![[prime]](/stockgif3/prime-script.gif)
OH), essential for ribozyme activity. The study of the origin of life covers many areas of expertise and requires the input of various scientific communities. In recent years, this research field has often been viewed as part of a broader agenda under the name of “exobiology” or “astrobiology.” In this review, we have somewhat narrowed this agenda, focusing on the origin of modern terrestrial life. The adjective “modern” here means that we did not speculate on different forms of life that could have possibly appeared on our planet, but instead focus on the existing forms (cells and viruses). We try to briefly present the state of the art about alternative hypotheses discussing not only the origin of life per se, but also how life evolved to produce the modern biosphere through a succession of steps that we would like to characterize as much as possible. ©2007 HFSP Publishing
Traditionally, two approaches have been employed to understand how terrestrial life originated (Fig. 1). The bottom-up approach, exemplified by Miller's experiment, attempts to reconstruct the conditions of the primitive Earth in order to imagine how the main components of living organisms came into being. This is the realm of astrophysics, geophysics and chemists. The top-down approach is favored by biologists, who try finding in modern organisms the relics of their ancestors in order to reconstruct ancient metabolic pathways and molecular processes. Neither of these two approaches can be successful alone, and the final goal of any “origin-of-life” program should be to bring together all these lines of research to build up a coherent scenario leading from inorganic chemistry to Darwinian evolution. In that sense, the quest of our origin is intrinsically interdisciplinary and should bring together various expertises to deal with the same issues.
Figure 1. Despite the difficulty of the topic, great advances have been made during the last decade in understanding the origin of modern life. A major issue that remains to be solved is the origin of RNA, since this is where the bottom-up and top-down approaches meet. We definitely know, from the resolution of the ribosome structure, that modern proteins were “invented” by RNA (Steitz and Moore, 2003). This means that, once upon a time, RNA was the master of life, covering both the genetic and catalytic properties today performed by DNA and proteins, respectively. However, the formation of a bona fide ribonucleotide has so far never been successfully achieved in the laboratory, and the formation of oligoribonucleotides from monomers is extremely difficult to achieve. In this review, keeping in mind that the origin of RNA is the central issue, we will briefly review the state of the art and the recent controversies in the fields, and we will try to identify the most promising areas of research for the next decade.
Plausible mechanisms for the formation of the solar system have now been formulated, especially explaining the accretion mechanism that could have led to the formation of a terrestrial-like planet (Montmerle et al., 2006). The formation of Earth is quite precisely dated at 4 56 Ga ago, based on the dating of a particular type of meteorites called “ordinary chondrites.” The accretion mechanism was probably rapid (about 100 Myr), leading in a first time to a very hot planet with a magma ocean. The formation of oceans and continents took place probably more rapidly than previously thought (between 4.5 and 4.4 Ga) (Hawkesworth and Kemp, 2006). This is inferred from the study of the oldest rock, a 4.4 Ga old zircon from Australia that gives evidence for an interaction between water and rock at temperatures below 100 °C (Wilde et al., 2001). An atmosphere would also have formed quite early from volatile elements (such as nitrogen) contributed by extraterrestrial material on the surface of the Earth. Astrophysics has taught us that life is not alien to the universe, since its fundamental fabric—organic chemistry—is a ubiquitous component of the interstellar space. Complex organic molecules, as well as silicates, hydrocarbons, and various forms of ice have been found in extrasolar clouds (Bernstein, 2006). Therefore, as temperature decreased, organics, either produced on Earth or coming from meteorites or micrometeorites (cosmic dust), could have started accumulating on the surface. For some authors, the conditions for the emergence of life (liquid water, continental crust, atmosphere) were already in place at 4.4–4.3 Ga. However, the habitability of the early Earth was seriously compromised by multiple giant impacts. In particular, around 3.9 Ga the Earth was subjected to an impressive episode of bombardment, called the late heavy bombardment (LHB) (Cohen et al., 2000).
The craters observed on the surface of the Moon and other planets whose surface was not remodeled by erosion, sedimentation, and plate tectonics (Mars, Venus) testify for the diameter of the giant meteorites (more than 100 km and up to 5000 km) that hit the Earth's surface during the LHB [for a recent review, see (Claeys and Morbidelli, 2006)]. This dramatic event could have been triggered by the migration of giant planets that took place after the dissipation of the gaseous circumsolar nebula (Gomes et al., 2005). The LBH may have lasted from 20 to 200 million years, with a frequency of impact that is highly debated (from one each 10,000 years to one every 20 years). Models predict that such impacts would have almost completely resurfaced our planet, leading to evaporation of the oceans, melting of the crust down to at least 1000 ms, and loss of the atmosphere. It might be significant that the oldest terrestrial continental crust (Isua, Greenland) dates exactly to the end of the LHB, at 3.8 Ga. In our opinion, it is unlikely that any forms of life, if already present, would have survived the devastating impacts of the LHB. If this view is correct, it implies that the path to modern life would have (re)started after 3.9–3.8 Ga. The presence of sedimentary rocks testifies that oceans had already reformed by that time. However, putative isotopic traces of life found in these rocks are now believed to be artifactual (see below), consistent with the idea that modern life might have indeed originated after the LHB.
