ABSTRACT
My twenty-five year fascination with membrane ATPases grew out of my experiences in the laboratories of André Jagendorf and Efraim Racker. André introduced me to photosynthetic phosphorylation and Ef, to whose memory this article is dedicated, convinced me that ATPases had much to do with ATP synthesis.
Introduction
My twenty-five year fascination with membrane ATPases grew out of my experiences in the laboratories of André Jagendorf and Efraim Racker. André introduced me to photosynthetic phosphorylation and Ef, to whose memory this article is dedicated, convinced me that ATPases had much to do with ATP synthesis.
Astounding progress has been made in the H+-ATPase field in just two decades. By the early 1970s, it was generally recognized that oxidative and photosynthetic ATP synthesis were catalyzed by membrane enzymes that could act as H+-ATPases and that the common intermediate between electron transport and phosphorylation is the electrochemical proton gradient. At that time, it had been shown that a cation-stimulated ATPase activity was associated with plasma membrane preparations from plant roots. The endomembrane or vacuolar ATPases were unknown.
The application of improved biochemical methods for membrane isolation and purification, as well as membrane protein reconstitutions, led rapidly to the conclusion that there are three major classes of membrane H+-ATPases, P, V and F. P-ATPases, which will not be considered further in this article, are phosphorylated during their catalytic cycle and have a much simpler polypeptide composition than V-or F-ATPases. The plasma membrane H+-ATPase of plant, yeasts and fungal cells is one example of this class of enzymes (see Pedersen and Carafoli, 1987, for a comparison of plasma membrane ATPases).
Biochemical and gene sequencing analysis have revealed that V-and F-ATPases resemble each other structurally, but are distinct in function and origin. The ‘V’ stands for vacuolar and the ‘F’ for F1FO. Fi was the first factor isolated from bovine heart mitochondria shown to be required for oxidative phosphorylation. Fo was so named because it is a factor that conferred oligomycin sensitivity to soluble Fi. Other F-ATPases are often named to indicate their sources. For example, chloroplast F1 is denoted CF1 (see Racker, 1965, for early work on F1). Recent successes in reconstitution of vacuolar ATPase have led to a V1 Vo nomenclature for this enzyme as well.
The term ‘ATP synthase’ is now in general use to describe F-ATPases. This term emphasizes the facts that although F-ATPases function to synthesize ATP, they do not catalyze, normally, ATP hydrolysis linked to proton flux. In contrast, V-ATPases are very unlikely to operate as ATP synthases. Thus, F-ATPases are proton gradient consumers, whereas V-ATPases generate proton gradients at the expense of hydrolysis.
In this brief review, 1 will compare the structures of F-and V-ATPases. Also, I give some insight into the mechanisms that help prevent wasteful ATP hydrolysis by the chloroplast ATP synthase (CF1F0).
Structure of F-and V-ATPases
Both F-and V-ATPases are structurally complex. CF1F0, for example, contains nine different proteins and a total of about twenty polypeptide chains (Jagendorf et al. 1991). The Mr of V-and F-ATPases is approximately 550000 –650000. The complexity of these enzymes is puzzling. This is especially true in view of the fact that P-ATPases can be as simple as a single polypeptide chain of about 100000Mr-Genetic studies with Escherichia coli have established that each of the eight E. coli F1F0 polypeptides is required for function (Futai et al. 1989).
F-ATPases may be readily separated into two parts – F1, the catalytic sector and F0, the proton translocating part. F1 can be removed from many coupling membranes either by physical means or simply by diluting membranes into media of low ionic strength that contain EDTA. Fi is, thus, an extrinsic membrane protein. Once removed from the membrane, F1 is soluble to at least 100 mg ml−1. From 2 kg of spinach leaves, 400 mg of CF1, greater than 95 % pure, may be prepared in a day and a half. This ease of preparation makes one less reluctant to engage in calorimetric or equilibrium nucleotide binding studies that can require as much as 100 mg of CF1 for each experiment. Depending on the source, the Mr of F, ranges from 365 000 to 400000.
