- Email: [email protected]
F0F1-ATP synthase: general structural features of ‘ATP-engine’ and a problem on free energy transduction
Biochimica et Biophysica Acta 1458 (2000) 467^481 www.elsevier.com/locate/bba
Review
F0 F1-ATP synthase: general structural features of `ATP-engine' and a problem on free energy transduction Eiro Muneyuki a , Hiroyuki Noji b , Toyoki Amano Masasuke Yoshida a; * a
b;1
, Tomoko Masaike a ,
Research Laboratory of Resources Utilization, R-1, Tokyo Institute of Technology, 4259 Nagatsuta, Midori-ku, Yokohama 226-8503, Japan b CREST (Core Research for Evolutional Science and Technology) `Genetic Programming' Team 13. Teikyo University Biotechnology Research Center 3F, 907 Nogawa, Miyamae-ku, Kawasaki 216-0001, Japan Received 19 May 1999; received in revised form 26 August 1999; accepted 26 August 1999
1. Introduction F0 F1 -ATP synthase plays a central role in ATP synthesis during oxidative or photophosphorylation. This enzyme is special in that it utilizes the protonic electrochemical potential created by electron transfer reaction by light or respiration to drive thermodynamically unfavorable ATP synthesis as predicted by Mitchell. It can also drive uphill proton transport by hydrolyzing ATP. These properties are clearly in contrast with those of usual enzymes which promote only thermodynamically favorable reactions by lowering activation energy. In addition, this enzyme has an unusual mechanistic property that the rotation of a part of the enzyme complex relative to the rest plays an essential role in its function as predicted by Boyer. Yet this enzyme has common sequence motifs called Walker motifs A and B which exist in a large number of proteins. These proteins, which we call Walker ATPase, are `ATP-engine' which exert a variety of mechanistic works fueled by ATP. Recent
* Corresponding author. Fax: +81-45-924-5277; E-mail: [email protected] 1 Present address, Department of Biology and Geosciences, Faculty of Science, Shizuoka University, 836 Ohya, Shizuoka, 422-8529, Japan.
crystallographic studies also indicated that nucleotide binding domain in the catalytic subunit of ATP synthase has a protein folding similar to other ATP utilizing proteins. In this article, we ¢rst focus on the common aspect of the ATP utilizing proteins in relation to our previous prediction on the catalytic carboxylate. Based on the recent results, we suggest that the position of catalytic carboxylate may have some correlation with the function of the protein. Then we discuss the rotation of the Q-subunit in the K3 L3 -cylinder of the thermophilic F1 -ATPase from the viewpoint of kinetics and energetics. Finally, an attempt is made to identify some of the important questions that remain unanswered with an emphasis on the problem of free energy transduction. 2. Catalytic carboxylate in F1 -ATPase and other Walker ATPases 2.1. Catalytic carboxylate in F1 -ATPase Activation of a water molecule in the catalytic center of F1 -ATPase by a general base is required for ATP hydrolysis. Using F1 -ATPase from a thermophilic Bacillus PS3, we have demonstrated that: (1) ATPase activity was inactivated completely when Glu-190 of the L-subunit was labeled by dicyclohexylcarbodiimide (DCCD) [1]; (2) substitution of Glu-
0005-2728 / 00 / $ ^ see front matter ß 2000 Elsevier Science B.V. All rights reserved. PII: S 0 0 0 5 - 2 7 2 8 ( 0 0 ) 0 0 0 9 5 - 5
BBABIO 44854 24-5-00 Cyaan Magenta Geel Zwart
468
E. Muneyuki et al. / Biochimica et Biophysica Acta 1458 (2000) 467^481
190 by Gln resulted in the loss of ATPase activity [2] and introduction of di¡erent carboxylate-containing side chains at position 190 caused drastic loss of ATPase activity [3]; (3) the DCCD labeled F1 ATPase lost the ability to synthesize enzyme-bound ATP in the presence of dimethylsulfoxide [4]; and (4) Glu-190 might not be involved in binding of nucleotide because DCCD labeled F1 -ATPase showed normal nucleotide binding [5]. No or little contribution of this residue to nucleotide binding was more clearly shown later by the study of mutant Escherichia coli F1 -ATPase [6]. As to point 1, the equivalent mutant of E. coli F1 -ATPase was at ¢rst reported to have a weak ATPase activity [7], but the critical role of this residue was proved later [8]. Indeed, the same E. coli mutant could not catalyze even a uni-site ATP hydrolysis [9]. The catalytic role of Glu-190 as a general base in ATP hydrolysis was ¢nally veri¢ed by the crystal structure of mitochondrial F1 -ATPase in which Glu-188 of the L-subunit carrying AMPPNP (equivalent to Glu-190 of the thermophilic F1 ATPase) is located at such a position that it can activate a water molecule for an in-line attack of the Q-phosphate of nucleotide (Fig. 2A) [10]. All three L-subunits in F1 -ATPase should have Glu at position 190 to continue catalytic turnover because a chimeric (but homogeneous) mutant K3 L3 Q-complex of thermophilic F1 -ATPase, in which one of the Lsubunits has Gln at position 190, but the rest two Glu, can mediate a uni-site catalysis, but cannot propagate bi- and tri-site catalysis [11]. To be a general base, pKa value of Glu-190 should be in the neutral pH range and indeed the value pKa 6.8 was estimated from NMR titration of the carboxylate at position 190 [12]. 2.2. Previous prediction of catalytic carboxylate in Walker ATPases A large number of proteins which catalyze, or are supposed to catalyze ATP-triggered reactions have a set of Walker sequence motifs A (GXXXXGKT/S; X can be varied) and B (ZZZZD; Z is a hydrophobic residue) [13]. When the crystal structure of RecA protein, one of the Walker ATPases, was solved and its Glu-96 was assigned to be the catalytic carboxylate [14], we were surprised at the similarity of the relative location of the catalytic carboxylate of
RecA and F1 -ATPase L-subunit; 24 and 26 residues downstream from Lys of motif A, respectively. Even though the structure of F1 -ATPase was not known at that time and many F1 -ATPase researchers regarded adenylate kinase as a structural model of catalytic domain of F1 -ATPase, we predicted that the structure of catalytic domain of F1 -ATPase would resemble that of RecA [15]. This prediction soon turned out to be true when the F1 -ATPase structure was solved [10]. The prominent structural feature common to the two proteins is that the catalytic domain consists of one K-helix surrounded by six parallel L-strands and the Asp in motif B and the catalytic carboxylate are located at the exits of adjacent two L-strands. Encouraged by this success, we extended the prediction to other Walker ATPases [16]. According to the prediction, the Asp or Glu conserved in many Walker ATPases at 25 þ 2 residues downstream from Lys of motif A, should be a critical catalytic residue. Since then, mutageneses of the predicted catalytic carboxylate have been carried out for several Walker ATPases and, very recently, crystal structures of two important members in the prediction list were determined. Now we can judge the validity of the previous prediction. 2.3. RecA-F1 type ATPase SecA protein is an essential factor for protein export across the membranes in the E. coli cell. We predicted that the Asp-133, located at 24 residues from Lys in motif A, is a catalytic carboxylate. Indeed, when we replaced the Asp-133 with Asn, the mutant SecA lost protein translocation-dependent ATPase activity and did not mediate translocation [17]. Therefore, the prediction was supported and tertiary structure of the nucleotide-binding domain of SecA is most likely similar to that of RecA. We call this type of proteins RecA-F1 protein, a subfamily of Walker ATPases. 2.4. AAA + protein We have also tested the prediction for Lon protease, an ATP-dependent intracellular protease in E. coli. The mutant, in which the predicted catalytic carboxylate, Asp-386, was replaced with Asn, showed no di¡erence from the wild-type; fully active
BBABIO 44854 24-5-00 Cyaan Magenta Geel Zwart
E. Muneyuki et al. / Biochimica et Biophysica Acta 1458 (2000) 467^481
in ATPase and ATP-dependent protease (T. Amano, M. Yoshida, K. Takahashi, T. Muramatsu, unpublished result). The proteins grouped into putative ATPases in the previous list came to be called AAA family of ATPases, and recently, the extended version of AAA protein family called AAA protein family which includes Lon protease has been proposed [18]. AAA protein may not use the previously predicted carboxylate as a catalytic residue. 2.5. ABC transporter Proteins in the family of ATP-binding cassette (ABC) transporters have Walker motifs A and B, and we indicated two or more candidates for the catalytic carboxylate. For instance, we predicted that Glu-64 and Glu-94 of MalK, ATP-hydrolyzing subunit of Salmonella typhimurium ABC transporter, were catalytic carboxylates [16]. However, Stein et al. reported that substitution of both residues of MalK with Gln had no deleterious e¡ect on maltose transport [19]. Ueda and his colleagues replaced the predicted carboxylates of human P-glycoprotein (MDR1) with corresponding Gln or Asn, but ATPase activity and drug-transport activity of the mutants were not impaired (K. Ueda, personal communication). 2.6. Catalytic carboxylate of AAA + protein and ABC transporter Our prediction on the location of catalytic carboxylate in AAA protein and ABC transporter has turned out to invalid. Therefore, we omit AAA protein and ABC transporters from our previous list and propose a revised list as in Fig. 1. Then, where is a catalytic carboxylate in AAA protein and ABC transporter? Recent reports on crystal structures of a member of AAA protein (NSF D2 domain, Fig. 2B) [20] and ABC transporter (HisP, Fig. 2C) [21] have provided a likely answer. The structures of nucleotide binding domains of these two proteins have an apparent structural similarity to the RecA-F1 protein, that is, they are composed of a central K-helix (Fig. 2, yellow) and a surrounding `wall' of L-sheet made up from at least ¢ve parallel Lstrands (green). Their order in the primary sequences is L1 K1 VL2 VL3 VL4 VL5 as schematically shown in
469
Fig. 2D. Walker A motif is present in a loop connecting L1 and K1 (P-loop, blue). Walker B motif is located in L3 and the Asp of ZZZZD is at its exit. The L1 -strand is sandwiched by L4 and L5 in the wall of the L-sheet. Behind the wall each of the loops connecting two L-strands, forms one or several Khelices which function as conformational transducers or switches responding nucleotide turnover occurring at the front side of the wall. These structural features are typically observed in AAA protein. RecA-F1 type protein and ABC transporter have additional unique characteristics; RecA-F1 protein has an additional parallel L-strand beside L2 , making the total number of parallel L-strands in the wall six, and ABC transporter has a large insert domain between K1 and L2 . As described, the catalytic carboxylate, Glu-188, exists at the exit of L2 in mitochondrial F1 -ATPase (Fig. 2A). Contrary, in the case of NSF-D2 (AAA protein) and HisP (ABC transporter), the Asp-612 and Glu-179, adjacent residues to the Walker B motif are positioned at appropriate sites to act as a general base in ATP hydrolysis. Consistently, the sequence, ZZZZD-D/E, is well conserved in AAA protein and ABC transporter. In F1 -ATPase L-subunit and RecA, the corresponding residues are Asn and Ser, respectively. 2.7. Other variations of Walker ATPase We included nitrogenase Fe protein (NifH), a protein catalyzing ATP dependent reduction of N2 to ammonia, in our previous list. Its Asp-43 was predicted as a catalytic carboxylate. In the crystal structure of the nitrogenase complex (Fe-protein and MoFe-protein) from Azotobacter vinelandii stabilized by ADPAlF3 4 , the predicted Asp-43 (designated as Asp-39 in the literature) is really very close to the aluminum atom [22] and therefore a strong candidate for the catalytic carboxylate. However, it should be pointed out that, di¡erent from the prototype folding topology in Fig. 2D, the L1 -strand in the Fe-protein is sandwiched by L3 and L4 , rather than L4 and L5 . Another variation of Walker ATPase is the DExx box helicase. The crystal structure of catalytic domain of the helicase is very similar to that of RecA [23]. Nonetheless, like AAA protein and ABC protein, Glu-224, a residue next to Walker motif B, is assigned to be a catalytic carboxylate.
BBABIO 44854 24-5-00 Cyaan Magenta Geel Zwart
470 E. Muneyuki et al. / Biochimica et Biophysica Acta 1458 (2000) 467^481
BBABIO 44854 24-5-00 Cyaan Magenta Geel Zwart Fig. 1. Revised list of putative RecA-F1 subfamily. Lys in Walker motif A, putative catalytic carboxylate, and Asp in Walker motif B are shown by boxed characters.
