A Truncated Tropo-Myosine-Related Kinase B Receptor, T1, Regulates Glial Cell Morphology via Rho GDP Dissociation Inhibitor 1

Koji Ohira,1,2 Haruko Kumanogoh,1 Yoshinori Sahara,1 Koichi J. Homma,3 Hirohisa Hirai,2 Shun Nakamura,1 and Motoharu Hayashi2 1Department of Biochemistry and Cellular Biology, National Institute of Neuroscience, National Center of Neurology and Psychiatry, Tokyo 187-8502, Japan, 2Department of Cellular and Molecular Biology, Primate Research Institute, Kyoto University, Inuyama, Aichi 484-8506, Japan, and 3Department of Molecular Pathology, Faculty of Pharmaceutical Sciences, Teikyo University, Kanagawa 199-0195, Japan


Introduction
Brain-derived neurotrophic factor (BDNF) is enriched in the CNS and plays pivotal roles in neural survival, differentiation, and plasticity (Bibel and Barde, 2000;Thoenen, 2000). The effects of BDNF are transduced through the tropo-myosine-related kinase B (TrkB) receptor (Barbacid, 1994). There are three TrkB receptor isoforms in the mammalian CNS (Barbacid, 1994). The full-length isoform (TKϩ) is a typical tyrosine kinase receptor and transduces the BDNF signal (Kaplan and Miller, 2000). In contrast, two truncated isoforms (TKϪ: T1 and T2) possess the same extracellular domain, transmembrane domain, and first 12 intracellular amino acid sequences as TKϩ. However, the C-terminal sequences are the isoform specific (11 and 9 amino acids, respectively) (Barbacid, 1994).
Currently, TKϪ, especially T1, is hypothesized to be a dominant-negative form of TKϩ and is involved in negative functions against TKϩ, such as the TKϩ phosphorylation (Knüsel et al., 1994), the calcium efflux (Eide et al., 1996), the cell survival activity (Haapasalo et al., 2001), and gene expression by BDNF (Offenhäuser et al., 2002). According to this hypothesis, TKϪ is postulated to form the homodimer or heterodimer with TKϩ, which prohibits TKϩ signaling or limits the availability of BDNF to the neural tissue by trapping excess BDNF.
In contrast, there are several findings that provide evidence against the hypothesis that T1 is a dominant-negative form of TKϩ. For example, the expression of T1 increases markedly at various important periods in the developing mammalian CNS, such as axonal remodeling and synaptogenesis (Allendoerfer et al., 1994;Fryer et al., 1996;Ohira et al., 1999Ohira et al., , 2001. The specific alignment of the intracellular domain of T1 is completely identical among mice, rats, and humans (Klein et al., 1990;Middlemas et al., 1991;Shelton et al., 1995), suggesting that the alignment plays a unique role. In addition, T1 is capable of binding BDNF at the same level as does TKϩ (Biffo et al., 1995). As regards the physiological function of T1, it is involved in the control of the elongation of distal dendrites of cortical pyramidal neurons (Ya-coubian and Lo, 2000) and the BDNF-induced calcium entry in astrocytes (Rose et al., 2003). Based on these results, we considered a new hypothesis, namely, that T1 binds to proteins through its C-terminal-specific sequence, which elicits a unique type of signal transduction other than the well understood regulation of the tyrosine kinase pathway. In fact, T1 has been reported to mediate signal transduction (i.e., the acid metabolite release from cells) (Baxter et al., 1997).
To clarify the T1 signaling cascade, in this study, we first performed affinity purification with a T1-specific sequence and then identified Rho GDP dissociation inhibitor 1 (GDI1) as a T1 binding protein from the rat brain. Rho GDI1 is an inhibitory regulator of Rho GTPases that can regulate cell morphology via the remodeling of the cytoskeleton. Furthermore, we provide evidence that T1 is capable of ligand-mediated signaling through Rho GDI1 and of regulating astrocytic morphology in primary cultures.

