Densin-180, a synaptic protein, links to PSD-95 through its direct interaction with MAGUIN-1

Authors


  • Communicated by: Fumio Hanaoka

*Correspondence: E-mail: minagaki@aichi-cc.jp

Abstract

Background: Densin-180, a brain-specific protein highly concentrated at the postsynaptic density (PSD), belongs to the LAP [leucine-rich repeats and PSD-95/Dlg-A/ZO-1 (PDZ) domains] family of proteins, some of which play fundamental roles in the establishment of cell polarity.

Results: To identify new Densin-180-interacting proteins, we screened a yeast two-hybrid library using the COOH-terminal fragment of Densin-180 containing the PDZ domain as bait, and we isolated MAGUIN-1 as a Densin-180-binding protein. MAGUIN-1, a mammalian homologue of Drosophila connector enhancer of KSR (CNK), is known to interact with PSD-95 and has a short isoform, MAGUIN-2. The Densin-180 PDZ domain bound to the COOH-terminal PDZ domain-binding motif of MAGUIN-1. Densin-180 co-immunoprecipitated with MAGUIN-1 as well as with PSD-95 from the rat brain. In dissociated hippocampal neurones Densin-180 co-localized with MAGUINs and PSD-95, mainly at neuritic spines. In transfected cells, Densin-180 formed a ternary complex with MAGUIN-1 and PSD-95, whereas no association was detected between Densin-180 and PSD-95 in the absence of MAGUIN-1. MAGUIN-1 formed a dimer or multimer via the COOH-terminal leucine-rich region which is present in MAGUIN-1 but not in -2. Among the PDZ domains of PSD-95, the first was sufficient for interaction with MAGUIN-1.

Conclusion: These results suggest that the potential to dimerize or multimerize allows MAGUIN-1 to bind simultaneously to both Densin-180 and PSD-95, leading to the ternary complex assembly of these proteins at the postsynaptic membrane.

Introduction

Exquisite protein networks asymetrically distributed to pre- and postsynaptic plasma membranes maintain axono-dendritic specialization and homeostasis of chemical synapses and mediate rapid and efficient signal transduction. The afferent postsynaptic part is characterized by a submembraneous electron-dense meshwork, the postsynaptic density (PSD) (Ziff 1997). Densin-180, a brain specific O-sialoglycoprotein, was isolated from the PSD of the rat forebrain and was postulated to act as an adhesion molecule at the PSD (Apperson et al. 1996). Densin-180 interacts with the Ca2+/calmodulin-dependent kinase II α-subunit and α-actinin (Strack et al. 2000; Walikonis et al. 2001). We have recently reported that Densin-180 participates, as a cytosolic protein, in organization of the synaptic cell–cell junction through interaction with δ-catenin/neural plakophilin-related armadillo-repeat protein (Izawa et al. 2002a). In dissociated hippocampal neurones Densin-180 co-localizes with PSD-95, a prototypical membrane-associated guanylate kinase (MAGUK) protein at the PSD (Apperson et al. 1996; Hata & Takai 1999). However, the relationship between these proteins remains unclear. Densin-180 has a unique structure, in that it contains 16 leucine-rich repeats (LRRs) and one PSD-95/Dlg-A/ZO-1 (PDZ) domain (Apperson et al. 1996). Recently, further proteins containing both types of domains have been identified and thus named LAP (leucine-rich repeats and PDZ domain) proteins (Bilder & Perrimon 2000; Legouis et al. 2000; Borg et al. 2000; Saito et al. 2001). To date, a variety of binding partners for these LAP proteins has been found, including the ErbB2/Her2 tyrosine kinase receptor, hemidesmosomal proteins, and p120 related catenins (Borg et al. 2000; Jaulin-Bastard et al. 2001, 2002; Huang et al. 2001; Favre et al. 2001; Laura et al. 2002; Izawa et al. 2002a,b). The striking features of LAP proteins in epithelia are that they play essential roles in defining the shape and apical-basal polarity of epithelial cells as well as carrying their interacting partner(s) to specific plasma membrane domains (Bryant & Huwe 2000; Bilder 2001).

Less is known about the functions of LAP proteins in neurones. Among LAP proteins, Densin-180 expression is restricted to the brain, which suggests that Densin-180, as a LAP protein, triggers functions which are specific to neurones (Apperson et al. 1996). In the course of a yeast two-hybrid screen of a human brain cDNA library—using the Densin-180 COOH terminus containing a PDZ domain as bait—we isolated MAGUK-interacting protein (MAGUIN-1) as a potential interacting partner for Densin-180. MAGUIN-1 was a recently identified protein that binds to PDZ domains of the synaptic scaffolding molecule (S-SCAM)/MAGI-2 and PSD-95 (Hirao et al. 1998, 2000; Hata et al. 1998; Yao et al. 1999). MAGUIN-1 possesses one sterile α motif (SAM), one conserved region in connector enhancer of KSR (CNK), one PDZ, and one pleckstrin homology (PH) domain, as well as having a short isoform, MAGUIN-2 (Yao et al. 1999; Iida et al. 2002). MAGUIN-1 recruits S-SCAM and PSD-95 into the Triton X-100 insoluble fraction in transfected CHO cells (Yao et al. 1999). In addition, a recent study showed that the over-expressed COOH-terminal PDZ domain-binding region of MAGUIN-1 inhibits the synaptic targeting of PSD-95 (Iida et al. 2002). MAGUINs are also a mammalian homologue of a Drosophila CNK, a Raf-interacting protein involved in the regulation of eye development (Therrien et al. 1998, 1999; Yao et al. 2000). CNK together with KSR is requisite for stimulus-dependent Raf activation (Anselmo et al. 2002).