E. Muneyuki et al. / Biochimica et Biophysica Acta 1458 (2000) 467^481
2.8. Mechanistic implication Even belonging to the same Walker ATPase, RecA-F1 protein utilizes a di¡erent carboxylate as a catalytic base when compared with AAA protein and ABC transporter. This might be a re£ection of the possible unique function of RecA-F1 protein; it can push a large molecule to one direction. F1 -ATPase L-subunit undergoes a large open-to-close domain movement during ATPase turnover and the concerted movement of three L-subunits in F1 -ATPase drives the central shaft Q-subunit to one direction, that is, rotation. RecA protein can transform the helical pitch of DNA chain and can rotate DNA chain. SecA protein holds a polypeptide stretch of the protein to be secreted, carries it to the opposite side of the membrane by inserting the arm domain, releases the polypeptide, deinserts the arm from membrane, holds next stretch, and repeats the cycle. On the contrary, the motion of other two classes might be less dynamic, though possibly more ¢nely regulated. AAA protein might work as a chaperone-like ATPase associated with the assembly and disassembly of protein complexes. ABC transporters, present at membranes, act as pumps, as regulators of other membrane proteins, and as channels. The contention described here is more speculative than likely, but accumulation of knowledge on mechanism of each protein may give a mechanistic reason of the di¡erence. 3. Rotation of the Q-subunit in relation to the kinetics and energetics 3.1. The rotation of the Q-subunit couples the ATP hydrolysis/synthesis by F1 and proton translocation by F0 In order to achieve e¤cient energy transduction, energy transducing proteins must have a mechanism to transmit the energy released by the exergonic reaction such as ATP hydrolysis to drive the endergonic reaction, such as active transport or mechanical work. As the ¢rst step, it is necessary to conserve the energy released at the active site. Keeping the energy from rapid dissipation by protein molecule is by itself an important subject [24]. Then the energy
471
must be utilized to drive the coupled endergonic reaction which takes place at another distant active site. The rotation of the Q-subunit within the K3 L3 -cylinder of the F1 -ATPase was ¢rst visualized by Noji et al. as a rotation of £uorescently labeled actin ¢lament attached to the Q-subunit (Fig. 3) [25]. The actin ¢laments rotated invariably counter-clockwise when viewed from above in Fig. 3, which is consistent with the direction predicted by X-ray crystal structure [10] and Boyer's Binding Change Mechanism [26^28]. This rotation is a part of the whole reaction coupling ATP hydrolysis and active transport by F0 F1 holoenzyme, yet it is very important in view of the fact that we can directly observe the individual event of the energy transduction (coupling of ATP hydrolysis to mechanical rotation) by a single molecule. The observed rotations were always under the in£uence of hydrodynamic friction against the rotating actin ¢lament, but there is further advantage of the use of actin ¢lament in that we can clearly identify the rotation and calculate the torque. The detailed analyses of the rotation will provide insights into the general aspects of the energy transduction by proteins. In the following part of this article, we will describe the characteristics of the rotation of the Qsubunit observed in single molecule experiments in relation to the knowledge obtained in the usual biochemical experiments in kinetics and energetics and what should be experimentally shown in the future. 3.2. Kinetic aspects of the ATP hydrolysis and ATP hydrolysis-driven rotation of the Q-subunit The kinetic scheme of ATP hydrolysis by F1 - or F0 F1 -ATPase may be written as in Scheme 1. There are negative substrate binding cooperativity (Km1 6 Km2 6 Km3 ) and positive catalytic cooperativity (Vmax1 6 Vmax2 6 Vmax3 ). In total, F1 - and F0 F1 ATPase exhibit apparent negative cooperativity which is characterized by an upward concave double reciprocal plot or downward concave Eadie^Hofstee plot [29^34]. The reaction in the ¢rst line in Scheme 1 is known as `uni-site catalysis' [32,35]. The a¤nity to the substrate ATP in this step is extremely high (Kd = 10312 M in the case of MF1 [32,35^37], 1039 ^ 10310 M in the case of EF1 [38^41]) and the turnover rate (as designated by k3 ) is extremely low [32,35^41].
BBABIO 44854 24-5-00 Cyaan Magenta Geel Zwart
472
E. Muneyuki et al. / Biochimica et Biophysica Acta 1458 (2000) 467^481
Fig. 2. Nucleotide binding domains of three subfamilies of Walker ATPase. The central helix is shown by yellow, ¢ve L-strands by green, and P-loop by blue. The catalytic carboxylates are indicated. (A) RecA-F1 protein ; bovine mitochondrial F1 -ATPase L-subunit with bound AMP-PNP [10]. (B) AAA protein; Chinese hamster ovary NSF D2 domain with bound AMP^PNP [20]. (C) ABC transporter; Salmonella typhimurium HisP with bound ATP [21]. (D) A schematic model of nucleotide binding domain of Walker ATPase.
BBABIO 44854 24-5-00 Cyaan Magenta Geel Zwart
E. Muneyuki et al. / Biochimica et Biophysica Acta 1458 (2000) 467^481
Fig. 2 (continued).
BBABIO 44854 24-5-00 Cyaan Magenta Geel Zwart
473
474
E. Muneyuki et al. / Biochimica et Biophysica Acta 1458 (2000) 467^481
Fig. 3. The experimental system used for the observation of the Q-subunit in the K3 L3 Q-subcomplex of thermophilic F1 -ATPase. A Histag composed of 10 His residues was introduced at the N-terminal of each L-subunit in order to immobilize the complex on a NiNTA bead. A Cys residue was introduced to the Q-subunit and modi¢ed by biotinylated maleimide. Biotinylated and £uorescently labeled actin ¢lament were connected to the Q-subunit via streptavidin. The rotation of the Q-subunit was observed as a rotation of the actin ¢lament under a £uorescent microscope [25,50].
As ATP concentration increases, the reaction pathways shown in the second and third line become dominant. They are called bi-site and tri-site catalysis, respectively [32]. The apparent Km values for the bi-site catalysis and tri-site catalysis are estimated to be 1^30 and 100^300 WM [29^34]. They seem to re£ect ATP binding to the catalytic sites, but not to the non-catalytic sites, because a mutated enzyme, which completely lacks nucleotide binding at non-catalytic nucleotide binding sites, also showed typical uni-site catalysis using TNP^ATP as a substrate and apparent negative cooperativity with Km values of 4 and 135 WM in ATP hydrolysis [42]. A single turnover of the uni-site catalysis may not accompany the rotation of the Q-subunit because cross-linking between the Q- and L-subunits does not inhibit the single turnover of uni-site catalysis
[43]. This contention is supported by the results that a single turnover of the uni-site catalysis was not suppressed when the conformational change of two L-subunits was suppressed by cross-linking [44] and that K^Q cross-linking did not a¡ect the ATP binding to the single high a¤nity site [45]. On the other hand, it was previously shown that the bi-site catalytic cycle was coupled to proton translocation [46]. Under the bi-site condition where two catalytic sites are simultaneously occupied by Mg adenine nucleotide, the inhibition by NaN3 [47], and conversion of the enzyme to the MgADP inhibited form [42] are promoted. Bi-site catalytic cycle may accompany large structural changes because transition from uni-site to bi-site catalysis promotes product release from the ¢rst high a¤nity site and hydrolytic reaction on the enzyme (chase promotion). Recently, a
BBABIO 44854 24-5-00 Cyaan Magenta Geel Zwart
E. Muneyuki et al. / Biochimica et Biophysica Acta 1458 (2000) 467^481
Scheme 1. The ¢rst, second and the third lines represent unisite, bi-site, and tri-site catalysis, respectively. Km1 is de¢ned as (k2+k3)/k1. Vmax1 is de¢ned as (total enzyme) k3. Km2 , Km3 , Vmax2 and Vmax3 are similarly de¢ned. b
conformational change involving a direct contact between two L-subunits by simultaneous occupation of two Mg adenine nucleotide was clearly shown by cross-linking between genetically introduced Cys residues [44]. The catalytic cycle where three catalytic sites are operating is called tri-site catalysis (third line in Scheme 1). Full hydrolytic activity is attained in the tri-site catalysis [48,45]. There is little doubt about the coupling of the tri-site cycle to proton translocation [46,49]. In the single molecule experiments, the rotation of the Q-subunit occurred even at 20 nM ATP where bi-site catalytic cycle predominantly operates [50]. This ¢nding is consistent with the data on proton translocation. Due to the attenuation of the rotational rate by long actin ¢lament, coupling of the tri-site catalysis with the rotation of the Q-subunit has not yet been con¢rmed, but it
475
seems most likely that the rotation should occur in the tri-site cycle. As stated above, both bi-site and tri-site cycles are coupled with proton translocation and hence with the rotation of the Q-subunit. However, which step in the catalytic cycle is coupled to the proton translocation or rotation is not clear. Extensive analyses of oxygen exchange reactions and the e¡ect of uncouplers on the exchange reactions suggested that the step of a¤nity change of the catalytic sites (conversion of the high a¤nity site to the low a¤nity site and low a¤nity site to the high a¤nity site) is coupled to energy transduction [51^53] rather than the cleavage or formation of the chemical bond. Consistently, spontaneous formation of bound ATP in the absence of membrane energization [54,55] and dissociation of tightly bound nucleotide upon membrane energization [56] was reported, but the implication of the latter observations is not very clear due to the MgADP inhibition and activation caused by membrane energization [57]. It has been shown that the Km of tri-site catalysis is not a¡ected by membrane potential [58], but Vmax is a¡ected [59]. Similar results were obtained using proteoliposomes co-reconstituted with bacteriorhodopsin [60,61]. A vpH clamp method was applied for analysis of steadystate kinetics of photophosphorylation [62]. In their study, Km of ATP in ATP hydrolysis and Km of ADP in ATP synthesis were found to be insensitive to the change of vpH [63].
Fig. 4. Stepwise rotation at 20 nM ATP. (A) Revolutions versus time. It can be seen that the unitary step of the rotation is 120³. (B) Trace of the centroid of the actin image. The triangular distribution indicates that the actin ¢lament tends to stop with 120³ intervals [50].