From all sources, mitochondrial, chloroplast and eubacterial plasma membranes, the subunit composition of F1is similar. There are five different proteins, labelled α-ϵ in order of decreasing Mr. There is a consensus that these polypeptides are present in a 3:3:1:1:1 stoichiometry (Table 1). The primary sequence of a large number of F1 polypeptides has either been deduced from gene sequences or determined chemically. From sequence comparisons, three striking facts were revealed. First, the sequence of the βsubunit has been remarkably conserved, even across Kingdom lines (Hudson and Mason, 1988; Walker and Cozens, 1986). For example, the β subunit of CF1 and E. coli F1 have 76% sequence identity. Second, the α and β subunits are related and it seems quite possible that α arose by gene duplication. Third, the smaller subunits, γ, δand ϵ, are not well conserved. In fact, ϵ of mitochondrial F1 does not resemble CF1 ϵ or E. coli F, e at all and mitochondrial F1 δ is also unique (Walker and Cozens, 1986). The bacterial and chloroplast δ subunits resemble oligomycin sensitivity conferral protein (OSCP) which is considered to be an Fo subunit. Thus, generalizations about subunit function based simply on Mr can be problematic.
From image reconstruction electron microscopy (Boekema et al. 1988), distance mapping by fluorescence energy transfer (McCarty and Hammes, 1987) and X-ray diffraction (Bianchet et al. 1991), the structure of F1 is emerging (Fig. 1). The structure is dominated by the α/βhexamer. The α subunits are in contact with the β subunits, but the α subunits are closer to the membrane surface. The position of the γ subunit is not exactly known, but it is likely that it is centrally located in an asymmetrical manner. The e subunit interacts strongly with γ, and δis placed close to the membrane surface because of its role in F1 binding to Fo.
The inherent structural asymmetry of F1 is intriguing. The α and β subunits cannot be in equivalent environments. That is, the interactions of a given α or β subunit in the complex with the smaller subunits must differ from those of another α and β subunit. There is evidence to suggest that CF1 lacking the δ and ϵ subunits retains its structural asymmetry (Shapiro et al. 1991). Thus, interactions of the y subunit with α and β may be the major asymmetry inducers.
The structural asymmetry of Fi probably explains the nucleotide binding site asymmetry as well as the unusual reactivity of Lys-378 of the a subunit of CF1. Only one of the three a subunits of CF1 reacts at this position rapidly with Lucifer Yellow vinyl sulfone. The six nucleotide binding sites of Fi show quite different properties. The most plausible model for the mechanism of Fi is one in which two or three catalytic sites change their properties during the catalytic cycle (Boyer, 1989). In this alternating site or binding change mechanism, catalytic sites alternate between very high affinity and much lower affinity. The binding of substrate to one catalytic site promotes catalysis and product release from another. If this mechanism holds, and there is much evidence in its favor, F1 cannot be permanently frozen in one asymmetrical state. The incubation of CF1 with Mg2+-ATP was found to cause two nucleotide binding sites on CF1 to switch their properties (Shapiro and McCarty, 1990). This observation is consistent with the binding change mechanism. How substrate binding affects changes in the enzyme is unknown.
Although all five E. coli F1 or CF1 subunits are required for ATP synthesis, α3β3γ complexes exhibit the highest ATPase activity. An intact γ subunit, however, is not required for ATPase activity and some ATPase activity is seen in α/β complexes (see, for example, Kagawa et al. 1989). Although it probably bears the catalytic sites, isolated β is a very poor ATPase. In CF1, γ definitely plays a regulatory role and has been proposed to be involved in proton translocation. The δ subunit is required for the binding of E. coli F1to Fo and is necessary for the functional binding of CF1 to CFO. In E. coli and chloroplast F1, 6 is an inhibitory subunit, probably involved in regulation of the enzymes. These studies are summarized by Futai et al. (1989) and Jagendorf et al. (1991).
