1. CHANCE AND NECESSITY

Forty years ago, Jacques Monod published his seminal work, Chance and Necessity (Monod, 1971). Monod's bleak perspective, that the origin of life was a freak accident, and that we are almost certainly alone in an empty universe, was opposed by Christian de Duve, who has argued that life is a "cosmic imperative", and as such will arise throughout the universe (de Duve, 1995). It is emblematic of the state of play that either view could still be right. A similar dialectic exists between evolutionists. Steven Jay Gould advocated the contingency of evolution: wind back the clock, and life will evolve again along radically different lines (Gould, 1989). Simon Conway Morris argues that convergence is the stronger force – the principles of engineering dictate that the same, or similar, solutions will emerge repeatedly (Conway Morris, 2003). Wind back the clock and humanity will enjoy a second coming.

In the 40 years since Chance and Necessity, we have certainly learned a great deal about the origin of life. But Monod would perhaps be surprised to discover that many of the same concepts are still bandied around. The Miller-Urey experiment is still regularly cited, rightly – it is the cornerstone of all experimental research into the origin of life (Miller, 1953). Primordial soup is equally venerable, dating back to Haldane and Oparin in the 1920s (Haldane, 1929; Oparin, 1924); and soup still stains the pages of modern text books. As one of the pioneers of molecular biology, notably the discovery of messenger RNA (Jacob and Monod, 1961), Monod would surely have been delighted by the RNA world. Promoted mostly by molecular biologists, with no love of geochemistry, the RNA world is usually taken to play out in soup, if only for simplicity (Orgel, 2008). There is little that Monod would find unfamiliar with this world. No wonder he saw life as an improbable freak. Apart from the absence of evidence for soup, or the absence of heterotrophs that might have fermented soup from anywhere near the base of any phylogenetic tree, the lack of thermodynamic driving force makes the emergence of life from soup improbable in the extreme (Lane, Allen and Martin, 2010). If any doubts remain, Say and Fuchs (2010) have shown that the enzymes involved in synthesizing sugars predate those that break sugars down.

But Monod knew nothing about vents. Black smokers were not discovered until 1977, seven years after his book, and one year after Monod died. There's nothing of the equilibrium about smokers, but they suffer equally intractable problems (Bada, 2004). Black smokers are short-lived and unstable (Kelley et al., 2002), intensely hot, and lend meaning to the word 'vent': a chimney pumping effluent straight out into the ocean. Today, life in black smokers depends on oxygen, and therefore indirectly photosynthesis: not heat, not pyrites, not surface chemistry, but the reaction of H₂S with O₂.

Four billion years ago there was little, if any, oxygen. Wächtershäuser's conception of surface chemistry pulled by pyrites formation (Wächtershäuser, 1988; 2006) would have surprised Monod, as it did everyone else, but ingenious as the idea is, it bears no relationship to life as we know it today. If the failing of soup is to ignore thermodynamics, the failing of surface catalysis is to ignore microbiology, the chemistry of real life. Another big failing of surface catalysis is that, as soon as a product is released from the surface, it is flushed out of the vent to disperse in the ocean. Again, the sheer improbability of life seems little short of a miracle.

But there is an alternative, pioneered over the last 20 years by Michael Russell and colleagues, which turns the probability of life's origins on its head (Russell et al., 1993; 1994; Russell and Hall, 1997; Martin and Russell, 2003; Martin and Russell, 2007; Russell and Kanik, 2010). It lacks only for a colourful name to distinguish it from black smokers. White non-smoker will not do; but "alkaline hydrothermal vent" is rather less memorable than "black smoker", and not sufficiently distinct. Every other article on alkaline hydrothermal vents seems to be illustrated with pictures of black smokers (see Figure 1). That will not do either.

Alkaline hydrothermal vents are formed not by volcanism but by the process of serpentinization, in which water reacts with ultramafic rocks from the Earth's upper mantle, like peridotite, containing magnesium and iron-rich minerals, in particular olivine (Sleep et al., 2004; Bach et al., 2006). Such minerals are common across the oceanic crust, and serpentinization is a global process, likely to occur on any wet rocky planet (Fyfe et al., 1994; Russell and Kanik, 2010). The water reacts: it physically binds to olivine, hydroxylating the mineral, thereby metamorphosing it into serpentinite, named after its mottled green colouration and sinuous appearance. The reduction of water by olivine releases hydrogen gas and hydroxide anions, via exothermic reactions, the heat of which drives the alkaline fluids back to the surface, where they emerge into the ocean as warm alkaline hydrothermal vents, with a pH of around 9 to 11 (Martin et al., 2008). Four billion years ago, such vents would have been the ideal hatcheries for life, as they provide essentially all the requirements needed for sustained abiotic chemistry and the beginnings of natural selection.

