1. Metabolism Evolution

1.1 The Primordial Metabolism. We still do not know precisely when and how life originated on Earth. Nevertheless, it is commonly assumed that early organisms inhabited an environment rich in organic compounds spontaneously formed in the prebiotic world. This heterotrophic origin of life received partial support by the experiments of Stanley Miller in the 1950's (Miller, 1953), and is frequently referred to as the Oparin-Haldane theory (Oparin, 1924, 1936; Lazcano and Miller, 1996). According to this theory, primordial microorganisms could count on external sources of nutrients and had to perform only a minimum of biosynthesis.

Nevertheless, comparative and functional genomics have revealed that most of all living (micro)organisms possess an intricate network of metabolic routes to drive cellular processes. How these pathways have originated and evolved is still under debate (Copley, 2000). In fact, if we assume that primordial organisms were heterotrophic, at least two questions can be asked: why and how did primordial cells expand their metabolic abilities and genomes? The increasing number of early cells thriving in a primordial soup would have quickly led to the exhaustion of essential nutrients imposing a progressively stronger selective pressure that favoured those (micro)organisms that were able to perform more complex metabolic reactions. These organism flourished, diversified, and evolved. Thus, the origin and the evolution of basic metabolic pathways represented a key step in molecular and cellular evolution, since it rendered the primordial cells less dependent on the external supply of nutrients. Concerning the timing, it is not possible to assign a precise chronology to the development of biochemical pathways; however, of them were likely assembled in a DNA/protein world (Fig. 1).

1.2 From Ancestral to Extant Genomes. The first living systems probably stemmed directly from the primordial soup and evolved up to a common ancestor, usually referred to as LUCA (Last Universal Common Ancestor), an entity representing the divergence starting-point of all the extant life forms on Earth (Fig. 1). It has been suggested that LUCA could have been a community of organism rather than a particular organism or a single organismal lineage (Woese, 1998). This community evolved into a smaller number of more complex cell types, which ultimately developed into the ancestor(s) of all the extant life domains. At the beginning, the major evolutionary driving force was probably horizontal gene transfer (HGT) (Joseph 2009). This, together with the inaccuracy of the first information processing, determined the "high genetic temperature" in which, over time, these primordial (micro)organism evolved into a smaller number of increasingly complex cell types leading to three primary groupings of extant organisms.

According to this view, contemporary genomes are the result of 3.5-4 billions of years of evolution. The increasing number of available genetic sequences from organisms belonging to the three domains of life (Bacteria, Archaea and Eukarya) and the continuous implementation of powerful bioinformatic tools provided considerable insight into both the size and the gene content of the genomes of the first living cells, suggesting that ancestral genomes were probably composed by about 1000-1500 genes (Ouzounis et al., 2006). However, despite this small gene content, ancestral genomes were probably fairly complex, enabling a variety of functional capabilities including metabolic transformation, information processing, membrane/transport proteins and complex regulation (Ouzounis et al., 2006).

Although genome size appears highly variable among organisms with the same level of morphological complexity (Salvetti, 2001), it seems well-established that the vast majority of modern-day organisms possesses much more than the hypothetical gene content of LUCA and displays a great complexity (gene regulatory and protein interaction networks, mobile genetics element, etc.). Hence, starting from a common pool of highly conserved genetic information, still shared by all the extant life forms, genomes have been shaped to a considerable extent during evolution. This raises the question: which are the molecular mechanisms that drove the evolution of the earliest genes and genomes.

As a general framework, the evolution of genes and genomes requires (at least) two main steps, 1) the acquisition of new genetic material and 2) its modification to (eventually) develop a new function. The first step is usually achieved via duplication of genes or stretches of DNA, whereas the second step is generally gained through evolutionary divergence. This may be accomplished through several different molecular mechanisms such as, for examples, mutations in the catalytic or regulatory domains or fusions involving two (or more) cistrons (Fondi et al., 2007a; Fondi et al., 2009b).

1.3 Molecular Mechanisms Involved in the Expansion of Metabolic Capabilities. Different molecular mechanisms may have been responsible for the expansion and the shaping of early genomes and the evolution of metabolic pathways. These include gene elongation, duplication and/or fusion, the modular assembly of new proteins, cell fusion (synology) and HGT (xenology) (Fig. 2). The next sections will deal with three of them, i.e. gene duplication, gene fusion and HGT.