Materials and Methods
Affinity chromatography. All experimental procedures for animals were performed in accordance with our institutional guidelines (1996). Young adult (4-week-old) Wistar rat whole brains (10 g) were homogenized in 10 vol of homogenization buffer (0.32 M sucrose, 5 mM Tris-HCl, pH 7.5, and 150 mM NaCl containing 1 mM PMSF, 10 g/ml leupeptine, and 20 g/ml aprotinin). After centrifugation at 100,000 ϫ g at 4°C for 1 h, the supernatant was adjusted to a concentration of 1 mg/ml protein, and this solution was defined as the cytosolic fraction. Eleven synthesized amino acid residues (FVLFHKIPLDG) of the C terminal of T1 were conjugated to a poly-␤-hydroxybutyrate-Tenta Gel S (Shimadzu, Kyoto, Japan) matrix. The affinity column was equilibrated with the homogenization buffer. Another column, without the synthetic peptides, was prepared as a control column. Ten milliliter aliquots of the cytosolic fraction were applied to the control column and then were loaded onto the affinity column. After the affinity column was washed with homogenization buffer containing 500 mM NaCl, the bound proteins were eluted in one step with 50 mM glycine, pH 2.5. Ten microliters of each fraction (200 l) were subjected to SDS-PAGE (10% gel). The proteins in the gels were then silver stained.
Amino acid sequence analysis. The fractions containing the 28 kDa protein, which had been obtained from 20 independent affinity chromatography experiments, were concentrated using Centricon YM-10 (Millipore, Bedford, MA). The proteins were transferred to a polyvinylidene difluoride (PVDF) membrane (Millipore) using a buffer containing 10 mM 3-[(3-cholamidopropyl)dimethylammonio]-1-propanesulfonate and 10% methanol, pH 11. The membranes were stained with 0.1% Ponceau S in 1% acetic acid and were destained with distilled water. For the peptide sequence, immobilized protein bands were cut with 2 mg of CNBr and placed in 200 l of 70% formic acid in an Eppendorf (Eppendorf Scientific, Westbury, NY) tube overnight. The resulting solution and membranes were dried and then boiled in SDS sample buffer. Tricine-SDS-PAGE was used to segregate the small peptides (Ͻ10 kDa) (Ploug et al., 1989). The cut peptides were transferred to PVDF SQ membranes (Millipore), which were then stained with 0.1% Ponceau S in 1% acetic acid and destained with distilled water. The bands were applied to a Sequencer (476A protein sequencer; Applied Biosystems, Foster City, CA).
Cell cultures. For the cell cultures, human embryonic kidney 293 (HEK293) cells were kept in DMEM supplemented with 10% FBS. Astrocytic primary cultures were prepared from neonatal rat pups (Sprague Dawley). The hippocampi were cut into 1 mm slices, incubated in activated papain (20 U/ml, 20 min), and dissociated by gentle trituration (Sahara and Westbrook, 1993). Dissociated astrocytes from neonatal rats were plated at 300,000 cells per dish on cover glasses coated with poly-Llysine (Sigma, St. Louis, MO). The culture medium contained MEM (Invitrogen, Carlsbad, CA), 0.6% glucose, 5% heat-inactivated FBS, and penicillin-streptomycin (Invitrogen). At 3 d in vitro (DIV) after plating, the expression plasmid vectors (see below) were transfected into astrocytes with Lipofectamine 2000 (Invitrogen). At 4 h after transfection, the culture media were exchanged to DMEM containing N2 supplement (Invitrogen). After 24 h, the cells were used for the experiments. For the long-term cultured astrocytes, cells were incubated in DMEM with 5% FBS, and the medium was exchanged every 3 d. At 72 h before the experiments, the culture media were replaced with DMEM containing the N2 supplement. Additionally, the media of all cultures were exchanged for fresh media at 2 h before the experiments.
DNA constructs and transfection. The T1 cDNA obtained from the adult rat cortex was subcloned into the EcoRI sites of p internal ribosomal entry site (IRES) 2-enhanced green fluorescent protein (EGFP) (Clontech, Palo Alto, CA), and the resulting construct was designated as pT1-IRES-EGFP. Mutant constructs of T1 were prepared by a PCR mutagenesis method. Briefly, we used the 5Ј primer for T1, GGTCT-GCCGTCTGCACGTCTG, and the 3Ј primers A, CGCGGATCCCTA-GAGCAGAAGCAGCATC; B, CGCGGATCCTTAACCTTTCATGCC; C, CGCGGATCCCTAAACAAAACCTTTC; D, CGCGGATCCCTAAT-GAAACAAAACAAAAC; and E, CGCGGATCCCTAGGGGATCT-TATG. The underlined sequences indicate the BamH1 sites. The boldface letters represent mutation sites. PCR was performed at 94°C for 30 s, 58°C for 30 s, and 72°C for 60 s for 30 cycles, followed by a final elongation at 72°C for 5 min with an Expand High Fidelity PCR system (Roche Applied Science). The amplified PCR products were separated on 2% agarose gel. The PCR products were digested at the 5Ј BstPI and the 3Ј BamH1 sites, and the digested products were subcloned into pT1-IRES-EGFP. The vectors of enhanced cyan fluorescent protein (ECFP)-⌬11 and ECFP-intracellular domain (ICD) were also prepared by PCR, using the 5Ј primer GGTCTGCCGTCTGCACGTCTG and the 3Ј primer of ⌬11, CGCGGATCCTTAACCTTTCATGCC or the 3Ј primer of ICD, CGCGGATCCCCCAGCCTTGTCTTTCCTTTATC. The deletion mutants are shown in Figure 3. The underlined sequences indicate the BamH1 sites, and the boldface letters represent mutation sites. PCR was performed at 94°C for 30 s, 58°C for 30 s, and 72°C for 60 s for 30 cycles, followed by a final elongation at 72°C for 5 min with an Expand High Fidelity PCR system (Roche Applied Science). The amplified PCR products were separated on 2% agarose gel. The PCR products were digested by the 5Ј and 3Ј BamH1 sites and were then subcloned into these sites of pECFP-C1 (Clontech). These constructs were transfected into HEK293 cells and astrocytes with Lipofectamine 2000.
Production of glutathione S-transferase-fusion proteins and in vitro binding assay. Constructs of glutathione S-transferase (GST)-fusion proteins were prepared by a PCR method. For the preparation of GST-T1-ICD, we used the above pT1-IRES-EGFP as a template, the 5Ј primer CAAGAATTCCTCCAAGTTGGCGAGACATTCC, and the 3Ј primer GTTGTCGACTTGTCTTTCCTTTATCTCAG. The single-underlined sequence indicates the EcoRI site, and the double-underlined sequence indicates the SalI site. PCR was performed at 94°C for 30 s, 58°C for 30 s, and 72°C for 60 s for 25 cycles, followed by a final elongation at 72°C for 5 min with an Expand High Fidelity PCR system. The digested PCR product was subcloned into the pGEX-5X-3 bacterial expression vector (Amersham Biosciences, Piscataway, NJ). To obtain the cDNA of Rho GDI1, we used a cDNA library of the adult mouse cortex as a template, the 5Ј primer CACGAATTCTAGGGCAGAACAGGACC, and the 3Ј primer GTTGTCGACTAGGTAGGGGGTTAG. A single-underlined sequence indicates the EcoRI site, and a double-underlined sequence indicates the SalI site. The boldface letter in the 5Ј primer of GST-Rho GDI1 is the point mutation site. The methods used for PCR preparation and subcloning into the pGEX-5X-3 vector were the same as those used for T1. After the OD 600 reached 0.6, 1 mM isopropyl-1-thio-␤-Dgalactopyranoside was added to the cultures, and Escherichia coli were grown for an additional 16 h at 25°C (Yamashita and Tohyama, 2003). After the cells were collected, they were resuspended in PBS and sonicated. To the cell lysates, 0.5% Triton X-100 was added, and the samples were incubated for 30 min at 4°C. After centrifugation at 10,000 ϫ g for 5 min, glutathione-Sepharose 4B (Pharmacia, Piscataway, NJ) was added to the supernatants, which were then incubated for 30 min at 4°C. After centrifugation at 10,000 ϫ g for 5 s, the beads were washed three times in PBS containing 0.5% Triton X-100. The purity of the proteins was determined by SDS-PAGE. Then, glutathione-Sepharose 4B with GST-T1-ICD was used for the binding assay. To remove the GST moiety from GST-Rho GDI1, Factor Xa (Novagen, Darmstadt, Germany) was added to the glutathione-Sepharose 4B with GST-Rho GDI1, and the samples were incubated for 16 h at 20°C. After centrifugation at 10,000 ϫ g for 5 min, Xarrest agarose (Novagen) was added to the supernatants, and the samples were incubated for 10 min at room temperature. After centrifugation at 1000 ϫ g for 5 min, the supernatants were designated as the recombinant Rho GDI1 without GST. To check the cleavage of GST-Rho GDI1, glutathione-Sepharose 4B was added into the supernatant of Rho GDI1, and the samples were incubated for 30 min at 4°C. After centrifugation at 10,000 ϫ g for 5 min, the precipitates were washed three times in PBS containing 0.5% Triton X-100, and they were then boiled in SDS sample buffer (see supplemental Fig. 1, available at www.jneurosci.org as supplemental material).
For the in vitro binding assay, recombinant Rho GDI1 solution was added to the glutathione-Sepharose 4B with GST-T1-ICD, and the samples were incubated for 1 h at 4°C with agitation. After centrifugation at 10,000 ϫ g for 5 s, the precipitates were washed three times in PBS containing 0.5% Triton X-100, and then they were boiled in SDS sample buffer.
Precipitation assays. After each incubation with reagents, the cells were lysed with 0.15 ml of lysis buffer. For Rho GDI1-T1 coimmunoprecipitation, lysis buffer A (10 mM triethanolamine, 10 mM iodoacetoamine, pH 7.8, 150 mM NaCl, 2 mM EDTA, 1% digitonin, 1 mM PMSF, 10 g/ml leupeptin, and 20 g/ml aprotinin) was used. The lysates were centrifuged at 10,000 ϫ g at 4°C for 20 min. Then, 50 l aliquots of resulting supernatants were designated as total protein samples. Normal mouse IgG and protein G-Sepharose were added to the remaining supernatants, which were incubated at 4°C for 1 h with gentle rotation. After centrifugation at 5000 ϫ g at 4°C for 1 min, mouse monoclonal anti-pan-TrkB (2 l; Transduction Laboratories, Lexington, KY) or rabbit polyclonal anti-Rho GDI1 (3 l; Santa Cruz Biotechnology, Santa Cruz, CA) was added. In the competitive assays with the synthetic peptides of the T1 C terminal, the peptides (final concentration, 100 M and 1 mM) were added to the lysates and incubated at 4°C for 1 h. The samples were incubated at 4°C for 2 h with antibody and then were incubated with protein G-Sepharose at 4°C for 1 h with gentle rotation. The precipitates were washed four times with lysis buffer A and boiled in 40 l of SDS sample buffer for 3 min.
For the RhoA, Rac1, and Cdc42 pull-down assay, we used lysis buffer B [50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 5 mM MgCl 2 , 0.5% Triton X-100, 1 mM PMSF, 10 g/ml leupeptin, 20 g/ml aprotinin, and 10 nM microcystin LR (Leu and Avg)]. The lysates were centrifuged at 10,000 ϫ g at 4°C for 20 min. We then performed the RhoA pull-down assay with Rhotekin beads (Upstate Biotech, Charlottesville, VA) according to the method of Ren et al. (1999) and the Rac1 and Cdc42 pull-down assay with p21-activated kinase (PAK) beads (Upstate Biotech). Then, the Rhotekin (Upstate Biotech) or PAK beads (30 l) were added to the lysates (1 mg of protein/ml) and incubated at 4°C for 45 min. The beads Figure 1. Affinity purification of a binding protein of T1. A, The cytosolic fraction from rat brain was loaded onto an affinity column with the T1 synthetic peptide. One-step elution with glycine (50 mM; pH 2.5) was performed, and elution samples from the column were separated; an aliquot of each eluate was subjected to SDS-PAGE (10% gel). The resulting silver-stained gel image is shown here. The arrowhead indicates the position of the 28 kDa protein. B, Western blot analysis of the eluate from the control column (lane 1) and the affinity column (lane 2) with anti-Rho GDI1. The arrowhead indicates the position of Rho GDI1 (28 kDa).

