1. The Emergence of Sparse Metabolic Networks

Metabolites in extant metabolic networks represent an extremely sparse sample of chemical space. It is not clear to what extent the emergence of a sparse network was a necessary condition for the emergence of life, and to what extent it was a consequence of the emergence of life. Further, the process by which the network of extant metabolism emerged is one of the major puzzles in the origin of life field. Metabolism may have been constructed only after the emergence of macromolecular RNAs (Joyce 2002), or a proto-metabolic network that generated amino acids and nucleotides (although possibly not ribonucleotides) may have set the stage for the emergence of the RNA World. These questions lie at the core of debates over the environment in which life arose, the source of the organic molecules that supplied the earliest form(s) of life, and the origin of extant metabolism.

Estimates of the size of chemical space vary widely. Bohacek et al. (1996) estimate that there are more than 10⁶⁰ molecules with molecular weights < 500 Da. A more careful analysis that eliminates compounds with small rings and triple bonds predicts that there are > 26 million small molecules containing 11 atoms of C, N, O and F (Fink and Reymond 2007). Regardless of the exact number, it is clear that chemical space is immense. The KEGG database lists only 16,255 small molecule metabolites in all characterized organisms (Kanehisha et al. 2010). Exact enumeration of the metabolome of any individual organism is difficult at this point (Kind et al. 2009); metabolic reconstructions typically underestimate the number of metabolites due to the large number of enzymes of unknown function encoded in every genome. However, current estimates are likely to be correct within a factor of perhaps two. The metabolome of S. cerevisiae contains 800 recognized compounds (Kind et al. 2009). The metabolome of the chemolithoautotrophic bacterium Aquifex aeolicus is even smaller – on the order of 300 compounds (Srinivasan and Morowitz 2009), excluding metabolites involved in cofactor synthesis pathways. Despite its small genome (only 1512 genes) (Deckert et al. 1998), this bacterium can synthesize all of the building blocks of macromolecules from CO₂, H₂, NH₃ and inorganic sources of phosphorous and sulfur. Thus, the core metabolome necessary to support life as we know it under the most demanding conditions (autotrophic growth) is an extremely sparse set of the possible small molecules. Here we will explore the importance of sparseness, how it might have emerged, and how considerations of sparseness inform the debate over the origin of life. Our discussion necessarily spans from the earliest stage of geochemistry in an abiotic world, through the stage of the last universal common ancestor (LUCA), to the final and better understood stage of cells subject to Darwinian selection. Clearly a number of imperfectly understood criteria influenced the inclusion of molecules in the network characteristic of extant metabolism. These include, but are undoubtedly not limited to, the boundary conditions imposed by geochemistry, structural features such as charged functional groups, the availability of catalysts, compatibility within the context of a network that must function within cells as well as within ecosystems, and the potential for assembly into structures of greater complexity such as proteins, nucleic acids, and membranes. The challenge is to understand how the interplay of these various factors led to the emergence of the metabolic network of extant life.

2. Why Sparseness is Important

A high level of chemical diversity might on first consideration seem advantageous for the emergence of life, as it would provide a large number of compounds with different structural and chemical properties, from which the most useful could be selected. However, sparseness in a pre-biotic environment has a number of important advantages for the emergence of life. Given a limited amount of carbon, a sparse network allows accumulation of individual components at higher concentrations than could be achieved if the carbon were distributed among an enormous number of components. The increased concentrations of specific compounds result in a proportional increase in the rates of second-order reactions involving those compounds. Further, the number of potential reaction partners in a sparse network is limited, so there is less potential for unproductive side reactions.

As later stages, sparseness becomes important for additional reasons. Macromolecules that perform catalytic or regulatory functions must discriminate between the correct ligands and similar molecules that interfere with function; both the process of discrimination and the evolution of binding sites with adequate discriminatory abilities are more difficult in a complex mixture. Furthermore, within the confines of a cell, the total number of macromolecules is limited. In a sparse network, concentrations of individual catalysts can be higher and consequently fluxes through pathways can also be higher.

