybridized in a buffer (total 5 ml) containing 32 P-labeled probe (ϳ4 ϫ 10 8 cpm/g of cDNA), 5 ϫ SSC, 5 ϫ Denhardt's solution, 0.5% SDS, 50% formamide, and 100 g/ml salmon sperm DNA at 42°C overnight. It was then washed with 0.5 ϫ SSC, 0.5% SDS at 55°C three times each for 30 min and exposed to Kodak X-Omat AR film at Ϫ70°C with an intensifying screen.
Subcellular Fractionation-Adult brains were homogenized in a homogenizing buffer containing 0.32 M sucrose, 4 mM HEPES/NaOH, pH 7.4, 5 mM EDTA, 5 mM EGTA, 20 units/ml Trasylol, and 0.1 mM phenylmethylsulfonyl fluoride with a glass-Teflon homogenizer as described previously (33). Briefly, the ground tissue was centrifuged at 800 ϫ g for 10 min, and the supernatant was designated the homogenate (S1). The homogenate was centrifuged at 9,000 ϫ g for 15 min, yielding P2 (the crude synaptosomal fraction) and S2. The P2 fraction was resuspended in the homogenizing buffer and used for coimmunoprecipitation studies. To purify PSDs, the resuspended P2 pellet was subjected to another centrifugation at 10,000 ϫ g for 15 min, and the pellet was lysed by hypoosmotic shock in water, rapidly adjusted to 1 mM HEPES/NaOH, pH 7.4, and stirred on ice for 30 min. The lysate was then centrifuged at 25,000 ϫ g for 20 min, yielding P3 and S3. The P3 pellet was resuspended in 0.25 M buffered sucrose, layered onto a discontinuous sucrose gradient containing 0.8 M/1.0 M/1.2 M sucrose, and centrifuged for 2 h at 65,000 ϫ g in a Beckman SW-28 rotor. The gradient yielded a synaptosomal plasma membrane (SPM) fraction at the 1.0 M/1.2 M sucrose interface. The SPM fraction was solubilized with 0.4% Triton X-100 in 0.5 mM HEPES/NaOH, pH 7.4, yielding an insoluble PSD fraction and a soluble SPM extract after a centrifugation at 65,000 ϫ g for 20 min. Synaptophysin, a synaptic vesicle protein, was fractionated with the SPM fraction but was not present in the PSD fraction (see Fig. 8).
Immunoprecipitation, Pull-down Assays, and Immunoblotting-Cell lysates (ϳ400 g of protein) were incubated directly without or with indicated antibodies for 1 h at 4°C. They were then incubated with protein A-agarose beads overnight at 4°C on a rotating platform. After centrifugation, beads were washed four or five times with the modified RIPA buffer. Bound proteins were eluted with SDS sample buffer and subjected to SDS-PAGE. Immunoprecipitation of Erbin from rat brain P2 fraction was done as described previously (13).
The GST fusion protein containing the PDZ of Erbin (amino acids 1241-1371) was induced in BL21 cells with 1 mM isopropyl ␤-D-thiogalactopyranoside and purified using glutathione-agarose beads (Roche Molecular Biochemicals, Indianapolis). Equal amounts of GST fusion protein beads (ϳ50 g of protein) were incubated with cell lysates overnight at 4°C on a rotating platform. After centrifugation, beads were washed four or five times with wash buffer (150 mM sodium chloride, 10 mM sodium phosphate, 1% Triton X-100, pH 7.4). Bound proteins were eluted with SDS sample buffer and subjected to SDS-PAGE and immunoblotting.
Proteins resolved on SDS-PAGE were transferred to nitrocellulose membranes (Schleicher & Schuell). Nitrocellulose blots were incubated at room temperature for 1 h in Tris-buffered saline with 0.1% Tween (TBS-T) containing 5% milk followed by an incubation with 1% milk with the indicated antibodies except the anti-phosphotyrosine antibody, which required 3% bovine serum albumin in the blocking buffer and 1% bovine serum albumin in the blotting buffer. After washing three times for 15 min each with TBS-T, the blots were incubated with horseradish peroxidase-conjugated donkey anti-mouse or anti-rabbit IgG (Amersham Pharmacia Biotech) followed by washing. Immunoreactive bands were visualized with enhanced chemiluminescence substrate (Pierce).
