Early bioenergetic evolution

Our focus here is not the primordial synthesis of bases, nor how bases or nucleosides might react with one another, nor what an RNA world might have looked similar in detail, nor the origin of the genetic code. We are interested in the chemical environment and the energetic processes that might have made those important evolutionary steps possible, as exergonic reactions or, more likely, as side products of a central exergonic reaction involving H2–CO2 disequilibrium (similar to the case in modern microbial metabolism). Three points can be noted.

First, DNA is synthesized from RNA in metabolism, which has long served as an argument that RNA came before DNA in evolution. By the same reasoning, we infer that amino acids came before RNA in evolution, because in metabolism RNA bases are synthesized from amino acids: glycine, aspartate, glutamine, CO2 and some C1 units in the case of purines; aspartate, ammonia and CO2 via carbamoyl phosphate in the case of pyrimidines (figure 2c). Amino acids are thermodynamically favoured over nucleotides and thus more likely to accumulate in larger amounts in a hydrothermal setting anyway [35,36]; similarly in metabolism, some might spill over into RNA-like bases. If an RNA world-like system arose in the same environment where the acetyl-CoA pathway arose—a logical consequence of the idea that the acetyl-CoA pathway is the most ancient pathway of carbon fixation [73]—then the RNA world arose in a chemical environment that was very rich in CO2, reactive C1 moieties and reactive methyl groups. Two of the five carbon atoms in the purine backbone come from reactive C1 moieties: 10-formyl-H4F or formyl phosphate [87], and one of the five comes from CO2, while one of the four carbon atoms of the pyrimidine backbone comes from CO2 via carbamoylphosphate synthase reaction (figure 2c).

This might be a biosynthetic fossil, an imprint of the chemical environment in which the genes arose that catalysed the first, genetically specified, enzymatic base biosynthesis. That is, the substrates of the enzymes that catalyse base synthesis were probably in existence before the origin of those genes, and the four carbon atoms that stem from CO2 or formyl moieties in all modern purine and pyrimidine bases (figure 2c) might be relics of life's autotrophic origin. There are limits as to how far one can go with such reasoning, but it is something to consider.

Second, the RNA world as it is currently construed in the laboratory, operates with pure mixtures of usually four bases [98]. But, in the beginning there is just no way that base synthesis could have been highly specific. Any RNA world that might have really existed must have consisted of an ill-defined mixture of bases synthesized spontaneously without the help of either genes or pure chemicals. The chemical environment central to our considerations here is replete with reactive C1 moieties and methyl groups (on their way to becoming methane, acetyl groups and acetate); modern ribosome–tRNA interactions at the heart of the genetic code require myriad base modifications in tRNA and rRNA [99], the vast majority of which are methylations [100]. This possible significance of tRNA modifications [101] becomes all the more evident when those common to all prokaryotes that allow the code to function at the wobble base [102] are considered. Or, we can just look at the nature of the chemical modifications that are present in the 28 modified bases common to archaebacteria and eubacteria [100]. They include 19 methyl-, one acetyl-, five thio-, two methylthio-, one seleno-, two methylaminomethyl- and four carbamoyl-moieties [103]. Those modifications might be a chemical imprint—a molecular fossil—of the environment in which tRNA–rRNA interactions (genes and proteins) arose that is preserved in the chemical structure of modified tRNA and rRNA bases. Assuming that any RNA world-like reaction nexus ever really existed in early evolution (a reasonable premise), it would by necessity have consisted of spontaneously and unspecifically synthesized mixtures of bases.

The substantial number of enzymes that life forms have carried around for almost 4 billion years to perform these cumbersome and specific modifications, suggests that the modifications are essential, otherwise they would have been discarded. That those enzymes recreate the ancestral state of the RNA world is something to consider.

