- Open Access
Amino acid fermentation at the origin of the genetic code
© de Vladar; licensee BioMed Central Ltd. 2012
Received: 24 November 2011
Accepted: 10 February 2012
Published: 10 February 2012
There is evidence that the genetic code was established prior to the existence of proteins, when metabolism was powered by ribozymes. Also, early proto-organisms had to rely on simple anaerobic bioenergetic processes. In this work I propose that amino acid fermentation powered metabolism in the RNA world, and that this was facilitated by proto-adapters, the precursors of the tRNAs. Amino acids were used as carbon sources rather than as catalytic or structural elements. In modern bacteria, amino acid fermentation is known as the Stickland reaction. This pathway involves two amino acids: the first undergoes oxidative deamination, and the second acts as an electron acceptor through reductive deamination. This redox reaction results in two keto acids that are employed to synthesise ATP via substrate-level phosphorylation. The Stickland reaction is the basic bioenergetic pathway of some bacteria of the genus Clostridium. Two other facts support Stickland fermentation in the RNA world. First, several Stickland amino acid pairs are synthesised in abiotic amino acid synthesis. This suggests that amino acids that could be used as an energy substrate were freely available. Second, anticodons that have complementary sequences often correspond to amino acids that form Stickland pairs. The main hypothesis of this paper is that pairs of complementary proto-adapters were assigned to Stickland amino acids pairs. There are signatures of this hypothesis in the genetic code. Furthermore, it is argued that the proto-adapters formed double strands that brought amino acid pairs into proximity to facilitate their mutual redox reaction, structurally constraining the anticodon pairs that are assigned to these amino acid pairs. Significance tests which randomise the code are performed to study the extent of the variability of the energetic (ATP) yield. Random assignments can lead to a substantial yield of ATP and maintain enough variability, thus selection can act and refine the assignments into a proto-code that optimises the energetic yield. Monte Carlo simulations are performed to evaluate the establishment of these simple proto-codes, based on amino acid substitutions and codon swapping. In all cases, donor amino acids are assigned to anticodons composed of U+G, and have low redundancy (1-2 codons), whereas acceptor amino acids are assigned to the the remaining codons. These bioenergetic and structural constraints allow for a metabolic role for amino acids before their co-option as catalyst cofactors. Reviewers: this article was reviewed by Prof. William Martin, Prof. Eörs Szathmáry (nominated by Dr. Gáspár Jékely) and Dr. Ádám Kun (nominated by Dr. Sandor Pongor)
I propose that an alternative ancient function for the amino acids in an earlier RNA world was to harvest energy. I propose a scenario in which coding is required in order to carry out the catabolic degradation of amino acids, and which may have easily arisen from initial random assignments.
Amino acids involved in Stickland fermentation in Clostridium spp.
e- Donors (reductants)
e- Acceptors (oxidants)
Donors or acceptors (depending on partners)
Valine * 
Aspartic acid • 
Glutamic acid• 
Secondly, amino acids are readily produced under numerous potential scenarios for the abiotic synthesis of the building blocks of life. The famous Miller experiment  and other Miller-Urey syntheses [11–13] employ energy sources to generate amino acids from gases. The "iron-sulfur" chemo-autotrophic theory , which is compatible with the conditions of black-smoker hydrothermal vents (extremely hot and highly acidic), employs the reductive power of sulphides and high temperatures to synthesise organic molecules, including amino acids, via thio esters. In the warm but alkaline hydrothermal vents (the white smokers) a similar reduction of CO2 by sulphides occurs, and the reactions that take place are analogous to the acetogenesis (Wood-Ljungdahl) pathway, where the nitrogen-fixation branches lead to amino acids, which are, in turn, also used as precursors to synthesise nucleotides .
The third observation is the crucial one; anticodons that are complementary often code for amino acids that have conjugated Stickland pairs. This is particularly true for the "Miller" amino acids (which are the simplest ones), and also holds for complementary anticodons that involve non-canonical G-U pairs. The correlation between these two forms of complementation (the metabolic or Stickland complementation, and the anticodon complementation) is relevant because in the evolution of the proto-adapters, RNA complementation played a role in diversifying the anticodons [6, 17, 18].
These three independent facts suggest that amino acid fermentation could have played a role in the establishment of the genetic code. The purpose of this article is to develop this idea in more detail, to state the hypothesis in a testable way, and to analyse the implications that it has on our current understanding about the context of the early metabolism and on the factors that established the translational machinery.
Abiogenesis of amino acids
In the 1920's Oparin  and Haldane  proposed that under anoxic conditions and with suitable energy sources, organic matter would spontaneously form, which would be the basis for the first forms of life. Later, Miller  synthesised amino acids from a mix of four elementary gases: ammonia (NH3), methane (CH4), water (H2O), and carbon dioxide (CO2). This mixture forms a "reductive atmosphere", which means that the compounds that are formed have a significant potential to undergo redox reactions. Some compounds are good electron donors, such as ammonia and methane. Carbon dioxide and molecular oxygen are powerful electron acceptors. The presence of electron acceptors is critical for organisms, since they allow the oxidation of carbon sources (e.g. glucose) to water and carbon dioxide, and the concomitant release of free energy is employed for vital metabolic processes. However, contemporary understanding suggests that methane and ammonia were absent from the early Hadean atmosphere (about 4 billion years ago). Instead, molecular nitrogen (N2) was abundant . In this "weakly reducing" atmosphere N2 is a poor electron donor, which limits both the diversity and the yield of the compounds that can be abiotically formed.
In any case, applying energy to these reducing mixtures synthesises amino acids that are biologically significant. This is true for several of the variants of Miller's experiment . Miller-like experiments usually give relatively high yields of glycine, aspartic acid, alanine and valine, amongst others, depending on the energy source and initial mix [11–13, 22]. Figure 4 depicts the relative yields of the reducing atmosphere, in which some important amino acids are readily synthesised.