Figure 2.
Direct binding of Rho GDI1 to T1 and binding motif in the T1-specific sequence. A, The GST moiety or the GST-T1-ICD was mixed with recombinant Rho GDI1 in Eppendorf tubes. The GST proteins were precipitated with glutathione-Sepharose 4B. The coprecipitated Rho GDI1 was detected with anti-Rho GDI1. The arrowhead indicates Rho GDI1 (28 kDa). B, The Rho GDI1 binding site in the intracellular domain of T1. A display of T1 and its deletion mutants is shown in the top panel. The white boxes represent the extracellular and transmembrane domains, and the underlined sequence indicates the specific amino acid sequence of T1. The constructs were transfected into HEK293 cells. After 24 h, coimmunoprecipitation with anti-Rho GDI1 was performed. N, No vector; V, empty vector expressing GFP; T1, normal T1; T1-⌬ICD, T1 without its ICD; T1-⌬n, T1 deletion mutants lacking the indicated number of amino acids from the C terminal. IB, Immunoblot; IP, immunoprecipitation.
were washed three times with lysis buffer B. The pellets were then mixed with 40 l of SDS sample buffer and boiled for 3 min. In the control assays using GTP␥S-and GDP-loaded lysates, we confirmed our assay systems (supplemental Fig. 2, available at www.jneurosci.org as supplemental material). Briefly, 3 l of 0.5M EDTA (10 mM) and then 1.5 l of GTP␥S (100 M) or GDP (1 mM) was added to a 0.15 ml aliquot of each cell extract. The extracts were incubated at 30°C for 30 min. To stop the loading of GTP␥S and GDP, we added 9 l of 1 M MgCl 2 (60 mM). The procedures that were subsequently performed have been described above.
Western blot analysis. Using an aliquot of astrocytic culture lysates, we also performed a Western blot analysis of protein expression. The samples (5 g total protein per lane, except for TrkB, Cdc42, and Rac1, as follows: 100 g total protein per lane for TrkB, Cdc42, and Rac1 and 10 l per lane for the precipitates) were subjected to SDS-PAGE and then were blotted onto PVDF membranes. The membranes were blocked for 1 h in 5% skim milk in PBS [containing (in mM): 137 NaCl, 8.1 Na 2 HPO 4 ⅐7H 2 O, 2.7 KCl, and 1.5 KH 2 PO 4 ]. After incubation with the primary antibodies at room temperature for 1 h, the blots were incubated for 1 h with secondary antibodies conjugated with HRP and then were visualized by the ECL system (Amersham Biosciences). For the primary antibodies, we used anti-pan-TrkB Immunohistochemistry. Young adult rats (4-week-old Wistar rats) were anesthetized and perfused with 4% formaldehyde in phosphate buffer. The brains were postfixed for 6 h and cryoprotected in 30% sucrose in PBS. The brains were mounted in Tissue-Tek (Miles, Elkhart, IN), frozen rapidly on dry ice, and stored at Ϫ30°C. The sections were cut to a thickness of 35 m with a cryostat (Leica, Wetzlar, Germany). The sections were mounted on glass slides coated with 3-aminopropyltriethoxysilane, washed for 30 min with PBS, and then preincubated with PBS-GB [4% normal goat serum (Vector Laboratories, Burlingame, CA) and 1% bovine serum albumin in PBS] for 2 h at room temperature. The sections were incubated for 48 h at 4°C with antibodies. We used the following primary antibodies: rabbit polyclonal anti-TKϩ (1:800) and anti-T1 (1:800) and mouse monoclonal anti-neurofilament (1:1000; clone SMI32; Sternberger Monoclonals, Lutherville, MD), anti-glutamic acid decarboxylase (GAD; 1:3000; Affinity Research Products, Exeter, UK), and anti-glial fibrillary acidic protein (GFAP; 1:1000; Chemicon, Temecula, CA). The sections were incubated for 1 h at room temperature with the following secondary antibodies: anti-mouse IgG cyanine 3 (Cy3; 1:200; Chemicon) and anti-rabbit IgG Alexa 488 (1:200; Molecular Probes, Eugene, OR). The sections were embedded with Permafluor (Thermo Shandon, Pittsburgh, PA). We used a confocal microscope (TCS SP2; Leica) to analyze the samples.
Morphological assays. The cells were stimulated for the indicated periods at 37°C with 20 ng/ml BDNF (PeproTech, Rocky Hill, NJ) or 100 ng/ml NGF (PeproTech) or vehicle. The cell samples were also incubated for 20 min with anti-BDNF (5 g/ml; Santa Cruz Biotechnology) and then were incubated for 30 min with 20 ng/ml BDNF. For treatment with Toxin A (Biogenesis, Poole, UK) and C3 toxin (Calbiochem, La Jolla, CA), we performed the procedures according to methods described previously (Just et al., 1995;Maekawa et al., 1999). Toxins (20 ng/ml of Toxin A, 30 g/ml of C3 toxin) were added to the 30 DIV cells, and the cells were incubated for 24 h at 37°C. The cells were washed twice in PBS and then were fixed in 4% formaldehyde in PBS for 1 h at room temperature. The cells were preincubated with PBS-GB and were incubated with anti-GFAP for 48 h at 4°C. After being washed in PBS, the cells were incubated with anti-mouse IgG Cy3. The cells were then embedded with Permafluor. We used a confocal microscope for the analysis. For the time-lapse analysis, the astrocytic cultures were set on the confocal microscope (TCS SP2; Leica) with oxygen supply. Phase-contrast images were taken using a 40ϫ water-immersion objective at indicated time. Each cell area was measured by AquaCosmos (Hamamatsu Photonics K.K., Hamamatsu, Japan).

