The Rotary Machine in the Cell, ATP Synthase

adenosine 5′-(β,γ-imino)- triphosphate piconewtons dicyclohexylcarbodiimide ATP synthase, a major ATP supplier in the cell, is a rotary machine found next to the bacterial flagella motor in the biological world. This enzyme is composed of two motors, F0 and F1, connected by a common rotor shaft to exchange the energy of proton translocation and ATP synthesis/hydrolysis through mechanical rotation. Rotation of the isolated F1 motor driven by ATP hydrolysis was directly observed with an optical microscope, and its marvelous performance has been revealed. The motor rotates with discrete 120° steps, each driven by hydrolysis of one ATP molecule with nearly perfect energy efficiency. Apparently a cooperative domain bending motion of the catalytic β subunits initiated by ATP binding generates the torque. In the F0 motor, which we know less about, it has been proposed that torque may be generated by the large twist of one helix of F0 c subunits or by the change in electrostatic forces between rigid subunits. ATP synthase is a ubiquitous enzyme that is located in the inner membranes of mitochondria, thylakoid membranes of chloroplasts, or the plasma membranes of bacteria. As implicated by the binding change mechanism proposed by Boyer (1Boyer P.D. Biochim. Biophys. Acta. 1993; 1140: 215-250Crossref PubMed Scopus (928) Google Scholar), ATP synthase employs mechanical rotation to convert the electrochemical potential energy of protons across the membranes (Δμ̃ H+), built up by respiration or a photoreaction, to the chemical energy of ATP synthesis. This enzyme is comprised of two motors sharing a common rotor shaft (Fig. 1 A). The F1 motor, a subcomplex of the ATP synthase corresponding to the protruding portion from the membrane, can generate rotary torque using the energy of ATP hydrolysis (Fig. 1 B). Its subunit composition is α3β3γ1δ1ε1and the M r is ∼380,000. The F0motor, a membrane-embedded subcomplex, generates torque coupled with proton movement down (Δμ̃ H+) (Fig. 1 C). Bacterial F0 has the simplest subunit structure (a 1 b 2 c 10–14(?)) with an M r of ∼150,000. The eukaryotic F0 contains several kinds of subunits. The γ and ε of F1 constitute a rotor shaft and are attached to the F0 c subunits. A stator stalk, made up of δ and F0 b 2, also connects F1and F0 keeping the stators (α3β3 and F0 a) from spinning with the rotor. Under physiological conditions where the driving force for the F0 motor is larger than that for the F1 motor, the F0 motor rotates the common shaft in its intrinsic direction so as to reverse the F1 motor enforcing the ATP synthesis (Fig. 1 A). When the driving force for the F1 motor is larger, the F1 motor reverses the F0 motor to pump protons to the opposite side of a membrane. F1 can be easily and reversibly dissociated from F0 as a soluble enzyme that only hydrolyzes ATP and is often called F1-ATPase. The catalytic sites are mainly located on the β subunit, but the minimum stable ATPase-active complex is the α3β3γ subcomplex (2Matsui T. Yoshida M. Biochim. Biophys. Acta. 1995; 1231: 139-146Crossref PubMed Scopus (86) Google Scholar). The crystal structures of α3β3γ of the bovine mitochondrial F1 show that three αs and βs are alternatively arranged in a hexamer ring forming a large central cavity in which half of the long coiled-coil structure of γ is inserted (3Abrahams J.P. Leslie A.G. Lutter R. Walker J.E. Nature. 1994; 370: 621-628Crossref PubMed Scopus (2764) Google Scholar). According to the recently reported structure of the F1-F0 c complex of yeast ATP synthase (4Stock D. Leslie A.G. Walker J.E. Science. 