It is debatable however, under which atmospheric conditions the origin of the genetic code occurred, and therefore which were the amino acids that were relevant for an early metabolism. One possibility is that the genetic code was established just after the origin of life, in a pre-RNA world in a prebiotic-soup setting, under a weakly reducing atmosphere with low yields of glycine, aspartic acid and alanine . This is an unlikely possibility because an abiotic metabolism which included nucleic acids had to exist. The geological conditions during this period were also very harsh, characterized by hostile volcanic activity and high temperatures. These conditions are thought to be too adverse for the establishment of early forms of life, although it is not clear whether autotrophic metabolisms could have actually existed. Another possibility, is that the genetic code was established in a late RNA world, after the atmosphere had already changed, perhaps to include methane and ammonia produced from the metabolism of methanogenic protobionts. This, as we understand it, happened during the Archaean period, about 3.5 billion years ago. This is a more likely scenario because in an already existing metabolism, an emerging code would result from the reorganisation of established processes. In other words, the pre-adaptations that were required for the genetic code to be established are consistent with early metabolisms.
Wächtershäuser summarised and pointed out several reasons why Miller's "prebiotic soup" does not work as a model for prebiotic evolution . Although these reasons are debatable, it is worth considering that a prebiotic soup may in fact not be the most adequate scenario. One strong argument is that the composition of the primitive atmosphere  would not consist of the gases needed to synthesise primordial amino acids with enough yield. As an alternative, Wächtershäuser proposed what has been called the "primitive pizza" [1, pp. 32-33]. This theory posits the prebiotic reactions occurring not in solution, but on the surface of pyrites, minerals rich in iron. Ferrous cations (positively charged) and sulphur anions (negatively charged) are adsorbed in the mineral surface and react to form pyrite. This reaction, i.e. Fe2+ + 2H2S →FeS2 + 4H+ + 2e-, liberates two electrons and is exergonic, making energy readily available for other chemical reactions to happen. Amongst the possible chemical pathways, the surface reactions can form amino acids. Most amino acids could in principle be formed. However, the first amino acids that would be formed on pyrite surfaces, and which by tentative implication would be the first to be included in an ancient genetic code, would be serine and aspartic acid. Serine would be immediately cleaved to produce glycine . In experimental syntheses of Wächtershäuser's theory, some amino acids were actually obtained: glycine, alanine and serine , and alanine, glutamic acid, phenylalanine, and tyrosine [23, 24].
Another chemo-autotrophic scenario that is worth discussing actually occurs at the alkaline hydrothermal vents . Arguably, this setting comprises the most comprehensive view of the origin of energy metabolism. Geosynthesis in the Earth's crust powers serpentinization: silicates of magnesium and iron are oxidised by water and produce H2. This molecular hydrogen is later released at the hydrothermal vents, where it reacts with CO2 and iron sulphide (FeS), forming thio esters. These compounds are the basis for the synthesis of many compounds and cofactors, and also couple the activation of acetate with phosphate to give acetyl-phosphate, which can be further hydrolysed to end up in acetate (CH3COOH). As in the iron-sulfur scenario, FeS can also catalyse the oxidative amination of α-keto acids, resulting in alanine, glycine, and aspartic and glutamic acids. Notably, purines and pyrimidines can also be formed from aspartic acid and glycine (a pathway that involves acyl-phosphates) .
In summary, these different scenarios provide contextual frameworks that support the synthesis of amino acids. The question of the yield amino acid production still stands; even if these theories reveal plausible pathways for de novo amino acid synthesis, experimental assessments have shown that the yield is rather poor compared to the requirements of modern organisms. But was the relatively low abiotic yield of amino acids enough to sustain protobionts based on RNA metabolisms? The answer to that question strongly depends on the role of amino acids at the moment when the genetic code was established.
The Stickland reaction
The role of the amino acids is inevitably associated with the structural and catalytic nature of proteins. Given their small size, it would be absurd to consider that free amino acids could have a structural role. But because of their physical-chemical properties, free amino acids associated with RNA adaptamers could aid catalysis . The theory of the coding coenzymatic handles (CCH) argues that the amino acids were covalently attached to trinucleotides, in a way that is reminiscent of metabolic cofactors, such as nicotinamide adenine dinucleotide (NADH) or ATP for example [7, 26]. The trinucleotides served as "handles" through which ribozymes non-covalently attached to the amino acids. In this way, the amino acids could be re-used to aid catalysis . Any unambiguous association between triplets and the amino acid repertoire would ensure that correct catalytic factors would be used for specific functions. This assumes that a modestly-varied repertoire of amino acids was already available. The synthesis of most catalytically important amino acids is very elaborate, and their abiotic yield is negligible (except for aspartic acid), a fact that necessarily postpones the catalytic functions of the amino acids to later historical stages.
Another fundamental role of protein and amino acid metabolism is nutrition. Amino acids are oxidised via the citric acid cycle and converted into urea (which is disposed of) and pyruvate and other keto acids of the citric acid cycle which are used in the anabolic synthesis of other compounds. Amino acids have a similar oxidation state to that of glucose, suggesting that they can undergo fermentation, and in extant organisms their breakdown fosters ATP synthesis.
There are several catabolic pathways of amino acid degradation. Amongst these, the Stickland reaction is the most efficient, by coupling the oxidation of one amino acid with the reduction of another. The amino acids that are oxidised undergo a deamination by the amino acid dehydrogenases, losing in addition two electrons and two protons [28–30]; the resulting keto acid then loses one carbon to CO2, and leads to ATP synthesis by substrate-level phosphorylation (Figure 3A) [9, 29–31]. This reaction is then coupled to the "reductive branch" by transferring the electrons and the protons to the other amino acid, which undergoes a reductive deamination, catalysed by a reductase [32–35]. The reduced keto acid has the same number of carbons as the acceptor amino acid. Normally, the reductive branch does not lead to ATP (unless both substrates are glycine), (Figure 3B) .
The overall amount of free energy varies according to the specific pair of amino acids undergoing fermentation, and to the stoichiometry of the reaction. For example, fermenting four moles of glycine leads to 3 moles of ATP; fermenting glycine and alanine (in a stoichiometric ratio of 3:1 moles) leads to 1.7 moles of ATP. Both result in lower yield than the fermentation of two moles of glucose, which leads to 5 moles of ATP, but still efficient when compared with the yield of fermenting other carbohydrates (e.g. 3 moles of lactate gives a yield of 2.3 moles of ATP) [[36, 37], Ch. 12].