Results
Rho GDI1 is a T1-binding protein T1 exhibits a characteristic developmental expression pattern in the mammalian CNS (i.e., the expression of T1 is known to be remarkably increased after birth and is a major product among TrkB subtypes in adults) (Allendoerfer et al., 1994;Fryer et al., 1996;Ohira et al., 1999). Thus, we purified T1-binding proteins from the cytosolic fraction of adult rat brains, using an affinity column conjugated with the C-terminal-specific sequence of T1. A 28 kDa protein was eluted from the column as a sharp peak under low pH conditions using glycine buffer, pH 2.5 (Fig. 1 A). The fractions containing the protein eluted from the affinity column were concentrated by a centrifugal concentrator. Then, the 28 kDa protein was purified as a single band blotted on a PVDF membrane and cleaved by CNBr. The resulting peptides were separated by gel electrophoresis and blotted on a PVDF membrane to purify each band. The N-terminal sequence of one peptide was determined as KYIQHT according to the Edman degradation method. Consequently, this sequence was found to match the inner sequence of Rho GDI1, a Rho guanine nucleotide dissociation inhibitor that can stabilize the inactive, GDP-bound form of Rho GTPase (Takai et al., 2001). Western blot analysis identified the 28 kDa protein as Rho GDI1 (Fig. 1 B).