1999; 286: 1700-1705Crossref PubMed Scopus (1093) Google Scholar), the other half of the coiled-coil of γ extends to touch the F0 c subunits. The ε subunit binds to the side surface of the lowest part of the coiled-coil. In ATP synthase, ε also has close contact with F0 c. The δ subunit, the last subunit whose atomic structure is not known, is likely to sit on top of the α3β3 ring (5Wilkens S. Zhou J. Nakayama R. Dunn S.D. Capaldi R.A. J. Mol. Biol. 2000; 295: 387-391Crossref PubMed Scopus (70) Google Scholar). Three catalytic sites on the βs are different in nucleotide binding states; the first is occupied by Mg·AMP-PNP1 (an analog of ATP), the second is occupied by Mg·ADP, and the third is empty (no bound nucleotide); these sites are termed βT, βD, and βE, respectively. These structural features are quite consistent with what the binding change mechanism predicted; the three catalytic sites should be in three different nucleotide states at a given moment, and cooperative interconversion of the states causes the rotation of γ. To visualize the rotation, F1 molecules from a thermophilic bacterium (Bacillus strain PS3) were fixed on the glass surface of a coverslip, and a large marker, a fluorescently labeled actin filament, was attached to γ (Fig.2 A) (6Noji H. Yasuda R. Yoshida M. Kinosita Jr., K. Nature. 1997; 386: 299-302Crossref PubMed Scopus (1974) Google Scholar). Dependent on ATP, the rotation of the actin filaments with a length of 1–4 μm at 0.2–10 revolutions per s was seen under an optical microscope. The rotation continued for several minutes with hundreds of revolutions. The direction of the rotation was always anti-clockwise viewed from the F0 side, consistent with the crystal structure in which one β undergoes transition from βT to βD to βE. The F1s from Escherichia coli(7Noji H. Hasler K. Junge W. Kinosita Jr., K. Yoshida M. Engelbrecht S. Biochem. Biophys. Res. Commun. 1999; 260: 597-599Crossref PubMed Scopus (84) Google Scholar, 8Omote H. Sambonmatsu N. Saito K. Sambongi Y. Iwamoto-Kihara A. Yanagida T. Wada Y. Futai M. Proc. Natl. Acad. Sci. U. S. A. 1999; 96: 7780-7784Crossref PubMed Scopus (126) Google Scholar) and the chloroplast (9Hisabori T. Kondoh A. Yoshida M. FEBS Lett. 1999; 463: 35-38Crossref PubMed Scopus (62) Google Scholar) are also shown to be a rotary motor by applying the same technique. No obvious difference among the γ rotations was observed. The mechanical properties of the F1motor described below seem to be conserved among species. Because of the hydrodynamic friction, at high ATP concentrations the rotation of an actin filament is the slowest step in the catalytic turnover. The rates of rotation of the filaments with the same length were, therefore, leveled off above 2 μm ATP. At ATP concentrations below 600 nm, the slowest step is the ATP binding and actin filaments showed a stepwise rotation; F1waits for ATP at the fixed position, makes a 120° rotation upon arrival of the ATP, and waits for the next ATP (10Yasuda R. Noji H. Kinosita Jr., K. Yoshida M. Cell. 1998; 93: 1117-1124Abstract Full Text Full Text PDF PubMed Scopus (718) Google Scholar). Obviously, a 120° step is a reflection of the 3-fold arrangement of the catalytic β subunits in the α3β3 hexamer. The histogram of the duration time between 120° steps obeys an exponential function, and the estimated apparent rate constant of ATP binding to F1 agrees well with the rate obtained in a bulk F1 solution. This confirms that the hydrolysis of one ATP molecule suffices for making one 120° step. Interestingly, F1 occasionally makes a backward step as fast as forward steps and too fast to be ascribed to a thermal fluctuation. Presumably, the molecular machine makes a mistake in the order of ATP binding or product release. The rotational rate became slower with an increasing filament length because of the increased viscous friction. However, when the rotary torque is calculated from the frictional drag coefficient and the rotation rate, it becomes clear that the F1 motor generates a constant torque of 40 pN·nm