Molecular Features of Energy Coupling in the F0F1 ATP Synthase

Robert K. Nakamoto


H+ translocation is coupled to ATP synthesis in the F0F1 ATP synthase via a rotary mechanism. Catalytic turnover, site-site cooperativity, and H+ transport obligatorily involve rotation of a set of subunits. The transport domain in the membranous F0 and the catalytic domain in the F1 are mechanisms designed for generating torque.

One of the most frequently occurring reactions in biology is the synthesis of ATP. The vast majority of this production is carried out by a nearly ubiquitous multisubunit complex. Whether in mitochondria, chloroplasts, or eubacteria, the F0F1 ATP synthase couples the energy stored in the electrochemical gradient of protons (the proton motive force or ΔμH+ generated by electron transport chains) to form a high-energy phosphoanhydride bond from ADP and inorganic phosphate (Pi). The process of coupling the “downhill” transport of protons to the ATP synthesis reaction may be approached with the same thermodynamic considerations as the “classic” pumps, which are more commonly thought of as active transporters, such as the P-type sodium/potassium or calcium ATPases. In fact, bacteria in anaerobic conditions utilize the F0F1 in the reverse role as a primary proton pump using ATP derived from glycolysis to drive a ΔμH+. Structurally and mechanistically, however, the F0F1 is quite different from the P-type pumps.

Understanding the molecular features of energy coupling in all active transport proteins remains an elusive goal. The long-standing nature of this problem has been caused in part by a dearth of structural information describing the large, integral membrane proteins, some of which are multiple-subunit complexes. Recently, the structural knowledge of the F0F1 ATP synthase has dramatically increased. Although only a partial structure of the soluble F1 domain was obtained from bovine mitochondria (1), the X-ray crystallographic structure at 2.8-Å resolution has led to several insights into the catalytic and coupling mechanisms. Furthermore, the structure has provided the basis for novel experimental approaches that have demonstrated the mechanical rotation of at least two subunits driven by the chemical reaction of ATP hydrolysis. From these experiments and others, it has become clear that rotary movement of a core of subunits is an integral part of the catalytic mechanism and probably transport as well. Models derived from these experiments hypothesize that free energy gained by proton transport down its electrochemical gradient is converted into torque that is used to drive ATP synthesis. As a proton pump, the enzyme works in the opposite direction by reverse of the same pathway; free energy released from hydrolysis of ATP is applied to generate a torque that drives proton translocation. On the heels of the revelation of rotational catalysis, it has been quickly realized that coupling to transport is not purely mechanical and there is specific conformational information transmitted between the two functions. This review will focus on the molecular features of the rotary complex and how it functions in coupling catalysis to transport.

Structural elements of rotor and stator

The F0F1 complex can be separated into two sectors that are well defined biochemically: the membrane-extrinsic F1 and the membrane-intrinsic F0. In Escherichia coli, the system for which the most mutagenic data have been obtained, the soluble F1 sector consists of five different subunits with three copies each of the homologous α- and β-subunits and single copies of γ, δ, and ϵ (Fig. 1). The F0 sector has three different subunits, all of which are integral membrane polypeptides: a single copy of a, two b, and 9–12 subunit c. F1, which can be reversibly dissociated from F0, by itself catalyzes hydrolysis of ATP. There is a nucleotide binding site at each of the α-β and β-α interfaces. The three sites found predominantly in the α-subunits do not carry out catalysis, and exchange nucleotides only very slowly, whereas those found in the β-subunits catalyze ATP synthesis or hydrolysis and turn over nucleotides in a kinetically competent manner. F0 mediates passive H+ translocation upon removal of F1. Coupled transport occurs only when F1 is bound to F0, and at least in the case of the E. coli complex, assembly of the functional complex requires all eight subunits. F0F1 from mitochondria have additional subunits, but homologs of the bacterial subunits are found in the mitochondrial or chloroplast complexes. Based on extensive structural and biochemical analyses, the homologous subunits carry out similar roles, and mechanistic information is generally relevant to all F0F1, regardless of the source.

