Chapter 144

Evolution of Metabolic Pathways in Enteric Bacteria

ROY A. JENSEN

INTRODUCTION

Enteric bacteria form a cohesive phylogenetic cluster whose common ancestor diverged from other γ-subdivision members of the gram-negative purple bacteria (77). This divergence occurred quite recently on the grand scale of evolutionary time: it was an event that correlated with the emergence of mammalian organisms. The repertoire of complex metabolic machinery now utilized by enteric bacteria and other contemporary groups must have been preceded by considerably less sophisticated systems encoded by fewer genes.

Ancient microbes very likely possessed unregulated genes encoding broad-specificity enzymes, each of which utilized a family of related substrates and produced a family of related products (41, 84). This assumption is the basis of the recruitment hypothesis (41), which states that genes for most newly evolved enzymes have been recruited from preexisting genes encoding enzymes having similar mechanisms of action. The expansion and refinement of metabolic pathway capability in contemporary organisms is postulated to have occurred through gene duplication and a subsequent, differential narrowing of substrate specificities. Thus, minor products formed fortuitously by ancestral proteins could have been amplified and preserved by selection whenever the new products proved useful. Mutations in individual gene copies leading to specialization for a given substrate would enlarge the gene family. Repetitions of such a process would generate families of homologs catalyzing parallel reactions, e.g., anthranilate synthase and 4-aminobenzoate synthase. In the modern era of molecular genetics, the outlines of many new and unexpected gene families are emerging. It is intriguing to realize that when these gene families are all recognized in the near future, each family may represent an ancient gene from which many specialist members radiated.

It is not unusual for modern enzymes to have retained low affinities for "nonphysiological" substrates, but this broad specificity can be effectively masked by the regulatory mechanisms that have evolved. As a striking example, the overall regulatory mechanisms in force for control of isoleucine-valine-leucine biosynthesis mask the considerable potential of the first leucine pathway enzyme (α-isopropylmalate synthase) to utilize inappropriate substrates. The masking effect is illustrated by the dramatic and far-reaching consequences of a single deregulating mutation releasing threonine deaminase, the initial enzyme of isoleucine biosynthesis, from feedback inhibition by isoleucine (50). Allosteric insensitivity results in an abnormal diversion of pyruvate to isoleucine and therefore a limited production of valine and leucine. An imbalanced ratio of pathway keto acids occurs, with elevation of α-ketobutyrate and ketoisoleucine levels but diminution of pyruvate, ketovaline, and ketoleucine levels. The entire leucine biosynthetic pathway derepresses in response to leucine starvation, and α-isopropylmalate synthase encounters low levels of its normal substrate (ketovaline) and high levels of α-ketobutyrate and ketoisoleucine (2-keto-3-methylvalerate). The last two false substrates are converted via the four leucine biosynthetic pathway enzymes to the antimicrobial agents norvaline and homoisoleucine. That all of this is avoided in wild-type cells when both the level and the activity of α-isopropylmalate synthase are regulated normally exemplifies how overall regulatory mechanisms and different relative affinities for alternative substrates can constrain broad-specificity enzymes to operate as specific catalysts under most conditions while still retaining high capacity for evolutionary (and physiological) adaptations.

The variety of sophisticated metabolic systems now in existence could have emerged in various ways. Functionally similar enzymes (e.g., catalyzing identical, parallel, or reverse reactions) could have arisen independently. Such proteins are termed analogous if their similar functions originated by convergence. When gene duplication has produced homologous proteins, they are termed paralogous. Where speciation is responsible for homologous proteins, the relationship is termed orthologous. If a homologous relationship is due to horizontal gene transfer, a xenologous relationship exists.

In this chapter, an initial effort is made to organize the emerging information that bears upon the evolutionary relationships of biochemical pathways in the enteric bacteria. These evolutionary considerations require knowledge of the biochemical features of nonenteric organisms over a phylogenetic span appropriate to the time of divergence of a given character state. A sufficient amount of comparative data is thus far approached only for the pathway of aromatic amino acid biosynthesis, and this system is highlighted to exemplify prospects for deducing the evolutionary histories of other pathways. In addition, a rapidly expanding database of gene sequences is leading to the recognition of new families of paralogous genes that identify evolutionary relationships between different biochemical pathways. Thus, it is exciting to realize that we are at the threshold not only of following the set of genes that specify a given biochemical-pathway unit backward through evolutionary time but also of being able to identify whole gene families that span a larger metabolic gestalt and that contain individual members possessing highly specialized functions.

IDENTICAL REACTIONS PARALLEL REACTIONS, AND HOMOLOGY RELATIONSHIPS

Proteins that catalyze the same reaction in one organism, e.g., sets of regulatory isoenzymes such as the Escherichia coli 3-deoxy-d-arabinoheptulosonic acid 7-phosphate (DAHP) synthases or aspartokinases, can be expected to belong to a common gene family. Parallel reactions (utilizing related substrates and producing related products) can reasonably be expected to be catalyzed by homologous proteins. The power of molecular genetic analysis, however, has revealed surprises at both extremes of expectation. Some same-reaction proteins have failed to show evidence of homology, while some different-reaction proteins have proven to be unexpected homologs. Inevitably, functional similarities of unexpected different-reaction homologs that were not obvious until the fact of homology became known have become appreciated.

The individuality of various reaction mechanisms, ranging from facile catalysis to highly restrictive catalytic requirements, means that the selective pressure exerted to conserve protein sequence will vary in different gene families. At one extreme, chorismate mutase catalyzes a reaction that is sufficiently undemanding that divergence may mask homology. Thus, the monofunctional periplasmic chorismate mutase (CM-F) of Erwinia herbicola shows neither clear homology with the chorismate mutase domains of PheA (CM-P) or of TyrA (CM-T) from E. coli or Erwinia herbicola nor with the monofunctional chorismate mutase of Bacillus subtilis (81). In such cases, it should be possible to decide between the alternatives of independent origin and rapid divergence by analyzing sequences from a progression of intermediate, closely spaced organisms.