2. THE ORIGIN OF LIFE IN ALKALINE HYDROTHERMAL VENTS

Life is notoriously difficult to define, but much easier to describe. All life today is made of reduced carbon compounds, typically formed by reducing CO₂; all life uses enzymes to catalyse metabolism; all life replicates using DNA as a hereditary template; all life is composed of cells which concentrate molecules and discriminate between inside and outside; and all life is powered by proton gradients (occasionally sodium gradients) which are tapped by chemiosmotic coupling across a membrane.

Plainly life needs a thermodynamic driving force, which today is always redox chemistry: the transfer of electrons from donors to acceptors, glaringly absent in soup. The simplest such transfer is from H₂ to CO₂, to form organics such as methane and acetate. There are only five known primary pathways of carbon assimilation across all of life (Thauer, 2007). The simplest and cheapest in energetic terms is the acetyl CoA pathway in acetogens and methanogens, which runs all the way to pyruvate (Fuchs, 1989; Berg et al., 2010).

Everything else requires more ATP, or some other form of energy such as sunlight, to reduce CO₂, and is therefore less suited as a primordial pathway (Fuchs and Stupperich, 1985; Shock et al., 1998). Alkaline hydrothermal vents are the only major source of hydrogen on the planet (Sleep et al., 2004). As in the acetyl CoA pathway today, the energy released by the reaction of H₂ with CO₂ can drive the formation of thioesters (simpler acetyl thioesters like acetyl methyl sulfide perhaps substituting for acetyl CoA in the primordial pathways) and acetyl phosphate (Martin and Russell, 2007). In principle such energy-rich compounds could drive intermediary metabolism, such as the reverse Krebs cycle and amino acid formation; and indeed acetyl phosphate is still used alongside ATP in many prokaryotes today (Wolfe, 2010).

Then life requires catalysis. In the strongest version of the RNA world, the first catalysts were RNA molecules, which raises the serious question of where all the RNA comes from. A weaker version of the RNA world allows for RNA catalysis, but not as the only catalyst. Minerals containing transition metals, such as Fe-S clusters, are found even today at the active sites of critical enzymes, including respiratory complexes and hydrogenase enzymes (Russell et al., 2008). Many of these are identical to mineral structures found in alkaline vents, including almost all the enzymes of the acetyl CoA pathway in methanogens and acetogens, giving these pathways "rocky roots" (Russell and Martin, 2004). These minerals, by catalysing the formation of organics that enter into intermediary metabolism (driven by phosphorylation), give rise to a richer catalytic spectrum of amino acids, polypeptides, nucleotides and simple cofactors, many of which are nucleotides (Copley et al., 2007; Yarus, 2010). Again, the continuous flow of hydrogen and reduced carbon compounds should drive these reactions forward, with the catalysts providing direction (the beginnings of later metabolic pathways) as much as speed (Copley et al., 2007).

So, alkaline vents provide the thermodynamic driving force, abundant raw materials, sustained over scores of millennia, and the catalysts needed for each step in the same form as modern enzymes, if initially without their protein matrix. (Proteins speed up reactions enormously; but it is the cofactor or mineral cluster that determines the nature of the reaction, and therefore the primordial metabolic pathways; Copley et al., 2007.) They also provide reduced nitrogen, such as ammonia (Martin and Russell, 2007). Phosphate is stable in alkaline conditions, likely bound to Mg^2+M (Mellersh and Smith, 2010). Much pioneering research on abiotic chemistry, envisaged to occur beneath a hypothetical reducing atmosphere (Saladino et al., 2008; Powner et al., 2009; Costanzo et al., 2009) applies equally to alkaline vents, except that the steady supply of reduced precursors is real and measurable, even in modern vent systems like Lost City (Proskurowski et al, 2008; Bradley et al., 2009). The details are beyond the scope of this paper, but are discussed in other articles in this edition of the Journal of Cosmology.

Life is cellular. Without restricted, enclosed spaces, it is not possible to concentrate the monomers needed for protein and nucleic acid polymerization; they simply disperse into the oceans at far too low a concentration to be any use. Alkaline vents are a honeycomb of interconnected micropores, on roughly a cellular scale, through which reduced fluids percolate, their slow circulation driven by convection currents and thermal diffusion (Russell and Hall, 1997; Martin and Russell, 2003; Baaske et al., 2007). In modern vent systems such as Lost City, the pores are enclosed by delicate aragonite walls (Kelley et al., 2001), but four billion years ago the walls would have been catalytic, composed of bubbly membranous iron-sulfur minerals, perhaps mingling with aragonite (Russell and Kanik, 2010). Today, the oceans are oxygenated, so iron rapidly precipitates out as rusty sediments. In the past, this process occurred on a vast scale to form banded iron formations, or BIFs (Lane, 2002). But four billion years ago, BIFs were all in the future, and the oceans were saturated with ferrous iron (hence iron was available at those primordial vents delivered from the Hadean Ocean, but it is not today). Sulfide could have been carried to the vents in the alkaline solution (Mielke et al., 2010). Fischer-Tropsch syntheses take place in modern vents (McCollom et al., 1999; Konn et al., 2009) and would certainly have been more common if catalysed by transition metals like iron. Such reactions are likely to have coated the pore walls with hydrophobic hydrocarbons, giving rise to the first rudimentary, and probably fragmentary, membranes, which did not hinder flow, but did foster the beginnings of membrane chemistry (Lane et al., 2010).

