Phylogenetic distribution and evolutionary relationships of F- and V-ATPases

Sequence and structural comparisons of F- and V-type ATPases revealed homology between the subunits of the catalytic hexamer (α and B, β and A, respectively), membrane-embedded c/K-oligomers, and, perhaps, some subunits of the peripheral stalk (see Fig. 1), whereas the subunits of the central stalk have been found to be unrelated [8, 35–37]. Based on this distribution of homologous and non-homologous subunits in the structure of F- and V-type ATPases, we recently outlined a scenario according to which F/V-ATPases evolved from ATP-dependent RNA/protein translocases that contained the translocated polymer at the place of the central stalk and, initially, functioned within primordial membranes that were permeable for small ions, but not for RNA or protein [37].

Furthermore, the absence of homology between the subunits of the central stalk indicates a major dichotomy between the F-type ATPases, on the one hand, and V-type ATPases, both prokaryotic and eukaryotic, on the other hand. In contrast, all subunits of prokaryotic V-type ATPases (A-type ATPases, according to refs. [5, 40, 41]) have their counterparts in the vacuolar V-ATPases of eukaryotes. Eukaryotic V-ATPases are more complex than their archaeal and bacterial counterparts owing to the presence of several additional subunits [7, 9]. A similar phenomenon, however, is observed in F-ATPases, where the mitochondrial enzyme has several extra subunits as compared to the bacterial one [2, 44]. Sequence analysis also indicates that the primary evolutionary split lies between the F-type and V-type ATPases followed by secondary splits between the prokaryotic and eukaryotic enzymes in each of the branches: in phylogenetic trees, bacterial, archaeal and eukaryotic V-ATPase subunits invariably cluster together and separately from the F-type ATPases [35, 42, 45]. This is a remarkable, rather rare case where the evolutionary affinity between archaea and eukaryotes is seen outside the information processing systems. Thus, consideration of A-type ATPases as a separate class of rotary ATPases does not seem fully justified, not to mention that it obscures the use of the archaeal enzyme as a model for studying the eukaryotic V-ATPase.

Structural superposition of the Na⁺-binding sites of F- and V-ATPases

Proton-translocating and Na⁺-translocating forms are found among both F- and V-type ATPases. Notably, Na⁺-dependent ATPases are capable of translocating protons in the absence of Na⁺ [29, 34], whereas H⁺-dependent ATPases cannot translocate Na⁺ [46]. The apparent cause of this asymmetry is that sodium translocation is more structurally demanding: the coordination number of Na⁺ is 6, hence multiple, specific amino acid ligands are required to draw a sodium cation from water and to keep it in the non-polar inner space of the membrane.

The structures of the membrane-spanning, rotating oligomers of two Na⁺-translocating ATP synthases have been recently reported. The oligomer ring of the Ilyobacter tartaricus F-type ATP synthase consists of 11 c-subunits [47] whereas the ring of the Enterococcus hirae V-ATP synthase is composed of 10 K-subunits [38]. These structures show the same overall configuration of the Na⁺-binding sites (Fig. 2). All the four Na⁺-binding amino acid residues of I. tartaricus have readily identifiable counterparts in the E. hirae ATPase, and the two sets of ligands perfectly superimpose in space (Fig. 3). Even the position of a non-ligating tyrosine that stabilizes the principal, Na⁺-binding glutamate [34] is conserved. The K subunit of E. hirae ATPase contains an additional Na⁺-binding ligand, Gln65 [38], which is not shown on Fig. 3. The corresponding residue in the I. tartaricus c-subunit, Thr67, is much smaller, but, apparently can bind Na⁺ ion through a water molecule. Sequence alignment of c/K subunits (Additional File 1) shows that all (predicted) Na⁺-translocating ATPases contain, in this position, a polar residue, Ser, Thr, Gln, or Arg, that is capable of coordinating the Na⁺ ion either directly (Gln or Arg) or, in the case of Ser and Thr, via a water bridge [48]. In contrast, in all (predicted) H⁺-translocating ATPases, the corresponding position is occupied by an aliphatic residue incapable of making a hydrogen bond (Leu, Ile or Val). The sixth coordinating bond for the Na⁺ ion, in all likelihood, is provided by a water molecule. In enzymes of both types, this water molecule might occupy the pocket between the conserved Pro/Ser28, Gly/Ala29 (see Fig. 4), and the Na⁺ cation.

