Portal de Periódicos da CAPES

Structures of Yeast ARF2 and ARL1

2001; Elsevier BV; Volume: 276; Issue: 45 Linguagem: Inglês

10.1074/jbc.m106660200

1083-351X

Juan Carlos Amor, J.R. Horton, Xinjun Zhu, Yi Wang, Cameron Sullards, Dagmar Ringe, Xiaodong Cheng, Richard Kahn,

RNA and protein synthesis mechanisms

Structures were determined by x-ray crystallography for two members of the ADP-ribosylation factor (ARF) family of regulatory GTPases, yeast ARF1 and ARL1, and were compared with previously determined structures of human ARF1 and ARF6. These analyses revealed an overall conserved fold but differences in primary sequence and length, particularly in an N-terminal loop, lead to differences in nucleotide and divalent metal binding. Packing of hydrophobic residues is central to the interplay between the N-terminal α-helix, switch I, and the interswitch region, which along with differences in surface electrostatics provide explanations for the different biophysical and biochemical properties of ARF and ARF-like proteins. Structures were determined by x-ray crystallography for two members of the ADP-ribosylation factor (ARF) family of regulatory GTPases, yeast ARF1 and ARL1, and were compared with previously determined structures of human ARF1 and ARF6. These analyses revealed an overall conserved fold but differences in primary sequence and length, particularly in an N-terminal loop, lead to differences in nucleotide and divalent metal binding. Packing of hydrophobic residues is central to the interplay between the N-terminal α-helix, switch I, and the interswitch region, which along with differences in surface electrostatics provide explanations for the different biophysical and biochemical properties of ARF and ARF-like proteins. ADP-ribosylation factor ARF-like GTPase activating protein interswitch region root mean square root mean square deviation switch I switch II N-terminal α-helix N-terminal loop ADP-ribosylation factor (ARF)1 and ARF-like (ARL) proteins comprise the ARF family within the Ras superfamily of regulatory GTPases. Members of this superfamily function as molecular nodes in signaling that can directly activate one or more enzymatic activities or coordinate the recruitment and assembly of more elaborate multisubunit complexes. The conformational changes that accompany the binding of GDP and GTP can lead directly to changes in the affinity of the GTPase for proteins, lipids, and membranes. For Ras superfamily members these conformational changes are centered in two functional regions, referred to as switch I (SW1) and switch II (SW2) (1Lacal J.C. McCormick F. The Ras Superfamily of GTPases. CRC Press, Boca Raton, FL1993Google Scholar). Members of the ARF family have an additional nucleotide-sensitive region, an extension at the N terminus and a covalently attached myristate that together work as a "myristoyl switch" to coordinate activation (GTP binding) with translocation onto a membrane (2Beraud-Dufour S. Balch W.E. Methods Enzymol. 2001; 329: 245-247Crossref PubMed Scopus (7) Google Scholar). In addition to its role in membrane binding, there is also evidence that the N terminus may influence effector binding or activities (3Kahn R.A. Randazzo P. Serafini T. Weiss O. Rulka C. Clark J. Amherdt M. Roller P. Orci L. Rothman J.E. J. Biol. Chem. 1992; 267: 13039-13046Abstract Full Text PDF PubMed Google Scholar, 4Zhu X. Boman A.L. Kuai J. Cieplak W. Kahn R.A. J. Biol. Chem. 2000; 275: 13465-13475Abstract Full Text Full Text PDF PubMed Scopus (24) Google Scholar). Residues in SW1 and SW2 make direct contacts with effectors and GTPase-activating proteins (GAPs) in a variety of GTPases, including ARFs (5Goldberg J. Cell. 1998; 95: 237-248Abstract Full Text Full Text PDF PubMed Scopus (439) Google Scholar, 6Goldberg J. Cell. 1999; 96: 893-902Abstract Full Text Full Text PDF PubMed Scopus (248) Google Scholar). It has been proposed that the two switches in ARF may act to coordinate the binding and activities of two different proteins simultaneously, a GAP and an effector, to effectively coordinate recruitment of vesicle cargo and GTP hydrolysis (6Goldberg J. Cell. 1999; 96: 893-902Abstract Full Text Full Text PDF PubMed Scopus (248) Google Scholar). However, the N-terminal region (residues 1–17) of ARF have been shown to be sufficient to confer ARF activity and increased GTPase activity to an ARL protein despite the distance between the N terminus and switch or nucleotide binding region (3Kahn R.A. Randazzo P. Serafini T. Weiss O. Rulka C. Clark J. Amherdt M. Roller P. Orci L. Rothman J.E. J. Biol. Chem. 1992; 267: 13039-13046Abstract Full Text PDF PubMed Google Scholar). These data highlight the complexity of