The energetic yield is in fact sufficient to sustain organisms that employ exclusively amino acids as energy sources. The most compelling example is the genus Clostridium which comprises chemoautotrophic, anaerobic, bacteria . Various species may or must use amino acids as a carbon source [9, 30, 35, 39, 40]. However the genus Clostridium is not a monophyletic group [40, 41], which indicates that the Stickland reaction may have independent origins (an unlikely possibility because of the sequence similarity of the dehydrogenates and reductases ), has been lost in other genera, or was acquired by lateral transfer. Amongst the species we find thermophilic and alkaliphilic, some of which are associated with hydrothermal vents [37, 38, 40, 42, 43]. This hardly places the extant Stickland reaction pathways in a deep ancestral branch of the tree of life, but it is conceivable that analogous, if not ancestral, versions of Stickland reaction pathways not only existed before the code was established, but could also have played a role in the establishment of the code, before the catalytic role was implemented and translation evolved.
Evolution of the adapters
The problem of the usage of amino acids is only one side of the coin. The other side is how amino acids were assigned to coding triplets. Even if the role of amino acids were taken for granted, as with CCH or the Stickland fermentation, the question remains as to which factors determined such assignments. In any case, it can be assumed that at some stage the assignment involved a covalent bond between the amino acid and the proto-adapters. It has been debated whether the handles composed ancient codons or anticodons [5, 26], but stereochemical associations favour the idea that these were anticodons [5, 6]. These co-enzymatic handles later elongated to form mini-helices. Helical structures could have conferred structural stability to the proto-adapters (tRNA precursors), possibly for coding functions. The ideas introduced in this article assume a mini-helical stage for the proto-adapters. The mini-helices eventually evolved by successive and recursive duplications to form the modern adapters with their cloverleaf structure, the tRNAs (Figure 2, ). This idea is founded in a recently proposed mechanism. It argues that the ancestor to the adapters was a primordial palindrome gene composed of 11 nucleotides. This small hairpin would be composed of two block sequences: a coding triplet (the CCH, say, GCC), and a replication tag (something like a promoter sequence) with sequence 5'-DCCA-3' (where D is either an A or a U). If the pre-coding triplet is linked between the tag and the tag's complementary sequence, a small hairpin of 11 nucleotides is formed: UGGDGCCdCCA (where d denotes the complement of D). Now, the tag plus triplet (7 nucleotides) is complementary to the 11 nucleotides block, and can be ligated at one of its ends and form an 18 nucleotides hairpin. If this pattern is iterated, then the 76 nucleotides long molecule, the tRNA, can be built . Notably, in each iteration of this duplication-elongation mechanism two proto-adapter molecules are co-evolving (one in each strand), which supports the long-standing hypothesis that amino acids were included in the code in pairs [6, 18, 44–47].
Presentation of the hypothesis
Complementary anticodons are assigned to amino acids that are conjugated Stickland pairs.
In other words, the suggestion is that the origin of the genetic code traces back to the mutual redox deamination of the amino acid pairs to synthesise high-energy intermediates, such as acetyl phosphate and other related phosphate compounds.
Testing the hypothesis
The association between Stickland pairs and complementary anticodons is expected to hold more strongly for the primordial amino acids. Later additions could have been affected by other factors, especially when the bioenergetic pathways had already evolved, and departed from amino acid fermentation.
This initial amino acid composition is supported by the bias in the amino acid use of ribosomal proteins . In other words, when the transition to proteins came, the simpler amino acids were used preferentially over the simpler ones; in particular glycine, alanine, and asparagine are overrepresented on the deep branches, although the case for valine, lysine and arginine remains unsolved (i.e. there seems that there was no significant increase in their usage, and that this has remained more or less constant) . Here we find some evidence that supports the hypothesis; some of the anticodons for glycine (NCC) are complementary to some anticodons of alanine (NGC). This pair of amino acids is Stickland reactive (in fact amongst the most efficient pairs). A second step in the extension of the code is given by including valine and aspartic acid, which allow the associations gly-ala, asp-val and asp-ala, which are Stickland reactive, and their templates (NGC and NUC, respectively) form a small cross-catalytic system (Figure 6A-C). Thus the addition of two more adapters results not in two reactive pairs, but in three.
If an amino acid has the tendency to give a good energetic yield, it is expected to pair with several other amino acids of conjugate role. Strictly speaking, amino acids that tend to be good donors or good acceptors are expected to be assigned to anticodons that are flexible in pairing with other anticodons. For instance, it has been proposed that originally, glycine was encoded by the anticodon NCN, which would match the original anticodon of alanine, NGN. The latter additions valine and aspartic acid would have NAN and NUN respectively.
Considering an extension of the amino acids included in the code (at least as implied by the products of Miller-Urey synthesis) reinforces the correlation amongst complementary anticodons and reactive amino acid pairs. For example the addition of the next two amino acids, namely isoleucine and threonine (electron donors in the Stickland reaction), is accompanied by the implementation of new anticodons (NAU and NGU), which are still complementary to pre-existing anticodons of the electron acceptors (i.e. aspartic acid, NUN, and glycine, NCN; Figure 7). At this point, the "symmetry" of the primaeval code would be broken at the third position of the anticodons. However the anticodons for the electron acceptors (which are more numerous that donors), would be left intact. In this way, the multiplicity of the reactions would be maintained. Again, the number of possible reactions is increased, in a combinatorial manner.
The overall yield of ATP is simply the sum of the yield of each reaction. One mol of ATP gives roughly ΔG° = -80 KJ/mol.
Sequential bootstrapping was also performed in order to evaluate whether there is a preferential order of addition of amino acids. The different 18 combination of Stickland donors and acceptors of the Miller amino acids all lead to distributions of ATP yields with means that are different from zero (p < 10-30), which is to be expected. However, further additions of individual donors or acceptors, or pairs of them, resulted in a significant increase of the mean yield of ATP, whilst adding non-reactive amino acids significantly decreased the mean yield of ATP. This is reassuring, because it reflects the fact that the amino acids that can undergo fermentation, and can drive the expansion of the proto-code. But based only on the distribution of yield, this test does not allow us to derive any conclusions about the sequential increase of the amino acid repertoire.
In order to account for historical factors, a preliminary evaluation of the role of selection on variation was performed by implementing a Monte Carlo simulation. The central assumption is that there is a mutation mechanism that generates variants of the code, and a mechanism to select amongst these variants. For the mutations a swap-or-replace mechanism is assumed; two anticodon triplets are randomly selected, and if they bear different amino acids, the assignments are swapped. If they are the same, one of the amino acids is randomly substituted by any other present in the code. Selection acts according a Boltzmann factor, exp [-80(Δ[ATP])/RT], where the Δ[ATP] is the difference in ATP yield amongst the code that is implemented at any given time, and a mutated one; R is the gas constant and T the temperature. Then, "evolution" is allowed to proceed as a Metropolis algorithm.
