ATP Synthase: Two rotary molecular motorsworking together

George OsterHongyun Wang

University of California, Berkeley

Introduction

ATP synthase—also called FoF1 ATPase, or simply F-ATPase—is the universal protein that

terminates oxidative phosphorylation by synthesizing ATP from ADP and phosphate. Nearlyidentical proteins are found in eukaryotic mitochondria and bacteria, and they all operate on thesame principle. Electron driven ion pumps set up concentration and electrical gradients across amembrane. ATP synthase utilizes the energy stored in this electrochemical gradient to drivenucleotide synthesis. It does this in a surprising way: by converting the electromotive force into arotary torque that is used to promote substrate binding and to liberate ATP from the catalytic sitewhere it was formed. Remarkably, this process can be reversed in certain circumstances: ATPhydrolysis can drive the engine backwards so that F-ATPase functions as a proton pump. Indeed,the vacuolar V-ATPases—the most ubiquitous intracellular proton pumps—are structurally similarto ATP synthase and appear to operate according to the same principles.

ATP synthase is composed of at least 8 subunit types, whose stochiometry is denoted with

subscripts: (a3, b3, g, d, e, a, b2, c10-14), that combine into two distinct regions. The geometric

arrangement of the subunits is shown schematically in Figure 1. The F1 portion is soluble and

consists of a hexamer, denoted a3b3. This hexamer is arranged in an annulus about a central shaft

consisting of the coiled-coil g subunit. Subunits d and e generally isolate with F1 as well. The Fo

portion consists of three transmembrane subunits: a, b2 and c10-14. 10-14 copies of the c-subunit,

depending on the species, form a disk into which the g and e subunits insert. The remainder of Fo

consists of the transmembrane subunits a, and b2; the latter is attached by the d subunit to the a3b3

hexamer so that it anchors the a subunit to F1. Thus there are two ‘stalks’ connecting Fo to F1: ge

and b2d.

The key to understanding how ATP synthase carries out its catalytic and synthetic roles lies in thisgeometric organization. The entire protein can be divided into two operational regions denoted

2

suggestively as the ‘rotor’ and ‘stator’ for reasons that derive from the rotary mechanism by which

the protein operates. The rotor consists of gec10-14; the stator consists of a3b3db2a. Indeed, it turns

out that ATP synthase is two rotary engines in one. The Fo motor converts transmembrane

electrochemical energy into torque on the g shaft, and F1 uses ATP hydrolysis to turn the g shaft in

the opposite direction. Since they are connected, one drives the other in reverse: when Fo dominates,

the rotor turns clockwise (looking from the membrane, i.e. upwards in Figure 1) so that F1

synthesizes ATP. When F1 dominates, Fo is driven counterclockwise, whereupon it can pump

protons against an electrochemical gradient. Deciphering how this remarkable dual energytransduction works is one of the great triumphs of modern chemistry.

Figure 1.

ATP synthesis in F1 is driven by the rotation of the g-shaft

We begin with the F1 motor because we now know precisely what it looks like. This is due to John

Walker and his x-ray crystallography group at Cambridge, who worked out the structure of the

a3b3g assembly [1]. Figure 2A shows the view from the membrane (looking upwards in Figure 1)

of a3b3g. There are 6 nucleotide binding sites on the a3b3 hexamer, all lying at the interfaces

between the a and b subunits. The three catalytic sites alternate with three non-catalytic sites. Thecatalytic sites lie mostly in the b subunit, while the noncatalytic sites lie mostly in the a subunit.The role of the noncatalytic sites is uncertain, but may help hold the hexamer together. Figure 2Bshows a side view of one a and one b and g. Walker‘s structure revealed essential asymmetries inthe molecule’s structure that were the key to understanding its mechanism.

