In recent publications involving bioinformatic analysis of aromatic amino-acid biosynthesis [9,10,11,12,13,14], we have implemented a nomenclature that attempts logical and consistent naming of genes and their gene products for different organisms. The problem is exemplified by the contemporary naming in two model organisms of 3-deoxy-D-arabino-heptulosonate 7-phosphate (DAHP) synthase, the initial enzyme step of aromatic amino-acid biosynthesis. In Bacillus subtilis the gene, appropriately enough, has been named aroA, but in Escherichia coli the equivalent function is represented by genes encoding three differentially regulated paralogs. These E. coli genes have been named aroF, aroG and aroH. In B. subtilis these latter three gene designations refer to 5-enolpyruvylshikimate-3-phosphate synthase (step 6 of chorismate synthesis), chorismate synthase (step 7) and chorismate mutase (initial step of phenylalanine and tyrosine biosynthesis). Even in B. subtilis, where the naming was intended to follow an orderly progression in terms of order of reaction steps, there is the complication that DAHP synthase is expressed as a fusion of two catalytic domains, one being a class of chorismate mutase called AroQ. This requires naming at the level of domain and the designation aroQ• aroA was implemented to denote such a fusion [9,10,11,12]. Thus, a single enzymatic function in one organism is accommodated through the cumulative expression of three paralog genes, but in another organism is only encoded by a portion of a single gene. A universal nomenclature is needed that labels at the level of domain, that labels in synchrony with order of reaction steps as much as possible, and that labels isofunctional paralogs at the same hierarchical level but with discriminating identifiers.

The status of nomenclature for the tryptophan pathway is not so chaotic as the genes in almost all prokaryotes have been named in line with E. coli, but even here distinct problems have arisen because gene fusions that exist in E. coli are often absent elsewhere. Much of the literature up to at least 1996 used five gene designations for the seven protein domains [15]. In B. subtilis these seven protein domains are encoded by separate genes [16]. Thus, E. coli trpD encodes the equivalent of B. subtilis TrpG and TrpD. The E. coli gene fusion has been re-designated with the convention of a bullet to represent a fusion: trpG• trpD [17]. Likewise, the E. coli trpC encodes the equivalent of B. subtilis TrpC and TrpF, and the E. coli gene fusion has been re-designated trpC• trpF. The nomenclature we advocate is more easily remembered because genes are named in the order of pathway reaction. Subunits for a given reaction are named at the same hierarchical level: anthranilate synthase (first reaction) consists of TrpAa and TrpAb, and tryptophan synthase (fifth reaction) consists of TrpEa and TrpEb. The implementation of a logical and consistent nomenclature, as illustrated in Figure 1, should be most helpful in the long term.

Initial amino-acid alignments were generated using ClustalW software, version 1.4 [18]. Manual adjustments were made through visual inspection to bring conserved motifs and residues into register. This was implemented by use of the BioEdit multiple alignment tool [19]. Inferences about the evolutionary relationships within the TrpEa and TrpEb protein families were made using the PHYLIP package of programs [20]. The Protpars program was used to generate a maximum parsimony tree, and the neighbor-joining and Fitch programs were used to generate a distance-based tree. The distance matrix used in the latter programs was produced using the program Protdist with a Dayoff PAM matrix. The Seqboot and Consense programs were then used to assess the statistical strength of the tree using bootstrap resampling. Neighbor-joining (PHYLIP), Fitch and Margolash (Fitch in PHYLIP), and maximum parsimony methods [21] all produced trees consistent with one another. Despite low bootstrap values at many individual internal nodes, the clusters formed and arrangement of taxa within them were largely identical. Ninety TrpEb and 63 TrpEa sequences were analyzed.

TrpEb proteins divide into two distinctly different groups, as illustrated by the unrooted tree shown in Figure 2. The major group, denoted TrpEb_1, includes the well-studied enzymes from such organisms as E. coli, Salmonella typhimurium, and B. subtilis. The minor group, denoted TrpEb_2, is represented heavily, but not exclusively, by archaeal proteins. Among the current inventory of completed archaeal genomes, only Methanococcus jannaschii lacks TrpEb_2. Seven archaeal genomes possess both TrpEb_1 and TrpEb_2 (Methanosarcina barkeri possesses two paralogs of TrpEb_1 in addition to one species of TrpEb_2). Six archaeal genomes possess TrpEb_2, but not TrpEb_1.

