Structure of GABARAP in Two Conformations

Neuron, Vol. 33, 63–74, January 3, 2002, Copyright 2002 by Cell Press

Structure of GABARAP in Two Conformations: Implications for GABAA Receptor Localization and Tubulin Binding Joseph E. Coyle,1 Seema Qamar,1 Kanagalaghatta R. Rajashankar,2 and Dimitar B. Nikolov1,3 1 Cellular Biochemistry and Biophysics Program Memorial Sloan-Kettering Cancer Center 1275 York Avenue New York, New York 10021 2 Brookhaven National Laboratory Brookhaven, New York 11973

Summary GABARAP recognizes and binds the ␥2 subunit of the GABAA receptor, interacts with microtubules and the N-ethyl maleimide sensitive factor, and is proposed to function in GABAA receptor trafficking and postsynaptic localization. We have determined the crystal structure of human GABARAP at 1.6 A˚ resolution. The structure comprises an N-terminal helical subdomain and a ubiquitin-like C-terminal domain. Structure-based mutational analysis demonstrates that the N-terminal subdomain is responsible for tubulin binding while the C-terminal domain contains the binding site for the GABAA. A second GABARAP crystal form was determined at 1.9 A˚ resolution and documents that GABARAP can self-associate in a head-to-tail manner. The structural details of this oligomerization reveal how GABARAP can both promote tubulin polymerization and facilitate GABAA receptor clustering. Introduction The GABAA receptor and the glycine receptor (GlyR) are the principal fast inhibitory neurotransmitter receptors of the vertebrate nervous system (Moss and Smart, 2001). They are both chloride channels that belong to the superfamily of pentameric, ligand-gated ion channels exemplified by the nicotinic acetylcholine receptor (nAChR) (Ortells and Lunt, 1995; Schofield et al., 1987). Individual receptor subunits within this superfamily share a common membrane topology comprising four transmembrane segments (TM1–4) and extracellular N and C termini (Unwin, 1998). Clustering of neurotransmitter receptors at the postsynaptic membrane is a critical requirement for efficient neurotransmission and appears to be mediated by intracellular proteins that can, either individually or as part of multiprotein complexes, physically link the membranelocalized receptors to the cytoskeleton (Colledge and Froehner, 1998; Moss and Smart, 2001). A large (90 amino acids) intracellular loop, connecting TM3 and TM4, is thought to contain the binding sites for cytoplasmic proteins which are required for ligand-gated ion channel receptor trafficking and clustering (Hanley et al., 1999; Maimone and Enigk, 1999; Meyer et al., 1995; Ramarao and Cohen, 1998). The GABAA receptor-associated pro3

Correspondence: [email protected]

tein (GABARAP) was identified in a yeast two-hybrid search for proteins that bind to this loop of the GABAA receptor (Wang et al., 1999). The 14 kDa GABARAP binds in vitro to GST fusion proteins containing the GABAA receptor ␥2 intracellular loop, coimmunoprecipitates with GABAA receptor subunits from brain extracts, and colocalizes with the punctate staining of GABAA receptors in cell bodies and neuronal processes on cultured cortical neurons (Kneussel et al., 2000; Wang et al., 1999). GABAA receptors are assembled from multiple subunit classes into heteromeric pentamers: the most common subtypes are thought to consist of 2␣, 2␤, and 1␥ subunits (Chang et al., 1996; Farrar et al., 1999; Tretter et al., 1997). Analyses of the subunit specificity for GABARAP association with the GABAA receptor have revealed that GABARAP interacts with ␥2 and ␥1 isoforms but not with ␣1, ␤, ␥3, or ␦ isoforms (Kneussel et al., 2000; Wang et al., 1999). The ability of GABARAP to promote clustering of GABAA receptors has been observed directly in transfected quail fibroblast cells coexpressing GABARAP and GABAA receptors (Chen et al., 2000). Interestingly, GABARAP-dependent receptor clustering alters the channel kinetics of the GABAA receptor: clustered receptors have a lower affinity for GABA, deactivate faster, and exhibit slower desensitization at a given GABA concentration. Other studies have revealed that the majority of GABARAP puncta are seen in intracellular compartments (putative ER and Golgi structures) of spinal cord and cultured cortical neurons (Kittler et al., 2001; Kneussel et al., 2000; Okazaki et al., 2000). Therefore, GABARAP may participate in the intracellular trafficking of GABAA receptors, e.g., targeting receptors to the plasma membrane via the Golgi apparatus or endocytotic sequestering of surface receptors. GABARAP interacts with both soluble and polymerized forms of tubulin in vitro, and colocalizes with microtubules in vivo (Wang et al., 1999; Wang and Olsen, 2000). The same in vivo study revealed that GABARAP also associates with microfilaments, although this interaction appears to be indirect (Wang and Olsen, 2000). Pulldown assays using a GST-GABARAP fusion protein have revealed that GABARAP can simultaneously interact with tubulin and the GABAA receptor (Wang and Olsen, 2000). Thus, GABARAP can potentially link GABAA receptors to the cytoskeleton, consistent with it playing a role in GABAA receptor trafficking and/or localization at the postsynaptic membrane. GABARAP has been shown to associate with gephyrin, a protein with a central role in coordinating GlyR and GABAA receptor postsynaptic localization (Essrich et al., 1998; Feng et al., 1998; Kirsch et al., 1993; Kneussel et al., 1999), both in vitro and in transfected PC12 cells, where GABARAP promotes the recruitment of gephyrin to the plasma membrane (Kneussel et al., 2000). The ability of gephyrin to directly interact with the large intracellular loop of the GlyR ␤ subunit and with microtubules is essential for postsynaptic clustering of the GlyR (Kirsch and Betz, 1995; Kirsch et al., 1991; Meyer et al., 1995). Gephyrin is also required for GABAA receptor

Neuron 64

Table 1. Summary of Crystallographic Analysis Crystal Data Collection Resolution (A˚) Unique Completeness (%) Redundancy Rmerge (%) (last shell) Phasing Sites Rcullis (anomalous) Rcullis (isomorphous) Phasing power Figure of merit (pre/post DM)

Native

Pt

Ir

Br

30.0–1.6 14767 99.8 4.6 4.4 (30.5)

20–2.4 4704 99.9 6.9 6.9 (24.4)

30–2.5 4102 100 6.8 5.5 (29.1)

22–2.0 8492 100 6.9 7.5 (36.2)

3 0.96 — — 0.32/0.76

5 0.91 0.92 0.91

4 0.92 0.87 1.09

7 0.96 0.95 0.49

Rms Deviations

Refinement: Crystal Form

Resolution (A˚)

Reflections ⬎ 1.0 ␴ Working/Test

Water Molec.