In the early 1980’s, Paul Boyer at UCLA, proposed the surprising theory that, in the catalytic sites

of F1, ATP was in chemical equilibrium with its reactants, ADP and phosphate [2]. So the

formation of ATP in the catalytic sites was essentially without energetic cost. However, since eachATP, when hydrolyzed under cellular conditions, liberates about 12 kcal/mol, this energetic price

must be paid at some point. Boyer proposed that F1 pays this price in the mechanical work

necessary to liberate the nucleotide from the catalytic site. Further, ATP release proceeds

sequentially and cyclically around the a3b3 hexamer because the synthetic reactions are

synchronized via the rotation of the g shaft and cooperative coupling between the three catalytic

3

sites. Boyer’s ‘binding change’ mechanism neatly fits the Walker structure. The way in which this

works is found in the shape of the a3b3 hexamer and the g shaft.

Figure 2.

The g subunit is asymmetric and bowed. It fits into a central annulus in the a3b3 which is itself

asymmetric (Figure 2B). At the top of the a3b3 hexamer is a hydrophobic ‘sleeve’ in which the g

shaft rotates. Further down, however, the annulus is offset from the center, so that as g rotates, itsequentially pushes outwards on each catalytic site. Figure 3 shows the relation between the grotation and the conformational change of b and catalytic site. In the hydrolysis direction, bendingof the lower part of b toward the upper part drives the counterclockwise rotation of the g (lookingfrom the membrane, upwards in Figure 3): panels A Æ B Æ C in Figure 3. In the synthesisdirection, clockwise rotation of the g forces the opening of the b and the catalytic site, reducing theATP binding affinity: panels C Æ B Æ A in Figure 3.

The catalytic sites do not act independently; rather they are synchronized so that each site traversesthe synthetic cycle in a more or less fixed phase with respect to the others. This synchronization isorchestrated in two ways. As the g shaft rotates, it not only stresses each catalytic site, but it alsointeracts electrostatically with the b subunits at two locations [3]. These interactions may mediatephosphate and nucleotide binding, the necessary precursors to synthesis. In addition, the catalyticsites appear to be elastically coupled so that the occupancy of one site affects the other two sites.The consequence of this coupling is that, when ATP concentrations are low enough so that only onesite is occupied, hydrolysis proceeds much more slowly than when more than one site is occupied.

Together with the F1 molecular structure, the binding change model strongly supported the idea that

catalysis involved rotation of the g subunit. However, dramatic visual confirmation was provided by

in vitro experiments in which the a3b3g subunits were isolated and attached to a bead. A

florescently tagged actin filament was attached to the g shaft and, when ATP was supplied, thefilament could clearly be seen to rotate. In fact, a complete revolution took place in 3 steps, andconsumed a single ATP per step [4]. Recently, more detailed experiments indicated that each 120˚rotation appears to take place in two steps: an initial 90˚ rotation that depends on the solutionconcentration of ATP, followed by a 30˚ rotation that is ATP independent [5]. The initial rotation isassociated with ATP binding, and the second rotation with product release.

4

The viscous drag on the actin filament could be estimated, which allowed the torque developed by

the F1 motor to be computed and compared with the free energy available from ATP hydrolysis.

The startling result was that the motor generated an average torque of more than 40 piconewton

nanometers (40¥10-12 N ¥ 10-9 m), more than six times the maximum force developed by kinesin

or myosin. More impressively, the motor operated near 100% mechanical efficiency; thisprecludes any sort of heat engine that would be limited by the Carnot efficiency [6]. Severalmodels have been proposed that address the issue of torque generation and efficiency (see Box 1)[6-9].

The energy to drive rotation derives from the hydrolysis cycle of ATP at the catalytic site.Moreover, the conformational change that drives the hydrolysis motor must be nearly the reverse ofthe motion that frees ATP from the catalytic site during synthesis. Examination of Walker’sstructure reveals that the major conformational change is a hinge-bending motion of about 30˚ in theb subunits. The lower part of each b below the nucleotide binding site rotates inward and upwardsduring which it pushes on the bowed g subunit, turning it much like one cranks an automobile jack(Figure 3) [8].

Figure 3

5

Box 1: How is torque generated?

