Early bioenergetic evolution

Furthermore, several proteins directly implicated in sulfate/sulfite reduction are strikingly related to HdrABC subunits. Sulfate reducers contain menaquinone, but its role in respiration was for long a mystery because the terminal reductases, APS reductase (AprBA) and DsrAB, are soluble cytoplasmic enzymes. A link between the two started to emerge with the identification of two membrane complexes having quinone-interacting subunits, QmoABC and DsrMKJOP, as probable electron donors to AprBA and DsrAB [191–193]. Both these complexes have Hdr-related subunits. The QmoABC complex is essential for sulfate reduction [194], and two of its subunits, QmoA and QmoB, are both flavoproteins closely related to the bifurcating subunit HdrA. In the DsrMKJOP complex, the two DsrMK subunits are closely related to the membrane-associated HdrED enzyme present in cytochrome-containing methanogens. DsrK contains the special catalytic FeS cluster that in HdrB is responsible for reduction of the heterodisulfide, which in the case of the sulfate reducers is proposed to be the small protein DsrC that is involved in sulfite reduction by DsrAB [195].

How energy is conserved by the QmoABC and DsrMKJOP membrane complexes has not been clearly established, but a recent proposal suggested a new mechanism of electron confurcation—bifurcation in the reverse direction [196]—involving menaquinone in the case of Qmo [197]. In this proposal, the endergonic reduction of APS by menaquinol is coupled to its exergonic reduction by a low-potential soluble electron donor (such as Fd). In contrast to electron bifurcation where a single-electron donor (NADH, H2 or formate) is used to reduce two electron acceptors, one of high and one of low potential (most commonly Fd), in this quinone-linked confurcation two electron donors, of high (menaquinone) and low potential (Fd?), are used to reduce a single-electron acceptor. Soluble confurcation processes have already been described, such as those involving multimeric hydrogenases (NADH and Fdred as electron donors for H2 production) [196] and the NfnAB transhydrogenase (NADH and Fdred as electron donors for NADP+ reduction) [89,198]. Having menaquinone as one of the confurcation players would allow for transmembrane electron transfer and charge separation to effectively couple the process to chemiosmotic energy conservation. This could not occur if only a soluble low-potential donor was involved. It is possible that this, so far speculative, idea might also apply to other anaerobes for which the role of menaquinones is so far unclear, and where MK does not have a sufficiently low reduction potential to participate in core energy or respiratory metabolism. The cytochrome-containing acetogens such as Moorella, in particular, come to mind in this respect, because they contain menaquinone, but there is yet no obvious role for it in core metabolism [89].

The involvement of electron bifurcation in sulfate reducers points further to the antiquity of sulfate/sulfite reduction as an ancient anaerobic physiology and is consistent with the observation that a number of sulfate reducers use the acetyl-CoA pathway [199–201].

Electron bifurcation in methanogens and acetogens permits generation of membrane potential from H2 and CO2, which can then be used to drive net ATP synthesis via the ATP synthase, or in the case of M. marburgensis, to reduce ferredoxin via the energy converting hydrogenase, Ech. Thus, electron bifurcation helps cells generate what alkaline hydrothermal vents geochemically provide for free: ion gradients. Given the foregoing, the continuous reduction of CO2 to organics could have ultimately driven the emergence of RNA, proteins and genes within the vent pores: what we shall call protocells, meaning the organic contents of inorganic pores within the microporous labyrinth of alkaline hydrothermal vents, but not, as yet, fully functional or independently dividing cells. How (and why) did these confined protocells, dependent on natural pH gradients, begin to generate their own gradients, enabling them ultimately to leave the vents as self-sufficient, free-living cells?

One possible solution depends on the counterintuitive properties of membranes in the presence of natural ion gradients [43]. Apart from the remarkable sophistication of modern respiratory chains, a second reason that chemiosmotic coupling is usually considered to have arisen late relates to the permeability of early membranes. Chemiosmotic coupling today requires an ion-tight membrane, so that ions pumped out return mostly through the ATP synthase, driving ATP synthesis. If ions can instead return through the lipid phase of a leaky membrane, ATP synthesis is said to be uncoupled—literally, short-circuited—and most of the energy consumed in pumping is simply dissipated as heat. Thus, pumping protons across a leaky membrane costs a lot more energy than can be conserved as ATP, making it worse than useless. The problem at the origin of life is that early membranes would certainly have been leaky, hence pumping protons across them could offer no advantage, especially in the presence of small organic acids which naturally dissipate proton gradients. Pumping protons at the vent is pointless anyway, as the proton gradient already exists. A second related problem is that archaebacteria and eubacteria do not have homologous membranes; hence we have argued that their last ancestor, confined to vents, did not have a modern membrane. While simple phase separation of amphiphiles, fatty acids, alkanes and other hydrophobics would tend to produce hydrophobic layers within the vent pores, probably even lipid bilayers, these were likely to be very leaky to small ions.