BBABIO 44854 24-5-00 Cyaan Magenta Geel Zwart
476
E. Muneyuki et al. / Biochimica et Biophysica Acta 1458 (2000) 467^481
In the single molecule experiment, at low ATP concentrations (sub-WM ATP), the rotational rates were limited by the rate of ATP binding rather than the hydrodynamic friction against the rotating actin ¢lament. Under this condition, actin ¢lament showed clear 120³ steps during rotation (Fig. 4) [50]. The analysis of dwell times between steps indicated that the onset of the step motion was a ¢rst order reaction with a rate constant consistent with the rate of ATP binding in solution. From this analysis, it was suggested that one ATP caused one step. This result indicates that the rate of ATP binding is not signi¢cantly attenuated by the load of actin ¢lament in the rotational experiment under bi-site condition. Together with the observations that protonic electrochemical potential did not a¡ect the apparent Km for ATP in ATP hydrolysis cited above, it seems that ATP binding per se is not directly coupled to proton translocation or rotation of the Q-subunit in ATPase reaction. 3.3. Slow transitions between distinct enzymatic states are directly observed by the single molecule technique The single molecule experiment has a potential to discern distinct states of individual molecules which are essentially indistinguishable in ensemble-averaged experiments. By this technique, slow £uctuation of enzymatic reaction rate in cholesterol oxidase was detected and analyzed [64]. In the case of F0 F1 and F1 -ATPase, slow conversion between active and inactive states has been known for a long time [65]. The inactive form of the ATPase is generated from enzyme with bound MgADP [66^71] in a catalytic site [72]. Due to this inhibition, the rate of ATP hydrolysis measured in the presence of ATP regenerating system decays to approximately less than half of the initial uninhibited rate within several minutes. In ensemble-averaged experiments, it is di¤cult to conclude whether this half-inhibited state stands for the equilibrium between completely active and completely inactive enzymes or the rate of individual enzyme molecule is attenuated to the half of its initial value. Single molecule experiments may give a clear answer on this point. If the latter is the case, the rate of rotation would decrease with time. If the former is the case, the continuous rotation would be inter-
vened by occasional stops as the progress of the inhibition. At present, the former possibility seems more probable. Slow state-transition of a protein molecule which is clari¢ed in the single molecule experiments may have some mechanistically and physiologically important roles. 3.4. How does the energy transduction occur in the F0 channel? Recently, it was shown that membrane potential was essential in ATP synthesis and Na concentration gradient exerted minor in£uence in Na translocating F0 F1 -ATP synthase from Propionigenium modestum [73]. Similar results were reported for E. coli enzyme [74] and the results were discussed in relation to the putative rotation of the ring structure of the c subunits [73]. In the case of chloroplast enzyme, the rate of ATP synthesis by vpH was shown to be dependent on the pKa of the proton binding sites facing inside and outside of reconstituted proteoliposomes [75]. Does it mean, however, that the concentration gradient of the transported ion cannot serve as an energy source for ATP synthesis in the absence of membrane potential? This is a special case of a more general question asking how an energy transducing protein can utilize free energy of entropic origin (vpH) as well as enthalpic origin (v8). It is known that £agellar motor can rotate utilizing the proton concentration gradient as well as membrane potential as the energy source [76]. These components are thermodynamically equivalent. Therefore, they must be equivalent as the energy source for an energy-transducing protein even if they are kinetically inequivalent. At least, at the equilibrium point of ATP synthesis and hydrolysis, membrane potential and vpH must be completely interchangeable. In order to understand the transduction of the energy of entropic origin by F0 F1 -ATP synthase at a molecular level, the essential role of thermal £uctuation should be considered [77] in relation to the function of F0 portion. The rotation of the c-subunit ring of the F0 portion has so far never been directly observed. When the individual step motion of F0 portion driven by vpH and/or v8 is observed, the results will give a hint on the above question. Recently, 5.3-nm substeps were resolved during a
BBABIO 44854 24-5-00 Cyaan Magenta Geel Zwart
E. Muneyuki et al. / Biochimica et Biophysica Acta 1458 (2000) 467^481
477
motion of a single myosin head along an actin ¢lament [24]. It was concluded that hydrolysis of one ATP caused multiple steps. This observation is very interesting when we think of the coupling between F1 and F0 . In the K3 L3 Q-subcomplex, the rotation of the Q-subunit showed clear 120³ steps corresponding to the three KL pairs at low ATP concentration. When the rotation coupled to the ring structure of F0 portion containing 10^12 c-subunits is observed, the energy derived from one ATP hydrolysis may be divided to drive substeps corresponding to each c-subunit. 3.5. Mechanics and energetics of the rotation of the Q-subunit The most prominent advantage of the direct observation of the rotating Q-subunit using actin ¢lament is that we can estimate the force and energy transduced from ATP hydrolysis, particularly, in a single event. Unique information can be obtained by these single molecule experiments. At high ATP concentrations, the rotational rates were limited by the hydrodynamic friction against the rotating actin ¢lament and the rotation was slower with longer actin ¢laments. When the rotational rates were plotted against the actin length, the highest rates which were most reliable at each actin length were on the iso-torque line corresponding to 40 pN nm (Fig. 5) [50]. Thus the K3 L3 Q-subcomplex produced constant force irrespective of the load. This is in contrast with the case of kinesin whose force decreases as load increases. In the case of the K3 L3 Qsubcomplex^actin ¢lament system, the torque may be ¢rst stored in a softest elastic element in the system and then released to induce the rotational motion of the actin ¢lament. The force was calculated to be as high as 40 pN from the torque on the assumption that it was generated at the L^Q-interface [50]. This value was by far the greatest among known molecular motors, such as myosin (3^6 pN) [78^80] and kinesin (5 pN) [81]. At low ATP concentrations, the rotational rates were limited by the rate of ATP binding as stated above. Under this condition (sub-WM ATP), actin ¢lament showed clear 120³ steps during rotation (Fig. 4) [50]. The discrete 120³ steps suggest that there is only a little slip between the Q- and K3 L3 -
Fig. 5. Load dependence of the rotational rate. Average rotational rates versus the length of actin ¢laments are shown. White circles, [ATP] = 2 mM; ¢lled circles, [ATP] = 2 mM, [ADP] = 10 WM, and [Pi] = 0.1 mM (calculated vGATP = 110 pN nm); ¢lled rectangles, [ATP] = 2 mM, [ADP] = 10 WM, and [Pi] = 10 mM (calculated vGATP = 90 pN nm). The curves in the ¢gure stand for iso-torque lines. In spite of the substantial scattering of the data, the experimental points distribute around the iso-torque line of 40 pN nm. The torque of 40 pN nm corresponds to the work of 80 pN nm by one ATP hydrolysis (40 pN nmW2Z/3 [ATP] = 80 pN nm).
cylinder. The rate of the rotation within one step seemed constant, suggesting that the K3 L3 -cylinder exerted a constant force during the travel of the Qsubunit over 120³. The energy used for the rotation of the actin ¢lament within one step was calculated to be about 80 pN nm, which was consistent with the value calculated from the rate of the continuous rotation observed at high ATP concentrations [50]. This value was virtually independent of the co-existence of ADP and Pi provided that the free energy of ATP hydrolysis was above 90 pN nm (Fig. 5). Comparing with the free energy change of ATP hydrolysis in vivo, it was suggested that the e¤ciency of the energy conversion from ATP hydrolysis to rotational motion under physiological conditions was nearly 100%. This is much higher than the e¤ciencies estimated for other molecular motors [82]. 3.6. What is the relationship between the input energy and output energy? Then, what happens if the free energy provided by one ATP hydrolysis was signi¢cantly lower than 80
BBABIO 44854 24-5-00 Cyaan Magenta Geel Zwart
478
E. Muneyuki et al. / Biochimica et Biophysica Acta 1458 (2000) 467^481
pN nm, but greater than 0? If the angular velocity during 120³ step was constant to yield a work of 80 pN nm, it seems as if this situation would violate the second law of thermodynamics. There are at least ¢ve possibilities. (1) The energy of the rotation of the actin ¢lament over 120³ step is constantly 80 pN nm, and the number of ATP hydrolysis per one step increases (variable stoichiometry between the ¢xed mechanical output and the number of ATP). (2) The mechanical work of the rotation of the Q-subunit depends on the free energy of ATP hydrolysis and therefore continuously changes (variable output energy). (3) Both the mechanical output and the number of ATP hydrolyzed in an unitary mechanical step change. (4) The rotation simply stops when the vGATP is lower than 80 pN nm. Under this condition, ATPase also stops and there is no problem of energy balance. (5) It might be possible that when
one ATP is hydrolyzed, 80 pN nm of the mechanical work is constantly derived. In this case, the supplementary energy is derived by thermal £uctuation and the energy and entropy balance is satis¢ed by the dissipation of energy through the rotation of actin ¢lament. In any case of the ¢ve possibilities, the second law of thermodynamics is seemingly satis¢ed, but the ¢rst four possibilities have common problems of how a protein molecule senses the concentration of ATP, ADP, and Pi whose ratio determines the free energy of ATP hydrolysis. In addition, the problem of how to switch the stoichiometry or output energy depending on the vGATP emerges. The problem of what determines the reversal point of ATP hydrolysis and ATP synthesis in the holoenzyme, F0 F1 also emerges in the ¢rst to third cases. These problems are complicated because the mechanical work of
Fig. 6. Schematic illustrations of a potential-transition model. U1(x) (red line) and U2(x) (blue line) describe the potential energy pro¢le of the de-energized and energized states, respectively. The abscissa is regarded as the reaction coordinate of interest (angle, translational coordinate, etc.) The ordinate is internal energy. (A) Biased di¡usion model (thermal ratchet). One of the potential pro¢les (U2(x)) is completely £at. P(x) in black line describes the probability density of a particle (a part of a motor protein complex). When the system is de-energized (potential pro¢le U1(x)), the probability density localizes at the n-th valley of U1(x) (lower P(x)). Upon transition to the energized state (potential pro¢le U2(x)), the probability density distributes widely on the £at potential pro¢le due to thermal di¡usion. When the system is de-energized again, the fraction located in the right part on U2(x) (shaded in blue in upper P(x)) will most probably redistributes to the (n+1)-th valley on U1(x) (shaded in blue in lower P(x)). Successive transitions between de-energized and energized states will drive the probability density to the right side. (B) Power stroke model. Both U1(x) (red line) and U2(x) (blue line) have tooth-like shapes whose phases are di¡erent (B). Successive transitions drive the particle to the right according to the slope of the potentials. There are many variations di¡ering in the shapes of potential pro¢les and the modulations of transition frequency [83^85,89]. In order to induce unidirectional motion, one of the potential pro¢les before or after the transitions must be asymmetric along the reaction coordinate. The transition between the two states must deviate the distribution of the two states from the Boltzmann distribution [89]. In these models, the energy input is the elevation of the potential of the particle by the transition of the potential pro¢le and the output of the energy is the downhill motion of the particle along the slope of the potential pro¢le. The force acting on the particle is determined by the slope of the potential.
BBABIO 44854 24-5-00 Cyaan Magenta Geel Zwart
E. Muneyuki et al. / Biochimica et Biophysica Acta 1458 (2000) 467^481
the rotation of the Q-subunit is not conservative, but dissipative. Simple application of thermodynamics without distinction of the conservative and dissipative work may yield a nonsense answer. When a clear answer is given experimentally, it would contribute to the understanding of energy transduction by proteins on the molecular level. 3.7. General problem of free energy transduction by a single molecule The answer to the above questions should be experimentally given in the future. Before then, it may merit to consider some general models to understand the problem of energy transduction. One such models assumes two states which have di¡erent potentials of interaction between two components of a motor protein complex (see Fig. 6) [83^89]. According to this type of model, unidirectional motion of a particle (a protein component of the motor complex) is induced by the transitions of the potential of interaction (see Fig. 6). In one extreme case of the models, one of the potential pro¢les is completely £at and the directional motion is induced by biased thermal di¡usion (Fig. 6A). This case is often called `thermal ratchet'. The other extreme case assumes two tooth-like potentials whose phases are di¡erent (Fig. 6B). This case is called `power stroke model'. In this case, the directional motion requires no di¡usive step. At the molecular level, completely £at potential may not exist and thermal £uctuation cannot be avoided. Therefore the real situation is somewhere between these extreme cases. There are some examples of the application of such models to explain properties of the molecular motors [84,89,90,91] and active ion transport [92]. One of the most elaborate models to describe the rotation of F1 -ATPase was developed by Wang and Oster [93]. They calculated several potential pro¢les. By making appropriate assumptions on the potential pro¢les and the probabilities of transitions, they could reproduce properties of the molecular motion of the Q-subunit known at that time. They included vGATP in the potential pro¢le (Fig. 3B in [93]), which is in a way a little di¡erent from Fig. 6 here. Fig. 6 represents only the internal energy which is independent of vGATP . In their case, the energy input
479
was the vertical distance between the two potential pro¢les of empty L-subunit. This was vGATP by definition. The output energy was the downhill motion of the particle on the slope of the potentials. Therefore, the e¤ciency of the energy transduction could be easily up to 100% by adjusting vGATP . By including vGATP in the potential pro¢le according to the de¢nition of thermodynamics, they could avoid the question if it is possible for a motor protein to sense the free energy of ATP hydrolysis or not. However, thermodynamics is based on averaging on time or ensemble of many particles. On the other hand, we are now observing a single event by a single molecule. If a protein can sense the concentration of substrates and re£ect the concentrations in a single event, it must spend some time to interact with the substrates and keep the memory of many interactions. As substrate concentrations decrease, the time required to sense the concentrations becomes longer. After all, the question raised in the previous section `How does the Q-subunit rotate when vGATP is less than 80 pN nm?' seems to remain here. We feel that the potential pro¢les of internal energy (Fig. 6) are more helpful than the potential pro¢les including vGATP in understanding the individual events carried out by a single molecule. In that case (Fig. 6), the e¡ect of vGATP appears as a change in the probability of transitions between the potential states through the concentration of ATP, ADP and Pi. The concept that the molecular motion of motor proteins or active transport systems can be explained as a progress along £uctuating potential surfaces is very intriguing and important, but it should be noted that the de¢nitions of the potential pro¢les of di¡erent models are not always the same. In addition, it is very important to discriminate conservative work and dissipative work when we compare the theory and real experimental results obtained by observing a single event of a single protein molecule and try to understand the free energy transduction. 4. Epilogue In this article, we ¢rst focused on the catalytic carboxylate and structural features of F0 F1 -ATPase in relation to the other ATP utilizing proteins which
BBABIO 44854 24-5-00 Cyaan Magenta Geel Zwart
480
E. Muneyuki et al. / Biochimica et Biophysica Acta 1458 (2000) 467^481
have Walker motifs. Then, we surveyed the present knowledge on the rotation of the Q-subunit in the K3 L3 Q-subcomplex of F1 -ATPase and kinetics and some structural change. So far, the relationship between them is hardly established. One of the most important future problems is to make a clear framework which accommodates the kinetics, structural changes, and rotation of the Q-subunit. It is needless to say that the proton translocation (and the rotation of F0 subunits) should be involved ultimately in the framework. In trying to highlight the unanswered questions on the free energy transduction by protein molecules, we took some liberties in speculating on possible mechanisms or models in the last part. The conclusion and establishment of any concept should await experimental results, but we hope these speculations would stimulate further discussion and alternative hypothesis, and are ¢nally veri¢ed by experiments directly observing the single molecules. The single molecule observation technique will open a new romantic era in the ¢eld of bioenergetics. Acknowledgements We are grateful to Dr. Sung-Hou Kim for providing us with the coordinate of HisP structure. We thank Dr. Kazuhiko Kinosita and George Oster for helpful comments on this article.