With consideration of the rotational catalysis model, subunits may be grouped in a different manner as belonging to the stator or rotor. The α- and β-subunits alternate in a hexamer “like the segments of an orange” (Fig. 1) as revealed by the X-ray crystallographic structure of bovine mitochondrial F1 at 2.8 Å (1). Significantly, the structure also revealed the relationship of the terminal helixes of the γ-subunit, which together form a coiled-coil structure that occupies the core of the orange. The observation of the α3β3 hexamer poised on the coiled coil strongly suggested that the γ-subunit could act as a shaft for rotation of the hexamer.

Although present in the crystal of mitochondrial F1, most of the γ-subunit and the entire δ (homolog of bacterial ϵ)- and ϵ (no homolog in bacteria)-subunits were not represented in the electron density map (1). The structural relationship of these subunits has been surmised from biochemical studies. Dunn (6) showed that the ϵ-subunit is tightly associated with the γ-subunit and is therefore likely a part of the rotor. One other subunit, F0 subunit c, may also be a part of the rotor. Subunit c is a simple hairpin loop with two membrane spanning α-helical segments connected by a short polar loop. Functional interactions between ϵ- and c-subunits were first recognized by identification of intergenic suppressor mutants. The effects of subunit c mutation cGln42 to Glu were suppressed by replacements of ϵGlu31 to Gly, Lys, or Val (15). Subsequently, cysteine replacements of the same residues were found to form disulfide bonds confirming their structural proximity. The γ-subunit is also involved in interactions in this region of the complex. Watts et al. (14) found that γ-subunit residue γTyr205 to Cys formed disulfide bonds with cysteine replacements of subunit c residues 42, 43, and 44 (some of the same residues that formed disulfides with ϵ-subunit). Furthermore, the Cys in place of γTyr205 also formed disulfides with cysteines at ϵ-subunit positions 38 and 43. A functional coupled enzyme with c-subunits cross-linked to the γ- or ϵ-subunit would be strong evidence that they are part of the rotor. Unfortunately, because of the chemistry required to induce formation of disulfides and the tendency of the subunit c cysteines to cross-link forming homodimers, enzymes could not be functionally tested and it was not clear whether the ϵ-c or γ-c disulfide-linked enzymes were capable of coupled transport.

Despite the lack of strong experimental evidence that the c-subunits are part of the rotor, recently published models of the arrangement of F0 subunits and the mechanism of transport accommodate this notion. Models such as that first presented by Vik and Antonio (13), and elaborated by others, suggest that the multiple copies of c-subunits form a ring with subunit a on one side (Fig. 1). An important feature of this model is that the c-subunits and their essential carboxylic acid group in the middle of the bilayer (cAsp61 in E. coli) face the lipid milieu, except for the two that face subunit a. H+ transport is proposed to be mediated between a and the ring of c-subunits, a mechanism that involves rotation of the ring relative to subunit a (see below). Certainly, knowledge of the quaternary structure of the F0 complex would help a great deal in judging whether or not such models are reasonable. A number of attempts have been made to visualize or model the structure of the F0 subunits, but major questions remain to be resolved.

A consequence of this model is that subunit a must be part of the stator. Recent evidence from several laboratories suggests that subunit a and the α3β3 hexamer are attached through F0 subunit b and F1 δ-subunit. The part of the stator linkage that is yet to be demonstrated is the presumed interaction between subunits a and b.

Evidence for rotational catalysis

Boyer first proposed the concept of rotational catalysis based principally on 32P and 18O exchange studies (reviewed in Ref. 4). These studies suggested that all three of the β-subunit catalytic sites participated equally in steady-state catalysis and in a sequential manner. In conditions of ATP hydrolysis, the first nucleotide binds with very high affinity, and catalysis in this first site can be studied in “unisite” conditions where substoichiometric amounts of ATP are allowed to react with the enzyme. In this manner, rate constants for the elementary reaction steps can be determined in the absence of the complication of site-site cooperativity. These steps are

Embedded Image

When only one site is occupied, the rate constants for release of Pi and ADP are very slow (kcat ≈ 1×10-3 s-1). These steps are accelerated ~105-fold when ATP binds to the other two sites. According to the “binding change mechanism,” this cooperative “multisite” catalysis proceeds in a rotational fashion sequentially through each of the three sites (4).

ATP synthesis follows the reverse of the hydrolytic catalytic pathway (3). In the absence of a ΔμH+, or for F1 alone, there is essentially no binding of Pi. To achieve ATP synthesis, the presence of a ΔμH+ increases the affinity for Pi approximately seven orders of magnitude, thus conferring tight binding in the catalytic site and proper coordination of the substrates for formation of ATP. The ΔμH+ also causes a decrease in the affinity for newly formed ATP so that the enzyme can release the product from the tightly bound state and proceed to the next cycle.

