The Structure of the Mammalian Signal Recognition Particle (SRP) Receptor as Prototype for the Interaction of Small GTPases with Longin Domains*

The eukaryotic signal recognition particle (SRP) and its receptor (SR) play a central role in co-translational targeting of secretory and membrane proteins to the endoplasmic reticulum. The SR is a heterodimeric complex assembled by the two GTPases SRα and SRβ, which is membrane-anchored. Here we present the 2.45-Å structure of mammalian SRβ in its Mg2+ GTP-bound state in complex with the minimal binding domain of SRα termed SRX. SRβ is a member of the Ras-GTPase superfamily closely related to Arf and Sar1, while SRX belongs to the SNARE-like superfamily with a fold also known as longin domain. SRX binds to the P loop and the switch regions of SRβ-GTP. The binding mode and structural similarity with other GTPase-effector complexes suggests a co-GAP (GTPase-activating protein) function for SRX. Comparison with the homologous yeast structure and other longin domains reveals a conserved adjustable hydrophobic surface within SRX which is of central importance for the SRβ-GTP:SRX interface. A helix swap in SRX results in the formation of a dimer in the crystal structure. Based on structural conservation we present the SRβ-GTP:SRX structure as a prototype for conserved interactions in a variety of GTPase regulated targeting events occurring at endomembranes.

the stable NG domain, which binds to the respective NG domain of SRP (SRP54, Ffh in eubacteria). Several structures of the isolated NG domains revealed the basis for the SRP GTPase cycle (for reviews, see Refs. 2-4 and 9), and the complex of the two NG domains shows a remarkably symmetric heterodimer with the nucleotides in direct contact at the center of the interface (10,11). The N-terminal part of SR␣ is responsible for tethering SR␣ to the ER membrane bound SR␤ (12). The interaction localizes to the globular SRX domain of SR␣ comprising the first 130 residues (13). SRX has been described as effector for SR␤ and only binds to the GTP-bound form of the GTPase (13,14). The SRX domain belongs to the SNARE-like superfamily including the N-terminal domains of non-syntaxin SNAREs, also known as longin domains (13,15). Longin domains have been proposed to regulate a variety of membrane trafficking processes (16). Members of this superfamily with known three-dimensional structures include the SNAREs Sec22b (17) and Ykt6 (18), the component SEDL of the transport protein particle (TRAPP) (19), and the clathrin adaptor proteins AP-and AP-N (20,21).
SR␤ is a classical small Ras-GTPase most similar to Arf (ADP-ribosylation factor) and Sar1 (secretion-associated and Ras-related 1) proteins with an accordingly low K D of ϳ30 nM for GTP (6,22). Phylogenetically, SR␤ together with Arf and Sar1 separated from other small Ras-GTPases already in the earliest branching event indicating the functional importance of an ancestral SR␤ in eukaryotic evolution (23). Typical for small GTPases, SR␤ is characterized by five conserved sequence elements (G1-G5), which form the nucleotide-binding site (24 -26). The so-called switch I (residues 85 B to 96 B in SR␤; to distinguish between SR␤ and SRX we use the subscript letters B and X, respectively), interswitch (97 B to 115 B ) and switch II (116 B to 130 B ) regions are known to change most in conformation during a GTPase cycle (26). In the switch II region of SR␤, the position of a critical catalytic residue (Gln-61 in Ras, Gln-71 in Arf) is occupied by a histidine (His-119 B ), which is conserved among the SR␤ and in the Sar1 family (His-79). A special feature of SR␤ is its predicted membrane spanning helix, which is unique but dispensable for SR function (27). In comparison, proteins of the Arf and Sar1 family have an extra N-terminal helix that becomes either myristoylated in Arf (28) or is preceded by an N-terminal hydrophobic patch in Sar1 (29). The GTPases are anchored in the GTP-bound state to their target membrane.