Here we demonstrate that Densin-180 directly interacts with MAGUIN-1 and forms a ternary complex in vivo with MAGUIN-1 and PSD-95. We also characterized molecular mechanisms for ternary complex assembly, in which the potential to dimerize or multimerize allows MAGUIN-1 to engage both Densin-180 and PSD-95 simultaneously. In addition, the self association of MAGUIN-1 depends on the COOH-terminal leucine-rich region which is present in MAGUIN-1, but not in -2. Furthermore, the interaction between PSD-95 and MAGUIN-1 depends on the first PDZ domain of PSD-95.

Results

Identification of MAGUIN-1 as a Densin-180-binding protein

To search for proteins that interact with Densin-180, we screened a human brain cDNA library, using the yeast two-hybrid method and the COOH terminus of human Densin-180 containing a PDZ domain as bait. The resulting positive clones contained three independent clones encoding the COOH terminus of human MAGUIN-1. It has been shown that the messages of both Densin-180 and MAGUIN-1 are restrictively detected in the brain and that both proteins are mainly found in PSD fractions of the rat brain (Apperson et al. 1996; Yao et al. 1999), which prompted us to see whether a biochemical and in vivo interaction exists between Densin-180 and MAGUIN-1. We carried out Blast searches to identify cDNAs encoding human MAGUIN-1, using the rat MAGUIN-1 cDNA sequence (EMBL accession no. AF102853). We identified KIAA0902 cDNA (EMBL accession no. AB020709) encoding the unknown NH2-terminal sequences of human MAGUIN-1 cDNA (amino acid residues 1–858). The protein sequence of human MAGUIN-1 is 99% identical to that of rat MAGUIN-1. We used the cDNA encoding rat MAGUIN-1 in the following experiments. While this study was in progress, DNA sequences identical to those of human MAGUIN-1 and -2 were deposited as a connector enhancer of KSR (CNK)2A (accession no. AF418269) and -2B (accession no. AF418270), respectively.

The Densin-180 PDZ domain interacts directly with the COOH terminus of MAGUIN-1

PDZ domains of mammalian LAP proteins belong to class I domains that interact with peptides containing a COOH-terminal (S/T)XV (in single letter amino acid code, where X is any residue) motif (Hung & Sheng 2002). MAGUIN-1 contains a COOH-terminal THV motif consistent with a canonical class I PDZ domain binding motif. Therefore, the Densin-180 PDZ domain was speculated to bind the COOH-terminal THV motif of MAGUIN-1. We next examined interactions between several constructs of Densin-180 and MAGUIN-1, using yeast two-hybrid methods (Fig. 1A). The COOH-terminal region of Densin-180 lacking the PDZ domain (C-ΔPDZ) did not bind to the wild-type MAGUIN-1 COOH terminus (C-THV (wild)). Conversely, deletion of the COOH-terminal THV motif in MAGUIN-1 abolished this interaction (Fig. 1B). These results indicate that Densin-180 interacts with the COOH terminus of MAGUIN-1 in a PDZ domain-mediated manner.

Figure 1.

The Densin-180 PDZ domain interacts directly with the COOH terminus of MAGUIN-1. (A) Schematic description of various constructs of Densin-180 and MAGUIN-1. Positions of the leucine-rich repeats (LRRs), the PDZ domain (PDZ), the sterile α motif domain (SAM), and the pleckstrin homology domain (PH) are shown. The cDNAs encoding amino acid residues 1242–1537 (C) and 1242–1442 (C-ΔPDZ) of human Densin-180 were cloned into the yeast bait vector (pGBD-C1). The cDNAs encoding amino acid residues 1–667 (N), 668–1032 (C-THV (wild)), and 668–1028 (C-ΔTHV) of rat MAGUIN-1 were inserted into the yeast prey vector (pGAD-C1). (B) Densin-180 PDZ domain interacts with the COOH terminus of MAGUIN-1 containing a PDZ domain-binding motif in yeast two-hybrid system. Yeast strain Y190 co-transformed with the bait and the prey vectors, harbouring various cDNA fragments as shown in (A), were plated on Trp-/Leu-/His- medium containing 25 mm 3-amino-1,2,4-triazole (3-AT). A plus sign (+) indicates growth on selective medium and positive β-galactosidase activity; a minus sign (–) indicates no growth on selective medium and negative β-galactosidase activity. (C) The Densin-180 PDZ domain binds directly to the COOH terminus of MAGUIN-1 in vitro. cDNAs encoding MAGUIN-1 C-THV (wild) and C-ΔTHV fragments as shown in (A), were ligated into the pCMV-Tag3B (Myc) expression vector for coupled in vitro transcription/translation in rabbit reticulocyte lysates, using the TNT kit (Promega). Labelled MAGUIN-1 C-THV (wild-type) and C-ΔTHV polypeptides in lysates were detected using Transcend non-radioactive translation detection systems (Promega) (lane 1). In vitro-translated products were incubated with GST (lane 2) or GST fusion protein encompassing Densin-180 C-terminus fragment, as shown in (A) (GST-Densin-C) (lane 3), and precipitated using glutathione beads. GST-Densin-C bound to MAGUIN-1 C-THV (wild) but not to C-ΔTHV. Molecular size markers (in kiloDaltons) are shown on the left.

We next asked if Densin-180 interacts directly with MAGUIN-1 in vitro. Indeed, GST fusion protein encompassing the PDZ domain of Densin-180 bound specifically to the in vitro-translated MAGUIN-1 COOH terminus (C-THV (wild)), but not with the MAGUIN-1 C-ΔTHV (Fig. 1C). These results indicate that the bacterially expressed Densin-180 PDZ domain interacts directly with the in vitro-translated MAGUIN-1 COOH terminus containing a PDZ domain-binding motif.