Such microporous honeycombs have been shown to concentrate single nucleotides and oligomers of RNA and DNA up to thousands of times their starting concentration (Baaske et al., 2007; Budin et al., 2009; Mast and Braun, 2010), a unique system making an RNA world plausible for the first time (or specifically, a weak form in which RNA functions as both catalyst and replicator, but not alone). Cyclic nucleotides have been shown to spontaneously polymerise into RNA in aqueous solution, when at high concentration and temperatures of around 80oC (Costanzo et al., 2009) - conditions that practically define alkaline vents, but no other setting. Interestingly, modern RNA polymerase is an Mg²⁺ enzyme (Zaychikov et al, 1997), which again brings to mind relatively cool vent fluids, bearing Mg²⁺ dissolved from olivine. It is not implausible that Mg²⁺, chelated by self-selecting amino acids, might exhibit some RNA polymerase activity under vent conditions. The catalytic action of many ribozymes is also dependent on Mg²⁺ (Uchimaru et al., 1993).

Replication is the next step. Thermal currents in hydrothermal vents simulate thermal PCR (Baaske et al., 2007): double-stranded DNA melts at higher temperatures, and anneals at lower temperatures. In the presence of a polymerase enzyme, around 1000 cycles of replication take place in an experimental vent system in a single day (Mast and Braun, 2010).

Equally important, the honeycomb cells in alkaline vents provide an ideal setting for the beginnings of natural selection (Koonin and Martin, 2006). Selection for replication speed of RNA invariably generates small RNAs capable of replicating at high speed, as Spiegelman showed with his "monster" (Oehlenschläger and Eigen, 1997). The theoretical question of how selection begins to act on encoded properties, such as suites of metabolic traits, rather than on replication speed alone, is perhaps the first major transition in evolution. Most models suggest that selection for metabolic traits, as in Eigen's hypercycle model (Eigen and Schuster, 1977), and Szathmáry's metabolic replicator model (Száthmary and Demeter, 1987) is quickly overwhelmed by faster-replicating parasites, and the system swiftly collapses (Branciamore et al., 2009). The only way that selection can act at the level of metabolism is through selection for a higher entity like a cell, which excludes parasitic RNA. The magical appearance of cells, of course, brings its own well-known problems; but Szathmáry has recently shown that compartmentalization per se – specifically the honeycomb pores of alkaline vents – provides a way of temporarily excluding the parasitic fast-replicating RNA, so enabling the group selection of metabolic traits (Branciamore et al., 2009). The origin of genomes in alkaline vents has been addressed in detail by Koonin and Martin (2006). Thus, alkaline vents potentially solve the origin of selection for cells and metabolism, as well as the origin of the replicators themselves.

For these reasons alone, alkaline hydrothermal vent systems should be seen as not merely the most promising setting for the origin of life, but as the only model that makes the emergence of life look like a probable and deterministic outcome of geology, geochemistry and thermodynamics (Table 1). Alkaline vents are electrochemical reactors that provide integrity, concentration, catalysis, replication and a suitable environment for selection. It is hard to imagine life not emerging in such a system, especially if one pictures the oceanic crust as host to practically contiguous vent systems, as might indeed have been the case (Russell and Arndt, 2005). One could even imagine a natural selection of vents, in which life emerges from the system with the best balance of H₂, flow and catalysis, perhaps going on to infect nearby systems. But the most compelling evidence for the singular importance of alkaline vents is the role of chemiosmotic coupling in life today.

3. THE ORIGIN OF THE PROTON-MOTIVE FORCE

All life is chemiosmotic. This is shocking and highly significant. The term "chemiosmotic coupling" refers to the mechanism of energy transduction in cells. Electrons are passed from a donor to an acceptor, by way of electron-transport chains that act, in effect, like electrical wires. The flow of electrons powers the transfer of protons across a membrane, to form a proton gradient over the membrane. The flow of protons through the ATPase – an amazing protein turbine – powers the synthesis of ATP. Thus the oxidation of an electron donor is coupled to ATP synthesis by way of a proton gradient. The concept was pioneered over 30 years by Peter Mitchell, against widespread resistance, for which he won the Nobel Prize in 1978 (Lane, 2005; Prebble and Webber, 2003). Since then, the exact mechanisms by which electron flow is coupled to proton transfer, and proton-motive force to ATP synthesis (Abrahams, et al., 1994) have been elucidated at atomic resolution, most recently in the remarkable piston action of complex I (Efremov et al., 2010). But the wider questions that originally motivated Mitchell himself have received surprisingly little attention.