Third and finally, there is a rather severe and very general problem for all theories on the origin of life. Namely, before there was any kind of genetic feedback via replicating molecules, whether self-replicating or replicated by auxiliary catalysts, we currently have no option but to accept the premise that some replicating chemical entity, catalytic and able to specify synthesis of self, arose spontaneously, which is theoretically possible in the context of autocatalytic networks [104–106]. Morowitz has observed that the chemistry of life is deterministic in many ways, but that does not mean that its emergence is inevitable, because a number of contingent factors, such as the nature of catalysts present, are involved

The origin of life is a deterministic event, the result of the operation of the laws of nature of a physical chemical system of a certain type. This system evolves in time, is governed by physical principles and eventually gives rise to living forms. The details need not be totally deterministic in every respect, but the overall behaviour follows in a particular way. [94, p. 3]

In a world of gene and proteins, evolution via variation and selection can readily supplant chemical determinism. But until we get to things that evolve via variation and selection, all we have are spontaneous reactions under thermodynamic and kinetic control. In this way, thermodynamics is very much chemistry's deterministic version of natural selection [107]. Getting from rocks and water to genuinely evolving systems remains a big problem. But energetics and thermodynamics can help a lot, because they place rather narrow contraints on which paths chemistry can traverse en route in evolving systems that ultimately spawn free-living cells.

That bottom-up perspective has a top-down corollary [108]. If we search among modern microbes and their metabolism for ancient physiological phenotypes, the acetogens and methanogens stand out [44,109]. These obtain their energy from making organic compounds out of CO2. In fact, they are often diazotrophic and as such they just live from gases: CO2, CO, H2, N2 and H2S. At the level of overall chemical similarity, core carbon and energy metabolism in methanogens and acetogens has quite a lot in common with geochemical processes at alkaline hydrothermal vents [43], and that is probably not coincidence.

Again, we have little to contribute to the nature of an RNA-like world [98], other than the insights that it probably arose in the same environment where the acetyl-CoA pathway did, a setting full of amino acids, reactive C1 moieties and CO2, and that molecular fossils of that environment might be preserved in RNA base modifications today. Nor do we have anything to contribute to the origin of the genetic code and translation, other than agreeing with our colleagues that it is a very significant and difficult problem [110–112], that it did occur, that the code increased in complexity during evolution [113,114], that the simplicity of the code in methanogens is intriguing [115] and that the origin of the code could have occurred at alkaline hydrothermal vents [116], supported by a continuous flux of harnessable carbon and harnessable energy.

It should be explicitly stated that, for the purposes of this paper, we are assuming that the origin of genes and proteins could have been fuelled with acyl phosphates (acetyl phosphate) as the central currency of ‘high-energy bonds’, that these were generated by substrate-level phosphorylation via reactions as outlined earlier (figure 2b), and that this requires a steady input of spontaneously synthesized methyl groups of geochemical origin [109], either from serpentinization or from redox processes at the vent–ocean interface [43]. The standard free energy of hydrolysis (ΔG°′) of acetyl phosphate is –43 kJ mol−1, meaning it can phosphorylate ADP to ATP (ΔG°′ –31 kJ mol−1), and by the same token it can drive phosphorylations in intermediary metabolism as well as ATP does. However, the actual ΔG°′ provided by a reaction depends on the level of disequilibrium between rates of formation and breakdown—specifically the steady-state ratio of acetyl thioesters to acetyl phosphate to acetate, which is ultimately driven far from equilibrium by the chemical disequilibrium generated by serpentinization.

Once increasing chemical complexity transforms into conventional evolution of genetically encoded proteins, the invention of new functions potentially becomes a very fast process. The jump we just took, to genes and proteins, is a big one, but all models for the origin of life entail such a step, and ours does too. Recalling that its mechanisms are, as stated earlier, not our focus, we proceed. At some early point, the advent of ATP as the universal energy currency was an important step in bioenergetic evolution, displacing (we posit) acetyl phosphate. However, while ATP is universal across lineages, it is not the sole energy currency within the metabolism of individual cells by any means [80]. This is an interesting and possibly significant point, suggesting that much went on in early biochemistry before ATP became a common currency. The simplest explanation for ATP's rise to prominence is that it was a consequence of the substrate specificity of the rotor–stator-type ATPase, a protein that is as universal among cells as the code, and that is unquestionably an invention of the world of genes and proteins [42]. Given genes and proteins, the origin of molecular machines such as the ATPase is an admittedly impressive, but not conceptually challenging, evolutionary step; it is a far less problematic increment than the hurdles that had to be surmounted at the origin of the ribosome and the code [110,117].