But arguments against early chemiosmotic coupling share a common flaw: they assume that pumping arose before harnessing. In the presence of natural proton gradients, this is not the case. We have argued that the ATPase arose before pumping mechanisms. The passage of protons through the ATP synthase today is about 4–5 orders of magnitude faster than through the lipid phase of the membrane. But with a free, geochemically provided gradient, membrane permeability could be orders of magnitude greater, while still allowing net ATP synthesis. All that is required is that protons should pass through the ATP synthase as well as through the lipid phase (and that they be removed by the alkaline effluent). Any improvements in the ion-tightness of the membrane would therefore be beneficial, as they would funnel a larger proportion of protons through the ATPase. Hence if there were any genetic component to early membranes, we would predict that coupling would steadily improve, which is to say that organic membranes would tend to become less permeable to small ions over time. On the face of it, that seems a reasonable statement, but it leads to an unexpected problem.

The problem relates to the transfer of charged particles—protons—across the membrane. In vents, neither hydrothermal fluids nor ocean waters carry a net charge: the excess of H+ in the ocean is balanced by HCO3–, whereas the excess of OH– in hydrothermal fluids is balanced by Mg2+. Thus, the transfer of a single proton across the membrane down a concentration gradient also transfers a positive charge, and this opposes the further transfer of charge—which is to say, more protons. Soon, a Donnan equilibrium is established, in which the electrical charge across the membrane balances the concentration gradient, annulling the proton-motive force. This problem is only unmasked if the membrane can actually hold an electric charge, which is to say, only if the membrane has become impermeable to small ions. Before that, however—if the membrane is leaky or discontinuous—the system is essentially open, and charge does not enter into it. Any H+ crossing the membrane will react with hydrothermal OH– to form water, which is swept away in the flow, negating the accrual of charge. In such an open system, the pH gradient is maintained not by pumping, but by the continuous circulation of hydrothermal fluids and ocean waters, continually juxtaposing phases in redox and pH disequilibrium. Thus, at an early stage, the proton-motive force can only exist in the presence of leaky membranes and an open system: far from being a problem, leakiness is actually a necessity.

What happens when the membrane tightens off to small ions and forms cell-like structures? Transfer of protons into an impermeable cell will result in a Donnan equilibrium that opposes further transfer unless counterions are also transferred, or unless the proton can be pumped out again. Pumping out a proton is problematic. Either the cell needs to invent machinery to pump out protons as soon as the membrane becomes impermeable (plainly impossible) or it needs to invent this machinery while the membrane is still permeable, i.e. before it needs it. This is also unlikely, as the cell would need to actively pump protons (costing energy) from an internal concentration of pH 9–10 (less than or equal to 1 nM) across a leaky membrane against a 1000-fold concentration gradient, which is immediately dissipated through the leaky membrane—a common objection to chemiosmotic coupling evolving early. The solution is surprisingly simple: a Na+/H+ antiporter, an otherwise unassuming protein that would hardly be a prime suspect for evolutionary heroics, except in alkaline hydrothermal vents.