Proton gradients are nearly universal, only occasionally being replaced by sodium gradients, which requisition the same apparatus (Lane et al., 2010). All three domains of life – bacteria, archaea and eukaryotes – use analogous systems (Lane, 2005). Both respiration and photosynthesis form ATP using proton gradients. The bacterial flagellum is proton powered, as is the movement of many solutes and metabolites in and out of the cell (Harold, 1986). Bacteriorhodopsin-based ATP synthesis in prokaryotes uses sunlight to translocate protons, via conformational changes in rhodopsin (Luecke et al., 1998). Even obligate fermenters, which do not use the proton gradient for energy transduction, still generate a proton gradient over the cell membrane to power membrane transporters (antiporters and symporters). ATP from fermentation drives the ATP synthase in reverse, to pump protons over the membrane (Harold, 1986). The fact that all cells use the same machinery, right down to the ATP synthase itself (which is homologous, if not identical, in archaea and bacteria; Mulkidjanian et al., 2007) means that chemiosmotic coupling is as universal as the universal code itself, a fundamental property of life on earth. But why?

Most models for the origin of life tack on chemiosmotic coupling as a peculiarity that just happens to need explaining somewhere down the line, rather than as a fundamental, perhaps even necessary, feature of life. Russell has long argued the significance of proton gradients for life, explaining their existence as a natural outcome of the origin of life in alkaline vents (Russell et al., 1993, 1994; Russell and Hall, 1997; Martin and Russell, 2007; Nitschke and Russell, 2009). The reason is simple: alkaline fluids, bubbling into a mildly acidic ocean (probably around pH 5 to 6, as CO₂ levels were as much as 1000 times higher four billion years ago than today) give rise to a natural proton gradient of 4 to 5 pH units, equating to a difference in proton concentration of about 10,000-fold across the vent walls. Such a proton gradient is of a similar magnitude (giving a membrane potential of around 200 mV) and polarity (outside acidic) to the proton gradients used by cells today (Russell et al., 1993, 1994; Lane et al., 2010; Figure 2). Plainly it is easier to tap a natural, pre-existing gradient than to generate one from scratch; indeed it is hard to imagine how, or why, life would have 'invented' such a profoundly counter-intuitive mechanism of energy generation, had proton gradients not been a pervasive aspect of the environment.

The details of how proton gradients might be tapped in a vent system are beyond the scope of this article, and have been discussed elsewhere (Nitschke and Russell, 2009; Lane et al., 2010). They are amenable to experiment. In terms of the probability of life, the question is this: did chemiosmotic coupling arise just because it could (a fluke aided by the environment) or because it must? Martin and Russell (2007) have demonstrated, on the basis of bioenergetics, that chemiosmotic coupling is in fact strictly necessary. Despite being the only pathway able to drive carbon and energy metabolism without an input of energy from anywhere else, the acetyl CoA pathway nonetheless depends on chemiosmotic coupling (Thauer et al., 2008). Substrate-level phosphorylation (the direct transfer of phosphate from one moiety to another) does not provide enough energy for growth in either methanogens or acetogens (Martin and Russell, 2007; Thauer et al., 2008). The reaction of H₂ with CO₂, despite being exergonic overall, requires surmounting a low kinetic barrier to form a methyl group (Martin and Russell, 2007; Nitschke and Russell, 2009; Lane et al., 2010). In vents, the critical intermediates are provided free of charge. Outside the vents, however, cells must pay themselves. Using substrate-level phosphorylation, cells must spend one ATP to gain one ATP, which precludes the possibility of growth. This, in fact, is a fundamental problem with chemistry, and boils down to a startling proposition: life cannot evolve by chemistry alone. The secret of chemiosmotic coupling is that it is not chemistry (Lane, 2005). Just as the DNA code enables information to replace chemistry, so too chemiosmotic coupling enables metabolism to escape the bounds of chemistry.

Chemistry is the reaction of molecules with each other, and can be defined by stoichiometry. It is not possible for one molecule to react with half a molecule, so chemical equations must be balanced to give whole numbers, to the chagrin of students. One of the molecular peculiarities of respiration, which foreshadowed the chemiosmotic hypothesis in the first place, is that it is not stoichiometric: the number of electrons transferred to oxygen does not correspond in a simple way to the amount of ATP synthesised. The interpolation of Mitchell's proton gradient explained all that. A gradient, by its nature, is composed of gradations, and so is not stoichiometric. A gradient, in effect, disconnects the exergonic reaction (which can be repeated indefinitely to generate a proton gradient) from ATP synthesis. Thus a reaction that releases enough energy to synthesise 1.5 ATPs is worthless in terms of stoichiometric chemistry, but pays off with a gradient. The 'small change' can be saved as protons towards the next ATP. Net growth is now possible. The remarkable conclusion is that life is obliged to be chemiosmotic. The fact that alkaline hydrothermal vents are naturally chemiosmotic is the defining reason to view them as the ideal incubator of life. One is tempted to say that life could not have arisen elsewhere, because no other environment provides dynamic proton (or Na+) gradients for 'free'.