Different coupling membranes seem to have tailored the Fo portion of the ATP synthase to suit their needs. By definition, Fo is that part of the ATP synthase that remains after removal of Fi. Unlike Fi, Fo is hydrophobic and can only be isolated from membranes by detergent extraction. E. coli Fo (Deckers-Hebestreit and Altendorf, 1992) and chloroplast Fo have been purified in active forms. Fo functions in at least two ways: it binds Fi and translocates protons across the membrane. The rates of proton conductance through CFo are high, suggesting that the Fo proton translocation mechanism is a proton-selective channel (Lili and Junge, 1989; Junge et al. 1992).
The number of different polypeptides in Fo varies with the source (Table 1). E. coli Fo is the simplest, with just three polypeptides, labeled a, b and c. Spinach CF0 has four polypeptides, denoted in Roman numerals, I-IV. Mitochondrial Fo can have as many as eight polypeptides. All Fo subunits contain a small (8000Mr) hydrophobic polypeptide, referred to as the proteolipid or DCCD-binding protein. There is no lipid covalently attached to the DCCD-binding protein. This polypeptide is present in 6–12 copies per ATP synthase. Remarkably, the reaction of just one copy of the protein with DCCD (N,N′-dicyclohexylcarbodiimide) at a specific glutamyl or aspartyl residue is sufficient to inhibit ATP synthesis totally. ATP synthesis is inhibited by DCCD because proton transport through Fo is blocked. All three E. coli Fo subunits are required for proton translocation and at least subunits III and IV of chloroplast Fo are necessary for proton transport.
Cross-linking studies indicate that several CFO subunits interact with CF1 Some CF0 subunits have hydrophilic domains that are predicted to extend out from the stromal side of the membrane; that is, towards CF1. Subunit HI of CF0 appears to bind to CF1 with sufficient strength to allow the purification of a CF1-subunit HI complex by a chromatographic procedure (C. Wetzel and R. McCarty, unpublished observations). This complex retains some of the properties characteristic of the CF1F0 from which it was prepared. Probably, all CF0 subunits are in contact with one or more CF1 polypeptides. The dominant interactions have yet to be determined.
Like F-ATPases, V-ATPases consist of an extrinsic component, called V1, and a membrane-associated part, now denoted Vo. The treatment of yeast, Neurospora crassa, or oat root vacuolar membranes and clathrin-coated vesicles with KI or KNO3 in the presence of Mg2+-ATP rapidly inactivates ATPase activity and associated proton pumping. Depending on the source, about five polypeptides are dissociated from the membrane by this treatment (see, for example, Ward et al. 1992). The dissociated polypeptides were devoid of ATPase activity by themselves, but, in the case of the clathrin-coated vesicles (Puopolo et al. 1992) and oat root systems (Ward et al. 1992), could be added back to the depleted membranes to give both ATPase activity and ATP-dependent proton pumping. It is of interest to note that membranes from which the V1 components had been removed were not leaky to protons. Thus, either Vo required a V1 polypeptide(s) for proton transport or the attachment of V1 helps to open a proton gating mechanism.
To an F1 person, the subunit composition of V-ATPases is somewhat confusing. Then again, to a V1 person, I would venture to guess that the composition of F-ATPases would also be confusing. As is the case for F-ATPases, V-ATPases seem to retain a similar subunit structure for V1, but tailor the membrane-associated components to fit the needs of the membrane (Nelson, 1989). The situation is made even more complicated by the possibility that a given tissue within an organism could possess a V-ATPase that differs from that in another tissue.
