Lateral gene transfer (LGT) undoubtedly occurs with high frequency [1]. But what is required for any given LGT event to escape transience and actually survive as a non-vertical contributor to the evolutionary history of a species? (i) Initially the incoming gene(s) must be replicated in the original recipient cell and its immediate progeny. (ii) Selective advantages of the LGT-gene(s) must be sufficient to foster eventual domination of the species population. (iii) During this time a crucial factor influencing survival will be whether the considerable demands imposed by the recipient genome for amelioration of alien genes can be met [2]. Obviously, there will be less amelioration pressure if the donor and recipient genomes are phylogenetically close and have similar guanine/cytosine base ratios, dinucleotide frequencies, promotor motifs, and so on. Even here a LGT insertion into a genome might be contra-selected for other reasons, for example, if the location of the insertion disrupts the symmetry and physical balance between the origin and terminus points of replication [3]. There is a balance in that one can expect alien sources of greatest novelty to be phylogenetically distant, yet genes from such remote sources will usually confront the greatest amelioration pressures. The most obvious candidates as successfully imported alien genes will encode novel functions that confer clear selective value, such as resistance to threatening environmental agents (e.g. antibiotics) or ability to utilize a new source of carbon and energy.

Enhanced probability of LGT success can be expected if only a single gene is needed for the novel function. This does not necessarily rule out the acquisition of a new function that is complex and multi-genic. Sometimes where existing functions are complex and multi-step, a single incoming gene can create a new metabolic linkage. For example, an organism having a multi-step pathway for tyrosine catabolism could acquire a previously unavailable ability to also catabolize phenylalanine through import of a gene encoding phenylalanine hydroxylase (which converts phenylalanine to tyrosine). In cases where the total repertoire of multiple steps that define a novel function are all absent in a given organism, acquisition of that function by LGT would be highly improbable were it not for the existence of operon modules in prokaryotes. Lawrence has proposed, in fact, that gene organization as operons largely exists because of "selfish" properties that operate at the hierarchical level of genes [4,5]. Although genes of L-tryptophan (Trp) biosynthesis, a classic operon system, are completely dispersed in some prokaryote genomes (e.g. Aquifex, Chlorobium, Wolinella and unicellular cyanobacteria such as Synechococcus, Synechocystis, and Prochlorococcus), they usually exist either as whole-pathway operons or as a combination of several partial-pathway operons [6].

Approaches for the detection of LGT events are either phylogenetic or parametric. Parametric approaches include the detection of nucleotide composition, dinucleotide frequencies and codon-usage biases in gene segments that are atypical of the recipient genome. Such parametric analysis is well illustrated by a study of the Escherichia coli genome [1]. However, such atypical parametric properties will drift with time toward those of the recipient genome (amelioration). Therefore, parametric analysis is limited to detection of genes that were acquired recently. We have found only a single example where trp genes exhibited parametric features suggesting LGT. In this case a low-GC gene block in Xylella fastidiosa contains seven genes, of which two encode the subunits of anthranilate synthase and one encodes a repressor, TrpR [7]. On the other hand, phylogenetic analysis can detect ancient events of LGT, and this approach has allowed the recognition of a number of whole-pathway and partial-pathway trp operons that were acquired by LGT. Presumably, the initial aberrant parametric properties associated with LGT have been ameliorated. Phylogenetic analysis also is a much more powerful indicator of the likely donor lineage following gene acquisition by LGT.

Ubiquitous pathways of primary biosynthesis, having a long history of genomic optimization and metabolic integration, are generally improbable candidates for replacement by LGT [8]. Among the challenges confronting successful LGT are that key regulatory genes may be spatially separated from the structural genes, and promoters and transcription signals vary between bacterial species, as also is the case for elements required for translation.

Exceptions can be envisioned, among them: (i) A gene of primary biosynthesis could, in fact, be synonymous with a gene of antibiotic resistance. Thus, if one or more primary-pathway enzymes is the target of an antimicrobial agent, then genes in nature that encode resistant versions of that enzyme(s) could confer strong selective pressure for replacement. The apparent displacement of an essential enzyme of isoprenoid biosynthesis (hydroxymethylglutaryl coenzyme A reductase) in Archaea by a statin-resistant enzyme of bacterial origin exemplifies this [9]. (ii) If primary pathway genes are lost by deletion, as often occurs with pathogens or symbionts, re-acquisition of the pathway might later become advantageous. In this case, no native pathway having a sophisticated history of amelioration is present to out-compete a pathway of alien origin. In a sense, the recipient has reverted to a pristine evolutionary state that is probably subject to instability and rapid change. (iii) An alien suite of enzymes might provide extraordinary properties of catalysis and/or regulation that are of sufficient selective value to offset a lack of ameliorative history.