Evolved proto-codes using the primaeval amino acids
# anticodons (SE)
G+U content (SE)
Evolved proto-codes, optimal solutions
# anticodons (SE)
G+U content (SE)
Evolved proto-codes, sub-optimal solution
# anticodons (SE)
G+U content (SE)
Summarising, the principle of amino acid fermentation via adapter pairing can lead to particular structures of the code, and explain some features of redundancy. The results are statistically significant and reproducible.
Possible experimental tests
There are two feasible experimental tests that can give support to the hypothesis above. First, notice that the "Stickland role" of any amino acid can depend on the actual partner. The classification shown in Table 1 is a compilation from the literature. In most of the experiments the amino acids were classified when tested against "universal" donors (usually alanine) or acceptors (usually glycine or proline). Ultimately, whether an amino acid acts as an electron donor or acceptor depends on its oxidation state relative to its partner, on their reactivity, and of the enzymes that catalyse the redox reactions. The electron-acceptor amino acids in all cases have higher oxidation states that their Stickland donors. This suggests that the roles of specific partners could be inverted, an observation supported by experimental results. For example, arginine can revert the roles of proline or lysine, and act as an electron acceptor when reacting with these [31, 52], which is reasonable when we consider that arginine's oxidation state is higher than that of proline or lysine. Therefore, testing the reactivity and free energy of specific pairs of amino acids in the extant species that carry the Stickland reaction can convey more information about (a) fermentation pathways and reaction mechanisms that could have also been employed before the code was established, and (b) the energetic yield of these reactions.
The second possible test is related to the catalysis of the reactions. It is assumed that before the establishment of the genetic code ribozymes performed catalytic functions. The redox pathways, even when energetically favourable tend to be limited by very slow reactions and must be aided by cofactors and ribozymes. It would be possible to test some of the steps of the Stickland reaction, such as the deamination of the amino acids and the phosphorylation of the keto acids. Could ribozymes be evolved in vitro to catalyse these functions? This is a very interesting question to be explored in the future, for which there is an enigmatic starting point: glutamate dehydrogenase (which oxidises glutamate, as in the first step of the Stickland reaction) is an enzyme that has notable sequence homology to the synthetases [47, 53]. This leads to the speculation that both functions could be somewhat related. It is a feasible idea to attempt to evolve the flexizymes (but in general, any other ribozyme) to have the dehydrogenase function.
Implications of the hypothesis
Why would the proto-adapters form complementary RNA complexes? The attachment of the amino acids to the proto-adapters can allow their spatial coordination in such a way that ribozymes and/or cofactors can catalyse their coupled redox deamination.
If this coordination does not include a covalent bond to the RNAs, it is hard to rationalise any structural mechanisms, due to the immense possibilities of coordination for which we currently know no constraint. Furthermore a non-covalent mechanism could be limiting because of the slow speed of the reaction rates. But if the amino acids are covalently attached, then mechanisms for this reaction can evolve due to steric constraints.
Amino acids attached to an RNA complex (say, a double strand) could be more easily oriented and prone to be catalysed. Current tRNAs have their amino acids attached at the 3' end of the molecule, i.e., at position 76, whereas the encoding triplet in the stem is located between positions 70 and 72. In a double stranded RNA of 11 nucleotides the amino acids that are attached at the 3' end of the RNA do not lie close to each other.
Figures 11C,D show a structural detail of the anchoring nucleotides facing each other, and arrows indicate the 2' point of attachment.
Towards a reaction mechanism
The structural constraint posed above, although reasonable from the structural point of view, poses two problems. The first is that the deaminations are less likely to happen than for free amino acids. Current dehydrogenases do not form a covalent bond with the amino acid, for which oxidative deamination is more likely . Thus if the Stickland reaction were mediated by ribozyme in an analogous way as the dehydrogenases, then the structural constraints described above would complicate matters. In fact, we can understand why. Because the ester bond displaces the electron cloud, the α-carbon is less electronegative; this impedes its deprotonation, making it unlikely that the nitrogen transiently forms a double bond with the α-carbon (the latter double bond is required for the nucleophilic attack of oxygen from water). Thus if covalently bonded, the amino acids are more stable than the keto acids . However this does not take into account that the side chain of the partner amino acid can aid deprotonation by interacting with the nitrogen moiety. Naturally, there must be some cofactors (such as NAD+, or NADPH+) to mediate the electron transfer, and ribozymes that allow the reaction to proceed.
The following step is the reductive deamination of the other amino acid. This is a complicated reaction, and how the precise mechanism proceeds remains unclear even for the extant reductases [32–34]. The central limitation is that this deamination requires a very strong oxidising agent, which in the reductases involves intermediates between selenium esters and the deaminated glycine  (in principle an analogous bond could be established by thioesters).
The second complication is that after the deamination, the keto acids remain attached to the ribose. This is an important factor, because ATP can be synthesised only from the remaining keto acids, which need to be detached from the ribose. The ester bond between the carboxyl group and the ribose has to be broken (to result in a deoxyribose), by direct or indirect oxidative phosphorylation of the keto acid.
Relation to the stereochemical hypothesis
The stereochemical theory suggests a physical and structural relation between the coding triplets (either codons or anticodons) and their cognate amino acids [, Chs. 6-7]. Based on sequence and structural data of riboswitches, it has been calculated that for some amino acids (arginine, histidine, isoleucine, leucine, phenylalanine, tryptophan and tyrosine) the chances that the coding triplets are evenly distributed in the binding site are small. The implication is that there is a significant association between the amino acids and some of their coding triplets (either codons or anticodons). Although only a fraction of the possible coding triplets (21%) show significant associations , these could have been an important determinant for coding, at least for some of the assignments [4, 6].
Given that the Stickland pairing is correlated to anticodon pairing, in any random sample containing complementary anticodons the chances that Stickland pairs are formed are high. Thus the appearance of the two Stickland pairs in the stereochemical associations might just be casual. For instance, of the anticodons that are significantly associated with their amino acids in the riboswitch data, two complementary anticodon pairs appear: GUG (histidine) with AUA (leucine), and AUA with leucine (AUA) with GUA (tyrosine), although there is no data on whether these amino acid pairs are Stickland reactive. On the other hand, this would make the two theories compatible.