In an alkaline hydrothermal environment, the ATPase has immediate beneficial function. The naturally preexisting proton gradient [9] at the interface of approximately pH 10 alkaline effluent (alkaline from serpentinization) with approximately pH 6 ocean water (slightly acidic by virtue of high CO2 content) provides a geochemically generated chemiosmotic potential that ‘just’ needs to be tapped, that is, converted into chemically harnessable energy in the form of high-energy bonds, exactly what rotor–stator-type ATPases do. That would have put a very high-energy charge on the contents of the inorganic compartments within which we presume that this was all taking place, accelerating biochemical innovations, energetically financing gene inventions, and the selective pressure on evolving proteins to adapt to a new energy currency is evident. ATP-binding domains are so prevalent in genomes [118], not because ATP is a constituent of RNA, but because it became the most popular energy currency. The ATPase transduced a geochemically generated ion gradient into usable chemical energy, and since the energy was free, the means to harness it as ATP ‘just’ required a suitable protein for the job, a complicated protein [119], but a protein.

One might object that the multi-subunit rotor–stator-type ATPase is too complicated as a starting point, and that it is hence preferable to assume that some simpler form of energy transducing protein, for example, a pyrophosphatase, preceeded it [120]. But the observation from modern microbes to be accounted for, in the evolutionary sense, is that all prokaryotes harbour the rotor–stator-type ATPase. Similar to a handful of only 30 or so other proteins, it is as universal as the ribosome; hence it was probably present in the common ancestor of all cells, which in this model was not free-living but rather was confined to the system of inorganically formed compartments within which it arose. Consistent with that, prokaryotic ATPases show a clear dichotomy into two types: the eubacterial (or F-type) ATPase and the archaebacterial (or A-type) ATPase [121]. That dichotomy is also mirrored by a number of other cellular, molecular and gene homology attributes, indicating that the deepest, most ancient split in the prokaryotic world is that separating eubacteria from archaebacteria [111,116,122].

That suggests that the last universal common ancestor (LUCA), which we posit to be a geochemically confined ancestor (figure 1), possessed among other things genes, proteins, the code, along with biosynthetic pathways for amino acids, bases and cofactors to support functional ribosomes and an ATPase that could tap the naturally chemiosmotic gradient at the vent–ocean interface. But it was far short of being a free-living cell. Many attributes of archaebacteria and eubacteria are so fundamentally different that, for lack of similar chemical intermediates in the pathway or for lack of subunit composition or sequence similarity, independent origin of the genes underlying those differences is the simplest explanation.

Such differences include: (i) their membrane lipids (isoprene ethers versus fatty acid esters) [123], (ii) their cell walls (peptidoglycan versus S-layer) [124], (iii) their DNA maintenance machineries [116,125], (iv) the 31 ribosomal proteins that are present in archaebacteria but missing in eubacteria [126,127] (v) small nucleolar RNAs (homologues found in archaebacteria but not eubacteria) [128], (vi) archaebacterial versus eubacterial-type flagellae [129], (vii) their pathways for haem biosynthesis [130,131], (viii) eubacterial- versus archaebacterial-specific steps in the shikimate pathway [132,133], (ix) a eubacterial-type methylerythrol phosphate isoprene pathway versus an archaebacterial-type mevanolate isoprene pathway [134], (x) a eubacterial-type fructose-1,6-bisphosphate aldolase and bisphosphatase system versus the archaebacterial bifunctional aldolase-bisphosphatase [135], (xi) the typical eubacterial Embden–Meyerhoff (EM) and Entner–Doudoroff (ED) pathways of central carbohydrate metabolism versus the modified EM and ED pathways of archaebacteria [136], (xii) differences in cysteine biosynthesis [137], (xiii) different unrelated enzymes initiating riboflavin (and F420) biosynthesis [138], and (xiv) in very good agreement with figure 2b, different, unrelated, independently evolved enzymes in core pterin biosynthesis [139], to name a few examples. The pterin biosynthesis example is relevant because the cofactors H4F, H4MPT and MoCo, which are central to the eubacterial and archaebacterial manifestations of the Wood–Ljungdahl pathway are pterins (figure 2d), suggesting that methyl synthesis occurred geochemically (non-enzymatically) for a prolonged period of biochemical evolution.