The proteins of purine biosynthesis provide examples of homologous and nonhomologous relationships that at first sight are unexpected. E. coli possesses two species of glycinamide ribonucleotide transformylase, PurN and PurT. Both produce β-formyl glycinamide ribonucleotide from glycinamide and formate. PurN and PurT do not exhibit homology, but PurT is a homolog of an enzyme in a later step of purine biosynthesis, PurK (57). PurK is a subunit of aminoimidazole ribonucleotide carboxylase and uses aminoimidazole ribonucleotide, bicarbonate, and ATP to produce carboxy aminoimidazole ribonucleotide. Thus, PurT is not a homolog of PurN even though it catalyzes the same overall reaction, but it is a homolog of PurK, which catalyzes a different, albeit similar reaction. The reactions of PurT and PurK are related in that both transfer a single carbon unit to a ribonucleotide derivative, the donor being formate in one case and bicarbonate in the other. The basic difference between PurN and PurT can be appreciated at the level of mechanism in that PurN is dependent upon N ¹⁰-formyl-tetrahydrofolate for transformylation, whereas the PurT reaction (like PurK) is not folate dependent.

Two similar energy-coupled transport systems coexist in E. coli for uptake of molecules too large for porin channels. Ferric siderophores, vitamin B₁₂, and group B colicins are taken up by the TonB-ExbB-ExbD system, whereas group A colicins are taken up by the TolA-TolQ-TolR system (17). These transport systems exemplify parallel, multicomponent gene product assemblies that exhibit clear homology.

Two hypothetical scenarios accounting for the evolutionary origin of prephenate dehydrogenase (PDH) in Tyr biosynthesis (Fig. 1) illustrate a basis for expectation that PDH would or would not be homologous with cyclohexadienyl dehydrogenase (CDH), an enzyme capable of catalyzing the same reaction. (See Table 1 for a comprehensive list of abbreviations and definitions used in consideration of the evolution of aromatic amino acid biosynthesis.) In the most straightforward scenario, PDH could have originated from an ancestral CDH via gene duplication and narrowed substrate specificity, thus generating PDH and CDH homologs. Alternatively, a fusion of two hypothetical primitive genes could have yielded a PDH. Prephenate (PPA) is known to form 4-hydroxyphenyllactate under nonenzymatic conditions at alkaline pH (27). Thus, existence of an enzyme catalyzing such a "hydroxyphenyllactate synthase" reaction is feasible. Fusion of the gene encoding such a hydroxyphenyllactate synthase with a gene encoding a broad-specificity type of lactate dehydrogenase would result in the overall conversion of PPA to 4-hydroxyphenylpyruvate. Broad-specificity lactate dehydrogenase enzymes that utilize 4-hydroxyphenyllactate are known (29). This scenario predicts that a PDH derived in this manner would have a modular organization, with one domain corresponding to the sequence of broad-specificity lactate dehydrogenase. The mechanism would demand absolute specificity for PPA, precluding l-arogenate (AGN) as an alternative substrate. (Since TyrA of enteric bacteria exhibits a minor capability for utilizing AGN as substrate [2], the gene fusion scenario is unlikely for this group.)

Fig. 1Composite of the biochemical flow routes to tyrosine and phenylalanine that exist in nature. See Table 1 for abbreviations.

Source:

Table 1Aromatic amino acid biosynthesis: abbreviations and definitions

To summarize, homologous proteins often catalyze mechanistically identical or similar reactions, but proteins that catalyze similar reactions are not necessarily homologous and may have arisen by separate evolutionary paths. In either case, sequence homologs can be expected to provide evidence for the events of evolutionary recruitment.

EVOLUTION OF AROMATIC AMINO ACID BIOSYNTHESIS

In enteric bacteria, the evolution of aromatic amino acid biosynthesis has been dynamic. It is striking that even though E. coli is widely used as a reference organism for prokaryotes, neither its post-PPA pathways (47) nor its tryptophan pathway (see Gene Fusion section) are the most typical overall arrangement to be found in enteric bacteria. Figure 1 highlights the post-PPA pathways of Phe and Tyr biosynthesis. Appreciation of the evolutionary dynamism and complexity is critically dependent on the availability of comparative data from the phylogenetic neighbors of enteric bacteria. The systematic choice of appropriately spaced organisms on the phylogenetic tree within the γ subdivision provides a comparative context that allows us to draw conclusions about which character states evolved recently and which ones evolved long ago. For example, analysis allows us to appreciate that the Phe-inhibited isoenzyme of DAHP synthase (DS-Phe) (6, 44) and the bifunctional T protein (5) emerged recently in the enteric lineage shortly after divergence of their common ancestor from the rest of the γ subdivision. On the other hand, the bifunctional P protein has an ancient origin and arose in a common ancestor of subdivisions γ and α shortly after its divergence from the β subdivision (1).

Aspects of aromatic amino acid pathway evolution will be considered here as they bear on the common pathway and the Tyr and Phe branches. The intriguing paralogous relationships of both the aminase and the glutamine-binding subunits of anthranilate synthase and p-aminobenzoate synthase of the tryptophan and p-aminobenzoate branches are considered by Nichols in chapter 143 of this volume.

The Initial Pathway Step

All enteric bacteria elaborate three differentially regulated isoenzymes of DAHP synthase. DS-Phe, DS-Trp, and DS-Tyr are feedback inhibited by Phe, Trp, and Tyr, respectively. A systematic and detailed analysis (8) of DAHP synthase isoenzymes has led to the conclusion that the common ancestor of the γ subdivision possessed two isoenzymes, DS-Tyr and DS-O, an isoenzyme of DAHP synthase that is insensitive to allosteric control. DS-O, present in contemporary Acinetobacter calcoaceticus, is insensitive to allosteric control. DS-O is proposed to be the ancestor of the DS-Trp isoenzyme of enteric bacteria. Most γ-subdivision bacteria (but not enteric bacteria) possess a tryptophan-sensitive isoenzyme that exhibits a lesser sensitivity to inhibition by chorismate [DS-Trp(CHA)]. Therefore, it was suggested that fusion of the gene encoding DS-O with a gene duplicate of that encoding anthranilate synthase (large subunit) might have produced DS-Trp(CHA). Recruitment of an anthranilate synthase domain would explain the presence of binding sites for both tryptophan (the best inhibitor) and chorismate. Perhaps only the inhibition by tryptophan is relevant physiologically in γ-subdivision bacteria that possess DS-Trp(CHA), and the sensitivity to chorismate may be an evolutionary remnant.

However, the Xanthomonas lineage exemplifies a most interesting case in which the sensitivity to feedback inhibition by chorismate has been maximized to give an isoenzyme species denoted DS-CHA(Trp). In this case, DS-Tyr has been discarded, and DS-CHA(Trp) exists as the sole DAHP synthase protein that operates as a key component of a novel pattern of sequential feedback inhibition (76). A comparison of the sequences of trpE (anthranilate synthase) and aroH (DS-Trp) from E. coli did not reveal any homology (66). However, since sensitivity of E. coli DS-Trp to chorismate has been eliminated altogether, the evidence for homology may now be minimal. A better test of the possible recruitment of a regulatory domain in DS-Trp isoenzymes from a trpE duplicate would come from sequencing genes encoding the DS-Trp(CHA) type of isoenzyme, since both the tryptophan- and chorismate-binding sites have been retained.

