Abbreviations used : CHO, Chinese hamster ovary ; GABARAP, GABAA receptor-associated protein ; GST, glutathione S-transferase ; LC3, light chain 3 ; MAP, microtubule-associated protein ; PAGE, polyacrylamide gel electrophoresis ; PC-tubulin, tubulin purified from MAP-rich tubulin by phosphocellulose chromatography and free of MAPs ; PEM buffer, 80 mM piperazine-N,N'-bis(2-ethanesulfonic acid) sequisodium salt (pH 6.8), 1 mM MgCl2, and 1 mM EGTA ; SDS, sodium dodecyl sulfate ; TBS-T, 20 mM Tris, 137 mM NaCl, and 0.1% Tween 20 (pH 7.6).
Address correspondence and reprint requests to Dr. R. W. Olsen at Department of Molecular and Medical Pharmacology, CHS 23-120, UCLA School of Medicine, Los Angeles, CA 90095-1735, U.S.A. E-mail : firstname.lastname@example.org
GABAA receptor-associated protein (GABARAP) was isolated on the basis of its interaction with the γ2 subunit of GABAA receptors. It has sequence similarity to light chain 3 (LC3) of microtubule-associated proteins 1A and 1B. This suggests that GABARAP may link GABAA receptors to the cytoskeleton. GABARAP associates with tubulin in vitro. However, little is known about the mechanism for the interaction, and it is not clear whether the interaction occurs in vivo. Here, we report that GABARAP interacts directly with both tubulin and microtubules in a salt-sensitive manner, indicating the association is mediated by ionic interactions. GABARAP coimmunoprecipitates with tubulin and associates with both microtubules and microfilaments in intact cells. The cellular distribution is altered by treatment with taxol, nocodazole, and cytochalasin D. The tubulin binding domain was located at the N terminus of GABARAP by using synthetic peptides and deletion constructs and is marked by a specific arrangement of basic amino acids. The interaction between GABARAP and actin might be mediated by other proteins. These results demonstrate the GABARAP interacts with the cytoskeleton both in vitro and in cells and suggest a role of GABARAP in the interaction between GABAA receptors and the cytoskeleton. Such interactions are presumably needed for receptor trafficking, anchoring, and/or synaptic clustering. The structural arrangement of the basic amino acids present in the tubulin binding domain of GABARAP may aid in recognition of the potential of tubulin binding activity in other known proteins.
GABAA receptors mediate the majority of rapid inhibitory transmission in the CNS. Sequence analysis indicates that they belong to a receptor superfamily, including glycine receptors and nicotinic acetylcholine receptors (Macdonald and Olsen, 1994). Studies on glycine receptors and nicotinic acetylcholine receptors revealed several molecules functioning in receptor clustering. Gephyrin, a 93-kDa microtubule-associated protein (MAP) (Kirsch et al., 1991 ; Prior et al., 1992), causes redistribution of glycine receptors in heterologous systems (Meyer et al., 1995). Loss of glycine receptor clustering was observed when gephyrin expression was knocked out by antisense oligonucleotides (Kirsch et al., 1993) or targeted gene disruption (Feng et al., 1998). Rapsyn, a 43-kDa actin-binding protein (Hill, 1992), promotes nicotinic acetylcholine receptor clustering when coexpressed in fibroblast cells (Phillips et al., 1991 ; Ramarao and Cohen, 1998).
Immunocytochemistry studies on GABAA receptors demonstrated that receptor clustering required a specific subunit composition (Craig et al., 1994 ; Nusser et al., 1995 ; Koulen et al., 1996). Both gephyrin and rapsyn have been proposed to cluster GABAA receptors (Yang et al., 1997 ; Essrich et al., 1998 ; Kneussel et al., 1999). However, attempts to demonstrate direct interaction between GABAA receptors and gephyrin were not successful (Meyer et al., 1995), and the low expression level of rapsyn in brain suggests that rapsyn may not be a major element for GABAA receptor clustering. The interaction between GABAA receptors and tubulin has been reported (Item and Sieghart, 1994 ; Kannenberg et al., 1997). This suggests the possible involvement of cytoskeleton in GABAA receptor clustering, but how the cytoskeleton system mediates receptor clustering is not clear. Recently, MAP 1B and GABAA receptor-associated protein (GABARAP) were shown to interact specifically with the ρ1 and γ2 subunit of GABA receptors, respectively, suggesting possible mechanisms for GABAA receptor anchoring and clustering (Hanley et al., 1999 ; Wang et al., 1999). GABARAP was found in some neurons to be colocalized with GABA receptor synaptic clusters (Wang et al., 1999), and GABARAP promotes clustering of GABAA receptors when coexpressed in recombinant systems (Chen et al., 1999). Preliminary observations show GABARAP association with plasma membranes in several cell expression systems (Kneussel et al., 2000) but also inside some neurons, associated with membraneous organelles (A. Triller and R. W. Olsen, unpublished data).