It has been accepted for a long time that the atmosphere of the early Archaean was anoxic and probably weakly reducing, and dominated by oxidative species such as CO2, N2, CO, and H2O, with small amounts of H2, that would have escaped rapidly to the outer space (Kasting, 1993). Reduced gases supplied by volcanic outgassing, such as CH4 and NH3, would have been destroyed by UV (photodissociation), and may have subsisted only locally around hydrothermal vents. However, a recent theoretical model has estimated that the hydrogen escape rates were lower than previously assumed in the early archaean atmosphere, suggesting that hydrogen may have been abundant (Tian et al., 2005). This would be good news for models in which life originated at the surface of our planet, since a reducing atmosphere would have favored “traditional” prebiotic chemistry. However, these recent estimations have already been criticized (Catling, 2006), and the debate is ongoing. It was noticed early on that the early Earth was in danger of freezing due to the low luminosity of the Sun, which was about 30% less than it is today (the “faint young Sun” paradox) (Sagan and Chyba, 1997). Several authors have suggested that high CO2 concentrations (or a mixture of CO2 and CH4) in the early atmosphere were required to prevent (via a greenhouse effect) Earth from freezing (Pavlov et al., 2000). Indeed, the presence of 3.5 Ga old sedimentary rocks excludes a global-scale glaciation of the planet at least by that time. The study of organic carbon isotopes indicates that oxygen concentrations became significant (but still very low) only at 2.7 Ga and then started to rise steadily (up to 1% of the present level) from 2.4 Ga, what is called the great oxidation Event (GOE) (Holland, 2006). Interestingly, this period coincides with two possible snowball Earth episodes around 2.9 and 2.4 Ga, which is assumed to have been triggered by the accumulation of biologically produced oxygen (and consequently the removal of methane and its greenhouse effect) following the emergence of oxygenic photosynthesis (Farquhar et al., 2000; Holland, 2006; Kasting and Ono, 2006). The isotopic fractionation of elements such as sulfur in archaean deposits points to an anoxic ocean during the whole archaean period and beyond, up to 1.8 Gyr. The oceans would have then gone through a euxinic phase (hydrogen-sulfide rich) and finally become fully oxygenated around 0.75 Gyr (Kump, 2005). Oxygen and silicon isotope data from archaean cherts indicate that ancient oceans may have been warmer than today, with temperatures as high as 70 °C around 3.3 Ga (Knauth, 1998; Robert and Chaussidon, 2006). However, the interpretation of isotopic data remains controversial since this would imply that archaean hot and acidic rainwater would have produced intense weathering that is not observed in the paleoweathering record. Furthermore, a hot ocean is difficult to reconcile with a first global glaciation that could have occurred at 2.9 and 2.4 Ga [for a critical review of these data, see (Kasting and Howard, 2006)].
The first and now popular descriptions of life traces in the Archaean regard layered structures very similar to present-day stromatolites from a 3.4 Ga old Australian Apex chert. These structures contain putative microfossils presenting morphological characteristics resembling present-day filamentous bacteria [for a review see (Schopf, 2006)]. However, their biologic nature remains hotly debated. For instance, it was shown that many of these structures are produced abiogenically in the laboratory under particular conditions [reviewed in (Brasier et al., 2006)]. Organic matter has been detected in these structures by in situ laser Raman spectroscopy (Schopf, 2006), although abiogenic structures also can absorb organic inclusions that give the typical Raman spectrum of a microfossil (Brasier et al., 2006). The earliest stromatolite formations of unambiguous biological origin thus remain for the time being those from around 2.6 Ga (Schopf, 2006). The question of the biogenic or abiogenic nature of earlier Archaean microfossils will have to await future methodological developments [for recent reviews see (Lopez-Garcia et al., 2006; Westall, 2005)].
The isotopic composition of different elements is affected by biological processes and can thus indicate the presence of particular metabolisms. Isotopic signatures of different elements (carbon, sulfur, nitrogen, and more recently iron) have therefore been extensively studied to search for life signatures in ancient rocks and to identify specific ancient metabolisms (Tice and Lowe, 2004) (Ueno et al., 2006). In particular, the carbon isotope values from apatites in Isua banded iron formations (3.8 Ga) have often been considered to be the earliest signatures of life on Earth (Mojzsis et al., 1996). However, all the data obtained remain vigorously debated (Fedo and Whitehouse, 2002; Mojzsis and Harrison, 2002). Some authors have argued, in particular, that some results are indeed compatible with an abiotic origin of isotopic composition from hydrothermal activity [for an extensive critical and well-balanced review on this topic, see Lollar and McCollom (2006)].
Finally, molecular fossils (kerogens) derived from the transformation of lipids have also been used to tentatively determine the age for the emergence of various life forms. However, it is very difficult to extract kerogens from Archaean rocks, and not all lipids are equally resistant. For example, lipids from archaea are very fragile and have not been found in rocks older than 1.8 Ga (Summons et al., 1988). The older biomarker record regards the presence of hopanes, lipids that today are distinctive of cyanobacteria, in 2.7 Ga old rocks from Australia (Brocks et al., 1999). The presence of eukaryotic-type steranes in the same ancient rocks (Brocks et al., 1999) is more controversial since some bacteria can produce sterols as well (Pearson et al., 2003; Tippelt et al., 1998), although not of the complexity of those found by Brocks et al. (Summons et al., 2006).