If both factors are taken into account, (Stickland pairing and stereochemical affinity), coding is more constrained; whilst the choice of some anticodons might have been determined stereochemically , their complementary pairs can be assigned to an amino acid that is an efficient Stickland partner. If this were true, it could explain why only a fraction of the amino acids are stereochemically assigned to their anticodons. Furthermore, it is also possible that the stereochemical factors bias the choice, purging the code of some naturally occurring amino acids that were not biochemically useful (for energy or otherwise). The analyses to support the stereochemical theory did not include the amino acids that seem to be most relevant for the initiation of the code, which are the simplest to synthesise abiotically, namely glycine, alanine, aspartic acid and valine, and other Miller amino acids. This does not discredit the stereochemical theory, but it does lead to some conflicts.
The stereochemical theory is an ideal precursor to the CCH hypothesis since it establishes some non-covalent associations that could become covalently linked in a later stage of evolution. In this way the non covalent associations would give birth to the catalytic cofactors. This view has its merits, but in the context of this article the sequence of events is somewhat reversed. Simple amino acids were used as energy sources (rather than for catalysis), and were initially assigned randomly. The more complex amino acids were included later in the code when biosynthetic pathways were available to synthesise them, most likely for catalysis. This is supported by phylogenetic analyses that showed that most of the species that have strong stereochemical associations to their coding triplets increased their representation much later in evolution, after the last universal common ancestor was established .
Other amino acids favoured in the stereochemical theory are underrepresented in the deep branches indicating that despite having good catalytic properties, they were hardly used in the nascent proteins. Thus, by the time these amino acids were included in the code, a more primitive code had to exist, and therefore the covalent associations to the proto-adapters (or even ancestral versions of the modern adapters) had to be readily established. Also, if it turns that the simple amino acids do not have a significant stereochemical association with their coding triplets, it is hard to argue for the case arginine and lysine, since they cannot be formed by simple abiotic processes, and thus cannot impose constraints in the early evolution of coding.
Relation to the abiotic and early metabolisms
It has been argued that the citric-acid cycle might have originated in the opposite way [15, 56, 57]; instead of liberating CO2, it would employ it in a biosynthetic and reductive direction (as cyanobacteria and many extremophiles do), in accordance with the reductive atmosphere scenario [11, 21]. This is similar to the scenario of the alkaline vents where CO2 is employed for the synthesis of the organic precursors. This theory is appealing not only because of it chemical basis [15, 25], but also because of the geological context, which is know to exist [25, 58]. It was mentioned in the introduction that it is sound to assume that the code originated around the time of the LUCA. If that is the case, were the early protobionts free from the geochemical substratum, or were they still strongly dependent on it? It makes a big difference for the following reasons:
This means that under hydrothermal conditions the Stickland hypothesis does not work. What is worse, the whole theory of the concerted origin breaks down on energetic grounds, because proto-adapters smaller than about 60-70 nucleotides would not attain a stable fold (I am assuming the best case scenario where the proto-adapters are completely self-complementary, and therefore would have 30 to 35 paired bases, giving a melting temperature of about 100°C; interactions with ions and cofactors are being ignored). The situation is different away from the hydrothermal vents, where the double stranded proto-adapters are stable (the melting temperature is higher than the environmental), and amino acid fermentation is energetically favourable [59, 60].
This allows us to hypothesise that there had to be a period where the autotrophic protobionts had to experience a cooler environment. This could have been a geologically cooler period, or perhaps autotrophs that gained independent compartmentalisation eventually migrated away from the mineral substrates at the vents towards cooler environments. In cooler environments the reaction rates are slower and any factor that increase catalysis, like the complex amino acids, are selectively advantageous, and thus proto-adapters could evolve to become CCHs. These organisms that were conveyed with the new catalytic innovations increased in frequency and spread. Upon confronting warmer environments again, an increase of the size of the adapters was necessary in order to allow thermal stability. Somewhere during those transitional stages, peptides, rather than cofactors, took over metabolic functions.
Evolution of the genetic code
The code within the codon  reveals that codons GNN code for the four most abundant of the Miller amino acids, hinting at an ancient origin. As has been discussed above, these four molecules, if provided in abundant quantities, are enough to fuel metabolism. Although other Stickland partners can be found in the Miller amino acids, their negligible yield makes it unlikely that they could fully complement the bioenergetic requirements. But the code within the codon also reveals that in each biosynthetic pathway, amino acids derived from the primordial ones are assigned to codons that conserve their first position. For instance, codons starting with A code for amino acids synthesised from aspartic acid (asparagine, threonine, isoleucine, methionine, and lysine). Codons starting with C code for amino acids synthesised from glutamic acid (glutamine, proline and arginine). Codons starting with U code for amino acids synthesised by the Shikimate pathway (serine, phenylalanine, tryptophan and tyrosine). This strict pattern in the evolution of the code and of the amino acid repertoire was previously noticed and formulated as the coevolutionary theory of the genetic code [62, 63], which states that precursor-product amino acids are assigned to codons that differ in only one base pair. The coevolution theory explicitly suggests the the distinct amino acids were not included in a random way in the code, but rather, that their biosynthesis evolved.
Assignation of biosynthetically derived amino acids to new anticodons
Precursor amino acid
New assigned anticodon
Derived amino acid
This begs the question of whether the biosynthetic modifications improve the fermentation yield. By changing the oxidative potential in the biosynthesis, the Stickland role can be altered, as in the case of the interconversion of serine and glycine, or aspartic acid and asparagine. In the other cases the Stickland role is conserved. At the moment the evidence is not conclusive, although by considering posterior modifications to the code we might gain considerable insight. Nevertheless the biosynthesis of amino acids could have evolved after selective pressure was established by the use of amino acids that were scarce but beneficial, and which were therefore already assigned to some anticodons.
The Oparin-Haldane hypothesis, and Miller's experiment were major contributions to the understanding of the material basis of life. However, after the discovery of the genetic code, it became obvious that it is much simpler to synthesise the simple building blocks of life, than to assemble them in units that are capable of evolution. The RNA world has become one of the central paradigms for this early stage of life, but how it emerged remains unexplained. The origin of genetic code is one of the most puzzling questions about the transition from this RNA world to the modern modes of life. The pervasiveness of the genetic code in the tree of life, with its minor deviations, suggests that it must date back to the last universal common ancestor.