In spite of the sequence similarity between GABARAP and light chain 3 (LC3) of MAP 1 (Mann and Hammarback, 1994), how GABARAP interacts with the cytoskeleton is not clear. GABARAP mRNA and immunoreactivity are observed in all tissues (Wang et al., 1999), so it clearly has functions in addition to those related to GABAA receptors, presumably involving microtubules. In this study, we demonstrated direct interaction between GABARAP and microtubules both in vitro and in cells and showed that the manner of interaction is similar to that of other MAPs. GABAA receptors, GABARAP, and tubulin appear to be present in the same complex. It is important that we identified a novel tubulin binding motif in GABARAP. GABARAP may also associate with microfilaments, possibly through other proteins. Our data suggest a mechanism for the involvement of cytoskeleton in GABAA receptor trafficking, anchoring, and/or clustering and GABARAP as a linker molecule.
MATERIALS AND METHODS
Molecular cloning of GABARAP cDNA from rat and mouse
Rat and mouse cDNAs of GABARAP were cloned by RTPCR. Poly(A)+ RNA was extracted from mouse or rat brain by using the FastTrack mRNA isolation kit (Invitrogen). Firststrand cDNA was synthesized using SUPERSCRIPT II (GibcoBRL). Oligonucleotide complementary to nucleotides 552-575 of human GABARAP cDNA (Wang et al., 1999) was used to prime the first-strand synthesis. The RT-PCR products were resolved by electrophoresis, and DNA of the predicted size was excised, purified, and cloned to pCR II (Invitrogen). Multiple clones were sequenced. Human, rat, and mouse GABARAP cDNA sequences were compared and analyzed.
Protein expression and purification
Full-length GABARAP and GABARAP36-117 were cloned in pGEX-2T (Pharmacia) and expressed as fusion proteins with the glutathione S-transferase (GST) moiety on the N terminus. Escherichia coli strain DH5α or BL21 containing the expression plasmid was grown to OD560 of 0.4-0.6, induced by 0.1 mM isopropyl β-d-thiogalactopyranoside for 1 h, homogenized in phosphate-buffered saline (137 mM NaCl, 2.7 mM KCl, 10 mM Na2HPO4, and 1.7 mM KH2PO4, pH 7.4) with 1% Triton X-100, 10% glycerol, and protease inhibitors, and loaded on glutathione-agarose (Sigma). The GST function proteins were eluted by 5 mM glutathione/50 mM Tris, pH 8.0. To remove the GST leader, GST fusion proteins were cleaved by thrombin (Boehringer ; 10 : 1, wt/wt) at room temperature for 8 h in 50 mM Tris (pH 8.8), 10 mM NaCl, and 2.5 mM CaCl2. Undigested GST fusion protein and GST molecules were removed by absorbing the digestion reaction with glutathione-agarose. Protein concentration was determined by the BCA protein assay kit (Pierce). The purified proteins were analyzed by sodium dodecyl sulfate (SDS)-polyacrylamide gel electrophoresis (PAGE).
Generation and purification of antibodies
Purified recombinant GST-GABARAP or GABARAP was used as antigen to immunize white New Zealand rabbits. Antibody activity was checked by western blot after the second injection. Rabbits were killed 10 days after the fifth injection. Anti-GABARAP antibodies were further purified by antigen affinity column chromatography. The affinity column was made by covalently attaching GST-GABARAP to Affi-Gel 10 (Bio-Rad) following the manufacturer's instructions. The crude serum was loaded on the column, washed, and eluted by a combination of acidic and basic elution.
Western blot analysis
Proteins were separated by SDS-PAGE and transferred to polyvinylidene difluoride membranes (Bio-Rad) by a semidry method. The membrane was then incubated in 20 mM Tris, 137 mM NaCl, and 0.1% Tween 20 (pH 7.6 ; TBS-T) with 3% nonfat milk and primary antibody at room temperature for 2 h. Five washes by TBS-T were performed before incubation with horseradish peroxidase-conjugated secondary antibody in TBS-T with 3% milk. After a 45-min incubation with the secondary antibody, the membrane was washed five times with TBS-T and visualized by enhanced chemiluminescence (ECL ; Amersham).
