View metadata, citation and similar papers at core.ac.uk brought to you by CORE

provided by Elsevier - Publisher Connector

Research Article 567

Structure of bovine mitochondrial F1-ATPase inhibited by Mg2+ADP and Kerstin Braig1, R Ian Menz2, Martin G Montgomery3, Andrew GW Leslie4* and John E Walker3*

1 Background: The globular domain of the membrane-associated F1Fo-ATP Addresses: Institut für Organische Chemie und synthase complex can be detached intact as a water-soluble fragment known as Biochemie, Albert-Ludwigs Universität Freiburg, α, β, γ, δ ε, Albertstr. 21, D-79104 Freiburg in Breisgau, F1-ATPase. It consists of five different subunits, and assembled with Germany, 2Department of Biochemistry, University the stoichiometry 3:3:1:1:1. In the of bovine F1-ATPase of Sydney, New South Wales 2006, Australia, determined previously at 2.8 Å resolution, the three catalytic β subunits and the 3Medical Research Council Dunn Human Nutrition three noncatalytic α subunits are arranged alternately around a central α-helical Unit, Hills Road, Cambridge CB2 2XY, UK and 4Medical Research Council Laboratory of Molecular coiled coil in the γ subunit. In the crystals, the catalytic sites have different Biology, Hills Road, Cambridge CB2 2QH, UK. nucleotide occupancies. One contains the triphosphate form of the nucleotide, the second contains the diphosphate, and the third is unoccupied. *Corresponding authors. Fluoroaluminate complexes have been shown to mimic the transition state in E-mail: [email protected] several ATP and GTP hydrolases. In order to understand more about its catalytic [email protected] mechanism, F -ATPase was inhibited with Mg2+ADP and aluminium fluoride and 1 Key words: aluminium fluoride, F1-ATPase, the structure of the inhibited complex was determined by X-ray crystallography. hydrolysis, X-ray crystallography

Results: The structure of bovine F -ATPase inhibited with Mg2+ADP and Received: 3 February 2000 1 Revisions requested: 10 March 2000 aluminium fluoride determined at 2.5 Å resolution differs little from the original Revisions received: 27 March 2000 structure with bound AMP-PNP and ADP. The nucleotide occupancies of the α Accepted: 28 March 2000 and β subunits are unchanged except that both aluminium trifluoride and 2+ β Published: 23 May 2000 Mg ADP are bound in the nucleotide-binding site of the DP subunit. The presence of aluminium fluoride is accompanied by only minor adjustments in the Structure 2000, 8:567–573 surrounding protein. 0969-2126/00/$ – see front matter Conclusions: The structure appears to mimic a possible transition state. The © 2000 Elsevier Science Ltd. All rights reserved. coordination of the aluminofluoride group has many features in common with other aluminofluoride–NTP hydrolase complexes. Apparently, once nucleotide is bound to the catalytic β subunit, no additional major structural changes are required for catalysis to occur.

Introduction a thermophilic bacterium [1]. During ATP synthesis in the γ ATP synthase (F1Fo-ATPase) catalyses the synthesis of intact ATP synthase it is assumed that the rotation of the ATP from ADP and , using the proton-motive subunit is driven by the Fo domain, using the proton-motive force generated by oxidative phosphorylation or photosyn- force as an energy source. Thus rotation of the γ subunit

thesis to drive the reaction. Catalysis takes place in the couples the proton-motive force to ATP synthesis in F1. extrinsic membrane domain, known as F1, which is com- α, β, γ, δ ε, posed of five polypeptides, and assembled with In the crystal structure of F1-ATPase from bovine heart the stoichiometry 3:3:1:1:1. The F1 domain is a roughly mitochondria [2], the three catalytic subunits have different spherical assembly about 100 Å in diameter in which the nucleotide-binding properties. One binds the triphosphate α β β three subunits and the three subunits are arranged alter- form of the nucleotide (denoted TP), the second the α β β nately around an antiparallel coiled coil of two helices in diphosphate ( DP), and the third is unoccupied ( E), despite the γ subunit. Both α and β subunits bind nucleotides, but the presence of substantial concentrations of nucleotides in the catalytic nucleotide-binding sites are in the β subunits at the mother liquor surrounding the crystals. It appears from the interface with the adjoining α subunits. The α-helical examination of the structure that these different γ α β γ coiled coil in the subunit protrudes from the 3 3 complex nucleotide-binding properties are imposed by the subunit, and interacts directly with the membrane domain, Fo. In the and that rotation of this subunit will take each of the three bovine mitochondrial enzyme, it is likely that the δ and ε catalytic sites through the three states represented in the subunits are associated with this protrusion. ATP-depen- structure, in a concerted manner. The molecular details of γ α β dent rotation of the subunit relative to the 3 3 complex these interconversions are obscure, and it would be helpful α β γ has been demonstrated directly for the 3 3 complex from if the structures of intermediate states, and particularly that 568 Structure 2000, Vol 8 No 6

