Introduction

Differentiation and classification of extant bacterial taxa have accelerated over the last two decades with the increased use of molecular phylogenetic data (Charlebois et al. 2003; Gould 1996; Martinez et al. 2004). DNA sequencing has transformed the classification of prokaryotes from broad groupings based on resemblances in morphological, biochemical, and physiological characteristics to detailed phylogenies resulting from sequences unique to bacterial species and even strains (Olsen and Woese 1993). DNA sequences from the 16S rDNA gene have been used to determine numerous prokaryotic phylogenies at all taxonomic levels (Christensen et al. 2004; Davidov and Jurkevitch 2004; Dorsch et al. 1992; Woese 1987). This universal applicability of 16S rDNA sequences has enabled microbiologists to confirm broad classifications, such as the γ-bacteria at the phylum level, as well as relationships of species that reside within the same family or genus. The wide application of 16S rDNA sequences stems from the highly conserved and variable components that compose the ribosomal subunits (Olsen and Woese 1993), making 16S rDNA one of the most widely exploited genes for use in microbial phylogenetic studies.

It is notable that 16S rDNA gene sequences have been useful in reconstructing the phylogenies of closely related species and strains in the family Enterobacteriaceae (Wertz et al. 2003). The most familiar species of the family are human pathogens and include pathogenic Escherichia coli, Shigella flexneri, and Yersinia enterocolitica, all of which can cause severe gastrointestinal illnesses (Toth et al. 2006). Other species, such as Erwinia amylovora, the causative agent of fire blight in rosaceous plant hosts, are agriculturally significant (Krieg and Holt 1985; van der Zwet and Beer 1995). In a family as diverse as the Enterobacteriaceae, deciphering the evolutionary relationships among animal and plant pathogen members is crucial, especially since they both employ common and conserved virulence mechanisms to elicit infection on their specific host cells.

Historically, classification studies have focused solely on either animal (Moran et al. 2005; Wertz et al. 2003) or plant (Fessehaie et al. 2002; Kwon et al. 1997) pathogen representatives of the family. Hauben et al. (1998), however, completed a comprehensive phylogenetic analysis on a large number of enterobacterial animal and plant pathogen species based on 16S rDNA gene sequences. As a result of this study, the taxonomy of the plant pathogen genera was emended. A new generic scheme was proposed that partitioned the phytopathogenic species into four distinct clusters: Erwinia, Brenneria, Pectobacterium, and Pantoea (Hauben et al. 1998), with Brenneria being basal to both animal and plant pathogen clusters.

Studies using 16S rDNA to determine microbial phylogenies (Christensen et al. 2004; Davidov and Jurkevitch 2004; Dorsch et al. 1992; Hauben et al. 1998; Woese 1987) have positioned this gene as an ideal candidate for reconstructing robust phylogenies of Erwinia, Brenneria, Pectobacterium, and Pantoea. Such phylogenies, in association with other housekeeping genes, could serve as a reference point for identifying instances of horizontal gene transfer among virulence genes and other auxiliary sequences in the genomes of these species. In this paper we evaluate the robustness and reliability of 16S rDNA gene sequences in phylogeny reconstruction of plant pathogen enterobacterial species by assessing how the resulting phylogenies are affected by the use of different plant pathogen accessions and varying numbers of animal pathogen representatives. To investigate these effects, we compiled and explored four data sets, two of which are from previously published phylogenies (Hauben et al. 1998; Kwon et al. 1997). Each data set focused on the enteric plant pathogens Erwinia, Brenneria, Pectobacterium, and Pantoea, with the first two comprised of the same 27 plant pathogen sequences. The last two data sets consisted of sequences published by Kwon et al. (1997) and Hauben et al. (1998), which were downloaded from GenBank. All four data sets were analyzed using the same methods in order to eliminate disparity resulting from the use of different phylogenetic analysis methods. The animal pathogen and outgroup strains used are the same as those of Kwon et al. (1997) in the first data set and a subset of Hauben et al. (1998) in the second. The resulting trees were inconsistent in the taxonomic placement of both plant and animal pathogen representatives in the four data sets. Additionally, they all displayed poor maximum likelihood (ML) bootstrap and Bayesian support, indicating that 16S rDNA, while effective for both species-level and family-level phylogenetic reconstruction, may underperform for genus-level phylogenetic analyses.

