The hidden universal distribution of amino acid biosynthetic networks: a genomic perspective on their origins and evolution
- Georgina Hernández-Montes†1,
- J Javier Díaz-Mejía†1, 2, 3,
- Ernesto Pérez-Rueda1 and
- Lorenzo Segovia1Email author
© Hernández-Montes et al.; licensee BioMed Central Ltd. 2008
Received: 4 December 2007
Accepted: 9 June 2008
Published: 09 June 2008
Twenty amino acids comprise the universal building blocks of proteins. However, their biosynthetic routes do not appear to be universal from an Escherichia coli-centric perspective. Nevertheless, it is necessary to understand their origin and evolution in a global context, that is, to include more 'model' species and alternative routes in order to do so. We use a comparative genomics approach to assess the origins and evolution of alternative amino acid biosynthetic network branches.
By tracking the taxonomic distribution of amino acid biosynthetic enzymes, we predicted a core of widely distributed network branches biosynthesizing at least 16 out of the 20 standard amino acids, suggesting that this core occurred in ancient cells, before the separation of the three cellular domains of life. Additionally, we detail the distribution of two types of alternative branches to this core: analogs, enzymes that catalyze the same reaction (using the same metabolites) and belong to different superfamilies; and 'alternologs', herein defined as branches that, proceeding via different metabolites, converge to the same end product. We suggest that the origin of alternative branches is closely related to different environmental metabolite sources and life-styles among species.
The multi-organismal seed strategy employed in this work improves the precision of dating and determining evolutionary relationships among amino acid biosynthetic branches. This strategy could be extended to diverse metabolic routes and even other biological processes. Additionally, we introduce the concept of 'alternolog', which not only plays an important role in the relationships between structure and function in biological networks, but also, as shown here, has strong implications for their evolution, almost equal to paralogy and analogy.
Metabolism represents an intricate set of enzyme-catalyzed reactions synthesizing and degrading compounds within cells. It is likely that a small number of enzymes with broad specificity existed in early stages of metabolic evolution. Genes encoding these enzymes probably have been duplicated, generating paralog enzymes that, through sequence divergence, became more specialized, giving rise, for instance, to the isomerases HisA (EC:18.104.22.168) and TrpC (EC:22.214.171.124), which act in histidine and tryptophan biosynthesis, respectively [1–4]. Additionally, gene duplication can promote innovations, generating enzymes catalyzing functionally different reactions, such as HisA, HisF (EC:2.4.2.-) and TrpA (EC:126.96.36.199). The classic view of metabolism is that relatively isolated sets of reactions or pathways are enough for the synthesis and degradation of compounds. The new perspective views metabolic components (substrates, products, cofactors, and enzymes) as nodes forming branches within a single network [5, 6].
In the past few years, an increasing amount of information on metabolic networks from different species has become available [7–10], allowing for comparative genomic-scale studies on the evolution of both specific pathways [11, 12] and whole metabolic networks [13–16]. Collectively, these studies highlight the contribution of gene duplication in the evolution of metabolism. Nevertheless, analog enzymes - those catalyzing the same reaction, even belonging to different evolutionary families - have been suggested to play an important role on this process as well . This results, for instance, in three different types of acetolactate synthases (EC:188.8.131.52) acting in the biosynthesis of L-valine and L-leucine in Escherichia coli. Additionally, the modern perspective of metabolic processes has shown that evolutionary studies must include not only phylogenetic relationships among enzymes, but also the influence of some topological properties of metabolic networks [5, 6, 18–20]. One of these properties is the capability of metabolism to circumvent failures - for example, mutations promoting unbalanced fluxes - using alternative network branches and enzymes. Here, we introduce the term 'alternolog' to refer to these alternative branches and enzymes that, proceeding via different metabolites, converge in a common product. Some authors have suggested that alternative branches can contribute to genetic buffering in eukaryotes to a degree similar to gene duplication , but the role of these alternologs in the evolution of metabolism in other phylogenetic groups remains to be solved. In evolutionary terms, one can assume that the universal occurrence of some pathways and branches in modern species suggests that they existed in the last common ancestor (LCA). The evolution of these pathways and the emergence of paralogs, analogs and alternologs reflect an increased metabolic diversity as a consequence of increasing genome size, protein structural complexity and selective pressures in changing environments. In the evolution of amino acid biosynthesis, for instance, alternative pathways synthesizing L-lysine via either L,L-diaminopimelate or alpha-aminoadipate have been suggested to have developed independently in diverse clades [21–23]. The evolution of these pathways is closely related to the biosynthesis of L-arginine and L-leucine [22–24] and even to the Krebs cycle , but the origin of all these pathways is still under discussion. Diverse studies [6, 25, 26] have suggested that amino acids could be among the earliest metabolic compounds. However, two main questions have emerged from these studies: from what did their biosynthetic networks originate and how did they evolve? And how did gene duplication (paralogs), functional convergence (analogs) and network structural alternatives (alternologs) contribute to these processes? The purpose of this work is to broach these questions, combining both a network perspective and a comparative genomics approach. For this purpose we consider that the architecture of proteins preserves structural information that can be used to identify their relative emergence during the evolution of metabolism. Specifically, we identified a set of enzymes and branches that originated closer to the existence of the LCA, delimiting a core of enzyme-driven reactions that putatively catalyzed the biosynthesis of at least 16 out of the 20 amino acids in early stages of evolution. Additionally, we determined the contributions of biochemical functional alternatives to this core (paralogs, analogs, and alternologs) during the evolution of amino acid biosynthesis in diverse species.
Results and discussion
Biological distribution of amino acid biosynthetic networks
Nine amino acid biosynthetic pathways are widely distributed across the three domains of life, and eight of their branches probably occurred in the LCA
There are four branches to synthesize L-glycine. Two of them, involving the degradation of L-threonine (Figure 1), are partially distributed in Bacteria and Eukarya (Figure 2). In contrast, the other two branches, interconnected through 5,10-methylene-tetrahydrofolate, involve either the glycine-cleavage system or serine hydroxymethyltransferase (EC:184.108.40.206). Both branches are widely distributed across the three cellular domains (Figure 2). Indeed, EC:220.127.116.11 is one of the most widely distributed enzymes across all the species, probably as it also participates in folate biosynthesis, another broadly distributed pathway. Collectively, the distribution of these enzymes suggests that the LCA synthesized glycine via the branch of 5,10-methylene-tetrahydrofolate.