of the transition-state complex, could be established. In atoms is 0.66 Å). Data processing and refinement statistics recent years, structures have been described of a number of are presented in Table 1. This model has been used as a NTPases (GTPases and ATPases) complexed with NDP- basis for all comparisons with the aluminium-fluoride- α fluoroaluminate. They include the subunits of hetero- inhibited complex (AlF3–F1). trimeric G proteins [3,4], myosin [5], nitrogenase [6], nucleoside diphosphate kinase [7], p21ras.GAP [8], the ther- Quality of the structure of the aluminium-fluoride-inhibited mosome [9] and UMP/CMP kinase [10]. Various features of complex the structures of these complexes suggest that they repre- Crystals of AlF3–F1 diffracted unusually well, allowing sent transition states in NTP hydrolysis. data collection to 2.5 Å resolution. The structure was solved by molecular replacement, using the refined frozen

F1-ATPases from both mitochondria and the thermophilic native F1 structure as a starting model. Data processing Bacillus PS3 are inhibited in a quasi-irreversible manner and refinement statistics (Table 1) show that there is little by ADP-fluoroaluminate in the presence of Mg2+ [11–13]. difference in the quality of the two structures. The higher

The inhibited mitochondrial enzyme contained two ADP- resolution of the AlF3–F1 structure allowed a greater fluoroaluminate complexes per F1 domain [12]; however, number of solvent molecules to be placed in this model both enzymes were slowly and completely inactivated by relative to the frozen native model (Table 1). The only one ADP-fluoroaluminate per F1 domain [13], Ramachandran plot statistics for the two structures are although two ADP-fluoroaluminate groups were bound very similar, with 88.8% and 88.9% of residues in most when the enzymes were pre-incubated with excess ADP. favoured regions, 10.9% and 10.7% in additionally allowed The mitochondrial enzyme was also inhibited by GDP- regions, 0.3% and 0.4% in generously allowed regions and fluoroaluminate, and, as GDP binds to catalytic no residues in disallowed regions for the AlF3–F1 and nucleotide-binding sites and not to noncatalytic sites, the frozen native structures, respectively. The final models for β α α inhibition is due to binding to subunits [12]. The affin- both structures comprise residues E 24–510, TP 24–401, 2+ α β β β γ ity of Mg ADP for one catalytic site increases markedly 410–510, DP 19–510, DP 9–475, E and TP 9–474, and in the presence of fluoroaluminate in the Escherichia coli 1–44, 77–90 and 209–272. enzyme. This effect is abolished when key catalytic residues are mutated, consistent with ADP-fluoroalumi- Table 1 nate acting as a transition-state analogue [14,15]. Crystallographic analysis statistics. This paper describes the crystal structure of a form of Frozen native F1-ATPase + AlF3 bovine mitochondrial F1-ATPase inhibited by ADP- fluoroaluminate. The structure shows the presence of a Diffraction data β β single ADP-fluoroaluminate bound to the subunit ( DP) Space group P212121 P212121 identified as adopting the catalytically active conformation Cell parameters (Å) in the original F -ATPase structure [2]. The observed a, b, c 280.8, 107.4, 139.6 278.6, 106.7, 137.9 1 Resolution (Å) 2.61 2.48 mode of binding explains the very high affinity of ADP- Measured intensities 368,683 320,630 fluoroaluminate and clarifies the roles of several key Unique reflections 121,408 136,913 residues in the catalytic mechanism. Completeness (%) 94.8 (99.1) 93.8 (74.6) Multiplicity 3.0 (2.6) 2.3 (2.0) R (%) 6.1 (20.2) 9.6 (28.2) Results and discussion merge