The above description of the F1 motor begs the basic question of how the chemical energy liberated

in the hydrolysis cycle is converted into a mechanical torque. One solution (the ‘binding zipper’model) proposes that, after an ATP diffuses into the catalytic site, the torque is generated during thetransition from weak to tight binding of the ATP onto the catalytic site [10] [9]. Recent worksupports this model [5, 11]. This accords with the observation that the power stroke of GroELaccompanies ATP binding [12]. Figure 4 illustrates the ‘binding zipper’ model. What remainsunsettled is the precise role of the hydrolysis step. It appears that both ATP binding at one catalyticsite and hydrolysis at a second catalytic site are required to complete a 120˚ power stroke [11].Moreover, the power stroke may proceed in two steps: a 90˚ step associated with ATP binding, anda 30˚ step associated with hydrolysis or product release [5], What is not clear is whether thehydrolysis step actually contributes to torque generation, or is simply permissive for the completionof the binding transition of ATP at another site. The binding zipper model posits that theelectrostatic repulsion of the hydrolysis products is used not for torque generation, but to loosen thetight binding of ATP to the catalytic site so that they can be released to repeat the power cycle.There may be an elastic ‘recoil’ that contributes to the power stroke as the b subunit relaxes back toits open conformation [10] [9].

Figure 4

Fo converts protonmotive force into rotary torque

There is currently no direct observation of rotation in the Fo portion of ATP synthase [13].

However, current thinking is that the Fo assembly converts the energy contained in the

transmembrane protonmotive force into a rotary torque at the interface between the a and the csubunits (Figure 1). This torque turns the rotor (the c subunits, g and e subunits) which couples to

the F1 synthetic machine.

6

The c assembly consists of 10-14 subunits, each consisting of two transmembrane a-helices [14][15] [16]. There is one essential acidic amino acid (Asp61 in the E. Coli ATP synthase) whichbinds protons. Since there are variants of ATP synthase that can operate on sodium rather thanprotons, the interaction between the c subunit and the translocated ion has the property of anelectrostatic carrier mechanism [17].

The a subunit consists of 6 transmembrane a-helices which contain at least one essential basicresidue (Arg210 in E. Coli) [16, 18]. The interaction of these rotor and stator charges is essentialfor torque generation, and several proposals have been put forward for how this could work [19]

[20] [21] [22] [23, 24]. Whatever the mechanism, the Fo motor must generate a torque sufficient to

liberate 3 ATP’s from the three catalytic sites in F1 for each revolution.

Under anaerobic conditions, the ATP synthase of the bacteria E. Coli can reverse its operation,hydrolyzing ATP and turning the c subunits backwards so that it functions as a proton pump. Thisis not surprising, since the F-ATPases are structurally similar to the most common proton pumps,the vacuolar, or V-ATPases [13]. These pumps may have been the evolutionary precursors of ATPsynthase [25]. A striking difference between the two is that the F-ATPases have 10-14 acidic rotorcharges, whereas the V-ATPases have ~6. It can be shown that this enables the V-ATPases tofunction more efficiently as ion pumps, at the expense of relinquishing their capability to synthesizeATP [26].

Summary

Both the F1 and Fo motors can operate in both directions. F1 is a hydrolysis-driven 3-cylinder

engine which can be driven in reverse to synthesize ATP from ADP and phosphate. Fo is an ion-

driven rotary engine which can be driven in reverse to function as an ion pump. The F-ATPases arestructurally similar to, and presumably evolutionarily related to, the V-ATPase ion pumps [25].Most ion pumps are thought to function by an ‘alternating access’ mechanism whereby an ion isfirst bound strongly on the dilute side, then energy is supplied to move the ion such that itcommunicates with the concentrated side and to weaken its binding affinity [27]. However, incontrast with other ion pumps, the F and V-ATPases accomplish this by a rotary mechanism that is

driven indirectly by nucleotide hydrolysis, rather than by direct phosphorylation [28]. The Fo

motor is thought to be related also to the bacterial flagellar motor. Both can operate on sodium,

although the flagellar motor has 8 or more ‘stators’ and develops far more torque than Fo [29] [30].