References [1] M. Yoshida, J.W. Poser, W.S. Allison, F.S. Esch, J. Biol. Chem. 256 (1981) 148^153. [2] M. Ohtsubo, M. Yoshida, S. Ohta, Y. Kagawa, M. Yohda, T. Date, Biochem. Biophys. Res. Commun. 146 (1987) 705^ 710. [3] T. Amano, K. Tozawa, M. Yohsida, H. Murakami, FEBS Lett. 348 (1994) 93^98. [4] M. Yoshida, W.S. Allison, J. Biol. Chem. 261 (1986) 5714^ 5721. [5] M. Yoshida, W.S. Allison, J. Biol. Chem. 258 (1983) 14407^ 14412. [6] S. Lobau, J. Weber, S. Wilke-Mounts, A.E. Senior, J. Biol. Chem. 272 (1997) 3648^3656. [7] K. Parsonage, S. Wilke-Mounts, A.E. Senior, Arch. Biochem. Biophys. 261 (1988) 222^225.
[8] A.E. Senior, M.K. Al-Shawi, J. Biol. Chem. 267 (1992) 21471^21478. [9] M. Futai, M.-Y. Park, A. Iwamoto, H. Omote, M. Maeda, Biochim. Biophys. Acta 1187 (1994) 165^170. [10] J.P. Abrahams, A.G.W. Leslie, R. Lutter, J.E. Walker, Nature 370 (1994) 621^628. [11] T. Amano, T. Hisabori, E. Muneyuki, M. Yoshida, J. Biol. Chem. 271 (1996) 17343^17348. [12] K. Tozawa, H. Ohbuchi, H. Yagi, T. Amano, T. Matsui, M. Yoshida, H. Akutsu, FEBS Lett. 397 (1996) 122^126. [13] J.E. Walker, M. Saraste, M.J. Runswick, N.J. Gay, EMBO J. 1 (1982) 945^951. [14] R.M. Story, T.A. Steiz, Nature 355 (1992) 374^376. [15] T. Amano, M. Yoshida, Y. Matsuo, K. Nishikawa, FEBS Lett. 351 (1994) 1^5. [16] M. Yoshida, T. Amano, FEBS Lett. 359 (1994) 1^5. [17] K. Sato, H. Mori, M. Yoshida, S. Mizushima, J. Biol. Chem. 271 (1996) 17439^17444. [18] A.F. Neuwald, L. Aravind, J.L. Spouge, E.V. Koonin, Genone Res. 9 (1999) 27^43. [19] A. Stein, S. Hunke, E. Schneider, FEBS Lett. 18 (1997) 211^ 214. [20] C.U. Lenzen, D. Steinmann, S.W. Whiteheart, W.I. Weis, Cell 94 (1998) 525^536. [21] L.-W. Hung, I.X. Wang, K. Nikaido, P.-Q. Liu, G.F.-L. Ames, S.-H. Kim, Nature 396 (1998) 703^707. [22] H. Schindelin, C. Kisker, J.L. Schlessman, J.B. Howard, D.C. Rees, Nature 387 (1997) 370^376. [23] H.S. Subramanya, L.E. Bird, J.A. Brannigan, D.B. Wigley, Nature 384 (1996) 379^384. [24] K. Kitamura, M. Tokunaga, A. Hikikoshi-lwane, T. Yanagida, Nature 397 (1999) 129^134. [25] H. Noji, R. Yasuda, M. Yoshida, K. Kinosita Jr., Nature 386 (1997) 299^302. [26] P.D. Boyer, FASEB J. 3 (1989) 2164^2178. [27] P.D. Boyer, Biochim. Biophys. Acta 1140 (1993) 215^250. [28] P.D. Boyer, Annu. Rev. Biochem. 66 (1997) 717^749. [29] S.D. Dunn, V.D. Zadorozny, R.G. Tozer, L.E. Orr, Biochemistry 26 (1987) 4488^4493. [30] M.J. Gresser, J.A. Myers, P.D. Boyer, J. Biol. Chem. 257 (1982) 12030^12038. [31] S.-Y. Wong, A. Matsuno-Yagi, Y. Hate¢, Biochemistry 23 (1984) 5004^5009. [32] R.L. Cross, C. Grubmeyer, H.S. Penefsky, J. Biol. Chem. 257 (1982) 12101^12105. [33] Y. Kato, T. Sasayama, E. Muneyuki, M. Yoshida, Biochim. Biophys. Acta 1231 (1995) 275^281. [34] C. Matsuda, E. Muneyuki, H. Endo, M. Yoshida, Y. Kagawa, Biochem. Biophys. Res. Commun. 200 (1994) 671^678. [35] C. Grubmeyer, R.L. Cross, H.S. Penefsky, J. Biol. Chem. 257 (1982) 12092^12100. [36] H.S. Penefsky, Methods Enzymol. 126 (1986) 608^618. [37] D. Cunningham, R.L. Cross, J. Biol. Chem. 263 (1988) 18850^18856. [38] T. Noumi, M. Taniai, H. Kanazawa, M. Futai, J. Biol. Chem. 261 (1986) 9196^9201.
BBABIO 44854 24-5-00 Cyaan Magenta Geel Zwart
E. Muneyuki et al. / Biochimica et Biophysica Acta 1458 (2000) 467^481 [39] T.M. Duncan, A.E. Senior, J. Biol. Chem. 260 (1985) 4901^ 4907. [40] A.E. Senior, R.S.F. Lee, M.K. Al-Shawi, J. Weber, Arch. Biochem. Biophys. 297 (1992) 340^344. [41] H. Hanada, T. Noumi, M. Maeda, M. Futai, FEBS Lett. 257 (1989) 465^467. [42] T. Matsui, E. Muneyuki, M. Honda, W.S. Allison, C. Dou, M. Yoshida, J. Biol. Chem. 272 (1997) 8215^8221. [43] J.J. Garcia, R.A. Capaldi, J. Biol. Chem. 273 (1998) 15940^ 15945. [44] S.P. Tsunoda, E. Muneyuki, T. Amano, M. Yoshida, H. Noji, J. Biol. Chem. 274 (1999) 5701^5706. [45] C. Dou, G. Fortes, W.S. Allison, Biochemistry 37 (1998) 16757^16764. [46] E. Muneyuki, H. Hirata, FEBS Lett. 234 (1988) 455^458. [47] E. Muneyuki, M. Makino, H. Kamata, Y. Kagawa, M. Yoshida, H. Hirata, Biochim. Biophys. Acta 1144 (1993) 62^68. [48] J. Weber, S. Wilke-Mounts, R.S.F. Lee, E. Grell, A.E. Senior, J. Biol. Chem. 268 (1993) 20126^20133. [49] M. Ro«gner, P. Gra«ber, Eur. J. Biochem. 159 (1986) 255^261. [50] R. Yasuda, H. Noji, K. Kinosita Jr., M. Yoshida, Cell 93 (1998) 1117^1124. [51] C. Kayalar, J. Rosing, P.D. Boyer, J. Biol. Chem. 252 (1977) 2486^2491. [52] D.D. Hackney, P.D. Boyer, J. Biol. Chem. 253 (1978) 3164^ 3170. [53] G. Rosen, M. Gresser, C. Vinkler, P.D. Boyer, J. Biol. Chem. 254 (1979) 10654^10661. [54] R.I. Feldman, F.D. Sigman, J. Biol. Chem. 258 (1983) 12178^12183. [55] M. Yohda, Y. Kagawa, M. Yoshida, Biochim. Biophys. Acta 850 (1986) 429^435. [56] H.S. Penefsky, J. Biol. Chem. 260 (1985) 13735^13741. [57] M.A. Galkin, A.D. Vinogradov, FEBS Lett. 448 (1999) 123^ 126. [58] E. Muneyuki, Y. Kagawa, H. Hirata, J. Biol. Chem. 264 (1989) 6092^6096. [59] H. Hirata, K. Ohno, N. Sone, Y. Kagawa, T. Hamamoto, J. Biol. Chem. 261 (1986) 9839^9843. [60] T. Takabe, G.G. Hammes, Biochemistry 20 (1981) 6859^ 6864. [61] T.G. Dewey, G.G. Hammes, J. Biol. Chem. 256 (1981) 8941^ 8946. [62] H. Strotmann, R. Thelen, W. Muller, W. Baum, Eur. J. Biochem. 193 (1990) 879^886. [63] G. Kothen, O. Schwarz, H. Strotmann, in: N. Murata (Ed.), Research in Photosynthesis, Vol. II, Kluwer Academic, Dordrecht, 1992, pp. 661^668. [64] H.P. Lu, L. Xun, X.S. Xie, Science 282 (1998) 1877^1882.
481
[65] J. Moyle, P. Mitchell, FEBS Lett. 56 (1975) 55^61. [66] A.F. Fitin, E.A. Vasilyeva, A.D. Vinogradov, Biochem. Biophys. Res. Commun. 86 (1979) 434^439. [67] E.A. Vasilyeva, I.B. Minkov, A.F. Fitin, A.D. Vinogradov, Biochem. J. 202 (1982) 9^14. [68] P.N. Lowe, R.B. Beechy, Biochem. J. 196 (1981) 443^449. [69] D.J. Hyndman, Y.M. Milgrom, E.A. Bramhall, R.L. Cross, J. Biol. Chem. 269 (1994) 28871^28877. [70] M.B. Murataliev, Y.M. Milgrom, P.D. Boyer, Biochemistry 30 (1991) 8305^8310. [71] V.V. Bulygin, A.D. Vinogradov, Biochem. J. 276 (1991) 149^ 156. [72] Y.M. Milgrom, P.D. Boyer, Biochim. Biophys. Acta 1020 (1990) 43^48. [73] G. Kaim, P. Dimroth, EMBO J. 17 (1998) 5887^5895. [74] G. Kaim, P. Dimroth, FEBS Lett. 434 (1998) 57^60. [75] F.E. Possmayer, P. Gra«ber, J. Biol. Chem. 269 (1994) 1896^ 1904. [76] S. Khan, M. Dapice, I. Humayun, Biophys. J. 57 (1990) 779^796. [77] F. Oosawa, J. Biochem. 118 (1995) 863^870. [78] J.T. Finer, R.M. Simmons, J.A. Spudich, Nature 358 (1994) 113^119. [79] H. Miyata, H. Yoshikawa, H. Hakozaki, N. Suzuki, T. Furuno, A. Ikegame, K. Kinoshita Jr., T. Nishizaka, S. Ishiwata, Biophys. J. 68 (1995) 286s^290s. [80] A. Ishijima, Y. Harada, H. Kojima, T. Funatsu, H. Higuchi, T. Yanagida, Biochem. Biophys. Res. Commun. 199 (1995) 1057^1063. [81] K. Svoboda, C.F. Schmidt, B.J. Schnapp, S.M. Block, Nature 365 (1993) 721^727. [82] K. Kinoshita Jr., R. Yasuda, H. Noji, S. Ishiwata, M. Yoshida, Cell 93 (1998) 21^24. [83] M.O. Magnasco, Phys. Rev. Lett. 71 (1993) 1477^1481. [84] R.D. Astumian, M. Bier, Phys. Rev. Lett. 72 (1994) 1766^ 1769. [85] C.R. Doering, E. Horsthemke, J. Riordan, Phys. Rev. Lett. 72 (1994) 2984^2987. [86] J. Rousselet, L. Salome, A. Ajdari, J. Prost, Nature 370 (1994) 446^448. [87] L.P. Faucheux, L.S. Bourdieu, P.D. Kaplan, A.J. Libchaber, Phys. Rev. Lett. 74 (1995) 1504^1507. [88] J. Madox, Nature 369 (1994) 181. [89] R.D. Astumian, M. Bier, Biophys. J. 70 (1996) 637^653. [90] Y. Okada, N. Hirokawa, Science 283 (1999) 1152^1157. [91] T. Mitsui, S. Kumagai, H. Chiba, H. Yoshimura, H. Ohshima, J. Theor. Biol. 192 (1998) 35^41. [92] E. Muneyuki, M. Ikematsu, M. Yoshida, J. Phys. Chem. 100 (1996) 19687^19691. [93] H. Wang, G. Oster, Nature 396 (1998) 279^282.