This should be qualified a little. Photosynthesis overcomes these problems, as it taps the energy of the sun. Chemiosmotic coupling is not strictly necessary, even if still useful. But photosynthesis is altogether more complex, and in terms of how life works today, sunlight could not have been the primordial source of energy. Likewise, aerobic reactions (such as H₂S with O₂) do produce enough energy for growth; but as already noted, there was little or no O₂ before oxygenic photosynthesis. Fermentation of primordial soup could also, in principle, provide sufficient energy, as the organic molecules in soup come 'free', formed, in theory, by UV radiation acting on methane and ammonia. But free soup is not a bargain if the pot is empty; and the absence of a thermodynamic driving force means that life did not arise by fermentation, as discussed earlier. So within this more restricted ambit, the origin of life by chemolithotrophy in anaerobic oceans is only possible with chemiosmotic coupling. The first prokaryotes that emerged from vents were necessarily chemiosmotic. But this mechanism, while freeing life from vents, also conceals hidden costs. These costs explain why life runs into a big problem getting past bacteria.

4. THE SINGULAR ORIGIN OF COMPLEX LIFE

Complex life arose only once on Earth in four billion years. This statement requires defending before continuing. The argument is based on three points: (i) all complex life is eukaryotic; (ii) prokaryotes show no tendency to evolve greater complexity; and (iii) intermediate forms exist, proving that the niche is viable (refuting the argument from extinction); but all these intermediate forms evolved by reductive evolution from more complex ancestors, rather than accruing greater complexity by point mutations or endosymbioses in prokaryotic populations (Lane and Martin, 2010).

All morphologically complex life on Earth is composed of eukaryotic cells; and all eukaryotes share a common ancestor. The nature of this common ancestor is inferred from the traits shared by all the major supergroups of eukaryotes. All, by definition, share the nucleus with its characteristic double membrane and large nuclear pore complexes. All have straight chromosomes with centromeres and telomeres; all share detailed chromatin and chromosomal structures, right down to the same introns in the same place in the same genes; all share dynamic cytoskeletons, endoplasmic reticulum, Golgi bodies, lysosomes, mitochondria (all those that do not have mitochondria today, had them in the past; see discussion below) and so on. All share dynamic processes like mitosis and meiotic sex. On average eukaryotic cells are much larger than prokaryotes (4 to 5 orders of magnitude) and have much larger genomes (on average 2 to 3 orders of magnitude, but larger eukaryotic genomes range up to 100,000 Mb, compared with just 13 Mb in cyanobacteria). In short, the last common ancestor of all eukaryotes was very different to any known prokaryote, whether bacterium or archaeon (Lane and Martin, 2010).

An interesting theoretical question is how the common ancestor of eukaryotes came to share so many traits (assuming that they cannot all have evolved simultaneously). It is likely that only meiotic sex in a small population of mating cells could have given rise to the last eukaryotic ancestor. While lateral gene transfer is common in bacteria and archaea, forming a network rather than a pure tree of life (Dagan and Martin, 2007, 2009), lateral gene transfer is non-reciprocal and so leads to greater variation, rather than less variation, in a population. Being reciprocal, sex does the opposite: it concentrates all shared properties within a single interbreeding population, restricting variance within the population, but increasing variance between populations (speciation). If the eukaryotic common ancestor had evolved by binary fission and lateral gene transfer, in bacterial fashion, it is likely that some eukaryotes would have a nucleus and others not; some would have an endoplasmic reticulum or mitochondria, others not, and so on. The very fact that all eukaryotes share so many fundamental traits almost certainly means that they arose in a small and rapidly evolving sexual population (Lane, 2009). That population had to be small, because a larger, geographically more structured population would inevitably diverge into different species, with different traits, as happened soon afterwards in the eukaryotic supergroups. The population had to evolve rapidly for much the same reasons: a leisurely pace of evolution would tend to promote branching and diversification through the dispersal and geographical structuring of stable populations. A high mutation rate in the earliest eukaryotes (caused by an early bombardment of mitochondrial genes and retrotransposons; Martin and Koonin, 2006) explains both the requirement for sex and the speed of evolution (Lane, 2009). No other mode of evolution could account easily for the large number of shared traits in the last eukaryotic ancestor.