irrespective of the length of the actin filament (10Yasuda R. Noji H. Kinosita Jr., K. Yoshida M. Cell. 1998; 93: 1117-1124Abstract Full Text Full Text PDF PubMed Scopus (718) Google Scholar). If the torque is produced at the β-γ interface at a radius of ∼1 nm from the central axis of the α3β3 hexamer, the force would amount to 40 pN. This is the highest value among reported nucleotide-driven motor proteins (3–5 pN for myosin/actin, 5 pN for kinesin/microtubule, and 14 pN for RNA polymerase/DNA) (11Kinosita Jr., K. Yasuda R. Noji H. Ishiwata S. Yoshida M. Cell. 1998; 93: 21-24Abstract Full Text Full Text PDF PubMed Scopus (145) Google Scholar). The torque of 40 pN·nm × 2π/3 radians (120°) or 80 pN·nm is the work done in a step against the viscous load. To define the free energy of the ATP hydrolysis, we purposely included 10 μm ADP and 10 mm Pi in addition to 2 mm ATP and measured the rotation rates (10Yasuda R. Noji H. Kinosita Jr., K. Yoshida M. Cell. 1998; 93: 1117-1124Abstract Full Text Full Text PDF PubMed Scopus (718) Google Scholar). The free energy of the ATP hydrolysis under the condition is 90 pN·nm per one ATP molecule, and the energy for the observed rotation was 80 pN·nm per 120° step. Therefore, F1 works with almost perfect efficiency. The high efficiency accords with the fully reversible nature of this motor. As another probe to visualize the F1 rotation, a single fluorophore was attached to γ, and its orientation was monitored (12Adachi K. Yasuda R. Noji H. Itoh H. Harada Y. Yoshida M. Kinosita Jr., K. Proc. Natl. Acad. Sci. U. S. A. 2000; 97: 7243-7247Crossref PubMed Scopus (186) Google Scholar). With this small marker, F1 can rotate almost without a load. Under this condition, F1 showed the 120° stepwise rotation at low ATP concentrations as seen in the experiment using an actin filament. Furthermore, the apparent rate of ATP binding is the same as that observed with actin filaments. This suggests that the torque-generating step in the ATPase catalytic cycle of F1 is not the ATP binding but step(s) afterit, including the interconversion of the β subunit that initially accommodates ATP from the “loose” binding state to the “tight” one. Knowledge of the exact kinetic sequence in the catalytic turnover of F1 is the prerequisite for any models of the rotation mechanism. Three catalytic modes are recognized when F1hydrolyzes ATP. At extremely low ATP (less than 1 nm) or a stoichiometric amount of ATP, only one ATP binds to the first catalytic site, and the hydrolyzed products are only released slowly (uni-site catalysis) (13Grubmeyer C. Cross R.L. Penefsky H.S. J. Biol. Chem. 1982; 257: 12092-12100Abstract Full Text PDF PubMed Google Scholar). Uni-site catalysis is not inhibited by the cross-link between γ and β (14Garcia J.J. Capaldi R.A. J. Biol. Chem. 1998; 273: 15940-15945Abstract Full Text Full Text PDF PubMed Scopus (54) Google Scholar) and is probably, if not exclusively, a process that does not couple with rotation. Binding of a second ATP to the next catalytic site significantly promotes the release of the products at the first site (13Grubmeyer C. Cross R.L. Penefsky H.S. J. Biol. Chem. 1982; 257: 12092-12100Abstract Full Text PDF PubMed Google Scholar). The apparent K m for this process is in the micromolar range (bi-site catalysis). Boyer's binding change mechanism has adopted the bi-site catalysis. We observed rotation in this ATP concentration range. It has been proposed that the third catalytic sites bind ATP to attain the maximum hydrolysis rate (tri-site catalysis). Actually, the ATPase activity of F1is usually saturated above 100 μm ATP. Using the fluorescence of tryptophan introduced near the catalytic site ofE. coli F1 as the signal of nucleotide binding, Senior's group (15Weber J. Senior A.E. Biochim. Biophys. Acta. 2000; 1458: 300-309Crossref PubMed Scopus (144) Google Scholar) found that the change was saturated at an ATP concentration above ∼100 μm and suggested that all three catalytic sites were filled by nucleotides. A similar observation has been made for the thermophilic F1 (16Dou C. Fortes P.A. Allison W.S. Biochemistry. 