R Value Rcryst/Rfree

Bonds (A˚)

Angles (⬚)

B Factors (A˚2)

Monomeric Oligomeric

20.0–1.6 20.0–1.9

13234/838 10684/519

138 126

19.9/24.4 22.4/26.5

0.006 0.006

1.45 1.37

2.10 2.44

Rmerge ⫽ ⌺|I ⫺ ⬍I⬎|/⌺ I where I ⫽ observed intensity, ⬍I⬎ ⫽ average intensity obtained from multiple observations of symmetry-related reflections. Rcullis ⫽ ⌺||FH(obs)|⫺|FH(calc)||/⌺|FH(obs)|, |FH(obs)| ⫽ observed heavy atom structure factor amplitude and |FH(calc)| ⫽ calculated heavy atom structure factor amplitude. Phasing power ⫽ root-mean-square(|FH|/E), |FH| ⫽ heavy atom structure factor amplitude, and E ⫽ residual lack of closure. Rms deviations in bond lengths and angles are the respective root-mean-square deviations from ideal values. Rms thermal parameter is the rms deviation between the B values of covalently bound atomic pairs.

clustering, (Essrich et al., 1998; Kneussel et al., 1999); however, this activity does not appear to be mediated through a direct interaction between the two proteins. Recent studies have demonstrated that GABARAP also associates with ULK-1 (Okazaki et al., 2000), a neuronal serine/threonine specific kinase implicated in neuronal outgrowth, and with N-ethylmaleimide sensitive factor (NSF) (Kittler et al., 2001), a protein involved in vesicle fusion, bringing the number of GABARAP-interacting proteins to at least five. The large number of protein partners, as well as its wide tissue expression pattern (Wang et al., 1999), suggest that GABARAP may function as an adaptor protein in a variety of cell processes. Here we describe crystal structures of GABARAP in two different conformations corresponding to monomeric and oligomerized protein. The structures provide a model for how GABARAP may facilitate clustering of GABAA receptors. To validate this model, we have used structure-based mutagenesis to examine the role of the individual domains of GABARAP in binding to the GABAA receptor and tubulin. Results and Discussion Structure Determination The human GABARAP protein was expressed in E. coli and purified using cation exchange and gel filtration chromatography. Crystallization screens identified two sets of conditions that produced well diffracting crystals. The initial crystals obtained diffracted to 1.6 A˚ resolution and the structure was determined utilizing multiple isomorphous replacement with anomalous scattering and data collected with iridium, platinum, and bromine derivatives (Table 1). The refined structure (Rfree ⫽ 24.4%, comprising amino acids 2–116) shows no significant GABARAP intermolecular interactions and is therefore referred to as the structure of monomeric GABARAP. The structure of a second GABARAP crystal form was determined to

1.9 A˚ resolution by molecular replacement using the first structure as a search model. This structure (Rfree ⫽ 26.5%, comprising amino acids 1–116) reveals a continuous head-to-tail GABARAP multimer (1322 A˚2 buried area at each intermolecular interface) and is therefore referred to as the structure of oligomeric GABARAP. Overall Structure of GABARAP GABARAP (Figure 1) is composed of two domains: a C-terminal domain (residues 27–117) and a small N-terminal subdomain (residues 1–26) that contains an ␣ helix (H2) and a 310 helix (H1). The C-terminal domain (highlighted in yellow on Figure 1A) shows structural similarity to ubiquitin, comprising a central fourstranded ␤ sheet with two helices (H3 and H4) and connecting loops packed against the concave side of the sheet. Despite having only 7% sequence identity, the C-terminal domain of GABARAP and ubiquitin superimpose with a root mean square deviation (rmsd) of 2.7 A˚ over 69 core ␣-carbon atoms. The N-terminal helical subdomain (highlighted in blue on Figure 1A), which is not present in ubiquitin, aligns on the opposite side of the C-terminal domain. Helix H2 angles away from the C-terminal domain, but a sharp turn around Pro-10 directs the N terminus including helix H1 back toward the central ␤ sheet. GABARAP Exists in Two Distinct Conformations The monomeric conformation presumably represents the structure of GABARAP at relatively low protein concentrations when it is not associated with other proteins or membranes, and indeed, gel filtration and dynamic light scattering experiments document that GABARAP is monomeric in solution in concentrations up to 100 ␮M and salt concentrations in the range of 0.05–1.0 M NaCl (data not shown). The oligomeric conformation of GABARAP in our crystals, on the other hand, is stabilized by the high salt conditions (2.4 M ammonium sulfate) which enhance the hydrophobic interactions dominating

Structure of GABARAP 65

Figure 1. Comparison of the Closed and Open Conformations of GABARAP Ribbon diagram of GABARAP in the closed, monomeric (A) and open, oligomeric (B) conformations. The N- and C-terminal domains are highlighted in blue and yellow, respectively. The views in (C) and (D) are rotated by 90⬚ relative to those shown in (A) and (B). (E) Superimposed C␣ traces of the closed (colored red) and the open (colored green) conformations of GABARAP. (F) Intermolecular contacts between neighboring molecules in the GABARAP oligomer stabilize the open conformation. The N terminus of one GABARAP molecule (colored purple) forms a parallel intermolecular ␤ sheet with the S2 strand of an adjacent molecule in the crystal. The locations of the proposed tubulin and GABAA interaction sites are indicated.

the multimerization interface. GABARAP oligomerization in vivo would presumably be induced and stabilized via interactions with other proteins, such as tubulin, or with membranes. The biophysical experiments discussed further below support our interpretation that the GABARAP oligomerization observed in the crystals is biologically relevant and not simply a result of the crystallization conditions. Overall, the structures of the two GABARAP crystal forms (monomeric form, Figures 1A and 1C; oligomeric form, Figures 1B and 1D) are very similar with an rmsd of 1.2 A˚ over ␣-carbons 10–116 (Figure 1E). They diverge significantly, however, around the first 10 amino acids (NH(10)). In the monomeric structure, the NH(10) residues form a 310 helix (H1) and a short stretch of extended structure orientated perpendicularly to H2 with the N terminus projecting down toward the surface of the

ubiquitin-like C-terminal domain. The resulting conformation is relatively compact and is subsequently referred to as the “closed” conformation. In the oligomeric crystal form, the NH(10) residues adopt an extended structure that, in contrast to the closed conformation, is projected away from the ubiquitin-like domain and binds a neighboring molecule in the crystal lattice (Figure 1F). Specifically, the N-terminal six amino acids of one molecule interact with the S2 strand of an adjacent molecule, giving rise to a head-to-tail GABARAP chain in the crystal lattice. This conformation is subsequently referred to as the “open” conformation. Interactions Stabilizing the Open and Closed Conformations of GABARAP The alternate conformations of the NH(10) sequence are stabilized by different networks of intramolecular

Neuron 66

Figure 2. Distinct Networks of Interactions Stabilize the Closed and Open Conformations of GABARAP (A) The hydrophobic mini-core at the interface between the N- and C-terminal domains of GABARAP in the closed, monomeric conformation. Side chains are shown in ball and stick representation. Tyr-5 and Phe-3 from the N-terminal domain are colored yellow. Ile-32, Phe-104, Tyr-106, and Ala-108 from the C-terminal domain are colored gray. (B) A salt bridge and hydrogen bonding interactions also stabilize the closed conformation. (C) The molecular surface of GABARAP in the open conformation is shown in green. The first six residues from an adjacent molecule in the crystal lattice are shown in stick representation and colored yellow. The intermolecular oligomerization interface is colored white. (D and E) Two views of the intermolecular contacts stabilizing the open, ologomeric conformation. One GABARAP molecule is colored green and a neighboring molecule in the crystal lattice is colored yellow. Side chains are shown in ball and stick representation. Hydrogen bonds are shown as dashed cyan lines. Oxygen and nitrogen atoms are colored red and blue, respectively.