In contrast, bacteria show no tendency to evolve greater complexity. Or rather, more perplexingly, they show a little tendency to evolve greater complexity, but not much. In essence, bacteria have made a start up every avenue of eukaryotic complexity, but then stopped short. So, for example, prokaryotes have evolved structures resembling a nucleus, recombination, straight chromosomes, internal membranes, multiple replicons, giant size, extreme polyploidy, dynamic cytoskeletons, predation, parasitism, introns and exons, intercellular signalling (quorum sensing) and even endosymbionts (Lane, 2007; Lane and Martin, 2010). What stopped them evolving true complexity? There is a serious question here about the nature of natural selection itself. In what way is the nucleus or phagocytosis different from an eye? In the case of the eye, the evolution of a fully functional image-forming eye can be modelled as a succession of steps from a light-sensitive spot, each step offering a small selective advantage over the last (Nilsson and Pelger, 1994; Figure 3). Not surprisingly, eyes have evolved essentially independently on scores of occasions in metazoans from a light-sensitive spot, albeit under the control of a small group of highly conserved regulatory genes such as Pax-6, which nonetheless independently recruited completely different sets of genes to construct a wide range of morphologically disparate eyes, from the mirror eyes of scallops to the convergent camera eyes of molluscs and mammals (Lane, 2009, Vopalensky and Kozmik, 2009)

But if each step to evolve the nucleus or phagocytosis also offers a slight selective advantage, then why don't such traits evolve repeatedly in prokaryotes? The trilobite eye probably evolved in a few million years; bacteria have had four billion years to evolve phagocytosis, and yet have apparently never done so, despite having a dynamic cytoskeleton (Vats et al., 2009), frequently losing the cell wall (Allan et al. 2009), and evolving internal compartments (for example in planctomycetes; Lindsay et al., 2001). The only reasonable conclusion is that natural selection does not tend to drive the evolution of complexity in prokaryotes, presumably because selection for fast replication and genomic streamlining invariably favours small and morphologically simply cells (Lane, 2005, 2007).

One alternative, frequently cited possibility, is that more complex prokaryotes, or proto-eukaryotes, arise regularly but are out-competed by the eukaryotes that already exist, and so are always driven to extinction, leaving no trace in the (imperfect) fossil record. According to this view, the eukaryotes emerged from some kind of environmental bottleneck (such as rising oxygen levels) and then drove any later arrivals, and for that matter their own more primitive ancestors, to extinction (de Duve, 2005, 2007; Kurland et al, 2006; Poole and Penny, 2007). Extinction is taken to be the fate of virtually every species anyway, so what could be more natural? In this reading it is true that all complex life is eukaryotic, and that eukaryotes arose only once; but in principle that says nothing about probability, because complex life could emerge repeatedly, with high probability, but when it does it is always outcompeted by the highly-adapted eukaryotes that already occupy the niche. The fact that mass extinctions often lead to evolutionary radiations, at least among plants and animals, seems to justify this view.

But this argument will not do. First, it is dubious whether any microbial group has ever fallen extinct. Methanogens and acetogens, energetically the most tenuous forms of life, have survived for nearly four billion years (Ueno et al., 2006). The continuity of the global geochemical cycles over that time – the sulfur cycle, nitrogen cycle, methanogenesis, and so on – attest to the fact that these microbial niches have been continuously occupied for nearly three billion years (Canfield et al., 2006). Thus, there is no evidence for extinction in the geochemical record, and the onus ought to be on those who invoke extinction to provide evidence. Second, niches are frequently shared by microbes, including eukaryotes. The tiny pico-eukaryotes, for example, discovered a decade ago (Moreira and Lopez-Garcia, 2002), compete on essentially prokaryotic terms with prokaryotes, yet are abundant (Piganeau, et al., 2008), not extinct as would surely be predicted by the bottleneck hypothesis. Likewise, the oceans are filled with mesophilic archaea that live alongside bacteria in the same niche; neither group shows any sign of extinction (Prosser and Nicol, 2008). Supposedly mesophilic eukaryotes are often found in extreme environments, once thought to be the sole preserve of extremophile archaea and bacteria, such as the Rio Tinto in Spain, with extreme acidity and heavy metal levels (Amaral Zetler et al., 2002), and the abysmal Mediterranean anoxic hypersaline lakes, such as L'Atlante basin (Danovaro et al., 2010). The biodiversity of life itself points to the ability of species, especially microbes, to coexist within ecosystems.

Third, and more specifically, there are more than 1000 known species of protist that do not have mitochondria: they occupy the intermediate complexity niche between prokaryotes and eukaryotes, a group dubbed 'archezoa' by Cavalier-Smith in the 1980s and taken to be exactly that, primitive eukaryotes that had not yet acquired mitochondria, or much of the rest of the eukaryotic paraphernalia, beyond the nucleus (Cavalier-Smith, 1987). Over the last 15 years or so, all of these groups turned out to have possessed mitochondria in the past, which later became specialised to anaerobic environments as ‘relict’ organelles such as hydrogenosomes or mitosomes (Tielens et al., 2002; Embley and Martin, 2006; van der Giezen, 2009). The general consensus is that, if there ever was a primitive amitochondriate eukaryote, an 'authentic' archezoon, then it has fallen extinct. But in light of the arguments above, why should it have fallen extinct? Because it was outcompeted to extinction in anaerobic environments by eukaryotic cells that are exactly the same (which is to say they are archezoa, if not primitively so)? If that were the case, the archezoa must have been outcompeted in anaerobic environments by eukaryotes with mitochondria that (according to the archezoa hypothesis) had previously been adapted to aerobic environments (Cavalier-Smith, 2002; 2009), hence were presumably at a disadvantage in these ancient and stable anaerobic environments.