1.4 Gene and Operon Duplication. The importance of gene duplication for the development of metabolic innovations was firstly discussed by Lewis (Lewis, 1951) and later by Ohno (Ohno, 1972). The coupling of duplication and divergence of DNA sequences of different size probably represents one of the most important forces driving the evolution of genes and genomes, since this process allows the formation of new genes from pre-existing ones. Genes descending from a common ancestor via a duplication event are called paralogs and they may undergo successive duplications leading to a paralogous gene family (Fig. 3). The comparative analysis of completely sequenced genomes has revealed that a very large proportion of the gene set of different organisms is the outcome of duplication events (Labedan and Riley, 1995; de Rosa and Labedan, 1998) predating or following the appearance of the LUCA (Pushker et al., 2004). These duplicative events may involve individual genes, networks of genes, or the entire genome (Joseph, 2009).

Portions of extant genomes might also be the outcome of DNA rearrangements involving a limited number (20-100) of starter types. Not all ancestral genes have arisen from duplication (Lazcano and Miller, 1996). These ancestral mini-genes might have undergone multiple rounds of gene elongation (i.e. an in tandem duplication followed by the fusion of the two copies) and/or duplication events leading to more complex and/or to paralogous genes. DNA duplications may also involve entire operons or parts thereof, and several examples of single or multiple operon duplications have been reported (Fani and Fondi, 2009) (Fig. 3b). All these events resulted in the (rapid) expansion of the metabolic abilities of primordial cells and the increase of their genome size.

1.5 Gene Fusion. In addition to gene duplication and rearrangements, another mechanism of molecular evolution is the fusion of independent cistrons leading to bi- or multifunctional proteins (Xie et al., 2003). Gene fusions provide a mechanism for the physical association of different catalytic domains or of regulatory structures (Jensen, 1976). Fusions frequently involve genes coding for proteins that function in a concerted manner, such as enzymes catalysing sequential steps within a metabolic pathway (Yanai et al., 2002). Fusion of such catalytic centers likely promotes the channelling of intermediates that may be unstable and/or in low concentration. This, in turn, requires that enzymes catalysing sequential reactions are co-localized within one cell (Mathews, 1993) and (transiently) interact to form complexes that are termed metabolons (Srere, 1987). Fusions have been disclosed in genes of many metabolic pathways, such as tryptophan (Xie et al., 2003), histidine (Brilli and Fani, 2004b) and others (Fondi et al., 2007a).

1.6 Horizontal Gene Transfer (HGT). The Darwinian view of evolution predicts that such process can be interpreted and represented by a "tree of life" metaphor. Any selectable evolutionary "invention", arising from gene or genome level molecular processes is vertically transmitted, and dies out only if lethal. Nevertheless, there are exceptions to the tree of life paradigm (Joseph, 2009). These include evolutionary landmark events of cellular evolution mediated by symbiosis (i.e., chloroplasts and mitochondria). Symbiotic-induced evolutionary events are examples of non-linear evolution. Such processes define the reticulate model of evolution (Gogarten and Townsend, 2005) that eventually took place along with the "classical" vertical transmission (Gribaldo and Brochier, 2009). From a molecular perspective HGT is carried out by different mechanisms and is mediated by a mobile gene pool comprising plasmids, transposons and bacteriophages (Frost et al., 2005; Joseph, 2009). HGT can involve single genes or longer DNA fragment containing entire operons and thus the genetic determinants for entire metabolic pathways conferring to the recipient cell new capabilities. It has been hypothesized that some genes are more likely than others to be transferred via HGT, such as those belonging to different functional categories, e.g. operational vs informational genes. Genes responsible for informational processes (transcription, translation, etc) are likely less prone to HGT than operational genes (Shi and Falkowski, 2008), even though the HGT of ribosomal operon has been described (Gogarten et al., 2002). This latter findings raise the question of the stability of bacterial genomes (Mushegian and Koonin, 1996; Itoh et al., 1999). It is therefore important for phylogenetic and evolutionary analysis to individuate the "stable core" and the "variable shell" in prokaryotic genomes (Shi and Falkowski, 2008).

It is also quite possible that, in addition to HGT (xenology), that early cells might have exchanged (or shared) their genetic information also through cell fusion (sinology). The latter mechanism might have been facilitated by the absence of a cell wall in the Archaean cells and might have been responsible for large genetic rearrangements and rapid expansion of genomes and metabolic activities. In this context, it has been shown how, at least in some cases, the occurrence of HGT might have introduced in the recipient cell important metabolic innovations (Emiliani et al., 2009), leading to crucial consequences in cellular evolution.