In a world of genes and proteins, a Na+/H+ antiporter is far easier to invent than an ATPase. It is a single protein, and an ancient one that is also a component of both Ech and Complex I [202,203]. But it confers an inordinate benefit. An antiporter transfers a positive charge in the opposite direction to the proton, meaning that overall the passage of a proton down a concentration gradient is charge neutral. Thus, with an antiporter, there is no problem with Donnan equilibria: the system does not bung up. What is more, the passage of H+ down a natural gradient drives the efflux of Na+ ions. This has two beneficial effects, one immediate and the other longer term. The immediate benefit is improved coupling, assuming, as we do, that ancient ATPases were promiscuous for H+ and Na+, as modern ATPases of methanogens are [204]. Modern membranes are 2–3 orders of magnitude more permeable to H+ than to Na+, and there is no reason to suppose that early membranes would have been any different, especially in vents, where the presence of small organic acids uncouples H+ gradients but not Na+ gradients. So an antiporter transduces a natural proton gradient into a biochemical sodium gradient, which immediately improves coupling at essentially no energetic cost. Na+ extrusion is powered by the free geochemical proton gradient, not by active pumping (which consumes an energy budget that could otherwise be spent on carbon assimilation). Any improvement in coupling, of course, requires that the ancient ATPase could use Na+ ions, similar to how the modern one can [204,205], perhaps because hydronium ions (H3O+) have an identical charge and similar ionic radius to Na+ (ionic radii for unhydrated H3O+ = 115 pm, and for Na+ = 117 pm).

A number of apparently ancient bioenergetic proteins found in both methanogens and acetogens, notably the ATPase and Ech, are promiscuous for H+ and Na+, which might be a relic of early coupling.

The longer term benefit relates to optimizing enzyme function by ionic balance. Most modern cells have low intracellular Na+ relative to K+, unlike the oceans, where Na+ concentration is much higher, 470 mM in oceans versus approximately 10 mM in cells [5]. Some universal and ancient proteins, notably those involved in translation, are optimized to this ionic balance, raising the question of how, if life arose in submarine alkaline hydrothermal vents, did intracellular Na+ concentration become lowered to this optimum? An H+/Na+ antiporter would readily explain how, but as the secondary consequence of a more immediate bioenergetic pressure.

The Na+/H+ antiporter also helps to solve the larger problem of the origin of active pumping. In the presence of free geochemical proton gradients, protocells in vents can adapt to a sodium-motive force, all the while being powered by a proton-motive force. Eventually, cell membranes became tight to Na+, if not H+. So how did active pumping originate?

Here, an important point needs to be made, lest the reader accuse us of invoking preadaptation with regard to pumping. There is no advantage to active pumping at regions of the vent where gradients are sharp by means of effluent flux. In distal regions of the vent, however, where alkaline effluent flux is impaired, energy coupling, and hence proliferation would cease. Only such protocells that ‘learned’—via the standard workings of natural variation and selection—to generate their own gradients in the context of their existing repertoire of redox reactions, could proliferate in such margins. And only such protocells could make the final transition to the free-living lifestyle. So there is clear proliferation benefit for the invention, by chance mutation, of proteins that can pump, and that benefit exists at regions of the vent where flux is impaired. This is likely the initial selective advantage of pumping.

What are the primordial active ion pumps? A simple solution is suggested by the sole coupling site of acetogens that lack cytochromes, Rnf. Electron transfer from ferredoxin to NAD+ drives the extrusion of Na+ via Rnf [147]. The problem of generating reduced ferredoxin in protoacetogens via flavin-based electron bifurcation at the hydrogenase step had been solved previously for the purpose of reducing CO2. The circuit was nearly complete, only redox balance remained to be achieved. Reoxidizing the NADH generated at the hydrogenase and Rnf reactions so as to synthesize methyl groups closes the stoichiometric loop, yielding an H2/CO2 dependent, acetate excreting entity that, upon cellularization (bubbling off, one might say) was energetically free to leave the vents.

Methanogens found a different and completely unrelated route, while also making use of flavin-based electron bifurcation, suggesting independent origins of active pumping. In the case of methanogens, the Na+-motive force is generated by pumping at the highly exergonic methyl transfer step (ΔG°′ = ca –30 kJ mol−1) from methyl-H4MPT to CoM by the methyl transferase (Mtr) complex, the only coupling site in methanogens that lack cytochromes [145]. Reoxidation of CoB-SH with methyl-CoM to generate the heterodisulfide CoM-S-S-CoB and methane establishes a redox balanced Na+ pumping circuit. In the protomethanogen lineage, this Na+ gradient could also be tapped via Ech for net CO2 to proceed. Regardless of the exact details, it appears that the deep divergence of overall design in methanogen and acetogen bioenergetics, despite their distinct commonalities, occurred within the vents.