1998; 37: 16757-16764Crossref PubMed Scopus (57) Google Scholar). However, the interpretation has been complicated by the “Mg-ADP-inhibited form,” a state observed for F1s from any sources in which one of the catalytic sites is stuck with a tightly bound Mg-ADP. F1 exerting the steady state catalysis is a dynamic mixture of the inhibited and active forms, and this equilibrium is dragged to the active form by the Mg-ATP binding to the noncatalytic α subunit of which the affinity is below 100 μm (17Jault J.M. Matsui T. Jault F.M. Kaibara C. Muneyuki E. Yoshida M. Kagawa Y. Allison W.S. Biochemistry. 1995; 34: 16412-16418Crossref PubMed Scopus (68) Google Scholar). Boyer has raised a question of whether deviation from simple kinetics at high ATP concentrations could be because of the Mg-ADP-inhibited form rather than the tri-site catalysis (18Milgrom Y.M. Murataliev M.B. Boyer P.D. Biochem. J. 1998; 330: 1037-1043Crossref PubMed Scopus (45) Google Scholar). Contrary to this, a mutant whose α subunits lost the nucleotide binding ability still showed kinetics best interpreted by tri-site catalysis (19Matsui T. Muneyuki E. Honda M. Allison W.S. Dou C. Yoshida M. J. Biol. Chem. 1997; 272: 8215-8221Abstract Full Text Full Text PDF PubMed Scopus (102) Google Scholar). The current results favor the tri-site catalysis as a physiological mode, but exclusive evidence is still needed to settle the argument. Noticeably, no obvious shift in the properties of the γ rotation was observed from 2 μmto 2 mm ATP where the transition from the bi- to tri-site catalysis should occur (10Yasuda R. Noji H. Kinosita Jr., K. Yoshida M. Cell. 1998; 93: 1117-1124Abstract Full Text Full Text PDF PubMed Scopus (718) Google Scholar). The source of energy for the γ rotation is ATP hydrolysis on the β subunits. The conformation changes occurring in β during the ATPase cycle should then be responsible for (or at least closely related to) the torque generation. In the crystal structure of the mitochondrial F1, both βT and βD are in the closed conformation in which the C-terminal domain is lifted to the nucleotide-binding domain (3Abrahams J.P. Leslie A.G. Lutter R. Walker J.E. Nature. 1994; 370: 621-628Crossref PubMed Scopus (2764) Google Scholar). The βE employs the open conformation with a wide crevice between the two domains. The crystal of the isolated β subunit takes the open conformation, 2K. Miki and M. Yoshida, unpublished result. and the addition of a nucleotide caused the transition from the open to closed conformation (NMR) (20Yagi H. Tozawa K. Sekino N. Iwabuchi T. Yoshida M. Akutsu H. Biophys. J. 1999; 77: 2175-2183Abstract Full Text Full Text PDF PubMed Scopus (21) Google Scholar). The binding energy of ATP to the β subunit facilitates an energetically unfavorable transition from the empty to closed conformation of the β subunit. When β in F1 is fixed in the closed conformation by cross-linking, ATP hydrolysis stops (21Ren H. Dou C. Stelzer M.S. Allison W.S. J. Biol. Chem. 1999; 274: 31366-31372Abstract Full Text Full Text PDF PubMed Scopus (14) Google Scholar). Thus, β appears to undergo a bending motion upon binding and the release of the nucleotide during catalysis. Like an automobile engine, the reciprocal motion of β in F1 is converted to the rotary motion of γ. For this to occur, three βs in F1 coordinate the motion, pushing and pulling the eccentric γ (22Wang H. Oster G. Nature. 