Regardless of the apparent heterogeneity of the subunit composition of V-ATPases, there are, as is the case for F-ATPases, recurrent themes. V-ATPases are oligomeric and contain both a peripheral, catalytic component Vi and an integral membrane part, called Vo by analogy to Fo (Table 2). All V-ATPases so far examined from animal, plant, fungal and yeast sources contain two easily discernable polypeptides of about 70000 (the A subunit) and 60000Mr (the B subunit). Based on covalent binding of -SH alkylating reagents and photoaffinity nucleotide binding, it is very likely that the A subunit is catalytic. The function of the B subunit is unknown. The A subunits of various V-ATPases share remarkable amino acid sequence similarity. Remarkably, the catalytic subunits of ATP synthases of archaebacterial ATPases are more closely related to V-ATPases than to F-ATPases. Nonetheless, the A subunits of both archaebacterial and V-ATPases share significant sequence homology with the β3 subunits of Fi (22–26 %) (Ihara et al. 1992). Moreover, the B subunits of V-ATPases are homologous with the α subunits of F1 (20–27%). See Nelson (1992) for a more complete description of the sequence homologies.
In addition to subunits A and B, V-ATPases from a number of membranes contain polypeptides ranging in relative molecular mass from 100000–115 000 to 12000–13000. Most V-ATPases contain polypeptides (four or five) in the 30000–45 000 Mr range. All so far contain a 16000 Mr protein that has the solubility properties of a ‘proteolipid’ and binds DCCD. As is the case for F-ATPases, DCCD inhibits V-ATPase activity and associated proton pumping. It is quite possible that the V-ATPase DCCD-binding protein arose by gene duplication. Some V-ATPases appear to contain a high Mr (100 000–115 000) polypeptide and others a low Mr ( 12 000-13 000 Mr) polypeptide.
V-ATPases are sensitive to inhibition by chaotrophic anions (I−, NO3− or SCN−) in a manner that is fascinatingly dependent on the presence of Mg2+-ATP) (Moriyama and Nelson, 1989). A number of polypeptides dissociate from the membrane as a result of this treatment; others are left behind. Concomitant with loss of ATPase activity and associated protein pumping, more or less spherical particles of 10-12 nm are removed from vacuolar membranes (Bowman et al. 1992).
The peripheral portion (V1) is defined as that part removed by chaotroph treatment in the presence of Mg2+-ATP. The A and B subunits, probably present in a 3:3 stoichiometry with respect to the smaller polypeptides, are clearly in this class. In addition, the polypeptides in the 30000–45 000Mr range are removed by this treatment. Recently yeast subunit C (Mr 42000) was cloned and sequenced (Beltran et al. 1992). Subunit C was shown to be required for assembly and exhibited no homology to the γ subunit of F1. What is left after chaotroph treatment is defined as Vo. There is a consensus that the 16 000Mr DCCD-binding protein, which is probably present in multiple copies, is part of Vo. The approximately 100000Mr polypeptide of some V-ATPases as well as smaller proteins (12 000–20000Mr) may also be part of Vo. Further analysis is required and will undoubtedly be forthcoming.
There are several observations that point to major differences between V-and F-ATPases, even though superficially they seem similar. Removal of F1 greatly increases the proton permeability of the membrane in which it resides. This effect is not observed for V-ATPases. Perhaps some V1 component remains associated with Vo after chaotroph treatment and blocks a proton conductance mechanism. Alternatively, a V1 component is required for proton transport through Vo. So far V1 has not been released from the membrane in a form that is active in ATP hydrolysis. This result is in distinct contrast to F1F0. Finally, Mg2+-ATP destabilizes V1V0, whereas it stabilizes F1 against coldinactivation in the presence of chaotrophic anions.
Regulation of CF1F0
Thylakoid membranes, as usually isolated, hydrolyze Mg2+-ATP in the dark, even at 37 °C, at rates close to 0.05μmol min−1 mg−1 protein. Much of this activity may be attributed to phosphohydrolases that contaminate the preparations. In contrast, the same membranes, when illuminated, can catalyze ATP synthesis at rates in excess of 5μmol min−1 mg−1 protein. It is clear, then, that light dramatically activates the chloroplast ATP synthase. The ATP synthases have evolved to be specialists in ATP synthesis and, through a number of different mechanisms, potentially wasteful ATP hydrolysis by CF1F0 is prevented.