3. A Sparse Metabolic Network Had Emerged By The LUCA

By the time the LUCA emerged, a core metabolic network similar to that in extant microbes was in place. This assertion is based upon the remarkable conservation of core biosynthetic pathways in the three domains of life (Caspi et al. 2010; Kanehisha and Goto, 2000; Kanehisha et al. 2006; Kanehisha et al., 2010). The pathways for synthesis of most of the twenty amino acids used in proteins and the four nucleotides used in RNA are identical or nearly identical in Archaea, bacteria and eukaryotes, suggesting that these pathways were inherited from the LUCA. The TCA cycle is present in representatives of all domains of life, although in some organisms only parts of the cycle are present, and in others the cycle runs in reverse. Note that there are cases in which non-homologous enzymes catalyze comparable reactions. For example, the glycolytic pathway in Archaea is similar to the glycolytic pathway in bacteria and eukaryotes, but a number of the Archaeal enzymes are not homologous to the analogous enzymes in bacteria and eukaryotes (Siebers and Schönheit 2005). Such discrepancies do not diminish the power of the argument. Non-homologous replacement can occur if a gene is lost during a period in which it is not required for viability, and then a non-homologous enzyme is recruited to serve the same function when it becomes important again. Thus, it appears that that the LUCA had the ability to synthesize the critical building blocks of life and did not rely on exogenous sources of these compounds. This supposition is supported by bioinformatic reconstructions of the genome of the LUCA (Ouzounis et al. 2006), although there is uncertainty inherent in any reconstruction based on uncertainties in estimates of the frequency of gene loss and horizontal gene transfer (Mirkin et al. 2003).

The existence of a sparse metabolic network in the LUCA begs the question: Was sparseness a pre-requisite for or a consequence of the emergence of the LUCA? It is difficult to imagine the formation of RNA, or an earlier version (proto-RNA) based upon a different backbone such as threonucleic acid (TNA) (Eschenmoser 1999; Schoning et al. 2000), peptide nucleic acid (PNA) (Nielsen 2007) or glycol nucleic acid (GNA) (Zhang et al. 2005), from an extremely diverse set of pre-biotic compounds present at low levels. This difficulty might have been overcome by a process of self-organization that selected only certain nucleosides (of some form) out of a highly diverse mixture. Hud et al. (2007) and Horowitz et al. (2010) have proposed that "molecular mid-wives", planar heterocyclic molecules with dimensions approximately matching those of Watson-Crick base pairs, may have promoted, via stacking interactions, assembly of nucleobase pairs into positions that allowed subsequent covalent reactions of attached functional groups to form a backbone. Alternatively, the emergence of a proto-RNA may not have been possible until a relatively sparse set of potential precursors had been generated by pruning of a complex proto-metabolic network. With respect to formation of proto-RNA, self-organization and a requirement for a relatively sparse set of precursors are not mutually exclusive, and indeed both may have been required. However, the LUCA also had a sparse network of amino acid and sugar biosynthetic pathways, and mechanisms for selforganization of these components analogous to the molecular midwives model proposed by Hud and coworkers for self-organization of nucleobases are more difficult to imagine.

4. What Was The Starting Point?

Any attempt to understand how the sparse network of metabolism arose must grapple with the unresolved question of where life arose in order to establish the relevant boundary conditions. The theory that has prevailed for several decades posits that life arose in ponds or lagoons supplied with organic compounds by in-fall from the atmosphere, interstellar dust particles, meteors and comets. This theory is based on the early ideas of Oparin (1938) and Haldane (1929), and gained momentum from the famous Miller-Urey experiment, which demonstrated that electric discharges in an atmosphere containing CH₄, NH₃, H₂O and H₂ generate a complex mixture of organic compounds, including 22 amino acids (Miller 1953, Johnson et al. 2008). The complexity of the product mixture arises because highly reactive free radicals are formed during the electric discharge; indeed, a large amount of intractable organic tar is also formed. The conditions used by Miller are now believed to be an inaccurate model of the atmosphere of the early Earth. The composition of the early atmosphere is still uncertain, but was more likely a mixture of N₂, CO₂, CO, H₂O and perhaps CH₄, and probably contained little NH₃ (Kasting 2008). However, Cleaves et al. (2008) have shown that amino acid precursors are generated when electrical discharges are passed through mixtures of these gases. Substantial levels of amino acids were observed only when the reactions mixtures were buffered to approximately neutral pH with CaCH₃ and were subjected to acid hydrolysis after the experiment in the presence of ascorbate (to prevent oxidative decomposition of amino acids by nitrite and nitrate formed during the spark discharge). On the pre-biotic Earth, formation of amino acids could have occurred by similar processes, although at a slower rate. In addition to delivery of organics formed by atmospheric processes, an enormous amount of material was delivered to the early Earth from space. Chyba and Sagan (1992) estimated that > 10⁹ kg of carbon was delivered per year from interstellar dust particles, >10⁶ kg by comets, and > 10⁴ kg by meteors.