Amino acid biosynthetic pathways, such as the Trp pathway (see Table 1 for nomenclature used), are not novel in the sense that they are ancient and widely distributed. When the Trp pathway is absent in prokaryotes, it is a consequence of gene loss (reductive evolution), usually, if not always, in pathogens or symbionts whose hosts or symbiont partners supply the Trp required. Table 2 provides a current listing of microbial genomes that have lost some or all of the trp genes. Such reductive changes can be quite recent, as illustrated by existence of some trp-gene degradation in Yersinia pestis KIM, but not in the other two completely sequenced Y. pestis genomes. Genomes having incomplete trp pathways are presumably in an intermediate state of genome reduction. However, see Xie et al. [10,11] for novel functional innovations of "incomplete" Trp pathways in chlamydiae. For example, the addition of two genes to the partial-pathway operon of Chlamydophila psittaci has been asserted to have created a novel operon responsible for a kynurenine-to-Trp pathway that is important in overcoming a mechanism of host defense during pathogenesis [1]. This hypothesis has recently been confirmed experimentally [12]. Also, see Barona-Gòmez and Hodgson [13] for their resolution of the mystery of why trpC is absent in the clade of actinomycete organisms, that is, they identified priA as a replacement for the otherwise universal trpC. A thorough overview of the phenomenon of "missing genes in metabolic pathways" and how this can be approached via comparative genomics has been provided by Osterman and Overbeek [14].

Trp is biochemically the most expensive of the amino acids synthesized by prokaryotes. Accordingly, it is not surprising that Trp biosynthesis is usually regulated with fine-tuned precision. Different organisms deploy completely different modes of regulation, for example, compare those of E. coli and Bacillus subtilis [15,16]. Pseudomonas aeruginosa exhibits yet a third distinctive system of regulation, part of which involves an activator gene (trpI) [17]. The sophisticated, complex, and highly distinctive regulation in each of the latter three organisms appears to be of recent origin based upon the relatively narrow clades possessing these particular regulatory systems. This is a very important point in support of the thesis [6] that modern, sophisticated genomes may be much less prone to displacement by LGT of trp genes than were ancient genomes. In this context, it will be quite important to determine whether organisms that lack known modes of regulation based upon current model organisms really have relatively unsophisticated control mechanisms, or whether unknown regulatory mechanisms are in place. For example, it has been initially surprising that trp genes of Streptomyces coelicolor are not regulated by feedback repression. However, an excellent and detailed study [18] has demonstrated the existence of regulation that is both growth phase-dependent and growth rate-dependent. This has been attributed to the oligotrophic lifestyle of Streptomyces. It would be interesting to know whether the clade defined by the Streptomyces mode of regulation is also narrow (and therefore of recent origin).

We refer to trp genes that produce Trp in general support of protein synthesis as genes of primary metabolism. Those trp genes responsible for the production of intermediates (or Trp) for any other purpose (e.g. as precursors of antibiotics or pigments) are referred to as genes of specialized (or secondary) metabolism. There is ample precedent for maintenance in a given genome of co-existing structural genes whose gene products catalyze the same reaction, but which function in differing temporal or spatial modes in different pathways. Such genes may be differentially regulated by distinctive control mechanisms or mechanisms that accomplish spatial separation. For example, S. coelicolor possesses trp paralog genes of identical catalytic function that exist in separate operons dedicated to primary biosynthesis, on the one hand, or to calcium-dependent antibiotic (CDA) production, on the other hand [6,19]. It is interesting that, in such cases, one or more of the pathway genes sometimes exist as a single copy and, therefore, must be shared between both functions.

