The Crystal Structure of Cdc42 in Complex with ...

doi:10.1016/j.jmb.2006.03.019

J. Mol. Biol. (2006) 359, 35–46

The Crystal Structure of Cdc42 in Complex with Collybistin II, a Gephyrin-interacting Guanine Nucleotide Exchange Factor Song Xiang1,2,3, Eun Young Kim1,2, Jessica J. Connelly1, Nicolas Nassar3 Joachim Kirsch4, Jan Winking5, Gu¨nter Schwarz5 and Hermann Schindelin1,2,6* 1

Department of Biochemistry SUNY Stony Brook, Stony Brook, NY 11794-5215, USA 2

Center for Structural Biology SUNY Stony Brook, Stony Brook, NY 11794-5215, USA 3

Department of Physiology and Biophysics, SUNY Stony Brook Stony Brook, NY 11794-8661 USA 4

Institute for Anatomy and Cell Biology, University of Heidelberg, Im Neuenheimer Feld 307 69120 Heidelberg, Germany 5

Institute for Plant Biology Technical University of Braunschweig, Spielmannstr. 7 38106 Braunschweig, Germany

The synaptic localization of ion channel receptors is essential for efficient synaptic transmission and the precise regulation of diverse neuronal functions. In the central nervous system, ion channel receptors reside in the postsynaptic membrane where they are juxtaposed to presynaptic terminals. For proper function, these ion channels have to be anchored to the cytoskeleton, and in the case of the inhibitory glycine and g-aminobutyric acid type A (GABAA) receptors this interaction is mediated by a gephyrin centered scaffold. Highlighting its central role in this receptor anchoring scaffold, gephyrin interacts with a number of proteins, including the neurospecific guanine nucleotide exchange factor collybistin. Collybistin belongs to the Dbl family of guanine nucleotide exchange factors, occurs in multiple splice variants, and is specific for Cdc42, a small GTPase ˚ resolution crystal structure of the belonging to the Rho family. The 2.3 A Cdc42–collybistin II complex reveals a novel conformation of the switch I region of Cdc42. It also provides the first direct observation of structural changes in the relative orientation of the Dbl-homology domain and the pleckstrin-homology domain in the same Dbl family protein. Biochemical data indicate that gephyrin negatively regulates collybistin activity. q 2006 Elsevier Ltd. All rights reserved.

6

Rudolf Virchow Center for Experimental Biomedicine and Institute of Structural Biology University of Wu¨rzburg Versbacher Str. 9, 97078 Wu¨rzburg, Germany *Corresponding author

Keywords: Cdc42; collybistin; Dbl-homology (DH) domain; gephyrin; pleckstrin-homology (PH) domain

Introduction Present addresses: G. Schwarz, Institute for Biochemistry, Otto-Fischer-Str. 12-14, University of Cologne, 50674 Cologne, Germany; S. Xiang, Department of Biological Sciences, 701 Fairchild Center, Columbia University, New York, NY 10027, USA. Abbreviations used: BSA, bovine serum albumin; GEF, guanine nucleotide exchange factor; DH, Dbl-homology; PH, pleckstrin-homology; GABAA, g-amino-butyric acid type A. E-mail address of the corresponding author: [email protected] or [email protected]

Activation of Rho GTPases Cdc42, Rac, Rho and their isoforms leads to a variety of cellular processes, including reorganizations of the actin cytoskeleton during cell cycle and motility.1 Rho GTPases are lipid-modified at their C termini, which targets them to the plasma membrane where signaling events occur.2 Small GTPases are activated by guanine nucleotide exchange factors (GEFs), which accelerate the exchange of GDP with GTP. Members of the Dbl family are GEF proteins specific for Rho GTPases and contain a catalytic core

0022-2836/$ - see front matter q 2006 Elsevier Ltd. All rights reserved.

36 composed of tandem Dbl-homology (DH) and pleckstrin-homology (PH) domains. The highly homologous DH domains interact directly with the respective Rho GTPase and constitute the minimal unit required for nucleotide exchange. The PH domains are more variable in their primary sequences and three-dimensional structures, and their role in the exchange reaction is not fully understood.3 Recently, a number of Dbl family structures have been determined,4–10 providing insights into the general exchange reaction mechanism, the molecular basis of selectivity towards different Rho-family GTPases and initial suggestions regarding the functional role of the PH domains. The precise location of synaptic ion channel receptors is essential for proper synapse function. In inhibitory synapses, the protein gephyrin plays a critical role in anchoring glycine and g-aminobutyric acid type A (GABA A ) receptors.11 Gephyrin binds to a cytoplasmic loop of the b subunit of the glycine receptor12–14 and to tubulin.15 In vivo inhibition of gephyrin synthesis prevents clustering of glycine receptors at postsynaptic membranes.16,17 Although no direct interactions between gephyrin and GABAA receptors have been demonstrated, loss of GABAA receptor clustering in gephyrin knockout mice also implicates gephyrin in anchoring these receptors.16,18 Gephyrin has been proposed to form a scaffold underneath the postsynaptic membrane, which links the receptors to the cytoskeleton.11,19 On the basis of the crystal structures of the MogA20 and MoeA21 proteins, which are related to the N-terminal G-domain and C-terminal E-domain, respectively, a model for the gephyrin scaffold has been proposed, in which the G-domain and E-domain of gephyrin form inter-molecular trimers and dimers, respectively.21 Collybistin/hPEM-222 was identified initially as a GEF specific for Cdc42. It was later found to be a gephyrin-binding partner in yeast two-hybrid assays,23 and, subsequently, the gephyrin-binding site on collybistin has been mapped.24 Collybistin is small compared to other Dbl family proteins and is expressed predominantly in the brain. Three alternatively spliced C termini combined with the presence or the absence of an N-terminal SH3 domain generate multiple collybistin isoforms.23,25 The shortest version, collybistin II, has few residues outside the DH and PH domains. Co-transfection experiments demonstrated that collybistin and gephyrin co-localize in vivo and that collybistin II and gephyrin form submembranous clusters in some cells, which recruit glycine receptors.23,25 Since Cdc42 regulates the reorganization of actin filaments and collybistin is an activator of Cdc42, it seems possible that collybistin controls the gephyrin scaffold by regulating proximal components of the cytoskeleton. In order to gain insights into collybistin function, we have determined the crystal ˚ structure of the Cdc42–collybistin II complex at 2.3 A resolution.