The hypothesis presented in this article complements the current understanding stating that the amino acids and proto-adapters were added to the code in pairs [6, 44, 45]. Stickland pairing is an appealing candidate mechanism because (a) it imposes constraints to the amino acids to be assigned to newly evolved proto-adapter pairs, and (b) it favours redundancy by assigning good electron donors or acceptors to proto-adapters in a way that allows it to react with several amino acids with complementary Stickland role. The evidence that supports or that is in favour of the hypothesis can be summarised in the following points:
Bacteria of the genus Clostridium employ pairs of amino acids as energy sources.
The energetic yield of amino acid fermentation is comparable to that of carbohydrate fermentation .
In the abiotic syntheses of amino acids, Stickland pairs are readily produced.
The composition of the ribosomal proteins of early organisms is biased towards the use of the simpler amino acids, specially alanine, asparagine, and glycine .
The code within the codon suggests the ancient origin of the four basic amino acids alanine, aspartic acid, glycine and valine .
A significant correlation exists between the number of complementary anticodons of an amino acid and the free energy of the Stickland reaction (Figure 8).
By increasing the number of amino acids used as substrate, the numbers of possible reactions is increased in a combinatorial manner.
The multiple complementarity of the proto-adapters (by considering U-G pairs) increases the synonymy of the reactions.
As mentioned above, we need to invoke a specific evolutionary model in order to account for plausible evolutionary histories. This in itself is an independent research subject; for example, we might make different assumptions about the modes of selection: did it act in a directional way, or was it stabilising? What kinds of costs could have been involved in the inclusion of new amino acids to the early metabolism? To what extent were the codes heritable? What is the distribution of mutation effects on the adapters? Questions such as these are relevant in determining the details of the associations. Fortunately, these are accessible by combining RNA models with state of the art of evolutionary genetics theory.
But even sparing from these details, the theory proposed here helps us to understand the early steps in the establishment of the code. It explains stages prior to the CCH in terms of bioenergetic advantages to a hypothetical protobiont living prior to proteins being implemented for enzymatic and structural roles. Although we can not know with certainty what actually happened, the ideas exposed in this article give a testable coherent picture about some factors that could have moulded the genetic code.
I am grateful to J.P. Bollback, C. Guet and M. Sixt for our discussions and comments on the manuscript, to J. Schwarz for his help with the chemistry, and to T. Ellis for copyediting the manuscript. I would like to acknowledge E. Szathmáry for kindly hosting me at the Collegium Budapest for a short visit, where the stimulating exchange on the origins of the genetic code was ignited. The author was supported by the ERC-2009-AdG Grant for project 250152 SELECTIONINFORMATION.
- Maynard Smith J, Szathmáry E: The major transitions in evolution. 1997, Oxford, UK: Oxford University PressGoogle Scholar
- Copley SD, Smith E, Morowitz HJ: A mechanism for the association of amino acids with their codons and the origin of the genetic code. P Natl Acad Sci USA. 2005, 102 (12): 4442-4447. 10.1073/pnas.0501049102.View ArticleGoogle Scholar
- Woese CR: The Genetic Code. 1967, Harper & RowGoogle Scholar
- Yarus M, Caporaso J, Knight R: Origins of the genetic code: the escaped triplet theory. Annu Rev Biochem. 2005, 74: 179-198. 10.1146/annurev.biochem.74.082803.133119.PubMedView ArticleGoogle Scholar
- Yarus M, Widmann JJ, Knight R: RNA-Amino Acid Binding: A Stereochemical Era for the Genetic Code. J Mol Evol. 2009, 69 (5): 406-429. 10.1007/s00239-009-9270-1.PubMedView ArticleGoogle Scholar
- Rodin AS, Szathmáry E, Rodin SN: On origin of genetic code and tRNA before translation. Biology Direct. 2011, 6: 14-10.1186/1745-6150-6-14.PubMedPubMed CentralView ArticleGoogle Scholar
- Szathmáry E: Coding coenzyme handles: a hypothesis for the origin of the genetic code. P Natl Acad Sci USA. 1993, 90 (21): 9916-9920. 10.1073/pnas.90.21.9916.View ArticleGoogle Scholar
- Schwartz A, Schäfer R: New amino-acids, and heterocyclic-compounds participating in stickland reaction of clostridium-sticklandii. Arch Mikrobiol. 1973, 93 (3): 267-276. 10.1007/BF00412026.PubMedView ArticleGoogle Scholar
- Barker HA: Amino Acid Degradation by Anaerobic Bacteria. Annu Rev Biochem. 1981, 50: 23-40. 10.1146/annurev.bi.50.070181.000323.PubMedView ArticleGoogle Scholar
- Miller SL: A production of amino acids under possible primitive earth conditions. Science. 1953, 117 (3046): 528-529. 10.1126/science.117.3046.528.PubMedView ArticleGoogle Scholar
- Miyakawa S, Yamanashi H, Kobayashi K, Cleaves H, Miller SL: Prebiotic synthesis from CO atmospheres: Implications for the origins of life. P Natl Acad Sci USA. 2002, 99 (23): 14628-14631. 10.1073/pnas.192568299.View ArticleGoogle Scholar
- Miller SL, Urey H: Organic compound synthesis on the primitive earth. Science. 1959, 130 (3370): 245-251. 10.1126/science.130.3370.245.PubMedView ArticleGoogle Scholar