There is imperfect overlap between the set of compounds delivered in this manner and the set of compounds that constitutes the core of metabolism. For example, over 100 amino acids have been found in meteorites (Pizzarello 2007). Of these, only some are α−amino acids and only 8 (Ala, Asp, Ile, Leu, Glu, Gly, Pro and Val) are used in extant life. Conversely, 12 of the proteinogenic amino acids are not found in meteors. In the spark discharge experiments reported by Cleaves et al. (2008), the proteinogenic amino acids Asp, Glu, Ser, Gly and Ala are formed, but γ-aminobutyrate, β-alanine, and α-aminoisobutyric acid, which are not found in proteins, were also observed. The lack of many proteinogenic amino acids is not a concern; certainly amino acids that were not delivered exogenously could have been synthesized later by macromolecular catalysts. The hypothesis that some of the complex amino acids were added to the genetic code later than the simpler ones is well accepted (Trifonov 2000). However, the presence of many non-biological components in the materials delivered from space and atmospheric discharges requires that, if life emerged in ponds or lagoons, the sparse network of metabolism must have emerged by substantial pruning of a complex network of abiotic compounds or by some process of self-organization that selected only certain molecules for inclusion in the proto-metabolic network but excluded others that had similar properties.

An entirely different possibility is the proposal that life originated at hydrothermal vents (Baross and Hoffman, 1985; Martin et al. 2008). Observations of lush ecosystems surrounding hydrothermal vents that are supplied with nutrients solely from the vent fluids, as well as an increased understanding of the nature of chemolithoautotrophic metabolism, have generated considerable interest in this hypothesis. At mid-ocean spreading centers, hot hydrothermal fluids containing CO₂, H₂, and H₂S, and low levels of CH₄ are vented through porous minerals, including transition metal sulfides such as pyrrhotine (Fes), pyrite (FeS₂), chalcopyrite (CuFeS₂) and sphalerite (Zn,Fe)S (Kelley et al. 2002). Clay minerals such as nontronite (Murnane and Clague, 1983) and phyllosilicates such as talc and chlorite (Dekov et al. 2008) are also found in these hydrothermal systems. A fundamentally different type of vent is found in peridotite-hosted systems (e.g. the Lost City) (Kelley et al. 2001; Kelley et al. 2005) found off the main spreading axis. At these sites, reactions between sea water and newly exposed mantle olivine ((Mg,Fe)₂SiO₄) generate serpentine (Mg₃Si₂O₅(OH)₄), magnetite (Fe₃O₄) and alkaline fluids that, upon mixing with cool sea water, generate towers of aragonite (CaCH₃) and brucite (Mg(OH)₂). Serpentine undergoes further reaction with CO₂ to give talc (Mg₃Si₄O₁₀(OH)₂). Vent fluids are cooler in these systems (< 90ºC), and are rich in H₂ and CH₄. In both types of vents, the walls are porous, allowing chemical reactions between constituents of vent fluids to occur in pores lined with catalytic minerals at a range of temperatures varying from that of the inner conduit to that of seawater. The types of structures formed at hydrothermal vents may have been different in the Hadean because the pH of the ocean was likely acidic (Grotzingen and Kasting 1993) and because Fe(III) was the primary oxidant present in sea water. For example, iron monosulphide bubbles may have precipitated at hydrothermal vents due to mixing of alkaline hydrothermal fluids derived from serpentinization with the acidic iron-laden ocean (Russell and Hall, 1997).

If life originated at hydrothermal vents, then the sparse network of metabolism must have emerged by elaboration of the extremely sparse network of small molecules found in vent fluids. Experimental investigation of abiotic reactions under hydrothermal vent conditions has lagged behind studies of reactions in the primitive atmosphere and surface waters, partly because hydrothermal vents were only discovered in 1979 (Corliss et al. 1979), 26 years after the Miller- Urey experiment jump-started the origin of life field. Experimental studies of catalysis of prebiotically relevant reactions by minerals found at hydrothermal vents are promising. For example, CO is reduced to methane thiol and other alkane thiols by phyrrhotite in the presence of H₂S at 100 ºC (Heinen and Lauwers, 1996), and carboxylic acids (Cody et al. 2004) and α-keto acids such as pyruvate (Cody et al. 2000) are formed from alkane thiols and CO by several transition metal sulfides.