interactions as well as the novel mechanisms of regulation of ARFs as compared with other members of the Ras superfamily. The distinction between ARFs and ARLs is based on both sequence relatedness and function (7Clark J. Moore L. Krasinskas A. Way J. Battey J. Tamkun J. Kahn R.A. Proc. Natl. Acad. Sci. U. S. A. 1993; 90: 8952-8956Crossref PubMed Scopus (111) Google Scholar). There are six mammalian ARFs and more ARLs, while the yeast Saccharomyces cerevisiae has only two ARFs and two ARLs. ARFs all share at least 60% identity in primary sequence and are active as (i) co-factor in the cholera toxin-catalyzed ADP-ribosylation of Gαs, (ii) direct activators or phospholipase D, and (iii) suppressors of lethality resulting from the deletion of the two yeast ARF genes. The biological roles of ARFs are central to many steps in vesicular traffic, particularly those involving the Golgi (8Roth M.G. Trends Cell Biol. 1999; 9: 174-179Abstract Full Text Full Text PDF PubMed Scopus (99) Google Scholar). In contrast, ARLs share 40–60% sequence identity with each other or with any ARF and lack each of the ARF activities described above (9Tamkun J.W. Kahn R.A. Kissinger M. Brizuela B.J. Rulka C. Scott M.P. Kennison J.A. Proc. Natl. Acad. Sci. U. S. A. 1991; 88: 3120-3124Crossref PubMed Scopus (124) Google Scholar). Much less is known about biological roles, protein effectors, or the importance of membrane-lipid interactions for the ARLs. That ARFs are essential and highly conserved throughout eukaryotic evolution is highlighted by the findings that at least one ARF has been found in every eukaryote tested (although not yet in any prokaryote) and that each one tested has demonstrated the ability to complement the arf1−rf2−double mutant in yeast. Structures have been determined by x-ray crystallography of the two most divergent human ARFs, ARF1 and ARF6 (10Amor J.C. Harrison D.H. Kahn R.A. Ringe D. Nature. 1994; 372: 704-708Crossref PubMed Scopus (254) Google Scholar, 11Menetrey J. Macia E. Pasqualato S. Franco M. Cherfils J. Nat. Struct. Biol. 2000; 7: 466-469Crossref PubMed Scopus (73) Google Scholar), and also for the ARF-like protein murine ARL3 (18Hillig R.C. Hanzal-Bayer M. Linari M. Becker J. Wittinghofer A. Renault L. Structure. 2000; 8: 1239-1245Abstract Full Text Full Text PDF PubMed Scopus (49) Google Scholar). These structures have revealed the fundamental conservation of structure between ARFs and other members of the Ras superfamily in the six β-strands at the core of the protein and surrounding helices. We report here structures of GDP-bound, S. cerevisiae ARF2 (ScARF2) and ARL1 (ScARL1), as determined by x-ray crystallography to 1.6 and 3.2 Å resolution, respectively. The structures of the two proteins show changes in and surrounding the N terminus. These differences give indications of how the diverse biochemical characteristics may be determined or mediated by the sequence and structure of the N terminus. In addition, comparison of surface electrostatics of the four ARF family members revealed surprisingly large differences that are discussed with respect to properties of each protein. ScARF2 and ScARL1 were expressed in BL21(DE3) bacteria using the pET3C expression plasmid as described previously for other ARF proteins (12Kahn R.A. Methods Enzymol. 1991; 195: 233-242Crossref PubMed Scopus (19) Google Scholar). Cells were grown to an A600 of 1.0 before protein expression was induced with isopropylthio-β-d-galactoside (0.5 mm). Cells were harvested by centrifugation after ∼12 h of growth at 37 °C. Cells were resuspended in 25 mmTris, 50 mm NaCl, 1 mm MgCl2, 1 mm dithiothreitol, 1 mm phenylmethylsulfonyl fluoride, 50 μm GDP, pH 7.5 before lysis using a French press. Bacterial cell lysate was passed over a low substituted Q-Macro column (Bio-Rad; 100-ml bed volume) in the same buffer, and the flow-through was applied to a high substituted Q-Macro column (Bio-Rad; 50-ml bed volume). A linear gradient (50–350 mm NaCl, 500 ml) in lysis buffer was used to fractionate proteins. Pooled fractions were concentrated to 5 ml and loaded on a Sephacryl S-100 HR (Amersham Pharmacia Biotech; 60 cm, 300-ml bed volume) gel filtration column. Protein purity was determined by denaturing (SDS) polyacrylamide gel electrophoresis, and the pooled fractions were concentrated to a final concentration of 1 mm protein for storage at −80 °C. Crystals of ScARF2 were grown by the hanging drop method combining equal volumes (2 μl) of the protein stock and mother liquor (20% polyethylene glycol 8000, 0.1 mcacodylate, pH 6.5) at room temperature. ScARF2 crystals grew as thin plates (0.3 × 0.06 × 0.02 mm3) from a central hub and had to be separated before vitrification in 30% polyethylene glycol 8000, 0.1 m cacodylate, pH 6.5 and 15% glycerol using liquid propane. ScARL1 crystals were also obtained by the hanging drop (20% polyethylene glycol 8000, 15% glycerol, 0.1 mTris, pH 8.5). Statistics for the structure determination are shown in Table I. For ScARF2, data collection to better than 1.6 Å was performed at the Brookhaven National Laboratory X12b beam line (γ = 1.008 Å). One-degree oscillation frames were integrated with DENZO, then merged, and scaled with SCALEPACK from HKL (13Otwinowski Z. Minor W. Methods Enzymol. 