Structure of the Cdc42–Collybistin II Complex

Results and Discussion Overall structure Collybistin and Cdc42 were over-expressed separately in Escherichia coli, purified, and functionally characterized by a nucleotide exchange assay (Figure 1). Under these experimental conditions collybistin II is more effective in catalyzing nucleotide exchange on Cdc42 than the collybistin I isoform. The purified Cdc42–collybistin II complex was crystallized in space group P212121 with two copies of the heterodimer in the asymmetric unit. The structure was determined by molecular replacement, using the Rac1–Tiam1 complex structure5 (PDB code 1FOE) as the search model. The PH domain was removed from the search model and was subsequently rebuilt during the later stages of refinement (see Material and Methods). The final structure was refined to an R-factor of 18.2% (RfreeZ ˚ to 2.3 A ˚ 23.2%) using diffraction data from 50 A (Table 1 and Figure 2(a)). The first complex contains residues 1–179 and 184–191 of Cdc42, and residues 37–401 of collybistin II, while the second copy in the asymmetric unit contains residues 1–179 and 185– 191 of Cdc42, and residues 45–397 of collybistin II, totaling 9027 protein atoms plus 627 solvent atoms. The overall architecture of the Cdc42–collybistin II heterodimer is similar to the structures of other Dbl family proteins in complex with Rho family GTPases (Figure 2(b)). The DH domain of collybistin II is a helical bundle composed of six major helices.6 The main body of the PH domain is composed of two anti-parallel b-sheets, capped by an a-helix at the C terminus.26 Two additional

Figure 1. Collybistin activates Cdc42. Mant-GDP loading onto Cdc42 is assayed by the change in its fluorescence following Cdc42 binding. Addition of 200 nM collybistin II (curve a, the arrow indicates addition) increases the exchange rate, compared to control (curve b, same volume of buffer added). Inset: collybistin I is less active than collybistin II in vitro. Curves c and e represent the effects of same amounts of collybistin II and I (200 nM) on Cdc42. In the experiment represented by curve d, collybistin I was added at a concentration of 1.2 mM.

37


Table 1. Data collection and refinement statistics A. Data collection ˚) Resolution (A Completeness Mean redundancy Rsyma hI/sIib B. Refinement ˚) Resolution (A Number of reflections Number of protein/solvent atoms Rcrystc Rfreed Deviations from ideal values in ˚) Bond distances (A Bond angles (deg.) Torsion angles (deg.) ˚ 3) Chiral-center restraints (A ˚) Plane restraints (A ˚) VDW repulsions (A ˚) Potential H bonds (A Ramachandran statisticse Most favored (%) Allowed (%) Generously allowed (%) Disallowed (%) Average B-factors ˚ 2) Main chain (A ˚ 2) Side-chain (A ˚ 2) Solvent atoms (A

50.0–2.3 0.978 (0.998) 5.0 (4.8) 0.081 (0.354) 25.9 (4.3) 50.0–2.3 60,100 9027/627 0.182 (0.192) 0.232 (0.269) 0.019 1.644 6.92 0.120 0.007 0.247 0.190 93.3 5.5 0.9 0.3 43.2 47.1 58.3

Values in parentheses are for the highest resolution shell; 2.38– ˚ for data collection and 2.36–2.3 A ˚ for refinement. 2.3 A P P P P a Rsym Z hkl i jIi KhIij= hkl i Ii where Ii is the ith measurement and hIi is the weighted mean of all measurements of I. b hI/sIi indicates the average of the intensity divided by its standard deviation. P P c RcrystZ kFojKjFck/ jFoj, where Fo and Fc are the observed and calculated structure factor amplitudes. d Rfree is the same as Rcryst but for 5% of the data randomly omitted from refinement. e Ramachandran statistics calculated by PROCHECK.42

domains reveals a large conformational change in the PH domain, which can be described as a 378 rotation (Figure 2(c)). PH domains exhibit considerable variability in their orientations relative to the DH domains among published Dbl family protein structures; in the most dramatic case of Sos1, the PH domain masks the Rho GTPase binding site.4 The Cdc42–collybistin II interface Collybistin II and Cdc42 in both heterodimers form an almost identical set of interactions, except for a conformational difference in the switch I region (see below). Only the DH domain of collybistin interacts with Cdc42, while the PH domain is not involved. The complex buries over ˚ 2 of accessible surface area,† involving 3000 A residues from the DH domain of collybistin II and residues from switch I (residues 26–36), switch II (residues 58–77) and neighboring regions of Cdc42. Two-thirds of the interface area is hydrophobic, but the interface also contains nine hydrogen bonds. The switch I and switch II regions of Cdc42 are altered in this nucleotide-free complex, compared to the GDP-bound Cdc42 structure.27 The conformations of the switch II and the P-loop regions are similar to that described for the Rac1–Tiam1 structure, featuring the intrusion of the side-chain of Ala59 into the Mg2C-binding site and a new hydrogen bond between Lys16 and Glu62,5 and will not be discussed in detail here. The minor difference is that the conserved Lys195 (Lys1195 in Tiam1) in collybistin II does not interact with the switch II region of Cdc42. Switch I interactions and conformations

structural features (a b-turn and a 310 helix) at the N terminus of the PH domain, unique to PH domains of Dbl proteins, are present also in collybistin II. The last a-helix at the C terminus of the PH domain, which folds back to interact with N-terminal regions of the PH domain, is not observed in PH domains of other Dbl family protein structures. The DH domain appears to be the most rigid structural unit, since the two DH domains in the asymmetric unit can be superimposed with a root˚ for all Ca mean-square (rms) deviation of 0.69 A atoms, while the corresponding values for the PH ˚ and 1.18 A ˚, domains and Cdc42 are higher at 0.90 A respectively. The larger rms deviation between the Cdc42 molecules is mostly due to movements of structural elements on the opposite side of the Cdc42–collybistin binding interface (Figure 2(c)), which is likely caused by crystal packing. The DH domains also display lower average temperature ˚ 2 for the Ca atoms of the ˚ 2 and 30.9 A factors (31.9 A two copies in the asymmetric unit) compared to the corresponding numbers for the PH domains ˚ 2) and Cdc42 molecules ˚ 2 and 39.4 A (58.6 A 2 2 ˚ ˚ (37.2 A and 32.7 A ). Superposition of the DH