We found the five L-tryptophan biosynthetic enzymes widely distributed across the three domains of life, confirming previous reports . Nevertheless, we did not identify orthologs for these enzymes in animals (Figure 2), with the exception of Nematostella vectensis, a cnidaria representative of early stages in animal evolution . This indicates that some animals had a secondary loss of the L-tryptophan biosynthetic enzymes and also explains why this amino acid is essential for humans. Thus, the LCA probably was able to synthesize L-tryptophan in a similar fashion to contemporary species.
There are at least six L-proline biosynthetic branches (Figure 1). Three of them converge in L-glutamate γ-semialdehyde and, judging from their TDs, ornithine-δ-aminotransferase (EC:18.104.22.168) is the most widely distributed enzyme within this pathway, even in some archaeal genomes (Figure 2). The other two branches have been biochemically characterized, although their catalyzing enzymes are unknown. The sixth branch, which directly converts L-ornithine to L-proline via ornithine cyclodeaminase (EC:22.214.171.124), was found in some Archaea and scarcely in Bacteria and Eukarya (Figure 2). Further analyses are necessary to corroborate experimentally the activities of these archaeal open reading frames, because the putative EC:126.96.36.199 enzymes do not have the canonical catalytic residues involved in this activity, and little information is known about the EC:188.8.131.52 activity. Thus, the archaeal biosynthesis of L-proline remains enigmatic and makes it difficult to infer if the LCA was capable of synthesizing L-proline.
The biosynthesis of L-leucine consists of five reactions following a mainly linear pathway (Figure 1). Using the E. coli and M. jannaschii sequences for BRHs, we detected that putative enzymes catalyzing the first three reactions are widely distributed (Figure 2). These three enzymes belong to a group of duplicated genes catalyzing consecutive steps in the biosynthesis of three amino acids, L-lysine, L-leucine and L-isoleucine (Figure 3). The evolutionary relationships between L-lysine and L-leucine biosynthesis have been documented previously [23, 24, 29]: we found that L-isoleucine biosynthesis is also implied in this phenomenon. These duplicates together with those from L-arginine/L-lysine biosynthesis support our previous report on the importance of the retention of duplicated genes as groups, instead of as single entities, in the evolution of metabolism . The fourth reaction occurs spontaneously and does not require a catalyzing enzyme. Complementarily, the fifth step in E. coli is catalyzed by one out of the two analog branched-chain amino acid transferases (EC:184.108.40.206); one of them belongs to the D-amino acid aminotransferase-like PLP-dependent superfamily and is widely distributed across the three domains, including some animals. In contrast, the second EC:220.127.116.11 belongs to the PLP-dependent transferases superfamily and is sparsely distributed across genomes. Collectively, these observations suggest that the LCA was able to synthesize L-leucine-like contemporary species. Further biochemical characterization of animal open reading frames is necessary, as L-leucine is an essential amino acid for humans.
Structurally speaking, L-histidine and L-tryptophan biosynthesis are similar; both are mainly linear pathways diverging from anthranilate using EC:18.104.22.168 (Figure 1) and, given their wide distribution, they have been proposed to be ancient pathways. The L-histidine biosynthesis enzyme histidinol-phosphatase (EC:22.214.171.124) is the only enzyme from this pathway partially distributed across genomes (Figure 2). This is probably due to the existence of two analog EC:126.96.36.199 enzymes (S. cerevisiae- and E. coli-types). Both types are highly divergent in sequence, and when we relaxed the stringency of BRH analysis (increasing the threshold E-value from 10-6 to 10-1), we detected orthologs in 84% and 40% of the analyzed genomes for the S. cerevisiae and E. coli types, respectively. The other enzymes analyzed in this study are not affected by the stringency of BRHs. Additionally, we found that animals, with the exception of N. vectensis, have experienced a secondary loss of the L-histidine biosynthetic machinery (Figure 2). Taking these results together, we suggest that the LCA had the same L-histidine synthesis pathway as extant species.
Two out of the three L-threonine biosynthetic enzymes from E. coli were found across the three domains. We did not find any orthologs in Archaea when we performed a genome scan with the E. coli threonine synthase (EC:188.8.131.52) as seed. Alternatively, when we used as seed an M. jannaschii paralog with the same function, we identified orthologs in Archaea (Figure 2). Again, this finding reinforces the importance of using enzymes from multiple species as seeds. Some animals apparently lost the biosynthetic machinery for this amino acid, but N. vectensis retained it. We suggest that the LCA could synthesize L-threonine like contemporary species.
L-glutamine and L-glutamate
As depicted in Figure 1, the inter-conversion of L-glutamine and L-glutamate can be performed by many alternolog enzymes. Both paralog glutamate synthases, the NADH dependent (EC:184.108.40.206) and the NADPH dependent (EC:220.127.116.11), produce L-glutamate from L-glutamine, and are widely distributed across the three domains (Figure 2). In the reverse direction, from L-glutamate to L-glutamine, we found that glutamine synthetase (EC:18.104.22.168), which is ATP dependent, is also widely distributed across the three domains. This suggests that the LCA was able to inter-convert L-glutamine and L-glutamate. But it leaves one open question: was the LCA capable of producing these amino acids independently of each other? Similarly to glutamate synthases, both paralog glutamate dehydrogenases, the NAD(P)+-dependent (EC:22.214.171.124) and the NADP+-dependent (EC:126.96.36.199) enzymes, produce L-glutamate from 2-oxoglutarate and ammonia, and are also widely distributed across the three domains. On the other hand, all other reactions synthesizing L-glutamine use L-glutamate as substrate and are sparsely distributed. In summary, we suggest that the LCA was able to synthesize L-glutamate from 2-oxoglutarate and inter-convert it with L-glutamine, but it is difficult to determine if the LCA was able to produce this last amino acid independently of the former one.