In vitro binding assay
To examine whether T1 directly binds to Rho GDI1, we performed an in vitro pull-down assay using recombinant proteins (i.e., the GST-T1-ICD and Rho GDI1). As shown in Figure 2 A, the GST moiety did not bind to Rho GDI1, whereas the GST-T1-ICD fusion protein precipitated Rho GDI1. The possible contribution of the direct binding of GST-Rho GDI1 to glutathione-Sepharose 4B as a result of incomplete cleavage could be excluded, because we detected the Rho GDI1 at 28 kDa but not the 54 kDa band of the fusion protein. Therefore, we concluded that T1 directly binds to Rho GDI1.

Binding motif of T1 with Rho GDI1
We further determined a specific motif of T1 binding to Rho GDI1 using deletion mutants of the C terminal of T1. Constructs of T1 lacking its intracellular domain (T1-⌬ICD) and T1 deletion mutants lacking the indicated number of amino acids from the C-terminal domain (T1-⌬n) were transfected into HEK293 cells, and coimmunoprecipitation with anti-Rho GDI1 antibody was performed at 24 h after transfection. We detected the bands of both T1-⌬3 and T1-⌬6 at ϳ95 kDa, which was comparable with those of normal T1 (Fig. 2 B). However, deletion mutants lacking nine or more amino acids were no longer able to bind to Rho GDI1. Thus, the present results suggested that LFH in the T1specific sequence (FVLFHKIPLDG) is responsible for binding to Rho GDI1.

Expression of BDNF receptors in astrocytic primary cultures and adult rat brains
T1 has been reported to be distributed in both neurons and glia (Frisén et al., 1993;Armanini et al., 1995;Ohira and Hayashi, 2003). In astrocytic primary cultures from the neonatal rat hippocampus, RT-PCR analysis did not reveal the mRNA expression of any of the TrkB subtypes or of p75 at 2 DIV after plating (Fig.  3A). Astrocytes cultured long-term (30 DIV) expressed T1 mRNA, whereas no TKϩ, T2, or p75 mRNA expression was detected. We also examined the distribution of T1 in the adult rat cortex using fluorescent double-staining histochemistry. In this series, the following cell markers were used: neurofilament for pyramidal neurons, GAD for GABAergic neurons, and GFAP for astrocytes. TKϩ immunoreactivity was localized in both pyramidal and GABAergic neurons but not in astrocytes (Fig. 3B). In contrast, T1 immunoreactivity was detected not only in both pyramidal and GABAergic neurons, but also in the astrocytes (Fig. 3C). The neurotrophin receptor p75 (p75) was only expressed in the pyramidal neurons (Fig. 3D). Together, these results indicate that astrocytes in the cortex and hippocampus of adult rats possess only T1 among the known BDNF receptors. Therefore, we used rat hippocampal astrocytes to investigate the signaling mechanism of T1.

Dissociation of Rho GDI1 from T1 in a BDNF-dependent manner
To determine whether Rho GDI1 dissociated from T1 in a BDNF-dependent manner in long-term cultured (30 DIV) astrocytes, we performed a pull-down assay of T1 with anti-Rho GDI1 antibody, and we detected T1 using anti-T1 antibody. As shown in Figure 4 A, the T1 band was reduced to ϳ60% of the control level at 20 min after BDNF treatment (20 ng/ml). The reduced levels of T1 bands were maintained for 60 min, and then a less significant reduction in T1 bands (70% of the control level) was observed at 120 min after the addition of BDNF. Moreover, the dissociation of Rho GDI1 from T1 appeared to occur in a dose-  3 and 4). D, Coimmunoprecipitation of Rho GDI1 with anti-pan-TrkB. Astrocytes were stimulated for 30 min by BDNF (20 ng/ml). The arrowhead indicates Rho GDI1 (28 kDa). Quantitative analysis of the bands in A-D. The control levels were taken as 100%. The asterisks indicate statistically significant differences ( p Ͻ 0.05; one-way ANOVA and Scheffé's post hoc test). In C, the # symbol indicates significant differences between the BDNF and the peptide treatments ( p Ͻ 0.05; one-way ANOVA and Scheffé's post hoc test). Values are given as means Ϯ SD of four independent experiments. IB, Immunoblot; IP, immunoprecipitation. dependent manner (Fig. 4 B). We found that 50 ng/ml BDNF stimulation for 30 min led to the adequate dissociation of Rho GDI1, which then reached a plateau level (50% of control level).
Furthermore, to determine whether the interaction between T1 and Rho GDI1 was specific, we performed the peptide competition assays using the T1-specific C-terminal peptide. The peptides were added to the lysates derived from 30 DIV astrocytes to a final concentration of 100 M or 1 mM. Both additions of 100 M and 1 mM peptides significantly inhibited the interaction (4.7 and 1.4% of control level, respectively) (Fig. 4C). Because Rho GDI1 was immunoprecipitated by anti-Rho GDI1 in the competitive assays (Fig. 4C,  lanes 3 and 4), which was comparable with the control level (Fig. 4C, lane 1), the peptides specifically blocked the T1 binding to Rho GDI1.
In addition, we performed a pull-down assay of Rho GDI1 using anti-pan-TrkB antibody, and we detected Rho GDI1 using anti-Rho GDI1 antibody; this approach thus reversed the use of the antibodies to confirm the interaction between T1 and Rho GDI1. Then, we found that the Rho GDI1 band was reduced by 60% of the control level by BDNF treatment (Fig.  4 D), which was comparable with the results of the pull-down assay performed with anti-Rho GDI1. Therefore, these findings suggest that treatment with BDNF can lead to the dissociation of Rho GDI1 from T1. In the subsequent experiments, to obtain sufficient dissociation of Rho GDI1 from T1, we applied 20 ng/ml BDNF for 30 min to astrocytic cultures. At the concentration of 20 ng/ml BDNF, p75 cannot function (Dechant and Barde, 1997), even through p75 expression remained below the levels that could be detected by PCR (Fig. 3A).