In some experiments, after visualizing an immunoactive protein, the nitrocellulose filter was incubated in a buffer containing 625 mM Tris/ HCl, pH 6.7, 100 mM ␤-mercaptoenthanol, and 2% SDS at 50°C for 30 min, washed with 0.1% Tween 20 in 50 mM TBS at room temperature for 1 h, and reblotted with different antibodies.
Immunohistochemistry-Normal or denervated (5 days postdenervation) muscles were rapidly dissected, stretched on a board, and frozen in isopentane cooled with dry ice. 10-m sections were prepared using a cryostat, thaw mounted on gelatin-coated slides, and stored at Ϫ80°C. Sections of adult rat muscles were incubated with 2% normal goat serum (Vector Laboratories, Burlingame, CA) in PBS for 1 h at room temperature to reduce background staining and then incubated with the affinity-purified antibodies against Erbin or preimmune serum in 2% normal goat serum in PBS overnight at 4°C. In some experiments, affinity-purified antibodies were preincubated with 10 nM GST-Erbin/ PDZ overnight at 4°C prior to immunohistochemical studies. After washing the sections five times with PBS, each for 30 min, the sections were incubated with a fluorescein isothiocyanate-conjugated anti-rabbit antibody (Zymed Laboratories Inc., San Francisco) and rhodamine-conjugated ␣-bungarotoxin (Molecular Probes, Eugene, OR). Fluorescent images of cells were captured on a Sony CCD camera mounted on a Nikon E600 microscope using Photoshop imaging software.
Labeling of Surface Proteins-To label surface proteins, cells were washed with cold PBS containing 1 mM MgCl 2 and 0.1 mM CaCl 2 and incubated with 0.5 mg/ml sulfo-NHS-LC-biotin in the same buffer at room temperature for 30 min. The labeling reaction was quenched by incubation with 100 mM glycine for 10 min at room temperature. Cells were then lysated in the modified RIPA buffer. Lysates were incubated with streptavidin-agarose beads (Molecular Probes) overnight at 4°C. Bound proteins were subjected to SDS-PAGE.
Protein Assay-The protein was assayed according to the method of Bradford (34) using bovine serum albumin as a standard.

RESULTS
Cloning of Erbin-The C terminus (-DVPV*) of ErbB2 fits the consensus site for PDZ binding. However, whereas ErbB4 binds strongly to PSD-95, ErbB2 has little or no affinity for PSD-95 (13). To identify proteins that bind to ErbB2, we generated several bait constructs composed of the ErbB2 C terminus in various lengths. Most of the baits showed autonomous transactivation activity in various yeast strains except the one with the last 10 amino acid residues (amino acids 1251-1260). Screens using this bait of mouse muscle, mouse brain, and human heart cDNA libraries led to isolation of cDNAs all of which encoded partial sequences of an apparently same protein with a PDZ domain in the C terminus. This protein was initially named as B2BP for ErbB2-binding protein because it only interacted with ErbB2 and not ErbB3 or ErbB4 (see below). While this study was in progress, Borg et al. reported Erbin (18). Sequence analysis indicated that B2BP was the mouse homolog of Erbin. Thus, this protein will be referred as Erbin in the rest of this manuscript. Erbin showed high homology to Densin-180, a protein identified previously as a postsynaptic component (35,36). Like Densin-180, Erbin had 16 LRR domains in the N terminus. In the C terminus, there is a PDZ domain of group I which is characterized by a conserved histidine residue (37). The homology between Erbin and Densin-180 was 73% in LRR domains, 71% in the PDZ domain, and 39% in the middle region.
Characterization of Erbin Binding to ErbB2 in Yeast-Nice clones were isolated from the libraries that encoded two fragments of Erbin: Erbin965 and Erbin1254 (Fig. 1A). The binding of Erbin to ErbB2 was dependent on the PDZ domain of Erbin because deletion of the PDZ domain prevented the interaction. Furthermore, the PDZ alone was sufficient to bind to ErbB2. Although Erbin showed high homology with Densin-180, ErbB2 did not interact with the PDZ domain of Densin-180 (Fig. 1A), nor did it interact with the PDZ domains of PSD-95 (Fig. 1A) and of Scribble, nNOS, or ␣1-, ␤1-, and ␤2-syntrophin (data not shown). The binding of Erbin to ErbB2 was dependent on the ErbB2 C terminus. Mutation of the valine residues at the Ϫ1 and Ϫ3 positions to alanine prevented ErbB2 from interacting with Erbin (Fig. 1B). On the other hand, Erbin interacted specifically with the C terminus of ErbB2 and did not interact with C termini of the ErbB3, ErbB4, or NR2A subunit of the N-methyl-D-aspartate receptor (Fig. 1C).