The conformational change in the switch I region compared to nucleotide-bound Cdc42 is anchored by a highly conserved Glu residue from collybistin II (Glu57), which is hydrogen-bonded to the hydroxyl group of a conserved Tyr at position 32 and the amide nitrogen atoms of Thr35 and Val36 of Cdc42. This interaction causes a movement of switch I towards the base of the nucleotide, moving Phe28 away from the guanine-binding pocket, presumably weakening nucleotide binding. Phe28 is conserved within the Ras superfamily of small GTPases, where it is engaged in a perpendicular aromatic–aromatic interaction with the guanine base in the nucleotide-bound state. In previously determined Rho family GTPase–Dbl protein complex structures, as well as in one Cdc42– collybistin II heterodimer, Phe28 is stabilized by Leu160 of Cdc42 or the equivalent residue in an arrangement resembling the nucleotide-bound state. However, in the other Cdc42–collybistin II complex, the adjacent residue Pro29 undergoes a rearrangement of the main chain, which moves Phe28 completely away from Leu160, opening the † http://www.biochem.ucl.ac.uk/bsm/PP/server/

38


Figure 2. Structure of the Cdc42–collybistin II complex. (a) Stereo view of a representative region of an omit Fo–Fc map covering the switch I region of Cdc42 with Phe28 in the flipped-out orientation (contoured at three times the rms deviation). The map was calculated omitting Cdc42 residues 14–38. Phe28 is highlighted. (b) Overall structure of the Cdc42–collybistin II complex with the DH domain in yellow, the PH domain in green and Cdc42 in cyan. Secondary structural elements were assigned with Promotif,46 and named according to a standardized nomenclature.6 The same coloring scheme is used throughout unless indicated otherwise. The switch I, the switch II regions and the P-loop are highlighted in red. (c) Structural changes of the two complexes in the asymmetric unit. The two complexes are superimposed according to their DH domain. The PH domain and Cdc42 of the second complex are represented in gray. The rotational movement (red/green circle) around the axis (brown) that relates the PH domain in the open (green) and closed (gray) conformations is highlighted. Part (a) was prepared with PyMOL (http://pymol.sourceforge.net/). Parts (b) and (c) and Figures 3, 4(a), 5(b) and 6(b) were generated with MOLSCRIPT47 and Raster3D.48 Figures 4(b) and (c), 5(a)

39


Figure 3. Structural changes in the switch I region of Cdc42. (a) Comparison of the two complexes present in the asymmetric unit. The DH domain is depicted in yellow, Cdc42 from the two complexes in cyan and red and GDP-bound Cdc42 (PDB code 1A4R)27 in gray. Leu132 from a symmetry-related collybistin II, which presumably stabilizes the flipped-out conformation of Phe28, is shown in green. GDP-bound Cdc42 was superimposed onto the complex according to the Ca atoms of residues 1–177, excluding the switch I, switch II and neighboring regions (residues 27–35 ˚ . The Cdc42 molecule from the second complex in the and 59–76, respectively), resulting in an rms deviation of 0.87 A crystal was aligned on the basis of its DH domain. Important residues, as well as the GDP and Mg2C are highlighted. Dotted lines indicate hydrogen bonds. (b) Comparison with the Ras-Sos1 switch I region. Collybistin II and Cdc42 with Phe28 in the flipped-out position are represented as in (a). The switch I region of Ras from the Ras–Sos1 structure is represented in blue, and helix aH of Sos1 in transparent green. Sos1-bound Ras was superimposed on Cdc42 on the basis of the Ca atoms of residues 1–178, excluding regions in the proximity of the switch I region and helix a3d (Cdc42 residues 26–39, 117–135, respectively) and a few other residues (Cdc42 residues 95, 96, 106, 107, 151, and 152), resulting in an rms ˚. deviation of 1.46 A

nucleotide-binding pocket (Figure 3(a)). Although the outward movement of the switch I region is more pronounced in this conformation, no additional interaction between the DH domain and Cdc42 is observed. A Leu residue from a neighboring molecule interacts with Phe28 in the flipped-out conformation, and might help to capture this conformation in the crystal. Both conformations of the switch I region are well defined in the electron density map. A representative portion of an omit map covering the switch I region with Phe28 in the flipped-out orientation is shown in Figure 2(a). The flexibility of the switch I region has been well documented. The flipped-out conformation of Phe28, however, only resembles what has been observed in the Ras–Sos1 structure, in which the insertion of an a-helix from the Sos1 protein causes the displacement of the switch I region of Ras, thus opening the nucleotide-binding pocket.28 In both cases, Phe28 is pointing away from the nucleotidebinding pocket, although in the case of Ras–Sos1 the displacement of the switch I region is more pronounced (Figure 3(b)). The flipped-out conformation of Phe28 appears to facilitate the release of the bound nucleotide and the access of the incoming nucleotide, and it seems that the GEF

proteins induce this conformational change, although by different means, to catalyze the reaction. In the case of the DH domain–Rho GTPases, this conformation might represent an extreme case of the outward movement of the switch I region, induced by a conserved set of interactions between the DH domain and the GTPase, and might exist transiently in solution during the nucleotide exchange reaction. A similar outward movement of the switch I region has been observed in the recently published Cdc42–SopE complex structure, but in this case again the flipped-out conformation of Phe28 is not observed. SopE is a bacterial toxin from Salmonella typhimurium, which activates Cdc42 by means of accelerating its nucleotide exchange. It is not related to the Dbl family proteins, but utilizes a similar set of interactions to cause the outward movement of the switch I region. These interactions involve hydrogen bonding between Tyr32, Thr35 and Val36 on Cdc42 and an Asp residue on SopE.29 Cdc42 specificity Collybistin has been demonstrated to be specific for Cdc42.22 A previous study,7 based on the then available crystal structures of complexes between

and 6(a) were generated with SPOCK (Christopher, J. A. (1998). SPOCK: the structural properties observation and calculation kit. The Center for Macromolecular Design, Texas A&M University, College Station), MOLSCRIPT and Raster3D.

40 Rho GTPases and tandem DH/PH GEFs, proposed a general mechanism for the selectivity of Dbl family proteins towards Rho GTPases. According to this proposal, Phe56 of Cdc42 plays a vital role in Dbl family proteins selecting this GTPase. The corresponding residue on RhoA and Rac1 is a Trp, and the larger side-chain prevents productive binding of these GTPases to Cdc42 specific RhoGEFs. This has been demonstrated by the structure of the Cdc42–intersectin complex, in which Phe56 forms favorable van der Waals contacts with Met1369 and Leu1376. Intersectin is Cdc42-specific, but the substitution L1376I, which presumably enlarges the hydrophobic pocket, renders it active towards Rac1. Furthermore, an Ile residue at the position corresponding to L1376 in intersectin is observed among Rac1-specific RhoGEFs.7 In collybistin, the residues corresponding to Met1369 and Leu1376 of intersectin are Ile180 and Leu187, respectively, and they accommodate Phe56 of