(closed conformation; monomeric GAPARAP) and intermolecular (open conformation; multimeric GABARAP) contacts. In the closed conformation, the NH(10) region is held in place through numerous interactions with residues located on the convex face of the central ␤ sheet (Figures 2A and 2B). Phe-3 and Tyr-5 from the NH(10) region make van der Waal’s contacts with Ile-32 from strand S1, and Phe-104, Tyr-106, and Ala-108 from strand S4, generating a hydrophobic mini-core at the interdomain interface (Figure 2A). Additional stability is provided by a salt bridge between Glu-100 and Lys-6 and through hydrogen bonding between the side chain carboxylate of Glu-34 and the backbone amide groups of Val-4 and Tyr-5 (Figure 2B). The NH(10) sequence in the open conformation is flipped almost 180⬚ relative to its position in the closed conformation. This dramatic movement hinges around Pro-10, which caps helix H2. Residues 1 to 6 in the open conformation interact with a neighboring molecule in the crystal lattice, binding to a conserved region of GABARAP that is implicated in protein-protein interactions (see below). These six residues (Met-1*, Lys-2*, Tyr-3*, Val-4*,

Phe-5*, and Lys-6*) form an extended structure (with three amino acids forming a short ␤ strand) that aligns parallel to the S2 strand of an adjacent GABARAP molecule in the crystal lattice. The intermolecular contacts observed in the open structure bury 1322 A˚2 of solventaccessible molecular surface and form a remarkably complimentary interface, both in shape and in the arrangement of polar and nonpolar surfaces (Figures 2C– 2E). Specifically, the side chain of Met-1* fits snugly into a hydrophobic pocket lined by the side chains of Ile-21, Pro-30, Leu-50, and Phe-104. (Figure 2C). The hydrophobic side chains of Tyr-3*, Val-4*, Phe-5*, and the C␣-C␥ stretch of Lys-2* form additional nonpolar interactions with Tyr-25, Tyr-49, Val-51, Phe-60, and Leu-55. Hydrogen bonds also stabilize the intermolecular interface observed in the open structure: there are three main chain ␤ sheet hydrogen bonds and the terminal guanidinium group of Arg-28 hydrogen bonds to the backbone carbonyl oxygen of Val-4*. In addition, Arg-28 forms an amino-aromatic interaction with the phenyl ring of Phe-5*, and Lys-6* forms a salt bridge with Asp-54 (Figures 2D and 2E). The hinge-like opening of the

Structure of GABARAP 67

N terminus in the open conformation exposes residues on the convex face of the central ␤ sheet that are occluded in the closed conformation. This surface provides a potential binding site for GABARAP-interacting proteins. The distinct networks of highly specific interactions stabilizing the N terminus of GABARAP in both the open and closed states indicate that these conformations are not simply the result of a flexible N terminus, but rather a biologically important conformational switch. The intermolecular interface observed in the open GABARAP crystal form could be mimicking a bona fide GABARAP protein-protein interaction; however, no match for the N-terminal MKYVFK sequence of GABARAP is observed in the sequence of known GABARAP binding proteins (GABAA receptor ␥2 subunit, gephyrin, tubulin, NSF, and ULK-1). We suggest, therefore, that the head-to-tail oligomer of GABARAP molecules represents a functionally relevant molecular assembly. The GABARAP Molecular Surface The distribution of conserved residues over the GABARAP surface in the closed and open conformations is shown in Figures 3A and 3B. Immediately apparent in both structures is the extended conserved hydrophobic surface (comprising Ile-21, Pro-30, Tyr-49, Leu-50, Val-51, Leu-55, Phe-60, Leu-63, and F-104) centered around the region of GABARAP that constitutes the oligomerization interface. The precise arrangement of these residues is slightly changed in the open protein conformation relative to the closed conformation. In particular, it is apparent that the hydrophobic pocket that accommodates M-1* is not so well defined in the closed structure, indicating that minor side chain rearrangements occur upon head-to-tail oligomerization of GABARAP molecules in the open conformation. It is likely that all members of the GABARAP family utilize this large hydrophobic patch in mediating associations with protein partners since clusters of conserved hydrophobic residues are a distinct feature of protein-protein interfaces (Janin and Chothia, 1990; Young et al., 1994). In addition, a highly conserved ridge of basic residues (Lys-46, Lys-48, Arg65, Lys-66, and Arg-67) runs alongside the large hydrophobic patch underscoring the important functional role of this region among GABARAP family members. Both the complimentary nature of the intermolecular contacts stabilizing the open conformation of GABARAP and the highly conserved, hydrophobic nature of the GABARAP surface mediating these contacts provides strong evidence that the head-to-tail oligomerization of GABARAP is functionally important. Possible biological roles for such an intermolecular arrangement are discussed below. Function of the GABARAP C-Terminal Domain In order to define the functional role of the N- and C-terminal domains of GABARAP, as well as the functional relevance of its observed multimerization, we constructed two mutant proteins, one lacking the first 10 amino acids (⌬N10) and the other lacking the entire N-terminal subdomain (⌬N27). We then examined the ability of these mutant proteins to bind the GABAA receptor and tubulin. In the original study identifying GABARAP

(Wang et al., 1999), a GABARAP binding site was localized to an 18 amino acid sequence near the C-terminal end of the large GABAA receptor intracellular loop. We, therefore, first measured the affinity of these mutant proteins for an 18 residue peptide (RTGAWRHGRIHIRIAKMD) corresponding to this receptor region. Since the peptide contains a single tryptophan residue and GABARAP has no tryptophans, we used intrinsic tryptophan fluorescence to measure the binding (Figures 4A and 4B). The saturating binding curve data fits a ligand binding equation with an apparent Kd value of 1.29 ⫾ 0.09 ␮M. Two other GABA receptor-derived peptides were assayed for binding to GABARAP: (1) a C-terminally truncatated 18-mer peptide corresponding to the 13-mer sequence— RTGAWRHGRIHIR—and (2) an N-terminally truncated 13-mer peptide corresponding to the 11-mer peptide sequence—GAWRHGRIHIR. All three peptides bind GABARAP with comparable affinities (Figure 4B), while a control peptide does not, demonstrating that the GAWRHGRIHIR sequence of the GABAA receptor constitutes the GABARAP recognition motif. A second set of experiments (Figure 4A) documents that the ⌬N10 GABARAP mutant protein binds the 18-mer GABAA receptor peptide with identical affinity to that observed for full-length GABARAP, while the Kd for 18-mer interaction with the ⌬N27 mutant is only slightly reduced (Kd ⫽ 6.10 ⫾ 0.29 ␮M). The data identifies the ubiquitin-like C-terminal domain of GABARAP as responsible for binding to the large intracellular loop of the GABAA receptor. This conclusion is consistent with yeast two-hybrid and GST fusion pulldown studies that have identified GABARAP residues 36–117 as important for GABAA receptor recognition (Chen et al., 2000; Wang et al., 1999). A model for GABAA receptor recognition by GABARAP can be inferred from the structural homology between the ubiquitin-like domain of GABARAP and the ubiquitinlike domain of Elongin B (rmsd of 2.2 A˚ over 68 core ␣-carbon atoms). Elongin B forms a heterodimeric complex with Elongin C as a component of the multiprotein VHL tumor-suppressor complex (Duan et al., 1995; Kishida et al., 1995). The crystal structure of a VHL:Elongin B:Elongin C ternary complex has revealed that the Elongin B:C interface is centered around an intermolecular parallel ␤ sheet involving the S2 strand of Elongin B and the S2 strand of Elongin C (Stebbins et al., 1999) (Figure 4D). Elongin B does not contain the N-terminal subdomain observed in GABARAP; however, helix H1 from Elongin C occupies a similar position to the H2 helix of GABARAP. The ubiquitin-like domains of RafGBD and RalGDS also display a similar intermolecular ␤ strand interface in complex with Ras, although in these cases, the ␤ sheet pairing is antiparallel (Huang et al., 1998; Nassar et al., 1995). We suggest that the GABARAP: GABAA receptor interface is structurally similar to that observed in the Elongin B:C complex. A comparison between the structure of two adjacent GABARAP molecules in the open conformation with the Elongin B:C complex is shown in Figures 4C and 4D (see also Figure 1F). The N-terminal residues of the Elongin B S2 strand make the majority of the intermolecular contacts with the S2 strand of Elongin C. In contrast, the C-terminal residues of the GABARAP S2 strand are required for binding to the sequence from a neighboring molecule