There are at least four ways to synthesize L-cysteine (Figure 1). The most widely distributed, proceeding via cystathionine, uses cystathionine β-synthase (EC:188.8.131.52) and cystathionine γ-lyase (EC:184.108.40.206) and is documented as being eukaryotic-type, yet we found it distributed across the three domains (Figure 2). Alternatively, cystathionine-β-lyase (EC:220.127.116.11), cystathionine γ-synthase (EC:2.5.1.-) and O-succinylhomoserine(thiol)-lyase (EC:18.104.22.168) catalyze equivalent reactions and they are widely distributed in Bacteria and Eukarya. In contrast, an alternolog branch using EC:22.214.171.124 via O-acetyl-L-serine is sparsely distributed across genomes (Figure 2), while another branch without assigned enzymes (nor genes) uses O-acetyl-L-homoserine. These findings suggest that not all the L-cysteine biosynthetic pathways occurred in the LCA, but that the contemporary eukaryotic-like type could.
Eight amino acid biosynthetic pathways are partially distributed across the three domains of life, and five of their branches probably occurred in the LCA
L-lysine biosynthesis has been used largely to exemplify the existence of alternolog branches in amino acid biosynthesis [21–23]. Six alternative pathways can be recognized for the biosynthesis of L-lysine (Figure 1), grouped in two superpathways proceeding via either L,L-diaminopimelate or alpha-aminoadipate. The superpathway involving L,L-diaminopimelate has four alternolog branches, corresponding to L-lysine biosynthesis types I, II, III and VI in MetaCyc; they share a common set of six reactions catalyzed by widely distributed enzymes. Four of these enzymes catalyze the upper steps of the superpathway, from aspartate kinase (EC:126.96.36.199) to dihydrodipicolinate reductase (EC:188.8.131.52), and form the pairs of duplicated genes between the biosynthesis of L-arginine/L-lysine (Figure 3). The other two enzymes (EC:5.1.17 and EC:4.1.120) catalyze the lower portion of the superpathway. The TDs of enzymes catalyzing intermediate steps in these alternologs are as follow. In the type I pathway (E. coli-type), which is catalyzed by three enzymes, only N-succinyl-L,L-diaminopimelate desuccinylase (EC:184.108.40.206) is widely distributed across the three domains. In the type II pathway (B. subtilis-type), catalyzed by the other three enzymes, only tetrahydrodipicolinate acetyltransferase (EC:220.127.116.11) is widely distributed in Bacteria, while it is absent in Archaea and Eukarya. The type III pathway of Corynebacterium glutamicum (EC:18.104.22.168) appears constrained to some actinobacteria and firmicutes, while the recently discovered type VI pathway, formed by a single enzyme, namely L,L-diaminopimelate aminotransferase (EC:2.6.1.-), seems to be specific for plants. These results illustrate a general finding of this work: linear pathways seem to be more widely distributed than bifurcating ones. As described above, L-histidine, L-tryptophan and L-leucine pathways support this observation, and correlate with previous studies showing that within amino acid biosynthesis, larger pathways tend to have lower rates of change in their structure than shorter pathways . However, further studies on whole metabolic networks are necessary to assess the generality of this property in the evolution of metabolism. On the other hand, the second superpathway, proceeding via the degradation of alpha-aminoadipate, is formed by lineage specific type IV and V pathways that share a core of five reactions from homocitrate synthase (EC:22.214.171.124) to α-aminoadipate aminotransferase (EC:126.96.36.199). This core contains the four enzymes forming pairs of duplicated genes between the biosynthesis of L-leucine/L-lysine (Figure 3). The type V pathway, using N-2-acetyl-L-lysine (RXN-5181 to RXN-5185), was characterized in the Thermus-Deinocuccus lineage, and its representatives were found in Archaea and some Bacteria, while the type IV pathway, proceeding via saccharopine (EC:188.8.131.52 to EC:184.108.40.206), appears restricted to Eukarya and some Bacteria. Collectively, the TDs of these two superpathways show that alternative pathways have led the origin of the biosynthesis of L-lysine. None of these alternologs appears to be universally distributed and, thus, the LCA probably was not able to produce L-lysine using the set of enzymes analyzed here. Interestingly, both L-lysine biosynthetic superpathways retain groups of duplicated genes for the biosynthesis of L-leucine and L-arginine (Figure 3), which, as detailed above, probably occurred in the LCA. Thus, there is a possibility that L-lysine biosynthesis was incorporated into metabolism from L-leucine and L-arginine biosynthetic routes.
L-valine and L-isoleucine
The terminal four steps in the biosynthesis of L-valine and L-isoleucine employ a common set of widely distributed enzymes, from EC:220.127.116.11 to branched-chain amino-acid aminotransferase (EC:18.104.22.168) (Figure 4). This set was not found, however, in animals, again with the exception of N. vectensis. Complementarily, five alternolog branches can catalyze the initial steps of L-isoleucine biosynthesis, converging in 2-oxobutanoate, which is, in turn, a substrate of acetolactate synthase (EC:22.214.171.124) (Figure 1). We found that the canonical E. coli branch carrying out these steps via propionate uses EC:126.96.36.199 and EC:188.8.131.52 and is sparingly distributed among bacterial genomes. In contrast, the alternolog branch characterized in spirochaetes, proceeding via (R)-citramalate (Figure 1), uses isopropylmalate isomerase (EC:184.108.40.206) and β-isopropylmalate dehydrogenase (no EC number assigned), and both enzymes are widely distributed across the three domains (Figure 4). These results clearly exemplify that the E. coli canonical pathways are not necessarily the most widely distributed ones and, thus, alternolog pathways must be included in evolutionary analysis. Additionally, this branch participates in the retention of a group of duplicated genes catalyzing consecutive reactions in the biosynthesis of L-lysine, L-leucine and L-isoleucine (Figure 3). Taking together the wide distribution of the spirochaetes-like branch and the enzymes shared between L-valine and L-isoleucine biosynthesis, we suggest that the LCA and even contemporary species could combine these branches to synthesize both amino acids.