SR␤ does not hydrolyze GTP (30,31) and also not in complex with SR␣ or SRX alone (Refs. 13 and 22 and this study). Like for other small Ras-GTPases, a GTPase-activating protein (GAP) and a guanine nucleotide exchange factor (GEF), which stimulate the low intrinsic GTPase activity and the release of GDP, respectively, are necessary to drive the GTPase cycle (26,32). RNCs interact with SR␤ in its GTP-bound state (22). GTP binding to SR␤ is stimulated by the translocon and is required to induce signal sequence release from SRP (33). The GAP function for trypsin-digested SR heterodimers that retain SR␤ and the N-terminal fragment of SR␣ including SRX has been attributed to the RNC complex (22). Interestingly, a GAP function of the RNC for the isolated SR␤ could not be found (30,31). In the yeast system, the GEF activity for SR␤ has been assigned to the two orthologues (Sbh1p, Sbh2p) of the Sec61␤ subunit of the translocon (34) and point mutations in the cytoplasmic loops of the yeast translocon severely affect the co-translational translocation pathway (35). However, the molecular details for the initiation of GTP hydrolysis and the release of GDP in SR␤ remain so far unclear.
Here we describe the crystal structure of the mammalian SR␤-GTP: SRX receptor complex. The comparison of the structure with the yeast homolog provides detailed insights into the family of the eukaryotic SR. The homology to other small GTPases and the underlying principles of regulation together with previous biochemical data allows attributing a co-GAP function to SRX for SR␤-GTP. We analyze the fundamentals of the longin domain family and suggest the interaction of small GTPases and longin domains to be important for targeting of large complexes or vesicles to the endomembrane system.

EXPERIMENTAL PROCEDURES
Protein Expression and Purification-The His 6 -tagged N-terminal 176 amino acids from human SR␣ (including SRX, residues 1-130) together with mouse SR␤ lacking the N-terminal transmembrane region (here referred to as SR␤, residues 58 -269) were expressed as a bi-cistronic construct from vector pET16b (Stratagene) in BL21(DE3) Escherichia coli cells (Stratagene). The protein was purified via affinity tag purification using Ni 2ϩ -loaded chelating Sepharose Fast Flow beads (Amersham Biosciences). The protein was further purified via ion exchange chromatography (Q-and SP-Sepharose) and finally via size exclusion chromatography (Superdex 200, Amersham Biosciences) using a low salt buffer (10 mM Tris/HCl pH 8.0, 150 mM NaCl, 5 mM MgCl 2 , and 1 mM dithiothreitol). The same buffer was also used for crystallization trials.
Crystallization and Data Collection-Initial crystallization conditions were obtained from hanging drop vapor diffusion at 20°C using the WIZARD sparse matrix screen (Emerald BioStructures) and a protein concentration of 12 mg/ml by mixing the protein in a 1:1 ratio with reservoir. Crystals grew within 4 -6 weeks over a reservoir containing 100 mM sodium citrate, pH 5.5, 2.0 M (NH 4 ) 2 SO 4 , and in the presence of 100 mM guanidinium chloride. The leaf-shaped crystal plates belong to space group I222 and diffract to 2.45 Å. The asymmetric unit contains one SR␤-GTP:SRX heterodimer corresponding to a solvent content of 55% and a Matthews coefficient of 2.7 Å 3 /Da. Crystals were flash-frozen in liquid nitrogen using 20% (v/v) glycerol as cryo-protectant. Diffraction data were measured at the European Synchroton Radiation Facility in Grenoble, France at beamline ID14 -4. Data were collected at 100 K  and a wavelength of 0.979 Å and processed using the HKL program package (36). Data statistics are listed in Table 1.
Structure Determination and Analysis-The structure was determined by molecular replacement using the ␤-subunit of the yeast SR␤-GTP:SRX complex (Protein Data Bank code: 1NRJ) as a search model in AMoRe (37). Model building was done with program O (38). The model including 104 water molecules was refined at 2.45 Å resolution using the CNS package (39) to an R-factor of 19.3% and an R-free factor of 23.2% (Table 1). Residues 41 X -47 X , 131 X -176 X , 208 B -219 B , and 248 B -254 B are not ordered and therefore missing in the model. All structural figures were created with program PyMOL. Program DALI (40) was used to search the Protein Data Bank for proteins similar to SR␤ and the angle between the two ␣1 X helices within the SR␤-GTP:SRX "dimer" was determined with program DYNDOM (41).

RESULTS AND DISCUSSION
Overall Structure-The overall structure of the refined SR␤-GTP: SRX model is depicted in Fig. 1A. SR␤ is a typical small GTPase and features a classical Rossmann fold with a central six-stranded (␤1 B -␤6 B ) mixed ␤-sheet packed in between five helices. SR␤ reveals highest similarity to the GTP-bound structures of Sar1 in complex with Sec23/ Sec24 (43) (23). Besides the N-terminal membrane anchoring regions, the most striking structural difference between SR␤ and Arf or Sar1 is an insertion between helix ␣4 B and strand ␤6 B (37 residues compared with Sar1). Helix ␣4 B is extended by two turns and protrudes from the protein core as described earlier (13). The insertion is partially disordered and no particular function has been attributed to it so far. In the SR␤-GTP:SRX complex SR␤ is in a state not competent for GTP hydrolysis as the catalytic histidine residue (His-119 B ) is pointing away from the active site (see below).