High-resolution crystal structures of H⁺-translocating c-oligomers are currently unavailable. However, sequence comparison indicates that they lack at least some residues that function as Na⁺ ligands except for the principal acidic (Glu or Asp) cation-binding residue (see Fig. 4 and Additional File 1).

The ancestral status of sodium bioenergetics inferred from phylogenetic analysis

The identical arrangement of the Na⁺-binding sites in the F-type and V-type Na⁺-translocating ATPases implies two mutually exclusive evolutionary scenarios:

1) the ancestor of the F/V ATPases was a H⁺-translocating enzyme, so the Na⁺-binding sites evolved independently in the two lineages via the recruitment of the same residues in the same positions, i.e., by convergence;

2) the common ancestor of the F/V ATPases was a Na⁺-translocating enzyme, but the ability to bind and translocate Na⁺ was independently lost in multiple lineages of F- and V-ATPases, yielding H⁺-translocating enzymes.

Convergence in the evolution of enzymes, manifest in the independent evolution of the same set of ligands conferring the same specificity in two lineages, is rare but not unheard of, as illustrated by the seminal study of Stewart et al. on the evolution of the langur monkey lysozyme that independently gained the same 7 ligands as the ungulate lysozymes during adaptation to a cellulolytic diet [49]. Thus, to distinguish between the two evolutionary scenarios, we constructed a phylogenetic tree of the catalytic subunits of the V- and F-ATPase and superimposed on it the known or predicted cation specificity, based on the available experimental data and presence or absence of the complete set of Na⁺-binding ligands (Fig. 5). The catalytic subunits were employed for this analysis because these are large proteins with numerous phylogenetically informative positions (see Additional File 2); by contrast, membrane-spanning c-subunits that harbor the cation-binding sites are short proteins that carry a weak phylogenetic signal, at best [45].

In the resulting tree, Na⁺-translocating ATPases did not form a clade within either the F or the V branch but instead comprised three distinct lineages among V-ATPases and at least three lineages among F-ATPases (Fig. 5). The monophyly of each clade of Na⁺-translocating ATPases was supported by moderate to high bootstrap values. In each of these clades, the same set of Na⁺ ligands was conserved (Fig. 4). The independent emergence of (nearly) identical ensembles of Na⁺ ligands in 6 distinct lineages of Na⁺-translocating ATPases can be, effectively, ruled out, leading to the conclusion that the common ancestor of V- and F-ATPases had a Na⁺-binding site. Unlike independent gain of the Na⁺ ligands in multiple lineages, multiple independent losses of subsets of these ligands appear plausible and could have led to the multiple transitions to proton translocation, provided that the membranes in the respective organisms became proton-tight.

The caveat beyond this conclusion is, of course, the uncertainty of the tree topology which remains a concern in any phylogenetic tree involving the deepest branchings in life's history. To address this concern, we performed statistical testing of alternative phylogenetic hypotheses using the approximately unbiased (AU) test. The AU test showed that the transition between Na⁺ and H⁺ translocation occurred at least twice, i.e., at least once in the V-lineage and at least once in the F-lineage: the topology with monophyletic Na⁺-translocating and H⁺-translocating ATPases was unequivocally rejected (Figure 6a). Moreover, the topology where the primary split in both the V-lineage and the F-lineage was between Na⁺-specific and H⁺-specific forms was rejected as well (Figure 6b). This result suggests that the switch, actually, occurred more than once in both V-type and F-type ATPases although topologies with a single switch in one of the principal branches could not be formally rejected (Figure 6c–h). The test failed to discriminate between the topologies where Na⁺-translocating or H⁺-translocating forms occupy the basal position in either the V-ATPase and F-ATPase clades (Fig. 6c,e,f,h). As discussed above, extraneous criteria, namely, the far greater likelihood of multiple losses of Na⁺ ligands compared to that of multiple, independent acquisitions of these ligands strongly point to the ancestral status of Na⁺ translocation.