2. Hypotheses on the Origin and Evolution of Metabolic Pathways

In the previous sections, we have illustrated how genomes might have assembled and how novel metabolic functions might have appeared during life evolution. Nevertheless, it remains to be shown how entire metabolic pathways and/or networks might have assembled and integrated in the course of time. Several theories have been proposed during the last decades (see Fani and Fondi, 2009). Two of them will be discussed in the following sections: the Retrograde Hypothesis proposed by Horowitz (1945) and the so-called Patchwork Hypothesis, independently developed by Ycas and Jensen in 1974 and 1976, respectively (Ycas, 1974; Jensen, 1976).

2.1 The Retrograde Hypothesis. In 1945 Horowitz (Horowitz, 1945) firstly attempted to provide an explanation of how metabolic pathways might have assembled and evolved, suggesting that biosynthetic enzymes emerged via gene duplication that took place in the reverse order found in current pathways (Fig. 4). This model, known as the Retrograde Hypothesis, states that if the contemporary biosynthesis of compound "A" requires the sequential transformations of precursors "D", "C" and "B" via the corresponding enzymes, the final product "A" was the first compound used by the primordial heterotrophs (Fig. 4). In other words, if a compound A was essential for the survival of primordial cells, the depletion of A from the primitive soup should have imposed a selective pressure allowing the survival and the reproduction of those cells that were become able to transform a chemically related compound "B" into "A" through an enzyme "a" that would have lead to a simple, one-step pathway. The selection of variants having a mutant "b" enzyme related to "a" via a duplication event and capable of mediating the transformation of molecule "C" chemically related into "B", would lead into an increasingly complex route, a process that would continue until the entire pathway was assembled (Fig. 4).

2.2 The Patchwork Hypothesis. The "patchwork" hypothesis (Ycas, 1974; Jensen, 1976) postulates that metabolic routes may have been assembled through the recruitment of primitive slow, non-specific enzymes that could react with a wide range of chemically related substrates enabling primitive cells containing small genomes to overcome their limited coding capabilities. A schematic three-step model of the patchwork hypothesis is reported in Fig. 5: a) an ancestral enzyme E0 endowed with low substrate specificity is able to bind to three substrates (S1, S2 and S3) and catalyze three different, but similar reactions; b) a duplication of the gene encoding E0 and the subsequent divergence of one of the two copies leads to the appearance of enzyme E2 with an increased and narrowed specificity; c) a further gene duplication event, followed by evolutionary divergence, leads to E3. In this way the ancestral enzyme E0, belonging to a given metabolic route is "recruited" to serve other novel pathways.

The evolutionary model proposed with the patchwork hypothesis might have permitted the expansion of the metabolic capabilities of primordial cells and also the evolution of regulatory mechanisms coincident with the development of new pathways (Brilli and Fani, 2004b). This idea is supported by both the analysis of completely sequenced genomes and by directed evolution experiments (Fani and Fondi, 2009) and it has been invoked to explain the evolution of several processes such as the urea and the Krebs cycle, nitrogen fixation (Copley, 2000; Fani et al., 2000) and some amino acid biosynthetic pathways, such as those of histidine (Fani et al., 1995), tryptophan (Xie et al., 2003) , lysine, arginine and leucine (Fondi et al., 2007b) , as well as older or newer catabolic pathways (Fani and Fondi, 2009 and references therein).

3. Histidine Biosynthesis: A Paradigm For The Study of the Evolution Of Metabolic Pathways

Histidine biosynthesis is an important metabolic cross-road and that plays a key role in cellular metabolism being interconnected to both the de novo synthesis of purines and to nitrogen metabolism. How the his pathway originated remains an open question, but the analysis of the structure and organization, as well as the phylogenetic analyses of the his genes in (micro)organisms belonging to different phylogenetic archaeal, bacterial and eukaryal lineages revealed that different molecular mechanisms played an important role in shaping this pathway. Therefore, the histidine biosynthetic pathway represents an excellent model for understanding the molecular mechanisms driving the evolution of metabolic routes.

3.1 Gene Elongation, Duplication: hisA and hisF. Two of the histidine biosynthetic genes, hisA and hisF, are particularly interesting from an evolutionary viewpoint. The two genes code for a [N-(5'-phosphoribosyl) formimino]-5- aminoimidazole-4-carboxamide ribonucleotide (ProFAR) isomerase and a cyclase, respectively, which catalyze two central and sequential reactions of the pathway. The comparative analysis of the HisA and HisF proteins revealed that they are paralogous and share a similar internal organization into two paralogous modules half the size of the entire sequence (Fani et al., 1994). According to the model proposed, an ancestral module half the size of the present-day hisA gene underwent a gene elongation event leading to the ancestral hisA gene, which in turn duplicated, giving rise to hisF (Fig. 6).