In vitro binding assay, immunoprecipitation, and cosedimentation assay
To test the binding to soluble tubulin in vitro, the purified GST fusion proteins were charged to glutathione-agarose or Affi-Gel 10 (Bio-Rad). Tubulin, purified from MAP-rich tubulin by phosphocellulose chromatography and free of MAPs (PC-tubulin ; Cytoskeleton, Denver, CO, U.S.A.), was incubated with the immobilized GST fusion protein in buffer A [20 mM Tris (pH 7.5), 5 mM MgCl2, 2 mM CaCl2, 1 mM dithiothreitol, 1% Triton X-100, and 1% nonfat milk], washed extensively, and eluted by SDS-PAGE loading buffer. In vitro binding assays with purified actin (Cytoskeleton) were done in buffer containing 20 mM Tris (pH 7.5), 0.2 mM ATP, 0.2 mM CaCl2, 0.2 mM dithiothreitol, 1% Triton X-100, and 1% nonfat milk. The bound proteins were resolved on SDS-PAGE and visualized by immunostaining with monoclonal anti-β-tubulin (1:1,000 ; Sigma) or anti-actin (1:500 ; Sigma). Similarly, brain homogenates were used to test the binding of endogenous tubulin, actin, and GABAA receptors to GABARAP. Rat brains were homogenized with 10Xvolumes of buffer containing 100 mM NaC1, 10 mM Tris (pH 7.5), 5 mM EDTA, 10 mM MgCl2, 0.5% NP-40, and 1% Triton X-100. Cell debris and unsolubilized membranes were removed by centrifugation.
For immunoprecipitation experiments, brain homogenates were incubated with preimmune serum or purified anti-GABARAP antibody charged to Affi-Gel 10. The bound proteins were detected by western blot assay using monoclonal anti-β-tubulin.
To perform the coassembly assay, tubulin was first incubated with 1 mM GTP and 20 μM Taxol in PEM buffer [80 mM piperazine-N,N'-bis(2-ethanesulfonic acid) sequisodium salt (pH 6.8), 1 mM MgCl2, and 1mM EGTA] at 37°C for 20 min to form microtubules. Then, 0.1 μM recombinant GABARAP was incubated with 1 μM polymerized microtubules in 100 μ1 of PEM buffer supplemented with 1 mM GTP and 20 μM Taxol at 37°C for 20 min. The reaction mixture was centrifuged at 100,000 g at 37°C for 20 min in a Beckman model TL-100 tabletop ultracentrifuge. The supernatant and pellet were analyzed by SDS-PAGE. The presence of tubulin and GABARAP was assayed by western blot. In some cases, 400 mM NaCl was present in PEM buffer, to test if the association of GABARAP with tubulin could be disrupted by high salt. Before each assay, all proteins were precleared by centrifugation at 100,000 g for 10 min to remove the aggregates.
PC-tubulin (0.8 mg/ml) and GABARAP (0.1 mg/ml) were mixed in 20 μ1 of PEM buffer and incubated at 37°C for 30 min after addition of 1 mM GTP and 20 μM Taxol. The reaction was then fixed by adding 180 μ1 of 1 mM GTP, 20 μM Taxol, 30% glycerol, and 0.5% glutaraldehyde in PEM buffer. After 10 min, 20 μ1 of 2 M NH4Cl was applied to terminate the fixation. The coassembly reaction was deposited on Formvarcoated copper grids and washed three times by PEM buffer supplemented with 1 mM GTP, 20 μM Taxol, and 10% glycerol. The grid was blocked by the same buffer with 1% normal goat serum and incubated with anti-GABARAP (1:500). The grid was washed three times, incubated with goat anti-rabbit colloidal gold-conjugated IgG (15 nm ; 1.75 ; Amersham), washed three times, and negatively stained with 1% uranyl acetate. The sample was then examined with a JEOL 100CX electron microscope.
Synthetic peptide GABARAP1-22, corresponding to the N-terminal 22 amino acids, and the scrambled peptide GABARAP1-22S (REKIEEHMKFPFESEGVYKKRR) were used in the assay. PC-tubulin at a concentration of 0.8 mg/ml was mixed with 300 μM peptide in PEM buffer at 30°C. The reaction was initiated by addition of 1 mM GTP, and the absorbance at 340 nm was measured in a 30-s interval (Gaskin et al., 1974).
