Phylogenomics and protein signatures elucidating the evolutionary relationships among the Gammaproteobacteria

Phylogenetic analysis of gammaproteobacteria

The availability of genomic sequences now makes it possible to examine evolutionary relationships based on concatenated sequences for large numbers of proteins. This approach is more reliable than analysis based on any single gene or protein (Rokas et al., 2003; Brown & Volker, 2004; Belda et al., 2005; Ciccarelli et al., 2006). We have performed phylogenetic analyses for gammaproteobacteria based on the combined sequences for 36 conserved proteins from 45 gammaproteobacterial species covering 13 of its 14 orders (all except for the order Acidithiobacillales). The ML phylogenetic tree for the gammaproteobacterial species based on this large dataset is shown in Fig. 1⇓. The proportion of puzzled quartets (ML analysis), or percentage bootstrap scores in MP analysis, which supported different nodes (only values >50 % are shown) are indicated. A NJ tree for this dataset is provided as Supplementary Fig. S2 (see IJSEM Online). The species from a number of orders of the class Gammaproteobacteria (e.g. the orders Enterobacteriales, Pasteurellales, Vibrionales, Aeromonadales, Legionellales and Xanthomonadales) formed distinct clades with good statistical support (i.e. relationships supported by >70 % bootstrap samples or puzzling quartets). Based on these trees, a clade consisting of the species of the order Enterobacteriales was found to be the most recently diverging lineage within the class Gammaproteobacteria. The late divergence of the order Enterobacteriales has also been observed in earlier studies (Olsen et al., 1994; Lerat et al., 2003; Brown & Volker, 2004; Ludwig & Klenk, 2005; Belda et al., 2005; Mrazek et al., 2006). Within this clade, various species corresponding to endosymbiotic bacteria (such as Buchnera aphidicola, Wiggelsworthia glossinidia and ‘Candidatus Baumannia cicadellinicola’) formed a distinct deeper-branching cluster. It has been previously shown by Belda et al. (2005) that the deep branching of these bacteria is most probably due to their faster rate of evolution in comparison with the free-living enteric bacteria.

The phylogenetic trees also strongly supported a close relationship of the order Enterobacteriales to the order Pasteurellales. The combined clade of these two orders was linked at a higher level to the clades consisting of species of the orders Vibrionales and Aeromonadales. Although a clade consisting of these four orders was supported by different phylogenetic methods, the relative branching of the orders Vibrionales or Aeromonadales with respect to the order Enterobacteriales–Pasteurellales clade was not resolved. In the NJ tree (see Supplementary Fig. S2 in IJSEM Online), but not in the ML or MP trees (Fig. 1⇑), these two orders were found to group together with strong bootstrap support. The phylogenetic trees also strongly indicated that the species of the order Alteromonadales formed an immediate outgroup of the above four orders. One additional clade that was reliably observed consisted of species from the above five orders as well as various species belonging to the orders Oceanospirillales and Pseudomonadales. It is noteworthy that species from the orders Oceanospirillales and Pseudomonadales did not form well-defined clades in the trees, indicating that these orders are phylogenetically heterogeneous. In comparison with these orders, the species from other gammaproteobacterial orders (such as the orders Thiotrichales, Cardiobacteriales, Legionellales, Chromatiales, Methylococcales and Xanthomonadales), consistently showed deeper branching in the trees and their relative branching positions were not resolved.