It has been argued that these deep differences between archaebacteria and eubacteria reflect comparatively recent environmental adaptations; a specific claim is that archaebacteria adapted very recently [140]. However, many eubacteria thrive in hyperthermophilic environments without going to the trouble of replacing all their membranes and walls, DNA replication, etc., while plenty of archaebacteria thrive in mesophilic environments such as the open oceans and soils. It is therefore not the case that archaebacteria are somehow better adapted to hyperthermophilic conditions whereas bacteria are better adapted to mesophilic conditions, even though the perception of archaebacteria as extremophiles persists. Alternatively, the differences of archaebacteria and eubacteria have been interpreted as adaptations to energy stress [141], but both acetogens and methanogens live at the lower end of the free energy spectrum whereby acetogens live from less energy than methanogens [142,143] but without having become archaebacteria, so the energy stress argument is clearly not robust. Given the almost equally wide environmental representation of both groups, both in energy-rich and in energy-poor environments [144], the deep divergence between them in all likelihood simply reflects the very early evolutionary divergence, in our view physical divergence, of replicating compartment contents within the mound that led to two stem lineages which became the common ancestors of eubacteria and archaebacteria, respectively (figure 1). Physiologically, those stem lineages gave rise to acetogens and methanogens. The similarities and differences in carbon metabolism (the Wood–Ljungdahl pathway) in acetogens and methanogens outlined earlier (figure 2) are paralleled by similarities and differences in their mechanisms of energy conservation (figure 3), which entail chemiosmotic coupling.

At first sight, the idea that chemiosmosis is a very ancient means of energy transduction might seem counterintuitive. More familiar to many is the old (and popular) doctrine that the most ancient pathway of energy metabolism is a fermentation such as glycolysis [77], an idea that goes back at least to Haldane [2] and hence arose long before anyone had a clue that biological energy can be harnessed in a manner that does not involve substrate-level phosphorylations and ‘high-energy’ bonds [149,150]. In modern life, all biological energy in the form of ATP comes ultimately from chemiosmotic coupling [151], the process of charge separation from the inside of the cell to the outside, and the harnessing of that electrochemical gradient via a coupling factor, an ATPase of the rotor–stator-type. It was not until the 1970s that it became generally apparent that Mitchell [152] was right, his Nobel prize coming in 1978, and it is hard to say when it became clear to microbiologists that all fermentative organisms are derived from chemiosmotic ancestors. We also note that Mitchell's consideration of the problem of the origin of life introduced key concepts of his later chemiosmotic hypothesis, including a definition of life as process, and the idea of vectorial catalysis across a membrane boundary that is inseparable either from the environment or from the organism itself [153].

The maxim that glycolysis is ancient might be an artefact of experience, since it was the first pathway both to be discovered and that we learned in college; in that sense, it really is the oldest. When one suggests that chemiosmotic coupling in methanogens or acetogens might be ancient, many listeners and readers shy away, mainly because the pathways are unfamiliar and often entail dreaded cofactor names. The familiar cytochrome complexes, quinones and quinone-based charge translocation are lacking in these apparently ancient life forms, which possess no complexes I, II or III, no terminal oxidases, and only a single coupling (ion pumping) site each. Their energy metabolic pathways differ so fundamentally from the familiar respiratory chain-type chemiosmotic pathways of E. coli or mitochondria that only recently have biochemists uncovered how their basic energetics works. Yet, despite these fundamental differences, and their unfamiliarity to most of us, the bioenergetics of methanogens and acetogens nonetheless feature bona fide chemiosmotic circuits, with coupling proteins, ion electrochemical gradients over membranes and the ATP synthase. These pathways are illustrated in figure 3, which recapitulates the schemes presented by Buckel & Thauer [143], Thauer et al. [40] and Poehlein et al. [146], for methanogens and acetogens that lack cytochromes. Acetogens and methanogens are unique among chemiosmotic organisms studied so far in that both types exist as forms that lack cytochromes and quinones.