Evolution of the c subunits of F/V-type ATPases by duplication and loss of Na⁺ligands

The c subunits of F-type and some prokaryotic V-type ATPases are formed by two transmembrane (TM) segments that fold into a single helical hairpin and are less than 100 amino acid residues long. Some of these proteins carry the complete set of Na⁺-ligands, e.g. in the c subunits of the F-ATPase of Thermotoga maritima and the V-ATPase of Nanoarchaeum equitans (see Additional File 1), whereas others have a partial set of Na⁺ ligands or even retain only the principal Glu/Asp cation-binding residue (e.g. F-ATPase from Escherichia coli or V-ATPase from Pyrobaculum aerophilum). Thus, H⁺-specific ATPases could have evolved from Na⁺-specific ATPases on multiple occasions and via somewhat different routes. The alignment of 2-helical c subunits shows patterns of loss/replacement of cation-binding residues that are generally consistent within bacterial and archaeal phyla, suggesting that these replacements occurred early in the evolution of the respective lineages (see Additional File 1). This finding corroborates the recent report of von Ballmoos and Dimroth on different modes of proton-binding by different F/V-type ATPases [34]. The pH-profiles of (i) the ATP hydrolysis and (ii) the modification of the principal carboxyl residue by dicyclohexylcarbodiimide (DCCD) were similar in the Na⁺-binding F-type ATPase (I. tartaricus and Propionigenium modestum) and H⁺-translocating V-type ATP synthase of Halobacterium salinarium. However, these profiles dramatically differed from the corresponding profiles of the F-type ATPases of E-coli, spinach chloroplasts, and bovine mitochondria. The revealed difference in the modes of proton-binding in the active site might reflect independent transition to H⁺-ATPase in different lineages.

In c/K subunits of other V-type ATPases, the 2-TM hairpins are duplicated, resulting in proteins with 4 TM segments (Fig. 2). Methanocaldococcus jannaschii and some other methanogens have a triplication of the 2TM hairpin, whereas Methanopyrus kandleri encodes a single protein of 1021 amino acid residues that encompasses 26 TM segments in 13 hairpins [50]. An early phylogenetic analysis of the two-hairpin c subunits has suggested that they evolved via several independent gene duplication and fusion events [45]. Assuming that the ancestral state was a Na⁺-translocating single-hairpin c subunit similar to the ones in T. maritima or N. equitans, it becomes possible to reconcile the history of duplications of the sequences encoding 2TM hairpins with the accompanying changes in cation specificity. Obviously, gene duplication per se could have occurred without any loss of the Na⁺-binding residues, as seen, e.g., in the proteins from Methanothermobacter thermautotrophicus or Methanobrevibacter smithii (see Additional File 1). In the M. kandleri protein, each of the 13 hairpins retains the complete set of Na⁺-binding residues such that one Na⁺ ion per hairpin can be bound (see Additional File 1). However, in other V-ATPases, the duplication of the 2TM hairpins is often accompanied by the loss of one or more Na⁺ ligands in one or both hairpins. In these cases, the hairpin duplication is not accompanied by the increase in the number of bound Na⁺ ions (as in E. hirae, see Fig. 2). While in single-hairpin c subunits, as discussed above, the loss of even one ligand is likely to hamper Na⁺-binding, in the double-hairpin but single Na⁺-ion-binding c/K subunits, replacement/reshuffling of Na⁺-binding residues appears to be possible, being, however, constrained so that residues from two different hairpins provide the full complement of ligands for a Na⁺ ion as seen in the E. hirae V-ATPase (Fig. 2).

Primordial sodium bioenergetics and evolution of membranes

The possibility that sodium bioenergetics antedated proton bioenergetics has been considered previously [28, 30, 51] but has not drawn much attention, and proton-dependent bioenergetics is widely believed to be ancestral [31, 34, 52]. This view is attractive considering the intrinsic chemical coupling between protonation/deprotonation events and redox reactions, in particular, those of water and diverse quinones [53–55]. However, proton energetics demands sophisticated, proton-tight membranes. In the scenario of the origin of F/V-ATPases from ATP-dependent RNA/protein translocases [37], with the translocated polymer initially taking the position of the central stalk, these primordial tranlocases were proposed to function within primitive, ancient membranes that were permeable for both H⁺ and Na⁺ but not for RNA or protein. Under this model, the evolution of F/V-ATPases was concomitant with the evolution of modern, complex, ion-tight membranes from primordial, semi-permeable membranes [37]. Conceivably, the ancient translocase employed sodium cations to crosslink and stabilize the subunits of the c-oligomer (as in I. tartaricus, see Fig. 3), preventing its destruction by the translocated polymer. Thus, even at the stage of the RNA/protein translocase, when the primordial membranes might have been leaky to both Na⁺ and H⁺, there could have been a mechanistic demand for Na⁺-binding and, accordingly, selection for the corresponding set of amino acid ligands. This scenario is supported by experiments demonstrating a dramatic decrease in the stability of the c-oligomers of Na⁺-translocating F-ATPases from I. tartaricus and P. modestum in the absence of Na⁺ [56].