Endogenous Densin-180 is associated with MAGUIN-1 and PSD-95 in the rat brain

To explore the in vivo interaction between Densin-180 and MAGUIN-1, we precipitated protein complexes containing Densin-180 from detergent extracts of crude synaptosomal fractions of the rat brain (Fig. 2A). Densin-180 was immunoprecipitated with an anti-Densin-180 antibody, but not with control rabbit IgG, from detergent extracts of the crude synaptosomal fractions. MAGUIN-1 co-immunoprecipitated with Densin-180, when using an anti-Densin-180 antibody. Two major bands of different molecular sizes were detected by immunoblotting with anti-MAGUIN serum, which may represent protein degradation, post-translational modifications, or alternative splicing isoforms, as previously described (Yao et al. 1999). Because MAGUIN-1 interacts with PSD-95 in the rat brain (Yao et al. 1999), we also asked if immunoprecipitates formed by the anti-Densin-180 antibody also contained PSD-95. Indeed, PSD-95 and MAGUIN-1 were co-precipitated by anti-Densin-180 antibody. In addition, immunoprecipitates from anti-MAGUIN serum contained Densin-180 and PSD-95, and precipitates from anti-PSD-95 antibody included Densin-180 and MAGUIN-1. These results indicate that Densin-180 is associated in vivo with MAGUIN-1 and PSD-95.

Figure 2.

Densin-180 is associated in vivo with MAGUIN-1 and PSD-95. (A) Co-immunoprecipitation of endogenous Densin-180, MAGUIN-1 and PSD-95 from the rat brain. Immunoprecipitates were prepared from detergent extracts of crude synaptosomal fractions of the rat brain (lane 1), using control rabbit IgG (lane 2), anti-Densin-180 antibody (lane 3), rabbit pre-immune serum (lane 4), anti-MAGUIN serum (lane 5), control mouse IgG (lane 6), and anti-PSD-95 antibody (lane 7). Each precipitate, separated into three parts, was subjected to immunoblotting with anti-Densin-180 antibody (IB: α-Densin), anti-MAGUIN serum (IB: α-MAGUIN), or anti-PSD-95 antibody (IB: α-PSD-95). A doublet of bands was detected by anti-MAGUIN serum as previously described (Yao et al. 2000). (B) Co-localization of Densin-180, MAGUINs and PSD-95 in dissociated hippocampal neurones. Rat dissociated hippocampal neurones were doubly immunostained with anti-Densin-180 antibody (a) and anti-PSD-95 antibody (b). The merged image of (a) and (b) is shown in (c). Hippocampal neurones were also double stained by anti-MAGUIN serum (d) and anti-PSD-95 antibody (e). The superimposed image of (d) and (e) is shown in (f). Bar = 10 µm.

Densin-180 co-localizes with MAGUINs and PSD-95 in dissociated hippocampal neurones

We next observed the subcellular localization of Densin-180, MAGUINs and PSD-95 in dissociated hippocampal neurones, using immunocytochemical staining (Fig. 2B). Hippocampal neurones plated at E18 were grown in culture for 2 weeks. Double staining with an anti-Densin-180 antibody and an anti-PSD-95 antibody confirmed the co-localization of Densin-180 and PSD-95, mainly at spine-like structures along neurites (Fig. 2B, a–c) as previously reported (Apperson et al. 1996; Izawa et al. 2002a). The staining of MAGUINs was shown to be distributed in the cell body and in neurites, as such being similar to that of N-methyl-D-aspartate (NMDA) receptor NR1 subunits in rat hippocampal neurones (Yao et al. 1999). We observed that the immunocytochemical staining for MAGUINs was also punctate along neurites, and that MAGUINs co-localized with PSD-95 at spine-like structures along neurites and the cell body (Fig. 2B, d–f). These findings suggest that Densin-180, MAGUINs and PSD-95 co-localize preferentially at neuritic spines in rat hippocampal neurones.

Association of Densin-180 with PSD-95 requires the COOH terminus of MAGUIN-1

To further characterize the interaction of Densin-180 with MAGUIN-1 and PSD-95, non-neuronal cells were triply or doubly transfected with full-length Densin-180, MAGUIN-1 COOH terminus (C-THV (wild)), and/or full-length PSD-95 (Fig. 3A). In these experiments, we did not use full-length MAGUIN-1, because the full-length MAGUIN-1 was mainly Triton X-100 insoluble in transfected COS7 cells (data not shown), as previously reported (Yao et al. 1999). Extracts of COS7 cells triply transfected with Densin-180, MAGUIN-1 and PSD-95 were immunoprecipitated with an anti-Densin-180 antibody or control IgG. Both MAGUIN-1 and PSD-95 were detected in the immunoprecipitates using an anti-Densin-180 antibody (Fig. 3A, lane 2), and as shown in Fig. 2A. In COS7 cells doubly transfected with Densin-180 and MAGUIN-1, Densin-180 was co-immunoprecipitated with MAGUIN-1 as expected (Fig. 3A, lane 4). Additionally, in COS7 cells doubly transfected with MAGUIN-1 and PSD-95, MAGUIN-1 co-immunoprecipitated with PSD-95 (Fig. 3A, lane 6) as previously reported (Yao et al. 1999). Next we asked if Densin-180 associates with PSD-95 in transfected cells, because ERBIN was reported to interact with PSD-95 in transfected cells (Huang et al. 2001). In COS7 cells doubly transfected with Densin-180 and PSD−95, no interaction was detected between Densin-180 and PSD-95 (Fig. 3A, lanes 8 and 10). Together with previous studies, these findings in non-neuronal cells (Fig. 3A) suggest the following: Both Densin-180 and PSD-95 bind to the COOH-terminal THV motif of MAGUIN-1, whereas no direct interaction between Densin-180 and PSD-95 exists. Furthermore, Densin-180 can form a tripartite complex with MAGUIN-1 and PSD-95, in which the COOH terminus of MAGUIN-1 is sufficient for ternary complex assembly.