DS-Phe evolved quite recently, emerging within the enteric lineage and being absent throughout the remainder of the γ subdivision (8). Since Phe and Tyr are close analogs, it seems likely that the new Phe-sensitive isoenzyme might have arisen via modifications of a gene duplicate encoding the Tyr-sensitive isoenzyme. This scenario predicts that DS-Phe in the γ subdivision originated independently of DS-Phe in the other proteobacterial subdivisions. This analysis exemplifies the point that without a systematic study of organisms appropriately spaced on the phylogenetic tree, it might not be appreciated that DS-Phe within the enteric lineage evolved recently and is distinct from DS-Phe isoenzymes in other gram-negative bacteria. [Likewise, it would not be appreciated that DS-Trp from E. coli is probably a closer homolog of DS-CHA(Trp) from Xanthomonas campestris than of DS-Trp from Zymomonas mobilis (3), even though the allosteric specificity of Z. mobilis is identical to the E. coli pattern and that from X. campestris is not.]

The Branch to Tyrosine

tyrA specifies a two-domain protein that catalyzes the initial two steps of tyrosine biosynthesis. The bifunctional T protein possesses an N-terminal CM-T flanked by a dehydrogenase domain. Although the latter has been named PDH, it is actually able to use either PPA or AGN as an alternative substrate (2) and is therefore a CDH. Since CM-T activity would generate PPA in the microenvironment of the dehydrogenase catalytic center, the ability to utilize AGN in vitro appears not to be realized appreciably in vivo. If it is not, then the gene fusion has effectively narrowed the substrate specificity. At the same time, the existing substrate ambiguity represents a potential for suppressor mutations. Moreover, the broad substrate specificity probably reflects the origin of the T-protein dehydrogenase domain from an ancestral CDH via gene fusion. This conclusion is consistent with observations that no organism possesses the T protein and CDH simultaneously and that the nearest relatives of enteric bacteria that lack a T protein possess a tyrosine-inhibited CDH.

It was originally supposed (42) that the bifunctional T protein evolved by fusion of genes encoding CDH and CM-F, in which case T-protein-containing enteric bacteria should lack CM-F as well as CDH. However, many enteric organisms, including Salmonella typhimurium (official designation, Salmonella enterica serovar Typhimurium) possess CM-F in addition to CM-T and CM-P (4). CM-F has proven to be a periplasmic protein whose deduced amino acid sequence does not exhibit obvious homology with CM-T (81). Therefore, CM-F is unlikely to be the progenitor. Since PheA and TyrA are homologous at their N termini (see chapter 28 in this volume), the CM-T domain of TyrA may have been recruited from the CM-P domain of PheA.

Cofactor specialization of tyrosine pathway dehydrogenases has proven to be a character state of evolutionary interest. Pyridine nucleotide-dependent dehydrogenases usually require NADPH if they catalyze reductive biosynthetic reactions and NAD⁺ if they catalyze oxidative catabolic reactions. However, oxidative biosynthetic reactions such as those catalyzed by the tyrosine pathway dehydrogenases are not subject to these rules and may utilize NAD⁺ or NADP⁺ or may accept either. Narrowed specificity for cofactor in tyrosine pathway dehydrogenases has been a recent evolutionary event in the Proteobacteria, occurring after the separation of the lineages leading to the α, β, and γ subdivisions. In all three subdivisions of the Proteobacteria, phylogenetic clusters that exhibit narrowed specificity for NAD⁺ or NADP⁺ exist (43). These clusters are cohesive, indicating that once this specialization has evolved, it is faithfully retained. A cofactor specialization for NAD⁺ exists for the TyrA dehydrogenase in enteric bacteria. Cofactor specialization is probably tied to the redox biology of the cell, and it would be interesting to see whether the same choice (NAD⁺ or NADP⁺) made for the tyrosine pathway dehydrogenase in a given organism is also made for other dehydrogenases (e.g., homoserine dehydrogenase) that also catalyze oxidative biosynthetic reactions.

The third and final step of Tyr biosynthesis is an aminotransferase reaction catalyzed by TyrB, which is encoded by tyrB, which (like tyrA) is repressed by TyrR. The gene family relationship of tyrB is considered in a later section.

Thus, the aromatic pathway branch to Tyr in enteric bacteria probably entailed a narrowing of substrate preference that converted CDH to a more effective PDH catalyst and then a fusion of the dehydrogenase-encoding gene with the gene fragment encoding CM-P.

The Branch to Phenylalanine

pheA specifies the initial two steps of the phenylalanine branch. PheA, known as the bifunctional P protein, carries domains for CM-P and PDT. The P protein appears to be present throughout the α and γ subdivisions of the Proteobacteria. CM-P and CM-T of the tyrosine pathway T protein are homologs, and the 5' portion of pheA may have given rise to the 5' portion of tyrA following gene duplication (81). In contrast, the dehydratase domain of PheA has no obvious homology to the dehydrogenase domain of TyrA.

The prephenate dehydratase (PDT) of many organisms is typically subject to very strong activation by Tyr (1). Indeed, Tyr is obligatory for activity in various γ-subdivision (10) and β-subdivision (71a) lineages. However, P proteins of enteric bacteria are not affected by Tyr. In enteric bacteria, the evolution of P proteins that are no longer activated by Tyr correlates perfectly with the appearance of the bifunctional T protein. With the evolutionary acquisition of the T protein in enteric bacteria, a balanced system of competing "channels" to Phe and Tyr was established, effectively moving the branchpoint backwards from PPA to chorismate. The tyrosine branch acquired direct end product control of Tyr formation, thus replacing the indirect mechanism that existed before the divergence of the enteric lineage (consult Fig. 2). Nonenteric Proteobacteria frequently possess a CDH that is not inhibited or is weakly inhibited by Tyr. In such systems, Tyr regulates its own biosynthesis indirectly via activation of PDT activity in the Phe pathway. Thus, a balance is maintained whereby PPA molecules generated by CM-P are preferentially utilized by CDH, but excess Tyr promotes activation of the PDT domain, which effectively channels PPA to Phe. In nonenteric organisms that lack CM-T, CM-P thus functions for both Phe and Tyr biosynthesis via what has been termed a channel shuttle mechanism (18). This indirect mechanism of Tyr regulation via activation of the Phe-branch PDT must be quite ancient, since it is widely distributed in nature, occurring in cyanobacteria and some gram-positive bacteria. Thus, in this respect, enteric bacteria are quite distinctive.