These two scenarios for the origin of life differ in important ways. Ponds and lagoons supplied with organic compounds from space have an over-abundance of chemical constituents, but a relatively sparse supply of catalysts. Hydrothermal vents have a sparse set of inputs, but an abundance of catalytic minerals. Photochemical reactions, which often produce complex mixtures due to formation of highly reactive free radicals, are relevant at the surface of the earth, but irrelevant at hydrothermal vents. Further, these scenarios differ substantially with respect to the potential for compartmentalization, which was undoubtedly important for concentrating and protecting metabolites and, eventually, macromolecules. Compartmentalization in ponds and lagoons might have arisen by cycles of evaporation and re-dissolution at the margins of lagoons and ponds that concentrated amphiphilic molecules capable of self-organization into primitive membranes (Segré et al. 2001). In contrast, compartmentalization could have been provided by pores in the walls of hydrothermal vents and the fractured subsurface surrounding vents even before the emergence of large amphiphilic molecules capable of forming membranes (Martin and Russell, 2003; Koonin and Martin, 2005).

5. The Critical Roles of Catalysts and Environmental Conditions in the Emergence of a Sparse Metabolic Network

Regardless of where life originated, the emergence of a sparse metabolic network depended critically on catalysis. The importance of rate acceleration by catalysts is obvious; nearly every reaction in extant cells is too slow to support life without a catalyst (see Table 1). The half-lives for the reactions listed in Table 1 apply to reactions at 25 ºC and pH 7; these reactions might be faster at higher temperatures and/or different values of pH, but the need for catalysis is still evident.

A less-appreciated role for early catalysts would have been to prune complex protometabolic networks. By accelerating reactions, catalysts can direct flux toward a single product at the expense of uncatalyzed competing reactions. In the context of a complex potential network space, catalysts determine the topology of the network and the products that are formed. Fig. 1 shows a hypothetical example in which catalysts prune a complex network, resulting in a much sparser suite of products than would be formed in the absence of catalysis. Note that the availability of catalysts for different steps in a network results in significantly different network topologies and accumulation of different products.

An elegant demonstration of the pruning of a complex network by a catalyst is the effect of phosphate on the reaction of cyanamide and glycolaldehyde (see Fig. 2) (Powner et al. 2009). In the absence of phosphate, this reaction forms a multitude of products, but in the presence of phosphate, which acts as a general base catalyst and accelerates the rates of two particular steps, 2-amino-oxazole is produced in over 80% yield. The ¹H-NMR spectrum shown in Fig. 2b shows a forest of peaks due to the many products formed in the absence of phosphate; the spectrum in Fig. 2c is much simpler because a single product dominates the mixture.

In addition to the profoundly important role played by catalysts, the structure of protometabolic networks would have been a function of geochemical conditions such as temperature and pH. Reactions involving proton transfers and/or or nucleophilic attack by groups containing oxygen, nitrogen, or carbon are particularly susceptible to pH. For example, the rate of hydrolysis of tetramethylsuccinanilic acid shows a complex dependence on pH (see Fig. 3).

6. The Earliest Catalysts

The importance of mineral surfaces in catalyzing pre-biotic reactions has been recognized by Bernal (1951), Wächtershäuser (1988) and Russell and Hall (2006), among others. At least 1000 mineral species were formed during planetary accretion and the subsequent reworking of the crust and mantle prior to 3.5 billion years ago (Hazen and Ferry, 2010), so many potentially catalytic minerals were available in a range of geochemical settings during the critical time in which life emerged.

Mineral surfaces could have played three important roles in the emergence of a sparse metabolic network. First, mineral surfaces would have concentrated proto-metabolites with functional groups, particularly carboxylates and phosphates, that promoted adsorption. In addition, mineral surfaces could have exerted catalytic effects by polarizing carbonyl groups and enhancing their electrophilicity, by providing nucleophilic hydroxyl groups, or by facilitating electron transfer reactions. Metal ions are used in similar roles in many protein enzymes, and may be relics of a time when metal ions performed similar roles in the absence of proteins. Several discoveries demonstrate that mineral surfaces can catalyze reactions of pre-biotic importance; some of these were mentioned above. In addition, montmorillonite clays accelerate condensation of activated nucleotides to form oligomers up to 50 in length (Ferris 2006), and magnetite catalyzes the reduction of N₂ to NH₃ in the presence of formate as a reducing agent (Brandes et al. 1998).