Neuron 68

Figure 3. Conserved Surface Residues in the Open and Closed Forms of GABARAP Sequence conservation projected onto the molecular surface of GABARAP in the closed, monomeric (A) and open, oligomeric conformation (B). The views on the right are rotated 180⬚ around the y axis. Conserved, surface-exposed residues are colored yellow (hydrophobic), blue (basic), and red (acidic). The sequence of GABARAP was compared with human GATE-16, human LC-3, and S. cerevisiae Aut7p. The following residues were considered similar: Lys and Arg, Glu and Asp, Phe, Met, Val, Leu, and Ile, Phe and Tyr.

in the oligomeric crystal form. Indeed, the binding sites for Elongin C and the MKFVYK sequence of GABARAP only slightly overlap on the superimposable S2 strands of Elongin B and GABARAP, respectively. Both monomeric and oligomeric GABARAP could, therefore, potentially associate with a protein partner in an analogous fashion to that observed in the elongin complex. Furthermore, the sequence of GABARAP between residues 36 and 68 has been implicated in binding to the GABAA receptor ␥2 subunit (Wang et al., 1999). This region of GABARAP encompasses both the S2 strand and a large number of the conserved surface exposed residues described above, underscoring the potential structural and architectural similarity between the GABARAP:GABAA and the Elongin B:C complexes. Function of the GABARAP N-Terminal Subdomain Using the ⌬N10 and ⌬N27 mutants, we examined further the functional role of the individual GABARAP domains using an in vitro tubulin polymerization assay. Full-

length GABARAP produces a rapid increase in the optical turbidity of a 9.1 ␮M (1 mg ml⫺1) tubulin solution (Figure 5A). Taxol induced assembly of microtubules under the same conditions exhibits slower kinetics and gives rise to a ⵑ5-fold lower optical turbidity signal (Figure 5A). The higher turbidity induced by GABARAP suggests that it can promote bundling of microtubules in a manner reminiscent of other microtubule-associated proteins (Brandt and Lee, 1993; Deka et al., 1998; Melki et al., 1991). To confirm this hypothesis, we examined the ability of GABARAP to augment the optical turbidity signal of a solution of taxol-induced microtubules. Sequential addition of taxol leads to the assembly of predominantly single microtubules (Yoon and Oakley, 1995), therefore any increase in the turbidity of a taxolinduced microtubule solution can be indicative of microtubule bundling. The results on Figure 5B reveal that GABARAP produces such an increase, and thus displays a microtubule bundling activity. The tubulin polymerization assay was also performed

Structure of GABARAP 69

Figure 4. GABAA Receptor Binding by GABARAP (A) Peptides corresponding to regions of the large intracellular loop of the GABAA receptor ␥2 subunit were assayed for binding to GABARAP using intrinsic tryptophan fluorescence. The sequences of the peptides were: 18-mer (RTGAWRHGRIHIRIAKMD), 13-mer (RTGAWRHGRIHIR), and 11-mer (GAWRHGRIHIR). Data were fitted using the equation: fluorescence signal ⫽ (Fmax [peptide])/(Kd ⫹ [peptide]), where Fmax ⫽ the maximum fluorescence signal and Kd ⫽ the dissociation constant. (B) Intrinsic tryptophan fluorescence was also used to measure the dissociation constant for 18-mer peptide binding to the ⌬N10 and ⌬N27 GABARAP mutant proteins. Structural comparison between (C) two adjacent GABARAP molecules in the open, oligomeric conformation (colored green and yellow) and (D) the Elongin B (green):Elongin C (yellow) complex.

with the ⌬N10 and ⌬N27 GABARAP mutant proteins (Figure 5A). Strikingly, there is no turbidity increase when tubulin is incubated with the ⌬N27 mutant, highlighting the absolute requirement for the GABARAP N-terminal subdomain in tubulin binding. Incubation of tubulin with the ⌬N10 mutant does lead to some increase in optical turbidity, although the kinetics of the reaction are significantly slower than observed for the full-length GABARAP protein. This data indicates, therefore, that residues 10–27 constitute the tubulin binding domain of GABARAP and that the NH(10) sequence, which mediates protein oligomerization, greatly enhances the ability of the tubulin binding region to promote microtubule assembly. Of course, the possibility that the whole N-terminal subdomain 1–27 constitutes the tubulin binding site and the ⌬N10 behavior simply reflects the removal of some of the tubulin-interacting residues cannot be excluded. Previous deletion studies have indicated that the N-terminal 35 amino acids of GABARAP may contain the site for tubulin binding (Wang et al., 1999; Wang and Olsen, 2000). In addition, a peptide corresponding to residues 1–22 from GABARAP has been shown to promote tubulin polymerization in vitro (Wang and Olsen, 2000), although with kinetics that were much slower than we observe for full-length GABARAP. Furthermore, the concentration of peptide required to induce tubulin polymerization was higher than we observed for GABARAP.

Together with these data, our results define the minimal tubulin binding region of GABARAP to the sequence from residues 10–22. Examination of the GABARAP structure reveals that residues 10–22 encompass helix H2 in both the closed and open conformations. The decreased ability of the 1–22 peptide in promoting microtubule formation, relative to full-length GABARAP, suggests that: (1) the 10–22 sequence must be constrained in a helical conformation for effective tubulin binding and/or (2) other regions of GABARAP may augment the tubulin polymerization activity of GABARAP. The N-terminal domain, particularly the H2 helix region, contains a high content of basic residues, most of which align on the exposed face of the helix (Figure 5C). Tubulin binding motifs identified in other proteins often have a net positive charge that is thought to be important for binding to the highly acidic C-terminal region of tubulin (Littauer et al., 1986; Marya et al., 1994; Serrano et al., 1984). This tubulin region is located on the outer face of microtubules and therefore provides an accessible, acidic binding site for the association of microtubule binding proteins (Nogales et al., 1999). Microtubules are comprised of tubulin heterodimers arranged in linear head-to-tail protofilaments that associate laterally to form ⵑ25 nm wide hollow cylindrical polymers (Amos, 2000). Interestingly, the intermolecular distance between adjacent tubulin monomers in the 2D

Neuron 70

Figure 5. The N-Terminal Domain of GABARAP Is Important in Promoting Tubulin Polymerization (A) The kinetics of tubulin polymerization monitored by measuring the increase in sample absorption at 340 nm. Microtubule assembly was measured in the presence of GABARAP (5 and 50 ␮M), ⌬N10 (100 ␮M), ⌬N27 (100 ␮M), and taxol (10 ␮M). (B) The turbidity of one sample containing 1 mg ml⫺1 tubulin after sequential additions of taxol at the final concentration indicated. The sample was incubated for 30 min after each addition step. Subsequent addition of GABARAP (50 ␮M) produces a large increase in sample turbidity, indicating that GABARAP may promote microtubule bundling.