BBABIO 44854 24-5-00 Cyaan Magenta Geel Zwart
Review
F0 F1-ATP synthase: general structural features of `ATP-engine' and a problem on free energy transduction Eiro Muneyuki a , Hiroyuki Noji b , Toyoki Amano Masasuke Yoshida a; * a
b;1
, Tomoko Masaike a ,
Research Laboratory of Resources Utilization, R-1, Tokyo Institute of Technology, 4259 Nagatsuta, Midori-ku, Yokohama 226-8503, Japan b CREST (Core Research for Evolutional Science and Technology) `Genetic Programming' Team 13. Teikyo University Biotechnology Research Center 3F, 907 Nogawa, Miyamae-ku, Kawasaki 216-0001, Japan Received 19 May 1999; received in revised form 26 August 1999; accepted 26 August 1999
1. Introduction F0 F1 -ATP synthase plays a central role in ATP synthesis during oxidative or photophosphorylation. This enzyme is special in that it utilizes the protonic electrochemical potential created by electron transfer reaction by light or respiration to drive thermodynamically unfavorable ATP synthesis as predicted by Mitchell. It can also drive uphill proton transport by hydrolyzing ATP. These properties are clearly in contrast with those of usual enzymes which promote only thermodynamically favorable reactions by lowering activation energy. In addition, this enzyme has an unusual mechanistic property that the rotation of a part of the enzyme complex relative to the rest plays an essential role in its function as predicted by Boyer. Yet this enzyme has common sequence motifs called Walker motifs A and B which exist in a large number of proteins. These proteins, which we call Walker ATPase, are `ATP-engine' which exert a variety of mechanistic works fueled by ATP. Recent
* Corresponding author. Fax: +81-45-924-5277; E-mail: [email protected] 1 Present address, Department of Biology and Geosciences, Faculty of Science, Shizuoka University, 836 Ohya, Shizuoka, 422-8529, Japan.
crystallographic studies also indicated that nucleotide binding domain in the catalytic subunit of ATP synthase has a protein folding similar to other ATP utilizing proteins. In this article, we ¢rst focus on the common aspect of the ATP utilizing proteins in relation to our previous prediction on the catalytic carboxylate. Based on the recent results, we suggest that the position of catalytic carboxylate may have some correlation with the function of the protein. Then we discuss the rotation of the Q-subunit in the K3 L3 -cylinder of the thermophilic F1 -ATPase from the viewpoint of kinetics and energetics. Finally, an attempt is made to identify some of the important questions that remain unanswered with an emphasis on the problem of free energy transduction. 2. Catalytic carboxylate in F1 -ATPase and other Walker ATPases 2.1. Catalytic carboxylate in F1 -ATPase Activation of a water molecule in the catalytic center of F1 -ATPase by a general base is required for ATP hydrolysis. Using F1 -ATPase from a thermophilic Bacillus PS3, we have demonstrated that: (1) ATPase activity was inactivated completely when Glu-190 of the L-subunit was labeled by dicyclohexylcarbodiimide (DCCD) [1]; (2) substitution of Glu-
0005-2728 / 00 / $ ^ see front matter ß 2000 Elsevier Science B.V. All rights reserved. PII: S 0 0 0 5 - 2 7 2 8 ( 0 0 ) 0 0 0 9 5 - 5
BBABIO 44854 24-5-00 Cyaan Magenta Geel Zwart
468
E. Muneyuki et al. / Biochimica et Biophysica Acta 1458 (2000) 467^481
190 by Gln resulted in the loss of ATPase activity [2] and introduction of di¡erent carboxylate-containing side chains at position 190 caused drastic loss of ATPase activity [3]; (3) the DCCD labeled F1 ATPase lost the ability to synthesize enzyme-bound ATP in the presence of dimethylsulfoxide [4]; and (4) Glu-190 might not be involved in binding of nucleotide because DCCD labeled F1 -ATPase showed normal nucleotide binding [5]. No or little contribution of this residue to nucleotide binding was more clearly shown later by the study of mutant Escherichia coli F1 -ATPase [6]. As to point 1, the equivalent mutant of E. coli F1 -ATPase was at ¢rst reported to have a weak ATPase activity [7], but the critical role of this residue was proved later [8]. Indeed, the same E. coli mutant could not catalyze even a uni-site ATP hydrolysis [9]. The catalytic role of Glu-190 as a general base in ATP hydrolysis was ¢nally veri¢ed by the crystal structure of mitochondrial F1 -ATPase in which Glu-188 of the L-subunit carrying AMPPNP (equivalent to Glu-190 of the thermophilic F1 ATPase) is located at such a position that it can activate a water molecule for an in-line attack of the Q-phosphate of nucleotide (Fig. 2A) [10]. All three L-subunits in F1 -ATPase should have Glu at position 190 to continue catalytic turnover because a chimeric (but homogeneous) mutant K3 L3 Q-complex of thermophilic F1 -ATPase, in which one of the Lsubunits has Gln at position 190, but the rest two Glu, can mediate a uni-site catalysis, but cannot propagate bi- and tri-site catalysis [11]. To be a general base, pKa value of Glu-190 should be in the neutral pH range and indeed the value pKa 6.8 was estimated from NMR titration of the carboxylate at position 190 [12]. 2.2. Previous prediction of catalytic carboxylate in Walker ATPases A large number of proteins which catalyze, or are supposed to catalyze ATP-triggered reactions have a set of Walker sequence motifs A (GXXXXGKT/S; X can be varied) and B (ZZZZD; Z is a hydrophobic residue) [13]. When the crystal structure of RecA protein, one of the Walker ATPases, was solved and its Glu-96 was assigned to be the catalytic carboxylate [14], we were surprised at the similarity of the relative location of the catalytic carboxylate of
RecA and F1 -ATPase L-subunit; 24 and 26 residues downstream from Lys of motif A, respectively. Even though the structure of F1 -ATPase was not known at that time and many F1 -ATPase researchers regarded adenylate kinase as a structural model of catalytic domain of F1 -ATPase, we predicted that the structure of catalytic domain of F1 -ATPase would resemble that of RecA [15]. This prediction soon turned out to be true when the F1 -ATPase structure was solved [10]. The prominent structural feature common to the two proteins is that the catalytic domain consists of one K-helix surrounded by six parallel L-strands and the Asp in motif B and the catalytic carboxylate are located at the exits of adjacent two L-strands. Encouraged by this success, we extended the prediction to other Walker ATPases [16]. According to the prediction, the Asp or Glu conserved in many Walker ATPases at 25 þ 2 residues downstream from Lys of motif A, should be a critical catalytic residue. Since then, mutageneses of the predicted catalytic carboxylate have been carried out for several Walker ATPases and, very recently, crystal structures of two important members in the prediction list were determined. Now we can judge the validity of the previous prediction. 2.3. RecA-F1 type ATPase SecA protein is an essential factor for protein export across the membranes in the E. coli cell. We predicted that the Asp-133, located at 24 residues from Lys in motif A, is a catalytic carboxylate. Indeed, when we replaced the Asp-133 with Asn, the mutant SecA lost protein translocation-dependent ATPase activity and did not mediate translocation [17]. Therefore, the prediction was supported and tertiary structure of the nucleotide-binding domain of SecA is most likely similar to that of RecA. We call this type of proteins RecA-F1 protein, a subfamily of Walker ATPases. 2.4. AAA + protein We have also tested the prediction for Lon protease, an ATP-dependent intracellular protease in E. coli. The mutant, in which the predicted catalytic carboxylate, Asp-386, was replaced with Asn, showed no di¡erence from the wild-type; fully active
BBABIO 44854 24-5-00 Cyaan Magenta Geel Zwart
E. Muneyuki et al. / Biochimica et Biophysica Acta 1458 (2000) 467^481
in ATPase and ATP-dependent protease (T. Amano, M. Yoshida, K. Takahashi, T. Muramatsu, unpublished result). The proteins grouped into putative ATPases in the previous list came to be called AAA family of ATPases, and recently, the extended version of AAA protein family called AAA protein family which includes Lon protease has been proposed [18]. AAA protein may not use the previously predicted carboxylate as a catalytic residue. 2.5. ABC transporter Proteins in the family of ATP-binding cassette (ABC) transporters have Walker motifs A and B, and we indicated two or more candidates for the catalytic carboxylate. For instance, we predicted that Glu-64 and Glu-94 of MalK, ATP-hydrolyzing subunit of Salmonella typhimurium ABC transporter, were catalytic carboxylates [16]. However, Stein et al. reported that substitution of both residues of MalK with Gln had no deleterious e¡ect on maltose transport [19]. Ueda and his colleagues replaced the predicted carboxylates of human P-glycoprotein (MDR1) with corresponding Gln or Asn, but ATPase activity and drug-transport activity of the mutants were not impaired (K. Ueda, personal communication). 2.6. Catalytic carboxylate of AAA + protein and ABC transporter Our prediction on the location of catalytic carboxylate in AAA protein and ABC transporter has turned out to invalid. Therefore, we omit AAA protein and ABC transporters from our previous list and propose a revised list as in Fig. 1. Then, where is a catalytic carboxylate in AAA protein and ABC transporter? Recent reports on crystal structures of a member of AAA protein (NSF D2 domain, Fig. 2B) [20] and ABC transporter (HisP, Fig. 2C) [21] have provided a likely answer. The structures of nucleotide binding domains of these two proteins have an apparent structural similarity to the RecA-F1 protein, that is, they are composed of a central K-helix (Fig. 2, yellow) and a surrounding `wall' of L-sheet made up from at least ¢ve parallel Lstrands (green). Their order in the primary sequences is L1 K1 VL2 VL3 VL4 VL5 as schematically shown in
469
Fig. 2D. Walker A motif is present in a loop connecting L1 and K1 (P-loop, blue). Walker B motif is located in L3 and the Asp of ZZZZD is at its exit. The L1 -strand is sandwiched by L4 and L5 in the wall of the L-sheet. Behind the wall each of the loops connecting two L-strands, forms one or several Khelices which function as conformational transducers or switches responding nucleotide turnover occurring at the front side of the wall. These structural features are typically observed in AAA protein. RecA-F1 type protein and ABC transporter have additional unique characteristics; RecA-F1 protein has an additional parallel L-strand beside L2 , making the total number of parallel L-strands in the wall six, and ABC transporter has a large insert domain between K1 and L2 . As described, the catalytic carboxylate, Glu-188, exists at the exit of L2 in mitochondrial F1 -ATPase (Fig. 2A). Contrary, in the case of NSF-D2 (AAA protein) and HisP (ABC transporter), the Asp-612 and Glu-179, adjacent residues to the Walker B motif are positioned at appropriate sites to act as a general base in ATP hydrolysis. Consistently, the sequence, ZZZZD-D/E, is well conserved in AAA protein and ABC transporter. In F1 -ATPase L-subunit and RecA, the corresponding residues are Asn and Ser, respectively. 2.7. Other variations of Walker ATPase We included nitrogenase Fe protein (NifH), a protein catalyzing ATP dependent reduction of N2 to ammonia, in our previous list. Its Asp-43 was predicted as a catalytic carboxylate. In the crystal structure of the nitrogenase complex (Fe-protein and MoFe-protein) from Azotobacter vinelandii stabilized by ADPAlF3 4 , the predicted Asp-43 (designated as Asp-39 in the literature) is really very close to the aluminum atom [22] and therefore a strong candidate for the catalytic carboxylate. However, it should be pointed out that, di¡erent from the prototype folding topology in Fig. 2D, the L1 -strand in the Fe-protein is sandwiched by L3 and L4 , rather than L4 and L5 . Another variation of Walker ATPase is the DExx box helicase. The crystal structure of catalytic domain of the helicase is very similar to that of RecA [23]. Nonetheless, like AAA protein and ABC protein, Glu-224, a residue next to Walker motif B, is assigned to be a catalytic carboxylate.
BBABIO 44854 24-5-00 Cyaan Magenta Geel Zwart
470 E. Muneyuki et al. / Biochimica et Biophysica Acta 1458 (2000) 467^481
BBABIO 44854 24-5-00 Cyaan Magenta Geel Zwart Fig. 1. Revised list of putative RecA-F1 subfamily. Lys in Walker motif A, putative catalytic carboxylate, and Asp in Walker motif B are shown by boxed characters.