BDNF effects on astrocytic morphology in 30 DIV cultures
Rho GTPases are involved in the regulation of cell morphology by remodeling the cytoskeleton, which contains microfilaments, intermediate filaments, and microtubules (Ridley, 2001;Etienne-Manneville and Hall, 2002). Rho GDI1 has been shown to selectively interact with the GDP-bound forms of the Rho GTPases and to inhibit their conversion from the GDP-bound inactive form to the GTPbound active form (Takai et al., 2001). Thus, we examined the morphological alteration of astrocytes by endogenous T1 in 30 DIV cultures. In serum-free medium containing N2 supplement, the form of 30 The cells were incubated in DMEM containing N2 supplement without BDNF stimulation and fixed in 4% formaldehyde. The immunoreactive cells for GFAP and T1 were stained by red and green, respectively. The cells coexpressing both GFAP and T1 were yellow. Note that almost all cells were coimmunopositive. Scale bar, 50 m. B, Time-lapse images of the 30 DIV astrocytic cultures by the phase-contrast microscope. BDNF (20 ng/ml) was added to the 30 DIV cultures at 0 min. The number in each photograph indicates the time lapse (in minutes) after BDNF treatment. Three cells were marked by the colored small letter, yellow "a," red "b," and black "c," and they were also outlined by the dotted lines of corresponding colors. Note that the more the time passed, the thinner the cell bodies and processes were. Scale bar, 20 m. C, Quantification of cell size. Values are given as the means Ϯ SD of the cell size analysis based on the results of four independent experiments (total cell counts: 0 min, n ϭ 944; 10 min, n ϭ 1003; 20 min, n ϭ 978; 30 min, n ϭ 981; 60 min, n ϭ 1105; 120 min, n ϭ 1054). The asterisks indicate significant differences ( p Ͻ 0.05; one-way ANOVA and Scheffé's post hoc test) compared with the value at 0 min. DIV astrocytes became fibrous (Fig. 5 A, B). Almost all astrocytes were immunoreactive for both GFAP and T1 (Fig. 5A). The timelapse analysis showed that BDNF stimulation led to a dynamic change in the shape of the astrocytes from fibrous to flat within 30 min (Fig. 5B). The cell surface area significantly increased 1.7fold at 10 min and reached a threefold plateau level at 30 min ( Fig.  5C). At 120 min after BDNF stimulation, the cells had decreased in size, albeit not significantly. At the same time, we measured the activities of RhoA, Rac1, and Cdc42, which are substrates of Rho GDI1, at 30 min when the change in astrocytic morphology reached a maximum. The active forms of all Rho GTPases were reduced by 60% of the control level (Fig. 6). The observed alterations in the shape of the astrocytes and the changes in Rho GTPase activity were both closely associated with the interaction between T1 and Rho GDI1 (Fig. 4), suggesting that BDNF-T1-Rho GDI1 signaling might control the Rho GTPases and consequently alter astrocytic morphology.
To elucidate the mechanism by which Rho GTPases regulate astrocytic morphology, we performed an inhibition assay of Rho GTPases using Toxin A and C3 toxin, which are known to inhibit all Rho GTPases (RhoA, Cdc42, and Rac1) and RhoA, respectively, in 30 DIV cultures. The astrocytic cultures were stimulated by BDNF at 20 ng/ml for 30 min. The results of the control cells were the same as the time-lapse analysis above (Fig. 7A). The astrocytes without BDNF treatment had long processes. Being treated by BDNF, the shapes of astrocytes were flat. In the Toxin A-treated cultures, the morphology of astrocytes was flat, regardless of BDNF treatment (Fig. 7A), which was similar to the shapes of the BDNF-treated control cells. The size of the cells was the same as that of BDNF-treated normal astrocytes (Fig. 7B). In contrast, C3 toxin treatment left the astrocytic morphology fibrous with fine filopodia-like processes (Fig. 7A). After BDNF treatment, the cells flattened; these findings were comparable with the observed morphology and size of the BDNF-treated con-trol cells and the Toxin A-treated cells (Fig. 7A). Therefore, endogenous T1 might alter astrocytic morphology (i.e., it renders astrocytes fibrous and flat) via the control of Rho GTPases and primarily through Cdc42 and Rac.

Involvement of T1 in the regulation of astrocytic morphology
We examined the molecular mechanism of the T1-induced morphological alteration of astrocytes by performing a transfection assay with T1 mutants. Because 30 DIV astrocytes were refractory to transfection (Ͻ5%), whereas short-term cultured astrocytes (2 DIV) were easily transfected at high transfection efficiencies (Ͼ90%), we used 2 DIV astrocytes, in which we observed no expression of BDNF receptors (Fig. 3A).
First, we examined the effect of BDNF on the morphology of astrocytes overexpressing normal T1 or a T1 deletion mutant. Both the untransfected control cells and empty vector (GFP)transfected cells showed a flat polygonal morphology with processes (Fig. 8). When these cells were treated with BDNF (20 ng/ml) for 30 min, we observed no changes in the cell morphology. On the other hand, the T1-transfected cells exhibited a fibrous, spindle morphology with long and narrow processes, even under the no-treatment condition. Interestingly, BDNF treatment rapidly and remarkably altered the cell morphology. Only 30 min after treatment with BDNF, the cells exhibited flat and wide cell bodies and stretched-out GFAP-positive fibers. The relative cell area of T1-transfected astrocytes treated with BDNF increased significantly, about fivefold, compared with that of T1transfected astrocytes without BDNF treatment (Fig. 8 B). In contrast, when T1-⌬11, a deletion mutant of a T1-specific sequence (Fig. 2 B), was transfected, the cell morphology observed was a flat polygon with processes similar to those of the control, and BDNF treatment was not found to induce any morphological changes such as those found in the T1-transfected cells. Recently, p75 has been reported to associate with Rho GDI1 and regulate Rho activity (Yamashita et al., 1999;Yamashita and Tohyama, 2003). However, high-concentration treatment with NGF (100 ng/ml), which is a p75 ligand, had no effect on cell morphology. Taken together with the evidence that there was no morphological change in nontransfected cells or in GFP-expressing cells, it was concluded that p75 had no effect on cell morphology.
Next, we investigated whether BDNF negatively regulates the Rho GTPases through Rho GDI1 released from T1. To this end, we performed a pull-down assay of the active forms of the Rho GTPases. In astrocytes expressing exogenous T1, BDNF treatment was found to reduce the amount of activated RhoA, Cdc42, and Rac1 by 55, 51, 55% of the control level, respectively (Fig. 9). In contrast, cells expressing T1-⌬11 and cells treated with NGF (100 ng/ml) were not associated with a decrease in the active forms of RhoA, Cdc42, and Rac1. These results are compatible with findings regarding the regulation of Rho GTPase activity by endogenous T1 (Fig. 6). Thus, the present results suggest that the specific C-terminal alignment of T1 is necessary for the control of Rho GTPases and for the observed morphological alteration of astrocytes.