From rather similar starting points (but with membranes that had already diverged), both methanogens and acetogens solved the problem of active pumping using an Na+/H+ antiporter, both relied on chemiosmotic ATP synthesis, and both evolved similar mechanisms of flavin-based electron bifurcation, using a similar set of hydrogenase proteins. But divergent stem lineages solved the details of the problem in different ways. And in both, flavin-based electron bifurcation gave rise to a balanced redox circuit and Na+ gradient generation via a single coupling site. Other elements of the more familiar respiratory chain, notably quinones, the Q cycle, and cytochromes b and c, arose later. This interpretation is supported by the remarkable homology of several subunits of complex I to the proteins discussed here: FeFe trimeric hydrogenases, NiFe hydrogenases, ferredoxin-reducing subunits, Ech and Na+/H+ antiporters [202]. The only further subunits required for true complex I function were quinone and NADH-binding subunits. We discuss the origin of quinones and cytochromes in §13, but it seems likely that they did indeed arise after Ech and Na+/H+ antiporters. In this case, it is also noteworthy that complex I itself possesses some of the same Na+/H+ promiscuity [206] as Ech and the ATP synthase do. The fact that the proton-motive force is more universal today than the sodium-motive force probably derives from evolutionary interpolation of quinones—which are always coupled to proton-dependent redox reactions—on these earlier, necessarily promiscuous origins [205].

Thinking things through, one additional aspect comes to the fore that is not immediately related to energetics. Namely, if the transition to the free-living state was truly independent in the archaebacterial and eubacterial lineages, as we suggest, then prokarytic cell division well could be predicted to have arisen twice independently. A look at the comparative genomics of cell division in archeabacteria and eubacteria provides evidence for the ubiquity of the FtsZ-derived system in eubacteria, and widespread occurrence of FtsZ in archaebacteria, but in addition the occurrence of ESCRTIII-related cell division machineries in some archaebacterial lineages [207]. Thus, while one can say that the ESCRTIII-related route is specific to archaebacteria, it is not obviously present in their common ancestor, while FtsZ (which seems to be a main cell division protein in methanogens) probably was [207]. Assuming that cell division had to evolve before cells escaped to the free-living state, what might cell division within rigid inorganic confines have looked like?

The example of spore formation in the acetogen Acetonema longum [208] is possibly instructive. Sporulation is a specialized case of cell division where the daughter cell grows inside the mother cell: some members of the clostridia, such as Epulopiscium, use the cell within a cell strategy for normal cell division [209]. During that process in Acetonema, the inner membrane of the mother cell becomes the outer membrane of the spore (the daughter cell), which grows within the mother cell and which initially posseses two membranes, inner and outer, one of which is ultimately shed [208]. This general kind of cell division has, not unreasonably, been suggested as the ancestral source of the outer membrane in Gram-negative bacteria; it might also represent an ancestral state of cell division reflecting a growth process within physical confinement.

Recently, Nitschke & Russell [210, p. 482] have argued that the Wood–Ljungdahl pathway ‘appears inadequate to drive reductions, condensations and biosyntheses required of an emergent metabolic cooperative’. The gist of this argument is that the H2–CO2 couple does not present enough free energy to get life started (although methanogens and acetogens do just fine). Their inference is that high-potential electron acceptors such as NO or NO3– with midpoint potentials near or exceeding O2 must have been involved at the very earliest stages of chemical evolution [210,211], and in the latest formulation entailing oxidant-driven methane oxidation as the first step towards the synthesis of organic molecules [212]. A severe problem with that view, though, is that it demands that existence of a 1 : 1 molar ratio of high-potential acceptors for every organically incorporated carbon atom at the onset of primordial biochemistry; that would shift the oxidation state of such an environment dramatically.

This is especially true when one recalls that at the onset of biochemistry, there was little or no specificity in catalysis, such that a vast molar excess of ‘non-biogenic’ (side product) reduced carbon compound was required for the synthesis of every ‘biogenic’ (biologically useful) one [43]. The consequence is that in the presence of abundant high-potential acceptors of the type that methane oxidation [212] would demand, organic synthesis would probably just ‘go up in smoke’ being pulled towards CO2. The reason is because amino acid synthesis and cell mass synthesis, while requiring little if any energy input or even being exergonic under strictly anoxic hydrothermal vent conditions [36], become extremely endergonic even under very mildly oxidizing conditions, such as microoxic conditions corresponding to only 1/1000th of present oxygen levels [35].