Figure 3.

Association between Densin-180 and PSD-95 requires MAGUIN-1, which dimerizes or multimerizes via the COOH terminus and not via the SAM domain. (A) In non-neuronal cells, Densin-180 forms a ternary complex with MAGUIN-1 and PSD-95, whereas no direct interaction exists between Densin-180 and PSD-95 in the absence of MAGUIN-1. The extracts of COS7 cells expressing Myc-tagged full-length Densin-180 (Densin-180), HA-tagged MAGUIN-1 C-THV (wild) fragment, as shown in Fig. 1A (HA-MAGUIN), and/or FLAG-tagged full-length PSD-95 (FLAG-PSD-95), as indicated (expression; plus or minus signs) were incubated with control rabbit IgG (lanes 1, 3, 5 and 7), anti-Densin-180 antibody (lanes 2, 4 and 8), rabbit anti-HA antibody (lane 6), control mouse IgG (lane 9), or mouse anti-FLAG antibody (lane 10). The immunocomplexes were immobilized on protein G Sepharose beads, and proteins attached to the beads were detected using an anti-Densin-180 antibody (IB: α-Densin), mouse anti-HA antibody (IB: α-HA), or anti-FLAG antibody (IB: α-FLAG). The presence of Myc-tagged Densin-180, HA-tagged MAGUIN-1, and/or FLAG-tagged PSD-95 in each cell extract was confirmed by immunoblotting with the appropriate antibodies (data not shown). (B) The intermolecular interaction between deletion mutants of MAGUIN-1 in the yeast two-hybrid system. Y190 cells co-transformed with the bait (pGBD-C1) and the prey (pGAD-C1) vectors, harbouring cDNA encoding MAGUIN-1 N or C-THV (wild) fragments (as shown in Fig. 1A) as indicated, were selected on Trp-/Leu-/His- medium containing 25 mm 3-AT and subjected to β-galactosidase filter assay. The interactions were determined as described in the legend to Fig. 1B. (C) MAGUIN-1 forms a dimer or a multimer via the COOH terminus in transfected cells. COS7 cells were co-transfected with Myc-tagged (Myc) and HA-tagged (HA) constructs of MAGUIN-1, as indicated (expression). The constructs of MAGUIN-1 ‘N’ and ‘C’ fragments correspond to the ‘N’ and ‘C-THV (wild)’ fragments, respectively, as shown in Fig. 1A. Cell extracts were incubated with control rabbit IgG (lanes 1, 4 and 7), rabbit anti-Myc antibody (lanes 2, 5 and 8), or rabbit anti-HA antibody (lanes 3, 6 and 9). Immunocomplexes were immobilized on protein G sepharose beads, and proteins attached to the beads were detected using a mouse anti-Myc antibody (IB: α-Myc) or a mouse anti-HA antibody (IB: α-HA).

MAGUIN-1 forms a dimer or a multimer via the COOH-terminal leucine-rich region present in MAGUIN-1, but not in -2

Given the interaction of Densin-180 with MAGUIN-1 and PSD-95 in the aforementioned experiments (Fig. 3A), the potential to form a dimer seems to be required for MAGUIN-1, especially the COOH terminus (C-THV (wild) fragment), to bind simultaneously to both Densin-180 and PSD-95. On the other hand, MAGUIN-1 has a SAM domain, and these have been found in a variety of proteins. Most of them mediate homodimerization and/or heterodimerization with other SAM domain-containing proteins (Stapleton et al. 1999; Chi et al. 1999; Sheng & Kim 2000). Next we determined if MAGUIN-1 forms a dimer via the COOH terminus, the SAM domain, or both. We examined the intermolecular interaction of MAGUIN-1 in yeast (Fig. 3B) and transfected cells (Fig. 3C). No self interaction of MAGUIN-1 NH2 termini encompassing the SAM, PDZ and PH domains was detected in either yeast or transfected COS7 cells, whereas the self association of the COOH termini of MAGUIN-1 was evident. In addition, no interaction was detected between the NH2 terminus and the COOH terminus of MAGUIN-1. Furthermore, in COS7 cells coexpressing Myc-tagged and HA-tagged constructs of the MAGUIN SAM domain (amino acid residues 1–157), no self association of MAGUIN SAM domains was detected (data not shown). These results suggest that MAGUIN-1 forms a dimer or a multimer via the COOH terminus, and that the MAGUIN SAM domain does not mediate homodimerization.

To explore the precise mechanism by which MAGUIN-1 forms a dimer or a multimer, we narrowed down the region responsible for self association within the COOH terminus of MAGUIN-1. We developed various truncations of the C-THV (wild) fragment of MAGUIN-1 (MAGUIN-1-C) (Fig. 4A), including the COOH terminus of MAGUIN-2 (Fig. 4A, b), and examined the interactions between the truncations and MAGUIN-1-C in transfected COS7 cells (Fig. 4B). Intriguingly, the MAGUIN-2 COOH terminus did not interact with MAGUIN-1-C (Fig. 4B, b), which implies that the potential to self-associate is restricted to the COOH-terminal stretch present in MAGUIN-1, but not in -2. We found that the amino acid residues 948–976 of MAGUIN-1 contain five hydrophobic amino acids (three leucines, one methionine and one valine) that are spaced every seventh residue, which reminded us of the leucine zipper (Busch & Sassone-Corsi 1990). However, the truncation lacking amino acid residues 946–982 interacted with MAGUIN-1-C (Fig. 4B, lane d). By contrast, truncations lacking amino acid residues 897–945 (containing seven leucines) lost their potential to interact with MAGUIN-1-C (Fig. 4B, lanes e and f), whereas truncations containing these residues did not do so (Fig. 4B, lanes c, d and g). These results suggest that the COOH-terminal leucine-rich region (amino acid residues 897–945), which is only present in MAGUIN-1 and not in -2, is requisite for the self association of MAGUIN-1.