Six organisms (all Archaea) possess intact tryptophan pathways, but they lack TrpEb_1. TrpEb in Thermoplasma volcanii, T. acidophilum, and Ferroplasma acidarmanus is represented only by a single species of TrpEb_2. Although Sulfolobus solfataricus, Aeropyrum pernix and Pyrobaculum aerophilum all possess two species of TrpEb, both are the TrpEb_2 variety. Thus, in all six of these lineages TrpEa either might be unable to form a tight complex with TrpEb_2, or might have evolved different protein-protein contacts. In the latter case, distinct TrpEa subgroupings might be expected in parallel with the two TrpEb subgroupings. On the contrary, all TrpEa sequences fall into a single group (Figure 4). However, in contrast to sequences present in those Archaea that do possess TrpEb_1 (for example, Archaeoglobus, species of Pyrococcus, Methanococcus, Methanobacterium, Methanosarcina and Halobacterium), the six archaeal lineages that possess only TrpEb_2 have very distinctive elongated branches on the TrpEa tree (Figure 4). This suggests an elevated rate of evolutionary divergence, due either to selection for new productive contacts of TrpEa with TrpEb_2 or to lack of constraint to maintain TrpEa residues previously important for contacts with TrpEb_1.

The long branch of the TrpEa sequence of Chlamydia trachomatis reflects its likely status as a pseudogene. This is consistent with the observation that C. trachomatis TrpEb_1 (Figure 2) also seems to be a pseudogene. One does not expect positive selection for maintenance of function in C. trachomatis as it lacks an intact tryptophan pathway. Indeed, the alteration in C. trachomatis of many otherwise invariant amino-acid residues is evident from the information given in Table 2.

Figure 5 shows an alignment of the amino-acid sequence of TrpEb_1 from S. typhimurium with TrpEb_2 from P. furiosus. Each sequence is shown as a template for its own subfamily, as extracted from a refined multiple alignment. Conserved residues deduced from a full multiple alignment (available from the author on request) are indicated, as are the gap positions present in the full alignment. Functional roles in catalysis and allosteric regulation are indicated for the S. typhimurium TrpEb_1 sequence in order to compare similarities and differences between TrpEb_1 and TrpEb_2 proteins. Residues that are ligands of pyridoxal phosphate or that interact with pyridoxal phosphate are scattered throughout the sequences, including the catalytic K87, and are highly conserved. Residue E109 has been shown to render indole more nucleophilic via proton abstraction from N1 [23]. The serine substrate-binding region is highly conserved, as is a monovalent cation (MVC) binding region [7] coordinating with G232, F/Y306, and G/A/S308. A number of indels (insertions/deletions) distinguish TrpEb_1 and TrpEb_2, and TrpEb_2 is about 50 residues longer overall than TrpEb_1. In addition to other residues conserved between both TrpEb_1 and TrpEb_2, each subgroup has its own repertoire of uniquely conserved residues. The COMM domain [24], a rigid but mobile domain as originally defined with S. typhimurium TrpEb_1 [25], differs from the corresponding region of the TrpEb_2 subfamily by the presence of an indel. Key TrpEb_1 regulatory residues (R141, K167) within this region as well as one residue near the MVC site (D305) are not conserved in the TrpEb_2 subgroup.

The tryptophan synthase of S. typhimurium is a rigorously documented example of substrate channeling in which indole generated as an intermediate is passed through an internalized tunnel ([7] and references therein). Ligand binding at the α-site and covalent transformations at the β-site accomplish mutually reinforcing overall allostery. The movable COMM domain is comprised of residues G102 to G189 in S. typhimurium TrpEb_2. COMM interacts with both the β-active site and with α-subunit loops 2 and 6 in response to allosteric signals. Within the COMM domain of TrpEb_1, S178 participates in intersubunit signaling with G181 of TrpEa. Competing allosteric conformations are mediated by alternative salt bridges between K167 of TrpEb_1 and D305 of TrpEb_1 on the one hand, or between K167 of TrpEb_1 and D56 of TrpEa, on the other. When D305 of TrpEb_1 is not occupied with K167, it forms an alternative salt bridge with R141, as shown in Figure 5.