Competitive assay with T1 intracellular peptides
We then investigated the effects of the T1-specific C-terminal peptide on the regulation of astrocytic morphology. To inhibit the T1 signaling cascade in a competitive manner, we cotransfected the expression vectors of T1 and each of the following: CFP, CFP-⌬11, and CFP-ICD. We expected that CFP-ICD, but not CFP or CFP-⌬11, would trap Rho GDI1 within the cytoplasmic region and inhibit the association of Rho GDI1 to the Rho

RhoA
Cdc42 Rac1 Active Total -+ -+ -+ BDNF 100 50 0 Relative activity (%) * * * Figure 6. Activities of Rho GTPases in 30 DIV astrocytic cultures. Top, The cells were stimulated for 30 min with BDNF (20 ng/ml), and the cell extracts were subjected to in vitro binding assays. The precipitates and total proteins (5 g for RhoA; 100 g for Cdc42 and Rac1) were loaded on 15% gel and were detected with specific antibodies. Bottom, Quantitative analysis of the bands shown in the above images. Each control level, without BDNF stimulation, was taken as 100%. The asterisks indicate statistically significant differences ( p Ͻ 0.05; paired Student's t test). Values are given as means Ϯ SD of four independent experiments.
GTPases, thereby resulting in the inhibition of the activity of BDNF. As shown in Figure 10, A and B, when both CFP and CFP-⌬11 were transfected with normal T1, we observed fibrous astrocytes under the condition lacking BDNF treatment. The addition of BDNF induced the morphological alteration of the astrocytes from fibrous to flat for 30 min. Namely, neither CFP nor CFP-⌬11 blocked the effects of BDNF, compared with the results obtained with the transfectant with T1-expression vector alone (Fig. 8). On the other hand, when CFP-ICD was overexpressed, the cells exhibited the same fibrous characteristics as were observed in the cases of the CFP-and CFP-⌬11transfected cultures. However, BDNF treatment was not found to induce morphological changes among the astrocytes that remained fibrous. Therefore, the T1-specific sequence was determined to be indispensable for the morphological alteration of these astrocytes.

Discussion
In the present study, we demonstrated that (1) a truncated TrkB receptor, T1, is capable of ligand-mediated signaling via Rho GDI1, which acts as a negative inhibitor in the Rho signaling cascade, and (2) the T1 signaling cascade regulates glial cellular morphology. A schematic representation is shown in Figure 10C. Our findings clearly indicate that T1 is not only the dominant-negative form of TKϩ, but is also the active receptor of BDNF itself.

Interaction between T1 and Rho GDI1
The in vitro binding analysis clearly showed the direct binding of Rho GDI1 and T1 (Fig. 2). One would question what fraction of T1 and Rho GDI1 contributes to the association between T1 and Rho GDI1. In the Western blot analysis in Figure 4, we loaded the 100 g total protein per lane for TrkB, which is approximately one-third amount of total protein in each lysate derived from a 3 cm dish. As shown in Figure 4 A-C, the level of the precipitated T1 in the control (at 0 min or no addition of BDNF) is comparable with the total level. In addition, in this study, the immunoprecipitations with anti-GDI or anti-pan-TrkB were performed with an efficiency of ϳ30%. The one-fourth of each precipitate was loaded on SDS-PAGE. Therefore, 44% of total T1 in an astrocyte bind to Rho GDI1. In Figure 4 D, ϳ 1 ⁄60 amount of total protein in each lysate (5 g total protein per lane) for Rho GDI1 was loaded on each lane. As described above about T1, we calculated the fraction of Rho GDI1 in the interaction between T1 and Rho GDI1. Consequently, ϳ2.2% of total Rho GDI1 in an astrocyte is involved in the binding to T1. It is a big surprise for us that the drastic change of astrocytic morphology is attributable to the low percentage of Rho GDI1 associating with T1. Rho GTPases are implicated in the important cell functions via remodeling cytoskeleton such as proliferation, migration, elongation of neurites, and membrane trafficking, suggesting that Rho GTPases are strictly regulated. Therefore, Rho GDI1 as a regulator of Rho GTPases are also severely controlled. Then, Rho GDI1 regulated by T1 may be small amount. A part of the rest of Rho GDI1 may be interacted with ERM consisting of ezrin, radixin, and moesin (Sasaki and Takai 1998) and might be bound to other unknown proteins.
Interaction between T1 and other proteins p75 has been reported to control the activity of RhoA in a Rho GDI1-dependent manner (Yamashita et al., 1999;Yamashita and Tohyama, 2003). In the present study, we demonstrated that T1 also binds directly to Rho GDI1 and that LFH residues in the T1-specific sequence is important for this type of binding. Be- Figure 7. Morphological changes in astrocytes treated with Toxin A and C3 in 30 DIV astrocytic cultures. A, The images of cells treatedbynothing(control),ToxinA,andC3.The30DIVastrocyteswereincubatedwithToxinA(20ng/ml)orC3(30g/ml)for24h.After being washed with serum-free medium, BDNF (20 ng/ml) was added to the cells, which were then fixed in 4% formaldehyde. The cells were stained with anti-T1 (green) and anti-GFAP (red). All displays show merged images (yellow). Scale bar, 50 m. B, Quantitative analysis of cell size. Values are given as means Ϯ SD of four independent experiments (total cell counts: control, without BDNF, n ϭ 914, with BDNF, n ϭ 991; Toxin A, without BDNF, n ϭ 948, with BDNF, n ϭ 1064; C3, without BDNF, n ϭ 1055, with BDNF, n ϭ 1072). The asterisks indicate statistically significant differences ( p Ͻ 0.05; paired Student's t test).
cause LFH is not contained in the intracellular domain of p75, both T1 and p75 may bind different regions of Rho GDI1.
Recently, Kryl and Barker (2000) reported that truncated TrkBinteracting protein (TTIP) is isolated from 15N neuroblastoma cells by using coimmunoprecipitation with GST fusion protein containing the intracellular juxtamembrane. TTIP has a molecular weight of 61 kDa, and T1 peptide competitively interrupted TTIP binding to T1, suggesting the direct binding interaction between them. However, the BDNF stimulation cannot modulate the interaction between T1 and TTIP. Kryl and Barker (2000) also analyzed TTIP by using matrix assisted laser desorption/ionization-mass spectrometry and described that TTIP is an unique protein. It is uncertain whether Rho GDI1 and TTIP bind directly to the different motifs in the T1-specific region or compete the same binding site. T1-mediated signaling may depend on its cellular compartment, because a fraction of T1 binds Rho GDI1. On the other hand, we detected proteins of 50, 60, and 72 kDa eluted from an affinity column (Fig. 1A); however, the correlation of each of these proteins with TTIP remains to be clarified.