Interaction of Erbin with ErbB2, Not ErbB3 or ErbB4 -To characterize further the interaction between Erbin and ErbB proteins, we examined the ability of Erbin's PDZ domain to bind to ErbBs in in vitro pull-down assays. Lysates from HEK 293T cells transfected with ErbB2, ErbB3, or ErbB4 were incubated with GST-Erbin fusion protein immobilized on agarose beads. Bound proteins were resolved on SDS-PAGE and immunoblotted with individual anti-ErbB antibodies. Consistent with the results from yeast two-hybrid assays, GST-Erbin was only able to pull down ErbB2 ( Fig. 2A). In contrast, ErbB3 or ErbB4 was undetectable in the Erbin complex. To determine whether ErbB2 interacts with Erbin in mammalian cells, we expressed ErbB proteins with or without Myc-tagged Erbin in HEK 293T cells. Lysates of transfected cells were incubated with individual anti-ErbB antibodies, and the resulting immunocomplex was blotted with anti-Myc antibodies. Erbin was detected in the immunoprecipitates from cells that had been cotransfected with ErbB2 and Erbin (Fig. 2B), suggesting that ErbB2 associates with Erbin in vivo. In contrast, Erbin was not detected in the ErbB3 or ErbB4 immunoprecipitates (Fig. 2C).
Expression of Erbin mRNA-Northern blot analysis was used to study mRNA expression of Erbin. The membrane loaded with mRNAs from multiple tissues was probed with a 32 P-labeled Erbin DNA fragment (encoding amino acids 1241-1371 plus the 3Ј-noncoding region). A major transcript at 7.5 kilobases was detected in various tissues (Fig. 3, top panel). The expression was high in the lung, heart, and kidney, moderate in the brain, skeletal muscle, and testis, and little, if any, in the spleen and liver. In contrast, expression of the Densin-180 mRNA was brain-specific as reported previously (35). The 7.4-kilobase transcript of Densin-180 was detected only in the brain, but not in any of tested periphery tissues (Fig. 3, middle  panel). Note the exposure time of blots for Densin-180 (10 days) and Erbin (1 day), whereas both used a similar amount of probes (5 ng/ml, 5 ml) with same specific activity (4 ϫ 10 8 cpm/g of DNA). These results suggest that the expression level of Erbin may be at least five times higher than that of Densin-180 in the brain.
Erbin Is a Protein Tightly Associated with Membrane-To study Erbin expression at the protein level, antibodies against Erbin were generated using as antigen the Erbin PDZ domain (amino acids 1241-1371). When affinity purified, the antibody detected a 180-kDa protein on Western blots of HEK 293T cell lysates (Fig. 4A, large arrow). The interaction of the 180-kDa protein and the serum was specific because it could be blocked by preincubation of serum with the antigen GST-Erbin/PDZ fusion protein (Fig. 4A, lane 3). The antibodies also recognized the transfected Erbin965 recombinant protein in HEK 293T cells (Fig. 4A, small arrow), whose expression was evident by blotting with an anti-Myc antibody (Fig. 4A, lane 5). Although there is a high homology between the PDZ domains of Erbin and Densin-180, the anti-Erbin antibody did not cross-react with Densin-180 (Fig. 4B, lane 7). Taken together, these results indicate that Erbin is not a human or mouse ortholog of Densin-180.