In conclusion, the fact that the oldest traces of life that are not controversial are only those from 2.6 Ga (Schopf, 2006) leaves open a wide window for the origin of modern life between 3.9 (end of the LHB) and 2.7 Ga. The quest for traces of life in this time interval is a rapidly expanding research field. New drilling projects have now started in order to obtain novel samples of archaean rocks. Isotopic and chemical techniques are being improved to detect the presence of organic matter with less ambiguity, and new in situ techniques start to be applied to the analysis of putative microfossils. Novel and more performing techniques of lipid extraction will hopefully push back the limit of detection of biomarkers to the early Archaean. In parallel, theoretical models for the early Earth will surely benefit from a better description of known metabolisms (see below) and metabolic consortia, and their current distribution in a wide range of environmental settings.
In the traditional “prebiotic soup” scenario, organic molecules would have first accumulated in the ocean or in smaller water bodies on the early Earth, either delivered by extraterrestrial sources (micrometeorites, dust) and/or produced by “Miller's type experiments” (especially if the early atmosphere was hydrogen rich, see above) (Bada and Lazcano, 2003). The first “living systems” would have then emerged from the gradual complexification of the prebiotic broth. The authors supporting this “heterotrophic theory” often argue that prebiotic chemistry is the prolongation on our planet of the cosmic chemistry, whose products (e.g., amino acids) indeed overlap with the building blocks of life. For them, the possibility to easily produce in prebiotic conditions simple amino acids, purines, sugars, fatty acids, and other small organic molecules essential to modern life is too striking to be fortuitous (de Duve, 2003). Proponents of the prebiotic soup scenario (especially the Bada and Miller school) have in general argued in favor of a slow (gradual accumulation) and cold origin of life (essential to the long-term stability of organic matter).
As an alternative to the heterotrophic theory, 20 years ago Wachtershauser proposed an autotrophic origin of life, in which an energy flux provided by chemical reactions at liquid–solid interfaces was used for carbon fixation (Wachtershauser, 1988) (Wachtershauser, 2006). A related model was proposed later on by Russell and Hall (1997). In this view, gradual accumulation and complexification of organic matter occurred either on mineral surfaces (i.e., a two-dimensional life) or in networks of mineral pores. Instead of linking cosmic chemistry with biochemistry, the proponents of an autotrophic origin of life try to link biochemistry with geochemistry. Wachterhauser specifically suggested that a primitive metabolism evolved at the surface of pyrite minerals from the reduction of carbon dioxide using hydrogen sulfide (H2S) over ferrous sulfide (FeS) as the reducing agent [“pioneer metabolism theory” (Wachtershauser, 1988) (Wachtershauser, 2006) and references therein]. The negatively charged organic molecules synthesized by this reaction would have been stabilized by binding to the positively charged pyrite surface, thus forming a two-dimensional network. The number and diversity of these molecules would have thus grown autocatalytically in situ by carbon fixation, leading to the self-organization of cyclic chemical reactions, producing more and more elaborated products. Russell and Hall (1997) suggested that carbon fixation first occurred inside mineral three-dimensional networks formed by the precipitation of iron monosulfide from the mixing of sulfide-rich hydrothermal fluid and the iron-containing water of an acidic ocean, the system being energetically driven by a naturally occurring geochemical pH gradient. The authors of autotrophic scenarios have been strongly influenced by the discovery of hydrothermal vents and hyperthermophiles in the late 1970s and early 1980s. In contrast to the proponents of the heterotrophic origin, they usually favor a hot origin of life, the initial reaction being driven by a geothermal energy source. In their models, the stability of organic molecules is no more an issue, since these would have been short lived. On the contrary, high temperature is supposed to have increased the rate of reactions at the surface of the minerals or inside mineral structures.
Although the autotrophic models for the origin of life are in theory experimentally realizable in toto (Huber and Wachtershauser, 2006), experimental programs designed to test these theories have succeeded up to now in producing only simple organic molecules (from C2 to C4). Furthermore, none of these reactions has been shown to be autocatalytic, a crucial requirement to start real chemical evolution (Orgel, 2000). The controversy between the proponents of heterotrophic and autotrophic theories thus remains lively (de Duve and Miller, 1991) (Bada et al., 2007). However, there is now a general agreement on the idea that minerals (especially clays) may have catalyzed prebiotic reactions and that metal sulfides have been an important source of electrons for the reduction of organic compounds (Bada and Lazcano, 2002). In particular, proponents of the heterotrophic theory now often agree that reactions occurring in a hydrothermal and/or in a volcanic setting may have enriched the prebiotic arsenal of organic molecules, or else suggest that the first organics useful for life were concentrated at mineral-water interfaces and/or into porous minerals. Volcanic activity might have been especially important for the production of phosphoric compounds that are essential for life (Yamagata et al., 1991) (Schwartz, 2006). Indeed, the first source of phosphate may have been polyphosphates, which are found in volcanic condensates and hydrothermal vents produced by volcanic activity (Yamagata et al., 1991). In order to reconcile the requirements of volcanic activity with an environment favoring molecular stability, it is tempting to suggest that life originated in an “Iceland-like” setting mixing ice and fire, in which a geothermal gradient could provide a stable and continuous energy source over long periods, whereas a cold environment could provide stability for the accumulation of organic molecules.