In brief, methanogens that lack cytochromes, as Methanobacterium marburgensis (figure 3a) use their version of the Wood–Ljungdahl pathway to synthesize methyl-H4MPT, but instead of donating it to CoFeS and CODH/ACS as in carbon metabolism (figure 2b), they donate it to the thiol moiety of CoM in an exergonic reaction (ΔG°‘ = ca –30 kJ mol−1) catalysed by a membrane-bound methyltransferase (MtrA-H in figure 3a), which conserves energy via the pumping of Na+ ions across the cytoplasmic membrane [40]. CoM-SH is regenerated from methyl-CoM by methyl-CoM reductase, which releases methane by condensing the CoM moiety with the thiol group of coenzyme B (CoB-SH) producing the heterodisulfide CoM-S-S-CoB. The CoM-SH and CoB-SH thiols are regenerated by reduction with electrons from H2 by a heterodisulfide reductase (Hdr in figure 3a), in a highly exergonic reaction (ΔG°‘ = –55 kJ mol−1) [145]. Hdr also generates Fdred, which serves as the low-potential reductant for the formyl-MF dehydrogenase step at the beginning of the WL-pathway, and the significance of this Hdr reaction will be discussed shortly. Per mol of methane, 2 mol Na+ ions are pumped, whereas the ATPase requires about 4 Na+ per ATP [145]. The overall reaction is that given in reaction (2.4), whereby, per mol methane, the cell is able to condense about 0.5 mol of ATP from ADP and Pi [145]. Aside from the anhydride bond in the ATP that is the product of the pathway, there are no thioesters, no acyl phosphates and no ATP consumption involved in methane production. The main energy currency is low-potential-reduced ferredoxin, Fdred [143].

Acetogens that lack cytochromes use their version of the WL-pathway to make methyl-H4F, as sketched in figure 3b, which is modified from Pohlein et al. [146] and Buckel & Thauer [143] for Acetobacterium woodii. ATP investment (a concept familiar from glycoysis) is required at the endergonic formyl-H4F synthetase step, formyl phosphate (figure 2c) being the ‘activated’ ATP-generated reaction intermediate, both in E. coli [154,155] and in the acetogenic enzyme [156]. The methyl group of methyl-H4F is transferred via CoFeSP to CODH/ACS where the acetyl-CoA is released. The energy in the thioester bond is conserved as acetyl phosphate in the phosphotransacetylase reaction. Acetate kinase converts acetyl phosphate and ADP into acetate and ATP, thereby recovering the ATP investment at the formyl-H4F synthetase step. Methyl synthesis consumes electrons from H2 according to reaction (2.3). The A. woodii formate dehydrogenase probably uses H2 directly as substrate [146], whereas the remaining two reduction steps are NADH-dependent [143,146]. The NADH stems from a soluble hydrogenase (Hyd) that, importantly, also generates Fdred in the process [148]. Fdred is the substrate for a membrane-bound protein called Rnf [147] that generates NADH and pumps, in the process, about one Na+ ion per electron transferred, thus generating the ion gradient that is harnessed by the ATPase, which requires about 4 Na+ ions per ATP. Per mol of acetate, about 0.25 mol of ATP is generated [143]. The ATP investment at formyl-H4F synthetase step and return via acetate kinase is strictly 1 : 1, so there is no net gain of ATP via that route. The currency of energy conservation is again Fdred, which powers pumping at Rnf to drive the ATPase.

Both acetogens and methanogens reduce CO2 to a methyl group using electrons from H2. This reaction actually should not proceed at equimolar concentrations, because the midpoint redox potential, E°′, of the H2/H+ couple is –414 mV, while the E°′ for the CO2/formate couple is –430 mV and for the formate/formaldehyde couple it is –580 mV. In simple terms, H2 is not a strong enough electron donor to reduce CO2 to the level of a methyl group; the electrons would have to go steeply uphill. This prompted Wächtershäuser [157, pp. 1790–1791] to conclude that H2 can be ‘excluded as the first source of electrons since its reducing potential is not sufficient for reducing CO2’. So how do chemolithoautotrophs reduce CO2 with electrons from H2? The simple answer would be that they generate low-potential Fdred with midpoint potentials of the order of –500 mV, but that is also still steeply uphill from H2, bringing us to the crux of the issue: how do they generate Fdred from H2? The answer, which only recently became apparent and is an exciting development in energetics, is that they perform a reaction called flavin-based electron bifurcation [143,158]. This electron bifurcation occurs at the reaction catalysed by Hdr in the methanogens, and at the reaction catalysed by Hyd in the acetogens (figure 3).