Structure determination of the AMP-PNP-ADP form of F1 Refinement from a single frozen crystal Resolution (Å) 20–2.61 20–2.5 Rcryst (%) 23.2 21.8 The structure of bovine mitochondrial F1-ATPase Rfree (%) 28.0 28.2 described by Abrahams et al. [2] was based on X-ray data Reflections 115,942 129,652 from crystals grown in the presence of AMP-PNP and Protein atoms 22,663 22,663 Solvent atoms 543 848 ADP (native F1). These data were collected at 4°C from a total of 17 different crystals. More recently, diffraction Nucleotide atoms 160 172 data have been collected from flash-frozen crystals. It is Geometry possible, however, that the presence of the cryoprotectant Bond lengths (Å) 0.011 0.010 (20% glycerol) might result in subtle conformational Bond angles (°) 1.4 1.2 changes. Therefore, a new data set for the native F1 Values in parentheses are for the highest resolution shell. Σ Σ Σ Σ complex was collected from a single flash-frozen crystal Rmerge = h i | I(h) – I(h)i |/ h i I(h)i, where I(h)i is the mean weighted Σ Σ (frozen native F ) to a resolution of 2.6 Å. Refinement intensity after rejection of outliers. Rcryst = h|Fo –Fc |/ hFo, where 1 F and F are the observed and calculated structure-factor amplitudes, using this new data has resulted in a model that is very o c respectively, and were determined using 95% of the data. Rfree was similar to the originally reported structure (root mean determined from the residual 5% of the data. Hydrogen atoms were square deviation [rmsd] in Cα position is 0.45 Å, and in all excluded. For geometry, rmsds from ideal values are given. 2+ Research Article F1-ATPase inhibited by Mg ADP-fluoroaluminate Braig et al. 569

Figure 1

Stereoview of the nucleotide-binding site of β Arg189 Arg189 the DP subunit of the AlF3–F1 complex. Atoms are coloured yellow, red, blue, pink, αArg373 αArg373 orange, green and white for carbon, oxygen, Wat370 Wat370 nitrogen, phosphorous, magnesium, Glu188 Glu188 ADP ADP and aluminium, respectively. AlF3 is shown modelled into the positive electron density of Mg Mg σ the Fo–Fc map (contoured at 4 ) calculated excluding the AlF3 group from the model. All β Lys162 Lys162 residues shown belong to the DP subunit except αArg373, which lies on the adjacent α DP subunit. The figure was produced using Structure BOBSCRIPT [32].

Overall conformational changes motif. The second fluorine lies 2.8 Å from an NH2 group α α A comparison of the refined structures of the frozen native of Arg373 on the neighbouring DP subunit. The third F1 and AlF3–F1 complexes revealed no significant changes fluorine also interacts with the guanidinium group of in tertiary or quaternary structure. The rmsd in Cα posi- αArg373, and, in addition, the guanidinium group of tions is 0.39 Å for the entire complex, and ranges from βArg189 and the Mg2+ ion. The Al3+ ion is 2.4 Å from the α β γ β 0.28 Å ( DP and TP) to 0.58 Å ( ) when individual sub- closest oxygen of the -phosphate of ADP, and 3.1 Å from units are compared. This variation between subunits a water molecule (Wat370), which in turn is 2.4 Å and reflects the variation in average B factor, which varies from 2.8 Å from the carboxylate oxygens of the sidechain of 2 α 2 γ α β 42 Å ( DP) to 70 Å ( ). The rmsds in C positions are Glu188. Similar interactions have been observed in the broadly similar to the expected coordinate errors in a structures of other aluminofluoride NTP hydrolase com- structure determined at this resolution. plexes, in particular with the P-loop lysine, the Mg2+ ion and the guanidinium group of a catalytically important The aluminofluoride-binding site arginine residue. After the initial round of refinement using the structure of α β the frozen native F1 complex as a starting model, the most Superimposition of the DP/ DP subunits of the frozen significant feature in the difference electron-density map native structure onto those of AlF3–F1 complex reveals β was a large positive peak in the vicinity of the -phosphate only small sidechain movements associated with AlF3 β ε β group of the ADP bound to the DP subunit (Figure 1). binding (Figure 3a). The -amino group of Lys162 and β Both AlF4 and AlF3 were modelled into the density, but the guanidinium group of Arg189 move by less than α AlF3 not only gave a better fit to the density but also 0.3 Å, and the NH1 and NH2 groups of Arg373 shift by resulted in interactions with neighbouring protein atoms that were more stereochemically reasonable (Figures 1 Figure 2 and 2). In aqueous conditions with similar concentrations of nucleotide, fluoride and aluminium present the pre- αArg 373 βArg189 dominant solution species at pH 7.5 is (H2O)2(HO)AlF3, whereas at pH 6.0 it is (H2O)(HO)AlF4 [16]. This observa- + + tion correlates well with the AlFx species reported in pub- H2N NH2 H2N NH2 2.9 lished X-ray structures. Furthermore, a high-resolution Mg2+ 2.1 3.0 study of UMP/CMP-kinase complexed with ADP and alu- 2.5 2.8 O 3.0 F F 2.8 O minofluoride revealed AlF4 bound at the active site of the AMP O P O Al OH βGlu188 2.4 3.1 2 structure determined at pH 4.5, and AlF3 bound in the O 2.4 O structure determined at pH 8.5 [17]. The exception to this 2.7 F 2.7 trend is the thermosome structure [9] in which AlF3 was + reported with crystals grown at pH 5.6, although in this NH NH3 case the aluminium fluoride was introduced by soaking βGly 159 βLys162 rather than co-crystallisation. Structure