Cell culture and immunocytochemistry
Chinese hamster ovary (CHO) cells were grown in Ham's F-12 medium supplemented by 10% fetal bovine serum at 37°C with 5% CO2. Cytoskeleton drugs were applied to study possible changes in the localization of GABARAP. Specifically, overnight application of Taxol at 5 μg/ml was used to reorganize the microtubule skeleton. Inclusion of nocodazole (0.5 μg/ml, overnight treatment) or cytochalasin D (1 μg/ml, 1.5-h treatment) in the medium was used to disrupt microtubules or microfilaments, respectively. In the process of immunostaining, the cells were first washed with PEM buffer. Then the cells were either fixed and permeabilized by a 10-min treatment with dry methanol at -20°C or permeabilized with 20% glycerol and 0.5% Triton X-100 in PEM buffer (30 s), followed by a PEM buffer wash, and fixed by 3% formaldehyde and 0.5% dimethyl sulfoxide in PEM buffer (30 min). After fixation, the cells were washed three times with phosphate-buffered saline and incubated with anti-GABARAP (1:350), anti-β-tubulin (1:500), or Texas Red-conjugated phalloidin (1:1,000 ; Molecular Probes) in phosphate-buffered saline with 3% bovine serum albumin at 37°C for 1 h. Appropriate secondary antibodies (fluorescein-conjugated goat anti-rabbit IgG at 1:100 dilution for GABARAP, Texas Red-conjugated horse anti-mouse IgG at 1:100 dilution for tubulin) were incubated in the same buffer at room temperature for 45 min, following three washes with phosphate-buffered saline. The cells were finally washed three times with phosphate-buffered saline, mounted in the presence of Vector Shield (Vector Laboratories), and analyzed by fluorescent microscopy.
GABARAP is highly conserved among human, rat, and mouse
The original GABARAP cDNA was isolated from a human brain library (GenBank accession no. AF161586). Here we report the cloning of rat and mouse cDNAs of GABARAP by RT-PCR. They were deposited into GenBank with accession nos. AF161588 and AF161587. Sequence analysis of human, rat, and mouse GABARAP cDNA demonstrated that there is 93% identity between the human and mouse sequence. Human and rat GABARAP cDNA also share 93% identity, whereas 98% identity exists between mouse and rat cDNAs (data not shown). At the amino acid level, all three code for polypeptides with exactly the same sequence. Northern blots of mouse and rat poly(A)+ RNA revealed a single ~900-bp band (data not shown), indicating that GABARAP mRNAs of the mouse and rat have structures similar to that of the human. This also supports the conclusion that the mouse and rat clones obtained by RT-PCR contain a full open reading frame. Furthermore, western blot analysis with rat and mouse brain homogenates detected an antigen comigrating with the recombinant GABARAP (data not shown). By searching the database, we found some other gene products with sequence similarity to GABARAP. In addition to LC3 of MAP 1A and 1B in the rat (accession no. Q62625) and the cow (accession no. 041515), a Saccharmoyces cerevisiae protein, Aut7p (P38182, 55% identity), and several other proteins were also found to have significant sequence similarity. A protein of identical size (117 amino acids) has 70% identity (accession no. AF020262) and was described in the cow as a “general Golgi transport factor p16” (Legesse-Miller et al., 1998), and an identical sequence in the human was named “ganglioside expression factor II.” Other homologues include two human clones (accession nos. AF044671 and AF067171 ; 100% identity), a rat protein (accession no. 008765, 57% identity), a Caenorhabditis elegans protein (accession no. Q09490, 83% identity), and an Arabidopsis thaliana protein (accession no. AC006220, 57% identity). However, no function was specified for them. The yeast Aut7p protein is essential for delivery of autophagic vesicles to the vacuole and associates with microtubules (Lang et al., 1998). The GABARAP gene was localized to human chromosome 17 (accession no. AC003688). Comparison of the genomic to cDNA sequence indicates that GABARAP cDNA was spliced from at least four exons. The intron-exon boundaries match the consensus sequence (data not shown) (Padgett et al., 1986).
Expression of GABARAP and antibody generation
To carry out in vitro studies and generate antibodies against GABARAP, recombinant GABARAP was expressed in E. coli as a GST fusion protein. GST-GABARAP could be purified to >90% homogeneity by a single-step elution (Fig. 1A). Because there is a thrombin cleavage site between GABARAP and GST, nonfusion GABARAP can be obtained by digestion of GST-GABARAP with thrombin. After absorbing GST-GABARAP and GST to gluathione-agarose, GABARAP could be purified (data not shown).