Phylogenetic trees were also constructed for many individual proteins (particularly those in our set with length >400 aa) and also on a smaller dataset of concatenated sequences for 13 large proteins from this set (see Methods). The relationships observed with this smaller dataset of concatenated protein sequences were identical to those shown here with the larger dataset and the results for the ML/MP tree for this dataset are provided as a supplementary figure (see Supplementary Fig. S3 in IJSEM Online). This smaller dataset of protein sequences was rooted using ‘Aquifex aeolicus’ and this rooting did not affect the branching pattern or interrelationships among different gammaproteobacterial orders. The phylogenetic trees for most individual proteins (RpoB, SecA, DnaK, gyraseA, IleRS, SecY, PheRS, RpoA, ArgRS) supported similar relationships as seen here (Fig. 1⇑ and Supplementary Fig. S2), but due to smaller number of positions in these alignments, the bootstrap scores for many nodes were low and not resolved (results not shown). However, in the phylogenetic trees for some proteins (for example, UvrD helicase, GTP binding protein, EF-G and O-sialylglycoprotein endopeptidase), the endosymbiotic bacteria (such as Buchnera aphidicola, W. glossinidia and ‘Ca. Baumannia cicadellinicola’) did not group with other members of the order Enterobacteriales and instead they branched deeply in the tree (results not shown). In all of these cases, the branches for these species were very long, which can lead to artefactual deeper branching in the trees (Felsenstein, 1978; Gribaldo & Philippe, 2002; Belda et al., 2005).

Conserved indels that are specific for gammaproteobacteria and their subgroups

Rare genomic changes (RGCs) such as conserved inserts and deletions in genes/proteins that are restricted to species from well-defined taxonomic groups provide a powerful means for inferring as well as confirming evolutionary relationships (Rivera & Lake, 1992; Gupta, 1998; Rokas & Holland, 2000). In many cases, these RGCs have been instrumental in elucidating relationships that were not resolved by phylogenetic trees (Rivera & Lake, 1992; Baldauf & Palmer, 1993; Gupta, 1998; Rokas & Holland, 2000; Kunisawa, 2001). We have identified a number of conserved indels in important housekeeping proteins that are helpful in clarifying the evolutionary relationships among gammaproteobacteria. We previously described two conserved indels in the proteins AICAR-transformylase (PurH) and ribosomal protein L16, which appeared to be restricted to gammaproteobacteria (Gupta, 2000). However, sequence information for these proteins at that time was available for a limited number of gammaproteobacterial species belonging to only certain orders. Hence, it was of importance to re-examine the species distribution of these indels.

Fig. 2⇓ shows the partial sequence alignment of the PurH protein showing the 2 aa deletion that is common to various gammaproteobacteria. As can be seen, this 2 aa deletion, located in a conserved region, is uniquely shared by different gammaproteobacteria, but it is not found in other classes of the phylum Proteobacteria or other bacterial phyla. The only gammaproteobacterium in which this indel is absent is Francisella tularensis, which corresponds to one of the deepest branches in the phylogenetic tree (Fig. 1⇑ and Supplementary Fig. S2). Although, F. tularensis is currently in the order Thiotrichales within the class Gammaproteobacteria, in phylogenetic trees where members of the class Betaproteobacteria are also included, this species forms an outgroup from all of the gamma- and betaproteobacterial species (results not shown). These results indicate that the placement of this species within the class Gammaproteobacteria is probably incorrect and that the absence of the PurH indel in this species may not constitute an exception. However, based upon these results, other possibilities (e.g. this indel occurred after the branching of F. tularensis or the purH gene was acquired by this species by LGT) cannot be excluded. Nevertheless, the shared presence of this deletion in all gammaproteobacteria except F. tularensis (sequence information for >200 gammaproteobacteria is currently available) and its absence in all other bacteria, indicates that the RGC responsible for this deletion probably occurred in a common ancestor of all or most gammaproteobacteria. This RGC thus provides a good molecular marker for this large and important class of proteobacteria. It is interesting to note that besides the class Gammaproteobacteria, a 2 aa deletion in this position is also present in three archaeal species belonging to the order Methanomicrobiales. The sequences for two of the members of the order Methanomicrobiales are shown in the sequence alignment in Fig. 2⇓. In a phylogenetic tree for the PurH sequences, the class Gammaproteobacteria and order Methanomicrobiales do not group together (results not shown), indicating that the shared absence of this indel in these two groups is not due to LGT, but is very probably due to independent genetic events.