β The planar AlF group makes a number of interactions Schematic representation of the nucleotide-binding site of the DP 3 subunit of the AlF –F complex, showing the coordination of the with protein ligands, shown schematically in Figure 2. 3 1 aluminofluoride group. Possible hydrogen-bond interactions are shown One of the three is 2.7 Å from the ε-amino group as dotted lines. All distances are in Ångstroms. of βLys162, which is the conserved lysine in the P-loop 570 Structure 2000, Vol 8 No 6

Figure 3

β Stereoview of the catalytic site of the DP (a) Arg189 Arg189 subunit in the AlF3–F1 complex (coloured αArg373 αArg373 atoms and bonds) superimposed on (a) the β Wat370 Wat370 DP subunit of the frozen native F1 (grey atoms β Glu188 Glu188 and bonds) and (b) the TP subunit of the ADP ADP AlF3–F1 complex (grey atoms and bonds). In Mg Mg both cases the superposition was based on the Cα positions for residues 152–164. There is no structural equivalent to Wat370 in the Lys162 Lys162 β frozen native DP subunit. The figure was produced using BOBSCRIPT [32].

(b) Arg189 Arg189 αArg373 αArg373 Wat370 Wat370 Glu188 Glu188 ADP ADP Mg Mg

Lys162 Lys162

Structure

0.5 Å. The carboxylate oxygens of βGlu188 are displaced distances are both 2.5 Å but the reaction mechanism, by 0.6 Å and 0.8 Å, and in addition the sidechain density which involves a histidine residue, is different in this is much better defined in the AlF3–F1 complex. This is enzyme [7]). The shorter distances are also more consistent reflected in the average sidechain temperature factor of with Al–O distances found in small molecule structures. 31 Å2 compared with 58 Å2 in the frozen native structure Nevertheless, the structure represents an intermediate (the mean temperature factors for atoms in the between the substrate (ATP) bound form, as found in the β β nucleotide-binding domains of the DP subunits in the TP subunit, and the product (ADP + Pi) complex. In view 2 2 two structures are 40 Å and 55 Å , respectively). The of the small movement of the AlF3 group that would be reduction in temperature factors for the other sidechains necessary to mimic these end states of the reaction, the involved in AlF3 binding are much smaller, presumably structure described here shows many features relevant to because they are involved in binding ADP in the native the transition-state complex. In particular, residues α β β structure. Although the sidechain movements in DP and involved in ligating the AlF3 group, Lys162, Glu188, β β α DP are small, they are large enough to have a significant Arg189 and Arg373 (Figure 2) are in a position to partici- 2+ influence on the affinity of Mg ADP-AlF3 binding to the pate actively in catalysis. β DP catalytic site, in accord with the quasi-irreversible nature of the inhibition [11–13]. This view is supported by a mutagenesis study on E. coli

F1-ATPase. The fluorescence signal from a tryptophan β β → Superimposing the TP subunit of the AlF3–F1 complex mutant ( Tyr331 Trp) has been used to follow β β onto the DP subunit shows that the AlF3 group occupies a nucleotide binding to the catalytic subunits. In the pres- position close to, but not coincident with, that of the ence of AlCl3 and NaF, a marked increase in the binding γ β 2+ phosphate of AMP-PNP bound to the TP subunit affinity of Mg ADP to the high-affinity catalytic site was 2+ µ (Figure 3b). The position of the Mg ADP-AlF3 group rel- observed (Kd << 1 nM, compared with 0.07 M in the ative to the rest of the structure is shown in Figure 4. absence of fluoroaluminate) [14]. The very high affinity suggests that the Mg2+ADP-fluoroaluminate complex rep-