Chorismate is not an amino acid itself, but it is a key compound in the biosynthesis of aromatic amino acids and we consider the distribution of their catalyzing enzymes particularly interesting. The biosynthesis of chorismate comprises seven steps, the last two being catalyzed by two widely distributed enzymes, 3-phosphoshikimate-1-carboxyvinyltransferase (EC:220.127.116.11) and chorismate synthase (EC:18.104.22.168). Complementarily, the first two steps are catalyzed by enzymes widely distributed in Bacteria and some Eukarya, but absent in Archaea. A recent report suggesting a novel pathway for the biosynthesis of aromatic amino acids and p-aminobenzoic acid in the archaeon Methanococcus maripaludis helps to understand this distribution . Additionally, three intermediate steps are catalyzed by scarcely distributed analog and alternolog enzymes as follows. First, the transformation of 3-dehydroquinate to 3-dehydro-shikimate can be catalyzed by two analog 3-dehydroquinate dehydratases (EC:22.214.171.124). B. subtilis possesses both analogs, while Archaea, some Eukarya and a few Bacteria carry only the type II enzyme (Figure 4) belonging to the aldolase (TIM-barrel) superfamily. In contrast, the majority of Bacteria, including E. coli, uses the type I enzyme (Figure 4) belonging to the 3-dehydroquinate dehydratase superfamily. Second, in E. coli there are two paralogs catalyzing the conversion of 3-dehydro-shikimate to shikimate. One of them, NADP+-dependent EC:126.96.36.199, is widely distributed, while EC:188.8.131.522 (using either NAD+ or NADP+, and either quinate or shikimate) is sparsely distributed. In contrast, B. subtilis has only the NADP+-dependent shikimate dehydrogenase and, when its sequence is used as a seed for BRHs, we found more orthologs than with the E. coli counterparts (Figure 4). This finding is probably caused by cross-matches between the E. coli paralogs during the construction of TDs. Third, the transformation of shikimate to shikimate-3-phosphate can be catalyzed by two analog shikimate kinases (EC:184.108.40.206). The archaeal-type belongs to the GHMP kinase superfamily, while the bacterial/eukaryotic-type belongs to the superfamily of P-loop containing nucleoside triphosphate hydrolases. Interestingly, there is an almost perfect anti-correlation between the TDs of these enzymes (Figure 4). Animals, including N. vectensis, have lost all enzymes catalyzing intermediate steps in chorismate biosynthesis, supporting the fact that aromatic amino acids (L-histidine, L-trypthopan, L-phenylalanine, and L-tyrosine) are essential for humans. Summarizing, we found that the lower portion of chorismate biosynthesis, converting 3-dehydro-shikimate to chorismate, is widely distributed across the three domains, suggesting that it probably occurred in the LCA. In contrast, the upper and intermediate portions of this route appear to have originated independently in specific lineages during evolution.
L-aspartate and L-asparagine
The biosynthesis and inter-conversion of L-aspartate and L-asparagine are mediated by a diverse set of alternolog enzymes (Figure 1), most of which have been characterized in E. coli and are sparsely distributed. Nevertheless, aspartate aminotransferase (EC:220.127.116.11) and pyruvate carboxylase (EC:18.104.22.168) are able to produce L-aspartate from pyruvate, via oxaloacetate, and both enzymes are widely distributed across the three domains (Figure 4). Complementarily, the conversion of L-aspartate to L-asparagine can be carried out by three asparagine synthetases, two of which are glutamine dependent (EC:22.214.171.124) while the other is ammonia dependent (EC:126.96.36.199). Both EC:188.8.131.52 type 1 and EC:184.108.40.206 belong to the adenine nucleotide alpha hydrolases-like superfamily and are widely distributed across the three domains (Figure 4). In contrast, the production of L-aspartate and L-asparagine via 3-cyano-L-alanine, which is mediated by β-cyano-L-alanine-synthase (EC:220.127.116.11) and two paralog nitrilases (EC:18.104.22.168), appears to be restricted to plants, cyanobacteria and α-proteobacteria (Figure 4). This distribution could be the product of horizontal gene transfer among these clades, probably by symbiosis - as some α-proteobacteria are symbionts and parasites of plants - or by endosymbiosis - because cyanobacteria are considered descendants of plastid ancestors in plants. We did not detect any other possible horizontal gene transfer events in these routes using a database of putative horizontally transferred genes in prokaryotic complete genomes . Finally, the two analog asparaginases (EC:22.214.171.124), converting L-asparagine to L-aspartate, show anti-correlated TDs. One of them, from the glutaminase/asparaginase superfamily, was found in Archaea, some Bacteria, Fungi and Animals (Figure 4), while the second one, from the superfamily of amino-terminal nucleophile aminohydrolases shows a distribution similar to that of EC:126.96.36.199 and EC:188.8.131.52. In summary, the LCA probably was not able to produce either L-aspartate or L-asparagine via the modern canonical alternologs (nitrilase and asparaginase), but could via the degradation of oxaloacetate using the branches described above.