1997; 276: 307-326Crossref PubMed Scopus (38609) Google Scholar). Initial phases for the structure factors were determined by molecular replacement with AMoRe (14Navaza J. Acta Crystallogr. Sect. A. 1994; 50: 157-163Crossref Scopus (5030) Google Scholar) using HsARF1-GDP (10Amor J.C. Harrison D.H. Kahn R.A. Ringe D. Nature. 1994; 372: 704-708Crossref PubMed Scopus (254) Google Scholar) as the search model. For both proteins the final model was established by successive rounds of structure building using O (15Jones T.A. Zou J.Y. Cowan S.W. Kjeldgaard M. Acta Crystallogr. Sect. A. 1991; 47: 110-119Crossref PubMed Scopus (13014) Google Scholar) and refinement with CNS (16Brunger A.T. Adams P.D. Clore G.M. DeLano W.L. Gros P. Grosse-Kunstleve R.W. Jiang J.S. Kuszewski J. Nilges M. Pannu N.S. Read R.J. Rice L.M. Simonson T. Warren G.L. Acta Crystallogr. Sect. D Biol. Crystallogr. 1998; 54: 905-921Crossref PubMed Scopus (16979) Google Scholar) (Table I).Table IStatistics for x-ray structures reported hereStatistics for x-ray structure determination and refinementScARF2ScARL1Space groupP21212P31Unit cell parametersa = 61.8, b = 69.5, c = 40.9a = b = 104.3, c = 45.7α = β = γ = 90.0α = β = γ = 90.0Measured reflections127,58130,535Unique reflections23,0349,179Completeness (%)96.497.5Resolution range (Å)35–1.6028–3.17Rsym(%)4.57.1R-factor (%)20.420.6Rfree(%)23.524.2r.m.s. deviations Bond lengths (Å)0.00490.0077 Bond angles (°)1.21.3Average B-factor (Å2)20.437.4Number of water molecules20719Metals (Mg2+)10Nucleotide3228Protein atoms1,4221,341 (with NCS)NCS, noncrystallographic symmetry. Open table in a new tab NCS, noncrystallographic symmetry. Nucleotide bound to ScARF2 was released by the addition of ice-cold HClO4 to a final concentration of 0.5 m, incubation at 0 °C for 10 min, removal of precipitated protein by centrifugation, and neutralization of the supernatant by addition of 16 volume of 1 mK2HPO4, 0.5 m acetic acid and 16 volume of 3 m KOH at 0 °C. The solution of the extracted nucleotide was then exchanged into deionized water and lyophilized, and the residue was resuspended in a 20% propanol, 1% triethylamine, water mixture. All experiments were performed on a PE Sciex API 3000 triple-quadrupole mass spectrometer equipped with a turboionspray source. Product ion spectra were acquired to detect the (M − H)− ions of m/z 442.3 or 522.4. Data was acquired for a total of 1.7 min, thus each spectrum was the signal-averaged sum of ∼100 scans. Precursor ion spectra were acquired to detect either m/z 79.0 or 150.0. Data was acquired for a total of either 2.5 or 5.0 min, respectively, thus each spectrum was the signal-averaged sum of 300 scans. Nitrogen was used to collisionally activate precursor ion decomposition. ScARF1 and ScARF2 are over 97% identical and share about 74% identity with mammalian orthologs, e.g. HsARF1 or HsARF6 (see Fig. 1 A). ScARF1 and ScARF2 were each expressed in bacteria, purified to homogeneity, and crystallized as described under "Materials and Methods." Crystals of ScARF2 diffracted to higher resolution (1.6 Å) and the resultant structure provided details not seen in previous ARF structures. Comparisons between yeast (ScARF2) and human (HsARF1 (10Amor J.C. Harrison D.H. Kahn R.A. Ringe D. Nature. 1994; 372: 704-708Crossref PubMed Scopus (254) Google Scholar) and HsARF6 (11Menetrey J. Macia E. Pasqualato S. Franco M. Cherfils J. Nat. Struct. Biol. 2000; 7: 466-469Crossref PubMed Scopus (73) Google Scholar)) structures provide insights into conserved features throughout this evolutionary distance. We have also crystallized and solved a structure for yeast ARL1 (ScARL1) at a lower resolution (3.2 Å). Data and model statistics for each new structure are shown in TableI. Because of the functional differences between ARFs and ARLs (9Tamkun J.W. Kahn R.A. Kissinger M. Brizuela B.J. Rulka C. Scott M.P. Kennison J.A. Proc. Natl. Acad. Sci. U. S. A. 1991; 88: 3120-3124Crossref PubMed Scopus (124) Google Scholar) and greater divergence in sequence (40–60% identity) the variations between the structures provide details of structural divergence of likely