Fig. 2Evolution of aromatic amino acid biosynthesis in enteric bacteria. The superfamily B (? subdivision) ancestor possessed a pathway whose allosteric control was dominated by Tyr. Tyr inhibited (-) the DS-Tyr isoenzyme of DAHP synthase, inhibited CDH (whether using alternative substrates PPA or AGN), and was required for activation (+) of the PDT component of the bifunctional P protein. PDT was subject to feedback inhibition by Phe, allowing PPA generated by CM-P to diffuse to CDH and/or cytoplasmic aminotransferase ([AT]c). Periplasmic chorismate could be transformed to Phe via CM-F, CDT, and periplasmic aminotransferase ([AT]p). The altered character states that occurred prior to divergence from the common ancestor of enteric bacteria are shown. The pathway that emerged in enteric bacteria exhibited a new symmetry of allosteric control whereby all three amino acids control cognate DAHP synthase isoenzymes and the PDT requirements for Tyr was lost. A new symmetry of competing bifunctional proteins at the chorismate branch point also evolved. The periplasmic pathway of Phe biosynthesis was lost, retained, or partially retained as discussed in the text. See Table 1 for abbreviations and definitions.

Source:

Many properties of the P protein vary in the cases for which detailed studies have been done. These properties include the formation of dimers and tetramers in the presence of allosteric effectors. pheA in E. coli is subject to transcriptional regulation by an attenuation mechanism, and this mechanism is highly conserved in Erwinia herbicola (82). However, the P protein of Pseudomonas stutzeri exhibits no evidence of attenuation control (30). A systematic study of intervening organisms would indicate whether pheA attenuation arose relatively recently or whether it was simply abandoned in the Pseudomonas lineage.

The Periplasmic Pathway of Phenylalanine Biosynthesis

Table 2 lists enteric organisms and indicates the presence or absence in them of CM-F and/or cyclohexadienyl dehydratase (CDT). Both CM-F and CDT possess cleavable signal peptides that are responsible for their translocation to the periplasm in Erwinia herbicola (81) and Pseudomonas aeruginosa (85). The increased accumulation of PPA in a tyrosine auxotroph of S. typhimurium compared to that in similar auxotrophs of E. coli has been attributed to the presence of CM-F (14). An aromatic aminotransferase has also been located in the periplasmic fractions of Erwinia herbicola and S. typhimurium (unpublished data). These three enzymes potentially make up an intact branch of Phe biosynthesis that is not subject to feedback inhibition. Indeed, a mutant of P. aeruginosa possessing an allosterically insensitive species of DAHP synthase converts the excess chorismate produced into Phe (31).

Table 2Distributiona of periplasmic enzymes of phenylalanine biosynthesis in enteric bacteria and their closest phylogenetic relatives (Aeromonas and Alteromonas spp.)

Wild-type strains do not excrete Phe, presumably because only minimal amounts of chorismate are available in the periplasm. The normal function of these periplasmic proteins, which appear to be localized within the polar caps of the periplasm, is completely unknown. Interestingly, CDT is homologous with a class of periplasmic binding proteins specific for polar amino acids. CDT may serve as a receptor for chemotaxis and transport, and Tam and Saier (72) have cited this function as an example of an evolutionary link between cell surface receptors and enzymes.

Unlike all of the other character states that have been examined, the distribution of CM-F and CDT with respect to the phylogenetic tree (Table 2) is erratic, such that the evolutionary scenario deduced requires multiple events of loss and gain (4). It is therefore possible that the corresponding genes (aroQ and pheC) have been subject to horizontal transfer. Indeed, the codon usage of E. herbicola aroQ corresponds best with the class III coding sequences of E. coli (59a). Class III coding sequences make up about 16% of the total E. coli genes and are postulated to be genes inherited by horizontal transfer. Alternatively, the activities could be cryptic or low in some of the organisms in which they have not been detected.

Overview Perspective

The evolutionary history of aromatic amino acid biosynthesis in enteric bacteria can be evaluated with a reasonable degree of certainty within the γ subdivision, especially in the ancestral stage immediately preceding the divergence of enteric bacteria (42). This ancestral stage is retained in present-day γ-subdivision bacteria such as P. aeruginosa. Figure 2 shows a comparison of the ancestral stage deduced and the alteration of character states now displayed by enteric bacteria. Almost simultaneously, enteric bacteria gained a third isoenzyme of DAHP synthase (DS-Phe), discarded the tyrosine activation character state of PDT, gained the CM-T isoenzyme with the gene fusion that produced the bifunctional T protein, and eliminated the sensitivity of DS-Trp to inhibition by chorismate.

All of this created a pleasingly symmetrical pattern of regulation quite different from that in the ancestor. Three isoenzymes of DAHP synthase are specifically controlled by one of the end products, and two bifunctional proteins compete for a common substrate at the chorismate branch point. In contrast, the ancestral pathway uses Tyr as the dominating regulatory agent, since it feedback inhibits CDH and DS-Tyr and is a potent activator of the PDT activity of the P protein. Under conditions of tyrosine insufficiency, PPA is not utilized by PDT (which requires Tyr for activity) and dissociates from the P protein for utilization by either CDT or aromatic aminotransferase.

The deduction of progressively earlier ancestral states becomes increasingly speculative because sufficient data are lacking, but tentative scenarios have been suggested (3, 48). Highlights include the similarities between the CDT and CDH reactions, which raise the possibilities that CDH arose via duplication of the gene encoding CDT followed by acquisition of a pyridine nucleotide-binding domain and that AGN (like ornithine) is an ancient amino acid residue of proteins that was later replaced (48).

GENE FUSIONS

Gene fusions provide a mechanism for the physical association of different catalytic domains or of catalytic and regulatory entities. Fusion of catalytic centers presumably promotes the channeling of intermediates that may be unstable and/or low in concentration. In some cases, the nature of the catalytic reaction may be altered, as exemplified by certain aminases, which when fused with glutamine-binding proteins (glutaminase) will function as amidotransferases utilizing glutamine rather than ammonia as a nitrogen donor. Regulatory domains responsible for allostery, attenuation, and covalent modifications may have been recruited by enzymes via gene fusion mechanisms.