As intersubunit signaling between TrpEb_2 and TrpEa is either lacking or involves different contacts, one might expect the important catalytic residues, but not the allosteric residues, to be conserved in comparison of TrpEb_2 with TrpEb_1. This comparison is shown in Figure 5. Likewise, in those TrpEa proteins that lack a TrpEb_1 partner, instead being forced to function in concert with TrpEb_2, one might expect retention of catalytic residues but loss of allosteric residues. This comparison is shown in Figure 6. Of the intersubunit signaling pair TrpEb_1 S178/TrpEa G181, S178 is not conserved within its own subfamily and equivalent residues seem to exist [26]. It is, however, striking that G181 is invariant in all TrpEa proteins, except for those belonging to the six archaeal lineages lacking TrpEb_1. It is even more striking that K167, which participates in the alternative β-β or β-α salt bridges is invariant in TrpEb_1, but absent in TrpEb_2. Likewise, the salt-bridge partner D305 is invariant in TrpEb_1, but absent in TrpEb_2. The salt-bridge partner D56 in TrpEa is invariant except for the six archaeal lineages that lack TrpEb_1. Figure 5 shows that TrpEb_2 sequences carry an insertion of 16 residues in the general region corresponding to the COMM domain of TrpEb_1 proteins.

The organization of tryptophan-pathway genes in the ten archaeal genomes and six bacterial genomes that are thus far known to possess TrpEb_2 are displayed on a 16S rRNA tree in Figure 7. Pyrococcus horikoshii has lost the entire pathway, and only trpEb_2 is present. In Aquifex and Chlorobium all trp genes are dispersed, and trpEa is not linked to either trpEb_1 or trpEb_2. In Bacteria and Archaea the gene order trpEb_1 → trpEa is one of the most highly conserved of genomic gene arrangements. Often translational coupling exists, as seen in Figure 7 for P. furiosus, P. abyssi, M. thermoautotrophicum, T. volcanium, Archaeoglobus fulgidus and T. maritima. In each of the latter genomes trpEb_2 is outside the trp operon. Both species of Thermoplasma possess an otherwise intact trp operon that lacks trpEb_1 (also not present elsewhere in the genome). Hence, by default it appears that the unlinked trpEb_2 must function for tryptophan biosynthesis in these organisms. In A. pernix and S. solfataricus all of the trp genes are adjacent, but trpEb_2 flanks trpEa, and trpEb_1 is absent from these genomes. In each case, a second unlinked copy of trpEb_2 is present. In Geobacter sulfurreducens, both trpEa and trpEb_1 (presumably partnered based on the results shown in Figure 4) are unlinked to one another, and both are outside an otherwise intact trp operon. In this case it seems curious indeed that the operon contains trpEb_2.

A snapshot of the incredibly dynamic alteration of tryptophan gene organization in prokaryotes is apparent from Figure 7. Of the organisms shown in Figure 7, the most consistent linkage is that of trpAa and trpAb. In A. fulgidus trpD and trpB are fused, whereas in Rhodopseudomonas palustris trpAa and trpAb are fused. Operons are sometimes incomplete as with P. aerophilum and G. sulfurreducens, or fragmented, as with R. palustris. In A. pernix an inverted arrangement yields a divergent transcription of trpEa → trpEb_2 → trpC → trpD and trpB → trpAa → trpAb. This is also one of the very few cases where the order is trpEa → trpEb instead of the usual trpEb → trpEa.

The Ferroplasma genome illustrates a case where a non-tryptophan pathway gene, aroA encoding DAHP synthase, is a member of the operon (translationally coupled). The implied transcriptional control of DAHP synthase by L-tryptophan potentially could produce growth inhibition by L-tryptophan because of limitation of precursors needed for L-phenylalanine and L-tyrosine biosynthesis. This is because no other genes encoding DAHP synthase appear to be present in the genome. It would be interesting to know whether the Ferroplasma DAHP synthase is sensitive to allosteric control or not. The phenomenon of growth inhibition triggered by exogenous amino acids is exemplified by the effect of exogenous L-phenylalanine upon DAHP synthase in Neisseria gonorrhoeae [27] and in other organisms ([27] and references therein).