The partial sequence alignment for the ribosomal protein L16 is presented in Supplementary Fig. S4 (see IJSEM Online). Unlike the indel in PAC formyltransferase, the 1 aa deletion in this protein is specifically present in various species from the orders Enterobacteriales, Pasteurellales and Vibrionales, and also several species of the order Alteromonadales, but it is not found in other members of the class Gammaproteobacteria or other bacterial phyla. This indel supports a close relationship between the species belonging to these orders. The presence of this indel in some species of the order Alteromonadales but not others suggests that the species from this order are not phylogenetically homogeneous, a feature also observed in our phylogenetic analysis. In the ML/MP tree shown in Fig. 1⇑, the clade corresponding to the order Alteromonadales is weakly supported only by ML analysis and it is not supported by MP analysis. In the NJ tree (see Supplementary Fig. S2), these species do not group together, with Idiomarina loihiensis branching deeper than other species of the order Alteromonadales.

Two other novel conserved indels that are specific for certain orders or subgroups of gammaproteobacteria were identified in the present study. In the β-subunit of RNA polymerase (RpoB), a 4 aa deletion was uniquely present in various species from the orders Enterobacteriales, Pasteurellales, Vibrionales, Aeromonadales and Alteromonadales (Fig. 3⇓), but it was not found in any other gammaproteobacteria or other groups of bacteria. The genetic change responsible for this indel most probably occurred in a common ancestor of these particular orders after the divergence of other gammaproteobacteria at a stage marked by the filled circle in Fig. 1⇑. Interestingly, this deletion in RpoB was not present in species of the genus Marinobacter, which are indicated to belong to the order Alteromonadales. However, in the phylogenetic trees shown in Fig. 1⇑ and Supplementary Fig. S2, Marinobacter aquaeolei did not group with other species of the order Alteromonadales, but branched outside of the clade comprising these Alteromonadales species as well as various species from the orders Enterobacteriales, Pasteurellales, Vibrionales and Aeromonadales. Both these observations indicate that the genus Marinobacter is a deeper branching genus when compared with other genera of the order Alteromonadales. Another useful indel for the gammaproteobacteria is present in the protein leucyl-tRNA synthetase (Fig. 4⇓). In this case, a 2 aa deletion is present in various species belonging to the orders Enterobacteriales, Pasteurellales, Vibrionales, Aeromonadales, Alteromonadales and Oceanospirillales, but is not found in other gammaproteobacterial orders or in other bacteria. This indel suggests that the species from the order Oceanospirillales are more closely related to the above orders in comparison with the order Pseudomonadales.

It should be acknowledged that in the present study we have not carried out a comprehensive sequence alignment of all gammaproteobacterial proteins to identify different conserved indels that might be specific for this class of bacteria or for its different subgroups. Hence, it is likely that in future many other conserved indels that are specific for different gammaproteobacterial subgroups will be identified, providing additional molecular markers and further insights into the evolution of these bacteria.

Comparative genomic studies to identify proteins that are specific for gammaproteobacteria

We have also performed systematic blastp searches on various ORFs in the E. coli K-12 genome to identify proteins that are unique to the gammaproteobacteria at a higher taxonomic level. This genome was chosen as the query because E. coli belongs to the order Enterobacteriales, which, based upon our phylogenetic analysis (Fig. 1⇑ and Supplementary Figs S2 and S3), is the most recently diverged group/order within the class Gammaproteobacteria. Hence, by using probes from this genome, which lies at the ‘tip’ of the phylogenetic tree, it should be possible to identify proteins that are specific for gammaproteobacteria at different phylogenetic depths. The genome of E. coli is also well annotated and extensive functional and gene mutation studies have been conducted on this organism (Blattner et al., 1997; Gerdes et al., 2003; Kang et al., 2004; Chen et al., 2006). The objective of our comparative genomic studies in this work was to identify proteins that were distinctive characteristics for either most species of the order Enterobacteriales or were uniquely present in this order as well as other orders of the class Gammaproteobacteria (see Methods). Because our query sequences were from E. coli K-12, these studies will not have detected certain proteins that might be present in other gammaproteobacteria, but absent in E. coli K-12. Likewise, these studies will also not have detected proteins that are specific for other orders of the class Gammaproteobacteria, but which are not found in E. coli K-12.