Nested Gene Fusions Identify Phylogenetic Clusters

Thus far, a given gene fusion has not been found in more than one phylogenetic cluster (45). Hence, although many multifunctional proteins have been described and the list is expanding, the gene fusions presumed responsible for their origin must be difficult to establish. Once these gene fusions have evolved, a great selective advantage is implied by the strong conservation of this character state within the lineages that radiated from the initial ancestor. If a gene fusion occurred recently, the phylogenetic cluster marked by its presence will be small, as exemplified by the distribution of the bifunctional T protein (TyrA). On the other hand, if a gene fusion occurred long ago, the phylogenetic cluster marked by its presence will be much larger, as exemplified by the bifunctional P protein (PheA).

Figure 3 shows the nested distribution of four gene fusions that led to bifunctional proteins present in modern-day E. coli. AS:PRT is limited to one group of enteric bacteria denoted as enterocluster 1 (9). The presence or absence of AS:PRT is an apt probe for a better classification of the heterogeneous group of enteric bacteria presently referred to as the Herbicola-Agglomerans complex (7). On the other hand, PRAI:IGPS and the T protein are present in all enteric bacteria, and the bifunctional P protein exists throughout the γ and β subdivisions of gram-negative bacteria. 16S rRNA sequencing provides the best basis for determination of phylogenetic trees (77). However, this method is not perfect and is vulnerable to the same flaws inherent in the use of other single molecules for tree making. Gene fusions define discrete clusters nested within other clusters. The repertoire of nested gene fusions could be greatly expanded if we knew the phylogenetic boundaries of other gene fusions known to exist in E. coli and S. typhimurium. These include aspartokinase I:homoserine dehydrogenase I (22), aspartokinase II:homoserine dehydrogenase II (22), phosphoribosyl-AMP cyclohydrolase:phosphoribosyl-ATP pyrophosphohydrolase (19), imidazole glycerol phosphate dehydratase:histidinol phosphatase (19), histidinol dehydrogenase:histidinal dehydrogenase (19), and the quadrifunctional 3-hydroxyacyl-coenzyme A (CoA) epimerase:Δ ³-cis-Δ ²-trans-enoyl-CoA isomerase:enoyl-CoA hydratase:3-hydroxyacyl-CoA dehydrogenase (83). The phylogenetic distribution of all of the latter bifunctional proteins beyond E. coli and S. typhimurium is thus far unknown, except for the report (51) that Serratia marcescens possesses an aspartokinase/homoserine dehydrogenase system of differentially regulated isoenzymes that is apparently identical to that of E. coli.

Fig. 3Nested gene fusions specifying bifunctional enzymes of aromatic amino acid biosynthesis. Enterocluster 1 possesses four gene fusions that produce the following bifunctional proteins: (fusion 1 AS:PRT (anthranilate synthase;anthranilate 5-phosphoribosy1 pyrophosphate phosphori-bosyltransferase); (fusion 2) PRAI:IGPS (Phosophoribosyl anthranilate isomerase:indoleglycerol phosphate synthase); (fusion 3) CM-T:CDH (T-protein); and (fusion 4) CM-P:PDT (P-protein). The T-protein is known to include Aeromonas and Alteromonas (genera not usually placed with enteric bacteria). This larger group has been referred to as the enteric lineage (2). As indicated by dashed lines, it is unknown whether the gene fusion responsible for the PRAI:IGP protein extends to Aeromonas and/or Alteromonas. The P-protein is distributed throughout superfamily B and superfamily A. None of these four gene fusions has been found in ?-subdivision members (superfamily C), but other gene fusions undoubtedly are uniquely clustered within this subdivision.

Source:

Enteric Ureases

Urease is a nickel-containing enzyme in some enteric bacteria that hydrolyzes urea to form ammonia and carbamate. The native enzyme consists of UreA (an 11-kDa subunit), UreB (a 12-kDa subunit), and UreC (a 60-kDa subunit). UreA and UreB are recognizable domains (on the criterion of amino acid sequence alignment) of a 27-kDa subunit present in Helicobacter pylori, which also possesses a 62-kDa homolog of UreC. UreA, UreB, and UreC also correspond to readily aligned domains present in the single 91-kDa subunit of jack bean ureases (23). Thus, in contrast to the enteric gene fusions considered above, enteric bacteria also possess unfused genes that have undergone fusion in other phylogenetic lineages.

GENE FAMILIES THAT SPAN DIFFERENT BIOCHEMICAL PATHWAYS

The pyruvate decarboxylase (PDC) superfamily has been selected for in-depth consideration of homologous proteins that have become specialized for function in different biochemical pathways. A few examples of other superfamilies whose members also function in different metabolic networks are then described briefly.

The PDC Superfamily

A feasible scenario representing the divergent evolution of the PDC superfamily is given in Fig. 4. It is based on the dendrogram derived by me from the multiple alignment of deduced amino acid sequences. The superfamily shown contains homologous Mg²⁺-thymine PP_i (TPP) enzymes that perform six distinct functional roles. Benzoylformate decarboxylase (BDC) from Pseudomonas putida (73) has no known functional equivalent in enteric bacteria. Although PDCs from Z. mobilis (24) and Saccharomyces cerevisiae (49) do not exist as monofunctional entities in enteric bacteria, a pyruvate dehydrogenase multienzyme complex is present whose E1 component is TPP dependent and catalyzes decarboxylation of pyruvate. PDC and the E1 component of pyruvate dehydrogenase are not obvious homologs, but a common origin followed by wide divergence has been suggested (35). If this in fact occurred, a superfamily membership expanded from that shown in Fig. 4 exists. The E1 components of pyruvate dehydrogenase and 2-oxoglutarate dehydrogenase participate in protein-protein interactions as components of homologous multienzyme complexes. Therefore, substantial evolutionary divergence might be accounted for by selection for sequence alterations needed to promote favorable protein-protein interactions within the complex. Comparable selective pressure would not exist for members of the currently constituted PDC gene superfamily, so wide divergence can be expected.

Fig. 4Proposed evolutionary events within the PDC gene superfamily. Steps: 1, gain: FAD-binding domain, ubiquinone-binding domain, and lipid anchor; 2, loss: lipid anchor; 3, gain: small subunit; increased pH optimum; increased affinity for ?-ke-tobutyrate; 4, gain: specificity for glyoxylate; 5, loss: FAD-binding domain, ubiquinone-binding domain. Abbreviations: Zm, Z. mobilis; Sc, Saccharomycescerevisiae; Ee, E.Coli; Lp, Lactobacillusplantarum; Kp, KlebsiellaPneumoniae; Pp, Pseudomonas putida; PDC, pyruvate decarboxylase; POX, Pyruvate oxidase; AHS, acetohydroxy acid synthase; GCL, glyoxylate carboligase; ALS, catabolic x-acetolactate synthase; BDC, benzoylformate decarboxylase.