sheets used to determine the atomic structure of tubulin (Nogales et al., 1998, 1999) is very close to the intermolecular distance observed between adjacent molecules of GABARAP in the oligomeric crystal form: ⵑ40 A˚ for tubulin and ⵑ38 A˚ for GABARAP. Thus, GABARAP may bind to microtubules with a stoichiometry of 1 GABARAP:1 tubulin monomer and promote microtubule assembly by crosslinking adjacent tubulin dimers. The observed headto-tail GABARAP oligomer (Figure 1F) may, therefore, represent the tubulin bound state of GABARAP. This hypothesis predicts that the NH(10) region of GABARAP is required for effective tubulin polymerization activity and, indeed, we now show that the ⌬N10 mutant displays a substantially diminished ability to promote tubulin polymerization relative to full-length GABARAP. Furthermore, the dependence of the tubulin polymerization on the GABARAP concentration (Figure 5A) indicates that the binding is stoichiometric rather than simply catalytic. Implications for GABAA Receptor Clustering by GABARAP The protein-protein interactions in which GABARAP is implicated suggest that it may have an important role in postsynaptic receptor clustering via physically linking the intracellular region of GABA receptors to the cytoskeleton. The exact functional role of GABARAP, how-

ever, is yet unclear with some studies showing that it is predominantly localized in intracellular compartments and only found at low levels in synapses (Kittler et al., 2001; Kneussel et al., 2000; Okazaki et al., 2000), and others (Chen et al., 2000; Wang et al., 1999) showing that it is localized at the plasma membrane and can promote GABAA receptor clustering. The head-to-tail oligomerization of GABARAP molecules in the crystals (Figure 1F) could directly relate to the proposed role of GABARAP in clustering GABAA receptors. Linear polymers of GABARAP could potentially tether multiple GABAA receptors into a localized assembly. Generating the two-dimensional arrays of receptors observed in electron micrographs of postsynaptic membranes, however, would require some mechanism to crosslink linear chains of GABARAP. In the case of the GABAA receptor, the effectiveness of this GABARAP-dependent clustering mechanism would be increased by the presence of multiple ␥2 subunits per GABAA receptor pentamer, i.e., facilitating the formation of a two-dimensional network of receptors. Measurements of the exact subunit arrangement and stoichiometry of GABAA receptors are controversial, with the most common receptor subtypes believed to be composed of ␣, ␤, and ␥ subunits in the ratio of 2␣:2␤:1␥ (Baumann et al., 2001; Chang et al., 1996; Farrar et al., 1999), but also with observed

Structure of GABARAP 71

Figure 6. Structure-Based Sequence Alignment of GABARAP Homologs Conserved residues are indicated by (*), similar residues are indicated by (:). Secondary structure elements from the closed GABARAP crystal structure are shown above the sequences. Residue positions are numbered according to the GABARAP sequence. The GABARAP, LC-3, and GATE-16 sequences are all human forms. The Aut7p sequence is from S. cerevisiae.

1␣:2␤:2␥ or 2␣:1␤:2␥ stoichiometry (Backus et al., 1993; Khan et al., 1994; Quirk et al., 1994). Even a relatively low abundance of receptor subtypes containing multiple ␥2 subunits would promote two-dimensional GABAA receptor clustering by the head-to-tail linked GABARAP model proposed above. Additional proteins could also modulate the ability of GABARAP to “crosslink” multiple GABAA receptors. Of course, the precise GABARAP role in receptor clustering and the function of the observed oligomerization can only be clarified in carefully planned in vivo studies. In this respect, our results provide a starting point for the design of GABARAP proteins with altered properties (like the ⌬N10 mutant which is defective in oligomerization and tubulin binding but retains wild-type GABA receptor affinity) which can be used, for example, in cell-based transfection assays or in mouse knockout experiments. GABARAP Family GABARAP has a high level of sequence similarity to an expanding family of proteins that have been implicated in a variety of cellular processes involving membranelocalized and/or microtubule-associated protein-protein interactions (Figure 6). The GATE-16 (Golgi-associated ATPase enhancer of 16 kDa) protein has been to shown to participate in intra-Golgi protein transport (LegesseMiller et al., 1998; Sagiv et al., 2000), and is the only other family member whose structure is known. The crystal structure of GATE-16 (Paz et al., 2000) is topologically similar to the closed, monomeric conformation of GABARAP, with an rmsd of 1.0 A˚ over 112 ␣-carbon atoms. Light chain 3 (LC3) of microtubule-associated protein 1B (MAP1B) and Aut7p/Apg8p also have significant sequence similarity to GABARAP. LC3 has been shown to bind directly to microtubules (Mann and Hammarback, 1994) and also to AU-rich elements in the 3⬘ UTR of fibronectin mRNA (Zhou et al., 1997).

Aut7p/Apg8p is a yeast protein involved in autophagy, a degradative pathway that sequesters cytoplasm and organelles for delivery to the vacuole or lysosome (Klionsky and Emr, 2000) and appears to regulate autophagosome size (Huang et al., 2000; Lang et al., 1998). Recently, a novel protein lipidation mechanism was discovered, in which Aut7p/Apg8p is covalently conjugated to phosphatidylethanolamine (Ichimura et al., 2000) leading to the protein’s tight association with the autophagosome membrane. The lipidation process involves proteolytic processing in which the C-terminal arginine of Aut7p/Apg8p is removed by the Aut2/Apg4 cysteine protease, leaving an essential glycine residue at the C terminus (Kim et al., 2001; Kirisako et al., 2000). Phosphatidylethanolamine is then covalently attached to the glycine by a ubiquitination-like system involving an E1 protein (Apg7) and an E2 protein (Apg3/Aut1). Intriguingly, this C-terminal glycine is absolutely conserved in LC-3, GABARAP, and GATE-16. In addition, these three homologs have been shown to be substrates for the human form of Apg7 (hApg7) (Tanida et al., 2001). This suggests that GABARAP may be lipid modified in a similar manner, a process that would provide a mechanism to dynamically regulate the subcellular location of GABARAP. Lipid modification and membrane localization would increase the affinity of GABARAP for the GABAA receptor and might facilitate GABARAP-dependent receptor trafficking. Moreover, the C-terminal glycine of GABARAP (Gly-116) is located at the opposite side of the molecule from the N-terminal subdomain (Figure 1) and phosphatidylethanolamine-mediated membrane association would not interfere with tubulin binding. Finally, the ⵑ43 kDa protein rapsyn promotes clustering of the nAChR by directly associating with the large intracellular loop of the receptor, and the membrane localization of rapsyn via a N-terminal myristoyl group enhances its clustering function (Phillips et al.,

Neuron 72

1991; Ramarao and Cohen, 1998). In addition, rapsyndependent receptor clustering is thought to involve selfassociation of rapsyn (Ramarao et al., 2001; Ramarao and Cohen, 1998). Hence, GABARAP and rapsyn likely share common functional features despite being structurally unrelated. The diversity of proteins known to associate with members of the GABARAP family suggests that the GABARAP fold has evolved to function as an adaptor module. Some members of this family have common protein partners: (1) GABARAP and GATE-16 bind NSF and ULK-1 (Kittler et al., 2001; Okazaki et al., 2000; Sagiv et al., 2000), (2) GABARAP and LC-3 bind tubulin (Mann and Hammarback, 1994; Wang and Olsen, 2000). However, differences in specificity are also evident: GATE-16 does not bind to gephyrin or to the GABAA receptor ␥2 subunit (Kneussel et al., 2000), and Aut7p does not appear to directly associate with tubulin (Lang et al., 1998). Finally, since NSF has been implicated in the trafficking of the ␤-adrenergic receptor and AMPA-type glutamate receptors (McDonald et al., 1999; Nishimune et al., 1998; Osten et al., 1998; Song et al., 1998), NSF association with GABARAP could be important for GABAA receptor transport, e.g., sorting and cycling of GABAA receptors between intracellular and cell surface pools. Postsynaptic receptor clustering and trafficking are complex phenomena, requiring the interplay of multiple proteins in a coordinated and dynamic fashion. The ability of GABARAP to interact with different synaptic and cytoskeletal protein partners suggests that it may play a central role in directing the assembly of multifunctional complexes during both GABAA receptor trafficking and clustering. Our results demonstrate that the N- and C-terminal domains of GABARAP have distinct roles in tubulin and GABAA receptor binding and show that GABARAP can fulfill a bridging function between tubulin and the GABAA receptor. In addition, the structure of GABARAP and the discovery that the molecule can selfassociate in a head-to-tail fashion provide insight into the mechanisms of GABARAP-dependent tubulin polymerization and GABAA receptor clustering, and serve as a starting model for further studies of these processes. Experimental Procedures Cloning, Expression, and Purification of GABARAP GABARAP was cloned from a human adult brain cDNA library (Clontech) using PCR into the Nde1 and BamH1 sites of the pET11a expression vector (Novagen). Escherichia coli BL-21 strain containing the plasmid was grown at 37⬚C to an OD600 of 0.6 and expression was induced by addition of 1 mM IPTG. After induction, the incubation temperature was reduced to 20⬚C and cells were allowed to grow for an additional ⵑ15 hr before harvesting. The cells were resuspended in lysis buffer (20 mM MES [pH 6.5], 50 mM KCl, 1 mM phenylmethylsulfonylfluoride, 0.5 mg ml⫺1 hen egg white lysozyme) and lysed by sonication. The lysate was centrifuged, the supernatant was loaded on a 150 ml SP-Sepharose column, and proteins were eluted with a linear gradient of 0.05–0.6 M KCl. Fractions containing GABARAP were pooled and concentrated to ⵑ20 mg ml⫺1. The protein was then loaded onto a preparative grade Superdex 75 gel filtration column equilibrated in 20 mM MES (pH 6.5), 200 mM KCl. Fractions containing pure GABARAP, which migrates as a monomer, were pooled, concentrated to 30–40 mg ml⫺1, and stored at ⫺80⬚C in 15% glycerol. The ⌬N10 and ⌬N27 mutants were generated through a single round of PCR using the full-length GABARAP construct as a template. Mutant protein expression and