functional importance. The secondary structures of GDP-bound ScARF2, ScARL1, and HsARF1 are essentially conserved (see Fig. 1 B); the topology includes seven β-strands, six α-helices, and 12 connecting loops. Of these, the protein core is an invariant six-stranded β-sheet surrounded by six α-helices. The seventh strand forms an edge to the core and corresponds to the functionally important SW1 region seen in regulatory GTPases (1Lacal J.C. McCormick F. The Ras Superfamily of GTPases. CRC Press, Boca Raton, FL1993Google Scholar). The N-terminal region of each protein is an α-helix (αNt) wedged between the C-terminal helix α-E and loop L-2/3, which connects strands β2 and β3, also referred to as the interswitch region (ISR). The presence of αNt is in contrast to the rather undefined secondary fold seen for the N-terminal sequence in the ARL3 model (18Hillig R.C. Hanzal-Bayer M. Linari M. Becker J. Wittinghofer A. Renault L. Structure. 2000; 8: 1239-1245Abstract Full Text Full Text PDF PubMed Scopus (49) Google Scholar). SW2 corresponds to L-3/B and αB, where L-3/B is on the opposite side of the N-terminal region and αB is opposite the bound GDP. The nucleotide binding site is on the surface of the protein between αNt and SW2 (Fig. 1 B). Higher affinity for GDP over GTP and lipid dependence on the binding of activating guanine nucleotides likely explain the presence of GDP in the nucleotide binding site of ARFs as purified from mammalian tissues or bacteria (17Kahn R.A. Goddard C. Newkirk M. J. Biol. Chem. 1988; 263: 8282-8287Abstract Full Text PDF PubMed Google Scholar). GDP was well fit in the binding site of previous ARF structures and in both yeast proteins described here. But there remained clear, but unexplained, density at the 3′ position of the ribose bound to ScARF2. Different orientations of the ribose were considered, but only a covalent modification at the 3′ position could be fit well (see Fig. 2 A). The presence of a modified GDP in the protein preparation used to produce the ScARF2 crystals was investigated using mass spectroscopy. The presence of GDP, with the predicted mass of 442 atomic mass units, was evident in the preparation as was the presence of a second species with a mass of 522 atomic mass units. The mass of the larger species was consistent with a modification to GDP that adds 80 atomic mass units and can be explained by an additional phosphate (HPO3). We estimate from the mass spectroscopy results and the electron density in the crystal structure that our preparation of ScARF2 has ∼70% GDP and 30% GDP-3′-phosphate bound. This is the first modified nucleotide found on a regulatory GTPase as purified, although no biological significance can yet be ascribed to this observation. The high resolution of the ScARF2 structure also allowed the definition of alternate conformations of side chains that were not evident in previous structures as well as the binding of a number of small molecules. For example, the invariant lysine (A2:K30), 2In discussing comparisons between the four ARF family members we use the numbering for HsARF1 as the benchmark for this family. Residue numbering in HsARF1 (PDB code 1HUR) is indicated by the abbreviation A1:"residue," in ScARF2 as A2:residue, in ScARL1 as L1:residue, and in HsARF6 (PDB code 1E0S) as A6:residue. Single letter abbreviations are used for amino acids in this numbering scheme. which contacts the α- and β-phosphates, has alternate conformations for the γ- and δ-carbons. Only one of these conformations is observed for all the ARF family members structures published to date, except the HsARF1·GDP·GAP structure (6Goldberg J. Cell. 1999; 96: 893-902Abstract Full Text Full Text PDF PubMed Scopus (248) Google Scholar), which has the other conformation. Our results show both conformations simultaneously present for Lys30 in ScARF2. In addition to more than 200 water molecules, the surface of the protein was found to bind two glycerol, one 1,3-propanediol, nine 1,2-ethanediol, and three ethanol molecules per protein monomer, all introduced by the use of glycerol as cryoprotectant. The most provocative of the binding sites for these molecules is the channel extending from the nucleotide and magnesium binding site (Fig. 2 B). The fact that this channel is lined by residues involved in magnesium binding (Glu54 and Asp67) and residues of the lower ISR (Asn52) as well as SW1 (Thr48 and Phe51) and SW2 (not shown) leads us to speculate a potential for this being a site of interaction with lipid, or other modulators, that could promote nucleotide exchange or conformational changes in one or both switches. Similar to murine Arl3-GDP (18Hillig R.C. Hanzal-Bayer M. Linari M. Becker J. Wittinghofer A. Renault L. Structure. 