Our analyses identified 75 gammaproteobacteria-specific proteins that met these criteria and a brief account of their species distribution and other relevant information is provided. The first five proteins in Table 1(a)⇓ are largely specific for the order Enterobacteriales. Except for one or two hits mainly from other gammaproteobacteria, all other hits for these proteins are for species of the order Enterobacteriales. The next three proteins in this Table are mainly found in various sequenced species of the orders Enterobacteriales and Pasteurellales. Of these, all significant blast hits for proteins b2343 and b3793 are from these two orders, whereas for protein b4481, two hits are also seen for species of the order Oceanospirillales. Of these three proteins, b3793, which is annotated as putative ECA polymerase, is essential for E. coli cells (Gerdes et al., 2003).

Table 1(b)⇑ lists 24 proteins that are mainly restricted to species from the orders Enterobacteriales, Pasteurellales, Vibrionales and Aeromonadales. Two of these proteins (b0919 and b4311) are only found in various members of the orders Enterobacteriales and Vibrionales, whereas protein b3790 is present only in the orders Enterobacteriales, Pasteurellales and Vibrionales. Some of the proteins listed in Table 1(b)⇑ are missing in the order Aeromonadales (for example, b0956, b1811, b2510 and b4372) or from both the orders Aeromonadales and Pasteurellales. The absence of some of these proteins in species of the order Pasteurellales, which are obligate parasites, is probably due to gene loss. The species distribution profile of these proteins suggests that species from the orders Enterobacteriales and Pasteurellales are more closely related to the orders Vibrionales and Aeromonadales when compared with members of the order Alteromonadales or other orders of gammaproteobacteria; this is supported by phylogenetic studies (Fig. 1⇑ and Supplementary Figs S2 and S3). Of the proteins listed in Table 1(b)⇑, three proteins, b0922 (MukF), b0923 (MukE) and b0924 (MukB), which are encoded by neighbouring genes, form a complex, MukBEF, which is involved in chromosome partition and DNA repair (Gloyd et al., 2007). Another protein SeqA (b0687), which shows similar species distribution to these three proteins [except that it is also present in some members of the order Alteromonadales: Table 1(c)⇑], also interacts with the MukBEF complex in the cell division process (Yamazoe et al., 2005). All four of these proteins, as well as two other proteins, b0467 and b4151, which are annotated as primosomal replication protein N and the fumarate reductase subunit D, respectively, are essential for the growth of E. coli cells (Gerdes et al., 2003). The species distribution profiles of the Muk and SeqA proteins indicate that this novel mechanism for chromosome partition, which is limited to only certain orders of gammaproteobacteria, evolved very late in evolution.

Table 1(c)⇑ lists 20 proteins that we consider to be mainly specific for members of the orders Enterobacteriales, Pasteurellales, Vibrionales, Aeromonadales and Alteromonadales. Several of these proteins, such as b0119, b0687, b0964, b2187, b2295, b2900 and b3938, are entirely specific for these orders, whereas a number of others are also found in one or two species from other groups. The absence of some of these proteins in the order Pasteurellales is again presumably due to gene loss. The species distribution profiles of these proteins suggest that species from these orders shared a common ancestor exclusive of other gammaproteobacteria. This conclusion is strongly supported by phylogenetic analyses (Fig. 1⇑ and Supplementary Figs S2 and S3) and by the conserved indel in the RpoB protein (Fig. 3⇑). Five of these proteins (b0163, b0466, b3764, b3999 and b4255) are also present in some of the species from the order Oceanospirillales, indicating that species from this order show a close relationship to these other orders, a conclusion also supported by the signature indel in the LeuRS protein (Fig. 4⇑). Of the proteins in Table 1(c)⇑, in addition to SeqA (b0687), three further proteins, b1610, b2900 and b3466, are essential for the survival of E. coli cells (Gerdes et al., 2003). Of these proteins, b1610 (Tus) is annotated as a DNA replication terminus site binding protein, whereas the functions of the other two are unknown.