Taking advantage of the crystallographic structure, investigators have developed a variety of approaches to demonstrate that rotational catalysis does indeed occur and that it is the γ-subunit that rotates relative to the α3β3 hexamer during multisite catalysis. Duncan et al. (5) found that a β-γ cross-link can be switched to other β-subunits upon hydrolysis of ATP, demonstrating that the γ-subunit changed its orientation towards different β-subunits during the course of turnover. Sabbert et al. (11) took a different approach and monitored rotation in real time by using polarized absorption relaxation, after photobleaching of an eosin probe attached to the γ-subunit in an F1 complex mounted on a Sephadex bead. Their experiments indicated that the ATP-dependent rotational rate was kinetically competent. Perhaps the most compelling experiments were done by Yoshida and co-workers (10). They tagged the “membrane” end of the γ-subunit with a fluorescently decorated actin filament and mounted the α3β3γ complex on a glass slide via the “top” (opposite from the membrane) of the complex. Rotation of the actin filament observed in the fluorescence microscope occurred only with hydrolysis of ATP, and was stopped by inhibitors of ATP hydrolytic activity (i.e., azide). Two significant properties of the rotation were observed. 1) The direction of rotation was uniform. Rotation was always counterclockwise as observed from the “bottom” or membranous side of the hexamer. 2) The rotary motion, observed in a small fraction of immobilized actin filaments, was generally not continuous and had stochastic elements. The stepped motion was not unexpected, because the rotation is driven by catalytic turnover and discrete rate-limiting transition states must be overcome during the course of catalysis.

Several corollaries can be derived from these experiments.

  1. Multisite catalysis obligatorily involves rotation of the γ-subunit. In each of the experiments described above, ATP hydrolysis was required to observe the effects of rotation. Cooperative catalysis and site-site interactions require the γ-subunit. Because the γ-subunit confers asymmetry to the three sites and each site must pass through the three conformations observed in the bovine F1 structure (1), the γ-subunit must rotate to drive each site through its catalytic conformations (Fig. 2).

  2. The direction of rotation must be in one direction for hydrolysis and the other for synthesis. The direction of rotation observed by Noji et al. (10) was predicted by the binding change mechanism as proposed by Boyer (reviewed in Ref. 4) and the relative positions of the three conformational states of the catalytic sites as observed in the F1 structure (1) (see below and Fig. 2).

  3. Because all three sites participate in steady-state catalysis, the catalytic cycle in each site must be offset from the others by 120°. There is one turnover (synthesis or hydrolysis of 1 ATP) per 120° of rotation or three turnovers per revolution.

The role of the rotor in catalysis

As mentioned above, rotation of the γ-subunit drives each of the three nucleotide sites through their catalytic cycles by presenting a different interface to each β-subunit and therefore imposing a different conformation on each β-subunit (Fig. 2). This role of the γ-subunit in catalysis is supported by two important observations. First, the γ-subunit is required for the asymmetry of the three β-subunits. In the absence of the γ-subunit, the β-subunits are identical, as found in the X-ray crystallographic structure of the Bacillus PS3 α3β3 complex (12). Secondly, the asymmetry is essential for attaining maximal rates of catalysis. α3β3γ is the minimal complex required for cooperative multisite catalysis (7). Without the γ-subunit, the α3β3 complex is capable of catalyzing ATP hydrolysis only at low rates and lacks the cooperative behavior that is a hallmark of the enzyme.

The above results suggest that specific γ-β interactions are critical for cooperative catalysis. The role of some of these interactions was demonstrated by analysis of mutant enzymes with substitutions of amino acids in the γ-β interface. The two critical areas of interactions as defined by mutagenesis are the γ-subunit carboxyl terminus from γ269 to γ286 (E. coli numbering) and a portion of the γ-coiled coil involving residues γ18–35 and γ236–246 (see Fig. 3; reviewed in Ref. 9). The former region interacts with the β-subunits near the nucleotide binding sites. Mutations of two conserved residues in this part of the γ-subunit, γGln269 and γThr273, greatly reduce the catalytic turnover and appear to disrupt site-site cooperativity. The latter region interacts with the conserved β-subunit motif, β380DELSEED386, and, as discussed below, is involved in the transmission of coupling information.