purification were performed using the protocol for full-length GABARAP. The ⌬N27 mutant was concentrated only to 3 mg ml⫺1 due to aggregation problems at higher concentrations. Fluorescence Titration Steady state fluorescence emission was measured on a SPEX spectrofluorimeter using 1 cm pathlength fluorescence cuvettes in buffer containing 20 mM Hepes (pH 7.2), 50 mM KCl, and 5 mM MgCl2. All experiments were performed at 25⬚C with a circulating water bath. Fluorescence emission scans of the GABAA receptor peptides were recorded between 310 and 450 nm, using an excitation wavelength of 295 nm. Excitation and emission slit widths were 5 nm. Ligand binding titrations were performed with similar parameters, but in the time drive mode, measuring emission at a wavelength of 357 nm. Titrations were performed by adding small volumes of peptide to a cuvette containing 2 ml of a 2 ␮M solution of GABARAP. After each stage of addition, the solution in the cuvette was mixed thoroughly and allowed to equilibrate for ⵑ5 min. The total volume of ligand added never exceeded 10% of the total solution volume. An identical control experiment was performed for each titration by adding the peptide to a solution of buffer alone. This contribution was subtracted from the corresponding reading acquired in the presence of protein. Peptides were purchased from Research Genetics in crude resin cleaved form and purified by reverse-phase HPLC using a C18 column. Tubulin Polymerization Assay Tubulin polymerization was measured in a light scattering assay (Schiff et al., 1979). The turbidity was recorded as the optical absorbance at 340 nm using a Hewlett Packard HP Vectra XM UV-VIS spectrophotometer in 30 s intervals at 25⬚C. The polymerization reaction buffer comprised 80 mM PIPES (pH 6.8), 1 mM EGTA, 0.1 mM MgCl2, and 40 ␮M GTP. The polymerization reaction was initiated by adding 10 ␮l of a 10 mg ml⫺1 tubulin solution to 90 ␮l of a buffer containing either 5 or 50 ␮M GABARAP, 100 ␮M ⌬N10, 100 ␮M ⌬N27, or 10 ␮M taxol. Control experiments showed that buffer containing GABARAP alone or tubulin alone does not produce any significant signal. The maximal taxol-induced tubulin polymerization signal was determined by sequentially adding taxol to a final concentration of 75 ␮M. The sample was incubated for 30 min between separate taxol addition steps. Taxol (paclitaxel) was purchased from Sigma. Tubulin was purchased from Cytoskeleton Inc. as a 10 mg ml⫺1 solution. Crystallization and Structure Determination Crystals of GABARAP were obtained by hanging drop vapor diffusion. Protein at 15 mg ml⫺1 in 10 mM Tris (pH 7.2) was mixed with an equal volume of reservoir solution. The crystals containing GABARAP in the closed (monomeric) conformation were grown over a reservoir with 12% PEG MME 2000, 10 mM NiCl2, 20% glycerol, and Tris (pH 8.5). They belong to the P212121 space group (a ⫽ 29.9 A˚, b ⫽ 55.0 A˚, c ⫽ 65.4 A˚) with one molecule in the asymmetric unit. Heavy atom soaks were performed using either 5 mM K2IrCl6 or 3 mM K2Pt(SCN)6 for 72 hr or 0.5 M NaBr for 20 s. The open conformation (oligomeric GABARAP) crystal form was grown over a reservoir containing 2.4 M ammonium sulfate, 5% isopropanol, 0.1 M Hepes (pH 7.5). They belong to the P43212 space group (a ⫽ b ⫽ 36.0 A˚, c ⫽ 178.0 A˚) with one molecule in the asymmetric unit. All crystals were flash frozen in liquid nitrogen for data collection. Data Collection and Structure Determination Data were collected at the X-25 and X9B NSLS beamlines. Images were integrated, scaled, and merged using DENZO and SCALEPACK (Otwinowski and Minor, 1993). Subsequent calculations were performed with the CCP4 program suite (CCP4, 1994). The closed conformation GABARAP structure was determined using the MIRAS method (Table 1). Peak data were input into the program SnB to identify the position of the Pt atoms. The peaks were refined using MLPHARE (CCP4) in the resolution range 30–1.6 A˚. Ir and Br sites were located using anomalous-difference Fourier maps. Native data too were noticed to contain anomalous signal, which later we realized is due to two Ni ions bound to the protein. Derivatives suffered from nonisomorphism and the useful phases were provided mainly