1998; 396: 279-282Crossref PubMed Scopus (385) Google Scholar). A real time recording of the motion of the βs simultaneously with ATP hydrolysis and γ rotation is a challenge to prove the above contention. Residues playing key roles in the torque generation have been sought by mutagenesis (23Masaike T. Mitome N. Noji H. Muneyuki E. Yasuda R. Kinosita Jr., K. Yoshida M. J. Exp. Biol. 2000; 203: 1-8Crossref PubMed Google Scholar, 24Hara K. Noji H. Yasuda R. Kinosita Jr., K. Yoshida M. J. Biol. Chem. 2000; 275: 14260-14263Abstract Full Text Full Text PDF PubMed Scopus (61) Google Scholar). However, the F1 motor seems fairly robust against the mutations of the β subunit at the “hinge region” of the bending motion (23Masaike T. Mitome N. Noji H. Muneyuki E. Yasuda R. Kinosita Jr., K. Yoshida M. J. Exp. Biol. 2000; 203: 1-8Crossref PubMed Google Scholar) and the conserved “DELSEED region” that has a contact surface with γ in the closed conformation (24Hara K. Noji H. Yasuda R. Kinosita Jr., K. Yoshida M. J. Biol. Chem. 2000; 275: 14260-14263Abstract Full Text Full Text PDF PubMed Scopus (61) Google Scholar). F0 conducts proton movement across a membrane. F0 a is embedded in the membrane by five transmembrane helices. A dimer of F0 b is anchored to the membrane by a single transmembrane helix (25Dunn S.D. McLachlin D.T. Revington M. Biochim. Biophys. Acta. 2000; 1458: 356-363Crossref PubMed Scopus (82) Google Scholar). F0 c is a small hydrophobic protein with a hairpin structure, two transmembrane helices connected by a short polar loop (26Girvin M.E. Fillingame R.H. Biochemistry. 1993; 32: 12167-12177Crossref PubMed Scopus (77) Google Scholar). The F0 c subunits are arranged in a ring structure, but agreement has not been established for the number of subunits in a ring; 10, 12, 14, and variable copies have been proposed (4Stock D. Leslie A.G. Walker J.E. Science. 1999; 286: 1700-1705Crossref PubMed Scopus (1093) Google Scholar, 27Jones P.C. Fillingame R.H. J. Biol. Chem. 1998; 273: 29701-29705Abstract Full Text Full Text PDF PubMed Scopus (111) Google Scholar, 28Seelert H. Poetsch A. Dencher N.A. Engel A. Stahlberg H. Muller D.J. Nature. 2000; 405: 418-419Crossref PubMed Scopus (415) Google Scholar, 29Schemidt R.A. Qu J. Williams J.R. Brusilow W.S.A. J. Bacteriol. 1998; 180: 3205-3208Crossref PubMed Google Scholar). F0 a and F0 b 2 most likely exist outside of the F0 c ring. A carboxyl group located in the middle of the C-terminal helix of F0 c (glutamate in most cases but aspartate (Asp-61) in the case of E. coli) is proven to be essential for proton translocation (30Sebald W. Machleidt W. Wachter E. Proc. Natl. Acad. Sci. U. S. A. 1980; 77: 785-789Crossref PubMed Scopus (113) Google Scholar). This carboxyl group is specifically labeled with dicyclohexylcarbodiimide (DCCD), and the labeled ATP synthase loses the activity of the ATP hydrolysis/synthesis coupled with proton movement (31Hermolin J. Fillingame R.H. J. Biol. Chem. 1989; 264: 3896-3903Abstract Full Text PDF PubMed Google Scholar). Genetic studies indicated that several charged residues of F0 aare also essential and assumed to be components of a putative proton path of F0 (reviewed in Ref. 32Deckers-Hebestreit G. Greie J. Stalz W. Altendorf K. Biochim. Biophys. Acta. 2000; 1458: 364-373Crossref PubMed Scopus (33) Google Scholar). Although the assumption is widely accepted that the F0 c ring rotates together with γ and ε, it has not been proven yet by experiment. Actually, we observed the ATP-driven rotation of the actin filaments attached to the F0 c ring of the immobilized ATP synthase (33Tsunoda S.P. Aggeler R. Noji H. Kinosita Jr., K. Yoshida M. Capaldi R.A. FEBS Lett. 2000; 470: 244-248Crossref PubMed Scopus (72) Google Scholar). However, the detergent used in the experiments impaired the integrity of the enzyme, and the DCCD-labeled enzyme showed uninhibited rotation and ATP hydrolysis. The F0 c ring of the detergent-impaired ATP synthase could simply rotate by being dragged by the rotating γ without regard to whether the F0 c ring works as a stator in the native enzyme. Other groups also reported the same results using DCCD-insensitive preparations, but they thought that the rotation of the F0 c ring was proven (34Sambongi Y. Iko Y. Tanabe M. Omote H. Iwamoto-Kihara A. Ueda I. Yanagida T. Wada Y. Futai M. Science. 