- Ring D, Wolman Y, Miller SL, Friedman N: Prebiotic Synthesis of Hydrophobic and Protein Amino Acids. P Natl Acad Sci USA. 1972, 69 (3): 765-&. 10.1073/pnas.69.3.765.View ArticleGoogle Scholar
- Wächterschäuser G: Before Enzymes and Templates: Theory of Surface Metabolism. Microbiol Rev. 1988, 52 (4): 452-484.Google Scholar
- Martin W, Russell MJ: On the origin of biochemistry at an alkaline hydrothermal vent. Philos T R Soc B. 2007, 362 (1486): 1887-1925. 10.1098/rstb.2006.1881.View ArticleGoogle Scholar
- Huber C, Wächterschäuser G: alpha-hydroxy and alpha-amino acids under possible hadean, volcanic origin-of-life conditions. Science. 2006, 314 (5799): 630-632. 10.1126/science.1130895.PubMedView ArticleGoogle Scholar
- Eigen M, Schuster P: The Hypercycle: A principle of natural self-organization. Part C: The Realistic Hypercycle. Naturwissenschaften. 1978, 65 (7): 341-369. 10.1007/BF00439699.View ArticleGoogle Scholar
- Rodin SN, Ohno S, Rodin AS: On concerted origin of transfer RNAs with complementary anticodons. Orig Life Evol Biosph. 1993, 23 (5): 393-418. 10.1007/BF01582088.PubMedView ArticleGoogle Scholar
- Oparin AI: The origin of life. 1924, Moscow Worker PublishersGoogle Scholar
- Haldane JBS: The Origin of Life. Rationalist Annu. 1929, 148: 3-10.Google Scholar
- Kasting JF: Earth's early atmosphere. Science. 1993, 259 (5097): 920-926. 10.1126/science.11536547.PubMedView ArticleGoogle Scholar
- Rode BM: Peptide and the origin of life. Peptides. 1999, 20: 773-786. 10.1016/S0196-9781(99)00062-5.PubMedView ArticleGoogle Scholar
- Hafenbradl D, Keller M, Wächterschäuser G, Stetter KO: Primordial Amino Acids by Reductive Amination of alpha-Oxo Acids in Conjunction with the Oxidative Formation of Pyrite. Tetrahedron Lett. 1995, 36 (29): 5179-5182.View ArticleGoogle Scholar
- Huber C, Wächterschäuser G: Primordial reductive amination revisited. Tetrahedron Lett. 2003, 44 (8): 1695-1697. 10.1016/S0040-4039(02)02863-0.View ArticleGoogle Scholar
- Martin W, Baross J, Kelley D, Russell MJ: Hydrothermal vents and the origin of life. Nat Rev Microbiol. 2008, 6 (11): 805-814.PubMedGoogle Scholar
- Szathmáry E: The origin of the genetic code. Trends Genet. 1999, 15 (9): 223-229.PubMedView ArticleGoogle Scholar
- Kun A, Pongor S, Jordan F, Szathmáry E: Catalytic Propensity of Amino Acids and the Origins of the Genetic Code and Proteins. The Codes of Life: The Rules of Macroevolution. Edited by: Barbieri M. 2007, Springer, 39-59.Google Scholar
- Brunhuber N, Blanchard J: The Biochemistry and Enzymology of Amino-Acid Dehydrogenases. Crit Rev Biochem Mol. 1994, 29 (6): 415-467. 10.3109/10409239409083486.View ArticleGoogle Scholar
- Schneeberger A, Frings J, Schink B: Net synthesis of acetate from CO2 by Eubacterium acidaminophilum through the glycine reductase pathway. Fems Microbiol Lett. 1999, 177: 1-6. 10.1111/j.1574-6968.1999.tb13705.x.View ArticleGoogle Scholar
- Nisman B: The Stickland Reaction. Microbiology and Molecular Biology Reviews. 1954, 18: 16-42.Google Scholar
- Fonknechten N, Chaussonnerie S, Tricot S, Lajus A, Andreesen JR, Perchat N, Pelletier E, Gouyvenoux M, Barbe V, Salanoubat M, Le Paslier D, Weissenbach J, Cohen GN, Kreimeyer A: Clostridium sticklandii, a specialist in amino acid degradation: revisiting its metabolism through its genome sequence. Bmc Genomics. 2010, 11: 555-10.1186/1471-2164-11-555.PubMedPubMed CentralView ArticleGoogle Scholar
- Andreesen J: Glycine reductase mechanism. Curr Opin Chem Biol. 2004, 8 (5): 454-461. 10.1016/j.cbpa.2004.08.002.PubMedView ArticleGoogle Scholar
- Andreesen J: Glycine Metabolism in Anaerobes. Anton Leeuw Int J G. 1994, 66: 223-237. 10.1007/BF00871641.View ArticleGoogle Scholar
- Jackson S, Calos M, Myers A, Self WT: Analysis of proline reduction in the nosocomial pathogen Clostridium difficile. J Bacteriol. 2006, 188 (24): 8487-8495. 10.1128/JB.01370-06.PubMedPubMed CentralView ArticleGoogle Scholar
- Seto B: The Stickland reaction. Diversity of bacterial respiratory systems. Edited by: Knowles C. 1980, Boca Raton: CRC Press, 60-64.Google Scholar
- Lengeler J, Drews G, Schlegel H: Biology of the Prokaryotes. 12. 2009, Anaerobic Energy Metabolism, Wiley-BlackwellGoogle Scholar
- Orlygsson J, Houwen F, Svensson B: Thermophilic anaerobic amino-acid degradation - deamination rates and end-product formation. Appl Microbiol Biot. 1995, 43 (2): 235-241. 10.1007/BF00172818.View ArticleGoogle Scholar
- Wiegel J: Family I: Clostridiaceae. Bergey's manual of systematic bacteriology. 2009, Springer, 3: 736-738. 2Google Scholar
- Stickland L: Studies in the metabolism of the strict anaerobes (genus Clostridium). I. The chemical reactions by which CL. sporogenes obtains its energy. Biochem J. 1934, 28: 1746-1759.PubMedPubMed CentralView ArticleGoogle Scholar
- Brisbarre N, Fardeau M, Cueff V, Cayol J, Barbier G, Cilia V, Ravot G, Thomas P, Garcia J, Ollivier B: Clostridium caminithermale sp nov., a slightly halophilic and moderately thermophilic bacterium isolated from an Atlantic deep-sea hydrothermal chimney. Int J Syst Evol Micr. 2003, 53: 1043-1049. 10.1099/ijs.0.02471-0.View ArticleGoogle Scholar
- Rainey F, Hollen B, Small A: Genus I Clostridium. Bergey's manual of systematic bacteriology. 2009, Springer, 3: 738-828. 2Google Scholar
- Hernandez-Eugenio G, Fardeau M, Cayol J, Patel B, Thomas P, Macarie H, Garcia J, Ollivier B: Sporanaerobacter acetigenes gen. nov., sp nov., a novel acetogenic, facultatively sulfur-reducing bacterium. Int J Syst Evol Micr. 2002, 52: 1217-1223. 10.1099/ijs.0.01992-0.Google Scholar