Both heterotrophic and autotrophic theories are faced with the problem of ending up with a robust protometabolism that could provide the energy and monomers to establish the RNA world (de Duve, 2003). In a first step, it is important to consider how to transfer the energy acquired either from the outside (heterotrophic theory) or from the reactions in hydrothermal fluids (autotrophic theory) for further elaboration of the system inside protocells. Ferry and House (2006) recently proposed an interesting model in which the energy obtained from a geothermal energy flux is coupled to the formation of phosphorylated compounds. This model combines both features of the autotrophic and heterotrophic theories since the mechanism of energy conservation resembles those of modern heterotrophs that metabolize reduced organic compounds for the synthesis of adenosine triphosphate (ATP) by substrate-level phosphorylation. A major question is indeed whether the protometabolism can be inferred from the metabolism of modern cells. The proponents of the heterotrophic scenario have often considered that the first organic molecules were produced by reactions completely independent from modern metabolism. In particular, Orgel argued that the metabolisms of the RNA world would have been completely erased by the emergence of a new metabolism based on proteinenzymes (Orgel, 2003). On the contrary, the proponents of the autotrophic theory tend to directly link the protometabolism to modern (hyperthermophilic) proteins via the coevolution of RNA and peptides. In fact, as suggested by de Duve (2003) a metabolism entirely sustained by RNA catalysts can also be linked to the modern one, if one reasons in terms of Darwinian evolution (de Duve, 2003) by assuming that a protein enzyme could have initially only replaced the function of an existing ribozyme (i.e., transformation of a given substrate into a given product). Similarly, if ribozymes themselves replaced the function of more ancient catalysts, the metabolism of the RNA cells could have been built upon the more ancient protometabolism, especially if the RNA world itself originated in the framework of Darwinian evolution between competing protocells.
Some authors have suggested that Darwinian evolution could have already occurred prior to the existence of cellular entities, through the competition of isolated supramolecular assemblies concentrated on mineral surfaces or inside mineral pores (Wachtershauser, 2006) (Russell and Hall, 1997). However, compelling theoretical and experimental arguments suggest that cell formation occurred early in life evolution [see for instance (de Duve, 2003; Deamer et al., 2006) (Muller, 2006) (Lopez-Garcia et al., 2006; Forterre, 2005)]. The formation of “protocells” was probably essential for the evolution of RNA replicators (see below) and the establishment of any sustained energy-driven protometabolism by (i) keeping together RNA replicators and their corresponding genomic RNAs (i.e., only catalysts enclosed by membranes can benefit from their own reaction), (ii) excluding potentially competing external parasitic RNAs, and (iii) preventing the dilution of molecules and macromolecules. Furthermore, a protometabolism able to synthesize nucleotides for RNA production would have also been able to produce simple (amphiphilic) molecules that are rather easy to synthesize prebiotically and could have been abundant on early Earth [see (Muller, 2006) and references therein]. Lipid vesicles can be produced quite easily in vitro from fatty acids or even better from fatty acid glycerol ester. These vesicles have the ability to undergo several cycles of growth and division (Hanczyc et al., 2003). Mineral surfaces, such as montmorillonite, also stimulate the formation of lipid vesicles (Hanczyc et al., 2007). Interestingly, mineral catalysts are trapped inside vesicles during this process, suggesting that interactions between fatty acids and minerals on early Earth may have resulted in the enclosure of diverse arrays of mineral particles with catalytic properties.
Most interestingly, Szostak and co-workers have recently shown that vesicles encapsulating RNA grow preferentially by lipid capture at the expense of empty vesicles (Chen et al., 2004; Chen and Szostak, 2004) (Fig. 2). This is explained by the higher osmotic pressure inside RNA-containing vesicles due to the counterions screening the negative charges of RNA. This osmotic pressure is counterbalanced by membrane tension, driving the uptake of fatty acids. At an early stage, this mechanism could have favored vesicles containing charged molecules, such as ribose phosphate and/or polyphosphate, over those containing neutral molecules. Later on, the encapsulation of RNA replicators would have induced a primitive form of competition between the first RNA cells, since those containing more efficient replicators would have grown faster (Chen et al., 2004) (Fig. 2). In these scenarios, natural selection between competing protocells in the absence of genetic systems could have been originally driven by the physicochemical features of early systems. Finally, vesicle membrane growth generates a transmembrane pH gradient (Chen and Szostak, 2004), suggesting that some universal features of the living world could have their origin in fundamental physicochemical features. The perspective now would be to use such vesicles (with various mixtures of putative catalysts, minerals, peptides, or ribozymes) to test if they could favor to set up some form of protometabolism.
Figure 2. ATP and other NTPs, including many modified bases which were not included later on in RNA, probably originated first as energy conveyors in the protometabolism and as coenzymes of peptide catalysts before the origin of RNA itself (de Duve, 2003). Unfortunately, despite recent progress (see below) a single consecutive and convincing prebiotic process has not yet been experimentally demonstrated for their origin [for recent reviews, see (Joyce, 2002; Muller, 2006; Orgel, 2004) and references therein]. The main problem is the formation of ribose and nucleosides. Many sugars with four to six carbons can be produced at alkaline pH by the so-called formose reaction from formaldehyde and catalytic amounts of glycoaldehyde, two simple precursors that are present in interstellar space and were probably on early Earth as well. However, the products of the formose reaction are unstable, and ribose accounts for only a minor portion. Moreover, attempts to combine ribose with bases and/or phosphate in prebiotic conditions also produces complex mixtures of nonspecific products, generating many parasitic molecules that compete with the “normal” building blocks of a nucleotide in the assembly reaction. These observations have led many authors to conclude that ribose was not a prebiotic compound, but was “invented” by organisms living in a “pre-RNA world,” where the scaffold of the genetic material was not ribose but simple sugars [threofuranose nucleic acids (TNA)] or amino acids [peptide nucleic acids (PNA)] [for reviews see (Joyce, 2002; Orgel, 2004; Eschenmoser, 1999)]. However, these compounds are also difficult to produce by prebiotic chemistry and lack some of the interesting properties of RNA. In particular, PNA lacks the charged groups that allow RNA to favor the growth of RNA-containing vesicles versus RNA free vesicles in Szostak's experiments, whereas TNA lacks an activated oxygen (such as the ribose 2OH), essential for ribozyme activity.