This newly discovered mechanism couples an endergonic reaction to an exergonic one, analogously to the familiar case of coupling an unfavourable reaction to ATP hydrolysis so that the overall energetics of the reaction are favourable. But in electron bifurcation, no ATP is involved. Instead, it entails the splitting of electron pairs, from H2, for example (E°′ = –414 mV), at a flavin, such that one electron goes energetically uphill to generate reduced low-potential ferredoxins (Fdred) with a midpoint potential close to –500 mV, with the energy for that uphill climb being provided by the second electron of the pair going downhill to a high-potential acceptor, such as the heterodisulfide CoM-S-S-CoB (E°′ = –140 mV) of methanogens [145] or NAD+ (E°′ = –320 mV) at the electron-bifurcating hydrogenase of acetogens [148]. Electron bifurcation permits conservation of biochemical energy in the currency of reduced ferredoxin—a principle of energy conservation that departs from Lipmann's [149] high-energy bonds. Fdred contains no cleavable ‘high-energy’ bonds, but electrons at low potential, so it is clearly a currency of chemical energy [143]. Oxidation of Fdred by a pumping complex such as Rnf [147] or the Ech hydrogenase [159] couples the electron bifurcation mechanism to chemiosmotic energy conservation (figure 4).

The electrons in low-potential Fdred are crucial for CO2 reduction in organisms that use the acetyl-CoA pathway. By splitting the electron pair in H2 with one going uphill to Fd the other going downhill to a high-potential acceptor, these organisms can reduce CO2 under what might seem to be energetically hopeless conditions. The prevalence of electron bifurcation among strict anaerobes suggests that it is a very ancient biochemical mechanism [73,143], one that was apparently a prerequisite for a lifestyle of reducing CO2 with electrons from H2, which is how organisms that use the acetyl-CoA pathway for their carbon and energy metabolism survive.

In anaerobes, Fd is usually one of the most abundant proteins in the cell, pointing to the central importance of this small, FeS cluster containing protein in their energy metabolism. But Fd is not the only abundant FeS protein in these organisms. The proteins of anaerobes are generally rich in transition metals and transition metal sulfide clusters at the catalytic centres of their proteins, catalytic clusters which are often replaced by NAD+ and other organic cofactors in aerotolerant species. The early Earth was devoid of molecular oxygen [30]; hence the first life was anaerobic. There is a consistent and robust line of biochemical evolutionary reasoning that FeS clusters, chemolithoautotrophy and anaerobes tend to reflect antiquity [44,49,160,161]. That line of reasoning is still current [73,162,163] and is reinforced by geochemical data providing evidence for the antiquity of methanogenesis [164].

If FeS clusters are relics from the ancient past, it could be that the most ancient organisms preserved more FeS-containing proteins in their genomes. Major et al. [165] found that among the 120 genomes sampled then, the methanogens had by far the highest frequency of FeS proteins. Here, we have updated their survey with 1606 genomes, looking for the frequency of 4Fe-4S clusters of the type CX2CX2CX3C. We observed the distribution shown in figure 5 which clearly shows that methanogens, acetogens and sulfate reducers have the highest frequencies of FeS proteins. This fits well with our general premises, but it does not prove anything. There is a point to be made, though, in that the case has recently been argued that ZnS, rather than transition metal sulfides, might have been the starting point of prokaryotic biochemistry [5]. However, it should be stressed that there is a big difference between the view that life emerged from a ZnS-mediated process and the view that life emerged from FeS and other transition metal sulfide-mediated processes. In the model of Mulkidjanian et al. [5], ZnS has the role of providing electrons (photochemically), a role that we ascribe to serpentinization (geochemically), but that is not such a big difference. The big difference is that Zn is a substrate in their model, providing two electrons per atom in the one-off photochemical reaction, not a catalyst (which FeNiS is in our model). Zn is not catalytic in the way that Fe, Ni, Mo, W or other transition metals are anyway, for which reason modern anaerobes lack ZnS centres participating in redox reactions of carbon or energy metabolism. The Zn model lacks redox catalysis and similarity to modern metabolism.