Some γ-subunit mutations demonstrated that γ-β interactions play critical roles in energy coupling between catalysis and transport. One mutation in particular, replacement of the conserved γMet23 with Lys, was most revealing of this process, because this mutation caused inefficient energy coupling between catalysis and transport. From Arrhenius analysis of steady-state turnover, the mutation caused a dramatically increased activation energy, which suggested that an extra bond must be broken for the enzyme to achieve the rate-limiting transition state (2). This is true for both ATP hydrolysis and synthesis pathways. According to the crystallographic structure, a Lys residue in place of γMet23 is in a position to form an ionized hydrogen bond with the conserved β-subunit residue βGlu381 (in the β380DELSEED386 motif), hence the extra bond detected by the Arrhenius plots. Kinetic analysis suggests that the additional energy of interaction between the γ- and β-subunits impedes the rotation by adding a bond that must be broken before the rotation can proceed.

The above hypothesis was reinforced by the behavior of several second-site mutations that suppressed the effects of γMet23 to Lys. One change was γArg242 to Cys, several others were found between γGln269 and γVal280, and other changes were of βGlu381 (Ala, Asp, and Gln; see Fig. 3; reviewed in Ref. 9). The crystallographic structure revealed that each mutation affected γ-β interactions, including specific interactions between conserved residues that formed intersubunit hydrogen bonds (1, 2). Again, Arrhenius analysis proved useful in substantiating this observation. Several of the second-site suppressor mutations, including substitutions of βGlu381, reversed the increased energy of interaction between γ- and β-subunits that was originally caused by the γMet23 to Lys substitution.

Pre-steady-state kinetic analysis of the γMet23 to Lys mutant enzyme revealed that the γ-subunit also played critical roles in determining the conformation of the catalytic sites. The γMet23 to Lys mutant enzyme had dissociation rate constants for Pi and ATP that were increased ~50-fold and 8-fold, respectively. The mutation was found to prevent proper utilization of binding energy to drive catalysis and specifically caused an altered transition state structure for the binding and release of Pi. These results were consistent with the use of energy from ΔμH+ for increasing the affinity for Pi so that the substrate binds in a catalytically competent manner for synthesis of ATP. These analyses demonstrated that the interactions between γ-subunit and the β380DELSEED386 motif communicate specific information essential for the proper conformation of the catalytic sites. Even though these interactions are distant from the catalytic sites (an average of ~40 Å), subtle amino acids changes (e.g., βE381 to Asp) have significant effects on catalytic kinetics and thermodynamics.

The role of the rotor in transport

The mechanism by which torque is generated by ΔμH+ via the transport mechanism is much more speculative at this time. There is no direct evidence for rotatory motion, and, as described above, the quaternary arrangement of the F0 sector is still to be clarified. Nevertheless, evidence is mounting that the transport mechanism is within the F0 and that c-subunits are a part of the rotor having interactions with both γ- and ϵ-subunits (see above and Fig. 1). Vik and Antonio (13) have proposed a reasonable model for the proton pathway based largely on mutagenesis studies but without support of substantive structural information, especially for subunit a and the arrangement of c-subunits. Nevertheless, the mechanics for this model used to explain the generation of torque, as detailed by Elston et al. (8), are quite plausible.

Generation of torque in the transport mechanism relies on the energetically favorable rotation of a protonated and neutralized acidic group (cAsp61) into the bilayer (see Fig. 1). Protonation of cAsp61 occurs on the single subunit c that is exposed to a half-channel that has access from the outside or periplasmic space. This subunit c cannot rotate into the bilayer because of the high energetic cost of putting an unmasked charge in the lipid milieu. Furthermore, the subunit c is prevented from rotating in the reverse direction towards the other half-channel by the precise positioning of a stator positive charge on subunit a, possibly aArg210. Once protonated and neutralized, this subunit c can rotate away from subunit a into the bilayer. The proton pathway continues as the protonated cAsp61 rotates around until it comes into juxtaposition with the other half-channel, which has access to the opposite side of the membrane. Here, the proton must dissociate before the subunit c can rotate to face the first half-channel and pick up another proton. Movement restricted within the plane of the membrane is largely driven by thermal motions (kT). Importantly, the electrostatic factors introduced by the stator positive charge (aArg210) bias rotation in the correct direction and strictly couple rotation to proton transport. Furthermore, the model allows not only generation of ΔμH+-driven torque but also the reverse-direction proton pumping driven by torque generated by ATP hydrolysis.