Since the overall structure of the hisA and hisF genes are the same in all known (micro) organisms, it is likely that they were part of the genome of the LUCA. The biological significance of the hisA-hisF structure relies on the structure of the encoded enzymes; indeed, they contain a (TIM) (βα)8 barrel-like fold (Copley and Bork, 2000). The barrel structure is composed by eight concatenated (β-strand)-loop-(α-helix) units. Thus, the model proposed predicts (Fani et al., 1994) that the ancestral gene coded a half-barrel, which might assemble to give a functional enzyme by homo-dimerization. The elongation event leading to the ancestor of hisA/hisF genes resulted in the covalent fusion of two half-barrels producing a protein whose function was refined and optimized by mutational changes; once assembled, the "whole-barrel gene" underwent gene duplication, leading to the ancestor of hisA and hisF. The two genes represent a paradigmatic example of how evolution works at both molecular and functional level.

3.2 Gene Fusion and HGT: hisNB. The eighth and sixth steps of histidine biosynthesis are catalyzed by histidinol-phosphate phosphatase (EC 3.1.3.15) (HOL-Pase) and IGP dehydratase (EC 4.2.1.19), respectively (Winkler, 1987). Distinct HOL-Pases have been characterized in different organisms, whereas IGP dehydratase is the same in all known histidine synthesizing organisms. In E. coli the two activities are coded for by a single gene, referred to as hisNB (Brilli and Fani, 2004b): the N-terminal domain (HisN) is a phosphatase belonging to the DDDD family (Brilli and Fani, 2004b) and the C-terminal domain is responsible for IGP dehydratase activity. The evolutionary history of the hisNB gene has been recently reported by (Brilli and Fani, 2004b) who showed that hisNB gene fusions are present in most γ-proteobacteria, in some α-, δ- and ε-proteobacteria and in some representatives of CFB group (Fani et al., 2007). A deep phylogenetic analysis allowed to trace the fusion event in an ancestor of the γ-subdivision and was then horizontally transferred from some γ-proteobacteria to the other microorganisms. As shown by phylogenetic trees obtained with other His proteins, it is also quite possible that the transfer event might have involved the entire his operon or part thereof. This statement is also supported by the analysis of the organization of his genes in a representative sets of genomes harbouring the hisNB fusion, which revealed that all of them are localized within (more or less) compact operons (Fani et al., 2007).

3.3 A Model for the Assembly of Histidine Biosynthetic Pathway. Actually, an impressive series of well characterized duplication (Fani et al., 1995; Fani et al., 1998), elongation (Fani et al., 1994), and fusion (Fani et al., 1995; Brilli and Fani, 2004a; Fani et al., 2007) events has shaped this pathway. The integration of the whole body of molecular and evolutionary data, allows depicting possible scenarios accounting for the origin and evolution of histidine biosynthetic route in microbes. In fact, the phylogenetic distribution of the genes in Bacteria (Fani et al., 2007) and in Archaea (Fondi et al., 2009c) allows to infer that eight genes specifically involved in histidine biosynthesis (i.e. hisGDBHAFIE) were present in the genome of LUCA, coded for monofunctional enzymes (except HisD). Concerning the two remaining genes, hisC and hisN, it is quite possible that the two reactions carried out by the extant proteins they code for, might have been performed by enzymes with a broad substrate specificity. Concerning the organisation of histidine biosynthetic genes, it has been proposed (Fani et al., 2005) that, at least in Proteobacteria, the construction of a compact his operon would be a recent event in evolution and that it was piece-wisely assembled by the joining of short sub-operons comprising two-three genes.