Figure 4.

The COOH-terminal leucine-rich region present in MAGUIN-1 but not -2 is essential for the self association of MAGUIN-1. (A) Schematic description of various Myc-tagged constructs of the MAGUIN-1 COOH terminus. The letter (a–g) to the left of each construct corresponds to the letter of each lane in (B). The numbers refer to amino acid positions. The construct ‘b’ is also consistent with the COOH terminus of MAGUIN-2. (B) A series of Myc-tagged constructs of the MAGUIN-1 COOH terminus, as shown in (A), was expressed in COS7 cells together with HA-tagged MAGUIN-1 C-THV (wild) fragment (as shown in Fig. 1A) (HA-MAGUIN-C). The cell extracts were incubated with rabbit anti-HA antibody. The immunocomplexes were immobilized on protein G Sepharose beads, and proteins attached to the beads were detected using a mouse anti-Myc antibody (IB: α-Myc). The expression of Myc-tagged MAGUIN-1 species was confirmed by immunoblotting of the cell extracts with a mouse anti-Myc antibody (input). Immunoblotting with a mouse anti-HA antibody confirmed the presence of HA-MAGUIN-C precipitated by rabbit anti-HA antibody (IB: α-HA).

The first PDZ domain of PSD-95 is sufficient for interaction with MAGUIN-1

Which PDZ domain(s) of PSD-95 are responsible for interaction with MAGUIN-1 has remained unclear (Yao et al. 1999; Lim et al. 2002). PSD-95 uses three PDZ domains with a distinct specificity to bind and cluster glutamate receptors, ion channels, synaptic adhesion molecules and associated downstream signalling enzymes (Hata & Takai 1999). Furthermore, the third PDZ domain possesses binding requirements differing from those of the first two PDZ domains (Hata & Takai 1999; Lim et al. 2002). To further characterize the assembly of Densin-180-MAGUIN-1-PSD-95 complex, we determined the PDZ domain(s) of PSD-95 which were responsible for the interaction with MAGUIN-1. We constructed various truncations of PDZ domains of PSD-95 (Fig. 5A), and examined the interactions between truncations and the MAGUIN-1 COOH terminus in transfected COS7 cells (Fig. 5B). Truncations lacking the first PDZ domain lost their potential to interact with the MAGUIN-1 COOH terminus (Fig. 5B, lanes c, e and f). These results suggest that the first PDZ domain of PSD-95 is required for the interaction with MAGUIN-1.

Figure 5.

The first PDZ domain of PSD-95 is sufficient for interaction with MAGUIN-1. (A) schematic description of various Myc-tagged constructs of PSD-95 PDZ domains. Positions of the three PDZ domains (PDZ-1–3), the Src homology 3 domain (SH3), and the guanylate kinase domain (GK) are shown. The letters (a–f) to the left of each construct corresponds to the letter of each lane in (B). Numbers refer to amino acid positions. (B) A series of Myc-tagged mutants of PSD-95, as shown in (A), was expressed in COS7 cells together with HA-tagged MAGUIN-1 C-THV (wild) fragment (as shown in Fig. 1A) (HA-MAGUIN-C). The cell extracts were incubated with rabbit anti-HA antibody. The immunocomplexes were immobilized on protein G Sepharose beads, and proteins attached to the beads were detected using a mouse anti-Myc antibody (IB: α-Myc). The expression of Myc-tagged PSD-95 species was confirmed by the immunoblotting of cell extracts with a mouse anti-Myc antibody (input). Immunoblotting with a mouse anti-HA antibody confirmed the presence of HA-MAGUIN-C precipitated using a rabbit anti-HA antibody (IB: α-HA).

Discussion

The major findings of this study are as follows. First, Densin-180 interacts directly with the MAGUIN-1 COOH terminus and this interaction depends on the Densin-180 PDZ domain. Second, Densin-180, MAGUIN-1 and PSD-95 form a ternary complex in vivo at synapses, whereas no direct interaction between Densin-180 and PSD-95 exists. Third, the potential to dimerize or multimerize enables MAGUIN-1 to simultaneously engage both Densin-180 and PSD-95. Fourth, the SAM domain of MAGUINs does not mediate homodimerization, whereas the self association of MAGUIN-1 requires the COOH-terminal leucine-rich sequence between amino acid residues 897 and 945, which is only present in MAGUIN-1 and not in -2. Last, among the PDZ domains of PSD-95, the first mediates the interaction with MAGUIN-1.

Previous studies have indicated that the COOH-terminal PDZ domain-binding motif of MAGUIN-1 interacts with MAGUK family proteins, including S-SCAM, PSD-95, PSD-93/chapsyn-110 and SAP97/hDlg (Yao et al. 1999). Conversely, no LAP proteins have been reported to associate with MAGUIN-1 (CNK). We have obtained the first evidence for the in vivo association of a LAP protein with MAGUIN-1 in the brain. Furthermore, we have found that this novel protein complex also links to PSD-95 in vivo. Taken together with previous observations that Densin-180 and MAGUIN-1 are closely membrane-associated but are not integral proteins (Apperson et al. 1996; Yao et al. 1999; Izawa et al. 2002a), it seems likely that the assembly of these proteins may render their potential to associate with plasma membranes tighter and more stable. In addition, MAGUIN-1 binds the kinase domain of Raf-1, an activator of mitogen-activated protein kinase kinase/extracellular signal regulated kinase kinase (Yao et al. 2000); hence MAGUIN-1 may link Raf-1 to LAP protein Densin-180 or PSD-95 at the synaptic junction (Takai et al. 2001).