This model provides a solution for determination of the coupling ratio, which is 3–4 protons translocated per ATP synthesized or hydrolyzed. If there are 12 c-subunits in the ring, then 4 protons are translocated per ATP synthesized (4 c-subunits per 120° of rotation or per catalytic site). It is possible that variation in the number of c-subunits will alter the coupling ratio of protons per ATP. This principle may have already been demonstrated by the evolutionarily related V-ATPase, which operates with a coupling ratio of ~2 protons pumped per ATP hydrolyzed. The subunit c homolog, known as the 16-kDa subunit, has only one carboxylic acid per four membrane spanning segments. Assuming that the same total number of membrane spanning segments constitute the V-ATPase 16-kDa subunit oligomer, the number of protons translocated per rotation would be one-half that of the F0F1, or 2 protons per ATP hydrolyzed.


The above model for transport must also consider coupling to catalysis. As described above, specific interactions have been demonstrated between the γ-rotor and the α3β3 catalytic hexamer as well as the proton-transporting a-c complex. The rotor links the state of the catalytic sites to the transport mechanism, and these functions are not independent of each other. A major question is how torque is applied at the proper step of the catalytic pathway. In other words, how are four seemingly equivalent transport steps applied toward achieving discrete transition states of catalysis? In the same consideration, torque generated by ATP hydrolysis will differ for different c-subunits, depending on the relative position in the ring. A possible solution is to link the functions with a spring. This spring may be within the stator or connecting the halves of the rotor between F0 and F1, for example, in the γ- and ϵ-subunits near the interface with the c-subunits. A spring in the γ- and ϵ-subunits may explain the lack of electron density representing these subunits because the flexibility prevents conforming to the F1 crystal lattice. The spring allows storage of energy while the transport mechanism continues to generate torque, and the catalytic mechanism can use it as required. Future experiments will emphasize clarification of structural details of the F0 sector, testing of the rotation model for proton transport, and elucidating the roles of the transmission of conformational information between the catalytic and transport functions.


Subunit arrangement of the Escherichia coli F0F1 ATP synthase and the proposed proton pathway in F0. The α3β3 hexamer containing the catalytic sites in each of the β-subunits surrounds the rotor shaft made up of the γ-subunit coiled coil. The rest of the rotor consists of the ϵ- and c-subunits. The stator α3β3 and a-subunits are connected by the δ- and two b-subunits. The 10–12 c-subunits are believed to be arranged in a ring with subunit a on the side. Proton transport is mediated between the a- and c-subunits. The diagram indicates that the protons enter from the periplasmic space (right half-channel) and protonate the c-subunit, and the protonated c-subunit rotates counterclockwise until it meets the cytoplasmic-facing half-channel (left) and finally deprotonates.


Rotation of the catalytic β-subunit conformational states. For each 120° rotation of the γ-subunit (eccentric figure in center of each group of β-subunits), the catalytic site changes in conformation depending on which side of the γ-subunit faces it. Rotation of the γ-subunit counterclockwise drives ATP synthesis, and clockwise rotation is driven by ATP hydrolysis. βTP refers to the ATP binding form found in the bovine mitochondrial F1 X-ray crystallographic structure (1) and is assumed to be the tight nucleotide binding site carrying out reversible synthesis/hydrolysis. βDP refers to the ADP binding form and is the loose nucleotide binding site. βE refers to the empty binding form and is the site that utilizes coupled ΔμH+ energy to acquire conformational states that bind ADP and Pi on the way to synthesizing ATP (3, 4).


Conformational differences of two of the β-subunits imposed by the γ-subunit. The structures of the β-subunits (taken from Ref. 1) are shown in ribbon diagrams and the γ-subunit in a space-filling model. The heavy line points out the bound ADP in the βDP form. Regions of the γ-subunit that have specific interactions with the β-subunit are indicated by dark shading. The two thin lines indicate the region of interaction between γ- and β-subunits that are critical for communication of coupling information (see text).


This work was supported by National Institute of General Medical Sciences Grant R01–50957.


View Abstract