2000; 8: 1239-1245Abstract Full Text Full Text PDF PubMed Scopus (49) Google Scholar) a covalent homodimer was observed for ScARL1; however, here the disulfide bond is within the asymmetric unit and is linking the two Cys81 residues present in SW2 (Fig. 3). Dimerization was not the result solely of crystallization as it was detectible by either nondenaturing polyacrylamide gel electrophoresis or high-resolution gel filtration chromatography (results not shown) prior to crystallization. This is the only known example of a covalent homodimer being formed by a member of the Ras superfamily. Recently ARF1 and Ras were proposed to form homodimers on 3A. G. Rosenwald, M. Rhodes, H. Van Valkenburgh, G. Chapman, J. Shu, V. Palanivel, C. J. Testa, A. B. Boman, C. J. Zhang, and R. A. Kahn, submitted.membranes (19Zhao L.Y. Helms J.B. Brunner J. Wieland F.T. J. Biol. Chem. 1999; 274: 14198-14203Abstract Full Text Full Text PDF PubMed Scopus (71) Google Scholar,20Inouye K. Mizutani S. Koide H. Kaziro Y. J. Biol. Chem. 2000; 275: 3737-3740Abstract Full Text Full Text PDF PubMed Scopus (118) Google Scholar). Indeed membrane localization promotes dimer formation of Ras, which is essential, although not sufficient, for activation of its target molecule, Raf-1 (20Inouye K. Mizutani S. Koide H. Kaziro Y. J. Biol. Chem. 2000; 275: 3737-3740Abstract Full Text Full Text PDF PubMed Scopus (118) Google Scholar). Dimerization also appears to be involved in the regulation of the activity of Rho family GTPases (21Zhang B.L. Zheng Y. J. Biol. Chem. 1998; 273: 25728-25733Abstract Full Text Full Text PDF PubMed Scopus (56) Google Scholar). To test the functional importance of dimerization we constructed the C81S mutant of ScARL1 and integrated it at the ARL1 locus by homologous recombination. This point mutant was fully functional as defined only by rescue of the cold sensitivity seen in arl1−cells. Thus, there is no evidence to support a required role for dimerization, covalent or otherwise, of ScARL1. However, both ARFs and ARLs have conserved cysteines;e.g. all known ARL1 orthologs have a cysteine at the position homologous to L1:81, while ARL2 orthologs lack that cysteine but have one at position 103 (A1:102). The model for ScARL1 reported here includes a disordered gap at residues 71–74 because of the lack of density to model the structure appropriately. This region is the start of SW2 and typically has the highest crystallographic thermal values in GTPase structures. In contrast, the structure of the entire SW2 region of ScARF2 is well defined and completely determined, possibly due to the crystal contacts present in this area. Similarly, the entire SW2 region of the ARL3 structure could be modeled (18Hillig R.C. Hanzal-Bayer M. Linari M. Becker J. Wittinghofer A. Renault L. Structure. 2000; 8: 1239-1245Abstract Full Text Full Text PDF PubMed Scopus (49) Google Scholar). Comparison of the electrostatic surface of HsARF1, ScARF1, HsARF6, and ScARL1 revealed striking differences that are consistent with their membrane interactions. The electrostatic profiles of the four proteins fall into three categories: a patched surface, a primarily acidic surface potential, or a primarily basic surface potential (Fig. 4). HsARF1/ScARF2 comprise the group distinguished by the patched appearance of both anionic (negative potential in red) and cationic (positive potential in blue) areas as seen in Fig. 4, A and B. These patches of mixed charge, including SW1 and SW2, when positioning the N terminus toward the membrane, would provide a distinctive surface for protein interactions oriented toward the cytosol. Most prominent for HsARF1/ScARF2 is a horseshoe-shaped acidic cushion surrounding the N-terminal helix, which itself is predominantly basic (Fig. 4, A and B). Such a charged surface may assist in the orientation of the ARF1 to promote the interaction of the N terminus and the covalently bound myristate with membrane lipids. This may also help explain the ability of acidic lipids to promote nucleotide exchange through interaction with the N-terminal region (22Kahn R.A. Terui T. Randazzo P.A. J. Lipid Mediat. Cell Signal. 