- Pikuta EV, Hoover RB, Tang J: Microbial extremophiles at the limits of life. Crit Rev Microbiol. 2007, 33 (3): 183-209. 10.1080/10408410701451948.PubMedView ArticleGoogle Scholar
- Eigen M, Schuster P: The Hypercycle. 1979, Berlin, Germany.: Springer - VerlagView ArticleGoogle Scholar
- Rodin SN, Ohno S, Rodin AS: Transfer RNAs with complementary anticodons: could they reflect early evolution of discriminative genetic code adaptors?. P Natl Acad Sci USA. 1993, 90 (10): 4723-4727. 10.1073/pnas.90.10.4723.View ArticleGoogle Scholar
- Rodin SN, Rodin AS, Ohno S: The presence of codon-anticodon pairs in the acceptor stem of tRNAs. P Natl Acad Sci USA. 1996, 93: 4537-4547. 10.1073/pnas.93.10.4537.View ArticleGoogle Scholar
- Rodin AS, Rodin SN, Carter CW: On Primordial Sense-Antisense Coding. J Mol Evol. 2009, 69 (5): 555-567. 10.1007/s00239-009-9288-4.PubMedPubMed CentralView ArticleGoogle Scholar
- Rodin AS, Szathmáry E, Rodin SN: One ancestor for two codes viewed from the perspective of two complementary modes of tRNA aminoacylation. Biology Direct. 2009, 4: 4-10.1186/1745-6150-4-4.PubMedPubMed CentralView ArticleGoogle Scholar
- Rodin SN, Rodin AS: On the origin of the genetic code: signatures of its primordial complementarity in tRNAs and aminoacyl-tRNA synthetases. Heredity. 2008, 100: 341-355. 10.1038/sj.hdy.6801086.PubMedView ArticleGoogle Scholar
- Higgs PG: A four-column theory for the origin of the genetic code: tracing the evolutionary pathways that gave rise to an optimized code. Biology Direct. 2009, 4: 16-10.1186/1745-6150-4-16.PubMedPubMed CentralView ArticleGoogle Scholar
- Fournier GP, Gogarten JP: Rooting the ribosomal tree of life. Mol Biol Evol. 2010, 27 (8): 1792-1801. 10.1093/molbev/msq057.PubMedView ArticleGoogle Scholar
- Stadtman TC: On the metabolism of an amino acid fermenting Clostridium. J Bacteriol. 1954, 67 (3): 314-320.PubMedPubMed CentralGoogle Scholar
- LéJohn HB, Cameron LE, Yang B, Rennie SL: Molecular characterization of an NAD-specific glutamate dehydrogenase gene inducible by L-glutamine. Antisense gene pair arrangement with L-glutamine-inducible heat shock 70-like protein gene. J Biol Chem. 1994, 269 (6): 4523-4531.PubMedGoogle Scholar
- Illangasekare M, Sanchez G, Nickles T, Yarus M: Aminoacyl-RNA synthesis catalyzed by an RNA. Science. 1995, 267 (5198): 643-647. 10.1126/science.7530860.PubMedView ArticleGoogle Scholar
- Xiao H, Murakami H, Suga H, Ferré-D'Amaré AR: Structural basis of specific tRNA aminoacylation by a small in vitro selected ribozyme. Nature. 2008, 454 (7202): 358-361. 10.1038/nature07033.PubMedView ArticleGoogle Scholar
- Morowitz HJ, Kostelnik JD, Yang J, Cody GD: The origin of intermediary metabolism. Proc Natl Acad Sci USA. 2000, 97: 7704-7708. 10.1073/pnas.110153997.PubMedPubMed CentralView ArticleGoogle Scholar
- Fenchel T: Origin & early evolution of life. 2002, Oxford University PressGoogle Scholar
- Koonin E, Martin W: On the origin of genomes and cells within inorganic compartments. Trends Genet. 2005, 21 (12): 647-654. 10.1016/j.tig.2005.09.006.PubMedView ArticleGoogle Scholar
- Amend JP, Shock EL: Energetics of amino acid synthesis in hydrothermal ecosystems. Science. 1998, 281 (5383): 1659-1662.PubMedView ArticleGoogle Scholar
- Amend J, McCollom T: Energetics of biomolecule synthesis on early Earth. Chemical evolution II: from the origins of life to modern society. 2009, 63-94.Google Scholar
- Taylor FJ, Coates D: The code within the codons. Biosystems. 1989, 22 (3): 177-187. 10.1016/0303-2647(89)90059-2.PubMedView ArticleGoogle Scholar
- Wong JTF: A Co-Evolution Theory of the Genetic Code. P Natl Acad Sci USA. 1975, 72 (5): 1909-1912. 10.1073/pnas.72.5.1909.View ArticleGoogle Scholar
- Di Giulio M: The origin of the tRNA molecule: implications for the origin of protein synthesis. J Theor Biol. 2004, 226: 89-93. 10.1016/j.jtbi.2003.07.001.PubMedView ArticleGoogle Scholar
- Johnson AP, Cleaves HJ, Dworkin JP, Glavin DP, Lazcano A, Bada JL: The Miller volcanic spark discharge experiment. Science. 2008, 322 (5900): 404-10.1126/science.1161527.PubMedView ArticleGoogle Scholar
- Botta O, Glavin D, Kminek G, Bada J: Relative amino acid concentrations as a signature for parent body processes of carbonaceous chondrites. Origins Life Evol B. 2002, 32 (2): 143-163. 10.1023/A:1016019425995.View ArticleGoogle Scholar
- Bada J: Amino-Acid Cosmogeochemistry. Philos T Roy Soc B. 1991, 349-358.Google Scholar
- Rodin SN, Rodin AS: Origin of the genetic code: First aminoacyl-tRNA synthetases could replace isofunctional ribozymes when only the second base of codons was established. DNA and cell biology. 2006, 25 (6): 365-375. 10.1089/dna.2006.25.365.PubMedView ArticleGoogle Scholar
- Brown J, Doolittle W: Root of the universal tree of life based on ancient aminoacyl-tRNA synthetase gene duplications. P Natl Acad Sci USA. 1995, 92 (2441-2445): 2441-2445.View ArticleGoogle Scholar
- Betzel C, Lorenz S, Furste J, Bald R, Zhang M, Schneider T, Wilson K, Erdmann V: Crystal structure of domain A of Thermus flavus 5S rRNA and the contribution of water molecules to its structure. FEBS Letters. 1994, 351 (2): 159-164. 10.1016/0014-5793(94)00834-5.PubMedView ArticleGoogle Scholar
- Stickland L: Studies in the metabolism of the strict anaerobes (genus Clostridium). III. The oxidation of alanine by Cl. sporogenes. IV. The reduction of glycine by Cl. sporogenes. Biochem J. 1935, 29: 889-898.PubMedPubMed CentralView ArticleGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.