Cdc42 in a fashion similar to that observed in the Cdc42–intersectin structure. In addition, collybistin lacks the features that allow RhoGEFs to interact favorably with RhoA or Rac1. For instance, in the case of RhoA, the negatively charged residues Asp45 and Glu54 form ionic interactions with a positively charged residue from the RhoGEF, such as Lys758 in Dbs.7 The residues in Cdc42 corresponding to Asp45 and Glu54 of RhoA are not charged, and the residue equivalent to Lys758 in Dbs is Asp179 in collybistin. The PH domain The PH domains of Dbl family proteins adopt multiple orientations with respect to the DH domain (Figure 4(a)). Despite the documented conformational variability of the PH domain, the structure presented here is the first example where structural changes in the PH domain of the same Dbl protein are

Figure 4. Structural changes of the PH domain. (a) Structural comparison of the tandem DH/PH domains. Collybistin II, Tiam 1, Dbs, intersectin and Sos1 were aligned according to the conserved regions of their DH domains. In collybistin II, the PH domains are colored green in the open conformation and gray in the closed conformation. (b) Collybistin II in the open conformation. Residues that lose solvent-accessible surface area upon transition to the closed conformation are highlighted with positively charged residues in blue, negatively charged residues in red, polar residues in cyan and nonpolar residues in gray. (c) Collybistin II in the closed conformation (same color code). An additional salt-bridge between Asp136 (red) and Lys379 (blue) in the closed conformation is visible on the bottom of the DH/PH domain interface. (b) and (c) are aligned according to their PH domains. Residues highlighted in (b) with an arrow are disordered in the open conformation.


observed, although other Rho family GTPase–Dbl family protein complexes have been crystallized containing more than one protomer in the asymmetric unit.5–7 In the case of collybistin II, the two PH domains in the asymmetric unit are related by a 378 rotation after superposition of their DH domains and neither PH domain interacts with Cdc42. Here, we refer to these two conformations as open and closed (Figures 2(c), and 4(b) and (c)), depending on whether they have a less or more extensive interface, respectively. In the open conformation (Figure 4(b)), the interactions between the DH and PH domains are mediated through a6 on the DH domain, which interacts with the b-turn N-terminal to b1 and the linker between aC and aC1 of the PH domain. The interface contains seven hydrogen bonds and a total ˚ 2 of surface area is buried. In the closed of 850 A conformation (Figure 4(c)), most of the interactions observed in the open conformation are preserved, but ˚ 2 of surface area is buried and one an additional 430 A additional salt-bridge between Lys379 and Asp136 (Figure 4(b) and (c)) is formed. Both interactions between the DH and PH domains appear to be fairly weak, as judged by the relatively small buried surface area, but the closed conformation seems to be favored energetically, since more surface area is buried. The PH domain in the closed conformation also exhibits reduced mobility in the crystal, as reflected by its ˚2 average Ca atom temperature factor of 39.4 A 2 ˚ compared to 58.6 A for the PH domain in the open conformation. The structures of Cdc42 in complex with collybistin II and other DH–PH tandem exchange factors seem to suggest that the variability of the relative orientations of the DH and PH domains is a common feature among Dbl family proteins; however, the functional implication of this flexibility remains unclear. Membrane targeting Dbl proteins are targeted to the plasma membrane,30–33 as are Rho GTPases, which are lipidmodified near their C termini. PH domains bind to phosphatidylinositols with various specificities and affinities. The lipid-binding sites are formed by loops connecting strands b1/2 and b3/4, as revealed by crystal structures of PH domains in complex with phosphatidylinositol head-groups.26 The electrostatic potential of the Cdc42–collybistin II complex is strongly dipolar, with the putative lipid-binding site sitting on the electropositive side of the complex. This is observed for both the open and the closed conformations; however, for simplicity, here we will discuss only the open conformation (Figure 5). A similar situation has been reported for the Cdc42–Dbs structure, where a model of membrane association was proposed, in which the phosphatidylinositol-binding site on the PH domain and the C terminus of Cdc42 interact with the plasma membrane, and the DH domain is located on the opposite side of the Cdc42/ membrane interface.6 In our model (Figure 5(b)) the C terminus of Cdc42, however, is not in close

41

Figure 5. Membrane interaction model of the Cdc42– collybistin II complex. (a) The electrostatic potential of the open conformation of the Cdc42–collybistin II complex, calculated at zero ionic strength and contoured at 1.5 kT (blue) and K1.5 kT (red). (b) Model of the interaction between the plasma membrane and the complex in the open conformation. The observed conformation of the Cdc42 C terminus is colored in cyan, whereas the physiologically relevant conformations present in the Cdc42–Dbs and Cdc42–GDI complexes are colored in gray and red, respectively.

proximity to the plasma membrane. This is due to a disulfide bond between Cys188 and Cys105 of Cdc42, which is absent in vivo, where Cdc42 is lipid-modified at Cys188. In structures in which this disulfide bond is absent, the C terminus of Cdc42 is highly flexible, as revealed by structural studies of the Cdc42–GDI complex 34 and Cdc42–GEF complexes in which the C188S mutant of Cdc42 was used to prevent formation of the disulfide bridge.6,7 If the C terminus of Cdc42 is modeled on the basis of the conformation observed in these structures, it would interact with the plasma membrane according to this model (Figure 5(b)). It should be noted that the PH domains of Dbl proteins tested so far do not bind to phospholipids with sufficient affinity for membrane targeting, and in many Dbl proteins additional domains or binding partners are responsible for recruiting the protein to the plasma membrane. In the case of collybistin II, gephyrin is involved in its membrane targeting, since collybistin II and gephyrin are cytoplasmic when expressed individually, while the gephyrin–collybistin II complex has a submem-

42 branous location.23 This process is mediated by the PH domain of collybistin II,25 and is inhibited by the additional N-terminal SH3 domain in collybistin I,23 as well as in a recently identified collybistin isoform that is equivalent to collybistin II, but contains an N-terminal SH3 domain.25 The C terminus of collybistin II, which is rich in positively charged residues (Lys401-Gln-Lys-Val-Thr-Gln-Arg-LysTrp-His-Tyr), might also play a role in membrane targeting of the gephyrin–collybistin II complex. In our structure, this C-terminal region is disordered; ˚ away Lys401, the last residue visible, is about 26 A from the membrane based on the model. The missing ten residues and their side-chains can easily span this distance to present the positively charged residues near the C terminus to the negatively charged plasma membrane. It is likely that the positive charges at the collybistin II C terminus provide additional interactions between the gephyrin–collybistin II complex and the plasma membrane, thus facilitating the recruitment of gephyrin to submembranous locations for proper receptor anchoring. Two additional collybistin C termini, generated by alternative splicing have been identified,23,25 but they are not positively charged.