Our results also suggest that the first PDZ domain of PSD-95 is required for the interaction with MAGUIN-1. Previous studies have indicated that the first two PDZ domains of PSD-95 bind to Shaker-type potassium channels or NMDA receptor NR2 subunits (Hata & Takai 1999). Recent studies have revealed that all of the last five residues of proteins define the binding selectivity to the PDZ domains of PSD-95, and a COOH-terminal E/Q-S/T-X-V consensus sequence is the strongest binding to the first two PDZ domains of PSD-95 (Lim et al. 2002). This consensus sequence is also consistent with the MAGUIN-1 COOH terminus (ETHV). A sequence alignment of the first two PDZ domains of PSD-95 demonstrates a substantially greater degree of homology among themselves than with the third PDZ domain. It seems exceptional that the MAGUIN-1 COOH terminus binds restrictively to the first PDZ domain of PSD-95. Therefore, we cannot exclude the possibility that MAGUIN-1 may also interact in vivo with the second PDZ domain of PSD-95. It has been shown that PSD-95 forms a dimer or a multimer via NH2-terminal segments (El-Husseini et al. 2000). Therefore, it will be interesting to determine whether PSD-95 could bind simultaneously to both MAGUIN-1 and to ion channels or glutamate receptors at synapses and link them to Densin-180. As compared with PSD-95, little is known about the interacting partners for Densin-180 and MAGUIN-1, hence an identification of proteins that bind to other domains of Densin-180 and MAGUIN-1 (e.g. leucine-rich repeats of Densin-180, SAM and PDZ domains of MAGUIN-1) will aid in elucidating the precise function of the Densin-180-MAGUIN-1-PSD-95 ternary complex in its entirety.

Another interesting finding of this study was that MAGUIN-1 forms a dimer or a multimer via the COOH-terminal region which is present in MAGUIN-1 but not in -2, which implies that MAGUIN-2 cannot form a homodimer and that MAGUIN-1 and -2 cannot form a heterodimer. A previous report indicated the possibility that the SAM domain of Drosophila CNK mediates homodimerization and/or heterodimerization with other SAM domain-containing proteins (Therrien et al. 1999). Our finding raises the possibility that other as-yet unidentified proteins may bind to the SAM domain of MAGUINs. The PH domains and the COOH termini that are both contained in MAGUIN-1 and -2 are sufficient for interaction with Raf-1 (Yao et al. 2000). The COOH-terminal stretch which is only present in MAGUIN-1 most likely confers at least two additional functions, that is, the potential to interact with PDZ domain-containing proteins and to mediate homodimerization or -multimerization. The COOH-terminal residues 897–945 which are responsible for the self association of MAGUIN-1 that we narrowed down did not represent a sequence homologous to any other domains that are known to dimerize or multimerize. Further studies are necessary to elucidate the precisely related molecular mechanisms. In addition, MAGUIN-1 interacts in vivo not only with PSD-95, but also with S-SCAM (Yao et al. 1999); hence it will therefore be intriguing to determine if the MAGUIN-1 homodimer could simultaneously engage divergent MAGUK proteins.

Our studies also suggest the existence of a unique LAP-MAGUK (Densin-180-PSD-95) complex at the synapses, in that this protein complex assembly requires the MAGUIN-1 homodimer. The assembly of mammalian LAP-MAGUK complexes described to date is mediated by a direct protein interaction (Saito et al. 2001; Huang et al. 2001). Another LAP-MAGUK complex in neurones was exemplified by the binding of ERBIN to PSD-95 (Huang et al. 2001). The molecular mechanism underlying this protein interaction is interesting, in that PSD-95 PDZ domains interact directly with the ‘internal’ motif of ERBIN (between amino acid residues 965 and 1241) (Huang et al. 2001). Among the LAP proteins, Densin-180 shares a high homology with ERBIN in amino acid sequence and overall primary structure (Borg et al. 2000), and both Densin-180 and ERBIN are enriched in the PSD of the brain (Apperson et al. 1996; Huang et al. 2001). However, the PDZ domain of ERBIN interacts with the ErbB2 receptor, but that of Densin-180 does not (Huang et al. 2001), thereby suggesting a different role for these proteins in the brain. It is not surprising that two mammalian LAP proteins, ERBIN and Densin-180, associate with PSD-95 in a different manner, i.e. directly and indirectly. In addition, the LAP-MAGUK connection also exists in epithelia. Recent studies have shown that human Lano, an atypical LAP protein that lacks the PDZ domain interacts with hDLG (human discs large) at the baso-lateral membranes of epithelial cells (Saito et al. 2001). This interaction is also PDZ-domain dependent, in that the hDLG PDZ domain binds to the COOH-terminal TSV motif of Lano, matching a canonical class I PDZ domain binding site. Furthermore, a genetic interaction between LAP and MAGUK proteins in epithelia has been demonstrated in Drosophila and C. elegans. In Drosophila, loss of function of either the scribble or the dLg gene results in an overgrowth of epithelial cells of imaginal discs (Bilder et al. 2000). In C. elegans, mutations of let-413 or dLg-1 perturb the apico-basal polarity of epithelial cells and epithelial integrity (McMahon et al. 2001; Bossinger et al. 2001). In addition, recent studies have shown that let-413 and dLg-1 cooperatively regulate the proper localization of an apical junctional protein, and that let-413 may function upstream of dLg-1 (Koppen et al. 2001). Taken together, these findings indicate that interactions between LAP and MAGUK proteins exist in both epithelia and neurones and that in the mammalian brain at least two distinct LAP-MAGUK protein complexes exist, i.e. ERBIN-PSD-95 and Densin-180-MAGUIN-1-PSD-95 complexes. As the aforementioned genetic studies suggest that both LAP and MAGUK proteins belong to a common pathway crucial for maintaining cell polarity and tissue homeostasis in epithelia, neuronal LAP-MAGUK complexes may cooperatively define axono-dendritic specialization in neurones and maintain the homeostasis of synaptic junctions.