It must be emphasized that the effect of light is indirect. Transthylakoid electrochemical proton gradients drive ATP synthesis in total darkness and, thus, are also involved in the conversion of the ATP synthase to an active form. Light-driven electron flow generates proton gradients that serve two functions in photophosphorylation. First, the proton gradient is the driving force for ATP synthesis and, second, it is required for activation.
How this unusual ‘one way’ regulation occurs at the molecular level is beginning to be revealed. A coherent picture, summarized in Fig. 2, of what may happen in a dark-to-light transition can now be drawn. The e subunit is an inhibitor of the ATPase activity of CF1.
By necessity, it must also inhibit ATP synthesis. The e and γ subunits interact and it is probable that the inhibitory effects of e are mediated at least in part through the γ subunit. In the dark, e interacts very strongly with γand the ATPase activity is inhibited. Generation of the electrochemical proton gradient causes changes in CF1 that involve both the γ and ϵ subunits (at least). A movement of e relative to γ is indicated. In their new conformations, e inhibition could be nullified.
Redox regulation is also involved in the physiologically significant light activation of CF1 in chloroplasts. The γ subunit of CF1 from photosynthetic eukaryotes contains an insert of about twenty amino acids that is absent in E. coli or mitochondrial F1 subunits (Jagendorf et al. 1991). This insert contains a redox-active pair of Cys residues (Cys-199 and Cys-205 for spinach CF1γ subunit). In the dark, Cys-199 and Cys-205 form a disulfide bond, whereas in the light this disulfide (the only disulfide in CF1) is reduced. Thioredoxin, reduced by electrons from photosystem I, is probably the physiological reductant.
Although oxidized CF1 is active in ATP synthesis, reduced CF1 is even more active. At physiological values of the electrochemical proton gradient, phosphorylation could be enhanced as much as tenfold (Quick and Mills, 1986). Essentially, it appears that the energy cost for activation is significantly decreased by reduction of the y disulfide bond.
The activated state of oxidized CF1 in thylakoids decays quite quickly in the dark. Thus, after a period of illumination, little ATPase activity is observed in the dark. After reduction of the y disulfide bond, however, significant ATPase activity persists in the dark after a period of illumination (Bakker-Grunwald, 1977). This ATPase activity is coupled to inward proton fluxes and can generate ΔpH values of sufficient magnitude to permit ATP synthesis at a relatively low rate. Proton ionophores at concentrations that collapse the proton gradient completely abolish ATPase activity in the dark. The ΔpH generated by ATP hydrolysis in the dark by CF1 in reduced thylakoids is, thus, sufficient to keep the enzyme, at least in part, in an active form. During illumination, ATP hydrolysis is prevented by the high value of ΔpH which drives the reaction in the direction of synthesis (Davenport and McCarty, 1986).
In intact chloroplasts, the ability of reduced CF1 to hydrolyze ATP in the dark after a period of illumination is lost over a period of several minutes in a biphasic manner. The more rapid phase is probably a consequence of the binding of ADP to a nucleotide binding site on CF1. ADP at micromolar concentrations rapidly inactivates the ATPase of reduced CF1 in the dark, without causing oxidation of the y subunit, dithiol. To reconvert the CF1 to an active form after ADP has bound, ΔpH is required and ADP is released from the enzyme. The second, slower phase of the inactivation of CF|FO in intact chloroplasts in the dark is the oxidation of the dithiol by an unknown mechanism (Biaudet et al. 1988).