L-tyrosine and L-phenylalanine
There are at least five branches diverging from prephenate for the biosynthesis of L-tyrosine and L-phenylalanine. Two of them proceed via phenylpyruvate and use one of the two widely distributed analog prephenate dehydratases (EC:184.108.40.206). Another two branches proceed via L-arogenate and use either arogenate dehydrogenase (EC:220.127.116.11) to synthesize L-tyrosine or arogenate dehydratase (EC:18.104.22.168) to synthesize L-phenylalanine. EC:22.214.171.124 occurs in Bacteria and some Archaea, while EC 126.96.36.199 has no assigned enzyme (nor gene) sequences. The fifth branch uses prephenate dehydrogenase (EC:188.8.131.52) followed by an aromatic-amino acid aminotransferase (EC:184.108.40.206). E. coli, B. subtilis and S. cerevisiae have two EC:220.127.116.11 and all of them can be classified in the PLP-dependent transferase superfamily, with the exception of AroJ in B. subtilis, whose sequence is unknown. However, it is difficult to establish orthology relationships between these enzymes because none of them are BRHs and, thus, are putatively paralogs. Apparently, this high diversity is maintained by differential expression and multifunctional properties of these enzymes. For instance, TyrB in E. coli is approximately 1,000-fold more active on aromatic substrates than its paralog AspC, which is more specific for aspartate. Similarly, in B. subtilis, HisC is more active than AroJ on phenylalanine and tyrosine, in spite of its primary activity on histidinol-phosphate, and S. cerevisiae uses Aro8 preferentially in anabolism and Aro9 in catabolism [34, 35]. HisC could represent one of the most ancestral lineages in this family because it is the only member widely distributed across the three domains. This agrees with the fact that biosynthesis of L-histidine, the pathway in which HisC preferentially participates, is proposed to be ancestral (see above). The wide distribution of HisC and two analog EC:18.104.22.168 enzymes suggests that the biosynthesis of phenylalanine and tyrosine is also ancient. Nevertheless, the step preceding these enzymes, from chorismate to prephenate, can be catalyzed by one of the two analog chorismate mutases (EC:22.214.171.124); these show a sparse distribution, with some representatives in firmicutes, proteobacteria and plants. For instance, E. coli possesses two paralog EC:126.96.36.199 enzymes belonging to the chorismate mutase II superfamily and they are fused to domains catalyzing EC:188.8.131.52 and EC:184.108.40.206 activities, while B. subtilis also has two EC:220.127.116.11 enzymes, one from the chorismate mutase II superfamily and the other from the YjgF-like superfamily. Neither of these EC:18.104.22.168 domains is widely distributed; thus, it is difficult to establish whether the LCA was able to synthesize L-phenylalanine and L-tyrosine.
There are four alternolog single steps to synthesize L-alanine (Figure 1). One of them uses cysteine desulfurase (EC:22.214.171.124) to degrade L-cysteine and is widely distributed across the three domains. Given that L-cysteine biosynthesis probably occurred in the LCA (see above), we suggest that biosynthesis of L-alanine could occur in the LCA via this step. In contrast, alanine racemase (EC:126.96.36.199) isomerizes D-alanine to L-alanine and is constrained to Bacteria. Alanine aminotransferase (EC:188.8.131.52) converts L-glutamate and pyruvate to L-alanine and is widely distributed in Eukarya but poorly represented in Bacteria and Archaea, whereas valine-pyruvate aminotransferase (EC:184.108.40.206) degrades L-valine to L-alanine and was detected only in few Bacteria (Figure 4).
Finally, there are four proficient branches to synthesize L-serine. The first, proceeding via 3-phospho-hydroxypyruvate, is catalyzed by three enzymes, two of them, 3-phosphoglycerate dehydrogenase (EC:220.127.116.11) and 3-phosphoserine phosphatase (EC:18.104.22.168) are widely distributed, but the third, phosphoserine aminotransferase (EC:22.214.171.124) is restricted to Eukarya and some Bacteria. The second branch is a single step converting ammonia and pyruvate to L-serine by L-serine ammonia-lyase (EC:126.96.36.199), and is restricted to some Bacteria. The third and fourth branches are closely related to the biosynthesis of L-cysteine and L-methionine, inter-converting cystathionine and homocysteine by either EC:188.8.131.52 or EC:184.108.40.206, which are widely distributed across the three domains (see above). Given that L-cysteine and L-methionine could exist in the LCA, the biosynthesis of L-serine could also exist via these enzymes. In fact, this chain of widely distributed enzymes can be extended to the biosynthesis of L-alanine (Figure 1), and all of them together constitute the larger succession of reactions that probably existed in the LCA.
We have carried out a comprehensive analysis of the origin and evolution of amino acid biosynthesis. Our strategy combines genomic tools with a network perspective to identify a core of widely distributed enzymes that probably occurred in the LCA, synthesizing at least 16 out of the 20 standard amino acids. This proposal does not imply, however, that the full biosynthetic routes for these 16 amino acids appeared early, but only some of their branches that could satisfy the minimal biochemical and structural requirements. It is important to note that some species such as parasites and free living animals, including mammals, can lack significant portions of this 'universal' set because they can import amino acids from their hosts or include it in their diet. In parasites, these absences have been attributed to secondary loss. Our results show that most basal animal lineages and other Eukarya posses these universal branches and, thus, their absence in the animal kingdom apparently is also due to secondary losses. Further studies on the possible occurrence of the remaining four standard amino acid biosynthetic routes - for L-proline, L-lysine, L-phenylalanine and L-tyrosine - are necessary as some portions of these pathways are also widely distributed and some 'lost' reactions could fill the gaps.
One of the major biological roles of amino acids is that they are protein constituents; thus, an emerging question from our results is whether this core of amino acids could be sufficient for the LCA protein repertoire? Recently, Atchley et al. , grouped amino acids according to almost 500 attributes, ranging from structural to biochemical and biophysical properties, producing a multidimensional representation of amino acid variability. Mapping the putative core of 16 ancient amino acid biosynthetic branches onto the Atchley et al. plot (Figure 1 in Ref. ) suggests that the LCA was able to populate all the regions of amino acid variability space, so this core could be sufficient for protein functions in early biosystems. Most of the universal branches found in this work are connected to each other (Figure 1), allowing the possibility that they feedback and complete a minimal set of enzyme-driven reactions for the biosynthesis of amino acids. Interplay between the variability and selective pressures resulting from increasing genome sizes and protein structure complexity could promote the incorporation of novel amino acids to this core.