Specifically, the synthesis of cell mass requires 13-fold greater energy input (or even more if NO3– is the nitrogen source) under microoxic than under anoxic conditions [35]. That meshes with the observations of Heijnen & van Dijken [213] that anaerobic chemolithoautotrophs such as methanogens require of the order of 30–40 kJ to synthesize a gram of cell mass, whereas aerobes require of the order of 80–170 kJ to synthesize a gram of cell mass [35]. Thus, the argument that high-potential electron acceptors for methane oxidation were needed to get organic synthesis started [212] has the problem that their presence would tend to preclude accumulation of reduced organic compounds such as amino acids [35], because the reducing conditions would be gone. One might counter that these microoxic conditions apply to the oceans, whereas the internal compartments of hydrothermal vents are highly reducing, gases such as NO are as likely to traverse hydrophobic walls as CO2, and certainly more easily than protons, and would therefore inevitably alter the oxidation state within the vent. Indeed, McCollom & Amend [35] point out that ‘the primary control on the energetic requirements for biomass synthesis is the oxidation state of the environment’ which is consistent with the observations of Amend & Shock [34], who pointed out that in environments of high or even moderate reducing potential, the autotrophic synthesis of amino acids requires little energy input, in contrast to oxidizing environments. Thermodynamic considerations would strongly favour a reducing environment for the synthesis of life's first building blocks over an environment in which the synthesis of reduced carbon compounds had to take place against the workings of strongly oxidizing agents.

An additional complaint of Nitschke & Russell [210] that ‘acetate has defied chemical synthesis in aqueous solution directly from CO2 and H2 in the laboratory’ probably relates to electron bifurcation, whereby aqueous acetate synthesis from CO2 with Fe±0 is facile [62].

And when it comes to the topic of strong oxidants, there is a lot to be said for molybdenum. Mo is involved in many interesting reactions, among them CO2 reduction in the initial steps of the acetogen and methanogen methyl synthesis pathways (figure 2) and N2 reduction in nitrogenase, which was recently shown to possess a carbon atom with only Fe ligands in the centre of its FeS cluster [214]. Mo is also involved in some reactions where it was long thought that only a strong oxidant such as O2 could do the job. A recent example is the characterization of a molybdoenzyme from Sterolibacterium denitrificans that anaerobically hydroxylates a tertiary carbon atom in sterol [215], a reaction hitherto thought to require O2. Another example is the anaerobic hydroxylation of ethylbenzene by ethylbenzene dehydrogenase, where Mo(VI) performs the oxidative step in the reaction mechanism [216].

One might ask how phylogenomics stacks up against these concepts. Groups trying to work out the ‘phylogeny’ of prokaryotes [217] have come up with various schemes to classify groups, mainly based on the identification of some core set of genes that are concatenated and used to create a phylogenetic tree, a procedure laden with problems [218]. While such phylogenomic studies are worthwhile undertakings, they come with the heftiest of caveats. This is because even with massive amounts of data and refined molecular phylogenetic methods, it is a challenge to get the orders of mammals [219] or the orders of flowering plants [220] satisfactorily sorted out, and those processes span ‘only’ about 200 Myr into evolutionary time. Much, much harder are the early eukaryotic groups, which go back about 1.5 billion years [221], or even more difficult, linking eukaryotes to prokaryotes [222–224]. With that in mind, dismal appear the prospects of getting branching patterns sorted out for prokaryotes, which have been around for 3.8 billion years or more [30,164], assuming that there are any real branches in that phylogeny to be recovered in the first place [225], and the goal probably pushes phylogenetics beyond its limits.

What if we poll the genes that are used in such concatenation studies to see whether they individually tend to support the trees that they produce in concatenated analyses? We did that in figure 6. The result shows that the 48 core genes that can be distilled to be present in all of a small sample of 100 genomes recover, individually, the deep split between archaebacteria and eubacteria, and they recover recent phylogenetic signals at the tips of the tree (figure 6). The tree of life as currently constructed is based only on about 30 genes [217], which correspond to about 1 per cent of the genes found in an average prokaryotic genome; in that sense the tree of life is more the tree of 1 per cent [227]. But by viewing the tree in terms of how often different genes find the same branch for the concatenated tree they purport to uphold, what we see is not a tree of life, but a tree of tips [228]. In the deeper branches, not a single individual gene branches in the same way as the concatenated tree does. This underscores what Doolittle has been saying for some time [225,229], namely that scientists are unable to muster positive evidence that any given gene or set of genes has had the same history all the way back to the beginning of life's first cells.