A third way in which minerals could have played an important role in pre-biotic chemical networks is stabilization of certain products. For example, the half-life of leucine at 200 ºC is increased from <10 min to 30 h in the presence of troilite (Hazen et al. 2002). Ribose decomposes in minutes under alkaline conditions, but is stable for days in the presence of borate minerals. Further, borate minerals skew the products produced in the formose reaction, which involves aldol condensations of formaldehyde and glycolaldehyde at high pH and generally produces an intractable brown mixture. Borate stabilizes the intermediate glyceraldehyde, preventing enolization and forcing it to react only as an electrophile. Consequently, pentoses are formed in high yield (Ricardo et al. 2004).

Small molecules containing carboxylate, amino, thiol and phosphate groups may also have been important catalysts in pre-biotic reaction networks. Amino and thiol groups can act as general bases or nucleophiles, and carboxylates and phosphates as general acids or general bases, depending on the pH. For example, carboxylic acids act as general acid/base catalysts for many reactions, including nucleophilic substitution (Dietze and Jencks, 1989), hydrolysis (Covitz and Westheminer, 1966; Jencks and Carriuiolo, 1961) and nucleophilic attack of H₂O₂ on aldehydes (Sander and Jencks, 1968). Phosphate can also act as a general acid/base catalyst, as shown in Fig. 2 (Powner et al., 2009). The amino and carboxylate groups of amino acids can participate in catalysis even in the absence of a reactive side chain; for example, alanine and isovaline catalyze aldol condensations (Pizzarello and Weber, 2004). The simple amino acid proline is a remarkably good catalyst for a number of reactions, including aldol condensation, Michael reactions and Mannich reactions (List 2002).

Although mineral catalysts and small molecules were undoubtedly important in the earliest stages of the emergence of life, both have drawbacks that led to their eventual replacement with macromolecular catalysts. First, mineral catalysts are susceptible to poisoning as surfaces become coated with organic molecules. Second, both mineral and small molecule catalysts are inherently non-specific. Broad substrate specificity creates a potential for catalysis of undesirable reactions. Furthermore, catalysis of multiple reactions makes it difficult to optimize fluxes toward products that may be needed in very different quantities. Finally, catalytic power can be increased when a very specific substrate binding site allows optimal orientation of substrate with respect to the active site machinery.

7. Why This Particular Set Of Metabolites?

Molecules in extant metabolomes share a number of properties. They are hydrophilic and are usually, although not always, reasonably stable. They are generally inert with respect to nonenzymatic reactions with proteins and nucleic acids that would perturb the structure and/or function of these macromolecules. There are a few notable exceptions in which highly reactive intermediates are formed, but these are sequestered within protein tunnels that connect the active site at which they are formed with the active site at which they are used (Weeks et al. 2006).

Nearly all metabolites contain charged functional groups. In fact, only two metabolites in A. aeolicus (histidinol and histidinal) lack a carboxylate or a phosphate (Srinivasan and Morowitz, 2009). In extant cells these functional groups provide "handles" for binding to proteins and prevent diffusion through the membrane and thus loss of valuable metabolites. Such functional groups may have been important for different reasons during the emergence of life. The negative charges of carboxylates and phosphates promote adsorption to cations exposed at mineral surfaces, and amino groups can hydrogen bond with exposed oxygen atoms. By concentrating reactants near surfaces, adsorption can enhance the rates of second-order reactions as long as functional groups required for reactivity are not masked. Carboxylates, phosphates and amino groups might also have served as intramolecular catalysts. Substantial rate accelerations can be achieved via intramolecular catalysis. For example, hydrolysis of the amide bond shown in Fig. 4 is accelerated by 1.3 x 10⁵ in the presence of an intramolecular carboxylate (Higuchi et al. 1966). Thus, molecules with carboxylates, phosphates and amino groups may have been more likely to participate in a proto-metabolic network due to concentration on surfaces, and those with such functional groups in positions that allowed intramolecular catalysis may have been preferentially recruited into a network sustained by mutual catalysis.