Regulation of Rho proteins and astrocytic morphology by T1
The T1-interacting protein, Rho GDI1, is an inhibitory regulator of the Rho GTPases: Rho GDI1 is able to inhibit the activation of RhoA, Cdc42, and Rac1. On the other hand, the Rho GTPases are involved in the remodeling of the actin cytoskeleton: RhoA is involved in the formation of stress fibers; activated Cdc42 and Rac1 lead to lamellipodia and cell spreading, whereas activated Cdc42 induces filopodia (Hall, 1998). In this study, we demonstrated that Rho GDI1 released from T1 decreased the activities of the Rho GTPases, RhoA, Cdc42, and Rac1. However, it has remained unclear which Rho protein is related to the morphological changes in astrocytes triggered by BDNF. All Rho GTPases are known to be inhibited by Toxin A, and the form of astrocytes treated with Toxin A became flat. In contrast, the addition of C3, an inhibitor of RhoA, led to the formation of fibrous astrocytes with fine processes. Therefore, we were able to distinguish at least three types of morphology in this experiment. In the first type, the astrocytes became flat when all Rho GTPases were inhibited (Toxin A in Fig. 7A). In the second type, when only RhoA was inhibited by C3, the morphology of the cells was fibrous (C3 in Fig. 7A), which differed from the morphology of 30 DIV cells before the addition of BDNF (at 0 min in Fig. 5A and control in Fig. 7A), and the morphology of the fibrous cells also differed from that of T1-expressing 2 DIV cells (T1 in Fig. 8A). In the C3-treated condition, the processes of the astrocytes resembled filopodia; that is, fine processes extended from the bold processes of the astrocytes and the cell bodies (C3 in Fig. 7A). In the third type, the cells were not as fibrous as the second type of cell when the activities of the Rho GTPases remained at their basal levels. Typically, 30 DIV and T1-overexpressing astrocytes had spindle-shaped bodies or small, flat cell bodies and long processes (Figs. 5A, 7A, 8A). Thus, it appears that BDNF-T1 signaling suppressed the activity of all three Rho proteins and then induced morphological change leading to the flat type 1 cells. In addition, the cell flattening appears to be mediated primarily by the suppression of Cdc42 and Rac. In this context, it should be emphasized that extension of the astrocyte cell bodies was observed as a result of the inhibition of the Rho GTPases by T1-Rho GDI1 signaling. Recent studies have shown that the Rho GTPases control the remodeling of microfilaments, intermediate filaments, and microtubules (Ridley, 2001;Etienne-Manneville and Hall, 2002). Therefore, the regulation of cell morphology is not solely dependent on the microfilaments but instead depends on the well orchestrated control of various cytoskeletal proteins. More precise information regarding the mechanism of their regulation by BDNF remains to be obtained by additional study.

Functional role of T1 in astrocytes
In the present study, we demonstrated that astrocytes are able to alter their morphology rapidly and dramatically via the T1 Ͼ Rho GDI1 Ͼ Rho GTPase signaling cascade in a BDNF-dependent manner. In the mature mammalian CNS, BDNF is synthesized and secreted from presynaptic and/or postsynaptic sites, depend-  Fig. 2 B); T1 ϩ NGF, normal T1-expressing cells that were treated with 100 ng/ml NGF instead of with BDNF. The cells were stimulated with vehicle (Ϫ) or with 20 ng/ml BDNF (ϩ) and then were stained with anti-GFAP. Almost all of the cells were double positive for GFP (green) and GFAP (red), except for the No Vec cells. All displays show merged images (yellow-green). Scale bar, 30 m. B, Quantification of cell size in A. Values are given as the means Ϯ SD from the results of four independent experiments (total cell counts: No Vec, without BDNF, n ϭ 854, with BDNF, n ϭ 821; GFP, without BDNF, n ϭ 879, with BDNF, n ϭ 787; T1, without BDNF, n ϭ 698, with BDNF, n ϭ 731; T1-⌬11, without BDNF, n ϭ 687, with BDNF, n ϭ 634; T1 ϩ NGF, without BDNF, n ϭ 810, with BDNF, n ϭ 834). The asterisks indicate significant differences ( p Ͻ 0.05; one-way ANOVA and Scheffé's post hoc test) from the values obtained without BDNF treatment.