Additionally, we identified alternative branches and routes (paralogs, analogs and alternologs) reflecting the adoption of specific amino acid biosynthetic strategies by taxa, probably due to differences in their life-styles. Eleven out of the twenty amino acid biosynthetic routes revealed an important contribution of paralogy to the generation of diversity. In particular, we corroborated that the retention of gene duplicates as groups, instead of as single entities, is an important factor in the evolution of metabolism. Furthermore, analog enzymes contribute in eight out of the twenty standard amino acid biosynthetic routes, while alternolog routes participate in nine. This implies that analog enzymes and alternolog branches contribute almost as much as gene duplication to genetic buffering in the biosynthesis of amino acids. Further studies are necessary to determine the generality of these observations and to complement them with observations from alternative reactions modeling fluxes in metabolism . In conclusion, we suggest that despite a core of amino acid biosynthetic branches being inherited from ancient systems, the whole contemporary repertoire has been originated independently by lineages according to their environmental resources as reflected by the high diversity of anabolic branches.
In this sense, we consider that one of the goals of the two step strategy presented here (E. coli and multi-organismal TD seeds) is that it uses not only a traditional model organism for genomic analyses, but also as many species as available in current databases. This is important because 8 out of the 20 amino acid biosynthetic routes (L-cysteine, L-serine, L-alanine, L-isoleucine, L-arginine, L-aspartate, L-proline and L-methionine) were quite sparingly distributed from an E. coli-centric perspective, but widely distributed when adding the orthologs, paralogs, analogs and alternologs from other species, revealing the universal nature of some of these routes. Further studies are necessary to determine the generality of these findings, not only in metabolic networks but also in other biological processes.
Materials and methods
In bipartite metabolic networks, there are two sets of nodes - enzymes and compounds (substrates, products and cofactors) - and edges relating enzymes with compounds occurring in the same reaction. For instance, if reaction R1 consumes compound C1 and produces C2 and C3, and it is catalyzed by enzyme E1, the following edges are established: C1 → E1, E1 → C2, and E1 → C3. In reversible reactions a second group of links from products to enzymes and, in turn, from enzymes to substrates is added. In this work we reconstructed the bipartite networks derived from three metabolic databases: EcoCyc v8.0  for E. coli, MjCyc  for M. jannaschii and MetaCyc v8.0  for multi-organismal assignments. To obtain information concerning the nodes and edges for each reaction, we used the following files from EcoCyc and MetaCyc: reactions.dat (substrate/product), enzrxns.dat (reversibility) and reaction-links.dat (EC numbers). From MjCyc the corresponding information was retrieved manually from the database's web page. Networks derived from these databases were merged and prepared for presentation with Cytoscape v2.5.2 . Amino acids were highlighted (red triangles in Figure 1) to denote terminal points of pathways and branches into this network. For clarity in presentation, the most highly connected compounds (mainly cofactors) and the terminal non-amino acid metabolites were removed from the network. Additional data file 2 lists these compounds and contains the pairs of nodes used to construct Figure 1. Multifunctional enzyme sequences were split manually according to their functional domain assignments from Swiss-Prot . Thus, in Figure 1 each node represents one reaction catalyzed by a functional domain (or domains in multimeric enzymes). Analogue enzymes - those catalyzing the same reaction but possessing different folds - were detected by comparing the structural domain content among proteins according to the Superfamily database v1. 69  using HMMer . Additional data file 2 contains details for the final set of 537 enzyme functional domains analyzed in this work as well as 32 reactions without known gene/enzymes. Alternologs were detected by manual inspection of the network in Figure 1, looking for branches that, proceeding via different metabolites, converge in a given compound, generally in an amino acid.
Amino acid sequences from 537 enzymes (functional domains) were tracked in completely sequenced genomes using BLASTP (cutoff E-value = 10-20, and identity percentage >95), this was carried out to obtain the corresponding genomic sequences used as seeds for ortholog detection. Sixty-nine enzymes were excluded because they do not have assigned enzyme (gene) sequences (Additional data file 2). Each genomic sequence seed was used for ortholog detection across 410 non-obligate parasitic genomes (30 Archaea, 363 Bacteria and 17 Eukarya) representing 297 species from 192 genera (Additional data file 2), following the BRH criterion using BLASTP with a cut E-value of 10-5, and a minimum alignment coverage for query and/or subject sequence ≥ 50%. Genomes with less than 1,500 predicted open reading frames (mainly from obligate parasitic genomes) were eliminated from this analysis because they have experienced extensive secondary loses of anabolic enzymes in their genomes, which could introduce noise in TDs. Sequences in the same genome with >95% identity estimated with CD-HIT  were grouped into clusters. As reported , this procedure reduces the frequency of false-negative results caused by cross-matches between highly similar sequences within a genome. Given the redundancy of sequenced strains for some bacterial species, we systematically depurated the original set of genomes, attempting to obtain a normalized measure of ortholog distribution according to the following steps. In step 1, a TD of enzymes versus species was constructed assigning a value of1 to enzymes with orthologs (BRHs) in ≥ 50% of strains from each species. Otherwise a value of 0 was assigned to the corresponding bit. In step 2, a TD of enzymes versus genera was constructed. In this TD, each vector represents the percentage of species having a value of 1 (assigned in step 1) for each genus. In step 3, a TD of enzymes versus clades was constructed. In this TD each vector represents an average of the genera percentages obtained in step 2, for each clade. Clades correspond to the taxonomic categories from the Kyoto Encyclopedia of Genes and Genomes (KEGG) . This procedure provides a 'normalized average distribution' of enzymes across genomes. The final set of strains and genera for each clade is shown in Additional data file 2.
Additional data files
The following additional data are available with the online version of this paper. Additional data file 1 is a graph showing a detailed view of the bipartite network analyzed in this work. Additional data file 2 provides details of enzymes analyzed in this work. Additional data file 3 is a detailed view of pathways and branches analyzed in this work. Additional data file 4 is a bipartite graph of the final metabolic network analyzed in this work (after hub and end compounds removal).
best reciprocal hits
last common ancestor
We would like to thank Alejandra Covarrubias for helpful discussions and comments during the development of this project. We also want to thank Areli Morán for general help throughout this work. This work was partially supported by grant 43502 from the Mexican Science and Technology Research Council (CONACYT). GHM was supported by doctoral scholarship from CONACYT. JJDM has been partially supported by doctoral and post-doctoral fellowships from CONACYT, DGEP-UNAM and Fulbright-García Robles. EPR was partially supported by a grant (ASTF 224-2005) from EMBO. A substantial part of this work was conducted on the Cluster "Sputnik II" from Instituto de Biotecnología-UNAM sponsored by "Macroproyecto de Tecnologías de la Información y la Computación de la UNAM".