Source:

The pyruvate oxidase (POX) family diverged as distinctive peripheral-membrane redox flavoproteins by acquisition of a carboxy lipid anchor and FAD- and ubiquinone-binding capabilities. The C-terminal 20 amino acids (lipid anchor sequence) of E. coli POX are known to be essential for lipid binding. Related sequences have no counterpart to the lipid anchor sequence in multiple alignment comparisons. The Lactobacillus POX and the Escherichia POX have since experienced substantial catalytic divergence, as shown in Table 3.

Table 3Parallel reactions of pyruvate carboxylase superfamily

A subsequent signal event of evolution abolished the lipid anchor and redox function, generating the acetohydroxy acid synthase (AHS) subfamily. Following gene duplication of the ancestral poxB gene, one copy presumably suffered a deletion that eliminated the C-terminal amino acids responsible for membrane association. Indeed, when the anchor sequence of contemporary poxB was deleted in the laboratory, POX activity was completely lost, but the (poor) ability of this enzyme to catalyze the two AHS reactions was retained (20). AHS, glyoxylate carboligase (GCL), and α-acetolactate synthase (ALS) all condense two keto acids (Table 3) but exhibit different substrate specificities and regulation.

Although the evidence (20) that POX and the three AHS isoenzymes (large catalytic subunits) of isoleucine, valine, and leucine biosynthesis are homologs was initially surprising, the catalytic reactions shown in Table 3 for these enzymes plus other members of the protein family exhibit many common features, including utilization of pyruvate (or a closely related molecule) as substrate and formation of CO₂ as product as well as a catalytic requirement for TPP and Mg²⁺.

In the same subfamily with the AHS isoenzymes (Fig. 4) are GCL and catabolic ALS. The reaction catalyzed by E. coli GCL (21) parallels that catalyzed by the AHS isoenzymes, except that two-carbon rather than three-carbon substrates are condensed. GCL exhibits a closer sequence relationship with the AHS isoenzymes than with catabolic ALS (64), even though ALS catalyzes the same reaction as AHS. The production of 2,3-butanediol correlates with the presence of catabolic ALS in some but not all enteric bacteria. Enteric bacteria carry out a mixed acid and butanediol fermentation. If the fermentation balance of products produced by representatives of Escherichia, Shigella, and Salmonella spp. is compared with those from representatives of Klebsiella, Serratia, and Erwinia spp., the second group produce less acid but more CO₂ and ethanol and dramatically more 2,3-butanediol (78).

Since no overall redox reaction is catalyzed, it has been a long-standing curiosity that the three biosynthetic AHS isoenzymes contain FAD. This phenomenon has now been explained as the conserved binding of FAD by the AHS proteins during evolution, with FAD now functioning not as a coenzyme but as a component needed to maintain the structural integrity of the proteins (69). Exactly the same explanation would apply to the GCL in this family, a GCL that also contains FAD but does not catalyze an overall redox reaction. It may be relevant that chorismate synthase in the common pathway of aromatic amino acid biosynthesis also contains FAD, even though no overall redox reaction is catalyzed (37). Perhaps this curious fact will help identify an ancestral progenitor of the gene encoding chorismate synthase. Additional examples of this type of evolutionary remnant can be expected to emerge and to help track previously unanticipated evolutionary relationships. In addition to FAD, ubiquinone-binding sites are present in POX, the AHS isoenzymes, and GCL from E. coli. These sites were first recognized by the finding that AHS isoenzymes were sensitive targets of herbicides such as sulfometuron methyl, which act as ubiquinone mimics (69). The ubiquinone sites are no longer recognizable in the ALS lineage (71).

This analysis illustrates the power of a dendrogram matrix for guidance in conclusions about directional aspects of evolutionary events. For example, although PDC and BDC share three contemporary character states with catabolic ALS (i.e., FAD not bound, ubiquinone not bound, and no carboxy lipid anchor), this result was reached via two entirely different evolutionary paths. While PDC and BDC presumably never were associated with FAD, catabolic ALS is seen to have an evolutionary history of FAD binding gained followed by FAD binding lost. The same is true with respect to the character states for ubiquinone binding and a carboxy lipid anchor.

The AHS subfamily subdivides into three units of specialized metabolic function, all of which are represented in enteric bacteria: AHS, GCL, and ALS. In these lineages, appropriate kinetic parameters became fine-tuned for utilization of α-ketobutyrate, glyoxylate, and pyruvate, respectively. In the AHS lineage, an improved specificity for α-ketobutyrate (12) may have extended AHS function to the biosynthesis of l-isoleucine, an amino acid that might originally have been formed through a still-existing but limited pathway derived from l-glutamate (65). Even further subdivision of function is illustrated by the divergent regulatory isoenzymes of AHS in E. coli. Control of AHS isoenzymes (but not GCL and ALS) by feedback inhibition is probably intimately related to the existence of a unique small subunit associated with AHS isoenzymes (28).

Proline-Arginine-Pyrimidine-Lysine Biosynthesis

Biochemical relationships link arginine biosynthesis and the pathways responsible for proline, pyrimidine, and lysine biosynthesis. The beginning precursor substrate for both arginine and proline is glutamate. Successive kinase and dehydrogenase steps are commonly employed in nature to accomplish the reductive conversion of a carboxyl moiety to an aldehyde. The first and second steps of proline synthesis utilize a kinase (ProA) and a dehydrogenase (ProB) to produce glutamate semialdehyde, which spontaneously cyclizes to 1-pyrroline-5-carboxylate, the immediate precursor of proline. The spontaneous cyclization is avoided in the parallel steps of the arginine pathway by means of an initial acetylation that produces N-acetylglutamate. Subsequent kinase (ArgB) and dehydrogenase (ArgC) steps produce N-acetylglutamic semialdehyde, which cannot cyclize and is therefore available for transamination to form N-acetylornithine. Deacylation produces l-ornithine, an amino acid precursor of l-arginine. In addition to the kinase/dehydrogenase enzyme tandems that function in the proline and arginine pathways, E. coli employs aspartate kinase and aspartate semialdehyde dehydrogenase reactions in the common portion of the branched pathway generating lysine, methionine, and threonine. Surprisingly, none of these three pairs of enzymes catalyzing closely parallel reactions exhibits amino acid homology (63).