None of this sounds plausible. Because anaerobic eukaryotes are widely distributed across all six eukaryotic supergroups (Mentel and Martin, 2010), they most likely descend from the last eukaryotic common ancestor, which probably arose in the anaerobic 'Canfield' oceans 1 to 2 billion years ago (Knoll et al., 2006). Anaerobic environments have been continuously occupied by bacteria and presumably eukaryotes throughout that period (Lyons et al., 2009; Falkowski et al., 2008), so there has been plenty of time for genuinely primitive amitochondriate eukaryotes – real archezoa –to evolve repeatedly. Yet every single example evolved by reductive evolution from more complex ancestors that once had mitochondria, whereas not one single example evolved greater complexity by either point mutation or other primary endosymbiotic events in large populations of prokaryotes already well adapted to anaerobic environments. It is said to be unreasonable to conduct a statistical analysis of eukaryotic origins, because they only evolved once; but if chance were the only factor, from a purely statistical point of view the probability of all 1000+ species of archezoa being derived by reductive evolution from more complex ancestors, rather than simply evolving just a little more complexity, is about 1 in 10³⁰⁰ against. That constitutes good evidence that eukaryotic origins were genuinely improbable – a singular event – but once that threshold had been crossed, the niche was invaded repeatedly by reductive evolution from more complex eukaryotic ancestors. Chemiosmotic constraints explain why.

5. WHY ENDOSYMBIOSIS IN PROKARYOTES IS NECESSARY FOR COMPLEXITY

Prokaryotes respire over their plasma membrane, which is to say they generate a proton gradient over the plasma membrane. This means that prokaryotes face constraints in their surface-area-to-volume ratio. Double the dimensions of a bacterium, and the surface-area-to-volume ratio halves. Thus, the capacity for ATP synthesis relative to protein synthesis falls with increasing cell volume, placing large bacteria at a bioenergetic disadvantage. Most prokaryotes are indeed small, on average 4 or 5 orders of magnitude smaller than eukaryotic cells. If a spherical prokaryote were scaled up to the same size as an average eukaryotic cell, while continuing to respire over its plasma membrane, its energetic efficiency would fall off by around 40 to 50-fold. The actual difference is in fact several orders of magnitude, as discussed in detail elsewhere (Lane and Martin, 2010), because eukaryotes do not respire over their plasma membrane, but internally in mitochondria. Plainly, unless prokaryotes can find some way to adapt, larger cells will be penalised by selection.

To a point, prokaryotes do adapt to offset the disadvantages of size. For example, rods (bacilli) have a larger surface-area-to-volume ratio than spheres (cocci) so the larger the bacterium, the less likely it is to be spherical. Likewise, the density of respiratory complexes could be increased, and the rate-constants of respiratory enzymes increased. To a point this happens: on average, the metabolic rate per gram is faster in bacteria than in eukaryotes (Lane and Martin, 2010; Makarieva et al., 2005; Fenchal and Finlay, 1983). But the fact that the gap is not great (less than an order of magnitude) implies that other constraints, such as protein crowding and loss of membrane fluidity, offset the respiratory advantages. A third possibility, by far the most revealing, is simply to internalise respiration on intracellular membranes. This gets around the problem in essentially the same way as eukaryotes. Some bacteria do exactly this, including cyanobacteria and many nitrifying bacteria: their bioenergetic membranes are extensively invaginated within circumscribed areas of the cell (Pinevich, 1997). Nonetheless, even this approach appears to be limited in scale. Most prokaryotes remain small, even those with extensive invaginations of bioenergetic membranes. There would appear to be some other limiting factor.

This limitation is not to be found in something that prokaryotes do not have, but in something that eukaryotes do have: mitochondrial DNA. Mitochondrial respiration is bacterial, not surprisingly, as mitochondria were derived from bacteria (Sagan, 1967). Mitochondria almost always retain a small genome, the sequence and characteristics of which were instrumental in establishing the bacterial ancestry of mitochondria (Gray et al., 1999). Specifically, all eukaryotes that retain the ability to generate ATP by chemiosmotic coupling have retained the same small group of integral membrane proteins involved in coupling electron flow to proton pumping (Gray et al., 1999; 2004). Each eukaryotic supergroup has independently lost the vast majority of its ancestral genome, leaving each with an identical core set of genes encoding respiratory subunits (as well as a variable group of other genes in some cases). Conversely, if the entire genome has gone – usually the case in hydrogenosomes and mitosomes – the cell correspondingly loses the power to generate ATP by chemiosmotic coupling (van der Giezen, 2009). A simple and powerful hypothesis, long propounded by John F. Allen, is that mitochondrial genes are needed for the control of respiration – specifically for redox regulation of gene expression adjacent to the bioenergetic membranes: Co-localization for Redox Regulation, known as the CORR hypothesis (Allen, 1993; 2003). But regardless of the exact reasons for their retention, the same core group of respiratory genes would hardly be retained in all eukaryotes capable of chemiosmotic coupling by chance alone. It is a fair bet that they are necessary for respiration. Conversely, without these core genes, a large cell would lose control of respiration.