Seven additional proteins in Table 1(d)⇑ (b0411, b0953, b2792, b2944, b3995, b4550 and b4551) are commonly shared by most of the species from the orders Enterobacteriales, Pasteurellales, Vibrionales, Aeromonadales, Alteromonadales, Oceanospirillales and Pseudomonadales. The presence of these proteins suggests that species from these orders shared a common ancestor exclusive of other gammaproteobacteria, which is in accordance with our phylogenetic analyses (Fig. 1⇑ and Supplementary Figs S2 and S3). It is of interest to note that for a number of proteins in Table 1(c, d)⇑ and Table 2⇓, significant blast hits are also observed for a single alphaproteobacterial strain, sp. HTCC 2255. This species also contains a 2 aa deletion in the PurH protein, which is specific for the class Gammaproteobacteria, thus making a strong case for its grouping with the class Gammaproteobacteria rather than the class Alphaproteobacteria. Twelve additional proteins listed in Table 1(e)⇑ are also specific for gammaproteobacteria, but they are present sporadically in species from a number of different orders. The species distributions of these proteins can be accounted for by their evolution at various stages in the divergence of gammaproteobacteria (as noted above for other proteins) followed by gene losses in specific species or lineages.

Lastly, we describe four proteins (b0354, b1132, b1179 and b3033) that are present in most of the gammaproteobacteria, but which are not found in any other bacteria (see Table 2⇑). Except for a few orders that contain either intracellular or parasitic bacteria, these proteins are present in the majority of the sequenced genomes from other orders of gammaproteobacteria. These proteins are also not found in different strains of F. tularensis, again supporting our contention that the grouping of this species with gammaproteobacteria is incorrect. Of all the gammaproteobacteria-specific proteins identified in our analyses, these four proteins show the broadest species distribution and we suggest that their genes first evolved in a common ancestor of all of the gammaproteobacteria, followed by gene losses in certain groups, where their cellular functions were not required. Of these four proteins, b1132 and b3033 are essential for the survival of E. coli cells (Gerdes et al., 2003). Although some of these proteins have been assigned to specific COG groups (Tatusov et al., 2000), their cellular functions are not known at present.

Main inferences from phylogenomic and comparative genomic analyses

The results of our analyses indicate that the main orders within the class Gammaproteobacteria have branched or diverged in the following order (from earliest to most recent): Thiotrichales, Cardiobacteriales, Xanthomonadales, Chromatiales, Legionellales, Methylococcales >Pseudomonadales, Oceanospirillales >Alteromonadales >Aeromonadales, Vibrionales >Pasteurellales >Enterobacteriales. While the positions of the late branching orders are clearly resolved, the relationships amongst the early branching groups remain unclear. This branching order is supported not only by phylogenetic trees based on a large number of proteins, but it is also independently supported by the identification of a number of conserved indels and many proteins for which the RGCs or genes were introduced after some of the major branch points in this scheme.

The gammaproteobacteria have previously been characterized solely on the basis of their branching pattern in trees based on 16S rRNA gene sequences. However, our results show that the 2 aa deletion in the PurH protein is a distinctive characteristic of all gammaproteobacteria (>240 entries currently in the database), with the sole exception of F. tularensis, whose grouping with other gammaproteobacteria is questionable. The indel in the PurH protein thus provides the first known molecular marker that can be used to define and circumscribe the class Gammaproteobacteria. We have also identified four proteins (b0354, b1132, b1179 and b3033) that are uniquely found in most gammaproteobacterial species; the main exceptions being in the endosymbiotic or parasitic bacteria. Although these proteins are not present in all gammaproteobacteria, they also provide novel and useful molecular markers for this large and diverse group.

Of the 75 proteins that are specific for either the order Enterobacteriales or higher clades within the class Gammaproteobacteria, most are of unknown functions with a few exceptions. A number of these proteins are essential for the growth of E. coli cells (see Tables 1⇑ and 2⇑) (Gerdes et al., 2003). The remainder of these proteins, although they are not required for growth under laboratory conditions, are also expected to be important for these bacteria in their natural environments based on their high degree of conservation and persistence (Fang et al., 2005). It is thus of great importance to understand the cellular functions of these unique and broadly distributed proteins. In addition to providing significant insights in to possible novel biochemical or physiological characteristics that are common to many or all gammaproteobacteria, they may also provide potential drug targets for a large group of disease-causing bacteria which belong to this class.