Following the physiological line of reasoning on evolution a bit further, Decker et al. [44] suggested that the next physiological group to arise following the acetogens and the methanogens were the sulfate reducers, strict anaerobes, many of which are autotrophs that use the acetyl-CoA pathway [167–171]. This notion would fit very well within the present considerations, including the circumstance that there are archaebacterial and eubacterial sulfate reducers [172]. It is furthermore supported by geochemical isotope evidence that attests to the existence of processes performed by sulfate reducers—sulfate/sulfite reduction and sulfur disproportionation—in the early Archaean era, around 3.4 billion years ago [173–175].

The ancestral metabolism performed by sulfate reducers was likely to have been sulfite reduction or sulfur disproportionation—and perhaps also disulfide disproportionation, as newer findings suggest [176]—since sulfur and sulfite would be produced abundantly from volcanic and hydrothermal SO2, whereas sulfate, prior to the advent of oxygen, would probably have had only very limited, localized significance [177]. In line with that, the use of sulfate as a substrate by sulfate reducers involves soluble ATP consumption steps to produce sulfite [172]. On the early Earth, localized sulfate concentrations could be formed abiotically from atmospheric photolysis of SO2, or biologically from sulfide-dependent anoxygenic photosynthesis or sulfur disproportionation, but it was not until the increased oxygenation of the atmosphere that oceanic sulfate levels started to increase and sulfate began to play a prominent role in the global sulfur cycle [178]. Dissimilatory sulfite reduction is a more widespread trait than sulfate reduction, and the enzyme responsible for it, the dissimilatory sulfite reductase (DsrAB) is clearly an enzyme of very ancient origin, and it is closely related to the assimilatory enzyme present in all domains of life [179]. Both assimilatory and dissimilatory enzymes result from a gene duplication event that preceded the divergence of the archaebacterial and eubacterial domains [180–182], so a primordial sirohaem-containing sulfite reductase was most probably present in LUCA. Note that the presence of sirohaem does not mean the presence of haem in LUCA, because sirohaem synthesis merely entails the insertion of Fe2+ into sirohydrochlorin, an intermediate of anaerobic cobalamine synthesis (which acetogens and methanogens possess) and sirohaem is furthermore a precursor of the archaebacterial (alternative) haem biosynthetic pathway [183].

Something else that speaks rather strongly for the antiquity of methanogens, acetogens and sulfate reducers is their prevalence in the deep biosphere, where, as on the early Earth, there are few ways to make a living. Sulfate reducers, acetogens and methanogens appear to be the most common inhabitants of ecosystems situated deep within the crust [184–188]. They inhabit such environments today, suggesting that their ancestors might have as well. Curiously, though not involved in energy metabolism directly, primitive forms of Dsr-like proteins are apparently abundant in methanogens [189], with a role in sulfite assimilation or detoxification, further stressing the close ties between sulfate reducers and methanogens.

Another link between methanogens and sulfate reducers is the occurrence in the latter of a number of Hdr-related proteins, suggestive of electron bifurcation [172]. For methanogens, the critical bifurcation reaction, described earlier, is catalysed by a complex of a heterodisulfide reductase with a hydrogenase, the MvhADG–HdrABC complex [145]. Homologues of HdrABC are found in other organisms, including at least one cytochrome-containing acetogen, Moorella thermoacetica [89], and many proteobacterial and clostridial sulfate reducers [172]. In both cases, it is suspected that HdrABC-mediated electron bifurcation is participating in core carbon and energy metabolism, but this has not been directly shown so far. For Moorella, HdrABC homologues are encoded directly next to the gene for methylene-H4F reductase, which catalyses a very exergonic reaction and was therefore once suggested as a coupling site for energy conservation [39]. It remains possible that the Moorella methylene-H4F reducatase does conserve energy, but not via ion pumping as originaly envisaged, rather via electron bifurcation [89] and Fd reduction instead. Recently, an electron-bifurcating ferredoxin- and NAD-dependent [FeFe]-hydrogenase (HydABC) from Moorella was shown to catalyse the reversible bifurcation reaction [190]. For the sulfate reducers, several organisms contain HdrABC or HdrA-like proteins, encoded next to genes for possible electron donor proteins such as an MvhADG hydrogenase, a NAD(P)H dehydrogenase or a formate dehydrogenase [172], strongly suggesting that electron bifurcation is also involved in their energy conservation.