On the other hand, Price et al. (2006) showed that a unified hisGDC(NB)BHAF(IE) operon is present in some lineages. This result led the authors to hypothesize that this operon is ancient. The available data and the comparative analysis of his genes organization in Archaea allows re-addressing this point (Fondi et al, 2009b). In principle, if a given phylogenetic lineage includes microorganisms showing a different organization of genes belonging to the same metabolic pathway, i.e. complete scattering, compact operons or partial scattering/partial clustering, at least two hypothetical scenarios can be invoked to explain such a picture (Fig. 7):

1) LUCA harboured genes (partially) scattered throughout the genome, so that in some of the descendants the construction of clusters and/or operons occurred. According to this idea, histidine biosynthetic genes would have then been clustered independently in different archaeal lineages, that is Thaumarchaeota, Sulfolobales and (partially) in Thermoproteales. Conversely, all the other archaeal lineages would have maintained histidine biosynthetic genes (mainly) scattered throughout the chromosomes. The fact that in Thaumarchaeota the gene order resembles the γ-proteobacterial one may suggest that an event of HGT might have been responsible for the appearance of the his genes array in Thaumarchaeota. However, the representatives of this archaeal phylum are embedded in a well-supported monophyletic cluster in the archaeal part of the tree, ruling out the hypothesis of an HGT event. The same can be said for the other archaeal lineages displaying a (more or less) compact histidine operon (Sulfolobales and Thermoproteales). Noteworthy, according to this hypothesis, a strong selective pressure in recreating the same relative gene order (starting from scattered genes) must be invoked to account for the emergence of two nearly identical operons in two distantly related prokaryotic clades (Thaumarchaeota and γ-proteobacteria).

2) The genome of LUCA contained genes organized in operons and this organization was completely or partially destroyed during evolution in some of the descendants’ lineages (Fig. 5b). Although it is not possible to infer the gene order of this ancient gene cluster (it might be the hisGDC(NB)HAF(IE)) proposed by Price et al. (2006), this latter scenario would predict the disruption of the operon in different phylogenetic lineages (all the Euryarchaeal lineages and the Crenarchaeal groups of Thermoproteales and Desulfurococcales) and then the re-assembling of the his genes in a different way in respect to the ancestral state (in Sulfolobales). Since Thaumarchaeota likely represent the first emerging lineage in the Archaeal domain (Brochier-Armanet et al. 2008) their histidine genes organization may resemble the ancestral one. However, the fact that, within bacteria, only γ-proteobacteria display a similar organization (hisGDCNBHAFIE), whereas in others phyla the histidine genes are differently organized (e.g. hisDCBHA-impA-FI in Actinobacteria and hisZGDBHAFI in (most of) Firmicutes) would imply that histidine biosynthetic genes would have re-assembled differently in a number of bacterial phyla, while they would have been kept with the ancestral arrangement in γ-proteobacteria). Moreover, if this scenario is correct, this would mean that the destruction of operon organization should have given rise to scattered genes and/or mini-operons with the creation of new promoters upstream of each of them (Itoh et al. 1999) and the eventual reconstruction of a compact his operon with a different gene order in different archaeal lineages.

4. Conclusions

Metabolic pathways of the earliest heterotrophic organisms arose in conjunction with the exhaustion of the prebiotic compounds present in the primordial soup. In the course of molecular and cellular evolution different mechanisms and different forces might have concurred in the arisal of novel metabolic abilities and in the shaping of metabolic routes. It is possible that the ancestral forms of life might have expanded their coding abilities and their genomes by duplicating a small number of starter types mini-genes via a cascade of elongation, duplication, or fusion events, as well as duplication of catabolic or biosynthetic operons. The dissemination of metabolic routes between microorganisms might have been facilitated by horizontal transfer or cell fusions events, which could have been facilitated by the absence of a cell wall. The horizontal transfer of entire metabolic pathways or part thereof might have had a special role during the early stages of cellular evolution when, according to Woese (1998), the "genetic temperature" was high and might have been facilitated by the operon organization of early genes that would have permitted the transfer of entire metabolic routes.

There are many different schemes that can be proposed for the emergence and evolution of metabolic pathways depending on the available prebiotic compounds and the available enzymes previously evolved. Even though most of data coming from the analysis of completely sequenced genomes and directed-evolution experiments strongly support the patchwork hypothesis, we do not think that all the metabolic pathways arose in the same manner. In our opinion the different schemes might not be mutually exclusive. Thus, some of the earliest pathways may have arisen from the Horowitz scheme, some from the semi-enzymatic proposal, and later ones from Jensen’s enzyme recruitment.

The heterogeneous distribution organisation and structure of his genes within the microbial world revealed that they underwent several recombination events during evolution (including gene duplication/fusion and horizontal gene transfer) and this led to the different schemes of his genes organization that we observe in modern microorganisms. Finally, the organization of his genes in some extant archaeal and bacterial lineages speaks toward a piece-wise construction of his suboperons along with gene fusion and HTG events.