1999; 286: 1722-1724Crossref PubMed Scopus (421) Google Scholar). The loss of structural integrity of the ATP synthase in the detergent was unambiguously shown by the structure of the yeast ATP synthase crystals grown in detergent; the enzyme lost at least F0 a and F0 b 2. Whether the F0 c ring is a rotor or stator will be decided by demonstration of, for example, the DCCD-sensitive rotation of F0 c or by a clear biochemical result such as DCCD-sensitive proton translocation by ATP synthase containing a γ-ε-F0 ccross-link. 3Very recently, unequivocal evidence that F0 c belongs to the rotor part was obtained by linking the γ, ε, and F0 c subunits by disulfide bridges between cysteine residues introduced genetically at the interfaces (Tsunoda, S. P., Aggeler, R., Yoshida, M., and Capaldi, R. A. (2001) Proc. Natl. Acad. Sci. U. S. A. 98, in press). This fixing of the three subunits together had no significant effect on ATP hydrolysis, proton translocation, or ATP synthesis, and each of these functions retained sensitivity to DCCD. A monomer structure of F0 c in a water-saturated organic solvent, which mimics well the native structure, was determined by NMR (26Girvin M.E. Fillingame R.H. Biochemistry. 1993; 32: 12167-12177Crossref PubMed Scopus (77) Google Scholar). Using this method, a large conformational change of F0 c induced by deprotonation of essential Asp-61 was detected; the C-terminal helix rotates 140° as a unit with respect to the N-terminal helix, and the conformation of the loop region between two helices significantly changes (35Rastogi V.K. Girvin M.E. Nature. 1999; 402: 263-268Crossref PubMed Scopus (418) Google Scholar). If a deprotonated F0 c subunit is sandwiched by F0 c subunits in the F0 c the deprotonated Asp-61 close to the Asp-61 of the F0 and proton among the F0 and essential of F0 a will Although the are the process by which the protons the F0 motor may be mechanical than a simple rotational of the rigid F0 c ring driven by electrostatic to the above of F0 c is of the F0 c in found in membranes of and in eukaryotic are a dimer of F0 c composed of transmembrane helices (reviewed in Ref. M. J. Biol. Chem. 1999; 274: Full Text Full Text PDF PubMed Scopus Google Scholar). Interestingly, the dimer contains only a single essential in the second Because the ring structure of these c subunits in is made most likely using two helices as a the question as to these a rotary motion with essential two This the on any models to the common of the ATP synthase and The value has two the concentration difference and the transmembrane Although they are energetically they can be proton in F0 the force by and can the F0 motor. the does not any force to each proton in Using the ATP synthase from which of a group G. J. 1998; PubMed Scopus (72) Google Scholar, G. U. J. Exp. Biol. 2000; 203: Google Scholar) indicated that a of is always for ATP synthesis when is a major of suggested that the ATP synthesis in the transition experiment of the induced G. J. 1999; PubMed Scopus Google Scholar). It is to that β subunits of F1 the torque by the F0 motor as a and only can up the of the β as as observed. ATP synthase is a rotary motor enzyme. The evidence for the F1 rotation has Boyer's in the This is not the but the of The central are the motor generates force and the motor is have been However, we are to a We know The F0 motor a of The observation of rotation in a membrane using the membrane will be a challenge but is not an atomic structures are a prerequisite to the F1 and F0