The SRX domain (Fig. 1, A and B) belongs to the mixed ␣/␤ class proteins sharing topology (␤␤␣␤␤␤␣␣) and fold of the SNARE-like protein superfamily including the N-terminal domains of non-syntaxin SNAREs (longin domains). The fold is defined by a three-layer architecture with a central five-stranded antiparallel ␤-sheet packed against helix ␣1 X on the concave side of the ␤-sheet and two C-terminal anti-FIGURE 2. The SR␤-GTP:SRX dimer. The dimer formed by the SRX domains of two SR␤-GTP:SRX complexes as observed in the crystal structure reveals a domain swap of helix ␣1 X (rotation is indicated) and the formation of a trans ␤-sheet between the merged ␤3 x -␤4 x strands. The 2-fold axis between two monomers is highlighted in cyan. parallel helices ␣2 X and ␣3 X on the other side (secondary structure numbering is according to the Structural Classification of Proteins (SCOP) nomenclature, which is different to the nomenclature used for the yeast structure). At the N terminus the two anti-parallel ␤-strands ␤1 X -␤2 X are connected by a conserved ␤-hairpin. Helix ␣1 X locates almost perpendicular to the ␤-strands on the concave side and connects the peripheral ␤-strands of the ␤-sheet (␤2 X and ␤3 X ). The helix flanking loop regions are not conserved and only partially visible in the structure. Strands ␤3 X , ␤4 X , and ␤5 X are connected by short ␤-hairpin structures. The central strand ␤5 X is followed by the long helix ␣2 X , the ␣2 X -␣3 X loop in the plane of the ␤-sheet, and the C-terminal helix ␣3 X running anti-parallel to helix ␣2 X . Helix ␣2 X is kinked and wraps around the convex side of the ␤-sheet like a clamp and helix ␣L X is inserted in the ␣2 X -␣3 X loop.
The mammalian SR␤-GTP:SRX complex forms a crystallographic dimer due to an interaction of the SRX domains involving a domain swap of helix ␣1 X (ϳ50°rotation around helical N terminus) and the formation of a continuous trans ␤-sheet (Fig. 2). Here, strands ␤3 X -␤4 X of one "monomer" merge and align anti-parallel across the dimer interface. Dimerization leads to an additional buried interface of ϳ1000 Å 2 between the two SR␤-GTP:SRX monomers. To analyze the oligomerization state of mammalian SR␤-GTP:SRX in solution, we performed a sedimentation equilibrium experiment by analytical ultracentrifugation and determined a K D of 270 M for the dimer (data not shown). We cannot directly conclude from this result in solution to the state of the complete SR complex at the membrane, since full-length SR␣␤ is anchored to the membrane in vivo. The SR␤-GTP:SR␣ complex (without the transmembrane region) showed a tendency for aggregation and the K D could not be determined. Therefore, the physiological relevance for the dimerization of SR␤-GTP:SRX is not clear. The dimer might be as well enforced by crystal packing. The crystal symmetry favors the domain swap of the flexibly linked helix ␣1 (see below) due to steric hindrance. The simultaneous formation of the trans ␤-sheet stabilizes oligomerization by main chain hydrogen bonding. Interestingly, the comparison with the yeast structure (monomer, see below) showed that the domain-swapped helix ␣1 X of the second SRX molecule of the mammalian receptor superimposes with its corresponding position in the yeast monomer. Therefore, a "monomeric" mammalian receptor complex is used for further analysis.
The SR␤-GTP:SRX Interface-The SR␤-GTP:SRX interface involves the predominant effector-binding region of Ras-like GTPases (45) (Figs. 1 and 3). The buried surface between SR␤-GTP and SRX is 1850 Å 2 , which is similar to the yeast structure and other GTPase-effector complexes (13). SR␤ contributes to the interface with its G1 element (P loop, GLCDSGKT), switch I, interswitch, and switch II regions. The complete switch I region snugly binds into a hydrophobic groove of SRX and spans the whole interface. This groove is situated between the amphipathic helix ␣1 X and the hydrophobic concave surface of the SRX ␤-sheet. Although the protein interface forms a continuous surface, three regions of SRX organized in three layers contribute to the interface (Fig. 3): (i) helix ␣1 X , (ii) the ␤-hairpin between strands ␤1 X and ␤2 X , and (iii) the ␣2 X -␣3 X loop including the short helix ␣L X .