There is strong evidence that the rate of cell respiration does indeed depend on the copy number of mitochondrial DNA, hence the respiratory efficiency of eukaryotes depends on mitochondrial genes (Rocher et al., 2008). There is also strong evidence, as discussed already, that all eukaryotes either have mitochondria, or had them in the past and later lost them by reductive evolution (van der Giezen, 2009). Thus, the last common ancestor of all eukaryotes had mitochondria, making the origin of mitochondria and the origin of the eukaryotic cell plausibly one and the same event. Several rigorous, large-scale full genome studies support this position (Cox et al., 2008; Yutin et al., 2008; Pissani et al., 2007; Rivera and Lake, 2004) – eukaryotes are genetically chimaeras, with most informational genes (those that control gene expression) deriving from archaea, while operational genes (those that control metabolic flux) mostly derive from bacteria, the majority, presumably, from the ancestor of the mitochondria. The simplest explanation consistent with all of these findings is that the eukaryotic cell originated in an endosymbiosis between an archeaon and a bacterium, as proposed by William Martin and Miklos Müller in the hydrogen hypothesis (Martin and Müller, 1998), which has recently gained support from several studies (Atteia et al., 2009; Fritz-Laylin et al., 2010; Danovaro et al., 2010; Lane, 2010). The acquisition of mitochondria transformed the selection pressures acting on the host cell, enabling it to expand in cell size, and significantly in genome size, without a bioenergetic penalty (Lane, 2005; 2007; Lane and Martin, 2010). Prokaryotes are unable to co-localize the correct core contingent of genes alongside their internal bioenergetic membranes, and therefore cannot meet the large energy costs of any significant expansion in genome size. The detailed reasons for this are discussed elsewhere (Lane and Martin, 2010).

An endosymbiosis in prokaryotes reconciles virtually all the conflicting data about the origin of the eukaryotic cell, and more generally, complex life on Earth. It explains why the origin of eukaryotes was a singular event – endosymbioses in prokaryotes, which cannot phagocytose, are vanishingly rare occurrences, albeit one or two cases are known (Wujek, 1979; von Dohlen et al., 2001). These remain as endosymbionts rather than organelles, so plainly the conversion of an endosymbiont into an organelle is also problematic. If the evolution of complexity depends on a rare and stochastic event, that explains why natural selection does not continuously give rise to eukaryotic complexity in prokaryotic populations. The unique acquisition of mitochondria by a prokaryote transformed the course of evolution. By permitting cells to expand their cell volume and genome size without a crippling energetic penalty, eukaryotes could, and did, accumulate genes, ultimately the basis of complexity (Lane and Martin, 2010). It makes sense that prokaryotes are almost universally small and genetically streamlined, showing little tendency to evolve eukaryotic complexity (Lane, 2005, 2007). It makes sense that multicellular eukaryotes only arose once in 4 billion years of evolution. And it makes sense that the acquisition of mitochondria released the first eukaryotes from the energetic constraints that limit all prokaryotes, enabling the evolution of genetically and morphologically complex life on Earth (Lane and Martin, 2010).

6. CONCLUSION

This bioenergetic perspective provides an insight into the evolution of life, not just on Earth but also elsewhere in the universe. The origin of life is theoretically probable on any wet rocky planet, as the common geological process of serpentinization is predicted to give rise to alkaline vents on a global scale. Alkaline vents provide an ideal incubator for life, meeting all the conditions needed for life to evolve – specifically a thermodynamic driving force, stability, a steady supply of hydrogen, CO₂ and reduced nitrogen, the catalysts needed to drive primordial metabolism, a means of concentration in cell-like micropores, and natural proton gradients enabling the evolution of chemiosmotic coupling, the only means of powering free-living chemolithotrophic cells in anaerobic oceans. But even if the origin of prokaryotes is probable on any wet rocky planet, the very fact that prokaryotes necessarily depend on chemiosmotic coupling constrains evolution. Only a rare and stochastic endosymbiosis can free prokaryotes from the energetic constraints of respiring across the plasma membrane. On Earth this was apparently a singular event, and prokaryotes otherwise show little tendency to evolve greater complexity. The same constraints must face life on other planets too, because chemistry, and the principles of thermodynamics and energetics, are universal. The unavoidable conclusion is that the universe should be full of bacteria, but more complex life will be rare – and perhaps surprisingly similar to us. Aliens will have mitochondria too.