1996; 14: 209-214Crossref PubMed Scopus (10) Google Scholar). The surface of ScARL1 is distinguished by an overall negative electrostatic potential (see Fig. 4 C). Like HsARF1, both human and yeast ARL1 are N-myristoylated, yet we have been unable to demonstrate 4H. Van Valkenburgh, J. D. Sharer, and R. A. Kahn, unpublished observations. any GTP-dependent binding of HsARL1 to membranes or lipid vesicles that is a hallmark of ARFs (23Kahn R.A. Gilman A.G. J. Biol. Chem. 1986; 261: 7906-7911Abstract Full Text PDF PubMed Google Scholar). These observations lead us to predict that the binding of human and yeast ARL1 to membrane is more dependent on protein-protein interactions at the membrane than is the binding of ARFs. In contrast to the two profiles described above, the electrostatic potential of HsARF6 gives the protein a primarily cationic appearance (see Fig. 4 D). Interestingly the location on the surface of HsARF6 that surrounds and includes the N-terminal helix is strongly basic. An increased tendency to interact with acidic head groups of lipids would be predicted from such an electrostatic profile, and this may contribute to the finding that ARF6 more tightly associates with membranes even when GDP is bound. Despite the variation of the electrostatic potential for all four proteins as described above, there is one consistent observation among them. The highly conserved loop L-SW1/β2 is electrostatically isolated, both due to its exposure at the surface and its predominantly hydrophobic nature (especially noted for ScARL1 in Fig. 4 C). To identify structural differences between the four ARF family members, we aligned the structures of ScARF2, ScARL2, and HsARF6 against HsARF1. For an optimal alignment, one that would highlight the differences and maintain invariant features, we first superimposed the backbone Cα-atoms of the highly invariant fourth, fifth, and sixth β-strands. The r.m.s.d. for all Cα-atoms was then calculated and plotted against the sequence position as shown in Fig. 5. The largest differences were found at the N terminus (A1:2–19), the switch regions (SW1 = A1:36–52 and SW2 = A1:69–85), and the ISR (loop L-2/3, A1:57–62). Further prominent r.m.s. differences were found in regions A1:129–133 and A1:149–151. Amino acids 129–133 are adjacent to the highly conserved 126NKQD129sequence, which is directly involved in the binding of the guanine nucleotide base. Residues 149–151 are close to the GAP binding site as described by Goldberg (6Goldberg J. Cell. 1999; 96: 893-902Abstract Full Text Full Text PDF PubMed Scopus (248) Google Scholar). Structural changes with a smaller r.m.s.d. were found in regions A1:100–120, at A1:27, and centered around A1:162 (see Fig. 5). Residues 100–120 are involved in binding to ARF-GAP (6Goldberg J. Cell. 1999; 96: 893-902Abstract Full Text Full Text PDF PubMed Scopus (248) Google Scholar). Residue 27 is part of the putative phosphate binding sequence G X4GK. Residue 162 follows a group of residues involved in binding the guanine base as discussed in more detail below. That these differences in structure are scattered throughout the molecules and are mostly located in previously identified regions essential to function are consistent with these being functionally important regions and useful in defining specificity between the proteins. Membrane binding, nucleotide affinity, phospholipid dependence of GTP binding, and nucleotide hydrolysis by ARFs are all sensitive to the presence and sequence composition of the N-terminal 17 amino acids; they are the αNt and loop (L-Nt) comprising the "N-terminal region" (3Kahn R.A. Randazzo P. Serafini T. Weiss O. Rulka C. Clark J. Amherdt M. Roller P. Orci L. Rothman J.E. J. Biol. Chem. 1992; 267: 13039-13046Abstract Full Text PDF PubMed Google Scholar, 22Kahn R.A. Terui T. Randazzo P.A. J. Lipid Mediat. Cell Signal. 1996; 14: 209-214Crossref PubMed Scopus (10) Google Scholar). The structure and position of this region has only been determined for ARF(s) bound to GDP and is presumed to change dramatically upon binding GTP (5Goldberg J. Cell. 1998; 95: 237-248Abstract Full Text Full Text PDF PubMed Scopus (439) Google Scholar). How the presence or structure of the N-terminal region can so dramatically influence the nucleotide binding site has been the subject of much conjecture. A comparison of the four structures illustrates how the sequence of the N-terminal helix and the length of the following loop L-Nt can affect the structure of SW1 and the interswitch region and in so doing the nature of the nucleotide bound. In essence, the presence of the N-terminal region constrains the ISR (loop L-2/3), a primary determinant of the bound nucleotide. The largest differences detected from the comparison of the four structures were in these three elements: the N-terminal region, the ISR, and SW1. Differences in the N-terminal region (most obviously the lack of αNt) and the ISR were also reported for the recent structure of ARL3 (18Hillig R.C. Hanzal-Bayer M. Linari M. Becker J. Wittinghofer A. Renault L. Structure. 