This study is primarily a phylogenetic analysis and depends upon protein trees. However, sequences of proteins such as the Trp enzymes are not nearly as conserved as 16S rRNA, and it is well known that they are of limited value for making phylogenetic inferences over wide phylogenetic distances [20]. On the other hand, for a group of very closely related organisms, the 16S rRNA sequences can be so similar that there is a limited basis for discriminating order of branching. Here some protein trees may yield more refined branching relationships over short phylogenetic distances. Until recently, there have been relatively few sequenced genomes available that would provide a critical mass of closely related organisms as a source of Trp-protein sequences. The sequencing of new genomes is now beginning to provide phylogenetic groups of ever more dense genome representation. Health-related organisms heavily influence priorities, and wide gaps in the currently available tree exist. Phylogenetic regions where genome representation is sparse will undoubtedly persist for many years, and it would be helpful if new genomes for sequencing were selected specifically to fill phylogenetic gaps. One can anticipate that a given protein tree will progressively increase its useful phylogenetic span of discrimination as the gaps between regions of sufficient genome representation are filled.

Additional daunting problems (which are minimal for 16S rRNA trees) assail the phylogenetic validity of protein trees. These are: (i) unrecognized paralogy and (ii) xenology. Ancient paralogs that arose in a common cell will usually have diverged greatly in the contemporary organisms that house them. If one or the other paralog has been lost in various lineages in an erratic fashion and the genealogical history of these losses is not recognized, this situation has been termed "unrecognized paralogy". Differential loss of ancient paralogs in different lineages has the potential to place surviving homolog proteins of distantly related organisms closer to one another than to the surviving proteins of closely related organisms. In this case, xenology can be erroneously inferred. Although LGT has been increasingly recognized for genes in general, genes encoding 16S rRNA are thought to be recalcitrant to LGT (but see Gogarten et al. [21] and Doolittle [22].

Since particular lineages have optimized Trp biosynthesis to overall demands of primary and secondary metabolism that are both qualitatively and quantitatively individualistic, it seems that abandonment of native trp genes in favor of imported alien genes would not occur very often. If correct, then one would expect that the vertical trace of evolutionary descent for trp genes engaged in primary Trp biosynthesis would be demonstrable, as has indeed been shown following a very comprehensive analysis [6]. In the latter study, individual Trp-protein trees were found to be generally congruent with 16S rRNA trees in subtree regions that were supported by sufficiently dense phylogenetic representation.

Against this background of congruency (and, indeed, enabled because of the overall congruency), several clear examples were found of LGT of whole-pathway trp operons in which all of the native trp genes were displaced, occasionally leaving a surviving remnant behind. In addition, examples were found of LGT of partial-pathway trp operons that either displaced native genes of primary biosynthesis (one case) or became associated with a secondary function. In the literature relevant to LGT there is frequently little evidence about likely donor genomes, and even less indications of direction of transfer [23]. This paper provides detailed information and analysis not given in [6], and it provides an especially thorough documentation in that the occasional LGT transposition of the entire suite of seven genes responsible for Trp biosynthesis are evaluated. Importantly, we were able to identify donor lineages, to identify a gene remnant of the pre-existing Trp system in three divergent descendants of one of the recipient genomes, and to identify in one genome several evolutionary steps of specific operon perturbation that resumed in the vertical genealogy following LGT.

Individual Trp-protein domains vary in size and degree of conservation and thus confer varied amounts of phylogenetic information. Figure 2 shows a protein tree in which the seven protein domains responsible for the Trp pathway of biosynthesis have been concatenated in the order of the pathway steps (TrpAa/TrpAb/TrpB/TrpC/TrpD/TrpEa/TrpEb) prior to use of the tree-building program. The seven-domain Trp tree was constructed by using 47 organisms having a complete Trp pathway and by application of an analysis aimed at exclusion of divergent paralogs and xenologs that do not function in primary Trp biosynthesis (excluded paralogs and xenologs can be viewed in the succeeding figures). Each of the seven protein sections of a given concatenate has an impact on the overall tree that is roughly proportional to the number of amino acids encoded. Each of the seven nodes defining TCGs is supported by a bootstrap value of 100% (in contrast to the weakness of individual Trp-protein trees). Although the overall concatenated tree does not parallel the overall 16S rRNA tree, seven subtree regions of the concatenated tree do parallel 16S rRNA subtree regions. These seven groupings are not necessarily equivalent in phylogenetic span. TCG-6, for example, is a small, closely related grouping compared with the much looser TCG-7 assemblage. The emerging, fuller membership of these seven TCGs is hinted at in Table 3, where additional provisional TCG members that were not included in Figure 2 are listed in regular non-bold type. The latter were assigned (mostly with newly available genomes after completion of our primary analysis) by assessment of the best BLAST (Basic Local Alignment Search Tool) hits obtained after entering each Trp domain of the organisms listed as query sequences. Best-match methods are useful for rapid screening, but are subject to limitations [24]. Tentative TCG assignments were also assisted by other features. For example, Brucella melitensis, Agrobacterium tumefaciens, Bradyrhizobium japonicum, Sinorhizobium meliloti and Rhodopseudomonas palustris all possess in common the following split-pathway gene organization: trpAatrpAb (fused genes), a trpB/trpD operon, and a trpC/trpEb/trpEa operon [6].