Structure of GABARAP 73

by anomalous signal. The anomalous signal in the native was also used for phasing. The calculated phases had a figure of merit of 0.32, which was improved to 0.76 by density modification with the program DM (CCP4). The map was further improved using free atom refinement and the automatic chain tracing procedure of the wARP program (CCP4). The unambiguous tracing and sequence assignment of GABARAP was completed using O (Jones et al., 1991). Refinement of the model by conventional least-squares algorithms was done with X-PLOR (Brunger, 1993). Stereochemical analysis of the refined models using PROCHECK (CCP4) revealed main chain and side chain parameters better than or within the typical range of values for protein structures determined at the corresponding resolutions. None of the GABARAP residues fell within the disallowed region of the Ramachandran plot. The open crystal structure of GABARAP was determined using the Molecular Replacement method, with the closed GABARAP structure as a search model and the X-PLOR program. Acknowledgments We thank Dr. P. Jeffrey for help with X-ray measurements, and Drs. J. Himanen, M. Henkemeyer, and W. Barton for helpful discussion. D.B.N is a PEW Fellow and a Bressler Scholar. This work was also supported in part by the DeWitt Wallace Fund. Received October 5, 2001; revised November 14, 2001. References Amos, L.A. (2000). Focusing-in on microtubules. Curr. Opin. Struct. Biol. 10, 236–241. Backus, K.H., Arigoni, M., Drescher, U., Scheurer, L., Malherbe, P., Mohler, H., and Benson, J.A. (1993). Stoichiometry of a recombinant GABAA receptor deduced from mutation- induced rectification. Neuroreport 5, 285–288. Baumann, S.W., Baur, R., and Sigel, E. (2001). Subunit arrangement of {gamma}-aminobutyric acid type A receptors. J. Biol. Chem. 276, 36275–36280. Brandt, R., and Lee, G. (1993). Functional organization of microtubule-associated protein tau. Identification of regions which affect microtubule growth, nucleation, and bundle formation in vitro. J. Biol. Chem. 268, 3414–3419. Brunger, A.T. (1993). X-PLOR v.3.1 Manual. (New Haven, CT: Yale University). CCP4 (1994). The CCP4 suite: programs for X-ray crystallography. Acta Crystallogr. D 50, 760–763. Chang, Y., Wang, R., Barot, S., and Weiss, D.S. (1996). Stoichiometry of a recombinant GABAA receptor. J. Neurosci. 16, 5415–5424. Chen, L., Wang, H., Vicini, S., and Olsen, R.W. (2000). The gammaaminobutyric acid type A (GABAA) receptor-associated protein (GABARAP) promotes GABAA receptor clustering and modulates the channel kinetics. Proc. Natl. Acad. Sci. USA 97, 11557–11562. Colledge, M., and Froehner, S.C. (1998). To muster a cluster: anchoring neurotransmitter receptors at synapses. Proc. Natl. Acad. Sci. USA 95, 3341–3343. Deka, J., Kuhlmann, J., and Muller, O. (1998). A domain within the tumor suppressor protein APC shows very similar biochemical properties as the microtubule-associated protein tau. Eur. J. Biochem. 253, 591–597. Duan, D.R., Pause, A., Burgess, W.H., Aso, T., Chen, D.Y., Garrett, K.P., Conaway, R.C., Conaway, J.W., Linehan, W.M., and Klausner, R.D. (1995). Inhibition of transcription elongation by the VHL tumor suppressor protein. Science 269, 1402–1406. Essrich, C., Lorez, M., Benson, J.A., Fritschy, J.M., and Luscher, B. (1998). Postsynaptic clustering of major GABAA receptor subtypes requires the gamma 2 subunit and gephyrin. Nat. Neurosci. 1, 563–571. Farrar, S.J., Whiting, P.J., Bonnert, T.P., and McKernan, R.M. (1999). Stoichiometry of a ligand-gated ion channel determined by fluorescence energy transfer. J. Biol. Chem. 274, 10100–10104.

Feng, G., Tintrup, H., Kirsch, J., Nichol, M.C., Kuhse, J., Betz, H., and Sanes, J.R. (1998). Dual requirement for gephyrin in glycine receptor clustering and molybdoenzyme activity. Science 282, 1321–1324. Hanley, J.G., Koulen, P., Bedford, F., Gordon-Weeks, P.R., and Moss, S.J. (1999). The protein MAP-1B links GABA(C) receptors to the cytoskeleton at retinal synapses. Nature 397, 66–69. Huang, L., Hofer, F., Martin, G.S., and Kim, S.H. (1998). Structural basis for the interaction of Ras with RalGDS. Nat. Struct. Biol. 5, 422–426. Huang, W.P., Scott, S.V., Kim, J., and Klionsky, D.J. (2000). The itinerary of a vesicle component, Aut7p/Cvt5p, terminates in the yeast vacuole via the autophagy/Cvt pathways. J. Biol. Chem. 275, 5845–5851. Ichimura, Y., Kirisako, T., Takao, T., Satomi, Y., Shimonishi, Y., Ishihara, N., Mizushima, N., Tanida, I., Kominami, E., Ohsumi, M., et al. (2000). A ubiquitin-like system mediates protein lipidation. Nature 408, 488–492. Janin, J., and Chothia, C. (1990). The structure of protein-protein recognition sites. J. Biol. Chem. 265, 16027–16030. Jones, T.A., Zou, J.Y., Cowan, S.W., and Kjeldgaard (1991). Improved methods for building protein models in electron density maps and the location of errors in these models. Acta Crystallogr. A 47, 110–119. Khan, Z.U., Gutierrez, A., and De Blas, A.L. (1994). Short and long form gamma 2 subunits of the GABAA/benzodiazepine receptors. J. Neurochem. 63, 1466–1476. Kim, J., Huang, W.P., and Klionsky, D.J. (2001). Membrane recruitment of Aut7p in the autophagy and cytoplasm to vacuole targeting pathways requires Aut1p, Aut2p, and the autophagy conjugation complex. J. Cell Biol. 152, 51–64. Kirisako, T., Ichimura, Y., Okada, H., Kabeya, Y., Mizushima, N., Yoshimori, T., Ohsumi, M., Takao, T., Noda, T., and Ohsumi, Y. (2000). The reversible modification regulates the membrane-binding state of Apg8/Aut7 essential for autophagy and the cytoplasm to vacuole targeting pathway. J. Cell Biol. 151, 263–276. Kirsch, J., and Betz, H. (1995). The postsynaptic localization of the glycine receptor-associated protein gephyrin is regulated by the cytoskeleton. J. Neurosci. 15, 4148–4156. Kirsch, J., Langosch, D., Prior, P., Littauer, U.Z., Schmitt, B., and Betz, H. (1991). The 93-kDa glycine receptor-associated protein binds to tubulin. J. Biol. Chem. 266, 22242–22245. Kirsch, J., Wolters, I., Triller, A., and Betz, H. (1993). Gephyrin antisense oligonucleotides prevent glycine receptor clustering in spinal neurons. Nature 366, 745–748. Kishida, T., Stackhouse, T.M., Chen, F., Lerman, M.I., and Zbar, B. (1995). Cellular proteins that bind the von Hippel-Lindau disease gene product: mapping of binding domains and the effect of missense mutations. Cancer Res. 55, 4544–4548. Kittler, J.T., Rostaing, P., Schiavo, G., Fritschy, J.M., Olsen, R., Triller, A., and Moss, S.J. (2001). The subcellular distribution of gabarap and its ability to interact with NSF suggest a role for this protein in the intracellular transport of gaba(a) receptors. Mol. Cell. Neurosci. 18, 13–25. Klionsky, D.J., and Emr, S.D. (2000). Autophagy as a regulated pathway of cellular degradation. Science 290, 1717–1721. Kneussel, M., Brandstatter, J.H., Laube, B., Stahl, S., Muller, U., and Betz, H. (1999). Loss of postsynaptic GABA(A) receptor clustering in gephyrin-deficient mice. J. Neurosci. 19, 9289–9297. Kneussel, M., Haverkamp, S., Fuhrmann, J.C., Wang, H., Wassle, H., Olsen, R.W., and Betz, H. (2000). The gamma-aminobutyric acid type A receptor (GABAAR)-associated protein GABARAP interacts with gephyrin but is not involved in receptor anchoring at the synapse. Proc. Natl. Acad. Sci. USA 97, 8594–8599. Lang, T., Schaeffeler, E., Bernreuther, D., Bredschneider, M., Wolf, D.H., and Thumm, M. (1998). Aut2p and Aut7p, two novel microtubule-associated proteins are essential for delivery of autophagic vesicles to the vacuole. EMBO J. 17, 3597–3607. Legesse-Miller, A., Sagiv, Y., Porat, A., and Elazar, Z. (1998). Isolation

Neuron 74

and characterization of a novel low molecular weight protein involved in intra-Golgi traffic. J. Biol. Chem. 273, 3105–3109.

acetylcholine receptor cluster formation by rapsyn. Proc. Natl. Acad. Sci. USA 95, 4007–4012.

Littauer, U.Z., Giveon, D., Thierauf, M., Ginzburg, I., and Ponstingl, H. (1986). Common and distinct tubulin binding sites for microtubuleassociated proteins. Proc. Natl. Acad. Sci. USA 83, 7162–7166.

Ramarao, M.K., Bianchetta, M.J., Lanken, J., and Cohen, J.B. (2001). Role of rapsyn tetratricopeptide repeat and coiled-coil domains in self-association and nicotinic acetylcholine receptor clustering. J. Biol. Chem. 276, 7475–7483.