Collybistin–gephyrin interaction By combining interaction and mutagenesis studies, a short peptide stretch on collybistin has been identified by other researchers as the gephyrin-binding site,24 since its removal from collybistin abolishes the affinity for gephyrin. This peptide was proposed to form a loop located N-terminally to the DH domain. In addition, four charged residues in this region were proposed to be essential for the interaction with gephyrin, since a quadruple alanine variant abolished the interaction between the two proteins.24 In collybistin II this putative gephyrin-binding region corresponds to residues 44–57 and the four charged residues to Arg47, Asp48, Arg51 and Glu57. The Cdc42– collybistin II structure presented here indicates clearly that this region forms the N-terminal part of helix a1, and thus is an integral component of the DH domain (Figure 6(a) and (b)). Glu57 interacts with residues in the switch I region of Cdc42 as described above (Figures 3(a) and 6(b)). Arg47 and Asp48 are solvent-exposed, but are facing away from the Cdc42–collybistin interface. In contrast, Arg51 points to the interior of collybistin II and is

Figure 6. Interactions between gephyrin and collybistin. (a) The Cdc42–collybistin II complex, with Cdc42 in ribbon and collybistin II in surface representation. The previously proposed gephyrin-binding site on the DH domain is highlighted in red. The arrangement shown here led to the proposal that the binding sites for Cdc42 and gephyrin overlap. (b) A detailed view of the putative gephyrin-binding site. The Ca trace of the identified binding site is colored in red and Arg47, Asp48, Arg51 and Glu57, are shown with their side-chains and interacting residues. (c) Interaction studies with collybistin mutants. The yeast two-hybrid assays (top panel) were performed using LexA-fused gephyrin with either Gal4 activation domain (GAD)-tagged collybistin I wild-type, the R111A variant, or the R107A/D108A/ E117A triple variant. Yellow and blue indicate negative and positive protein interactions, respectively. In vitro pull-down assays (bottom two panels) with the collybistin II R47A/D48A/E57A triple variant also demonstrate that charged residues, which were previously identified as crucial for gephyrin binding, do not affect the interaction. (d) Gephyrin inhibits collybistin II activity in vitro. Collybistin II activity was assayed by the amount of activated Cdc42, which was pulled down by GST-PAK1BCD and detected by Western blot analysis. Corresponding band intensities were normalized against that of non-inhibited collybistin II. The experiment was repeated three times and the average values and standard deviations are plotted. Increasing amounts of gephyrin (units are mM) in the reaction cause a decrease in the amount of collybistin II-activated Cdc42, in contrast to BSA, which has no effect even when added at high concentration (C is a control lane in which BSA was added at a concentration of 8 mM).

43


completely buried. Its polar guanidinium group is stabilized by multiple interactions with several polar groups nearby; therefore, it is not likely to be involved in interactions with other proteins. To further investigate the contribution of these residues to the interaction with gephyrin, yeast twohybrid studies utilizing wild-type and collybistin I variants were performed (Figure 6(c)). A triple variant (R107A/D108A/E117A) in which the residues corresponding to Arg47, Asp48 and Glu57 of collybistin II were replaced simultaneously with Ala retained the ability to interact with gephyrin. On the other hand, substitution of the residue corresponding to Arg51 with Ala (R111A) abolished the interaction, which is most likely due to structural reasons. The absence of the Arg sidechain presumably leads to structural perturbations due to the absence of the multiple interactions that the guanidinium group of this residue participates in. To confirm the results of the yeast two-hybrid assays, intein pull-down experiments were conducted with wild-type gephyrin and collybistin II variants (R51A and R47A/D48A/E57A). While overall expression levels of both collybistin II variants were similar to that of the wild-type, the R51A variant showed a drastically decreased solubility, which may result from a structural instability. For this reason, only the triple variant could be tested in the pull-down assays and no difference in gephyrin binding compared to wildtype collybistin II was observed, which is in good agreement with the results of the yeast two-hybrid assays utilizing the collybistin I isoform (Figure 6(c)). Our interaction studies combined with the Cdc42– collybistin II structure contradict the biochemical data that identified the gephyrin-binding site on collybistin,24 since it is likely that local structural rearrangements, rather than loss of the binding site, contribute to the loss of affinity of collybistin to gephyrin in these experiments. Further investigations are therefore necessary to unambiguously pinpoint the gephyrin binding site on collybistin. Since the two-hybrid data and the results of pulldown assays were not available to us initially, the following experiment was carried out under the assumption that Cdc42 interacts with collybistin II near the N terminus. Based on this premise we concluded that the gephyrin binding site overlaps with the Cdc42 binding site, at least partially, and hypothesized that gephyrin negatively regulates collybistin activity by preventing Cdc42 from interacting with its DH domain. In order to test this hypothesis, we adopted a sensitive GEF activity assay, in which the amount of collybistin-activated Cdc42 was used as an indicator of collybistin activity. Activated Cdc42 was selectively pulled down by a glutathione-S-transferase (GST)-fused p21-activated kinase 1B (PAK1B) Cdc42/Rac1 interaction/binding domain (CRIB domain), and its amount was measured by Western blot analysis.35 In the control experiment (data not shown), increasing amounts of collybistin II caused an increase in the amount of activated Cdc42 protein,

which demonstrates the activity of collybistin II. When gephyrin was added to the reaction in increasing concentrations, a concentration-dependent inhibition effect on collybistin II activity was observed (Figure 6(d)), with a significant effect already obvious at a gephyrin concentration of 1 mM. Whether this inhibitory effect is the result of a direct competition of Cdc42 and gephyrin for the same binding site on collybistin, or due to an indirect effect such as a conformational change induced by gephyrin binding remains to be addressed in future experiments. According to a recently proposed model,36,37 the inhibitory postsynaptic sites undergo a multi-step maturation process. In the early stage, phosphatidylinositol 3-kinase (PI3K) is activated by an unknown factor released from the presynaptic site. Activation of PI3K results in an increase of the concentration of phosphatidyl inositol 3,4,5triphosphate in the plasma membrane, which recruits profilin and collybistin and subsequently initiates local actin cytoskeleton rearrangements. Gephyrin is then recruited to the postsynaptic site, and induces glycine receptor clustering.11 In line with this scheme, an inhibitory function of gephyrin on collybistin might play a role in terminating certain cellular processes of the Cdc42 signaling cascade, which are important during the early stages of synapse maturation, but are no longer necessary in later stages. One candidate for these cellular processes is the rearrangement of the actin cytoskeleton. It is probably involved only in the initial stages of inhibitory postsynaptic site maturation, since actin-depolymerizing agents failed to disrupt existing gephyrin clusters in highly mature hippocampal cultures.38