The few exceptions that violate the otherwise good congruency between TCGs and 16S rRNA subtree regions point strongly to instances of whole-pathway trp-operon displacement. These include the incongruent presence of Trp-protein concatenates from both Helicobacter pylori and the coryneform bacteria in TCG-1. This indicates that inter-genomic transfer of whole-pathway trp operons occurred, with the donor identifiable as a member of TCG-1.

Orange-highlighted dashed lines in Figures 1 and 2 mark lineages where closely related genomes have not yet been sequenced, that is, where there is sparse genome representation. There is no congruency of the branching positions of these ten orphan organisms when the 16S rRNA tree (Figure 1) is compared with the Trp-protein concatenate tree (Figure 2). It is generally believed that the position of these lineages on the 16S rRNA tree is reliable because 16S rRNA comparisons can discriminate very well over wide phylogenetic distances. In contrast, support for the position of branching obtained for these lineages with respect to the concatenated Trp-protein tree is not significant, that is, low bootstrap values. One can reasonably expect that, as more genomes are sequenced, new TCGs will emerge that include these current orphan organisms. For example, our preliminary data indicate that the Trp proteins of Bacteroides thetaiotaomicron will reside within a new TCG group with Cytophaga hutchinsonii.

In the following section, individual protein trees are shown for comparison with the much superior tree of Trp-protein concatenates shown in Figure 2. Each individual tree consists of 47 sequences that comprise segments of each concatenate string used for the tree shown in Figure 2. Included among the concatenates are proteins of LGT origin (from Helicobacter pylori, Corynebacterium glutamicum, and Corynebacterium diptheriae) that function for primary Trp biosynthesis. In addition to the foregoing sequences, Figure 3, Figures 5,6,7, and Figures 9,10,11 also display paralogs and xenologs that were excluded from the concatenate strings because they were not deemed to function for primary biosynthesis. It can be seen that TrpAb (Figure 5) and TrpC (Figure 7), being relatively short and not highly conserved sequences, are the least informative. TCG-2 and TCG-7 are frequently not visualized as entirely cohesive groupings when the individual trees are inspected. In each individual Trp-protein tree, primary-pathway Trp domains are designated by the organism acronym (e.g. Det), whereas unknown xenologs or paralogs are designated with a following number (e.g. Det_2). If functions are known or if names exist in the literature, for example, Sco_CDA or Pae_PhnA, respectively, these designations are used. Remnant proteins are denoted with an 'r', for example, Cgl_r (Figure 9).

Xie et al. [6] can be consulted for a detailed overview of the Trp pathway reactions (summarized here in Table 1) and for a perspective on the nomenclature issues. Table 4 is a comprehensive listing [see Additional File 1] that contains gi (gene identification) numbers (in bold) that correspond to each of the seven domains asserted to function in primary Trp biosynthesis. Paralog and xenolog gi numbers are presented in regular type. Each gi number is hyperlinked to the corresponding record at NCBI (National Center for Biotechnology Information).

Figure 12 depicts the evolutionary events proposed to account for the location in both H. pylori and coryneform bacteria of all seven domains that are associated with primary Trp biosynthesis within TCG-1, which otherwise is populated by members of the enteric lineage. Members of TCG-1 all possess whole-pathway operons with a trpD•trpC fusion. As an isolated observation, we cannot absolutely rule out independent fusions of trpD with trpC in Helicobacter and coryneform bacteria, followed by convergent evolution with respect to the remaining trpD•trpC fusions, to explain their joint locations in TCG-1 for TrpD and TrpC trees. However, the identical overall gene order of the 7-gene operon, the phylogenetic positioning of the remaining five protein domains that are not fused, and the existence of congruent trpD remnants in coryneform bacteria are observations that support the conclusion of LGT with conviction. Although the phylogenetic analysis pinpoints the LGT donor of the whole-pathway trp operon to a member within the enteric lineage, the subclade that includes E. coli, Salmonella typhimurium, Klebsiella pneumoniae, and Shigella flexneri can be excluded. This is because the latter four organisms all possess a recent gene fusion (trpAb•trpB) that is absent elsewhere in the enteric lineage (absent also, of course, in H. pylori and coryneform bacteria).