2000; 8: 1239-1245Abstract Full Text Full Text PDF PubMed Scopus (49) Google Scholar). A more detailed comparison of these areas is shown in Fig. 6. A key to the organization of these regions is the presence of hydrophobic residues, best viewed in HsARF1 (Fig. 6 A), which include Phe5, Phe9, Leu37, Leu39, Val56, Tyr58, Ile61, and Phe63, forming a hydrophobic cluster. Note, in particular, the extent to which Leu39 is inserted into the cluster and makes direct contact with the N-terminal helix. In HsARF6 (Fig. 6 B) the equivalent is Leu35, but it is much less well fit into this cluster due to the shortening of loop L-Nt, a four-residue deletion relative to HsARF1 (Fig. 1 A), and the presence of a leucine residue (Leu5) instead of a phenylalanine (Phe5), as in HsARF1(Fig. 6 A). These changes lead to a reorientation of the ISR and specifically Tyr54. The result is a large change in the structure of the N-terminal portion of SW1 of HsARF6. In contrast, loop L-Nt is much longer in ScARL1, allowing yet other structural differences in the ISR. When coupled with the change to L1:I40 (corresponding to A1:L39) this allows the first part of SW1 to make closer contacts and additional hydrogen bonds with the ISR (Fig. 6 C). Although ScARF2 (Fig. 6 D) has a shorter loop L-Nt than HsARF1, the conserved Leu39 and hydrophobic cluster prevent the change in SW1 seen in HsARF6 and ScARL1. In ARL3, differences in the structure of the N terminus and ISR result from differences in key residues,e.g. a lysine instead of glutamate at position 54 and the two phenylalanines in ARF1, Phe5 and Phe9, are serine and lysine, respectively, in ARL3 (18Hillig R.C. Hanzal-Bayer M. Linari M. Becker J. Wittinghofer A. Renault L. Structure. 2000; 8: 1239-1245Abstract Full Text Full Text PDF PubMed Scopus (49) Google Scholar). Thus, we conclude that the different packing of hydrophobic residues modified by contributions from the N-terminal helix and the ISR, in addition to the variation in length of loop L-Nt, lead to differences in the initial portion of SW1. The presence of this hydrophobic cluster, the degree of residue insertion, and the ability of backbone hydrogen bond formation are most likely the principal causes for the change in structure at the start of the SW1 region. Glutamate 50 of HsARF6 influences nucleotide exchange, probably through effects on the binding of Mg2+ (11Menetrey J. Macia E. Pasqualato S. Franco M. Cherfils J. Nat. Struct. Biol. 2000; 7: 466-469Crossref PubMed Scopus (73) Google Scholar). This glutamate (A6:E50, A1:E54) cannot approach the metal to make the contacts seen in ARF1 (see Ref. 10Amor J.C. Harrison D.H. Kahn R.A. Ringe D. Nature. 1994; 372: 704-708Crossref PubMed Scopus (254) Google Scholar and Fig. 7, A and B), and thus a weakening of the affinity for Mg2+ was predicted (11Menetrey J. Macia E. Pasqualato S. Franco M. Cherfils J. Nat. Struct. Biol. 2000; 7: 466-469Crossref PubMed Scopus (73) Google Scholar). We find similar structural variations in ScARL1 and ScARF2 that are also predicted to have an impact on the affinity for magnesium (Mg2+) and consequently nucleotides (Fig. 7, Cand D). This may explain the observations that indicate each of these proteins exchange nucleotides more rapidly than HsARF1. It may also explain the absence of bound Mg2+ for some of the published structures (11Menetrey J. Macia E. Pasqualato S. Franco M. Cherfils J. Nat. Struct. Biol. 2000; 7: 466-469Crossref PubMed Scopus (73) Google Scholar, 18Hillig R.C. Hanzal-Bayer M. Linari M. Becker J. Wittinghofer A. Renault L. Structure. 2000; 8: 1239-1245Abstract Full Text Full Text PDF PubMed Scopus (49) Google Scholar). The predicted differences in affinity for magnesium result from an increase in distance, i.e. decrease in interaction, between a conserved glutamate (A1:E54, L1:E55), involved in binding of the Mg2+, and a conserved tyrosine hydroxyl (A1:Y35, L1:Y36). This results from structural changes and sequence variations in the SW1 area, specifically at residues A1:40–42 (L1:41–43). In ScARL1 (Fig. 7 C), the modified positioning of L1:I40, including the flip of its ψ-angle, combines with the sequence difference of L1:V43 (A1:I42) to allow L1:Y36 (A1:Y35) to interact with L1:E55 (A1:E54). A 90° rotation around X1 of L1:Y36 in ScARL1 (from −172° in HsARF1 to −79° in ScARL1) enables the tyrosine hydroxyl to hydrogen bond with L1:E55 (3.0 Å) (see Fig. 7 C), moving L1:E55 away from its position as an effective Mg2+ coordinator (10Amor J.C. Harrison D.H. Kahn R.A. Ringe D. Nature. 