Whereas the formation of ribose has never been experimentally investigated in the framework of autotrophic theories, much effort has been done by proponents of the heterotrophic theory to increase the yields and specificity of the formose reaction. It was shown recently that several compounds( Pb++), cyanamide, or borate preferentially complex and stabilize aldopentose and/or especially ribose with respect to other sugars (Ricardo et al., 2004; Springsteen and Joyce, 2004; Zubay and Mui, 2001). The complex formed between ribose and boron is especially interesting since borate occupies the 2 and 3 position of the ribose thus leaving the 5 position available for reactions such as phosphorylation (Li et al., 2005). Borate minerals were probably present in the interstellar space and on early Earth. It was also suggested that ribose, together with purine bases, could have been synthesized in hydrothermal environments on the sea floor (favoring the formose reaction) that could be enriched in borate (Holm et al., 2006). Another recent finding that could be of great importance is that ribose permeates both fatty acid and phospholipid membranes more rapidly than other aldopentoses (Sacerdote and Szostak, 2005). The formation of nucleosides (ribose+base) is also very difficult to achieve in any prebiotic condition. Interestingly, the use of phosphorylated ribose instead of ribose facilitates the association between the base and the sugar, suggesting that phosphoribose might have been a major prebiotic intermediate [(Orgel, 2004) and references therein]. Future effort should thus be concentrated on the search for catalysts (including mixtures of minerals, peptides, and amino acids) that could produce ribonucleotides (and activated ribonucleotides such as NTP) from phosphorylated ribose and various bases, possibly inside lipid vesicles.
The polymerization of ribonucleotides in “prebiotic conditions” has only been achieved using nucleotide monophosphate activated by various amine compounds and using RNA primers. It has been shown that clays (montmorillonite) catalyze the condensation of such activated substrates to form RNA oligomers up to 40–50 nucleotides long [for recent reviews see (Muller, 2006) (Ferris, 2006) (Huang and Ferris, 2003)]. Importantly, the mineral catalysts increase the ratio of 3 to 5 over 2 to 5 phosphodiester bonds. A major problem for the establishment of a robust RNA world is the instability of RNA due to the reactive oxygen in 2 of the ribose (Forterre et al., 1995; Lazcano and Miller, 1996). RNA can be stabilized by a high concentration of monovalent salts (Hethke et al., 1999) (Tehei et al., 2002), but most ribozymes absolutely require millimolar concentrations of divalent salts (Woodson, 2005) which, in contrast, strongly increase RNA degradation at high temperatures (Ginoza et al., 1964). To solve this problem, Vlassov and co-workers have suggested that RNA occurred first in cold environments, where synthesis would have been favored over degradation, an “RNA world on ice” hypothesis (Vlassov et al., 2005). They reported that polymerization of nucleotides, ligation of small RNAs, and other critical prebiotic chemical reactions are indeed stimulated by freezing [(Vlassov et al., 2004) and references therein]. Interestingly, a 3–5 linkage between nucleotides is the major or even the only product formed under freezing conditions. Freezing probably accelerates some chemical reactions in aqueous solution because of the organization of frozen water and the concentration of reactants. In the RNA world on ice scenario, early ribozymes might have survived transport to more warm and wet environments by virtue of their synthetic power outpacing degradation (Vlassov et al., 2004).
The next problem is the production of polymers of sufficient length to harbor catalytic activity (minimal ribozymes). The smallest known “ribozyme” is a 7mer olinucleotide that can cleave itself at 37 °C [for reviews, see (Muller, 2006; Vlassov et al., 2005)]. A mini-RNA ligase of 29 nucleotides has also been obtained by in vitro selection (see below) (Landweber and Pokrovskaya, 1999). This shows that small ribozymes may support simple reactions of cleavage and ligation of other small RNAs. The production of large RNAs by successive ligation of small RNAs would have opened the way to the emergence of true ribozymes. The repertoire of catalytic activities accessible to RNA has been systematically explored in several laboratories using modern enzymes to produce libraries of random RNA oligomers. Large artificial ribozymes selected in vitro can catalyze a wide range of reactions such as RNA polymerization, aminoacylation of transfer RNA, and peptide bond formation [for reviews see (Brosius, 2005; Joyce, 2002; McGinness and Joyce, 2003; Muller, 2006)]. It has even been recently shown that RNA can be used to transport tryptophan across a membrane vesicle (Janas et al., 2004). A major goal of these approaches is to produce an RNA polymerase able to synthesize itself by carrying its own template [for reviews see (Muller, 2006; Orgel, 2004)]. However, whereas the most active RNA polymerase ribozyme (RPR) is 189 nucleotides long, it can only replicate a 14 nucleotide long template (Johnston et al., 2001). The next objectives are to increase the processivity of present RPRs and to introduce a helicase activity (an essential component of all modern polymerases). Future work will probably focus on the possibility of combining various RNA modules with different activities to produce a truly efficient RPR. There is no a priori obstacle to this, and workers in the field argue that powerful evolutionary search procedures using high throughput methodology should allow reaching the goal in the next decade (Muller, 2006).