The trp operon of E. coli (and presumably other members of the enteric clade) is subject to control at several levels [15,16,27,33]. These include (i) allosteric control of anthranilate synthase, whereby Trp acts as a potent feedback inhibitor; and (ii) an attenuator mechanism that is mediated by a Trp-rich leader peptide (encoded by trpL), and repression control mediated by trpR, which binds Trp as a corepressor moiety. Because sensitivity to feedback inhibition is built into the allosteric domain of TrpAa, and because trpL is located immediately upstream of the E. coli trp operon, it is not surprising that recipient organisms such as the coryneform bacteria possess both of the latter regulatory features. However, trpR is distant from the operon and was not co-transferred with the operon.

Perhaps the contemporary trp operon acquired by LGT offered selective advantages to the ancestor of coryneform bacteria, but Trp regulation in coryneform bacteria would appear to be relatively unsophisticated without trpR. In E. coli the impact of attenuation is relatively weak compared with the impact of repression [16,27]. Repression detects free Trp whereas attenuation detects uncharged tRNA^Trp. Since free Trp concentrations can be fairly low and still maintain highly charged tRNA^Trp, trpR-mediated repression acts over a large range of expression. Relief from attenuation ensues after maximal derepression has occurred. In E. coli, trpR also functions beyond the Trp pathway in that it binds to an operator for an initial gene of aromatic biosynthesis.

The existence of a leader peptide associated with regulation of the trp operon (and, indeed, the unexpectedly close relationship with the E. coli trp operon) was reported some time ago in C. glutamicum [34-36]. The leader-peptide region (encoded by trpL) upstream of the trp operons of C. glutamicum, C. efficiens, and C. diptheriae are shown (Figure 13) in comparison with the corresponding region of V. cholerae, a member of the enteric lineage that represents the LGT donor. In V. cholerae, C. glutamicum, and C. efficiens, trpL is immediately upstream of trpAa, the initial structural genes of the trp operon in these organisms. In C. diptheriae a copy of trpEb has recently been inserted between trpL and trpAa (Figure 12). The putative start codons of trpL in Vibrio cholerae, C. glutamicum, and C. diptheriae are ATG, TTG, ATG, and GTG, respectively. These yield leader peptides of 30 amino acids in length. The start codon is uncertain, and Heery and Dunican [35] have suggested a start codon (vertical black arrow) for C. glutamicum that would yield a 17-amino acid leader peptide.

In the case of H. pylori, the trp operon was either acquired without trpL, or trpL was soon lost. Thus, the trp pathway of H. pylori appears to lack regulation by both attenuation and repression. Assuming that the displaced pathway of H. pylori was similar to that of the contemporary C. jejuni or other epsilon-proteobacteria, it would be interesting to compare the relative efficiencies of the modern Trp pathways in these fairly close relatives.

P. aeruginosa possesses a partial-pathway operon consisting of genes encoding the two subunits of tryptophan synthase that are of xenologous origin. These trpEb/trpEa genes are engaged in primary biosynthesis. The donor is clearly a member of the TCG-3 clade. Thus, in the scenario considered [6] for gene separation of the ancestral trpC/trpEb/trpEa operon to yield a stand-alone trpC gene and a trpEb/trpEa operon, the native trpEb/trpEa must have been displaced by LGT, after the intragenomic translocation. Because the regulatory gene, trpI, is adjacent in divergent orientation to trpEb in the P. aeruginosa/P. syringae clade (and three other close relatives; see asterisks in Table 3), it will be interesting if any genomic members of TCG-3 to be sequenced in the future prove to have trpI adjacent to a trpEb/trpEa operon. (However, TrpI is a member of the ubiquitous and large family of LysR transcriptional activators, and could easily have emerged recently in the P. aeruginosa clade following gene duplication.)