Does the AlF3 complex represent the transition state? resents an analogue of the transition state. The enhanced The overall geometry of the AlF3 binding resembles that binding affinity in the presence of fluoroaluminate was seen in other complexes. However, the distances between not observed when βLys155 (equivalent to βLys162 in the 3+ β the Al ion and its apical ligands, the -phosphate oxygen mitochondrial enzyme MF1) was mutated to glutamine, (2.4 Å) and particularly the water molecule (3.1 Å), are sig- consistent with a role for this residue in stabilisation of the nificantly larger than previously reported values, which transition state. Similar results were obtained for mutants β β β β typically lie in the range 2.0–2.3 Å. (An exception is in the of Glu181 ( Glu188 in MF1), Arg182 ( Arg189 in MF1) α α structure of nucleoside diphosphate kinase, in which these [15] and Arg376 ( Arg373 in MF1) [18]. 2+ Research Article F1-ATPase inhibited by Mg ADP-fluoroaluminate Braig et al. 571

Figure 4 β bound to the catalytic nucleotide-binding site of the TP β subunit and no nucleotide bound to the E subunit. These N nucleotide occupancies are the same as in the original native structure [2]. Therefore, with respect to nucleotide

occupancy, the only difference between the AlF3–F1 and frozen native structures is that in the former the N β nucleotide-binding site of the DP subunit contains 2+ 2+ Mg ADP-AlF3 instead of Mg ADP.

Stoichiometry of aluminofluorate binding 2+ When F1-ATPase is inhibited by Mg ADP and aluminium fluoride at concentrations similar to those used to prepare the enzyme for crystallisation, two molecules of aluminium

fluoride are bound per F1 domain [12,13]. This suggests that one molecule of aluminium fluoride per F1 domain has been lost as a result of the crystallisation process, although this result was accidental rather than intentional. The observation that one bound ADP-fluoroaluminate group

per F1 domain is sufficient to fully inhibit the enzyme [13] validates the current structure as a model for the transition state, and demonstrates that the single occupied site corre- β sponds to the DP subunit. Examination of the crystallisa- tion protocol reveals two steps in which loss of an C aluminofluoride group might occur. First, two rounds of C dialysis are used to decrease the concentration of ADP from 660 µM to 5 µM, while the AMP-PNP concentration is α β increased from 0 to 250 µM. Second, there is no Al3+ or F– DP DP Structure present in solution during the dialysis procedure or the sub- α β sequent crystallisation. Issartel et al. [12] have also shown Location of ADP and fluoroaluminate in the DP/ DP subunit interface. β β ADP, AlF3 and the sidechains of Lys162, Glu188 (largely hidden), that when aluminiumfluoride-inhibited F1 is left in solu- βArg189 and αArg373 are shown in ball-and-stick representation, tion for 38 days, 50% of its activity is regained and at the using the same atom-colouring scheme as in Figure 1. The figure was same time 50% of the bound ADP is lost. This supports the produced using BOBSCRIPT [32]. β hypothesis that the TP catalytic site might have contained Mg2+ADP and aluminium fluoride at the time of inhibition Both the structural and mutagenesis results suggest spe- but that during the lengthy time of the crystallisation cific roles for several residues at the catalytic site. Amino process it was displaced by AMP-PNP. Equally, the nature β β β acids Lys162 and Arg189 both contribute to stabilisa- of the ADP fluoroaluminate species bound at the DP cat- tion of the transition state and to binding ATP, the sub- alytic site might have changed during this period. This strate for the hydrolysis reaction [14]. Amino acid αArg373 change might have been influenced by the loss of ADP flu- β also contributes to transition-state stabilisation, but in this oroaluminate from the TP catalytic site because communi- case mutagenesis data suggest that it has no significant cation between catalytic sites probably has an important role in binding ATP [18]. Its role is reminiscent of the role in the cooperativity of multi-site catalysis. arginine finger in the GTPase activating proteins [8]. In addition, its location at the active site makes it an attrac- Biological implications tive candidate for communication between active sites The hydrolysis of ATP to ADP and inorganic phos- through the noncatalytic α subunits as suggested previ- phate releases the energy required to drive many meta- ously [19,20], thereby contributing to the highly coopera- bolic processes in biology. ATP synthesis is carried out tive nature of catalysis. Finally, the carboxylate group of by the F1Fo-ATP synthase complex, which uses the βGlu188 is well placed to orient and polarise (or deproto- proton-motive force generated across the membrane by nate) a water molecule for nucleophilic attack on the oxidative phosphorylation or photosynthesis to drive the γ-phosphate group of ATP. synthesis of ATP from ADP and phosphate.