As observed with ErbB2, Erbin was present only in membrane, but not soluble, fractions (Fig. 5A). To determine how tightly Erbin associates with the membrane fraction, we treated brain membrane (P2) fractions with high concentrations of salt, high pH, or various detergents to solubilize Erbin. Solubilized or insolubilized fractions were immunoblotted with anti-Erbin antibodies. As shown in Fig. 5A, Erbin was resistant to wash with 1 M NaCl or high pH buffer (0.1 M NaCO 3 ), which disrupted protein interactions and extracted mainly peripheral membrane proteins, respectively. Except for partial solublization by 3% Nonidet P-40, Erbin was virtually insoluble in 2.5% CHAPS or 1% Triton, conditions under which many membrane proteins were solubilized. However, Erbin could be solubilized FIG. 2. Specific interaction of Erbin with ErbB2 but not ErbB3 or ErbB4. A, analysis of Erbin binding to ErbB2 by in vitro pull-down assays. HEK 293T cells were transfected with ErbB2, ErbB3, or ErbB4. Cell lysates were incubated with GST or the GST-Erbin fusion protein (containing amino acid residues 965-1371) immobilized on agarose beads. Bound proteins were resolved on SDS-PAGE and subjected to immunoblotting with the indicated specific anti-ErbB2, ErbB3, or ErbB4 antibodies. B, analysis of Erbin binding to ErbB2 by immunoprecipitation. ErbB2 was transfected without or with Myc-Erbin in HEK 293T cells. Immunoprecipitates (IP) with anti-ErbB2 antibodies (left panels) or preimmune serum (right panels) were resolved on SDS-PAGE and immunoblotted (IB) with an anti-Myc antibody. One-tenth of the input was used in the pull-down and coimmunoprecipitation experiments. C, analysis of Erbin binding to ErbB3 and ErbB4 by immunoprecipitation. ErbB3 or ErbB4 was transfected with or without Myc-Erbin in HEK 293T cells. Immunoprecipitates with the indicated specific anti-ErbB3 or ErbB4 antibodies were resolved on SDS-PAGE and immunoblotted with an anti-Myc antibody. One-tenth of the input was used in coimmunoprecipitation experiments. by 1% deoxycholate as has been described for various PSD proteins (38,39). The solubility pattern of Erbin was similar to that of ErbB2 (Fig. 5A). These results suggest that Erbin may be associated with the cytoskeletal structure, PSD. In a previous study, we have demonstrated that ErbB2 is localized at the PSD of the brain (13). It is interesting to note that Erbin seems to be a doublet in the brain in a previous study (18) but a singlet using our antibody (Fig. 5). The cause of this apparent difference in chromatographic behavior remains unclear. It may reflect the difference of the antibodies used in the two studies.
The finding that Erbin was present only in membrane fractions and insoluble in various detergents was very interesting because the hydrophobicity profile of the Erbin amino acid sequence did not predict the presence of a transmembrane domain (data not shown). We speculated that Erbin was associated with membranes by interacting with integral proteins such as ErbB2. To test this hypothesis, we biotinylated surface proteins in HEK 293T cells, which were then isolated with streptavidin-agarose beads. As shown in Fig. 5B, only 8.7% of total Erbin was present in the complex pulled down with the beads. In contrast, biotin labeled 35% of total ErbB2, a protein known to have a transmembrane domain. As a control, the amount of ERK1, a cytoplasmic kinase, in the pull-down complex was barely detectable. These results suggest that Erbin may not be as accessible to surface biotinylation in intact cells as the transmembrane ErbB2. Only a minimal amount of Erbin was present in the streptavidin-pull-down complex. To determine whether the presence of Erbin in the streptavidin-pulldown complex was caused by interaction with other proteins, we expressed Erbin and a mutant with deletion of the PDZ domain. As shown in Fig. 5C, the presence of transfected Erbin FIG. 6. Expression of Erbin in muscle cells and at the neuromuscular junction. A, Erbin was expressed in C2C12 and primary myoblasts and myotubes. Homogenates (50 g of protein) of rat brain, skeletal muscles and lysates of C2C12, and rat primary muscle cells were resolved on SDS-PAGE and subjected to immunoblotting with antibodies against Erbin or ErbB2. MB, myoblasts; MT, myotubes; Primary, primary muscle cells. B, C, and D, colocalization of Erbin with ␣-bungarotoxin (␣BTX) in skeletal muscles. Mouse diaphragm sections were incubated with affinity-purified anti-Erbin antibody (B), preimmune serum, or the anti-Erbin antibody preabsorbed with the immunogen. Rhodamine-conjugated ␣-bungarotoxin was added to label the acetylcholine receptor. The Erbin immunoactivity was visualized by fluorescein isothiocyanate-conjugated secondary antibody. Small arrows indicate acetylcholine receptor clusters; large arrows indicate sarcolemma.