E. Muneyuki et al. / Biochimica et Biophysica Acta 1458 (2000) 467^481
2.8. Mechanistic implication Even belonging to the same Walker ATPase, RecA-F1 protein utilizes a di¡erent carboxylate as a catalytic base when compared with AAA protein and ABC transporter. This might be a re£ection of the possible unique function of RecA-F1 protein; it can push a large molecule to one direction. F1 -ATPase L-subunit undergoes a large open-to-close domain movement during ATPase turnover and the concerted movement of three L-subunits in F1 -ATPase drives the central shaft Q-subunit to one direction, that is, rotation. RecA protein can transform the helical pitch of DNA chain and can rotate DNA chain. SecA protein holds a polypeptide stretch of the protein to be secreted, carries it to the opposite side of the membrane by inserting the arm domain, releases the polypeptide, deinserts the arm from membrane, holds next stretch, and repeats the cycle. On the contrary, the motion of other two classes might be less dynamic, though possibly more ¢nely regulated. AAA protein might work as a chaperone-like ATPase associated with the assembly and disassembly of protein complexes. ABC transporters, present at membranes, act as pumps, as regulators of other membrane proteins, and as channels. The contention described here is more speculative than likely, but accumulation of knowledge on mechanism of each protein may give a mechanistic reason of the di¡erence. 3. Rotation of the Q-subunit in relation to the kinetics and energetics 3.1. The rotation of the Q-subunit couples the ATP hydrolysis/synthesis by F1 and proton translocation by F0 In order to achieve e¤cient energy transduction, energy transducing proteins must have a mechanism to transmit the energy released by the exergonic reaction such as ATP hydrolysis to drive the endergonic reaction, such as active transport or mechanical work. As the ¢rst step, it is necessary to conserve the energy released at the active site. Keeping the energy from rapid dissipation by protein molecule is by itself an important subject [24]. Then the energy
471
must be utilized to drive the coupled endergonic reaction which takes place at another distant active site. The rotation of the Q-subunit within the K3 L3 -cylinder of the F1 -ATPase was ¢rst visualized by Noji et al. as a rotation of £uorescently labeled actin ¢lament attached to the Q-subunit (Fig. 3) [25]. The actin ¢laments rotated invariably counter-clockwise when viewed from above in Fig. 3, which is consistent with the direction predicted by X-ray crystal structure [10] and Boyer's Binding Change Mechanism [26^28]. This rotation is a part of the whole reaction coupling ATP hydrolysis and active transport by F0 F1 holoenzyme, yet it is very important in view of the fact that we can directly observe the individual event of the energy transduction (coupling of ATP hydrolysis to mechanical rotation) by a single molecule. The observed rotations were always under the in£uence of hydrodynamic friction against the rotating actin ¢lament, but there is further advantage of the use of actin ¢lament in that we can clearly identify the rotation and calculate the torque. The detailed analyses of the rotation will provide insights into the general aspects of the energy transduction by proteins. In the following part of this article, we will describe the characteristics of the rotation of the Qsubunit observed in single molecule experiments in relation to the knowledge obtained in the usual biochemical experiments in kinetics and energetics and what should be experimentally shown in the future. 3.2. Kinetic aspects of the ATP hydrolysis and ATP hydrolysis-driven rotation of the Q-subunit The kinetic scheme of ATP hydrolysis by F1 - or F0 F1 -ATPase may be written as in Scheme 1. There are negative substrate binding cooperativity (Km1 6 Km2 6 Km3 ) and positive catalytic cooperativity (Vmax1 6 Vmax2 6 Vmax3 ). In total, F1 - and F0 F1 ATPase exhibit apparent negative cooperativity which is characterized by an upward concave double reciprocal plot or downward concave Eadie^Hofstee plot [29^34]. The reaction in the ¢rst line in Scheme 1 is known as `uni-site catalysis' [32,35]. The a¤nity to the substrate ATP in this step is extremely high (Kd = 10312 M in the case of MF1 [32,35^37], 1039 ^ 10310 M in the case of EF1 [38^41]) and the turnover rate (as designated by k3 ) is extremely low [32,35^41].
BBABIO 44854 24-5-00 Cyaan Magenta Geel Zwart
472
E. Muneyuki et al. / Biochimica et Biophysica Acta 1458 (2000) 467^481
Fig. 2. Nucleotide binding domains of three subfamilies of Walker ATPase. The central helix is shown by yellow, ¢ve L-strands by green, and P-loop by blue. The catalytic carboxylates are indicated. (A) RecA-F1 protein ; bovine mitochondrial F1 -ATPase L-subunit with bound AMP-PNP [10]. (B) AAA protein; Chinese hamster ovary NSF D2 domain with bound AMP^PNP [20]. (C) ABC transporter; Salmonella typhimurium HisP with bound ATP [21]. (D) A schematic model of nucleotide binding domain of Walker ATPase.
BBABIO 44854 24-5-00 Cyaan Magenta Geel Zwart
E. Muneyuki et al. / Biochimica et Biophysica Acta 1458 (2000) 467^481
Fig. 2 (continued).
BBABIO 44854 24-5-00 Cyaan Magenta Geel Zwart
473
474
E. Muneyuki et al. / Biochimica et Biophysica Acta 1458 (2000) 467^481
Fig. 3. The experimental system used for the observation of the Q-subunit in the K3 L3 Q-subcomplex of thermophilic F1 -ATPase. A Histag composed of 10 His residues was introduced at the N-terminal of each L-subunit in order to immobilize the complex on a NiNTA bead. A Cys residue was introduced to the Q-subunit and modi¢ed by biotinylated maleimide. Biotinylated and £uorescently labeled actin ¢lament were connected to the Q-subunit via streptavidin. The rotation of the Q-subunit was observed as a rotation of the actin ¢lament under a £uorescent microscope [25,50].
As ATP concentration increases, the reaction pathways shown in the second and third line become dominant. They are called bi-site and tri-site catalysis, respectively [32]. The apparent Km values for the bi-site catalysis and tri-site catalysis are estimated to be 1^30 and 100^300 WM [29^34]. They seem to re£ect ATP binding to the catalytic sites, but not to the non-catalytic sites, because a mutated enzyme, which completely lacks nucleotide binding at non-catalytic nucleotide binding sites, also showed typical uni-site catalysis using TNP^ATP as a substrate and apparent negative cooperativity with Km values of 4 and 135 WM in ATP hydrolysis [42]. A single turnover of the uni-site catalysis may not accompany the rotation of the Q-subunit because cross-linking between the Q- and L-subunits does not inhibit the single turnover of uni-site catalysis
[43]. This contention is supported by the results that a single turnover of the uni-site catalysis was not suppressed when the conformational change of two L-subunits was suppressed by cross-linking [44] and that K^Q cross-linking did not a¡ect the ATP binding to the single high a¤nity site [45]. On the other hand, it was previously shown that the bi-site catalytic cycle was coupled to proton translocation [46]. Under the bi-site condition where two catalytic sites are simultaneously occupied by Mg adenine nucleotide, the inhibition by NaN3 [47], and conversion of the enzyme to the MgADP inhibited form [42] are promoted. Bi-site catalytic cycle may accompany large structural changes because transition from uni-site to bi-site catalysis promotes product release from the ¢rst high a¤nity site and hydrolytic reaction on the enzyme (chase promotion). Recently, a
BBABIO 44854 24-5-00 Cyaan Magenta Geel Zwart
E. Muneyuki et al. / Biochimica et Biophysica Acta 1458 (2000) 467^481
Scheme 1. The ¢rst, second and the third lines represent unisite, bi-site, and tri-site catalysis, respectively. Km1 is de¢ned as (k2+k3)/k1. Vmax1 is de¢ned as (total enzyme) k3. Km2 , Km3 , Vmax2 and Vmax3 are similarly de¢ned. b
conformational change involving a direct contact between two L-subunits by simultaneous occupation of two Mg adenine nucleotide was clearly shown by cross-linking between genetically introduced Cys residues [44]. The catalytic cycle where three catalytic sites are operating is called tri-site catalysis (third line in Scheme 1). Full hydrolytic activity is attained in the tri-site catalysis [48,45]. There is little doubt about the coupling of the tri-site cycle to proton translocation [46,49]. In the single molecule experiments, the rotation of the Q-subunit occurred even at 20 nM ATP where bi-site catalytic cycle predominantly operates [50]. This ¢nding is consistent with the data on proton translocation. Due to the attenuation of the rotational rate by long actin ¢lament, coupling of the tri-site catalysis with the rotation of the Q-subunit has not yet been con¢rmed, but it
475
seems most likely that the rotation should occur in the tri-site cycle. As stated above, both bi-site and tri-site cycles are coupled with proton translocation and hence with the rotation of the Q-subunit. However, which step in the catalytic cycle is coupled to the proton translocation or rotation is not clear. Extensive analyses of oxygen exchange reactions and the e¡ect of uncouplers on the exchange reactions suggested that the step of a¤nity change of the catalytic sites (conversion of the high a¤nity site to the low a¤nity site and low a¤nity site to the high a¤nity site) is coupled to energy transduction [51^53] rather than the cleavage or formation of the chemical bond. Consistently, spontaneous formation of bound ATP in the absence of membrane energization [54,55] and dissociation of tightly bound nucleotide upon membrane energization [56] was reported, but the implication of the latter observations is not very clear due to the MgADP inhibition and activation caused by membrane energization [57]. It has been shown that the Km of tri-site catalysis is not a¡ected by membrane potential [58], but Vmax is a¡ected [59]. Similar results were obtained using proteoliposomes co-reconstituted with bacteriorhodopsin [60,61]. A vpH clamp method was applied for analysis of steadystate kinetics of photophosphorylation [62]. In their study, Km of ATP in ATP hydrolysis and Km of ADP in ATP synthesis were found to be insensitive to the change of vpH [63].
Fig. 4. Stepwise rotation at 20 nM ATP. (A) Revolutions versus time. It can be seen that the unitary step of the rotation is 120³. (B) Trace of the centroid of the actin image. The triangular distribution indicates that the actin ¢lament tends to stop with 120³ intervals [50].