Methods Cloning, expression and protein purification Cdc42 was purified as described. 39 The murine collybistin II gene,23 starting from the second methionine residue, the proposed starting codon, was subcloned into the IMPACT system vector pTYB12 (New England Biolabs) and expressed in E. coli strain BL21 (DE3). Residues in collybistin were, however, numbered starting from the first methionine residue to be consistent with data in the literature. Cells in 50 mM Tris–HCl (pH 8.0), 300 mM NaCl, 10% (v/v) glycerol, 5 mM EDTA and a protease inhibitor cocktail (Roche) were lysed with a French pressure cell. The lysate was cleared by centrifugation, mixed with pre-equilibrated (50 mM Tris–HCl (pH 8.0), 1 M NaCl, 1 mM EDTA) chitin agarose beads (New England Biolab) and washed extensively with the same buffer. Following equilibration with cleavage buffer (10 mM Tris–HCl (pH 8.0), 100 mM NaCl, 1 mM EDTA, 50 mM DTT), beads were incubated overnight at room temperature. Collybistin II was eluted with 10 mM Tris– HCl (pH 8.0), 100 mM NaCl, 1 mM EDTA, loaded onto a source Q (Pharmacia) column and chromatographed using a NaCl gradient from buffer A (10 mM Tris–HCl, (pH 8.0), 1 mM EDTA, 5 mM DTT) to buffer B (buffer A plus 1 M NaCl). For purification of the complex,

44 collybistin II and Cdc42 were mixed at a molar ratio of 1:2, dialyzed overnight against gel-filtration buffer (10 mM Tris–H2SO4 (pH 8.0), 250 mM NaCl, 1 mM EDTA, 5 mM DTT), concentrated and loaded onto a Superdex 200 26/60 column (Pharmacia). The resulting complex was concentrated to 10 mg/ml, flash-frozen in liquid nitrogen and stored at K80 8C. Nucleotide exchange assay For biochemical assays, Cdc42 and collybistin II proteins were not further purified after the Ni-NTA and source Q column steps, respectively. Collybistin I was cloned into the IMPACT system vector pTXB1 and purified following the protocol described for collybistin II. Collybistin I and collybistin II were exchanged into buffer containing 20 mM Tris–HCl (pH 8.0) and 300 mM NaCl before the exchange assay, which was performed as described.5 Cdc42 (1 mM) was allowed to equilibrate with 200 nM Mant-GDP (Molecular Probes) for 15 min in reaction buffer (20 mM Tris–HCl (pH 8.0), 300 mM NaCl, 5 mM MgCl2) at room temperature. At the highlighted time-point, collybistin I and II at the indicated concentrations, or equal volumes of buffer were added to the reaction mixture. The exchange assays were performed on an ISSK2 multi-frequency phase fluorimeter, with excitation and emission wavelengths of 360 nm and 440 nm, respectively. Yeast two-hybrid assays The genes encoding rat gephyrin (P2 splice variant) and collybistin I (starting from the second methionine residue) were PCR amplified and subcloned into the XmaI/SalI restriction sites of pBTM116 and pGAD424, respectively. Mutations were introduced into wild-type pGAD424/ collybistin I with the QuikChangew Site-Directed Mutagenesis Kit (Stratagene). The pBTM116/P2 plasmid was cotransformed with either the empty pGAD vector or pGAD424/collybistin I variants into L40 yeast cells, and positive clones were selected by growth on LeuK/TrpK plates. Transformants from different combinations were further grown as separate patches for two to three days and replicated onto a nitrocellulose membrane followed by a b-galactosidase activity assay as described.40 Overexpression of each fusion protein was checked by Western blot analysis probed with either anti-LexA or anti-GAD antibodies. Pull-down assays For GTPase pull-down assays, the human PAK1B CRIB domain (residues 56–272, PAK1BCD) was cloned into pGEX-6P1 (Amersham Bioscience), expressed in BL21 (DE3), and purified by glutathione affinity chromatography. In order to saturate the nucleotide-binding site of Cdc42 with GDP, His6-tagged Cdc42 was incubated in 20 mM Tris–HCl (pH 8.0), 500 mM NaCl, 0.3 mM GDP, 10 mM EDTA overnight, followed by extensive dialysis against reaction buffer (20 mM Tris–HCl (pH 8.0), 300 mM NaCl, 20 mM MgCl2). For collybistin II activity assays, 20 ml of glutathione agarose beads charged with w40 mg of GST-PAK1BCD were mixed with 20 nM GMPPNP (Sigma), 40 nM Cdc42 and various amounts of collybistin II in 5 ml of reaction buffer. After rocking for 10 min at room temperature, the beads were washed three times with reaction buffer, and subjected to a Western blot using an anti-Cdc42 antibody (Cytoskeleton) to detect


bound activated Cdc42. An initial collybistin II activity assay indicated that the amount of activated Cdc42 correlates best with the amount of added collybistin II when a concentration of collybistin II of around 160 nM was employed, which was used in all subsequent experiments. Gephyrin or bovine serum albumin (BSA, negative control) was exchanged into reaction buffer with PD-10 columns (Amersham Bioscience), before adding to the reaction at the desired concentrations. The blots were scanned, and the intensities of corresponding bands were extracted with the program imageJ (NIH). For intein pull-down assays, intein-fused collybistin II wild-type and alanine substitutions were over-expressed as described earlier and immobilized on chitin beads. Then 15 ml of chitin beads with collybistin II wild-type and variants were incubated with 40 mM purified gephyrin13 in a total volume of 30 ml of binding buffer (10 mM Tris–HCl (pH 8.0), 250 mM NaCl) at 4 8C for 1 h. The supernatant was collected after centrifugation at 1250g for 5 min and beads were washed three times with 1 ml of binding buffer followed by SDS-PAGE analysis. For further confirmation, Western blot analysis was performed using an anti-gephyrin antibody. Crystallization, data collection and structure determination Plate-like Cdc42-collybistin II complex crystals were obtained within a few days by the hanging-drop, vapordiffusion method with a reservoir containing 5–9% PEG 8000, 50 mM KH2PO4 and 250 mM NaCl. The crystals ˚ , bZ147.5 A ˚, belong to space group P212121 with aZ58.0 A ˚ with two complexes in the asymmetric unit. cZ167.2 A Crystals were transferred into mother liquor containing increasing concentrations of glycerol up to a final concentration of 30% (v/v) in 5% increments, and were flash-cooled in liquid nitrogen. All diffraction data were collected on beamline X26C at the National Synchrotron Light Source (NSLS) at Brookhaven National Laboratory ˚ on a Quantum 4 ADSC CCD at a wavelength of 1.1 A detector (Table 1). Data were indexed, integrated and scaled using HKL.41 The Rac1–Tiam1 structure was used as a search model, with the side-chains and the PH domain removed before ˚ molecular replacement, which was performed at 3.5 A resolution with Molrep.42 A solution containing two copies of the search model in the asymmetric unit was identified with an R-factor of 55%. This initial model was subjected to rigid body refinement using Refmac,42 and torsion angle dynamics simulated annealing at 5000 K with CNS.43 Afterwards, the program ARP44 was used with the auto-trace option to build the missing PH domains, yielding a rather complete PH domain in one complex and maps improved sufficiently to allow sidechain assignment. The PH domain in the other complex was completed with the aid of the more complete PH domain. The model was refined using alternate cycles of model building with O45 and refinement with Refmac incorporating TLS refinement in the later refinement cycles. Protein Data Bank accession codes The atomic coordinates and structure factors (code 2DFK) have been deposited in the Protein Data Bank, Research Collaboratory for Structural Bioinformatics, Rutgers University, New Brunswick, NJ.