By way of summary, we propose that Densin-180 forms a ternary complex with MAGUIN-1 and PSD-95, closely associated with postsynaptic plasma membranes, as depicted in the Summary figure. We cannot exclude the possibility that ternary complexes also assemble partly in protein-sorting processes for the synaptic targeting of these proteins (El-Husseini et al. 2000, 2001). Future studies will shed light on how Densin-180 participates in organization of the protein machinery at the PSD of excitatory synapses, through interaction with MAGUIN-1 and PSD-95.

Experimental procedures

Yeast two-hybrid screening

cDNA encoding the COOH-terminal region of human Densin-180 (amino acid residues 1242–1537) (Densin-C) was cloned into the yeast GAL4 DNA-binding domain vector (pGBD-C1), which was kindly provided by P. James (University of Wisconsin, Madison, WI), as previously described (Izawa et al. 2002a). The resulting plasmid, pGBD-Densin-C, was used in a two-hybrid screen of a human brain cDNA library fused to the pACT2 vector (Clontech, Palo Alto, CA), following the Matchmaker two-hybrid system protocol (Clontech). Positive clones were screened for their potential to grow on selective medium and for the expression of β-galactosidase. Subsequent two-hybrid interaction analyses were carried out by co-transformation of plasmids, containing the GAL4 DNA-binding (pGBD-C1) and -activation (pGAD-C1) domains into Saccharomyces cerevisiae strain Y190, as previously described (Izawa et al. 2002a). The COOH-terminal region lacking the PDZ domain (amino acid residues 1242–1442) (C-ΔPDZ) of Densin-180 was constructed into the pGBD-C1 vector by means of enzyme digestion and use of polymerase chain reaction (PCR) techniques. The cDNAs encoding amino acid residues 1–667 (N) and 668–1032 (C-THV (wild)) of rat MAGUIN-1 were constructed into pGAD-C1 and pGBD-C1 vectors. The cDNA encoding amino acid residues 668–1028 (C-ΔTHV) of rat MAGUIN-1 were cloned into the pGAD-C1 vector. These DNAs were sequenced using an ABI 310 Genetic Analyser (Applied Biosystems, Foster City, CA).

In vitro translation, purification of recombinant protein and pull-down assays

cDNAs encoding amino acid residues 668–1032 (C-THV (wild)) and 668–1028 (C-ΔTHV) of rat MAGUIN-1 were inserted into the pCMV-Tag3B (Myc) expression vector (Stratagene, La Jolla, CA). These pCMV-Tag3B vectors harbouring various constructs of MAGUIN-1 COOH termini were used for coupled in vitro transcription/translation in rabbit reticulocyte lysates using the TNT kit (Promega, Madison, WI). The COOH-terminal fragment of Densin-180 (Densin-C) was expressed as glutathione S-transferase (GST) fusion protein (GST-Densin-C) in Escherichia coli, using the pGEX-4T-2 vector (Amersham Biosciences Corp., Piscataway, NJ) and purified as previously described (Izawa et al. 2002a). Equal amounts of GST or GST-Densin-C fusion protein (≈50 µg of protein) fixed on 20 µL of glutathione-Sepharose 4B beads (Amersham Pharmacia Biotech, Buckinghamshire, UK) were incubated with in vitro-translated products for 4 h at 4 °C on a rotating platform. After centrifugation, the beads were washed four times with a wash buffer containing 20 mm Tris-HCl at pH 7.5, 1% (w/v) Triton X-100 and 150 mm NaCl. Bound polypeptides were eluted with SDS sample buffer, subjected to SDS-PAGE and transferred on to a polyvinylidine difluoride (PVDF) membrane (Atto, Tokyo, Japan). Labelled MAGUIN-1 C-THV (wild) and C-ΔTHV polypeptides in lysates and on the beads were detected using Transcend non-radioactive translation detection systems (Promega).

Antibodies and immunoblotting

A rabbit polyclonal antibody raised against human Densin-180 was used as previously described (Izawa et al. 2002a). The rabbit polyclonal antibody directed against rat MAGUIN-1 was as previously described (Yao et al. 1999). The mouse monoclonal anti-PSD-95 antibody was purchased from Affinity Bioreagents Inc. (Golden, CO). The mouse monoclonal anti-Myc-tag antibody (9E10) was obtained from the American Type Culture Collection, the rabbit polyclonal anti-Myc-tag antibody was from Medical and Biological Laboratories Co. Ltd (Nagoya, Japan), the mouse monoclonal anti-HA-tag antibody was from Boehringer Mannheim (Mannheim, Germany), the rabbit polyclonal anti-HA-tag antibody was from Medical and Biological Laboratories Co. Ltd, and the mouse monoclonal anti-FLAG-tag antibody (M2) was from Sigma (St Louis, MO).

For the purposes of immunoblotting, proteins resolved on SDS-PAGE were transferred on to a PVDF membrane (Atto). The blots were incubated with the appropriate primary antibodies and then horseradish-peroxidase-conjugated secondary antibodies and immunoreactive bands were visualized, using chemiluminescence detection reagents (Renaissance; Perkin-Elmer Life Sciences).