Molecules containing carbonyl groups are ubiquitous in extant metabolic networks (Benner et al. 2004). Carbonyl groups are electrophilic due to unequal sharing of electrons in the C-O double bond. Thus, the partially positively charged carbon is susceptible to attack by a variety of electron-rich nucleophiles. Carbonyl compounds undergo a number of reactions that are critical for extant biosynthetic pathways, including aldol condensations that form new C-C bonds, reductive amination to form amino acids, reduction to form alcohols, and nucleophilic displacements such as those involved in formation of peptides from activated amino acids. No other functional group plays such a critical role in metabolic pathways. The carbonyl group undoubtedly played an important role in prebiotic chemistry, as well. While we can recognize features of metabolites that are useful in extant metabolism and that may have been carried over because they predominated in a surface-associated protometabolic network, many other molecules have similar features but were not incorporated into extant metabolism. Contingency based upon the availability of precursors, physical conditions of the environment such as pH and temperature that influence both the probability of synthesizing a molecule as well as its stability, and the types of available catalysts undoubtedly influenced the suite of molecules selected for metabolism. An example of a class of molecules that may have been excluded from an emerging metabolic network by such criteria might be beta amino acids. Polypeptides made of beta amino acids form helices and fold into stable structures (Daniels et al. 2007, Petersson and Schepartz, 2008). Polypeptides made from beta amino acids do not form beta sheets, so the richness of structures available to modern proteins, which display thousands of variations based upon combinations of helical and sheet structures, would not be available in life based upon beta amino acids. However, selection against beta amino acids may have occurred before the emergence of macromolecules or even peptides. In extant metabolism, amino acids are synthesized by transamination of keto acids, using another amino acid, most often glutamine, as an amino donor. Beta keto acids are less stable than alpha keto acids due to their tendency to decarboxylate (Jencks 1969). Therefore, the availability of beto keto acids would likely have been lower than the availability of alpha keto acids. This simple kinetic effect may have acted as a filter prior to any effect on Darwinian fitness conferred by peptides or proteins synthesized from amino acids.

Some molecules may have been weeded out at later stages due to disadvantageous features that became important only after more sophisticated macromolecules arose. Weber and Miller (1981) have given an insightful discussion of why certain amino acids are not found in proteins. For example, ornithine undergoes facile lactamization when the carboxylate is activated; this reaction would remove the amino acid from the 3’-OH of a tRNA to which it was attached, or if ornithine were added to a growing peptide chain, lactamization would remove the entire peptide from the peptidyl tRNA.

We also recognize that life could have evolved to use different amino acids in proteins and different backbones and bases in nucleic acids. Although the amino acids found in proteins each have specific advantageous properties (Weber and Miller 1981), this set is not likely to be a unique solution to the problem of generating stable proteins. Methods developed by Schultz and co-workers have allowed incorporation of more than 50 non-standard amino acids into proteins, although only at single positions (Young and Schultz, 2010). Incorporation of a non-standard amino acid at multiple positions would be expected to perturb the structure of a protein, but compensatory mutations at other positions would very likely allow recovery of proper folding and function. Such studies would yield fascinating insights into the potential for use of a different amino acid alphabet.

Alternative forms of nucleic acids have also been explored. Nucleic acids with intriguing alternative backbone structures include threonucleic acid (TNA) (Eschenmoser 1999; Schoning et al. 2000), peptide nucleic acid (PNA) (Nielsen 2007) and glycol nucleic acid (GNA) (Zhang et al. 2005). Further, several non-standard nucleobase pairs have been discovered that can be accommodated within a double helix, although some nucleobases are not ideal because they are prone to epimerization, oxidation, or formation of tautomeric structures that decrease the fidelity of replication (Benner 2004).

The selection of the amino acids found in proteins and the bases found in DNA in extant life on Earth presumably relied upon the availability of either these building blocks themselves or of the precursors and catalysts required to allow their synthesis. Under different geochemical conditions, a different suite of building blocks might well have emerged, and their adoption into macromolecules would necessarily have entailed a different set of metabolites in biosynthetic pathways. Indeed, some of these alternative solutions may have been explored before the emergence of the LUCA, and abandoned as certain building blocks "won" during the consolidation of the genetic code and the structure of metabolism and the emergence of translation in the communal progenote that preceded the LUCA (Woese 1998).