P. aeruginosa also possesses a partial-pathway trp operon encoding the two subunits of anthranilate synthase. Phylogenetic analysis indicates that these are trpAa trpAb xenologs that originated from within the enteric lineage. As with the whole-pathway operons just discussed, the E. coli/S. typhimurium/K. pneumoniae/S. flexneri clade can be excluded as the specific donor because this clade possesses a trpAb•trpB fusion. This xenologous two-gene operon of P. aeruginosa (originally denoted phnA and phnB by Crawford et al. [25,37]) has not displaced the corresponding trp genes that are engaged in primary biosynthesis. Although phnA and phnB were originally thought to function in phenazine production, as the naming implies, phenazine pigments are not derived from anthranilate [26]. Instead, the xenolog pair appear to have an unknown specialized function that is geared to stationary-phase physiology [38]. The LGT acquisition of the phnA/phnB xenolog operon by P. aeruginosa occurred after its divergence from close relatives such as P. putida, P. fluorescens, P. syringae and Azotobacter vinelandii because none of these relatives have the xenologous phnA/phnB operon. Hence, this acquisition appears to be quite recent.

Species of Xylella also possess a xenologous two-gene trpAa/trpAb partial-pathway trp operon that appears to have a specialized function that is distinct from homologs engaged in primary Trp biosynthesis. Xie et al. [7] have suggested that the gene denoted acl probably encodes an aryl-CoA ligase. Since acl appears to exist within the Xylella trpAa/trpAb operon, its gene product may have the specificity of anthranilate-CoA ligase. Activated anthranilate may then function as a key precursor for production of antibiotic, siderophore, and so on. Note that, if this is correct, the primary trp-pathway genes trpB/trpC/trpD/trpEa/trpEb are irrelevant to the specialized pathway. Thus, reference to the xenologous trpAa/trpAb pair as a "partial-pathway operon" could be somewhat a misnomer, and perhaps "hybrid operon" might prove to be more apt. The origin of the Xylella trpAa/trpAb genes might have been from a close relative of C. jejuni because the best matches of both genes are with C. jejuni. It is also suggestive that these genes have a low-GC ratio (about 38%), compared with a genomic GC ratio of 30.2% for C. jejuni. C. jejuni cannot have been the direct donor, however, because it possesses a trpAb•trpB fusion.

Refer to the Gold Genomes Online Database [49] for a list of published complete genomes and links to the corresponding references. Synonymous names include Rhizobium meliloti = Sinorhizobium meliloti; Corynebacterium glutamicum = Brevibacterium lactofermentum = Brevibacterium flavum = Brevibacterium divaricatum; Rhizobium meliloti = Sinorhizobium meliloti; Rhizobium loti = Mesorhizobium loti; Thermomonospora fusca = Thermobifida fusca; Chlamydophila psittaci = Chlamydophila caviae; Thiobacillus ferrooxidans = Acidithiobacillus ferrooxidans; Shewanella putrefaciens = Shewanella oneidensis; Sphingomonas aromaticivorans = Novosphingobium aromativorans.

16S rRNA subtrees were derived from the Ribosomal Database site [50,51].

Unrooted phylogenetic trees were derived from ClustalW alignment [52,53]. The neighbor-joining and Fitch [54] programs were employed to obtain distance-based trees. The distance matrix was obtained using Protdist with a Dayhoff Pam matrix. The Seqboot and Consense programs were then used to assess the statistical strength of the tree using bootstrap resampling. Neighbor-joining and Fitch trees yielded similar clusters and arrangement of taxa within them. Bootstrap values indicate the number of times a node was supported in 100 resampling replications.

Multiple sequence alignments were derived by input of the indicated homolog amino acid sequences into the ClustalW program (Version 1.4) [52,53]. Manual alignment adjustments were made as needed with the assistance of the BioEdit multiple alignment tool of Hall [55].

After each of the individual Trp-protein domain alignments were generated, both N-terminal and C-terminal unaligned regions were trimmed manually through visual inspection. Then the seven protein domains responsible for the Trp pathway of biosynthesis were concatenated in the order: TrpAa/TrpAb/TrpB/TrpC/TrpD/TrpEa/TrpEb. The resulting concatenated multiple alignment was used as input for generation of a phylogenetic tree using the program package PHYLIP [56].

Raw DNA contig sequences available from NCBI [57] and the TIGR Unfinished Microbial Genomes database [58] were screened using the built-in BLAST service. The protein sequences from GenBank were used as query entries. The BLAST 2.0 and the ORF Finder (Open Reading Frame Finder) offered by NCBI [59] were used to locate open reading frames and to confirm the similarity search result of the raw sequence.

Fusion protein sequences from GenBank and NCBI Microbial Genomes Blast Databases [57] were screened using the BLAST [60] program. Multiple alignments were obtained by input of single-domain and fusion-protein sequences into the ClustalW [52,53] program (version 1.4).