FIG. 5. Erbin was a membrane-associated but not integral protein. A, Erbin was tightly associated with membrane fractions in the brain. Rat brain synaptic plasma membranes (P2, 100 g of protein) were resuspended in PBS containing the indicated buffers with high concentration of salt, high pH, or various detergents for 30 min on ice and spun at 10,000 ϫ g for 15 min. The resulting pellets (P) and supernatant (S) were resolved on SDS-PAGE and subjected to immunoblotting with the respective antibodies. B, Erbin was not an integral protein. HEK 293T cells were labeled with sulfo-NHS-LC-biotin and lysed as described under "Experimental Procedures." Lysates (20 times of input) were incubated with streptavidin-agarose beads to pull-down biotinolated proteins or those associated with biotinolated proteins. Isolated proteins were subjected to Western blot for the indicated proteins. The left panel shows blots from a representative experiment; the right panel shows the results of densitometric analysis (mean Ϯ S.D. of three different samples) of autoradiograms, which were scanned and analyzed with NIH Imaging. C, dependence of Erbin presence in pulldown complexes on the PDZ domain. Cells were transfected with green fluorescent protein (GFP)-Erbin or GFP-Erbin⌬PDZ and labeled with sulfo-NHS-LC-biotin and lysed as described in B. Pull-down complexes were probed for transfected proteins with the indicated antibodies.
in the streptavidin-pull-down complex was dependent on the PDZ domain. Deletion of the PDZ domain in Erbin, which blocked binding to ErbB2 (Fig. 1), abolished its presence in the complex. These results suggested to us that Erbin may be a protein in the cytoplasm. It may be tightly associated membranes in a manner dependent on the PDZ domain, probably via interaction with ErbB2. These results, however, were unable to exclude the possibility that Erbin is a transmembrane protein with the extracellular domain somehow inaccessible to biotin labeling.
Localization of Erbin at the Neuromuscular Junction-As with in the brain, Erbin was expressed as a 180-kDa protein in the skeletal muscle, C2C12 mouse muscle cells, and muscle cells in primary culture (Fig. 6A). Expression of Erbin was at similar level in myoblasts and in myotubes, suggesting that differentiation of muscle cells had little effect on its expression. ErbB proteins including ErbB2 are concentrated in the postjunctional membrane at the neuromuscular junction (3, 14 -16, 40) and are required for neuromuscular junction formation (43). We proposed that Erbin could be enriched at the neuromuscular junction by interacting with ErbB2. To determine subcellular localization of Erbin in skeletal muscle, we stained mouse diaphragm sections with affinity-purified anti-Erbin antibody by immunofluorescence techniques. Rhodamine-conjugated ␣-bungarotoxin was used as a marker of the neuromuscular junction (31). The immunoreactivity of Erbin was visualized with a fluorescein isothiocyanate-conjugated secondary antibody. As shown in Fig. 6B, Erbin showed a pattern of labeling strikingly similar to that of the ␣-bungarotoxin staining. Merging the two images indicated that Erbin is localized in precise register with the acetylcholine receptor at the neuromuscular junction. In addition, the Erbin imunoreactivity was also present in the sarcolemma of the skeletal muscle (Fig. 6B, large arrows). Specificity of the staining of Erbin at the neuromuscular junction and sarcolemma was demonstrated by the fact that the preimmune serum produced no staining above the background (Fig. 6C). Furthermore, the Erbin staining was diminished by preabsorbing the antibody with the immunogen (Fig. 6D). These results indicated that Erbin is enriched at the neuromuscular junction and expressed at a low level in sarcolemma.
To determine that the Erbin staining signal at the neuromuscular junction was from the postsynaptic instead of presynaptic components, we studied the effect of denervation on Erbin expression. Denervation of the skeletal muscle causes rapid degeneration of presynaptic nerves. Until the development of pathological conditions (such as atrophy and inflammation), expression of postsynaptic proteins is normal or elevated to compensate the loss of presynaptic input (44). As shown in Fig. 7A, the level of Erbin protein was increased in denervated muscles, and so was that of ErbB2 as observed previously (5). These results suggest that Erbin expression may be regulated by electric activity. Remarkably, Erbin was detectable at the neuromuscular junction in denervated muscles (Fig. 7B), supporting the notion that Erbin was present in the postsynaptic membrane because presynaptic components degenerate in denervated muscles.