Occupancy of the other nucleotide-binding sites Much of our understanding of the catalytic mechanism of

In both the frozen native and AlF3–F1 structures, the three the complex has come from a detailed study of the cat- α noncatalytic nucleotide-binding sites in the subunits alytic domain F1. This domain contains five subunits in α β γδε contain AMP-PNP. Also, both structures have AMP-PNP the ratio 3 3 , and the structure of this domain has 572 Structure 2000, Vol 8 No 6

been solved at 2.8 Å [2]. This structure revealed that the Structure solution and refinement three catalytic β subunits alternate with the three non- The frozen native structure was solved by molecular replacement with AMoRe [24] using the original F -ATPase crystal structure ([2] PDB catalytic α subunits around the centrally located γ 1 code 1BMF) as a search model. subunit. The three catalytic β subunits all adopted differ- ent conformations and nucleotide occupancy. The posi- The structure of the aluminium-fluoride-inhibited enzyme was also tion and asymmetry of the γ subunit indicated that it solved by molecular replacement, but using the refined coordinates might rotate and be able to modulate the conformations of the frozen native structure as the search model. After initial rigid- β body refinement using TNT [25], a difference electron-density map of the subunits. It was subsequently shown that hydrol- using Fourier coefficients from SIGMAA [26] revealed a large posi- γ β β ysis of ATP drives rotation of the subunit [1]. There- tive peak adjacent to the -phosphate of ADP bound to the DP fore, the three different conformations and nucleotide subunit. AlF3 was built into the difference density with the graphics occupancies of the catalytic sites observed are ‘snap- programme O [27]. shots’ that occur during a single catalytic cycle. The Both structures were refined using a combination of rigid-body refine- more snapshots we have at different stages of the cat- ment using TNT [25], and positional and temperature-factor refine- alytic cycle, the better our understanding of the molecu- ment using X-PLOR [28]. For the calculations of the Rfree value [29], lar mechanism of ATP hydrolysis or the reverse 5% of the observed diffraction data were set aside and excluded from the entire refinement including the initial rigid-body refinement. To reaction, ATP synthesis. In the structure presented here avoid bias, the same reflections were used for the free R value as in we report on a new snapshot of the catalytic site with the original structure determination. All X-PLOR refinements were 2+ σ Mg ADP and AlF3 bound. It appears to mimic the tran- carried out using data with F > 2 between 6–2.6 Å and 6–2.5 Å res- sition state of the reaction. The binding of fluoroalumi- olution for the native and the inhibited enzyme, respectively. An initial energy minimisation without the X-ray residual term but with harmonic nate does not appear to require or induce any significant restraints on Cα coordinates was used to regularise the geometry. conformational changes either at the catalytic site or This was followed by a 250-step torsion-angle molecular dynamics remote from it. This observation suggests that, once refinement at 500K, a short Cartesian coordinate molecular dynamics nucleotide is bound at the catalytic site, only small con- simulation at 300K and a final energy minimisation to find the best formational changes are required to facilitate the hydrol- local minimum. Temperatures during simulations were maintained at constant values. After refinement of an overall anisotropic B factor, ysis reaction. further refinement was carried out in alternating cycles of torsion- angle dynamics and overall isotropic B-factor refinement. Subse- Materials and methods quently a group