7

The mechanism driving the F1 hydrolysis motor may carry hints for other nucleotide hydrolysis

fueled motors, such as kinesin, myosin and dynein. However, there are important structuraldifferences that may make the comparison difficult [6]. For example, the above mentioned motorsall ‘walk’ along a polymer track to which they bind tightly during a portion of their

mechanochemical cycle. The power stroke of the F1 motor is driven by the b subunit which pushes

on the g shaft, but does not appear to bind tightly to it; that is, it does not ‘walk’ around the g shaft.

Moreover, no other motor operates with nearly the efficiency as the F1 motor, implying that there

are important entropic steps in other motors that are absent in the F1 motor.

8

References1. Abrahams, J., et al., Structure at 2.8Å resolution of F1-ATPase from bovine heart

mitochondria. Nature, 1994. 370: p. 621-628.2. Boyer, P., The binding change mechanism for ATP synthase--some

probabilities and possibilities. Biochim. Biophys. Acta, 1993. 1140: p. 215-250.3. Al-Shawi, M., C. Ketchum, and R. Nakamoto, Energy coupling, turnover, and

stability of the F0F1 ATP synthase are dependent on the energy of interactionbetween g and b subunits. J. Biol. Chem., 1997. 272(4): p. 2300-2306.

4. Yasuda, R., et al., F1-ATPase is a highly efficient molecular motor that rotateswith discrete 120° steps. Cell, 1998. 93: p. 1117-1124.

5. Yasuda, R., et al., Resolution of distinct rotational substeps by submillisecondkinetic analysis of F1-ATPase. Nature, 2001. 410: p. 898-904.

6. Kinosita, K., et al., F1-ATPase: A rotary motor made of a single molecule. Cell,1998. 93: p. 21-24.

7. Oosawa, F. and S. Hayashi, The loose coupling mechanism in molecularmachines of living cells. Adv Biophys, 1986. 22: p. 151-183.

8. Wang, H. and G. Oster, Energy transduction in the F1 motor of ATP synthase.Nature, 1998. 396: p. 279-282.

9. Oster, G. and H. Wang, Why is the mechanical efficiency of F1-ATPase so high?Journal of Bioenergetics and Biomembranes, 2000. 32(5): p. 459-69.

10. Oster, G. and H. Wang, Reverse engineering a protein: The mechanochemistryof ATP synthase. Biochimica et Biophysica Acta (Bioenergetics), 2000. 1458(1-2): p. 482-510.

11. Menz, R., J. Walker, and A. Leslie, Structure of Bovine Mitochondrial F1-ATPasewith Nucleotide Bound to All Three Catalytic Sites: Implications for theMechanism of Rotary Catalysis. Cell, 2001. 106: p. 331-341.

12. Bukau, B. and A. Horwich, The Hsp70 and Hsp60 chaperone machines. Cell,1998. 92(3): p. 351-366.

13. Finbow, M. and M. Harrison, The vacuolar H+-ATPase: a universal proton pumpof eukaryotes. Biochem. J., 1997. 324: p. 697-712.

14. Stock, D., A. Leslie, and J. Walker, Molecular architecture of the rotary motor inATP synthase. Science, 1999. 286: p. 1700-1705.

15. Seelert, H., et al., Proton-powered turbine of a plant motor. Nature, 2000. 405: p.418-419.

16. Fillingame, R.H., Coupling H+ transport and ATP synthesis in F1F0-ATPsynthases: glimpses of interacting parts in a dynamic molecular machine. J.Exp. Biol., 1997. 200(Pt 2): p. 217-224.

17. Dimroth, P., Primary Sodium Ion Translocating Enzymes. Biochim. Biophys.Acta, 1997. 1318(1-2): p. 11-51.

18. Fillingame, R.H., Membrane sectors of F- and V-type H+-transporting ATPases.Current Opinion in Structural Biology, 1996. 6(4): p. 491-8.

19. Elston, T., H. Wang, and G. Oster, Energy transduction in ATP synthase. Nature,1998. 391: p. 510-514.

9

20. Dimroth, P., et al., Energy transduction in the sodium F-ATPase ofPropionigenium modestum. Proc. Natl. Acad. Sci. USA, 1999. 96(9): p. 4924-4929.