8. Why This Particular Set Of Reactions?

The set of metabolites constituting the extant metabolome is unlikely to be the only solution to the problem of synthesizing amino acids and nucleotides. Many alternative pathways could have reached the same endpoints via a different set of intermediates. This concept is illustrated by the two pathways for synthesis of the cofactor pyridoxal 5’-phosphate (PLP) shown in Fig. 5 (Kim et al, submitted for publication). E. coli and other gamma proteobacteria synthesize L-4-phosphohydroxythreonine, an intermediate in the synthesis of PLP, via the pathway shown on the left. When the gene encoding the second enzyme in this pathway is deleted, PLP can be synthesized if YeaB, which diverts an intermediate in serine biosynthesis to hydroxypyruvate, is overproduced. The pathway on the right is not a bona fide pathway that has evolved for this purpose, and is not known to occur in any living organism. Three of the four steps are catalyzed with rather poor efficiency by broad-specificity or promiscuous enzymes that normally serve other functions, but can be recruited to catalyze the indicated reactions when substrates that would not normally be present become available. The chemical reactions in this pathway are not intrinsically difficult or different from typical metabolic reactions. Thus, this pathway might have been perfectly suitable for producing PLP, but may have gone undiscovered because the extant pathway emerged first.

The assembly of a proto-metabolic or metabolic pathway given a set of catalysts depends on both "probabilistic" and "combinatorial" factors. The probabilistic aspect reflects the probability that a catalyst will be able to catalyze a particular reaction, which depends upon the nature of the substrate binding site and the ways in which the reaction can be accelerated. The combinatorial aspect relates to the number of ways available catalysts may be combined. Before the emergence of life, and even in the LUCA, there were fewer catalysts than in extant life, and they were both less efficient and less specific (Copley 2010). Thus, assembly of novel pathways early on may have depended primarily on the broad specificity of available catalysts, while in extant life, it depends primarily on novel combinations drawn from the larger number of available enzymes, but also on the existence of inefficient promiscuous activities found even in highly specific enzymes.

9. Exchange Of Organic Material Between Locales

It is unlikely that any single environment on the early earth was capable of generating all of the precursors for life. Synthetic chemists are well aware that temperature and pH, concentrations of starting materials and the presence and nature of catalysts profoundly affect the yield as well as the distribution of side-products of any reaction. An important and poorly understood issue is the extent to which molecules formed at different sites and under different conditions would have mixed on the early Earth. Exchange of material between ponds and lagoons on the surface of the earth would have been vectorial, with organic compounds moving down geophysical gradients and ultimately washing into the open ocean. In contrast, organic compounds formed in vent systems characterized by different mineral catalysts, temperature, and pH would likely have exchanged more readily. The crust of the early earth was highly fractured, and there would have been extensive circulation of fluid through vents and the surrounding crust. Thus, high temperature regions of vents might have been responsible for initial carbon fixation reactions, while formation of more fragile molecules might have occurred in the more moderate regions of the outer walls or in the fractured crust surrounding the vent. John Baross has referred to a pre-biotic system characterized by exchange of organic compounds over substantial distances as a "10 km cell".

9. Does The Sparse Metabolic Network of Extant Life Resemble A Proto-Metabolic Network?

As discussed above, biosynthetic pathways in extant organisms clearly resemble those in the LUCA. A more difficult question is whether the structure of metabolism in the LUCA reflects the structure of a pre-existing proto-metabolic reaction network, or replaced a preexisting proto-metabolic reaction network. In the first scenario, metabolic pathways would have remained largely the same while more efficient catalysts were recruited to facilitate individual reactions, leading to a smooth transition from the earliest stages of mineral and small molecule catalysis, through an intermediate stage involving proto-RNA and RNA catalysts (likely with catalytic auxiliaries provided by amino acids, peptides, and cofactors), and finally to protein enzymes. A point in favor of this argument is that it is undoubtedly easier to patch a single catalyst into a functioning pathway than to invent de novo an entirely different pathway whose efficiency surpasses that of a previously existing pathway.

A second hypothesis is that modern metabolic pathways have completely replaced primordial pathways due to the advent of more effective catalysts, probably at the stage of the RNA World. Benner et al. (2009) have proposed that modern metabolism is a palimpsest and that re-writing of modern metabolism has largely obscured previously existing proto-metabolic pathways. This viewpoint is based upon the assumption that a large number of highly effective catalysts arose in the RNA World, or at least by the LUCA, which together allowed flux through pathways that had never before been accessible. In the context of Fig. 1, this would correspond to a switching between sets of catalysts with consequent reconstruction of the topology of the network.