B-factor refinement was carried out with two B Crystallisation and data collection factors per residue (for mainchain atoms and sidechain atoms) and finally individual atomic B factors were refined. This model was Nucleotide-free bovine F1-ATPase was isolated as described previously [21] in a buffer containing 100 mM Tris-sulphate (pH 8.0), 50% (v/v) refined further using the maximum-likelihood target in REFMAC [30], glycerol and 4 mM EDTA. To prepare the aluminofluoride-inhibited including an overall anisotropic temperature factor, a bulk-solvent cor- enzyme, two volumes of the protein sample (protein concentration rection and all data between 20–2.5 Å resolution. Water molecules σ 20 mg/ml) were mixed with one volume of a solution containing 10 mM were added at the positions of positive density greater than 3 in the difference electron-density maps, providing the stereochemistry was MgSO4 and 2 mM ADP. After 20 min NaF was added to a final concen- tration of 6.25 mM, and after a further 20 min AlCl was added to a con- appropriate. All water molecules with refined B factors greater than 3 2 centration of 1.25 mM. After 1 h the F -ATPase was inhibited completely. 80 Å were deleted from the model. Manual adjustments to the model 1 were made using O [27]. The quality of the refined models was Crystals of both the native enzyme and the aluminofluoride-inhibited assessed with PROCHECK [31]. enzyme were grown by microdialysis in the presence of 250 mM AMP- PNP and 5 mM ADP [21]. Crystals appeared after 2–4 weeks and The crystallographic data are summarised in Table 1. reached a maximum dimension of approximately 0.5 mm in 6–8 weeks. Accession numbers Before data collection, the crystals were stabilised by slowly increas- The coordinates of the frozen native and AlF3-inhibited enzymes have ing the glycerol concentration of the crystallisation buffer from 0 to been deposited in the Protein Data Bank with accession codes 1E1Q 20% (v/v) in 5% increments. Crystals were frozen by plunging them in and 1E1R, respectively. liquid nitrogen. Acknowledgements Data for the native crystal grown in the presence of AMP-PNP and KB and RIM were supported by an EMBO fellowship and an HFSP fellow- ADP were collected to 2.6 Å resolution on beamline ID2 at the ESRF ship, respectively. We thank the staff at the SRS and ESRF synchrotron (Grenoble, France), λ = 0.91 Å. Data for the aluminium-fluoride-inhib- sites for support. ited form were collected to 2.5 Å resolution on station 9.6 at the SRS (Daresbury, UK), λ = 0.87 Å. Both datasets were collected at 100K from a single frozen crystal, using a Mar Research image plate detec- References 1. Noji, H., Yasuda, R., Yoshida, M. & Kinoshita K.J. (1997). Direct tor. Both the native and MgADP-AlF3 crystals belonged to the space observation for rotation of F1-ATPase. Nature 386, 299-302. group P212121 with unit-cell dimensions of a = 280.8 Å, b = 107.4 Å, 2. Abrahams, J.P., Leslie, A.G.W., Lutter, R. & Walker, J.E. (1994). c = 139.6 Å, and a = 278.6 Å, b = 106.7 Å, c = 137.9 Å, respectively. Structure at 2.8 Å resolution of F1-ATPase from bovine heart The diffraction images were processed with MOSFLM [22] and further mitochondria. Nature 370, 621-628. processing was performed with programs from the Collaborative Com- 3. Coleman, D.E., Berghuis, A.M., Lee, E., Linder, M.E., Gilman, A.G. & putational Project Number 4 Suite [23]. Sprang, S.R. (1994). Structure of active conformations of Gia1 and the mechanism of GTP hydrolysis. Science 265, 1405-1412. 4. Sondek, J., Lambright, D.G., Noel, J.P., Hamm, H.E. & Sigler, P.B. (1994). GTPase mechanism of G proteins from the 1.7-Å crystal structure of transducin α•GDP•AlF4–. Nature 372, 276-279. 2+ Research Article F1-ATPase inhibited by Mg ADP-fluoroaluminate Braig et al. 573