1994; 372: 704-708Crossref PubMed Scopus (254) Google Scholar) (see Fig. 7 B). The same rotation of A2:Y35 is observed in ScARF2 (X1 = −82°) as well as the presence of a valine at residue 42, but because the topology around SW1 is more similar to HsARF1, A2:Y31cannot approach A2:E54 (5.0 Å) (see Fig. 7 D). This trend is most exaggerated in ARF6. Again the same rotation of A6:Y31 is observed (X1 = −76°), but because of the topology divergence in SW1 around this area relative to HsARF1, A6:Y31 is even further from A6:E50 (8.0 Å) (Fig. 7 A). Instead A6:E50 hydrogen bonds with A6:S38 (equivalent to residue A1:I42). Thus, the most dramatic repositioning of the conserved glutamate is found in HsARF6 (11Menetrey J. Macia E. Pasqualato S. Franco M. Cherfils J. Nat. Struct. Biol. 2000; 7: 466-469Crossref PubMed Scopus (73) Google Scholar) followed by ScARL1 and ScARF2, and the nearest approach of glutamate and magnesium is seen in HsARF1. Although the direct coordination of magnesium with this conserved glutamate was not seen in the other HsARF1 structure (24Greasley S.E. Jhoti H. Teahan C. Solari R. Fensome A. Thomas G.M.H. Cockcroft S. Bax B. Nat. Struct. Biol. 1995; 2: 797-806Crossref PubMed Scopus (101) Google Scholar), all of these data are consistent with this glutamate playing a central role in the binding of magnesium to members of the ARF family and consequently to the affinity of guanine nucleotides. This argument was further supported by the recent ARL3-GDP model in which sequence variations in SW1, similar to those mentioned above, and the changes of both of the key residues (A1:E54and A1:Y35) to lysine result in altered binding of magnesium. In the ARL3-GDP model no Mg2+ is evident but a SO42− is interacting with both these lysines, and GDP binding, but not GTP binding, is Mg2+-independent (18Hillig R.C. Hanzal-Bayer M. Linari M. Becker J. Wittinghofer A. Renault L. Structure. 2000; 8: 1239-1245Abstract Full Text Full Text PDF PubMed Scopus (49) Google Scholar). Thus, a change in the binding of magnesium can result from either the replacement of the critical glutamate or its interaction with tyrosine, in each case resulting in the loss of a critical coordination of the magnesium. Nucleotides bind to ARFs through interactions above and below the plane of the base. On one face are side chains from the second conserved G-domain, NKQD (A1:N126–D129) in the ARF family, and on the other face is a less conserved residue (A1:T161). Given the importance of these interactions and their conservation among GTPases it was surprising to note that after the N-terminal region, SW1, and SW2, the region with the highest r.m.s. deviation between structures (>1 Å) is found in loop L-5/D, A1:130LPNA133, immediately following A1:D129. A1:K127 (not shown) interacts hydrophobically along the face of the base while Asp129forms a hydrogen bond at the edge of the base. While these interactions were conserved in all four structures the distances of the bonds formed and orientations of side chains differed in ways that may predict differences in nucleotide affinities or handling. Not only does the r.m.s. analysis (Fig. 5) show substantial differences between the backbones in this G-domain, the side chains are oriented differently as well. A change in the residue corresponding to A1:N132 is always coordinated with a concomitant change at residue A1:N95 (see Fig. 1 A). These residues are on loops that interact with each other, and it is the type of interaction, which can vary with sequence changes, that causes the larger r.m.s. deviation at the location of loop L-5/D. On the other side of the base is the less conserved sequence (A1:158TCAT161) that provides A1:T161 for base interactions, but this interaction appears much less strong. The density for A2:T161 indicates this residue is rotating through about 90°. It is also hydrogen bonding with a water molecule. The sequence at this position varies within the ARF family and can be a valine, isoleucine, leucine, or alanine in different ARLs. Structural variation at the guanine base sandwich due to A1:T161 and Asp129 (in turn due to structural perturbations in the subsequent loop residues 130–133) most likely influences the ability of the protein to bind the base of the nucleotide and therefore influences the affinity of the nucleotide, similar to differences in the binding of magnesium. A rank order in predicted affinity for GDP is difficult as it results from at least three factors that do not necessarily exert their effect in a coordinated fashion: the residue at position A1:T161, the proximity and mobility of residues A1:D129-L130, and the orientation of A1:N132 resulting from its interaction with A1:N95. Another ARL structure, that of ARL2(GTP) bound to the δ subunit of phosphodiesterase (PDEδ), was published during the review of this manuscript (29Renault L. Hanzal-Bayer M. Hillig R.C. Acta Crystallogr. Sect. D Biol. Crystallogr. 2001; 57: 1167-1170Crossref PubMed Scopus (18) Google Scholar).