On the other hand, the parallel reactions of ornithine transcarbamylase (arginine biosynthesis) and aspartate transcarbamylase (pyrimidine biosynthesis) are catalyzed by homologous proteins (75). In addition, acetylornithine deacetylase (ArgE) is a homolog of succinyldiaminopimelate desuccinylase (DapE) in the lysine biosynthetic pathway (16). These two deacylases in turn belong to a family that includes G2 carboxypeptidase from Pseudomonas sp. All three proteins display similar reactions and exhibit strikingly similar biochemical properties (16). ArgE exhibits broad substrate specificity, which might reflect retention of the substrate ambiguity of an ancestral protein. In fact, the broad specificity of ArgE is the basis of an interesting metabolic potential where a defect of proline biosynthesis (proAB) can be suppressed by a leaky argD mutation in S. typhimurium (52). Derepression of ArgE due to arginine limitation, a low supply of the normal substrate (N-acetylornithine) due to the leaky block, and an elevated level of accumulated abnormal substrate (N-acetylglutamic semialdehyde) behind the block result in conversion of N-acetylglutamic semialdehyde to the precursor of proline, glutamate semialdehyde. This conversion effectively bypasses the primary proAB defect.

Another set of cross-pathway homologs are aminotransferases. E. coli acetylornithine aminotransferase can transaminate ornithine as an alternative substrate, whereas catabolic ornithine aminotransferases from yeast cells, rats, and humans are specific for ornithine. These biosynthetic and catabolic aminotransferases are homologous (38). This homology contrasts with the lack of detectable homology between the aforementioned ArgB-ProB and ArgC-ProA pairs, even though the substrates differ in each case by only the N-acetyl moiety.

Two gene families of amino acid decarboxylases exist, each containing an ornithine decarboxylase. Ornithine, arginine, lysine, and histidine decarboxylases from enteric bacteria belong to one family of catabolic proteins (40, 58). This includes decarboxylases from Drosophila melanogaster, mammals, or plants that utilize glutamic acid, dihydroxyphenylalanine, tryptophan, or histidine (58). Specificity for a given substrate utilized may have evolved more than once, as exemplified by enteric (Morganella morganii) and mammalian (rat) histidine decarboxylases (40). This large cluster of amino acid decarboxylases is distinct from the biosynthetic diaminopimelate (DAP) decarboxylase encoded by lysA of E. coli and other prokaryotes (58). Surprisingly, mouse ornithine decarboxylase, a key enzyme of polyamine biosynthesis in mammals, belongs to the DAP decarboxylase family rather than to the bacterial ornithine decarboxylase family. Since DAP has no functional role in mammals, it has been suggested (58) that mouse ornithine decarboxylase was originally recruited from a DAP decarboxylase acquired from an endosymbiotic event that generated mitochondria. All of the amino acid decarboxylases are pyridoxal phosphate dependent, and it remains to be seen whether two highly divergent groups of common ancestry exist or whether the two groups are analogous.

Tryptophan-Cysteine-Threonine-Serine Biosynthesis

The beta subunit of tryptophan synthase (TrpB) in E. coli is homologous to O-acetylserine sulfydrylase A, which catalyzes the last step of cysteine biosynthesis (56). Although initially unexpected, a basis for common ancestry is now apparent in that both proteins bind pyridoxal phosphate and both have broad substrate specificities that overlap. In fact, the tryptophan synthase (α _2β2) can form cysteine from H₂S and l-serine in vitro. TrpB has also proven to be a homolog of threonine synthase, threonine dehydratase, and d-serine dehydratase proteins of E. coli (61, 62). These proteins are all pyridoxal phosphate dependent.

In contrast to these unexpected gene family affiliations for TrpB, a seemingly probable homology relationship that was predicted (41) did not materialize. Tryptophanase, another pyridoxal phosphate-dependent enzyme, possesses broad substrate specificity and is able to condense serine and indole to yield tryptophan and H₂O in vitro. Indeed, constitutive tryptophanase mutants are capable of suppressing trpB auxotrophs. However, no homology is detectable between trpB and tna (62), and these proteins are evidently analogous, with separate ancestries. Thus, although homologous proteins can be expected to exhibit similar mechanisms of catalytic reaction and perhaps even functional overlap, these similarities are no guarantee of homology.

Aminotransferase Gene Families

Some aminotransferases exhibit broad specificity for substrates, and this substrate ambiguity is reflected in vivo by the ability of a given aminotransferase to suppress a deficiency in another aminotransferase (46). E. coli ilvE (branched-chain aminotransferase), tyrB (aromatic aminotransferase), and aspC (aspartate aminotransferase) exhibit functional overlap in vivo (33). On this basis, one might expect (46) ilvE, tyrB, and aspC to make up a gene family. Indeed, tyrB and aspC have been proven to belong to the class I aminotransferase family, which contains the well-studied vertebrate aspartate aminotransferases (both mitochondrial and cytosolic). Unexpectedly, this gene family also contains hisC (encoding imidazole acetol phosphate aminotransferase) but not ilvE (59). Thus, ilvE exhibits functional overlap with tyrB and aspC as a result of convergence of nonhomologous genes (or perhaps as a result of a divergence of homologous genes that is so extensive that homology cannot be established with current data).

Although hisC exhibits no functional overlap with tyrB and aspC in E. coli, the corresponding gene in Bacillus subtilis (hisH) is capable of functioning for aromatic amino acid biosynthesis in vivo (60). In this case, hisH is the only histidine pathway gene to be located outside of the histidine operon. Thus, it seems possible that the tight repression control of hisC as a member of the E. coli his operon masks the potential of E. coli hisC to suppress deficiencies of tyrB and/or aspC in vivo. If this is so, perhaps his constitutive regulatory mutants indeed suppress tyrB and/or aspC mutant deficiencies.

Genes specifying other aminotransferases can be grouped into separate gene families (59). Once a sufficient number of sequences are obtained and representative crystal structures become available, some or all of these gene families may prove to be homologous members of one superfamily.

BASIS FOR PRESENT-DAY INTERLOCKING RELATIONSHIPS OF METABOLISM

The foregoing examples illustrate homology relationships that can extend throughout a complex metabolic network. These ancestral relationships can be the basis for suppressor capabilities, for global relationships of regulation between pathways, and for unexpected gene organizations such as complex operons (54).

Methionine-Peptidoglycan Relationship

Methionine and meso-DAP are generated by pathways that diverge from l-aspartate semialdehyde. Enzymes of the methionine pathway possess an intriguing latent capability for recruitment to function in replacing meso-DAP for peptidoglycan biosynthesis. This capability comes to the fore in mutants unable to synthesize normal meso-DAP.