Three rabbits were immunized. Rabbits 2626 and 2627 were injected with SDS-denatured GST-GABARAP. Rabbit 3411 was injected with nonfusion GABARAP not treated with any denaturing reagents other than adjuvants. The crude sera were purified by antigen affinity chromatography. Antibodies from all three rabbits demonstrated specific interaction with GABARAP in western blot assays. Antibodies from rabbit 2626 or 2627 recognized GST-GABARAP, GST, and GABARAP but not the control protein (interacellular loop of GABAA receptor α1 subunit). Antibodies from rabbit 3411 recognized GST-GABARAP and GABARAP, but not GST and the intracellular loop of α1 subunit (Fig. 1B). All anti-GABARAP antibodies recognized a polypeptide in brain tissue homogenate that comigrated with the recombinant GABARAP (Fig. 1B). Serum absorbed by antigen affinity column and preimmune serum did not give any specific staining for GABARAP (data not shown).
Both the soluble form and polymerized form of tubulin interact directly with GABARAP
To obtain biochemical evidence for the interaction between GABARAP and tubulin, we performed in vitro binding assay and coassembly assays. When brain homogenates were incubated with GST-GABARAP, we found that endogenous tubulin molecules bound GST-GABARAP but not GST (Fig. 2A). We further tested the association with purified tubulin. We include CaCl2 in the binding buffer to keep tubulin in soluble form, presumably tubulin dimers formed by α- and β-tubulin. The purified tubulin in soluble form specifically interacted with GST-GABARAP but not GST (Fig. 2B). From the western blot, we estimated that ~1% or 0.3 μg of tubulin in the brain homogenates stuck to 20 μg of GST-GABARAP, whereas ~6% or 3 μg of the purified tubulin associated with 20 μg of GST-GABARAP. This might be due to the lower concentration of tubulin in the brain homogenates or the presence of other MAPs that could compete for the binding sites on tubulin with GABARAP. We also tested the binding of GABARAP to microtubules in coassembly assays. The microtubules were assembled from soluble tubulin in the presence of GTP and Taxol. GST-GABARAP cosedimented with in vitro assembled microtubules. They were present in the pellet fraction along with microtubules (Fig. 2C, lanes 1 and 2). We also noticed that the interaction was disrupted when high salt (400 mM NaCl) was included in the coassembly buffer (Fig. 2C, lanes 5 and 6). When the pellet fraction was resuspended with high salt buffer, GST-GABARAP no longer cosedimented with microtubules (Fig. 2C, lanes 3 and 4). The assembly of tubulin was not affected by high salt. The salt-sensitive association of GABARAP with microtubules indicates that the binding is mediated by ionic interactions, in a manner similar to other MAPs (Mann and Hammarback, 1994). The tubulin molecules we have used in the in vitro binding assays (Fig. 2B) and the coassembly assays (Fig. 2C) are “PC-tubulin,” which is purified from MAP-rich tubulin (tubulin purified by cycles of heat-induced polymerization and cold-induced depolymerization that contains MAPs in addition to tubulin) by phosphocellulose chromatography. This tubulin is devoid of MAPs. Our results demonstrate that the binding of GABARAP to soluble tubulin and microtubules is direct, and it is not mediated by other MAPs. Nonfusion GABARAP gave results similar to those of GST-GABARAP in coassembly assay (data not shown). Furthermore, endogenous tubulin in brain homogenates coimmunoprecipitated with GABARAP (Fig. 2D), indicating their probable association in vivo.
GABAA receptors and tubulin bind GABARAP at the same time
When brain homogenates were incubated with the GST-GABARAP affinity column, we found that both GABAA receptors and tubulin associated with the column (Fig. 3). This indicates that they interact with GABARAP at the same time and that they may be components in the same complex. Judging from the western blot, we estimated that ~10-15% of the GABAA receptors in the brain homogenates were bound to the GST-GABARAP affinity column.
In vitro assembled microtubules are decorated by GABARAP
To visualize the association of GABARAP with in vitro assembled microtubules, we used immunoelectron microscopy. The microtubules assembled from PC-tubulin were observed in all cases (Fig. 4). When GABARAP molecules were coassembled with tubulin, they were close to or directly bound to microtubules (Fig. 4A, indicated by arrows). A small proportion of immunostained GABARAP was also observed in an area of negative staining that did not contain microtubules but presumably oligomeric tubulin structures (Fig. 4, open arrowhead). The immunogold-stained GABARAP and its decoration of microtubules were specific. Almost no gold particles were observed when tubulin was assembled alone (Fig. 4B) or when primary antibody was eliminated (Fig. 4C).