Materials and Methods

Strains and Sampling

Data set A is comprised of 33 samples (Table 1) including 27 plant pathogen isolates, 5 representative animal pathogen species, and 1 outgroup taxon (Kwon et al. 1997). Data set B includes 49 samples (Table 2), consisting of the same 27 plant pathogen isolates as in data set A, 18 representative animal pathogen species, and 4 outgroup taxa (Hauben et al. 1998). Data set C consists of the 22 16S rDNA gene sequences previously published by Kwon et al. (1997) (Table 3), and data set D consists of the 67 sequences published by Hauben et al (1998) (Table 4). The four data sets were used to characterize relationships within Erwinia, Brenneria, Pectobacterium, and Pantoea and to evaluate the placement of the four phytopathogenic genera within the Enterobacteriaceae.

Table 1 Bacterial strains and accessions for published and novel partial 16S rDNA gene sequences comprising data set A
Full size table
Table 2 Additional bacterial strain accessions for partial published 16S rDNA gene sequences used in conjunction with Table 1 to compose data set B
Full size table
Table 3 Accessions for partial 16S rDNA gene sequences published by Kwon et al. (1997) making up data set C
Full size table
Table 4 Accessions for partial 16S rDNA gene sequences published by Hauben et al. (1998) making up data set D
Full size table

Strains of Erwinia, Brenneria, Pectobacterium, and Pantoea listed in data sets A and B were obtained from the laboratories of Drs. Eric Brown and Amy Charkowski as frozen cultures and LB slants, respectively. Strain authentication was confirmed based on 16S rDNA sequence prior to their use.

Bacterial DNA Extraction and PCR Amplification

Total genomic DNA was extracted from the plant pathogen species listed in Table 1 using a silica-based matrix (Instagene; Bio-Rad Inc.) as described in the manufacturer’s directions. Briefly, bacterial cells from 24-h cultures grown on LB agar at room temperature were resuspended in Instagene matrix and heated at 56°C for 30 min. The preps were then vortexed vigorously and boiled for 10 min. Suspensions were then pelleted and the supernatant containing the DNA was decanted into a clean microtube. The optimum amount of eluate to be used in amplification reactions was determined to be 2 μl, after an optimization PCR was carried out using 20, 10, 5, 2, and 1 μl of DNA. Oligonucleotide primers were designed for amplification of the 16S rDNA gene from conserved sequences flanking variable regions of the genes, using Erwinia, Brenneria, and Pectobacterium 16S rDNA sequences from GenBank. The primers used were 16SF1 (5′-GCA GGC CTA ACA CAT GCA AGT CG-3′) and 16SR1 (5′-GCA ACC CAC TCC CAT GGT GTG ACG-3′). Amplification reactions were carried out with 2 μl of DNA template, 2.5 μl of 10× PCR buffer (Sigma), 1.5 μl of 25 mM MgCl2, 2 μl of 2.5 mM dNTPs, 2.5 μl of each 10 μM primer stock, 0.1 μl of 5 U/μl Taq DNA polymerase (Sigma), and purified H2O for a total reaction volume of 25 μl. PCR (Ehrlich et al. 1991) was performed in a PTC-200 thermal cycler (MJ Research) under the following conditions: initial denaturation at 94°C for 2 min, 94°C for 50 s, 58°C for 50 s, and 72°C for 45 s (35 cycles), ending with incubation at 72°C for 8 min.