- Horowitz NH: On the evolution of biochemical synthesis. Proc Natl Acad Sci USA. 1945, 31: 153-157. 10.1073/pnas.31.6.153.PubMedPubMed CentralView ArticleGoogle Scholar
- Ohno S: Evolution by Gene Duplication. 1970, New York: SpringerView ArticleGoogle Scholar
- Jensen RA: Enzyme recruitment in the evolution of new function. Annu Rev Microbiol. 1976, 30: 409-425. 10.1146/annurev.mi.30.100176.002205.PubMedView ArticleGoogle Scholar
- Aharoni A, Gaidukov L, Khersonsky O, Mc QGS, Roodveldt C, Tawfik DS: The 'evolvability' of promiscuous protein functions. Nat Genet. 2005, 37: 73-76.PubMedGoogle Scholar
- Jeong H, Tombor B, Albert R, Oltvai ZN, Barabasi AL: The large-scale organization of metabolic networks. Nature. 2000, 407: 651-654. 10.1038/35036627.PubMedView ArticleGoogle Scholar
- Wagner A, Fell DA: The small world inside large metabolic networks. Proc Biol Sci. 2001, 268: 1803-1810. 10.1098/rspb.2001.1711.PubMedPubMed CentralView ArticleGoogle Scholar
- Kanehisa M, Goto S: KEGG: kyoto encyclopedia of genes and genomes. Nucleic Acids Res. 2000, 28: 27-30. 10.1093/nar/28.1.27.PubMedPubMed CentralView ArticleGoogle Scholar
- Karp PD, Riley M, Saier M, Paulsen IT, Collado-Vides J, Paley SM, Pellegrini-Toole A, Bonavides C, Gama-Castro S: The EcoCyc Database. Nucleic Acids Res. 2002, 30: 56-58. 10.1093/nar/30.1.56.PubMedPubMed CentralView ArticleGoogle Scholar
- Krieger CJ, Zhang P, Mueller LA, Wang A, Paley S, Arnaud M, Pick J, Rhee SY, Karp PD: MetaCyc: a multiorganism database of metabolic pathways and enzymes. Nucleic Acids Res. 2004, 32: D438-442. 10.1093/nar/gkh100.PubMedPubMed CentralView ArticleGoogle Scholar
- Tsoka S, Simon D, Ouzounis CA: Automated metabolic reconstruction for Methanococcus jannaschii. Archaea. 2004, 1: 223-229.PubMedPubMed CentralView ArticleGoogle Scholar
- Huynen MA, Dandekar T, Bork P: Variation and evolution of the citric-acid cycle: a genomic perspective. Trends Microbiol. 1999, 7: 281-291. 10.1016/S0966-842X(99)01539-5.PubMedView ArticleGoogle Scholar
- Nishida H: Evolution of amino acid biosynthesis and enzymes with broad substrate specificity. Bioinformatics. 2001, 17: 1224-1225. 10.1093/bioinformatics/17.12.1224.PubMedView ArticleGoogle Scholar
- Teichmann SA, Rison SC, Thornton JM, Riley M, Gough J, Chothia C: The evolution and structural anatomy of the small molecule metabolic pathways in Escherichia coli. J Mol Biol. 2001, 311: 693-708. 10.1006/jmbi.2001.4912.PubMedView ArticleGoogle Scholar
- Alves R, Chaleil RA, Sternberg MJ: Evolution of enzymes in metabolism: a network perspective. J Mol Biol. 2002, 320: 751-770. 10.1016/S0022-2836(02)00546-6.PubMedView ArticleGoogle Scholar
- Light S, Kraulis P: Network analysis of metabolic enzyme evolution in Escherichia coli. BMC Bioinformatics. 2004, 5: 15-10.1186/1471-2105-5-15.PubMedPubMed CentralView ArticleGoogle Scholar
- Diaz-Mejia JJ, Perez-Rueda E, Segovia L: A network perspective on the evolution of metabolism by gene duplication. Genome Biol. 2007, 8: R26-10.1186/gb-2007-8-2-r26.PubMedPubMed CentralView ArticleGoogle Scholar
- Galperin MY, Walker DR, Koonin EV: Analogous enzymes: independent inventions in enzyme evolution. Genome Res. 1998, 8: 779-790.PubMedGoogle Scholar
- Kitami T, Nadeau JH: Biochemical networking contributes more to genetic buffering in human and mouse metabolic pathways than does gene duplication. Nat Genet. 2002, 32: 191-194. 10.1038/ng945. (corrigendum in December 2002).PubMedView ArticleGoogle Scholar
- Papp B, Pal C, Hurst LD: Metabolic network analysis of the causes and evolution of enzyme dispensability in yeast. Nature. 2004, 429: 661-664. 10.1038/nature02636.PubMedView ArticleGoogle Scholar
- Ravasz E, Somera AL, Mongru DA, Oltvai ZN, Barabasi AL: Hierarchical organization of modularity in metabolic networks. Science. 2002, 297: 1551-1555. 10.1126/science.1073374.PubMedView ArticleGoogle Scholar
- Hudson AO, Bless C, Macedo P, Chatterjee SP, Singh BK, Gilvarg C, Leustek T: Biosynthesis of lysine in plants: evidence for a variant of the known bacterial pathways. Biochim Biophys Acta. 2005, 1721: 27-36.PubMedView ArticleGoogle Scholar
- Miyazaki J, Kobashi N, Nishiyama M, Yamane H: Functional and evolutionary relationship between arginine biosynthesis and prokaryotic lysine biosynthesis through alpha-aminoadipate. J Bacteriol. 2001, 183: 5067-5073. 10.1128/JB.183.17.5067-5073.2001.PubMedPubMed CentralView ArticleGoogle Scholar
- Nishida H, Nishiyama M, Kobashi N, Kosuge T, Hoshino T, Yamane H: A prokaryotic gene cluster involved in synthesis of lysine through the amino adipate pathway: a key to the evolution of amino acid biosynthesis. Genome Res. 1999, 9: 1175-1183. 10.1101/gr.9.12.1175.PubMedView ArticleGoogle Scholar