The putative tubulin binding domain peptide promotes tubulin assembly
Tubulin bound to GST-GABARAP, but we did not observe any interaction between tubulin and GST-CABARAP36-117 in the in vitro binding assay. This indicated that the N-terminal 36 amino acids might possess the tubulin binding domain. Although tubulin binding motifs found in other MAPs were not detected within GABARAP, we noticed that the first 22 amino acids have a potential to fold into an α-helix by secondary structure prediction (Chou and Fasman, 1978 ; Garnier et al., 1978). It is interesting that five basic amino acids align on one side of the helix with four of them in consecutive turns (Fig. 5A). This structural arrangement may feature a tubulin binding motif. To test whether the N-terminal portion of GABARAP confers microtubule binding activity, we did further coassembly assays using GST-GABARAP36-117 lacking the putative tubulin binding motif. No detectable GST-GABARAP36-117 was found in the pellet (Fig. 5B, lane 4). The distribution of GST-GABARAP36-117 in the supernatant and the pellet was the same with or without microtubules (Fig. 5B). We further narrowed down the tubulin binding domain to the first 22 amino acids by using a synthetic peptide in the turbidimetric assay, which is a general method to identify tubulin binding domains (Gaskin et al., 1974). The synthetic peptide corresponding to the first N-terminal 22 amino acids promoted tubulin assembly (Fig. 5C), whereas the scarmbled peptide with the same amino acid composition as the test peptide had no significant effect (Fig. 5C), comparable to tubulin assembled alone (data not shown). Combining the results from the coassembly assay using GST-GABARAP36-117 and the turbidimetric assay using synthetic peptides, we conclude that the first 22 amino acids of GABARAP may represent a novel tubulin binding motif.
GABARAP colocalizes with both microtubules and microfilaments
CHO cells were chosen for studying colocalization of GABARAP with cytoskeleton structures. By western blot analysis of CHO cell homogenates, we detected an antigen that was recognized by GABARAP antibody and comigrated with recombinant GABARAP (data not shown). Because GABARAP is widely expressed in many tissues in addition to brain and it is highly conserved among the human, rat, and mouse, it is not surprising to find GABARAP in CHO cells. We stained CHO cells with GABARAP antibodies and found two different patterns of subcellular localization (Fig. 6). One was punctate and seemingly formed from a protein complex or aggregate. This punctate staining was not evenly distributed in the cell. It was concentrated around perinuclear regions and extended to peripheral regions of the cell. The extension was always observed with well-separated large cells and seemed to associated with certain kinds of skeletal structures. Another was fiber-like staining, possibly associated with microfilaments (Figs. 6A, arrows, and 7). When the cells were permeabilized and fixed at the same time by dry methanol, both forms of subcellular localization were observed. When cells were first permeabilized by PEM buffer (with 20% glycerol and 0.5% Triton X-100) and then fixed by formaldehyde, the stress fiber-like staining was still observed, whereas staining of the punctate pattern was dramatically decreased and became diffuse (Fig. 7). A likely interpretation is that the punctate form of GABARAP interacts dynamically with the cytoskeleton and that the stress fiber-like form of GABARAP is tightly associated with microfilaments. These two staining patterns were specific for GABARAP because the preimmune serum or anti-GABARAP serum absorbed by the GABARAP affinity column did not give any significant signal (data not shown).
As GABARAP interacted with both forms of tubulin in the test tube, we attempted to test their colocalization in intact cells by double-staining the cells with anti-GABARAP and anti-β-tubulin. The punctate staining pattern of GABARAP appeared to colocalize with subdomains of the microtubule network (Fig. 6A-C). The stress fiber-like staining did not overlap with any microtubules (Fig. 6A and B). We further confirmed the colocalization by treating the cells with different cytoskeleton drugs. When the cells were treated with Taxol, both the microtubule network and the distribution of GABARAP (punctate form) were reorganized, and they overlapped in certain regions (Fig. 6D-F). When the cells were incubated with the microtubule-disrupting agent nocodazole, the microtubules collapsed (Fig. 6H). Both the microtubules and GABARAP (punctate form) lost their extended forms of expression and redistributed to the same regions (Fig. 6G-I). The altered distributions of GABARAP (punctate form) after Taxol and nocodazole treatment was never observed in nontreated cells, indicating the effects of the microtubule drugs on GABARAP localization. The distribution of punctate GABARAP staining and microtubules was not affected by cytochalasin D, which is an actin filament-disrupting chemical (Fig. 6J-L). We also noticed that the stress fiber-like staining of GABARAP was not altered by Taxol or nocodazole treatment but was affected by cytochalasin D. These fibers collapsed and redistributed (Fig. 6J, arrowheads).