What is the function of TrpEb_2? As previously discussed, it seems clear that TrpEb_2 has sometimes been pressed into service with TrpEa to function as tryptophan synthase. What might be its function in those situations where trpEb_2 is isolated away from a typical tryptophan operon, which possesses closely linked or translationally coupled genes specifying trpEa and trpEb_1?

TrpEb_1 from S. typhimurium is the prototype member of a superfamily of pyridoxal phosphate (PLP)-dependent enzymes that are of remote relationship and that catalyze β-replacement and β-elimination reactions [28,29,30]. These include O-acetylserine sulfhydrylase, threonine deaminase, threonine synthase, cystathionine β-synthase, 1-aminocyclopropane-1-carboxylate deaminase, L-serine dehydratase, and D-serine dehydratase. Isolated TrpEb_1 does catalyze the reaction of L-serine dehydratase (deaminase) in vitro, but does not support significant levels of the other activities. It does not seem likely that TrpEb_2 catalyzes the latter reactions either, as Psi-Blast of TrpEb_2 sequences did not return hits for them any more avidly than was the case for TrpEb_1 queries.

Perhaps the most likely 'alternative' function of TrpEb_2 proteins is a catalytic activity already established as an in vitro activity of isolated TrpEb_1, of which there are two. The L-serine + indole → H₂O + L-tryptophan reaction might function alone for tryptophan biosynthesis in cells that acquire indole from the environment. Little information about the availability and utilization of indole in nature seems to exist. Model organisms such as E. coli and B. subtilis transport indole poorly and it tends to be toxic, but these organisms lack TrpEb_2.

As stated above, the second established in vitro activity of isolated TrpEb_1 dimers is L-serine dehydratase (deaminase): L-serine → pyruvate + ammonia. It is indeed suggestive that the presence of TrpEb_2 correlates almost perfectly with the absence of the primary L-serine deaminase (COG1760) used in nature (Table 3). This SdaA class of serine deaminase is an iron-sulfur protein, not a pyridoxal 5'-phosphate protein [31]. It is absent from the Archaea. Although widely distributed in the Bacteria, it is not present in Thermotoga, Aquifex, Chlorobium, Geobacter and Rhodopseudomonas. Mycobacterium smegmatis is the only TrpEb_2-containing organism to possess an SdaA homolog.

There are other candidates for carrying out the serine deaminase reaction. A serine deaminase that is PLP-dependent exists in eukaryotes, but seems to be absent or rare in prokaryotes. Threonine dehydratase has a lesser capability to carry out the serine dehydratase reaction in vitro. 'Biosynthetic' (IlvA) and 'catabolic' (TdcA) threonine deaminase are homologs, except that IlvA has a unique carboxy-terminal extension that provides an allosteric module. Most archaea lack ilvA, and only a limited number possess catabolic threonine dehydratase.

If archaea lack ilvA, what is the source of α-ketobutyrate for isoleucine biosynthesis? It appears likely that α-ketobutyrate is generated instead by the 'pyruvate' pathway in which citramalate is generated as the initial intermediate. A few tracer studies in the older literature have indicated derivation of α-ketobutyrate from pyruvate rather than from threonine [32,33], and (R)-citramalate synthase activity has been shown recently in M. jannaschii (MJ1392) [34]. This pathway was initially shown to exist in Leptospira [35], and it seems likely that more examples will surface. Indeed, it is interesting that an enteric bacterium has a latent potential to replace the threonine deaminase step with a pyruvate-derived pathway under appropriate selective conditions [36]. The citramalate pathway for conversion of pyruvate to 2-ketobutyrate involves a carbon-chain elongation mechanism that uses an initial step of condensation with acetyl-CoA, followed by rearrangement, oxidation and elimination of a carbon to produce a keto acid differing from the original substrate by the presence of an additional carbon. This mechanism is familiar in nature, analogous steps being used in the TCA cycle, the ketoadipate pathway of lysine biosynthesis, and in leucine biosynthesis. These analogous steps in different pathways were originally used in formulating the recruitment hypothesis for evolutionary acquisition of new function [37], and it is interesting that the citramalate synthase gene shown in M. jannaschii had previously been annotated as α-isopropylmalate synthase (which catalyzes the initial step of leucine biosynthesis).

Overall, the cumulative evidence indicates that established sources of L-serine dehydratase are low or absent in organisms that possess TrpEb_2, and therefore points to a plausible role for TrpEb_2 as L-serine dehydratase.