BBABIO 44854 24-5-00 Cyaan Magenta Geel Zwart
476
E. Muneyuki et al. / Biochimica et Biophysica Acta 1458 (2000) 467^481
In the single molecule experiment, at low ATP concentrations (sub-WM ATP), the rotational rates were limited by the rate of ATP binding rather than the hydrodynamic friction against the rotating actin ¢lament. Under this condition, actin ¢lament showed clear 120³ steps during rotation (Fig. 4) [50]. The analysis of dwell times between steps indicated that the onset of the step motion was a ¢rst order reaction with a rate constant consistent with the rate of ATP binding in solution. From this analysis, it was suggested that one ATP caused one step. This result indicates that the rate of ATP binding is not signi¢cantly attenuated by the load of actin ¢lament in the rotational experiment under bi-site condition. Together with the observations that protonic electrochemical potential did not a¡ect the apparent Km for ATP in ATP hydrolysis cited above, it seems that ATP binding per se is not directly coupled to proton translocation or rotation of the Q-subunit in ATPase reaction. 3.3. Slow transitions between distinct enzymatic states are directly observed by the single molecule technique The single molecule experiment has a potential to discern distinct states of individual molecules which are essentially indistinguishable in ensemble-averaged experiments. By this technique, slow £uctuation of enzymatic reaction rate in cholesterol oxidase was detected and analyzed [64]. In the case of F0 F1 and F1 -ATPase, slow conversion between active and inactive states has been known for a long time [65]. The inactive form of the ATPase is generated from enzyme with bound MgADP [66^71] in a catalytic site [72]. Due to this inhibition, the rate of ATP hydrolysis measured in the presence of ATP regenerating system decays to approximately less than half of the initial uninhibited rate within several minutes. In ensemble-averaged experiments, it is di¤cult to conclude whether this half-inhibited state stands for the equilibrium between completely active and completely inactive enzymes or the rate of individual enzyme molecule is attenuated to the half of its initial value. Single molecule experiments may give a clear answer on this point. If the latter is the case, the rate of rotation would decrease with time. If the former is the case, the continuous rotation would be inter-
vened by occasional stops as the progress of the inhibition. At present, the former possibility seems more probable. Slow state-transition of a protein molecule which is clari¢ed in the single molecule experiments may have some mechanistically and physiologically important roles. 3.4. How does the energy transduction occur in the F0 channel? Recently, it was shown that membrane potential was essential in ATP synthesis and Na concentration gradient exerted minor in£uence in Na translocating F0 F1 -ATP synthase from Propionigenium modestum [73]. Similar results were reported for E. coli enzyme [74] and the results were discussed in relation to the putative rotation of the ring structure of the c subunits [73]. In the case of chloroplast enzyme, the rate of ATP synthesis by vpH was shown to be dependent on the pKa of the proton binding sites facing inside and outside of reconstituted proteoliposomes [75]. Does it mean, however, that the concentration gradient of the transported ion cannot serve as an energy source for ATP synthesis in the absence of membrane potential? This is a special case of a more general question asking how an energy transducing protein can utilize free energy of entropic origin (vpH) as well as enthalpic origin (v8). It is known that £agellar motor can rotate utilizing the proton concentration gradient as well as membrane potential as the energy source [76]. These components are thermodynamically equivalent. Therefore, they must be equivalent as the energy source for an energy-transducing protein even if they are kinetically inequivalent. At least, at the equilibrium point of ATP synthesis and hydrolysis, membrane potential and vpH must be completely interchangeable. In order to understand the transduction of the energy of entropic origin by F0 F1 -ATP synthase at a molecular level, the essential role of thermal £uctuation should be considered [77] in relation to the function of F0 portion. The rotation of the c-subunit ring of the F0 portion has so far never been directly observed. When the individual step motion of F0 portion driven by vpH and/or v8 is observed, the results will give a hint on the above question. Recently, 5.3-nm substeps were resolved during a
BBABIO 44854 24-5-00 Cyaan Magenta Geel Zwart
E. Muneyuki et al. / Biochimica et Biophysica Acta 1458 (2000) 467^481
477
motion of a single myosin head along an actin ¢lament [24]. It was concluded that hydrolysis of one ATP caused multiple steps. This observation is very interesting when we think of the coupling between F1 and F0 . In the K3 L3 Q-subcomplex, the rotation of the Q-subunit showed clear 120³ steps corresponding to the three KL pairs at low ATP concentration. When the rotation coupled to the ring structure of F0 portion containing 10^12 c-subunits is observed, the energy derived from one ATP hydrolysis may be divided to drive substeps corresponding to each c-subunit. 3.5. Mechanics and energetics of the rotation of the Q-subunit The most prominent advantage of the direct observation of the rotating Q-subunit using actin ¢lament is that we can estimate the force and energy transduced from ATP hydrolysis, particularly, in a single event. Unique information can be obtained by these single molecule experiments. At high ATP concentrations, the rotational rates were limited by the hydrodynamic friction against the rotating actin ¢lament and the rotation was slower with longer actin ¢laments. When the rotational rates were plotted against the actin length, the highest rates which were most reliable at each actin length were on the iso-torque line corresponding to 40 pN nm (Fig. 5) [50]. Thus the K3 L3 Q-subcomplex produced constant force irrespective of the load. This is in contrast with the case of kinesin whose force decreases as load increases. In the case of the K3 L3 Qsubcomplex^actin ¢lament system, the torque may be ¢rst stored in a softest elastic element in the system and then released to induce the rotational motion of the actin ¢lament. The force was calculated to be as high as 40 pN from the torque on the assumption that it was generated at the L^Q-interface [50]. This value was by far the greatest among known molecular motors, such as myosin (3^6 pN) [78^80] and kinesin (5 pN) [81]. At low ATP concentrations, the rotational rates were limited by the rate of ATP binding as stated above. Under this condition (sub-WM ATP), actin ¢lament showed clear 120³ steps during rotation (Fig. 4) [50]. The discrete 120³ steps suggest that there is only a little slip between the Q- and K3 L3 -
Fig. 5. Load dependence of the rotational rate. Average rotational rates versus the length of actin ¢laments are shown. White circles, [ATP] = 2 mM; ¢lled circles, [ATP] = 2 mM, [ADP] = 10 WM, and [Pi] = 0.1 mM (calculated vGATP = 110 pN nm); ¢lled rectangles, [ATP] = 2 mM, [ADP] = 10 WM, and [Pi] = 10 mM (calculated vGATP = 90 pN nm). The curves in the ¢gure stand for iso-torque lines. In spite of the substantial scattering of the data, the experimental points distribute around the iso-torque line of 40 pN nm. The torque of 40 pN nm corresponds to the work of 80 pN nm by one ATP hydrolysis (40 pN nmW2Z/3 [ATP] = 80 pN nm).
cylinder. The rate of the rotation within one step seemed constant, suggesting that the K3 L3 -cylinder exerted a constant force during the travel of the Qsubunit over 120³. The energy used for the rotation of the actin ¢lament within one step was calculated to be about 80 pN nm, which was consistent with the value calculated from the rate of the continuous rotation observed at high ATP concentrations [50]. This value was virtually independent of the co-existence of ADP and Pi provided that the free energy of ATP hydrolysis was above 90 pN nm (Fig. 5). Comparing with the free energy change of ATP hydrolysis in vivo, it was suggested that the e¤ciency of the energy conversion from ATP hydrolysis to rotational motion under physiological conditions was nearly 100%. This is much higher than the e¤ciencies estimated for other molecular motors [82]. 3.6. What is the relationship between the input energy and output energy? Then, what happens if the free energy provided by one ATP hydrolysis was signi¢cantly lower than 80
BBABIO 44854 24-5-00 Cyaan Magenta Geel Zwart
478
E. Muneyuki et al. / Biochimica et Biophysica Acta 1458 (2000) 467^481
pN nm, but greater than 0? If the angular velocity during 120³ step was constant to yield a work of 80 pN nm, it seems as if this situation would violate the second law of thermodynamics. There are at least ¢ve possibilities. (1) The energy of the rotation of the actin ¢lament over 120³ step is constantly 80 pN nm, and the number of ATP hydrolysis per one step increases (variable stoichiometry between the ¢xed mechanical output and the number of ATP). (2) The mechanical work of the rotation of the Q-subunit depends on the free energy of ATP hydrolysis and therefore continuously changes (variable output energy). (3) Both the mechanical output and the number of ATP hydrolyzed in an unitary mechanical step change. (4) The rotation simply stops when the vGATP is lower than 80 pN nm. Under this condition, ATPase also stops and there is no problem of energy balance. (5) It might be possible that when
one ATP is hydrolyzed, 80 pN nm of the mechanical work is constantly derived. In this case, the supplementary energy is derived by thermal £uctuation and the energy and entropy balance is satis¢ed by the dissipation of energy through the rotation of actin ¢lament. In any case of the ¢ve possibilities, the second law of thermodynamics is seemingly satis¢ed, but the ¢rst four possibilities have common problems of how a protein molecule senses the concentration of ATP, ADP, and Pi whose ratio determines the free energy of ATP hydrolysis. In addition, the problem of how to switch the stoichiometry or output energy depending on the vGATP emerges. The problem of what determines the reversal point of ATP hydrolysis and ATP synthesis in the holoenzyme, F0 F1 also emerges in the ¢rst to third cases. These problems are complicated because the mechanical work of
Fig. 6. Schematic illustrations of a potential-transition model. U1(x) (red line) and U2(x) (blue line) describe the potential energy pro¢le of the de-energized and energized states, respectively. The abscissa is regarded as the reaction coordinate of interest (angle, translational coordinate, etc.) The ordinate is internal energy. (A) Biased di¡usion model (thermal ratchet). One of the potential pro¢les (U2(x)) is completely £at. P(x) in black line describes the probability density of a particle (a part of a motor protein complex). When the system is de-energized (potential pro¢le U1(x)), the probability density localizes at the n-th valley of U1(x) (lower P(x)). Upon transition to the energized state (potential pro¢le U2(x)), the probability density distributes widely on the £at potential pro¢le due to thermal di¡usion. When the system is de-energized again, the fraction located in the right part on U2(x) (shaded in blue in upper P(x)) will most probably redistributes to the (n+1)-th valley on U1(x) (shaded in blue in lower P(x)). Successive transitions between de-energized and energized states will drive the probability density to the right side. (B) Power stroke model. Both U1(x) (red line) and U2(x) (blue line) have tooth-like shapes whose phases are di¡erent (B). Successive transitions drive the particle to the right according to the slope of the potentials. There are many variations di¡ering in the shapes of potential pro¢les and the modulations of transition frequency [83^85,89]. In order to induce unidirectional motion, one of the potential pro¢les before or after the transitions must be asymmetric along the reaction coordinate. The transition between the two states must deviate the distribution of the two states from the Boltzmann distribution [89]. In these models, the energy input is the elevation of the potential of the particle by the transition of the potential pro¢le and the output of the energy is the downhill motion of the particle along the slope of the potential pro¢le. The force acting on the particle is determined by the slope of the potential.
BBABIO 44854 24-5-00 Cyaan Magenta Geel Zwart
E. Muneyuki et al. / Biochimica et Biophysica Acta 1458 (2000) 467^481
the rotation of the Q-subunit is not conservative, but dissipative. Simple application of thermodynamics without distinction of the conservative and dissipative work may yield a nonsense answer. When a clear answer is given experimentally, it would contribute to the understanding of energy transduction by proteins on the molecular level. 3.7. General problem of free energy transduction by a single molecule The answer to the above questions should be experimentally given in the future. Before then, it may merit to consider some general models to understand the problem of energy transduction. One such models assumes two states which have di¡erent potentials of interaction between two components of a motor protein complex (see Fig. 6) [83^89]. According to this type of model, unidirectional motion of a particle (a protein component of the motor complex) is induced by the transitions of the potential of interaction (see Fig. 6). In one extreme case of the models, one of the potential pro¢les is completely £at and the directional motion is induced by biased thermal di¡usion (Fig. 6A). This case is often called `thermal ratchet'. The other extreme case assumes two tooth-like potentials whose phases are di¡erent (Fig. 6B). This case is called `power stroke model'. In this case, the directional motion requires no di¡usive step. At the molecular level, completely £at potential may not exist and thermal £uctuation cannot be avoided. Therefore the real situation is somewhere between these extreme cases. There are some examples of the application of such models to explain properties of the molecular motors [84,89,90,91] and active ion transport [92]. One of the most elaborate models to describe the rotation of F1 -ATPase was developed by Wang and Oster [93]. They calculated several potential pro¢les. By making appropriate assumptions on the potential pro¢les and the probabilities of transitions, they could reproduce properties of the molecular motion of the Q-subunit known at that time. They included vGATP in the potential pro¢le (Fig. 3B in [93]), which is in a way a little di¡erent from Fig. 6 here. Fig. 6 represents only the internal energy which is independent of vGATP . In their case, the energy input
479
was the vertical distance between the two potential pro¢les of empty L-subunit. This was vGATP by definition. The output energy was the downhill motion of the particle on the slope of the potentials. Therefore, the e¤ciency of the energy transduction could be easily up to 100% by adjusting vGATP . By including vGATP in the potential pro¢le according to the de¢nition of thermodynamics, they could avoid the question if it is possible for a motor protein to sense the free energy of ATP hydrolysis or not. However, thermodynamics is based on averaging on time or ensemble of many particles. On the other hand, we are now observing a single event by a single molecule. If a protein can sense the concentration of substrates and re£ect the concentrations in a single event, it must spend some time to interact with the substrates and keep the memory of many interactions. As substrate concentrations decrease, the time required to sense the concentrations becomes longer. After all, the question raised in the previous section `How does the Q-subunit rotate when vGATP is less than 80 pN nm?' seems to remain here. We feel that the potential pro¢les of internal energy (Fig. 6) are more helpful than the potential pro¢les including vGATP in understanding the individual events carried out by a single molecule. In that case (Fig. 6), the e¡ect of vGATP appears as a change in the probability of transitions between the potential states through the concentration of ATP, ADP and Pi. The concept that the molecular motion of motor proteins or active transport systems can be explained as a progress along £uctuating potential surfaces is very intriguing and important, but it should be noted that the de¢nitions of the potential pro¢les of di¡erent models are not always the same. In addition, it is very important to discriminate conservative work and dissipative work when we compare the theory and real experimental results obtained by observing a single event of a single protein molecule and try to understand the free energy transduction. 4. Epilogue In this article, we ¢rst focused on the catalytic carboxylate and structural features of F0 F1 -ATPase in relation to the other ATP utilizing proteins which
BBABIO 44854 24-5-00 Cyaan Magenta Geel Zwart
480
E. Muneyuki et al. / Biochimica et Biophysica Acta 1458 (2000) 467^481
have Walker motifs. Then, we surveyed the present knowledge on the rotation of the Q-subunit in the K3 L3 Q-subcomplex of F1 -ATPase and kinetics and some structural change. So far, the relationship between them is hardly established. One of the most important future problems is to make a clear framework which accommodates the kinetics, structural changes, and rotation of the Q-subunit. It is needless to say that the proton translocation (and the rotation of F0 subunits) should be involved ultimately in the framework. In trying to highlight the unanswered questions on the free energy transduction by protein molecules, we took some liberties in speculating on possible mechanisms or models in the last part. The conclusion and establishment of any concept should await experimental results, but we hope these speculations would stimulate further discussion and alternative hypothesis, and are ¢nally veri¢ed by experiments directly observing the single molecules. The single molecule observation technique will open a new romantic era in the ¢eld of bioenergetics. Acknowledgements We are grateful to Dr. Sung-Hou Kim for providing us with the coordinate of HisP structure. We thank Dr. Kazuhiko Kinosita and George Oster for helpful comments on this article.