When considering the possibility that primordial bioenergetics functioned on a sodium gradient, a biogeochemical reality check is important: was there enough sodium in the ancient ocean? Currently, there is a consensus that the primordial ocean formed owing to the condensation of the atmospheric water vapor upon cooling of the Earth [57]. Hence, the primordial ocean was fresh at the time of its emergence. However, it became salty relatively fast as can be judged from the chemical composition of geologically trapped water (reviewed in [57, 58]). In particular, quartz crystals found in iron oxide structures from the 3.5-3.2 Ga Barberton greenstone belt, South Africa, contain fluid inclusions with a Na/Cl ratio that is the same as in the present-day ocean (0.858), but the total amount of sodium and chlorine in these inclusions is 1.6 times the present day value [59]. Similarly, primary fluid inclusions from intra-pillow quartz from the North Pole Dresser Formation Pilbara craton, Western Australia (3.490 Ga), contain saline fluid with 12 wt% equivalent NaCl and a Cl/Br ratio of 631 [60]. Since this Cl/Br ratio is very close to that of modern seawater (Cl//Br = 647), it appears likely that this saline fluid is, indeed, archaeal seawater. Its higher salinity as compared to present seawater is believed to be due to the more intensive leaching of sodium (and other metals) from the crust by hot hydrothermal fluids [57, 58]. Only starting from Proterozoic terrains, the water inclusions show salinity close to that of modern seawater [61], perhaps, reflecting certain calming of the hydrothermal activity.

Thus, the above data indicate that, by the time of emergence of the first cells, sodium concentration in the ocean was comparable to or even exceeded that in extant oceans. Therefore, the next stage of evolution can be envisaged as selection for tighter membranes that would maintain the ionic homeostasis of the evolving cells, which would be particularly important given the increasing ocean salinity, and concomitantly, would create the opportunity for the utilization of ion gradients. As suggested by Skulachev [28], sodium-tight membranes could precede proton-tight membranes as structurally less demanding (see also below). In response to the increasing salinity of the primordial ocean, cells would need a mechanism for pumping Na⁺ ions out of the cell, driving the transition from a protein translocase to the precursor of an ion-translocating membrane ATPase, as described elsewhere [37]. These ancestral ATPases would pump Na⁺ along with the Na⁺-transporting pyrophosphatase [62] and chemically-driven Na⁺-pumps, such as Na⁺-transporting decarboxylase [29, 63], which, being found in both bacteria and archaea, appear to antedate the divergence of the three domains of life. Unlike the other Na⁺ pumps, the common ancestor of the V/F-ATPases, by virtue of its rotating scaffold, would be able to translocate Na⁺ ions in both directions, so that, upon further increase in the external salinity, reversal of the rotation could result in Na⁺-driven synthesis of ATP by this primordial rotary machine. This would be the beginning of membrane bioenergetics: together with the ancient Na⁺ pumps, the V/F-type ATP synthase would complete the first, sodium-dependent bioenergetic cycle in a cell membrane (Figure 7).

The subsequent transition to the H⁺-dependent bioenergetics required proton-tight membranes. The conductivity of lipid bilayers for protons is by 5–7 orders of magnitude higher than the conductivity for Na⁺ and other small cations [64–66] and dramatically increases with temperature [32]. This difference stems from the unique mechanism of transmembrane proton translocation. Whereas transfer of other cations depends on their ability to penetrate the phospholipid membrane, crossing of the membrane/water interface by protons is not rate-limiting [64–67]: protons, being confined to the membrane ionizable groups [68], can apparently enter transient water clusters nested between the lipid hydrocarbon chains. The rate of proton transfer across the membrane is limited by proton "hopping" across the middle of the bilayer, with the protons transferred from one side of the membrane to the other via collisions between a protonated water cluster in one monolayer and a cluster in the opposite monolayer [65, 66]. The rate of proton translocation, while independent of pH and insensitive to the H₂O/D₂O substitution [64–67], is determined by the frequency of these collisions which explains the dependence of the proton transfer rate on the overall bilayer dynamics and on temperature. Proton leakage can be suppressed by decreasing lipid mobility and/or increasing hydrocarbon density in the midplane of the bilayer, which can be achieved by either branching the ends of the lipid chains or incorporating hydrocarbons with a selective affinity to the cleavage plane of the bilayer [66].