The answer to this question most likely lies somewhere between these two opposing theories. The idea that modern metabolism runs along pathways that were laid down before the emergence of the LUCA is appealing from the standpoint of continuity between pre-life and life, and because recruitment of catalysts one at a time is more likely than recruitment of several catalysts simultaneously to enable an entirely new pathway. However, recruitment of several catalysts simultaneously to enable a novel pathway can certainly occur once there is a sufficient collection of catalysts (see Fig. 5). Catalysts, even highly evolved modern enzymes, are never completely specific for catalysis of a single reaction. In some cases, promiscuous secondary activities can be patched together to constitute a new pathway. The serendipitous pathway for synthesis of 4-phosphohydroxythreonine described above is an example. A second example is the pathway for degradation of pentachlorophenol shown in Fig. 6. Pentachlorophenol is an anthropogenic pollutant that was first introduced as a pesticide in the 1930s. The soil bacterium Sphingobium chlorophenolicum has patched together a novel pathway that allows conversion of pentachlorophenol into a downstream metabolite in an existing catabolic pathway (Copley 2000; Dai et al. 2003; Xun et al. 1992; Xu et al. 1999; Xun et al. 1999). Although the pathway is inefficient and individual enzymes in the pathway perform poorly (Warner and Copley, 2007; Behlen, Pietari and Copley, unpublished), the pathway suffices to both detoxify pentachlorophenol and provide a novel source of carbon.

These examples clearly illustrate that multiple catalysts can be recruited simultaneously to generate a novel pathway. The key in these cases, and in others throughout the history of life, is whether a novel pathway provides an increase in fitness relative to a preexisting pathway. When there is no preexisting pathway (e.g. for degradation of pentachlorophenol), or when a preexisting pathway is disabled (e.g. the gene deletion that blocks the normal pathway for synthesis of PLP), a novel pathway may enhance fitness even if it is quite inefficient, and subsequent mutations can increase the efficiency of the pathway. However, if an existing pathway has already evolved to be reasonably efficient, most novel pathways will not be able to compete. This reasoning likely explains why the core biosynthetic pathways have survived essentially unchanged for over three billion years since the LUCA.

The argument in the previous paragraph is couched in terms of fitness, a concept that most readily applies in the post-LUCA era when Darwinian evolution has occurred. These ideas can, however, be extended to a pre-Darwinian world, however. The predecessor to the LUCA is believed to have been a communal organism in which metabolites and genetic information were shared (Woese 1998). Thus, even before the LUCA, there would have been the potential for metabolic novelty if a chance association of catalysts generated a pathway that was more efficient than an existing pathway and that enhanced the fitness of the communal organism.

10. Conclusions

A consideration of the mechanisms by which the sparse metabolic network characteristic of extant life emerged is complicated because molecules in a proto-metabolic or metabolic network must have satisfied multiple criteria simultaneously over the entire course of the emergence of life, even as these criteria were constantly changing. Initially, they must have been accessible from geochemical or astrophysical precursors by pathways that used reliably available catalysts or facile un-catalyzed reactions. If catalysts were compounds within the network (such as small molecules, metal ions chelated by small molecules, peptides or oligonucleotides (Copley et al., 2007), they must have been produced at sufficient rates relative to dilution or degradation to participate reliably in the network. It is a truism that evolution cannot anticipate the future usefulness of particular phenotypes. Thus, the earliest constituents of a proto-metabolic network must have provided some potential for the emergence of further complexity, or else life would not have emerged. Further, reasonable concentrations of at least a minimal set of building blocks must have been available before the emergence of macromolecules that were able to exert feedback on the system by catalyzing the synthesis of their own building blocks. Different processes would have influenced the sparse network of metabolism after the advent of macromolecules. At this stage, components that enhanced the functions of proto-RNA, RNA and, later, proteins would have been favored, but this more complex world would still have relied ultimately upon materials available from the environment. Thus, it is probably a false prejudice to cast the emergence of the sparse metabolic network of extant life as either a consequence of selection due to adsorption onto surfaces and kinetic effects before the emergence of life, or by selection in the Darwinian era, as discrete alternatives. A more nuanced picture of the emergence of metabolism must address how selective pressures varied over time and under changes in the complexity of the metabolic network, and which aspects of the composition and topology of extant metabolic networks arose by which mechanisms.

Acknowledgements: This material is based upon work supported by the National Science Foundation under Grant No. 0526747.