Erbin Is a Component of PSD in the Central Nervous System-Expression of Erbin was at similar levels in the cerebral cortex, hippocampus, cerebellum, and brain stem (Fig. 8A). To determine whether Erbin was present in the PSD fraction, we performed subcellular fractionation studies. As shown in Fig.  8B, ErbB2 and Erbin were present in synaptosomes and copurified into PSD fractions. The degree of Erbin enrichment in the PSD fraction correlated strongly with that seen with ErbB2. These biochemical studies demonstrate that Erbin is present in the PSD fraction and suggest that it is appropriately localized to form a protein complex in vivo with ErbB2. Next we determined whether Erbin interacts with ErbB2 in the central nervous system. The interaction between ErbB2 and Erbin was examined in rat brain synaptosomes. Synaptosomes were solubilized with 1% deoxycholate and incubated with antibodies against Erbin. As shown in Fig. 8C, immunoprecipitation of Erbin resulted in coimmunoprecipitation of ErbB2. In the lane where antibodies were missed, ErbB2 was absent in the precipitates (Fig. 8C). Furthermore, preabsorption of Erbin antibodies with the antigen blocked the coimmunoprecipitation of ErbB2 (data not shown). These results suggest that Erbin is associated with ErbB2 in vivo.
Erbin Interacted with PSD-95 in Synaptosomes and in Transfected Cells-Both ErbB2 and ErbB4 are proteins in the central nervous system synapses (13,17) and at the neuromuscular junction (40). Considering that PSD-95 binds to ErbB4 (13,17), we determined whether Erbin and PSD-95 were in the same complex in the central nervous system. Deoxylate-solubilized synaptosomes were incubated with anti-Erbin antibodies to isolate Erbin immunocomplexes. As shown in Fig. 9A, PSD-95 was detected in Erbin immunoprecipitates, suggesting an interaction of Erbin with PSD-95 in vivo. To identify the domains in PSD-95 and Erbin which are required for the in -FIG. 7. Effects of denervation on Erbin expression at the neuromuscular junction. A, expression of Erbin and ErbB2 was increased in denervated muscles. Denervated (Den, 5 days postsurgery) or control (Inn, sham-operated) muscles were homogenized. 50 g of protein was probed for Erbin, ErbB2, or ERK. B, colocalization of Erbin with ␣-bungarotoxin (␣BTX) in denervated muscles. Leg muscle sections were incubated with the anti-Erbin antibody that was visualized by fluorescein isothiocyanate-conjugated secondary antibody. On the immediate right is an image of ␣-bungarotoxin staining and on the far right, an image of overlays. Small arrows indicate acetylcholine receptor clusters; large arrows indicate sarcolemma.
teraction, HEK 293T cells were cotransfected with various constructs (Fig. 9B). The interaction of PSD-95 with Erbin was dependent on the PDZ domains of PSD-95 (Fig. 9D); in fact, the first and second PDZ domains were sufficient for interaction with Erbin (Fig. 9C). On the other hand, the Erbin interaction with PSD-95 did not appear to require Erbin's PDZ domain (Fig. 9E). Further analysis suggested that Erbin may interact with PSD-95 via the region between amino acids 965 and 1241 because an Erbin mutant with a deletion of the N-terminal 1-964 amino acid residues was able to interact with PSD-95 (Fig. 9F).