In the top layer, the amphipathic helix ␣1 X binds the switch I and II regions and the P loop of SR␤. The side chain of the conserved Asn-30 X forms hydrogen bonds to the side chain Gln-91 B (not conserved) and the main chain of Thr-92 B in switch I. One helical turn further, Arg-34 X forms a salt bridge to Asp-72 B in the P loop bridging the active site and forming a hydrogen bond to the side chain of Thr-92 B . Three residues of helix ␣1 X (Ile-33 X , Leu-37 X , and Leu-38 X ) are part of a hydrophobic pocket, which accommodates Ile-94 B and the aliphatic part of Gln-91 B in the center of the interface. Leu-38 X forms an additional hydrophobic interaction with Leu-122 B of switch II.
In the central layer, SRX exclusively interacts with the switch I region of SR␤. The central hydrophobic pocket is completed by Val-14 X and the aliphatic part of Lys-10 X . The conserved ␤-hairpin between ␤1 X and ␤2 X contributes a number of hydrophilic interactions, which are surrounded by a hydrophobic rim. All hydrophilic interactions are established by main chain atoms of the ␤-hairpin, which contains a conserved glycine (Gly-12 X ) at the tip. The carbonyl oxygen of Lys-10 X forms a hydrogen bond to the amide nitrogen of Ile-94 B . The carbonyl oxygen of Gly-11 X approaches the Mg 2ϩ binding site in SR␤ and forms a hydrogen bond with the side chain of Ser-93 B , which is essential for Mg 2ϩ coordination. Residues Gly-12 X to Val-14 X form a short stretch of an antiparallel trans ␤-sheet with residues Gln-91 B to Asp-89 B of switch I.
In the third layer, SRX binds to the switch I and interswitch regions of SR␤. Interactions are formed by the ␣2 X -␣3 X loop including the short helix ␣L X . Three hydrophobic side chains (Ala-103 X , Leu-104 X , and Leu-107 X ) from helix ␣L X interact with residues Phe-79 B , Val-80 B , Leu-83 B , and the hydrophobic methyl group of Thr-84 B from switch I as well as with Ala-99 B and Ile-100 B from the interswitch region. Hydrophilic interactions are established by main chain atoms of Ala-103 X and Leu- Comparison with SR␤-GTP:SRX from Yeast-The structures of mammalian and yeast SR␤-GTP (13) are conserved (r.m.s.d. of 1.16 Å over 158 C␣-positions, yeast is distinguished in the following by a "y" subscript). Differences include the lengths of the ␤-strands ␤2 B and ␤3 B that are almost twice as long in mammals and helix ␣4 B that is two turns shorter (Fig. 4A).
In contrast, there are significant differences in SRX (r.m.s.d. of 1.81 Å over 75 C␣-positions) (Fig. 4A). In the yeast structure there is no helix swap leading to a SR␤-GTP:SRX dimer. Instead, the central ␤-sheet of SRX y is extended by one strand (␤3 Xy ) between helix ␣1 Xy and strand ␤4 Xy (␤4 Xy corresponds to strand ␤3 X in our structure), which apparently stabilizes the position of helix ␣1 X and thereby prevents dimer formation. While helix ␣1 X and the ␤1 X -␤2 X hairpin in the interface superimpose very well, the central ␤-sheet and the connected helices ␣2 X and ␣3 X do not. Differences increase with distance from the SR␤-GTP:SRX interface.
Yeast SRX shows two major insertions (Fig. 4A). A 20-residue insertion elongates the central ␤-sheet by introducing the sixth ␤-strand (␤3 Xy : Glu-46 Xy to Ala-49 Xy ) and a loop touching helix ␣2 Xy on the convex side of the ␤-sheet. A 15-residue insertion changes the conformation of the loop between helices ␣2 X and ␣3 X and the herein inserted helix ␣L X is not present. The C-terminal helices (␣4 Xy and ␣3 X ) do not align, which might be due to a truncation of this helix in the yeast structure. With its insertions, yeast SRX appears unusual compared with other SRX domains.