In agreement with the hypothesis of independent evolution of proton energetics in multiple lineages, proton tightness of membranes appears to be achieved by radically different means in different organisms, namely, the mobility of side chains is restricted in distinct ways and diverse hydrocarbons are packed in the midplane of the H⁺-tight membranes [32, 66]. Specifically, the fatty acid chains in the membrane of many bacteria have branched termini (see [66] and references therein). In some thermoacidophiles, the lipid chains terminate with alicyclics (cyclohexane and/or cycloheptane), resulting in additional crowding of the midplane of the bilayer [69]. In archaea, lipids consist of two phytanyl (isoprenoid) chains which are linked via a ether bond to glycerol or other alcohols such as nonitol [70]. Extreme thermophilic and acidophilic archaea possess membrane-spanning lipids in which the phytanyl chains of two diether lipids are fused to a C40 core [32, 70]. Both the chain fusion and the involvement of a rigid ether bond restrict the hydrocarbon chain mobility. In addition, different organisms pack different hydrocarbons in the midplane of their H⁺-tight membranes, namely, phytosterols in plasma membranes of plants and protozoa, and hopanoids and squalane in bacteria (see [66] and references therein). In alkaliphilic bacteria, where H⁺-tightness is crucial, the squalane content is elevated as compared to mesophiles [71]. In eukaryotic mitochondria and chloroplasts, the middle of the bilayer is occupied by the ubiquinone and plastoquinone tails [72]. Remarkably, plasma membranes of animal cells, as emphasized in Fig. 4, have remained "sodium membranes" [28] inasmuch as they cannot maintain H⁺ potential.

Taken together, the abundance of Na⁺-dependent thermophiles among deeply branching archaea and bacteria [30, 32, 73], the existence of unrelated mechanisms that make membranes H⁺-tight in different lineages [66], the discovery of opportunistic, chemically driven Na⁺-translocating enzymes in several deeply branching organisms [29, 62, 63, 74], and the prevalence of the Na⁺-coupled membrane transport systems over H⁺-coupled ones in extremophiles [32] appear to provide strong support for the "Na⁺-translocating ATPase first" scenario. This scenario involves a gradual transition from the primordial membranes that were leaky to both Na⁺ and H⁺ to the currently predominant, more complex H⁺-tight membranes via Na⁺-tight, but H⁺-permeable membranes, which are still employed in animal cells, as well as, perhaps, in some bacteria and archaea with sodium energetics (Figure 7).

Considering the ubiquity of the F/V-ATPases in cellular life, their common ancestor is believed to have been present in the last universal common ancestor (LUCA) of cellular life forms [75]. Here we argue that this enzyme possessed a Na⁺-binding site. Thus, the LUCA either had leaky membranes such that the common ancestor of the F/V ATPases operated as polymer translocase with Na⁺ ions performing a structural role, or the membrane of the LUCA was tight to sodium but permeable to protons and, accordingly, the LUCA would have sodium energetics (Figure 7). It seems highly unlikely that the LUCA had modern, proton-tight membranes: as discussed above, such membranes, probably, evolved later, independently in different lineages, a feature that might underline the fundamental difference in the lipid structure between bacteria and archaea [76].

As noted above, proton energetics has a major advantage over sodium energetics because proton transfer can be chemically coupled to redox reactions, thus, enabling the advent of efficient redox-driven generators of proton potential, such as cytochrome bc₁ complex [55], cytochrome oxidase [54], or the oxygen-evolving photosystem II [77, 78]. Once membranes capable of maintaining proton gradient have evolved, separately in bacteria and archaea, the sodium-binding sites of the F/V ATPases became dispensable and deteriorated independently in multiple lineages.

Reviewer's report 1

Reviewer's report 2

Reviewer's report 3