Erbin Increased Surface Expression of ErbB2-As an initial step to identify the function of Erbin, we investigated the effect of Erbin on ErbB2 surface expression. HEK 293T cells were transfected with Myc-tagged Erbin or a PDZ domain deletion mutant. As shown in Fig. 10A, expression of Erbin had no effect on the total amount of ErbB2 in transfected cells. In contrast, Erbin increased the amount of biotin-labeled ErbB2. Such an increase was dependent on the intact C terminus of Erbin. In cells transfected with the Erbin mutant with deletion in the PDZ domain, the amount of biotin-labeled ErbB2 was similar to that in mock transfected cells (Fig. 10, A and B). These results suggest a possible role of Erbin in the regulation of ErbB2 surface expression. DISCUSSION The major findings of this study are the following. First, the novel PDZ domain-containing protein Erbin interacts specifically with ErbB2, not ErbB3 or ErbB4. Second, Erbin may not be an integral protein but is tightly associated with membranes. Third, this protein, like ErbB2 receptor tyrosine kinase, is enriched both in the postjunctional membrane at the neuromuscular junction and in the PSD of the brain. Fourth, in addition to ErbB2, Erbin also interacts with PSD-95, another PDZ domain-containing protein that interacts with ErbB4. Last, expression of Erbin increases the amount of biotin-la- FIG. 8. Erbin expression in the PSD and interaction with ErbB2 in the central nervous system. A, expression of Erbin in various brain regions. Homogenates (100 g of protein) were resolved on SDS-PAGE and subjected to Western blot using the anti-Erbin antibody. B, expression of Erbin in PSD. Rat brain homogenates (H) were subjected to sequential centrifugations to yield cytosol (S2) and synaptosomes (P2). Washed synaptosomes (P3) were fractionated further by discontinuous sucrose gradient centrifugation to generate synaptosomal plasma membrane (SPM) which was treated with 0.4% Triton X-100. The insoluble SPM was designated as PSD. Samples were separated by SDS-PAGE and subjected to immunoblotting with the respective antibodies. C, interaction between Erbin and ErbB2 in the central nervous system. Rat brain synaptosomes were solubilized with 1% deoxycholate. The resulting detergent extract (Input) was incubated with preimmune serum or antibodies against Erbin. Immunoprecipitates (IP) were resolved on SDS-PAGE and subjected to immunoblotting (IB) for ErbB2. 10 times of input were used for immunoprecipitations.

FIG. 9. Interaction between Erbin and PSD-95. A, interaction
between Erbin and PSD-95 in synaptosomes. Rat brain synaptosomes were solubilized with 1% deoxycholate. The resulting detergent extract (Input) was incubated with preimmune serum, antibodies against Erbin, or anti-Erbin antibodies that were preincubated with excess antigen. Immunoprecipitates (IP) were resolved on SDS-PAGE and subjected to immunoblotting (IB) for PSD-95 and Erbin. 10 times of input were used for immunoprecipitations. B, schematic diagrams of used expression constructs. All constructs were tagged with Myc epitope except ⌬NPDZ123 of PSD-95, which was tagged with a FLAG epitope. C and D, dependence of Erbin interaction with PSD-95 on the PDZ domains of PSD-95. FLAG-tagged ⌬NPDZ123 or Myc-tagged PDZ123, NPDZ123, or PSD-95 was transfected into HEK 293T cells. Erbin was immunoprecipitated and probed with anti-FLAG (in C) or anti-PSD-95 antibodies, which recognize all three PSD-95 recombinant proteins (in D). E and F, independence of Erbin interaction with PSD-95 on the PDZ domain of Erbin. PSD-95, Erbin, Erbin⌬PDZ, or Erbin965 (all Myc-tagged) was transfected or cotransfected into HEK 293T cells. Cell lysates were incubated with anti-PSD-95 antibodies, and the resulting immunocomplexes were subjected to blotting for the indicated proteins.
beled ErbB2 in mammalian cells, suggesting that Erbin was able to increase ErbB2 surface expression. Together with a recent study that suggests a role of Erbin in basolateral localization of ErbB2 in epithelial cells (18), our results suggest that ErbB receptor tyrosine kinases interact with a network of PDZ domain-containing proteins. The interaction between ErbBs and the intracellular PDZ domain-containing proteins may be essential for localized neuregulin signaling in a subcellular compartment including the neuromuscular junction and central synapses. Moreover, the PDZ domain-containing proteins, via interacting with ErbBs, may regulate neuregulin signaling.