The E. coli cell wall consists largely of a peptidoglycan whose disaccharide-pentapeptide polymer is cross-linked laterally by means of a peptide motif. meso-DAP is a critical amino acid component of cross-linking. In mutants that lack the ability to synthesize meso-DAP, a potential for substituting sulfur-containing amino acids (meso-lathionine, l-cystathionine, and l-allocystathionine) for meso-DAP exists (68). Surprisingly, such a remodeled peptidoglycan, which must possess gross stereochemical perturbations of peptidoglycan configuration, did not cause loss of viability.

The endogenous production of the sulfur-containing amino acids required overexpression of metB and inactivation of metC. Cystathionine synthase (MetB) exhibits broad substrate specificity, e.g., being able to convert l-cysteine to l-lanthionine. Cystathionase (MetC) normally converts l-cystathione to l-homocysteine during methionine biosynthesis and is highly efficient in utilization of l-lanthionine and l-allocystathione as substrates as well. Thus, the scavenging capabilities of wild-type MetC mask the potential of the cell to incorporate sulfur-containing amino acids into peptidoglycan in place of meso-DAP.

Latent Crossover Function of Homologs

Many gene families have diverged to produce subfamilies that have acquired specialized functions. TrpG and PabA exemplify such paralogous proteins that belong to different subfamilies. If one gene were lost, the potential might exist for reacquisition of the lost function following gene duplication and appropriate evolutionary remodeling of the remaining gene homolog. This would in fact be a case of extracistronic suppression. Such a crossover event would result in a new gene that would uniquely cluster in a subfamily of genes encoding proteins catalyzing a different (but related) reaction. Crawford (25) termed this situation "reticulate evolution" in consideration of TrpG of fluorescent pseudomonads, whose amino acid sequence clusters with PabA sequences of enteric bacteria rather than with other TrpG sequences.

Pyridoxine and Amino Acid Biosynthesis

Pyridoxine is a precursor of pyridoxal phosphate, a crucial coenzyme for many enzymes. Lam and Winkler (54) summarized relationships that tie pyridoxine and serine biosynthesis together in E. coli. The phosphorylated pathway of serine synthesis and part of the pyridoxine biosynthesis pathway follow parallel steps. SerA and PdxB are homologs that catalyze parallel dehydrogenase steps. They are, in fact, members of a larger family of 2-hydroxyacid dehydrogenases (34). SerA is capable of suppressing a PdxB deficiency. SerC is an aminotransferase that also functions directly to catalyze a parallel reaction in pyridoxine biosynthesis. Thus, SerC catalyzes a step in the biosynthesis of a cofactor that is obligatory for SerC function.

Metabolic relationships can be seen in some operon structures. serC coexists with aroA in a complex operon, i.e., an operon containing genes of apparently unrelated functions. This coexistence may reflect the input of both serine and chorismate for synthesis of the iron-binding agent enterochelin. Every molecule of tryptophan synthesized also requires one molecule of serine at the level of tryptophan synthase. Phenylalanine, tyrosine, and serine biosyntheses all require pyridoxal phosphate for essential aminotransferase steps as well. Thus, the complex operon may serve to coordinate a key step of aromatic biosynthesis with serine biosynthesis in the context of these relationships.

pdxB exists in another complex operon with hisT. Histidine biosynthesis involves a pyridoxal phosphate-dependent aminotransferase step. The range of repression-derepression for the histidine pathway enzymes in E. coli is very substantial, and the complex operon may be a mechanism that helps adjust pyridoxine supply to variations in the levels of expression for imidazole acetol phosphate aminotransferase. Other pdx genes remain to be identified and mapped. Perhaps one or more of these will prove to belong to other complex operons containing genes related to a pathway having an enzyme requiring pyridoxal phosphate as a cofactor.

CRYPTIC GENES

Cryptic genes can now be detected more readily than in the recent past by the use of molecular probes. Hall et al. (36) suggested that such genes are subject to periodic selection at the population level. The lack of expression of AHS II in E. coli K-12 (55) and AHS III in S. typhimurium (67) is due to point mutations that are readily reverted by single reversion events of mutation.

The E.coli gene argM encodes an enzyme able to catalyze the acetylornithine aminotransferase reaction, which is normally carried out by ArgD (26). This cryptic protein can be reactivated in argD-deficient mutants under selective conditions. ArgM has broad substrate specificity and is inducible by arginine. Since the ability of ArgM to transaminate succinylornithine and ornithine as alternative substrates in vitro is excellent, it has been suggested that ArgM is a remnant of a pathway of arginine catabolism (26). If it is, then other cryptic genes belonging to the entire catabolic pathway might be present.

E. coli does not ordinarily synthesize pyrroquinoline quinone (PQQ), a prosthetic group of some dehydrogenases that oxidize alcohols or sugars. However, E. coli expresses an inactive glucose dehydrogenase apoprotein that can be activated if PQQ is provided (39). The presence of genes encoding pqqE and pqqF in E. coli was indicated by functional complementation of Methylobacterium organophilum mutants (74). Indeed, mutants of E. coli that are capable of restored PQQ synthesis have been selected (13).

It was mentioned previously that the periplasmic enzymes CM-F and CDT exhibit an erratic distribution in enteric bacteria that might reflect horizontal gene transfer. However, a viable and testable alternative possibility is that these proteins are encoded by genes (aroQ and pheC, respectively) that are cryptic in some enteric bacteria.

The existence of other cryptic genes seems to be a distinct possibility whenever the distribution of a gene product is erratic within a closely related group of organisms such as the enteric bacteria. For example, tyrosine phenol-lyase is expressed in Erwinia herbicola (32). Although it is also produced in Escherichia intermedia (53), it is not expressed in the closely related E. coli K-12, which therefore might possess a cryptic gene.

SUMMARY

In this chapter, the PDC superfamily has been used to exemplify how evolutionary relationships can be traced across biochemical pathway boundaries, and aromatic amino acid biosynthesis has been highlighted to illustrate how the origin and refinement of a given biochemical pathway can be traced backward through evolutionary time. Genes that specify individual proteins that function within discrete biochemical units have been recruited from gene families whose members share common catalytic mechanisms, small-molecule-binding capabilities, novel physical properties, or regulatory features. Molecular genetic data have produced numerous examples of, on one hand, expected homologies that did not materialize and, on the other hand, totally unexpected homologies that did materialize. Unexpected homologies are inevitably explicable in terms of similar mechanisms. Ongoing refinement of analysis will probably show that many of the cases in which the expected homology was not initially apparent reflect extensive divergence from a common ancestor rather than an independent origin.

ACKNOWLEDGMENT

I thank Jian Song for his help with the illustration art work.