The observed structural differences between mammalian and yeast SRX are reflected by the low degree of conservation on the sequence level (14.2% identity, Fig. 5). Low sequence conservation is a general feature of the SRX family (13). One functionally important exception is the conserved Gly-12 X in the ␤1 X -␤2 X hairpin (Figs. 3 and 5). It facilitates the ␤-hairpin turn and a bulky side chain would sterically interfere with binding of SR␤. Position and amphipathic character of the important helix ␣1 X are conserved. Asn-30 X is conserved between human and yeast and interacts with SR␤ by hydrogen bonding to the switch I region. A polar residue one turn further appears to occupy a crucial position within helix ␣1 X . Arg-34 X forms a salt bridge with Asp-72 B in the P loop and thereby influences the position of the catalytic histidine (His-119 B ) with respect to the active site of SR␤ (Fig. 4b). Although this salt bridge is not conserved, a polar interaction is observed in the yeast structure between Ser-35 Xy and Gln-47 By within the P loop, suggesting a similar role.
SRX as Effector for SR␤-SRX occupies large parts of a typical GAP binding site (45) as it interacts with the P loop and the switch regions of SR␤-GTP resulting in the stabilization of switch II. However, in the SR␤-GTP:SRX complex the catalytic histidine (His-119 B ) in switch II of SR␤ (Gln-61 in Ras, Gln-71 in Arf) is in a "resting" position pointing away from the active site (Fig. 4b), the characteristic arginine finger of a GAP (46) is not present, and the complex is stable when bound to GTP. Therefore, the SR␤-GTP:SRX complex is not a GTPase:GAP complex, and for the stimulation of GTP hydrolysis an additional binding partner is needed. The RNC has been shown to stimulate GTP hydrolysis of SR␤-GTP:SRX (22). However, the RNC does not act as a GAP for SR␤-GTP alone (30). Therefore, the SRX domain can be assigned as co-GAP , and the SNARE hVamp7 are not known. Sequence numbering and secondary structure assignment are shown for human SRX above the aligned sequences (␤-strands in green, ␣-helices in orange). The secondary structure is indicated in all sequences. The conserved glycine in the ␤1 X -␤2 X loop is marked in red. The residue causing a conserved anomaly in strand ␤2 X is indicated in green, and the critical polar position in ␣1 X hydrogen bonding to the P loop in SR␤ is highlighted in blue.
for SR␤ which fulfils one part of the GAP function by stabilizing switch II. Examples for a split GAP function have been reported before. The GAP for the ␣-subunit of a heterotrimeric G protein (G i␣1 ) also stabilizes the switch regions, but the arginine finger is supplied in cis by an additional domain of the GTPase (47). A unique feature of the Arf1: ArfGAP1 structure is the exclusive stabilization of the switch II region (48). The switch I region is recognized by the heptameric coat protein complex (COPI) (49), which is found to stimulate GTP hydrolysis (48). Most likely an arginine finger is needed to trigger GTP hydrolysis in Arf1 (48), which might be the case as well in SR␤.
The co-GAP function can be envisaged by a comparison of SR␤-GTP: SRX with the structure of the Ras-GDP-AlF 3 :RasGAP transition-state complex (50). When SR␤ is superimposed with Ras, the loop of RasGAP containing the arginine finger (Arg-789 RasGAP ) fits between SR␤ and SRX (Fig. 6A). The only sterical clash concerns the arginine finger itself, which would interfere with the salt bridge between Arg-34 X and Asp-72 B (Fig. 6, A and B). In addition, the Ras-GDP-AlF 3 :RasGAP complex contains a second arginine (Arg-903 RasGAP ) in close proximity to Arg-34 X (Fig. 6B). Arg-903 RasGAP forms a salt bridge to Glu-63 Ras in the switch II region of Ras thereby stabilizing the switch II region. In SR␤- GTP:SRX the catalytic residue His-119 B is hydrogen bonded to the corresponding residue of Glu-63 Ras (Ser-121 B ) (Fig. 6B).