21. Junge, W., H. Lill, and S. Engelbrecht, ATP Synthase: An Electro-ChemicalTransducer with Rotatory Mechanics. Trends Biochem. Sci., 1997. 22(11): p.420-423.

22. Vik, S.B. and B.J. Antonio, A mechanism of proton translocation by F1F0 ATPsynthases suggested by double mutants of the a subunit. J. Biol. Chem., 1994.269: p. 30364-30369.

23. Fillingame, R.H., W. Jiang, and O.Y. Dmitriev, Coupling H+ transport to rotarycatalysis in F-type ATP synthases: structure and organization of thetransmembrane rotary motor. J. Exp. Biol., 2000. 203(1): p. 9-17.

24. Rastogi, V. and M. Girvin, Structural changes linked to proton translocation bysubunit c of the ATP synthase. Nature, 1999. 402: p. 263-268.

25. Cross, R. and L. Taiz, Gene duplication as a means for altering H+/ ATP ratiosduring the evolution of F0F1 ATPases and synthases. FEBS Lett., 1990. 259(2):p. 227-229.

26. Grabe, M., H. Wang, and G. Oster, The mechanochemistry of the V-ATPaseproton pumps. Biophys. J., 2000. 78(6): p. 2798-281.

27. Alberts, B., et al., Molecular Biology of the Cell. 3d ed. 1994, New York: Garland.28. Khan, S., Rotary chemiosmotic machines. Biochimica et Biophysica Acta, 1997.

1322(2-3): p. 86-105.29. Berg, H., Torque generation by the flagella rotary motor. Biophys. J., 1995. 68(4

Suppl): p. 163s-166s.30. Muramoto, K., et al., High-speed rotation and speed stability of the sodium-

driven flagellar motor in Vibrio alginolyticus. J. Mol. Biol., 1995. 251: p. 50-58.31. Pedersen, P.L., Y.H. Ko, and S. Hong, ATP Synthases in the Year 2000:

Evolving Views about theStructures of These Remarkable Enzyme Complexes. Journal of Bioenergetics and

Biomembranes, 2000. 32(4): p. 325-32.32. Uhlin, U., G.B. Cox, and J.M. Guss, Crystal structure of the epsilon subunit of

the proton-translocating ATP synthase from Escherichia coli. Structure, 1997.5(9): p. 1219-30.

33. Dmitriev, O., et al., Structure of the membrane domain of subunit b of theEscherichia coli F0F1 ATP synthase. Journal of Biological Chemistry, 1999.274(22): p. 15598-604.

34. Bianchet, M., et al., The 2.8-Å structure of rat liver F1-ATPase: Configuration of acritical intermediate in ATP synthesis/hydrolysis. Proc. Natl. Acad. Sci. USA,1998. 95: p. 11065-11070.

Supplementary Reading1. J. Weber, A. Senior, Biochim. Biophys. Acta 1319, 19-58 (1997).

2. M. Yoshida, E. Muneyuki, T. Hisabori, Nature Reviews, 1-21 (2001).

10

Figure CaptionsFigure 1. Structural models proposed for ATP synthases constructed by Pedersen, Ko and Hong

[31], based on the crystal structure for a3b3gc10 (1qo1) [14] and other crystal structures for

various subunits (1aqt, 1b9u, 1bmf, 1mab, 1c17) [32] [33] [1] [34] [24]. The c assembly consists of10-14 c subunits [14] [15]. The g subunit attaches to the c subunits. The a subunit is attached to

the a3b3 hexamer via the b2 and d subunits. Torque is generated by the protonmotive force at the

interface of the a and c subunits. This leads to the functional subdivision into two counter-rotating

assemblies, usually denoted as the ‘rotor’ and ‘stator’. The rotor consists of subunits gec10-14 and

the stator consists of subunits a3b3db2a.

Figure 2. (A) The view from the membrane (looking upwards in Figure 1) of a3b3g. (B) A side

view of only one a, one b and the g. The a subunit is in yellow, the two coils of the g subunit areblue and purple. The b subunit is in two colors: the stationary upper barrel segment is in green, andthe lower hinge segment is red. Three catalytic sites are indicated by arrows. In the Walker

structure, the three catalytic sites have different occupancies: one is empty (bE); another one is

occupied by an ADP (bD) and the third one is occupied by an AMPPNP (bT).