Acknowledgements We thank Dr Nils Schrader (Technical University of Braunschweig, Germany) for valuable discussions and assistance throughout the project, Tanja Otte (Technical University of Braunschweig, Germany) and Liqun Wang (SUNY Stony Brook) for technical assistance, and Dr Rolf Sternglanz for guidance with the yeast two-hybrid experiments. This work was supported by National Institutes of Health grants NS48605 to H.S. and GM28220 to Rolf Sternglanz, and American Heart Asocciation AHA0235522 to N.N. The NSLS in Brookhaven is supported by the Department of Energy and NIH, and beamline X26C is supported, in part, by the State University of New York at Stony Brook and its Research Foundation.

References 1. Hall, A. (1998). Rho GTPases and the actin cytoskeleton. Science, 279, 509–514. 2. Takai, Y., Sasaki, T. & Matozaki, T. (2001). Small GTPbinding proteins. Physiol. Rev. 81, 153–208. 3. Hoffman, G. R. & Cerione, R. A. (2002). Signaling to the Rho GTPases: networking with the DH domain. FEBS Letters, 513, 85–91. 4. Soisson, S. M., Nimnual, A. S., Uy, M., Bar-Sagi, D. & Kuriyan, J. (1998). Crystal structure of the Dbl and pleckstrin homology domains from the human Son of sevenless protein. Cell, 95, 259–268. 5. Worthylake, D. K., Rossman, K. L. & Sondek, J. (2000). Crystal structure of Rac1 in complex with the guanine nucleotide exchange region of Tiam1. Nature, 408, 682–688. 6. Rossman, K. L., Worthylake, D. K., Snyder, J. T., Siderovski, D. P., Campbell, S. L. & Sondek, J. (2002). A crystallographic view of interactions between Dbs and Cdc42: PH domain-assisted guanine nucleotide exchange. EMBO J. 21, 1315–1326. 7. Snyder, J. T., Worthylake, D. K., Rossman, K. L., Betts, L., Pruitt, W. M., Siderovski, D. P. et al. (2002). Structural basis for the selective activation of Rho GTPases by Dbl exchange factors. Nature Struct. Biol. 9, 468–475. 8. Worthylake, D. K., Rossman, K. L. & Sondek, J. (2004). Crystal structure of the DH/PH fragment of Dbs without bound GTPase. Structure (Camb.), 12, 1078–1086. 9. Kristelly, R., Gao, G. & Tesmer, J. J. (2004). Structural determinants of RhoA binding and nucleotide exchange in leukemia-associated Rho guaninenucleotide exchange factor. J. Biol. Chem. 279, 47352–47362. 10. Derewenda, U., Oleksy, A., Stevenson, A. S., Korczynska, J., Dauter, Z., Somlyo, A. P. et al. (2004). The crystal structure of RhoA in complex with the DH/PH fragment of PDZRhoGEF, an activator of the Ca2C sensitization pathway in smooth muscle. Structure (Camb.), 12, 1955–1965. 11. Kneussel, M. & Betz, H. (2000). Clustering of inhibitory neurotransmitter receptors at developing postsynaptic sites: the membrane activation model. Trends Neurosci. 23, 429–435.

45 12. Meyer, G., Kirsch, J., Betz, H. & Langosch, D. (1995). Identification of a gephyrin binding motif on the glycine receptor beta subunit. Neuron, 15, 563–572. 13. Schrader, N., Kim, E. Y., Winking, J., Paulukat, J., Schindelin, H. & Schwarz, G. (2004). Biochemical characterization of the high affinity binding between the glycine receptor and gephyrin. J. Biol. Chem. 279, 18733–18741. 14. Kim, E. Y., Schrader, N., Smolinsky, B., Bedet, C., Vannier, C., Schwarz, G. & Schindelin, H. (2006). Deciphering the structural framework of glycine receptor anchoring by gephyrin. EMBO J. 25, 1385–1395. 15. Kirsch, J., Langosch, D., Prior, P., Littauer, U. Z., Schmitt, B. & Betz, H. (1991). The 93-kDa glycine receptor-associated protein binds to tubulin. J. Biol. Chem. 266, 22242–22245. 16. Fischer, F., Kneussel, M., Tintrup, H., Haverkamp, S., Rauen, T., Betz, H. & Wassle, H. (2000). Reduced synaptic clustering of GABA and glycine receptors in the retina of the gephyrin null mutant mouse. J. Comp. Neurol. 427, 634–648. 17. Feng, G., Tintrup, H., Kirsch, J., Nichol, M. C., Kuhse, J., Betz, H. & Sanes, J. R. (1998). Dual requirement for gephyrin in glycine receptor clustering and molybdoenzyme activity. Science, 282, 1321–1324. 18. Kneussel, M., Brandstatter, J. H., Laube, B., Stahl, S., Muller, U. & Betz, H. (1999). Loss of postsynaptic GABA(A) receptor clustering in gephyrin-deficient mice. J. Neurosci. 19, 9289–9297. 19. Moss, S. J. & Smart, T. G. (2001). Constructing inhibitory synapses. Nature Rev. Neurosci. 2, 240–250. 20. Liu, M. T., Wuebbens, M. M., Rajagopalan, K. V. & Schindelin, H. (2000). Crystal structure of the gephyrin-related molybdenum cofactor biosynthesis protein MogA from Escherichia coli. J. Biol. Chem. 275, 1814–1822. 21. Xiang, S., Nichols, J., Rajagopalan, K. V. & Schindelin, H. (2001). The crystal structure of Escherichia coli MoeA and its relationship to the multifunctional protein gephyrin. Structure (Camb.), 9, 299–310. 22. Reid, T., Bathoorn, A., Ahmadian, M. R. & Collard, J. G. (1999). Identification and characterization of hPEM-2, a guanine nucleotide exchange factor specific for Cdc42. J. Biol. Chem. 274, 33587–33593. 23. Kins, S., Betz, H. & Kirsch, J. (2000). Collybistin, a newly identified brain-specific GEF, induces submembrane clustering of gephyrin. Nature Neurosci. 3, 22–29. 24. Grosskreutz, Y., Hermann, A., Kins, S., Fuhrmann, J. C., Betz, H. & Kneussel, M. (2001). Identification of a gephyrin-binding motif in the GDP/GTP exchange factor collybistin. Biol. Chem. 382, 1455–1462. 25. Harvey, K., Duguid, I. C., Alldred, M. J., Beatty, S. E., Ward, H., Keep, N. H. et al. (2004). The GDP–GTP exchange factor collybistin: an essential determinant of neuronal gephyrin clustering. J. Neurosci. 24, 5816–5826. 26. Ferguson, K. M., Lemmon, M. A., Schlessinger, J. & Sigler, P. B. (1995). Structure of the high affinity complex of inositol trisphosphate with a phospholipase C pleckstrin homology domain. Cell, 83, 1037–1046. 27. Rudolph, M. G., Wittinghofer, A. & Vetter, I. R. (1999). Nucleotide binding to the G12V-mutant of Cdc42 investigated by X-ray diffraction and fluorescence spectroscopy: two different nucleotide states in one crystal. Protein Sci. 8, 778–787.