Erbin belongs to a unique family of PDZ domain-containing proteins. These proteins include Densin-180; LET-413, an Erbin ortholog in Caenorhabditis elegans (45); and Scribble, a Drosophila protein essential for epithelial integrity (46). In addition to the PDZ domain in the C termini, the members of this family contain 16 LRRs in the N termini and have thus been named as LAP (for LRR and PDZ) proteins (18,45). Among the members of the LAP family, Densin-180 shares a high homology with Erbin in amino acid sequence and overall primary structure (18). Densin-180 is also a postsynaptic component and enriched in the PSD of the brain (35). However, the PDZ domain of Densin-180 does not interact with ErbB2, indicating substrate specificity of PDZ domains of Densin-180 and Erbin and suggesting a different role of these proteins in the brain. The C. elegans ortholog of Erbin, LET-413, is critical for normal assembly of adherens junctions. In LET-413 mutants, adherens junctions are abnormal, cell polarity is affected, and actin cytoskeleton is disorganized (45). Our results suggest that Erbin may play an important role in regulation of neuregulin signaling. Expression of Erbin increased biotin-labeled ErbB2 in transfected cells, indicating that Erbin can promote ErbB2 surface expression. This effect was specific in that it relied on the presence of the PDZ domain, through which Erbin interacted with ErbB2. The Erbin mutant without the PDZ domain did not increase ErbB2 surface expression.
At the neuromuscular junction, proteins essential for neurotransmission are densely packed at the postsynaptic membrane (1, 2, 47). This is caused and maintained at least in part by active transcription in synaptic nuclei. Neuregulin is a mol-ecule from motoneurons which stimulates acetylcholine receptor synthesis (48). In fact, one initially identified isoform of neuregulin is ARIA (for acetylcholine receptor inducing activity) (4). Neuregulin is synthesized in motoneurons (4) released from motoneurons and deposited in the synaptic cleft (49,50), activates ErbB receptors (3,6,51) and subsequent multiple signaling pathways (6,9,52,53) in the skeletal muscle to increase acetylcholine receptor expression. Results from studies of neuregulin-1 and ErbB2 mutant mice indicate that neuregulin is essential for the formation and maintenance of the neuromuscular junction (43,54). In support of this hypothesis are findings that ErbB protein tyrosine kinases and downstream signaling molecules are concentrated at the neuromuscular junction (3,15,16,40,55,56). However, the mechanisms underlying clustering of ErbB proteins in muscle cells remain unclear. We believe that ErbBs are clustered at the neuromuscular junction by interaction via the C termini with a network of anchoring proteins. Of the family of membrane-associated guanyl kinase-like proteins (MAGUK), PSD-95, SAP97, and SAP102 interacted with ErbB4 (13). 2 They are expressed in skeletal muscle cells. 2 Moreover, earlier studies suggested that SAP97 (57) and PSD-95 (58) may be localized at the neuromuscular junction. These proteins may play a role in clustering ErbB4 at the neuromuscular junction. Erbin may be a protein that anchors ErbB2 at the neuromuscular junction. The immunoreactivity of Erbin was enriched at the neuromuscular junction. Furthermore, denervation that destroyed presynaptic components had no apparent effect on Erbin staining in the skeletal muscle, suggesting that Erbin is present in the postsynaptic membrane of the neuromuscular junction where ErbB2 is localized. The hypothesis was supported further by results from the recent study of ErbB2 expression in polarized epithelial cells. ErbB2 is localized at the basolateral side of epithelial cells (18). This localization is dependent on the intact C terminus, the site of interaction with Erbin.
Another interesting finding of this paper is that PSD-95 interacts with Erbin. PSD-95 is a well characterized protein with three PDZ domains in the N terminus, an inactive guanylate kinase domain in the C terminus, and a SH3 domain in between (19 -22). PSD-95, via distinct PDZ domains, interacts with various proteins. In addition, it can form head-to-head multimers via disulfide linkage of its N terminus (41). The interacting proteins of PSD-95 include N-methyl-D-aspartate receptors, potassium channels, neuroligin, SynGap, and CRIP (42). Thus it is believed that PSD-95 is important for the assembly and maintenance of anatomic and/or functional synaptic complex. In this study, PSD-95 interacted with Erbin not only in the heterologous expression system but also in synaptosomes. The interaction of Erbin with PSD-95 did not depend on Erbin's PDZ domain, but a region between amino acids 965 and 1241, suggesting that Erbin may interact simultaneously with ErbB2 and PSD-95. Together with previous observations that PSD-95 interacts with ErbB4 (13,17), these results suggest that ErbB proteins may interact with a network of PDZ domain-containing proteins. It will be interesting to determine whether ErbB clustering requires both Erbin and PSD-95 or a PSD-95-like protein in central nervous system synapses and at the neuromuscular junction.