In addition to the punctate staining that colocalized with subdomains of the microtubule network, GABARAP seems to have another pattern of subcellular localization. It has a stress fiber-like staining and can be altered by actin drugs. To test its colocalization with microfilaments in cells, we double-stained CHO cells with anti-GABARAP and Texas Red-conjugated phalloidin. We chose to use staining conditions that eliminate most of the punctate staining, so the stress fiber-like staining of GABARAP could be better isolated and visualized. Compared with actin, GABARAP produced a lower level of immunostaining but colocalized with actin fibers (Fig. 7A-C). When cells were treated with nocodazole, the shape of cells changed, owing to the loss of the microtubule skeleton. However, the staining pattern of both actin and GABARAP persisted and colocalized (Fig. 7D-F). When cells were incubated with cytochalasin D, actin filaments were disrupted, redistributed, and collapsed (Fig. 7H). Accordingly, the staining pattern of GABARAP underwent a similar change, and the altered staining colocalized with the collapsed actin domains (Fig. 7G-I).
The interaction between GABARAP and actin is mediated by other cellular components
Because we detected colocalization of GABARAP and microfilaments in CHO cells, we further sought biochemical evidence for their interaction. We found that purified actin did not bind GABARAP ; however, endogenous actin in the brain homogenates interacted with GABARAP (Fig. 8). This indicates that their interaction is not direct and is probably mediated by other proteins present in the brain homogenates.
The involvement of cytoskeleton in anchoring and clustering of neurotransmitter receptors has been described. The cytoskeleton system could participate through direct interactions (van Rossum et al., 1999) with the receptor or by certain linker molecules such as the PDZ domain-containing proteins gephyrin and rapsyn (Hill, 1992 ; Prior et al., 1992 ; Sheng and Wyszynski, 1997). Interaction between tubulin and GABAA receptors has been reported (Item and Sieghart, 1994), and, furthermore, microtubule-disrupting drugs had effects on both GABAA receptor pharmacology and distribution (Whatley and Harris, 1996). Binding of MAP 1B and GABARAP to the ρ1 subunit and γ2 subunit of GABA receptors suggests a role of the cytoskeleton in receptor anchoring and clustering. The interaction between MAP 1B and microtubules has been well characterized (Noble et al., 1989). However, how GABARAP associates with the cytoskeleton is not clear. Studies on their interaction might facilitate the understanding of the functional role of the cytoskeletal network in clustering receptors at synapses. Here, we demonstrate that GABARAP interacts with microtubules both in vitro and in vivo. A novel tubulin binding motif is identified. We also find association of GABARAP with microfilaments in brain extracts, probably through the interaction with other cellular components.
Sequence analysis of GABARAP indicates its similarity to several proteins with known function, including LC3 of MAP 1A and 1B and Aut7p in yeast. LC3 can bind directly to microtubules and is a component of the MAP 1 complex. However, Aut7p is involved in vacuole trafficking but does not bind microtubules by itself (Lang et al., 1998). A bovine homologue is involved in microtubule-dependent Golgi protein transport (Legesse-Miller et al., 1998). This relationship of GABARAP to other trafficking proteins is consistent with its widespread expression and likely roles for GABARAP in trafficking different organelles or vesicular membrane proteins, among which are GABAA receptors. Among the receptor-clustering molecules, gephyrin and rapsyn bind tubulin and actin, respectively, whereas PSD-95 does not bind microtubules and seems not essential for NMDA receptor clustering (Migaud et al., 1998 ; Passafaro et al., 1999). CRIPT, a protein that interacts with both PSD-95 and microtubules, may mediate the connection between NMDA receptors and the cytoskeleton (Passafaro et al., 1999). By using in vitro binding assays, coassembly assays, and immunoelectron microscopy, we demonstrated the direct interaction between GABARAP and microtubules.
MAPs usually are rich in basic amino acids and possess a basic isoelectric point. They interact with the C-terminal acidic region of tubulin through ionic interactions (Maccioni and Cambiazo, 1995). Repeats of 18 amino acids had been identified as a tubulin binding motif for MAP-2 and τ (Lewis et al., 1988) and four basic amino acid repeats (KKEE or KKEI/V) for MAP 1B (Noble et al., 1989). These microtubule binding domains are positively charged. Except for MAP 1B, the mechanism for the binding of microtubules to the receptor-clustering proteins is not clear. Like other MAPs, GABARAP is a positively charged molecule and has an asymmetric structure. The basic amino acids are concentrated in the N terminus, and the C-terminal portion is acidic. GABARAP cosedimented with microtubules in a salt-sensitive manner, resembling other MAPs.