46


28. Boriack-Sjodin, P. A., Margarit, S. M., Bar-Sagi, D. & Kuriyan, J. (1998). The structural basis of the activation of Ras by Sos. Nature, 394, 337–343. 29. Buchwald, G., Friebel, A., Galan, J. E., Hardt, W. D., Wittinghofer, A. & Scheffzek, K. (2002). Structural basis for the reversible activation of a Rho protein by the bacterial toxin SopE. EMBO J. 21, 3286–3295. 30. Rossman, K. L., Cheng, L., Mahon, G. M., Rojas, R. J., Snyder, J. T., Whitehead, I. P. & Sondek, J. (2003). Multifunctional roles for the PH domain of Dbs in regulating Rho GTPase activation. J. Biol. Chem. 278, 18393–18400. 31. Pruitt, W. M., Karnoub, A. E., Rakauskas, A. C., Guipponi, M., Antonarakis, S. E., Kurakin, A. et al. (2003). Role of the pleckstrin homology domain in intersectin-L Dbl homology domain activation of Cdc42 and signaling. Biochim. Biophys. Acta, 1640, 61–68. 32. Fleming, I. N., Gray, A. & Downes, C. P. (2000). Regulation of the Rac1-specific exchange factor Tiam1 involves both phosphoinositide 3-kinase-dependent and -independent components. Biochem. J. 351, 173–182. 33. Fuentes, E. J., Karnoub, A. E., Booden, M. A., Der, C. J. & Campbell, S. L. (2003). Critical role of the pleckstrin homology domain in Dbs signaling and growth regulation. J. Biol. Chem. 278, 21188–21196. 34. Hoffman, G. R., Nassar, N. & Cerione, R. A. (2000). Structure of the Rho family GTP-binding protein Cdc42 in complex with the multifunctional regulator RhoGDI. Cell, 100, 345–356. 35. Sander, E. E., van Delft, S., ten Klooster, J. P., Reid, T., van der Kammen, R. A., Michiels, F. & Collard, J. G. (1998). Matrix-dependent Tiam1/Rac signaling in epithelial cells promotes either cell–cell adhesion or cell migration and is regulated by phosphatidylinositol 3-kinase. J. Cell. Biol. 143, 1385–1398. 36. Kneussel, M., Engelkamp, D. & Betz, H. (2001). Distribution of transcripts for the brain-specific GDP/GTP exchange factor collybistin in the developing mouse brain. Eur. J. Neurosci. 13, 487–492. 37. Kuriyama, K., Ohno, K., Shu, F., Ueki, T. & Sato, K. (2001). Developmental changes in gephyrin and collybistin mRNA expressions in the rat olfactory bulb. Brain Res. Dev. Brain Res. 131, 67–72.

38. Allison, D. W., Chervin, A. S., Gelfand, V. I. & Craig, A. M. (2000). Postsynaptic scaffolds of excitatory and inhibitory synapses in hippocampal neurons: maintenance of core components independent of actin filaments and microtubules. J. Neurosci. 20, 4545–4554. 39. Hoffman, G. R., Nassar, N., Oswald, R. E. & Cerione, R. A. (1998). Fluoride activation of the Rho family GTP-binding protein Cdc42Hs. J. Biol. Chem. 273, 4392–4399. 40. Chien, C. T., Bartel, P. L., Sternglanz, R. & Fields, S. (1991). The two-hybrid system: a method to identify and clone genes for proteins that interact with a protein of interest. Proc. Natl Acad. Sci. USA, 88, 9578–9582. 41. Otwinowski, Z. & Minor, W. (1997). Processing of X-ray diffraction data collected in oscillation mode. Methods Enzymol. 276, 307–326. 42. Collaborative Computing Project, Number 4. (1994). The CCP4 suite—programs for protein crystallography. Acta Crystallog. sect. D, 50, 760–763. 43. Bru¨nger, A. T. & Rice, L. M. (1997). Crystallographic refinement by simulated annealing: methods and applications. Methods Enzymol. 277, 243–269. 44. Perrakis, A., Morris, R. & Lamzin, V. S. (1999). Automated protein model building combined with iterative structure refinement. Nature Struct. Biol. 6, 458–463. 45. Jones, T. A., Zou, J. Y., Cowan, S. W. & Kjeldgaard, M. (1991). Improved methods for building protein models in electron density maps and the location of errors in these models. Acta Crystallog. sect. A, 47, 110–119. 46. Hutchinson, E. G. & Thornton, J. M. (1994). A revised set of potentials for beta turn formation in proteins. Protein Sci. 3, 2207–2216. 47. Kraulis, P. J. (1991). MOLSCRIPT—a program to produce both detailed and schematic plots of protein structures. J. Appl. Crystallog. 24, 946–950. 48. Merritt, E. A. & Bacon, D. J. (1997). Raster3D: photorealistic molecular graphics. Methods Enzymol. 277, 505–524.

Edited by R. Huber (Received 9 November 2005; received in revised form 6 March 2006; accepted 9 March 2006) Available online 29 March 2006