Cloning and DNA Sequencing

PCR products were purified using QIAquick gel extraction kits (QIAGEN) according to the manufacturer’s protocol. Briefly, completed PCR products were analyzed on 1% agarose gel and visualized with ethidium bromide and UV light. Positive PCR products were gel excised and incubated in buffer QG at 50°C for 10 min. DNA was bound in QIAquick spin columns by microcentrifugation, and purified DNA was eluted in 20 μl of buffer. Cloning reactions were completed using pGem-T Easy cloning kits (Promega) according to the manufacturer’s protocol except that the ligation and transformation reaction volumes were cut in half. The products were ligated to pGem-T vectors with T4 DNA ligase at 4°C overnight. Transformation proceeded via conventional heat shock methods (Manniatis et al. 1982) into competent E. coli (JM109) cloning cells. Cloned inserts were amplified directly from colonies using the original PCR primers and amplification conditions. PCR reactions were cleaned using exonuclease I and shrimp alkaline phosphatase by heating to 37°C for 15 min and then to 80°C for an additional 15 min (Mason-Gamer 2004). Purified PCR products were sequenced bidirectionaly using ABI Big Dye Terminator v3.1 and v1.1 Cycle Sequence kits (Applied Biosystems) according to the manufacturer’s directions, except that final reaction volumes were halved (10 μl) and the amount of BigDye was quartered (2 μl); reactions were run on an ABI 377 DNA sequencer according to the manufacturer’s directions. DNA sequences were edited and combined in Sequencer v.4.1 (Gene Codes Corp.). Sequences have been deposited in GenBank under accession numbers EU490590 to EU490613.

Sequence Alignment and Phylogenetic Analyses

DNA sequences for all four data sets were aligned using CLUSTAL W v1.5 (Thompson et al. 1992), followed by manual adjustments in MacClade v.4.0 (Maddison and Maddison 2000). Regions totaling approximately 100 bp were excluded from the beginning and end of each data set to avoid gaps due to missing data. Phylogenetic analyses for all data sets were performed using PAUP* (Swofford 2001) and GARLI (Zwickl 2006). The aligned sequences were first analyzed in PAUP* using maximum parsimony (MP) under heuristic search methods using tree bisection-reconnection (TBR) branch-swapping and 100 random taxon addition replicates. The shortest trees were then used as the starting topology for the evaluation of 16 nested models of sequence evolution (Frati et al. 1997; Sullivan et al. 1997; Swofford et al. 1996). Parameter space was searched for the best tree with simultaneous estimation for model parameters using a ML search conducted in GARLI. As recommended by Zwickl (2006), multiple runs were performed (∼20) to ensure that results are consistent as the algorithm is stochastic. The log likelihood values of each run were retained in order to compare the individual runs.

Branch support was determined by 100 ML bootstrap iterations in GARLI (Felsenstein 1985) and with Bayesian posterior probability (MrBayes v.3.1.2) approximation of 1 million generations discarding 25% (2500) of the tree samples as recommended in the MrBayes v.3.1.2 manual (Huelsenbeck and Ronquist 2001). For Bayesian character support methods, parameters of sequence evolution estimated from the final ML tree were used (Table 5). Only support values with a bootstrap score of 50 or better and a Bayesian probability of 0.95 or higher appear in Figs. 1–3.

Table 5 Maximum likelihood model parameters used in the phylogenetic analyses
Full size table
Fig. 1
figure 1

Maximum likelihood (ML) analysis of 16S rDNA sequences from plant and animal pathogen enterobacterial species. (a) Likelihood tree comprising 16S rDNA gene sequences from 27 plant pathogen and 5 animal pathogen enterobacterial strains. The log likelihood range for this analysis was −lnL: 5086.66695 to 5086.66606. (b) Likelihood tree comprising 16S rDNA gene sequences from 27 plant pathogen and 18 animal pathogen enterobacterial representatives. The log likelihood range for this analysis was −lnL: 5887.44756 to 5887.48735. ML solutions were generated in GARLI (reference) and visualized in PAUP* (reference). ML bootstrap nodal support was generated in PAUP* and subsequent values are reported for each node in parentheses. Bayesian nodal support values are presented for all nodes without parentheses

Full size image
Fig. 2
figure 2

Phylogenetic analysis of the Kwon et al. (1996) 16S rDNA enterobacterial sequence data set. The resultant maximum likelihood (ML) tree for this data set was produced in GARLI and viewed in PAUP*. The log likelihood range for this analysis was −lnL: 10118.59778 to 10118.73990. ML bootstrap nodal support was generated in PAUP* and subsequent values are reported for each node in parentheses. Bayesian nodal support values are presented for all nodes without parentheses