Maimone, M.M., and Enigk, R.E. (1999). The intracellular domain of the nicotinic acetylcholine receptor alpha subunit mediates its coclustering with rapsyn. Mol. Cell. Neurosci. 14, 340–354. Mann, S.S., and Hammarback, J.A. (1994). Molecular characterization of light chain 3. A microtubule binding subunit of MAP1A and MAP1B. J. Biol. Chem. 269, 11492–11497. Marya, P.K., Syed, Z., Fraylich, P.E., and Eagles, P.A. (1994). Kinesin and tau bind to distinct sites on microtubules. J. Cell Sci. 107, 339–344. McDonald, P.H., Cote, N.L., Lin, F.T., Premont, R.T., Pitcher, J.A., and Lefkowitz, R.J. (1999). Identification of NSF as a beta-arrestin1binding protein. Implications for beta2-adrenergic receptor regulation. J. Biol. Chem. 274, 10677–10680. Melki, R., Kerjan, P., Waller, J.P., Carlier, M.F., and Pantaloni, D. (1991). Interaction of microtubule-associated proteins with microtubules: yeast lysyl- and valyl-tRNA synthetases and tau 218-235 synthetic peptide as model systems. Biochemistry 30, 11536–11545. Meyer, G., Kirsch, J., Betz, H., and Langosch, D. (1995). Identification of a gephyrin binding motif on the glycine receptor beta subunit. Neuron 15, 563–572. Moss, S.J., and Smart, T.G. (2001). Constructing inhibitory synapses. Nat. Rev. Neurosci. 2, 240–250. Nassar, N., Horn, G., Herrmann, C., Scherer, A., McCormick, F., and Wittinghofer, A. (1995). The 2.2 A crystal structure of the Ras-binding domain of the serine/threonine kinase c-Raf1 in complex with Rap1A and a GTP analogue. Nature 375, 554–560. Nishimune, A., Isaac, J.T., Molnar, E., Noel, J., Nash, S.R., Tagaya, M., Collingridge, G.L., Nakanishi, S., and Henley, J.M. (1998). NSF binding to GluR2 regulates synaptic transmission. Neuron 21, 87–97. Nogales, E., Wolf, S.G., and Downing, K.H. (1998). Structure of the alpha beta tubulin dimer by electron crystallography. Nature 391, 199–203. Nogales, E., Whittaker, M., Milligan, R.A., and Downing, K.H. (1999). High-resolution model of the microtubule. Cell 96, 79–88. Okazaki, N., Yan, J., Yuasa, S., Ueno, T., Kominami, E., Masuho, Y., Koga, H., and Muramatsu, M. (2000). Interaction of the Unc-51like kinase and microtubule-associated protein light chain 3 related proteins in the brain: possible role of vesicular transport in axonal elongation. Brain Res. Mol. Brain Res. 85, 1–12. Ortells, M.O., and Lunt, G.G. (1995). Evolutionary history of the ligand-gated ion-channel superfamily of receptors. Trends Neurosci. 18, 121–127. Osten, P., Srivastava, S., Inman, G.J., Vilim, F.S., Khatri, L., Lee, L.M., States, B.A., Einheber, S., Milner, T.A., Hanson, P.I., and Ziff, E.B. (1998). The AMPA receptor GluR2 C terminus can mediate a reversible, ATP-dependent interaction with NSF and alpha- and beta-SNAPs. Neuron 21, 99–110. Otwinowski, Z., and Minor, W. (1993). In Data Collection and Processing, L. Sawyer, N. Isaacs, and S. Bailey, eds. (Warrington, UK: SERC Daresbury Laboratory), pp. 556–562. Paz, Y., Elazar, Z., and Fass, D. (2000). Structure of GATE-16, membrane transport modulator and mammalian ortholog of autophagocytosis factor Aut7p. J. Biol. Chem. 275, 25445–25450. Phillips, W.D., Maimone, M.M., and Merlie, J.P. (1991). Mutagenesis of the 43-kD postsynaptic protein defines domains involved in plasma membrane targeting and AChR clustering. J. Cell Biol. 115, 1713–1723. Quirk, K., Gillard, N.P., Ragan, C.I., Whiting, P.J., and McKernan, R.M. (1994). gamma-Aminobutyric acid type A receptors in the rat brain can contain both gamma 2 and gamma 3 subunits, but gamma 1 does not exist in combination with another gamma subunit. Mol. Pharmacol. 45, 1061–1070. Ramarao, M.K., and Cohen, J.B. (1998). Mechanism of nicotinic

Sagiv, Y., Legesse-Miller, A., Porat, A., and Elazar, Z. (2000). GATE-16, a membrane transport modulator, interacts with NSF and the Golgi v-SNARE GOS-28. EMBO J. 19, 1494–1504. Schiff, P.B., Fant, J., and Horwitz, S.B. (1979). Promotion of microtubule assembly in vitro by taxol. Nature 277, 665–667. Schofield, P.R., Darlison, M.G., Fujita, N., Burt, D.R., Stephenson, F.A., Rodriguez, H., Rhee, L.M., Ramachandran, J., Reale, V., Glencorse, T.A., et al. (1987). Sequence and functional expression of the GABA A receptor shows a ligand-gated receptor super-family. Nature 328, 221–227. Serrano, L., de la Torre, J., Maccioni, R.B., and Avila, J. (1984). Involvement of the carboxyl-terminal domain of tubulin in the regulation of its assembly. Proc. Natl. Acad. Sci. USA 81, 5989–5993. Song, I., Kamboj, S., Xia, J., Dong, H., Liao, D., and Huganir, R.L. (1998). Interaction of the N-ethylmaleimide-sensitive factor with AMPA receptors. Neuron 21, 393–400. Stebbins, C.E., Kaelin, W.G., Jr., and Pavletich, N.P. (1999). Structure of the VHL-ElonginC-ElonginB complex: implications for VHL tumor suppressor function. Science 284, 455–461. Tanida, I., Tanida-Miyake, E., Ueno, T., and Kominami, E. (2001). The human homolog of Saccharomyces cerevisiae Apg7p is a Proteinactivating enzyme for multiple substrates including human Apg12p, GATE-16, GABARAP, and MAP-LC3. J. Biol. Chem. 276, 1701–1706. Tretter, V., Ehya, N., Fuchs, K., and Sieghart, W. (1997). Stoichiometry and assembly of a recombinant GABAA receptor subtype. J. Neurosci. 17, 2728–2737. Unwin, N. (1998). The nicotinic acetylcholine receptor of the Torpedo electric ray. J. Struct. Biol. 121, 181–190. Wang, H., and Olsen, R.W. (2000). Binding of the GABA(A) receptorassociated protein (GABARAP) to microtubules and microfilaments suggests involvement of the cytoskeleton in GABARAPGABA(A) receptor interaction. J. Neurochem. 75, 644–655. Wang, H., Bedford, F.K., Brandon, N.J., Moss, S.J., and Olsen, R.W. (1999). GABA(A)-receptor-associated protein links GABA(A) receptors and the cytoskeleton. Nature 397, 69–72. Yoon, Y., and Oakley, B.R. (1995). Purification and characterization of assembly-competent tubulin from Aspergillus nidulans. Biochemistry 34, 6373–6381. Young, L., Jernigan, R.L., and Covell, D.G. (1994). A role for surface hydrophobicity in protein-protein recognition. Protein Sci. 3, 717–729. Zhou, B., Boudreau, N., Coulber, C., Hammarback, J., and Rabinovitch, M. (1997). Microtubule-associated protein 1 light chain 3 is a fibronectin mRNA-binding protein linked to mRNA translation in lamb vascular smooth muscle cells. J. Clin. Invest. 100, 3070–3082.