At some point, one has to assume that an efficient polymerase was not only able to replicate itself, but also to replicate templates producing catalysts (either ribozymes or peptides) useful for the metabolism of the RNA cell [for reviews and hypotheses on this period see (Jeffares et al., 1998; Poole et al., 1998)]. It is likely that many different types of ribozyme-catalyzed peptide synthesis arose, but that only one survived, leading to the modern translation apparatus with tRNA and ribosomes. Many authors have suggested that protein synthesis first appeared as a by-product of RNA replication and was later on selected based on the expanding chaperone and catalytic activities of longer and longer peptides (see below). For instance, by analogy with modern RNA viruses that contain tRNA-like structures at their 3 end used to initiate the replication of viral genomes, Maizels and Weiner (Maizels and Weiner, 1994) suggested that the amino-acid module of tRNA with its CCA end first originated as a tag for genomic RNA replication (functioning both as a telomer and as a marker for RNA to be replicated). All modern tRNAs are monophyletic, i.e., they originated from a single ancestral molecule that would have appeared in a particular RNA-cell lineage. They are made of two modules, the amino-acid binding module and the module carrying the anticodon. The amino-acid binding module probably originated first and was later on duplicated to produce the anticodon module (Maizels and Weiner, 1994). From the imagination of scientists, a great variety of scenarios have been proposed to explain the origin of the translation machinery (Schimmel and Henderson, 1994) (Poole et al., 1998) (Copley et al., 2005) (Taylor, 2006) (Szathmary, 1999) (Wolf and Koonin, 2007). A detailed presentation of these models is beyond the scope of this review. It is usually assumed that the primitive genetic code was simpler (for instance with a two-nucleotide codon and less amino acids) and expanded in the course of evolution. Two main theories have been proposed, suggesting either that codon choice was initiated by specific interaction between amino acids and anticodons (stereochemical theories) or that codon choice was set up parallel with the evolution of the amino acid biosynthetic pathways (historical theories) [for reviews see (Di Giulio, 2005; Ellington et al., 2000; Wong, 2005; Yarus et al., 2005) (Knight and Landweber, 2000)]. In any case, the modern genetic code is probably not a “frozen accident,” but seems to be optimized to minimize the deleterious consequences of mutations (Vogel, 1998) [for review see (Freeland et al., 2003)]. This indicates that the tendency to increase faithful translation was the major selection pressure that directed the evolution of the genetic code, as suggested early on by Woese (1965). Goldenfeld and co-workers have recently shown from in silico stimulation that an optimal code might have become universal in the frame of a communal evolution pervaded by intense horizontal gene transfer of coding sequences and coding system components among coevolving communities with different codes (Vetsigian et al., 2006). If correct, this suggests that mechanisms of gene transfer were operational very early, allowing genetic exchange between RNA-protein cells. Theories about the origin of the genetic code should now also accommodate structural data obtained for modern amino-acyl tRNA synthetases and ribosomes. For instance, from comparative structural analysis, it has been suggested that all modern amino-acyl tRNA synthetases evolved from two proteins whose initial role was to chaperone the tRNA (Ribas de Pouplana and Schimmel, 2001).
The first proteins were indeed probably short chaperonelike proteins that stabilized ribozymes and increased their catalytic activities. They would also have facilitated the transport of molecules (including nucleic acids) through the membranes of the RNA cells, (Jay and Gilbert, 1987). Longer genes and proteins may have originated by RNA recombination producing proteins of increasing size via a multistep combinatorial mechanism under the control of natural selection (de Duve, 2003). Starting from a small number of proteins of small size (corresponding to modern folds), this mechanism would have allowed the extensive exploration of the space sequence at each size level size. This period ended up with the establishment of all modern protein superfamilies by the various combinations of protein folds. Recent advances in comparative and structural genomics have provided fascinating insights on this process [see for instance many recent papers by the group of Koonin (Iyer et al., 2003) (Iyer et al., 2004)]. Complex protein enzymes, such as large RNA polymerases, ribonucleotide reductases, and thymydylate synthases, all required for the origin of DNA, likely only originated at the end of this process.
In the above scenario it is already very clear that DNA probably originated much later than RNA, i.e., in the ribonucleoprotein world (also called “the second age of the RNA world (Forterre, 2005)]. Indeed, it has been convincingly argued that the reduction of ribose is too complex in terms of chemistry to be catalyzed by a ribozyme (Freeland et al., 1999). One can safely assume that the first DNA molecules still contained uracil, because deoxythymidine triphosphate (dTMP) is produced in modern organisms by a modification (methylation) of deoxyuridine triphosphate (dUMP), a reaction catalyzed by thymydylate synthase. Interestingly, recent work has uncovered the existence of two nonhomologous thymydylate synthases, ThyA and ThyX, suggesting that modern DNA with thymidine may have been invented twice, and possibly independently (Myllykallio et al., 2002).
It is usually assumed that DNA replaced RNA because it is more stable and can be replicated more faithfully (Lazcano et al., 1988; Poole et al., 2001). As a consequence, DNA genomes would have become larger, allowing the evolution of complex cells. However, this cannot explain the selection of the first organisms with DNA because genome stability and fidelity was probably not a major problem for fast-replicating RNA cells with small genomes, and the first DNA cells could not have anticipated that their descendents would benefit from a larger genome. One of us has thus suggested that DNA first originated in viruses as a modified form of RNA to protect the viral genetic material against defense mechanisms of the infected cell (a direct selection pressure) (Forterre, 2002). Cellular RNA genomes would have then been transformed later on into DNA genomes following the recruitment by RNA cells of viral enzymes to produce and replicate DNA, or by the takeover of RNA cells by DNA viruses living in a carrier state (Forterre, 2005).