The comparison of SR␤-GTP:SRX with the Ras-GDP-AlF 3 :RasGAP complex suggests that upon the insertion of an arginine finger into the GTP binding pocket the salt bridge between Arg-34 X and Asp-72 B can be disrupted. The liberated Arg-34 X may then swing from the P loop toward Ser-121 B in switch II forming a hydrogen bond (Fig. 6C). His-119 B would therefore be released, the catalytic water can be positioned, and hydrolysis occurs. Mutants in which the salt bridge is disrupted (Asp-72 B 3 Gly and an Arg-34 X 3 Ala) still form the SR␤-GTP:SRX complex (data not shown) indicating that the missing GAP is essential to stimulate GTP hydrolysis. The large conformational changes that are typically observed in the switch regions upon GTP hydrolysis are expected to disrupt the SR␤:SRX interface and lead to the dissociation of the SR complex (13). In the context of the SRP cycle this could happen either before or after signal peptide release from SRP.
Longin Domains Revisited-SRX belongs to the superfamily of SNARE-like proteins with the longin domain fold (15). Sequence homology within the superfamily is low (Fig. 5), but the structural homology is high (Fig. 7) as illustrated by the comparison of SRX with SEDL (19), with the SNAREs Sec22b (17) and Ykt6 (18), and the 2 (N-terminal domain) and 2 adaptins (20). To determine conserved elements within the longin domain fold we prepared a structure based sequence alignment of structurally known longin domains and of important longin domain candidates (-COPI, N␦-COPI, VAMP7; Fig.  5). Among longin domains with known structures, SRX y reveals specific insertions like strand ␤3 Xy , whereas the mammalian structure is closer to other members of the superfamily.
Longin domains share the ␤␤␣␤␤␤␣␣ topology as described for SRX (Fig. 1B). The glycine residue (Gly-12 X in SRX) in the ␤1-␤2 hairpin is highly conserved (Fig. 5), and the hairpin adopts a similar conformation in all longin domain structures. Only in SEDL this glycine is exchanged for an aspartate, and the change is compensated by adjustments in the adjacent ␤-strands. Ykt6 comprises a unique insertion of three residues. Helix ␣1 is an essential component of the longin domains (see below). The amphipathicity of helix ␣1 is highly conserved, while there is no conservation on the sequence level and the length ranges from three (SRX) to six turns (Ykt6). The orientation of helix ␣1 with respect to the central ␤-sheet varies in the different longin domains (Fig. 7). Flexibility is reflected by elevated temperature factors in the loops connecting helix ␣1 to the ␤-sheet (not shown) and in the SRX structure the flexibility is responsible for dimer formation by the swap of helix ␣1 X (Fig. 2).
A conserved ␤-sheet anomaly (down-up-up-down) is the insertion of a bulky hydrophobic residue (Leu-15 X in SRX, Fig. 5) within strand ␤2. It seems to be important for stabilizing the protein core and indicates an evolutionary relationship between the longin domains. The C-terminal helix ␣3 differs in length and orientation between the individual structures and superimposes best for Sec22b, SEDL, and SRX. Helix ␣3 is truncated in the longin domains of the AP2 complex (N2, 2), which according to secondary structure predictions is also the case in other AP complexes and the COPI complex (N␦-, -COPI) (Fig. 5). Here, the longin domain fold is extended by a ␤-hairpin structure followed by another helix forming a fourth layer in the back of the longin domain fold (not shown). The length and the conformation of the loop regions vary significantly (Fig. 7).
GTPase:Longin Domain Complexes at Endomembranes-The localization of longin domains at the endomembrane system correlates with the presence of small membrane-associated GTPases like the Arf and Sar1 proteins, which are the closest relatives of SR␤. The structural conservation and the co-localization strongly suggest that other GTPase:longin domain interactions may exist. Two hydrophobic patches flanking helix ␣1 were noticed previously in longin domain structures and were proposed as protein-protein interaction surfaces (17)(18)(19). Interestingly, these patches are conserved in structurally determined longin domains (Fig. 7). In the SR␤-GTP:SRX complex, SR␤ binds to this interaction surface. SR␤ intercalates its switch I region between helix ␣1 X and the SRX ␤-sheet; one of the helix flanking hydrophobic patches is extended and forms a hydrophobic groove (Figs. 3 and 8). In free longin domain structures the hydrophobic groove is absent (Fig. 8). The opening of the groove can be envisaged by rolling the conserved amphipathic helix ␣1 onto the second hydrophobic surface patch on the other side of the helix. The flexibility of helix ␣1 is therefore a prerequisite for the interaction of longin domains with their respective GTPase.