The KKE tubulin binding motif in MAP 1B and τ is not obviously present in GABARAP. We have used a truncated form of GABARAP and turbidimetric assay to narrow down the tubulin binding domain to the N-terminal 22 amino acids. This region is rich in basic amino acids. It is predicted to fold into an α-helix by the GOR method (Garnier et al., 1978), as well as the CF method (Chou and Fasman, 1978). The basic amino acids align on the same side of the helix. These basic amino acids may participate in the ionic interaction with the helix formation from the acidic C terminus of tubulin (Maccioni and Cambiazo, 1995). As one of the important findings of this report, the tubulin binding motif we have identified in GABARAP is featured by a general structural organization of basic amino acids other than the specific amino acid sequences described for other MAPs. This organization was indicative in the tubulin binding domain of τ and MAP 1B. In MAP 1A, the SS1 domain contains three KKE motifs that are present in the tubulin binding domain of MAP 1B. However, the SS1 region could not bind microtubules. It is interesting that the SS2 region of MAP 1A is predicted to fold into α-helices with basic amino acids and acidic amino acids aligning on different sides of a helical wheel (Cravchik et al., 1994). Indeed, the SS2 region mediates the binding to microtubules. We also noticed such an amino acid arrangement in the tubulin binding domain of E-MAP-115, which has a predicted α-helical structure as well (Masson and Kreis, 1993). This novel tubulin binding motif in GABARAP marked by the specific structural organization may be useful in the recognition of tubulin binding potential of other known proteins.
Another important finding is that GABARAP associates with microtubules in intact cells, suggesting GABARAP links GABAA receptors and cytoskeleton in vivo. We found that the expression pattern of GABARAP in CHO cells is different from that in cultured neurons (Wang et al., 1999). Because GABARAP is highly conserved and expressed widely in different tissues, it should have other biological functions. These are likely to include microtubule-membrane interactions as mentioned above. Its subcellular localization and function may depend on the particular cellular environment. Although GABARAP staining is mostly intracellular in CHO cells, it is mostly membrane-bound when expressed in PC12 cells (Kneussel et al., 2000). Most receptor-clustering molecules do not show filamentous staining in neurons, and their interactions with the cytoskeleton are usually studied in other cell lines. For example, CRIPT and PSD-95 showed punctate staining in cultured neurons. In contrast, they have filamentous staining, associated with microtubules, in COS cells (Niethammer et al., 1998 ; Passafaro et al., 1999). In this study, we have chosen CHO cells to visualize better the cytoskeleton network and its colocalization with GABARAP. It is surprising that we found GABARAP also associated with microfilaments. In vitro binding assays for the interaction of GABARAP and actin indicated that the association might be mediated by other proteins. Some MAPs were found to interact with both microtubules and actin. It is interesting that MAP 2 and τ use the same domain for actin and tubulin interactions (Correas et al., 1990). Studies on a microtubule-interacting protein, Mip-90, have shown that Mip-90 has two expression patterns in human fibroblast cells : one colocalized with microtubules and another colocalized with microfilaments (Gonzalez et al., 1998). The association of GABARAP with microtubules and microfilaments is both intriguing and interesting. Both tubulin and actin were found to coimmunoprecipitate with GABAA receptors (Kannenberg et al., 1997), and the staining pattern of gephyrin was altered by both microtubule and microfilament drugs (Kirsch and Betz, 1995). Dual involvement of actin- and tubulin-based networks was suggested as well for the anchorage of NMDA receptors. NR1 subunits associate with both tubulin and actin, and NR2 subunits associate with tubulin (Matsuda and Hirai, 1999 ; van Rossum et al., 1999). Furthermore, in addition to the MAP CRIPT, α-actinin, which is an actin binding protein, interacts with NMDA receptors (Wyszynski et al., 1997). Our finding suggests that microtubule and microfilament may work differentially or/and cooperatively on GABAA receptor clustering through GABARAP, and studies on their localization are underway.
Together, our data provide the first evidence for direct interaction between a GABAA receptors binding molecule and the microtubules, as well as their association in vivo. Indirect association of GABARAP with microfilaments suggests that the actin-based network may also be involved in GABAA receptor clustering. The observations of this study suggest a possible mechanism for cytoskeleton-mediated postsynaptic specialization and organization of GABAA receptors.