The introduction of viruses in the early evolutionary scenario implies that viruses themselves originated at an early stage in life evolution. The concept of an ancient viral world was indeed first proposed by scientists who suggested that RNA viruses are relics of the RNA world [see, for instance (Maizels and Weiner, 1994)], and that retroviruses, with their RNA–DNA cycles, could give evidence for the transition from the RNA to the DNA world. This concept is now supported by the existence of viruses harboring homologous capsid proteins that infect cells from different domains (Archaea, Bacteria, Eukarya) (Akita et al., 2007; Bamford et al., 2005) suggesting that capsid proteins originated prior to the last universal common ancestor (LUCA). Several models have thus been recently proposed to explain the origin of viruses in the RNA world (Forterre, 2006). Interestingly, the concept of an ancient viral world implies that both modern RNA and DNA viruses might have preserved ancient molecular features from the pre-LUCA era. The study of viruses (especially the extensive exploration of their diversity) should thus be a major area for research on early life evolution in the next decade.
A major goal of the top-down approaches in the origin-of-life field is to reconstruct the common ancestor of all extant organisms to reach an intermediary stage between the origin of life and the present biosphere. The basic principle of cell division and membrane heredity (Cavalier-Smith, 2001) implies that all modern cells derive from a single cell. This historical entity was called the cenancestor (for common ancestor in Greek), the progenote, or the LUCA. This last term has the advantage to be both neutral (unlike the term progenote, which suggests a very primitive organism) and precise. It clearly states that LUCA should not be confused with the first cell, but was the product of a long period of evolution. Being the last means that LUCA was preceded by a long succession of older “ancestors.” In this framework, a plethora of cellular lineages that have left no descendants today may have existed before LUCA. It is important to consider that many of these were probably still present at the time of LUCA, and some have probably even coexisted for some time with its descendants, possibly contributing via horizontal gene transfer to some traits present in modern lineages (Fig. 3).
Figure 3. A consensus on the nature of LUCA is far from reached. For some authors LUCA was a very simple organism, even possibly acellular (Woese, 1998) (Russell and Martin, 2004), whereas others consider that LUCA was a modern-type bacterium (Cavalier-Smith, 2002) or even a primitive Eucaryote with a nucleus (Fuerst, 2005). Thanks to the advances of comparative genomics, some aspects of these hypotheses can now be tested. The identification of a set of genes present in Archaea, Bacteria, and Eukarya has led to the definition of a minimal gene content for LUCA (Delaye et al., 2005; Harris et al., 2003; Koonin, 2003). As expected from the universality of the genetic code, the minimal protein set includes a core of ribosomal proteins, tRNA synthetases, and translation factors (for both initiation and elongation) indicating that the translation apparatus was already well established in LUCA. Importantly, the minimal set includes the components of machineries that are intimately associated with the membrane, such as the signal recognition particle (SRP) and the Sec system—involved in protein secretion—and the complex ATP synthases—that function with a transmembrane proton gradient. These observations clearly indicate that LUCA was a cellular organism with a membrane rather similar to that of modern organisms (Jekely, 2006; Pereto et al., 2004). It remains to be explained why modern lipids are so different in Archaea compared to the “classical” lipids found in Bacteria and Eukaryotes (including an opposite polarity) [for discussion see (Pereto et al., 2004) (Xu and Glansdorff, 2002)]. Future experimental work should focus on the study of vesicles made of the two types of lipids and to the expression in bacteria of enzymes involved in the archaeal lipid pathway and vice versa.
Another controversial idea is that modern hyperthermophiles (i.e., organisms having an optimal growth temperature above 80 °C) could be the direct descendants of a heat-loving LUCA. Hyperthermophiles indeed appear as early diverging lineages in the rRNA universal tree and have relatively short branches (Stetter, 2006). However, this position might be due to the high guanine–cytosine content of their rRNAs, which could have reduced their rate of evolution (leading to shorter branches and artifactual grouping) (Forterre, 1996). Several attempts have been made to determine putative compositional biases in the rRNA, tRNA, or proteins from LUCA in order to determine the temperature at which these molecules were functional [see, for instance (Galtier et al., 1999) (Di Giulio, 2003)]. However, these approaches led to contradictory results and are hampered by the difficulty of reconstructing ancient phylogenies and uncertainties concerning the root of the tree of life (see below). In our opinion, a mesophilic LUCA fits better with the observation that hyperthermophiles are sophisticated organisms that have evolved specific mechanisms to thrive at very high temperatures [for a review see (Forterre and Philippe, 1999a; Xu and Glansdorff, 2002)]. In particular, phylogenomics analyses indeed suggest that reverse gyrase, an atypical DNA topoisomerase present in all hyperthermophiles, was absent in LUCA (Brochier-Armanet and Forterre, 2006; Forterre et al., 2000) whereas hot-temperature-adapted lipids are not homologous in Archaea and Bacteria, suggesting a secondary adaptation that occurred independently in each of these domains (Forterre and Philippe, 1999a; Xu and Glansdorff, 2002).