Immunoprecipitation from the rat brain

A rat crude synaptosomal fraction was prepared, as previously described (Hirao et al. 1998; Izawa et al. 2002a). In brief, one adult Sprague–Dawley (SD) rat brain was homogenized in 8 mL of Buffer A (0.32 m sucrose containing 4 mm HEPES/NaOH at pH 7.4) and centrifuged at 800 g for 10 min at 4 °C. The supernatant (S1) was centrifuged at 9200 g for 15 min at 4 °C to collect the pellet. The pellet was resuspended in 8 mL of Buffer A and centrifuged at 10 200 g for 15 min at 4 °C. The pellet (P2; the crude synaptosomal fraction) was resuspended with 4 mL of a buffer containing 20 mm HEPES/NaOH (pH 8.0), 100 mm NaCl, 5 mm EDTA, 1% (w/v) deoxycholic acid, 0.5% (w/v) Triton X-100, 10 µg/mL leupeptin, and 10 µm (p-amidonophenyl) methanesulphonyl fluoride hydrochloride (PMSF), and incubated for 30 min at 36 °C, followed by centrifugation at 100 000 g for 30 min at 4 °C to collect the supernatant. One-millilitre aliquots of the supernatant were incubated for 4 h at 4 °C with anti-Densin-180 antibody, anti-MAGUIN serum, anti-PSD-95 antibody, control rabbit IgG, control mouse IgG, or pre-immune rabbit serum, and immunocomplexes were immobilized on protein G sepharose beads (Amersham Pharmacia Biotech). After the beads were washed four times with buffer containing 20 mm 20 mm HEPES/NaOH at pH 8.0, 100 mm NaCl, and 1% (w/v) Triton X-100, immunoprecipitates on the beads were analysed by immunoblotting with an anti-Densin-180 antibody, anti-MAGUIN serum, or anti-PSD-95 antibody.

Cell culture

Hippocampal neurones prepared from embryonic day 18 (E18) SD rats were seeded on coverslips coated with poly-d-lysine and laminin (Becton Dickinson Labware, Bedford, MA) at a density of 103/mm2 and cultured in Neurobasal Medium (Life Technologies Inc., Rockville, MD) supplemented with B27 supplement (Life Technologies Inc.), 1 mm glutamine and cytosine β-d-arabinofuranoside in an atmosphere of 5% CO2 as previously described (Inagaki et al. 2000). COS7 cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% foetal bovine serum and penicillin in an air−5% CO2 atmosphere with constant humidity.

Immunofluorescence

Rat hippocampal neurones dissociated at E18 were grown in culture on coverslips for 14 days and fixed by incubation in 4% paraformaldehyde in phosphate-buffered saline for 10 min, followed by treatment with −20 °C methanol for 10 min. For double immunostaining for either Densin-180 or MAGUINs and PSD-95, cells were first incubated with anti-Densin-180 antibody or anti-MAGUIN serum followed by Alexa 488-labelled anti-rabbit antibody (Molecular Probes, Eugene, OR). Next the cells were incubated with anti-PSD-95 antibody followed by FluoroLink Cy3-linked anti-mouse antibody (Amersham Pharmacia Biotech). An Olympus LSM-GB200 microscope was used to examine the coverslips.

Mammalian expression vectors, transfection, and immunoprecipitation

A cDNA encoding full-length human Densin-180 (Izawa et al. 2002a) was cloned into the mammalian expression vector pRK5-Myc for the expression of Myc epitope-tagged protein. A cDNA encoding the C-THV (wild) fragment of MAGUIN-1 was cloned into the pRK5-HA vector for expression of haemagglutinin (HA) epitope-tagged protein. The pBJ plasmid containing FLAG-tagged human PSD-95 was kindly provided by Dr H. Saya (Kumamoto University, Kumamoto, Japan). A cDNA encoding full-length PSD-95 with FLAG tag was cloned into the pRK5-Myc vector. cDNAs encoding the N and C-THV (wild) fragments of MAGUIN-1 were inserted into the pRK5-Myc and pRK5-HA vectors. A series of truncations of MAGUIN-1 and PSD-95 were constructed into the pRK5-Myc vector by enzyme digestion using PCR techniques. Sequences of these constructs were verified by DNA sequencing.

COS7 cells grown on a 6-cm plate were transfected, using LipofectAMINE PLUS (Life Technologies Inc.) according to the manufacturer's protocols. Sixteen hours after lipofection, the cells were lysed on ice for 20 min in lysis buffer containing 20 mm Tris-HCl at pH 7.5, 1% (w/v) Triton X-100, 50 mm NaCl, 1 mm EDTA, 10 µg/mL leupeptin and 10 µm PMSF. Lysates, clarified by centrifugation at 14 000 g for 30 min at 4 °C, were then incubated with an appropriate antibody, and immunocomplexes were immobilized on protein G sepharose beads. After washing the beads four times with 20 mm Tris-HCl at pH 7.5 containing 1% (w/v) Triton X-100 and 100 mm NaCl, immunoprecipitates were analysed by immunoblotting with an appropriate antibody.

Acknowledgements

We thank H. Saya (Kumamoto University, Kumamoto, Japan) for generously providing the pBJ-FLAG-PSD-95 plasmid and P. James (The University of Wisconsin, Madison, WI) for kindly providing the pGBD-C1 and pGAD-C1 plasmids. We are grateful to M. Ohara for critical comments on the manuscript. This work was supported in part by Grants-in-aid for Scientific Research and Cancer Research from the Ministry of Education, Culture, Sports, Science and Technology of Japan; by the Japan Society of the Promotion of Science Research for the Future; by a Grant-in-aid for the Second Term Comprehensive 10-Year Strategy for Cancer Control from the Ministry of Health and Welfare, Japan; and by Research Grant of the Princess Takamatsu Cancer Research Foundation.

Ancillary