- Irvin SD, Bhattacharjee JK: A unique fungal lysine biosynthesis enzyme shares a common ancestor with tricarboxylic acid cycle and leucine biosynthetic enzymes found in diverse organisms. J Mol Evol. 1998, 46: 401-408. 10.1007/PL00006319.PubMedView ArticleGoogle Scholar
- Benner SA, Ellington AD, Tauer A: Modern metabolism as a palimpsest of the RNA world. Proc Natl Acad Sci USA. 1989, 86: 7054-7058. 10.1073/pnas.86.18.7054.PubMedPubMed CentralView ArticleGoogle Scholar
- Cunchillos C, Lecointre G: Integrating the universal metabolism into a phylogenetic analysis. Mol Biol Evol. 2005, 22: 1-11. 10.1093/molbev/msh253.PubMedView ArticleGoogle Scholar
- Xie G, Keyhani NO, Bonner CA, Jensen RA: Ancient origin of the tryptophan operon and the dynamics of evolutionary change. Microbiol Mol Biol Rev. 2003, 67: 303-342. 10.1128/MMBR.67.3.303-342.2003.PubMedPubMed CentralView ArticleGoogle Scholar
- Putnam NH, Srivastava M, Hellsten U, Dirks B, Chapman J, Salamov A, Terry A, Shapiro H, Lindquist E, Kapitonov VV, Jurka J, Genikhovich G, Grigoriev IV, Lucas SM, Steele RE, Finnerty JR, Technau U, Martindale MQ, Rokhsar DS: Sea anemone genome reveals ancestral eumetazoan gene repertoire and genomic organization. Science. 2007, 317: 86-94. 10.1126/science.1139158.PubMedView ArticleGoogle Scholar
- Velasco AM, Leguina JI, Lazcano A: Molecular evolution of the lysine biosynthetic pathways. J Mol Evol. 2002, 55: 445-459. 10.1007/s00239-002-2340-2.PubMedView ArticleGoogle Scholar
- Fani R, Lio P, Lazcano A: Molecular evolution of the histidine biosynthetic pathway. J Mol Evol. 1995, 41: 760-774. 10.1007/BF00173156.PubMedView ArticleGoogle Scholar
- Rutter MT, Zufall RA: Pathway length and evolutionary constraint in amino acid biosynthesis. J Mol Evol. 2004, 58: 218-224. 10.1007/s00239-003-2546-y.PubMedView ArticleGoogle Scholar
- Porat I, Sieprawska-Lupa M, Teng Q, Bohanon FJ, White RH, Whitman WB: Biochemical and genetic characterization of an early step in a novel pathway for the biosynthesis of aromatic amino acids and p-aminobenzoic acid in the archaeon Methanococcus maripaludis. Mol Microbiol. 2006, 62: 1117-1131. 10.1111/j.1365-2958.2006.05426.x.PubMedView ArticleGoogle Scholar
- Garcia-Vallve S, Guzman E, Montero MA, Romeu A: HGT-DB: a database of putative horizontally transferred genes in prokaryotic complete genomes. Nucleic Acids Res. 2003, 31: 187-189. 10.1093/nar/gkg004.PubMedPubMed CentralView ArticleGoogle Scholar
- Nester EW, Montoya AL: An enzyme common to histidine and aromatic amino acid biosynthesis in Bacillus subtilis. J Bacteriol. 1976, 126: 699-705.PubMedPubMed CentralGoogle Scholar
- Weigent DA, Nester EW: Purification and properties of two aromatic aminotransferases in Bacillus subtilis. J Biol Chem. 1976, 251: 6974-6980.PubMedGoogle Scholar
- Atchley WR, Zhao J, Fernandes AD, Druke T: Solving the protein sequence metric problem. Proc Natl Acad Sci USA. 2005, 102: 6395-6400. 10.1073/pnas.0408677102.PubMedPubMed CentralView ArticleGoogle Scholar
- Segre D, Deluna A, Church GM, Kishony R: Modular epistasis in yeast metabolism. Nat Genet. 2005, 37: 77-83.PubMedGoogle Scholar
- Shannon P, Markiel A, Ozier O, Baliga NS, Wang JT, Ramage D, Amin N, Schwikowski B, Ideker T: Cytoscape: a software environment for integrated models of biomolecular interaction networks. Genome Res. 2003, 13: 2498-2504. 10.1101/gr.1239303.PubMedPubMed CentralView ArticleGoogle Scholar
- Boeckmann B, Bairoch A, Apweiler R, Blatter MC, Estreicher A, Gasteiger E, Martin MJ, Michoud K, O'Donovan C, Phan I, Pilbout S, Schneider M: The SWISS-PROT protein knowledgebase and its supplement TrEMBL in 2003. Nucleic Acids Res. 2003, 31: 365-370. 10.1093/nar/gkg095.PubMedPubMed CentralView ArticleGoogle Scholar
- Gough J, Karplus K, Hughey R, Chothia C: Assignment of homology to genome sequences using a library of hidden Markov models that represent all proteins of known structure. J Mol Biol. 2001, 313: 903-919. 10.1006/jmbi.2001.5080.PubMedView ArticleGoogle Scholar
- Eddy SR: Hidden Markov models. Curr Opin Struct Biol. 1996, 6: 361-365. 10.1016/S0959-440X(96)80056-X.PubMedView ArticleGoogle Scholar
- Li W, Jaroszewski L, Godzik A: Tolerating some redundancy significantly speeds up clustering of large protein databases. Bioinformatics. 2002, 18: 77-82. 10.1093/bioinformatics/18.1.77.PubMedView ArticleGoogle Scholar
- Lozada-Chavez I, Janga SC, Collado-Vides J: Bacterial regulatory networks are extremely flexible in evolution. Nucleic Acids Res. 2006, 34: 3434-3445. 10.1093/nar/gkl423.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.