The minimal set of universal proteins includes a surprisingly small number of proteins that function in DNA replication, lacking in particular a DNA replicase, a primase, and a helicase. This is not due to unrecognized homology since the proteins performing these functions in Bacteria on one side, and Archea–Eukaryotes on the other, belong to different protein superfamilies (Bailey et al., 2006; Leipe et al., 1999). To explain this observation, Koonin and colleagues have suggested that LUCA had an RNA genome, but used DNA as a replication intermediate (much like a retrovirus) (Leipe et al., 1999). Alternatively, if LUCA had a DNA genome, the ancestral system might have been replaced in one lineage (probably in Bacteria) by a new system of viral origin (Forterre, 1999). Finally, if LUCA still had a bona fide RNA genome, Forterre suggested that the few universal proteins involved in DNA metabolism were independently introduced by DNA viruses in the three cellular domains (Forterre, 2006). The idea that LUCA still had a RNA genome has been recently boosted by the discovery of mechanisms for the repair of RNA damages and for enhancing the fidelity of RNA transcription and replication. These findings have suggested that RNA–protein cells may have reached a level of sophistication much more important than previously thought (Forterre, 2005; Poole and Logan, 2005).
Most authors assume that LUCA was identical to the last common ancestor of Archaea and Bacteria, either because it is commonly believed that the tree of life is rooted between the Archaea–Eukaryotes on one side and Bacteria on the other, or because of models where Eukaryotes originated from some kind of association between Archaea and Bacteria (Lopez-Garcia and Moreira, 1999; Martin and Muller, 1998; Rivera and Lake, 2004; Wachtershauser, 2006). However, the root of the bacterial tree and the origin of Eukaryotes remain highly controversial (Forterre and Philippe, 1999b; Gribaldo and Philippe, 2002), (Poole and Penny, 2007). If the root turned out to be in the eucaryotic branch (Philippe and Forterre, 1999), several features now exclusively present in Eukaryotes could already have been present in LUCA, whereas features common to Archaea and Bacteria might have originated in a common lineage to these two domains. At the moment, there is no definitive argument to conclude if the archaeal–eukaryal or even the unique eucaryotic features (e.g., the spliceosome and spliceosomal introns) are ancestral or derived. The same can be said for the features that are common to Bacteria and Archaea, such as the superoperons encoding ribosomal proteins. In any case, many puzzling observations that are difficult to fit in a single coherent scenario remain to be explained. The question of the topology of the universal tree of life is intimately linked to the problem of the origin of the three domains. The main questions to be solved are (i) why three canonical versions of the ribosome (or other universal traits) exist and (ii) how they are now so different from each other, but so similar inside each domain (Woese, 1987). Many contradictory scenarios have been proposed and are still actively debated (Lopez-Garcia and Moreira, 1999; Martin and Muller, 1998; Martin and Russell, 2003; Rivera and Lake, 2004; Woese, 2002) (Cavalier-Smith, 2002) (Forterre, 2006). Much more work has to be done in comparative biochemistry and molecular biology to test various evolutionary scenarios for all possible molecular machines present in modern organisms. In particular, it will be critical to analyze in depth the history of all universal molecular machines (especially the translation apparatus).
Although dramatic progress has been made these last 20 years concerning all aspects of research on the origin of life, there are still critical gaps, especially in the RNA world theory, and no experimental evidence for a consensus scenario. We still do not know how life originated on our planet, and we will possibly never know, since we address here a historical problem for which critical records may have completely disappeared. Furthermore, although the study of the origin of life is a popular subject, the number of laboratories truly working on the subject is extremely small. On the other hand, considering recent trends, we should be able in the near future to understand the physicochemical principles that supported the early emergence of life, and the particular path of evolution of the matter that produced life on our planet could be at least partly revealed by studying modern cells. A major bottleneck for further advances is that scientists working in the various origin of life fields are often isolated from each other either by the borders of their disciplines or by their own theoretical preferences. Research on the origin of life will thus surely benefit from interdisciplinary projects gathering all relevant disciplines to dive into our most distant past.
Work in our laboratory on DNA replication was funded by a HFSP grant.
REFERENCES
Full figure (51 kB)Fig. 1. Schematic of bottom-up and top-down approaches. Major events discussed in the text are highlighted. First citation in article
Full figure (41 kB)Fig. 2. Competition between vesicles in the early RNA world [adapted from Chen (2006)]. Lipid vesicles containing mineral catalysts (hexagons) and able to incorporate ribose (R) and polyphosphate (PP) grow by capturing lipids from vesicles containing amino acids (AA) only. The growth of vesicles induces a proton gradient (H+) that is used to facilitate the transport of various compounds, followed by the synthesis of small RNA oligomers (crosses). After division, vesicles containing RNA replicators (red crosses) grow at the expense of those containing RNA without self-replicating activity (blue crosses). These grow further using additional RNA (green barrel) to facilitate the transport of small polar molecules. First citation in article
Full figure (30 kB)Fig. 3. LUCA was the last bottleneck in a long series of ancestors to the three present-day cellular domains: Archaea, Bacteria, and Eukarya. Extinct lineages may have coexisted for some time with the descendants of LUCA, and transferred some features to them (yellow arrows). The emergence of a universal code in an earlier bottleneck organism may have been favored by preferential transfer between organisms sharing the same genetic code. First citation in article
CORRESPONDENCE: P. Forterre: forterre@pasteur.fr
S. Gribaldo: simo@pasteur.fr