References [1] M. Yoshida, J.W. Poser, W.S. Allison, F.S. Esch, J. Biol. Chem. 256 (1981) 148^153. [2] M. Ohtsubo, M. Yoshida, S. Ohta, Y. Kagawa, M. Yohda, T. Date, Biochem. Biophys. Res. Commun. 146 (1987) 705^ 710. [3] T. Amano, K. Tozawa, M. Yohsida, H. Murakami, FEBS Lett. 348 (1994) 93^98. [4] M. Yoshida, W.S. Allison, J. Biol. Chem. 261 (1986) 5714^ 5721. [5] M. Yoshida, W.S. Allison, J. Biol. Chem. 258 (1983) 14407^ 14412. [6] S. Lobau, J. Weber, S. Wilke-Mounts, A.E. Senior, J. Biol. Chem. 272 (1997) 3648^3656. [7] K. Parsonage, S. Wilke-Mounts, A.E. Senior, Arch. Biochem. Biophys. 261 (1988) 222^225.
[8] A.E. Senior, M.K. Al-Shawi, J. Biol. Chem. 267 (1992) 21471^21478. [9] M. Futai, M.-Y. Park, A. Iwamoto, H. Omote, M. Maeda, Biochim. Biophys. Acta 1187 (1994) 165^170. [10] J.P. Abrahams, A.G.W. Leslie, R. Lutter, J.E. Walker, Nature 370 (1994) 621^628. [11] T. Amano, T. Hisabori, E. Muneyuki, M. Yoshida, J. Biol. Chem. 271 (1996) 17343^17348. [12] K. Tozawa, H. Ohbuchi, H. Yagi, T. Amano, T. Matsui, M. Yoshida, H. Akutsu, FEBS Lett. 397 (1996) 122^126. [13] J.E. Walker, M. Saraste, M.J. Runswick, N.J. Gay, EMBO J. 1 (1982) 945^951. [14] R.M. Story, T.A. Steiz, Nature 355 (1992) 374^376. [15] T. Amano, M. Yoshida, Y. Matsuo, K. Nishikawa, FEBS Lett. 351 (1994) 1^5. [16] M. Yoshida, T. Amano, FEBS Lett. 359 (1994) 1^5. [17] K. Sato, H. Mori, M. Yoshida, S. Mizushima, J. Biol. Chem. 271 (1996) 17439^17444. [18] A.F. Neuwald, L. Aravind, J.L. Spouge, E.V. Koonin, Genone Res. 9 (1999) 27^43. [19] A. Stein, S. Hunke, E. Schneider, FEBS Lett. 18 (1997) 211^ 214. [20] C.U. Lenzen, D. Steinmann, S.W. Whiteheart, W.I. Weis, Cell 94 (1998) 525^536. [21] L.-W. Hung, I.X. Wang, K. Nikaido, P.-Q. Liu, G.F.-L. Ames, S.-H. Kim, Nature 396 (1998) 703^707. [22] H. Schindelin, C. Kisker, J.L. Schlessman, J.B. Howard, D.C. Rees, Nature 387 (1997) 370^376. [23] H.S. Subramanya, L.E. Bird, J.A. Brannigan, D.B. Wigley, Nature 384 (1996) 379^384. [24] K. Kitamura, M. Tokunaga, A. Hikikoshi-lwane, T. Yanagida, Nature 397 (1999) 129^134. [25] H. Noji, R. Yasuda, M. Yoshida, K. Kinosita Jr., Nature 386 (1997) 299^302. [26] P.D. Boyer, FASEB J. 3 (1989) 2164^2178. [27] P.D. Boyer, Biochim. Biophys. Acta 1140 (1993) 215^250. [28] P.D. Boyer, Annu. Rev. Biochem. 66 (1997) 717^749. [29] S.D. Dunn, V.D. Zadorozny, R.G. Tozer, L.E. Orr, Biochemistry 26 (1987) 4488^4493. [30] M.J. Gresser, J.A. Myers, P.D. Boyer, J. Biol. Chem. 257 (1982) 12030^12038. [31] S.-Y. Wong, A. Matsuno-Yagi, Y. Hate¢, Biochemistry 23 (1984) 5004^5009. [32] R.L. Cross, C. Grubmeyer, H.S. Penefsky, J. Biol. Chem. 257 (1982) 12101^12105. [33] Y. Kato, T. Sasayama, E. Muneyuki, M. Yoshida, Biochim. Biophys. Acta 1231 (1995) 275^281. [34] C. Matsuda, E. Muneyuki, H. Endo, M. Yoshida, Y. Kagawa, Biochem. Biophys. Res. Commun. 200 (1994) 671^678. [35] C. Grubmeyer, R.L. Cross, H.S. Penefsky, J. Biol. Chem. 257 (1982) 12092^12100. [36] H.S. Penefsky, Methods Enzymol. 126 (1986) 608^618. [37] D. Cunningham, R.L. Cross, J. Biol. Chem. 263 (1988) 18850^18856. [38] T. Noumi, M. Taniai, H. Kanazawa, M. Futai, J. Biol. Chem. 261 (1986) 9196^9201.
BBABIO 44854 24-5-00 Cyaan Magenta Geel Zwart
E. Muneyuki et al. / Biochimica et Biophysica Acta 1458 (2000) 467^481 [39] T.M. Duncan, A.E. Senior, J. Biol. Chem. 260 (1985) 4901^ 4907. [40] A.E. Senior, R.S.F. Lee, M.K. Al-Shawi, J. Weber, Arch. Biochem. Biophys. 297 (1992) 340^344. [41] H. Hanada, T. Noumi, M. Maeda, M. Futai, FEBS Lett. 257 (1989) 465^467. [42] T. Matsui, E. Muneyuki, M. Honda, W.S. Allison, C. Dou, M. Yoshida, J. Biol. Chem. 272 (1997) 8215^8221. [43] J.J. Garcia, R.A. Capaldi, J. Biol. Chem. 273 (1998) 15940^ 15945. [44] S.P. Tsunoda, E. Muneyuki, T. Amano, M. Yoshida, H. Noji, J. Biol. Chem. 274 (1999) 5701^5706. [45] C. Dou, G. Fortes, W.S. Allison, Biochemistry 37 (1998) 16757^16764. [46] E. Muneyuki, H. Hirata, FEBS Lett. 234 (1988) 455^458. [47] E. Muneyuki, M. Makino, H. Kamata, Y. Kagawa, M. Yoshida, H. Hirata, Biochim. Biophys. Acta 1144 (1993) 62^68. [48] J. Weber, S. Wilke-Mounts, R.S.F. Lee, E. Grell, A.E. Senior, J. Biol. Chem. 268 (1993) 20126^20133. [49] M. Ro«gner, P. Gra«ber, Eur. J. Biochem. 159 (1986) 255^261. [50] R. Yasuda, H. Noji, K. Kinosita Jr., M. Yoshida, Cell 93 (1998) 1117^1124. [51] C. Kayalar, J. Rosing, P.D. Boyer, J. Biol. Chem. 252 (1977) 2486^2491. [52] D.D. Hackney, P.D. Boyer, J. Biol. Chem. 253 (1978) 3164^ 3170. [53] G. Rosen, M. Gresser, C. Vinkler, P.D. Boyer, J. Biol. Chem. 254 (1979) 10654^10661. [54] R.I. Feldman, F.D. Sigman, J. Biol. Chem. 258 (1983) 12178^12183. [55] M. Yohda, Y. Kagawa, M. Yoshida, Biochim. Biophys. Acta 850 (1986) 429^435. [56] H.S. Penefsky, J. Biol. Chem. 260 (1985) 13735^13741. [57] M.A. Galkin, A.D. Vinogradov, FEBS Lett. 448 (1999) 123^ 126. [58] E. Muneyuki, Y. Kagawa, H. Hirata, J. Biol. Chem. 264 (1989) 6092^6096. [59] H. Hirata, K. Ohno, N. Sone, Y. Kagawa, T. Hamamoto, J. Biol. Chem. 261 (1986) 9839^9843. [60] T. Takabe, G.G. Hammes, Biochemistry 20 (1981) 6859^ 6864. [61] T.G. Dewey, G.G. Hammes, J. Biol. Chem. 256 (1981) 8941^ 8946. [62] H. Strotmann, R. Thelen, W. Muller, W. Baum, Eur. J. Biochem. 193 (1990) 879^886. [63] G. Kothen, O. Schwarz, H. Strotmann, in: N. Murata (Ed.), Research in Photosynthesis, Vol. II, Kluwer Academic, Dordrecht, 1992, pp. 661^668. [64] H.P. Lu, L. Xun, X.S. Xie, Science 282 (1998) 1877^1882.
481
[65] J. Moyle, P. Mitchell, FEBS Lett. 56 (1975) 55^61. [66] A.F. Fitin, E.A. Vasilyeva, A.D. Vinogradov, Biochem. Biophys. Res. Commun. 86 (1979) 434^439. [67] E.A. Vasilyeva, I.B. Minkov, A.F. Fitin, A.D. Vinogradov, Biochem. J. 202 (1982) 9^14. [68] P.N. Lowe, R.B. Beechy, Biochem. J. 196 (1981) 443^449. [69] D.J. Hyndman, Y.M. Milgrom, E.A. Bramhall, R.L. Cross, J. Biol. Chem. 269 (1994) 28871^28877. [70] M.B. Murataliev, Y.M. Milgrom, P.D. Boyer, Biochemistry 30 (1991) 8305^8310. [71] V.V. Bulygin, A.D. Vinogradov, Biochem. J. 276 (1991) 149^ 156. [72] Y.M. Milgrom, P.D. Boyer, Biochim. Biophys. Acta 1020 (1990) 43^48. [73] G. Kaim, P. Dimroth, EMBO J. 17 (1998) 5887^5895. [74] G. Kaim, P. Dimroth, FEBS Lett. 434 (1998) 57^60. [75] F.E. Possmayer, P. Gra«ber, J. Biol. Chem. 269 (1994) 1896^ 1904. [76] S. Khan, M. Dapice, I. Humayun, Biophys. J. 57 (1990) 779^796. [77] F. Oosawa, J. Biochem. 118 (1995) 863^870. [78] J.T. Finer, R.M. Simmons, J.A. Spudich, Nature 358 (1994) 113^119. [79] H. Miyata, H. Yoshikawa, H. Hakozaki, N. Suzuki, T. Furuno, A. Ikegame, K. Kinoshita Jr., T. Nishizaka, S. Ishiwata, Biophys. J. 68 (1995) 286s^290s. [80] A. Ishijima, Y. Harada, H. Kojima, T. Funatsu, H. Higuchi, T. Yanagida, Biochem. Biophys. Res. Commun. 199 (1995) 1057^1063. [81] K. Svoboda, C.F. Schmidt, B.J. Schnapp, S.M. Block, Nature 365 (1993) 721^727. [82] K. Kinoshita Jr., R. Yasuda, H. Noji, S. Ishiwata, M. Yoshida, Cell 93 (1998) 21^24. [83] M.O. Magnasco, Phys. Rev. Lett. 71 (1993) 1477^1481. [84] R.D. Astumian, M. Bier, Phys. Rev. Lett. 72 (1994) 1766^ 1769. [85] C.R. Doering, E. Horsthemke, J. Riordan, Phys. Rev. Lett. 72 (1994) 2984^2987. [86] J. Rousselet, L. Salome, A. Ajdari, J. Prost, Nature 370 (1994) 446^448. [87] L.P. Faucheux, L.S. Bourdieu, P.D. Kaplan, A.J. Libchaber, Phys. Rev. Lett. 74 (1995) 1504^1507. [88] J. Madox, Nature 369 (1994) 181. [89] R.D. Astumian, M. Bier, Biophys. J. 70 (1996) 637^653. [90] Y. Okada, N. Hirokawa, Science 283 (1999) 1152^1157. [91] T. Mitsui, S. Kumagai, H. Chiba, H. Yoshimura, H. Ohshima, J. Theor. Biol. 192 (1998) 35^41. [92] E. Muneyuki, M. Ikematsu, M. Yoshida, J. Phys. Chem. 100 (1996) 19687^19691. [93] H. Wang, G. Oster, Nature 396 (1998) 279^282.
BBABIO 44854 24-5-00 Cyaan Magenta Geel Zwart