There is good evidence that the e and y polypeptides of CF1 interact. They are physically close together (McCarty and Hammes, 1987) and alterations of the y subunit modify e-CF1 interactions (Soteropoulos et al. 1992). Reduction of the y disulfide decreases the affinity of e binding to CF1 about 20-fold and tryptic cleavage of a portion of the C terminus of y abolishes high-affinity e binding. Removal of the e subunit markedly enhances the rate of reduction of γ disulfide by dithiothreitol or thioredoxin (Dann and McCarty, 1992). The y subunit in CF1 depleted in e is easily attacked by proteases, with the initial major cut occurring close to the y disulfide to produce a γ fragment of about 27000Mr. If the γ disulfide is reduced, a second cut occurs that releases a peptide of Mr 1300. This peptide bears Cys-205.
The y subunit of CF1 in thylakoids in darkness is resistant to proteolysis. With the increase of ΔpH in the light, however, changes in the enzyme occur that render y very susceptible to partial proteolysis. The y subunit is cleaved to a fragment of 27 000 Mr and, if the disulfide is reduced, the same second cut that occurs in CF1 deficient in e is observed. Thus, with respect to the susceptibility of the γ subunit to protease attack, CF1 in illuminated thylakoids resembles CF1 deficient in e.
Mitochondrial F1F0 is regulated by a quite different mechanism from that of CF1F0. The 6 subunit of mitochondrial F| has no sequence similarity to that of CF1 or E. coli F1. Mitochondrial e does not appear to be an inhibitor of the ATPase activity of mitochondrial F1. Instead, mitochondrial Fi activity is regulated by an inhibitor protein (or proteins) that, unlike e, is not considered to be part of Fi (reviewed by Cross, 1981). The binding of the inhibitor protein is favored by ATP. In contrast, the inhibitor protein is released from the complex when the magnitude of the electrochemical proton gradient and the ADP/ATP ratio are high. Thus, ATP synthesis would be favored.
There is no evidence for an inhibitor protein other than e in either E. coli or chloroplasts. Moreover, it is very clear that e dissociation from CF1 cannot be a part of the activation process. CF1deficient in e binds to CF0 equally as well as CF1, but e-depleted CF1 cannot restore ATP synthesis (Richter et al. 1984). When CF1 is removed from thylakoid membranes the membranes become highly proton permeable because of proton leakage through CF0. The permeability is so high that even extremely rapid proton translocation by light-dependent electron transport does not generate a significant ΔpH. CF1 deficient in e fails to block the CF0 proton channel and, thus, cannot restore ATP synthesis to CF1-depleted thylakoids. If ϵ were to dissociate from CF1 during activation, therefore, ATP synthesis would be inhibited.
The e subunit is also an inhibitor of the ATPase activity of E. coli F1 and has been shown to interact strongly with γ (Dunn, 1982). The E. coli ATP synthase can operate as an H+-ATPase in vivo. For example, when respiratory chain activity is low, ATP hydrolysis by ATP synthase powers proton efflux to generate an electrochemical proton gradient. It is likely that the activity of the enzyme is regulated since the ATPase activity of the E. coli ATP synthase in its natural environment is much lower than the Vmax. An energy-dependent activation mechanism which involves the overcoming of e inhibition may occur in E. coli F1FO as well as in other ATP synthases.
The dual activities of the e subunit of CF1 (regulation and proton channel blocking) suggest an intriguing connection between activation and opening of a ‘proton gate’. The proton conductivity through CFiF0 is normally low when the enzyme is inactive, but is high (greater than 1 ms−1) during ATP synthesis. Could it be that activation entails - at least in part - an opening of the proton gate? In which part of the ATP synthase protons induce the conformational changes involved in activation and proton gate opening is unknown. Protonation of a CF0 subunit(s) in contact with CF1 could induce the changes. Alternatively, CF0 could act passively to deliver protons to a site on one or more of the CF1 subunits.
To date, little is known about regulation of V-ATPases. By analogy to the F|F0-ATPases, I predict that one or more of the smaller polypeptides of F1 will be a regulatory subunit. As is the case for mitochondrial F1, a dissociable inhibitory protein could be involved.