5. Fisher, A.J. et al., & Rayment, I. (1995). X-ray structures of the myosin Haven, CT. motor domain of Dictyostelium discoideum complexed with 29. Brünger, A.T. (1992). Free R value: a novel statistical quantity for MgADP•AlF4–. Biochemistry 34, 8960-8972. assessing the accuracy of crystal structures. Nature 335, 472-475. 6. Schindelin, H., Kisker, C., Schlessman, J.L., Howard, J.B. & Douglas, 30. Murshudov, G.N., Vagin, A.A., Lebedev, A., Wilson, K.S. & Dodson, C.R. (1997). Structure of ADP•AlF4–-stabilized nitrogenase complex E.J. (1999). Efficient anisotropic refinement of macromolecular and its implications for signal transduction. Nature 387, 370-376. structures using FFT. Acta. Crystallogr. D 55, 247-255. 7. Xu, Y., Moréra, S., Janin, J. & Cherfils, J. (1997). AlF3 mimics the 31. Laskowski, R.A., MacArthur, M.W., Moss, D.S. & Thornton, J.M. transition state of protein phosphorylation in the crystal structure of (1993). A program to check the stereochemical quality of protein nucleoside diphosphate kinase and MgADP. Proc. Natl Acad. Sci. structures. J. Appl. Crystallogr. 26, 283-291. USA 94, 3579-3583. 32. Esnouf, R.M. (1997). An extensively modified version of MolScript that 8. Scheffzek, K., et al., & Wittinghofer, A. (1997). The Ras–RasGAP includes greatly enhanced coloring capabilities. J. Mol. Graph. complex: Structural basis for GTPase activation and its loss in 15, 133-138. oncogenic ras mutants. Science 277, 333-338. 9. Ditzel, L., et al., & Steinbacher, S. (1998). Crystal structure of the thermosome, the archaeal chaperonin and homolog of CCT. Cell 93, 125-138. 10. Schlichting, I. & Reinstein, J. (1997). Structures of active conformations of UMP kinase from Dictyostelium discoideum suggest phosphoryl transfer is associative. Biochemistry 36, 9290-9296. 11. Lunardi, J., et al., & Vignais, P.V. (1988). Inhibition of H+-transporting ATPase by formation of a tight nucleoside diphosphate- fluoroaluminate complex at the catalytic site. Proc. Natl Acad. Sci. USA 85, 8958-8962. 12. Issartel, J.P., Dupuis, A., Lunardi, J. & Vignais, P.V. (1991). Fluoroaluminum and fluoroberyllium nucleoside diphoshate complexes as probes of the enzymatic mechanism of the mitochondrial F1- ATPase. Biochemistry 30, 4726-4733. 13. Dou, C., Grodosky, N.B., Matsui, T., Yoshida, M. & Allison, W.S. (1997). ADP-fluoroaluminate complexes are formed cooperatively at α β γ two catalytic sites of wild-type and mutant 3 3 subcomplexes of the F1-ATPase from the thermophilic Bacillus PS3. Biochemistry 36, 3719-3727. 14. Nadanaciva, S., Weber, J. & Senior, A.E. (1999). Binding of the transition state analog MgADP-fluoroaluminate to F1-ATPase. J. Biol. Chem 274, 7052-7058. 15. Nadanaciva, S., Weber, J. & Senior, A.E. (1999). The role of β-Arg- 182, an essential catalytic site residue in Escherichia coli F1-ATPase. Biochemistry 38, 7670-7677. 16. Martin, R.B. (1996). Ternary complexes of Al3+ and F– with a third ligand. Coordination Chem. Revs 141, 23-32. 17. Schlichting, I. & Reinstein, J. (1999). pH influences fluoride coordination number of the AlFx phosphoryl transfer transition state analog. Nat. Struct. Biol. 6, 721-723. 18. Nadanaciva, S., Weber, J. & Senior, A.E. (1999). Importance of F1- ATPase residue α-Arg376 for catalytic transition state stabilisation. Biochemistry, in press. 19. Soga, S., Noumi, T., Takeyama, M., Maeda, M. & Futai, M. (1989). Mutational replacements of conserved amino acid residues in the a subunit change the catalytic properties of Escherichia coli F1-ATPase. Arch. Biochem. Biophys. 268, 643-648. 20. Turina, P., Aggeler, R., Lee, R.S.-F., Senior, A.E. & Capaldi, R.A. (1993). The cysteine introduced into the α subunit of the Escherichia coli F1-ATPase by the mutation αR376C is near the α−β subunit interface and close to a noncatalytic nucleotide binding site. J. Biol. Chem. 268, 6978-6984. 21. Lutter, R. et al., & Walker, J.E. (1993). Crystallisation of F1-ATPase from bovine heart mitochondria. J. Mol. Biol 229, 787-790. 22. Leslie, A.G.W. (1992). Recent changes to the MOSFLM package for processing film and image plate data. Joint CCP4 and ESF-EACMB Newsletter on Protein Crystallography No. 26. Daresbury Laboratory, Warrington, UK. 23. Collaborative Computational Project Number 4. (1994). The CCP4 suite: programmes for protein crystallography. Acta Crystallogr. D 50, 760-763. 24. Navaza, J. (1994). An automated package for molecular replacement. Acta Crystallogr. A 50, 157-163. 25. Tronrud, D.E., Ten Eyck, L.F. & Matthews, B.W. (1987). An efficient general-purpose least-squares refinement program for macromolecular structures. Acta Crystallogr. A 43, 489-501. 26. Read, R.J. (1986). Improved Fourier coefficients for maps using phases from partial structures with errors. Acta Crystallogr. A Because Structure with Folding & Design operates a 42, 140-149. 27. Jones, T.A., Zou, J.Y., Cowan, S.W. & Kjeldgaard, M. (1991). ‘Continuous Publication System’ for Research Papers, this Improved methods for building protein models in electron density paper has been published on the internet before being printed maps and the location of errors in these models. Acta Crystallogr. A 47, 110-119. (accessed from http://biomednet.com/cbiology/str). For 28. Brünger, A.T. (1996). X-PLOR Version 3.1 Yale University, New further information, see the explanation on the contents page.