While the conservation of the hydrophobic patches suggests a similar mode of GTPase:longin domain interaction, the low degree of conservation reflects the special adaptations of the individual systems. For example, in all known longin domains the equivalent position of Arg-34 X within helix ␣1 X (Figs. 3 and 4B) seems to be occupied by a charged or polar residue (Fig. 5). In the respective GTPases the same is true for the residue at the position equivalent or adjacent to Asp-72 B in the P loop. Therefore, a polar contact between helix ␣1 and the P loop might be present in all GTPase:longin domain interactions. As discussed for the co-GAP function of SRX (see above), the residues corresponding to Arg-34 X could also participate in the stabilization of the switch II regions of the respective GTPases. The mammalian SR␤-GTP:SRX complex can thus be regarded as a structural prototype for a GTPase: longin domain interaction. Although there is no direct experimental proof, to our knowledge this idea does not contradict any previous data.
Structures of longin domains other than SRX have been determined as monomers (Sec22b, Ykt6, SEDL) or in context of the AP adaptin "trunc" complex. All clathrin adaptor complexes (AP1, -2, -3, -4) and COPI share a tetrameric trunc organization that consists of two large, a medium, and a small subunit (51). COPI and AP complexes contain two copies of longin domains (N␦and -COPI, and AP-N and -, respectively). In the structure of the AP2 complex (20), the two longin domains form the core of the trunc with the respective ␣1 helices being in close proximity. Therefore, a tandem GTPase:longin domain interaction might be an important feature in all these complexes.
A Molecular Explanation for a Genetic Disease-The GTPase:longin domain concept offers a structural explanation for the occurrence of spondyloepiphyseal dysplasia tarda, an X-linked skeletal disorder characterized by a short trunk (52). Point mutations in the human SEDL protein seem to be involved in a defect in cartilage transport from the ER FIGURE 8. An adjustable hydrophobic surface within longin domains. SR␤-bound SRX and free SEDL are shown in the same orientation with semitransparent surfaces. Helix ␣1 and the ␤1-␤2 hairpin are labeled. Colors are according to hydropathicity and hydrophobic groove and patches are indicated. A, SRX surface as prototype for a GTPase-bound longin domain. The switch I region of SR␤ (blue ribbon) binds into the hydrophobic groove created by the "packing" defect of helix ␣1 X and the central SRX ␤-sheet. Ile-94 B is inserted into a pocket in the center of the interface. B, SEDL as an example for a free longin domain. The hydrophobic groove is not present. to the Golgi apparatus (53). The yeast homologue of SEDL (Trs20p) has been shown to be part of the highly conserved transport protein particle I (TRAPP I) that is required to tether ER-derived vesicles to the Golgi (54) and consists of ten subunits (55). When the structure of SEDL is superimposed with SRX in the SR␤-GTP:SRX complex (Fig. 9), the pathogenic Asp-47 3 Tyr mutation in human SEDL would be located on the protein surface within helix ␣1 X in close proximity to the catalytic residue His-119 B and the interacting Ser-121 B of SR␤ (see Fig. 4B). There is no structure of the corresponding SEDL:GTPase complex; however, Ypt1p has been shown as the TRAPP interacting GTPase (55)(56)(57) and according to our model Gln-67 and Arg-69 in Ypt1p could form a favorable interaction with Asp-47 SEDL . Thus, the mutation most likely disturbs the GTPase regulation by interfering with the positioning of the catalytic residue.

CONCLUSIONS
The structure of the mammalian SR␤-GTP:SRX complex reveals the overall features of a GTPase-effector complex. Although the sequence identity is very low ((13) this study), the interaction is conserved between mammals and yeast. The mode of interaction, previous biochemical data (22,30), and the structural comparison with the Ras: RasGAP complex (50) point to a co-GAP function of SRX for SR␤. SRX stabilizes the switch II region but leaves room for an additional GAP, most likely the RNC-SRP complex, to insert an arginine-finger and stimulate GTP hydrolysis, which in turn would lead to the dissociation of the SR␤:SRX complex.
Small membrane-bound GTPases play a central role in all transport and targeting events at endomembranes (58,59). Given the importance of ribosome targeting for all protein secretion systems and the role of SR␤ in this process (23), the SR␤-GTP:SRX complex may be regarded as a structural and functional prototype for GTPase:longin domain interactions. The regulation of complex formation is indicative of a control point in the assembly of large transport complexes at the respective endomembrane system. The relevance of this concept needs to be tested experimentally in the future. The molecular explanation given for the genetic disease spondyloepiphyseal dysplasia tarda might be the first test case.