Thus, while concatenated gene sets (analysed individually or in some concatenated manner) readily provide groups (a classification), the branching order of those groups remains obscure. But more importantly, even if one were, by some means, to get a branching pattern sorted (or agreed upon by consensus) for some core prokaryotic gene set, there remain incontrovertible and fundamental differences between classification of eukaryotic groups and classification of prokaryotic groups: the former is largely natural, but the latter is not [229]. For example, we can generate an approximate phylogeny for the evolution of vertebrates and map onto that phylogeny the origin of jaws, fins, lungs, feet, wings and so forth in a meaningful manner. But the same is not possible for prokaryotes because they readily spread their genes via lateral gene transfer (LGT), and what we find in modern genomes are useful collections of genes that represent viable strategies for survival in existing niches [225]. In other words, we cannot take any given prokaryote phylogeny and map out the linear evolution of physiological traits, because their evolutionary patterns are too complex, often marked by very sparse distributions.

Notwithstanding, it is possible to generate an argument, based on phylogenetic trees, that acetogens and methanogens are among the true latecomers in evolution, recently evolved and derived specialists [212]. But the branches in phylogenetic trees built from sequence data depend on many things, mainly things that go on inside computers [222]. And for each tree that suggests methanogens and acetogens to be latecomers [211], there is one to show that methanogens [222,224,230] and clostridia (acetogens) [217] branch at the base of their respective domains.

Why is the split between eubacteria and archaebacteria in figure 6 (and in many other such trees) so deep and so clear, while the next branches disappear? Figure 6 summarizes trees of universal genes. While they do not recover the middle ages of prokaryote evolution, they do tell us what we already know, that archaebacteria and eubacteria are very different. Under our current premises, their divergence corresponds to a comparatively short period of time, less than ca 400 Myr, that separates the advent of water at 4.2 Ga and evidence for life at 3.8 Ga [11,30]. Perhaps the simplest interpretation for this deep divergence is that it reflects the circumstance that the stem archaebacteria and stem eubacteria diverged at a time when proteins were still being invented and those that existed were getting better at what they do. In modern evolution, proteins evolve neutrally on the whole, functional constraints causing sequence conservation are the norm, and have been for most of Earth's history. But at the origin of genes, evolution was anything but neutral—it was fierce competition for organization of available matter into things that could organize matter best.

The examples encountered so far for independent origin of genes in archaebacteria and eubacteria clearly indicate that the invention of the full gene repertoire needed to exist as a free-living cell had not gone to completion in the confined universal ancestor. In the two stem prokaryote lineages, proteins were getting better at what they do (positive selection quickly fixes new mutations) and coming to rest in different regions of the fitness landscape on the plane of sequence space. This causes rapid sequence evolution, which translates into many accumulated sequence differences per unit time, which translates into long branches. Once proteins reached a certain level of optimization at what they do, functional constraints causing sequence conservation (slower mutation accumulation) became the norm.

We can look for proteins that are of sufficiently wide distribution to assume that they might have been present in the common ancestor of archaebacteria and eubacteria (table 1). If we opt for strict criteria as ‘present in all genomes’ then we would come up with a list of the roughly 31 proteins that various groups [217,226,232] have examined. An extended version of that list would also include many proteins that are less than 20 per cent identical in many pairwise comparisons. Our criteria for ‘widely distributed’ are more relaxed with respect to gene presence or absence but a bit more strict with respect to sequence identity. The protein families used by Nelson-Sathi et al. [231] to construct deep phylogenetic trees are convenient. The criterion of sharing at least 30 per cent amino acid identity was applied, because alignment and phylogeny procedures produce severe artefacts with amino acid sequences that share substantially less identity [233]. Concerning gene presence or absence, one can relax the stringency a bit, allowing for some loss, and ask whether a protein is present in at least one representative of various higher prokaryotic taxa corresponding to phylum, for example. In the study of Nelson-Sathi et al. [231], there were 11 higher archaebacterial taxa (75 genomes total) and 30 higher eubacterial taxa (1143 genomes total) sampled. Table 1 shows how many genes fulfil the distribution criteria for being present in (only!) at least one member each of 11/11 archaebacterial groups and in at least one member each of 20/30 higher prokaryotic taxa and sharing more than 30 per cent amino acid identity.