Figure 3. Side views of one a, one b and the g. Three panels show the relation between the rotationof the g and the bending of the b. (A) The b is open. (B) The b is bent halfway upwards and the gis turned counterclockwise (looking upwards). (C) The b is bent more and the g is further turned inthe counterclockwise direction. The color scheme used here is the same as that in Figure 2. Abutthe catalytic site, there are two a helices: Helix B and Helix C. Helix B is associated with the lowerpart of the b; Helix C is associated with the upper part. In the hydrolysis direction, the ATPbinding transition drives Helix B towards Helix C, which drives the bending of b. The bendingmotion of b pushes on the eccentric g coiled-coil causing it to rotate within the barrel bearing

(panels A Æ B Æ C). In the synthesis direction, the torque from Fo drives the clockwise (looking

upwards) rotation of the g. The rotation of the eccentric g pushes on the lower part of b and forcesHelix B to move away from Helix C, which reduces the ATP binding affinity (panels C Æ B ÆA). Movies of the rotational sequence can be downloaded from the authors’ web site.

Figure 4. The binding zipper model. The hydrolysis cycle proceeds clockwise around the diagram,synthesis is counterclockwise. In the hydrolysis direction, an ATP in solution first diffuses to the

11

catalytic site and is weakly bound (ATP docking). The rate of this step is affected by the ATPconcentration in solution. The weakly bound ATP may dissociate from the catalytic site, returningto the solution. Occasionally, it proceeds from weak binding to tight binding (the bindingtransition). During the binding transition, the bonds between the ATP and the catalytic site formsequentially, and the binding affinity increases gradually. In this way the conformational change ofthe catalytic site is coupled to the binding affinity. Thus, during the multi-step ATP bindingtransition, the binding free energy is used efficiently to generate a nearly constant force to rotate theg shaft. ATP concentration in solution does not affect the binding transition, but only how oftenATP attempts docking to the catalytic site. After the binding transition, the ATP is in chemical

equilibrium with ADP and Pi. The transition ATP ‡ ADP + Pi weakens the ATP binding and

distributes it over ADP and Pi so that the hydrolysis products can be released and the cycle

repeated. Release may require the assistance of nucleotide binding to an adjacent catalytic site. Therotation of g concurrent with ADP and Pi release is driven by the ATP binding transition at theother two catalytic sites (not shown in the figure). In the binding zipper model, the rotation of the gshaft is continuous and the ATP binding affinity is continuously coupled to the g rotation [10] [9].The synthesis direction is counterclockwise around the diagram. A catalytic site first binds ADPand phosphate (not necessarily in that order). While trapped in the catalytic site, reactants (ADP and

Pi) and product (ATP) are in chemical equilibrium with equilibrium constant ~ 1. ATP is formed at

the catalytic site from ADP and Pi with little energy cost. However, the ATP is tightly bound at the

catalytic site. The mechanical torque from Fo on g is used to reduce the ATP binding affinity. In

this step, the mechanical torque is converted to the binding free energy of ATP. Most of the 12kcal/mol price of synthesis is paid in this step. When the binding affinity is low enough, the ATP isdislodged from the catalytic site by thermal fluctuations. Molecular dynamics simulations confirmthis sequence of events.

α3β3

γ

ε

δ

b2

a

c10-14

F1

Fo Membrane

βT

βE

βD

α

α

α

α

γβT

α γ β

Helix B

Helix C

Helix B

Helix C

Helix B

Helix C

γβ1

β2 β3

ATP

• • •

ADP

AT

P do

ckin

g

γβ1

β2β3

ADP

Multistep binding transitiondrives rotation

Hyd

roly

sis

Keq

~ 1

Pi

γβ1

β2 β3

ATP

γβ1

β2 β3

ADP+Pi

Phosphaterelease

ADPrelease

ATP

γβ1

β2 β3

γ

γγ