Full size image
Fig. 3
figure 3

Phylogenetic analysis of the Hauben et al. (1997) 16S rDNA enterobacterial sequence data set. The resultant maximum likelihood (ML) tree for this data set was produced in GARLI and viewed in PAUP*. The log likelihood range for this analysis was −lnL: 5149.07892 to 5149.07908. ML bootstrap nodal support was generated in PAUP* and subsequent values are reported for each node in parentheses. Bayesian nodal support values are presented for all nodes without parentheses

Full size image

Results

The 16S rDNA alignments of data sets A and B consisted of 33 and 49 taxa, respectively, and 1282 characters. The alignment of data set C (Kwon et al. 1997) consisted of 22 taxa and 1456 characters, while the alignment of data set D (Hauben et al. 1998) consisted of 67 taxa and 1472 characters. For all four alignments, longer sequences were truncated at the beginning and end in order to eliminate gaps from the termini of the data sets. The range of variation observed in log likelihood scores between runs was very small, and the tree and model parameters corresponding to the best score were used. The best ML trees are presented in Figs. 1–3.

In Fig. 1a, Pantoea forms a monophyletic clade, whereas Erwinia, Brenneria, and Pectobacterium are all polyphyletic. The Pantoea clade is derived from within a group of erwinias, with Erwinia lupinicola strain 1 as the basal taxon of the Erwinia–Pantoea clade. Erwinia rhapontici 2, Pectobacterium chrysanthemi 2, and Pectobacterium carotovorum ssp. carotovora 2 form a clade that is sister to the Erwinia–Pantoea clade. Finally, the two E. psidii strains and E. rhapontici 1 group with the Brenneria and additional Pectobacterium taxa.

In Fig. 1b, a markedly different phylogeny is revealed from the data set containing the 18 animal pathogen taxa. The E. rhapontici 2, P. chrysanthemi 2, and P. carotovorum ssp. carotovora 2 clade is no longer sister group to the Erwinia–Pantoea clade, however, Pantoea is still derived from within the erwinias. Pantoea, which remains monophyletic, has E. lupinicola 1 as its closest relative, and together they are sister group to E. amylovora 1, E. mallotivora 1 and 2, and E. tracheiphila 1. Taxa are also rearranged in the large plant pathogen clade of Erwinia, Brenneria, and Pectobacterium species, with some examples including Brenneria alni 1, E. psidii 1, and P. carotovorum ssp. carotovora 1, though there is little support for conflict in this large clade. Thus, from these two trees, which include the same 27 plant pathogen sequences and only differ in the number of animal pathogen representatives and outgroup taxa, it is apparent that placement of the enterobacterial phytopathogenic species is affected by the number of animal pathogen taxa included in the phylogenetic analyses.

Figures 2 and 3 also depict different phylogenetic patterns, though many nodes lack support. In both trees, Erwinia, Brenneria, and Pantoea each form monophyletic clades, whereas Pectobacterium is polyphyletic. Additionally, Erwinia and Pantoea species consistently group as each other’s closest relatives, as do Brenneria and Pectobacterium. However, the placement of Pectobacterium cypripedii varies in the two phylogenies, with P. cypripedii serving as the closest relative to the Erwinia clade in Fig. 2 and to the Brenneria clade in Fig. 3. The two trees also differ from the previously published phylogenies of both Kwon et al. (1997) and Hauben et al. (1998). However, because support values for most of the nodes are very low in both the previously published and the reanalyzed trees, these differences can likely be attributed to poorly supported and unstable nodes, rather than to differences among analytical methods.

It is noteworthy that none of the phylogenies presented in this study are consistent with the current placement of the plant pathogen genera within the Enterobacteriaceae. The trees resulting from the first two data sets consisting of Erwinia, Brenneria, Pectobacterium, and Pantoea sequences generated in our laboratory support multiple independent origins for the phytopathogenic enterobacterial species (Fig. 1). The third tree, derived from a reanalysis of the Kwon et al. (1997) data (data set C), supports a single origin for the phytopathogens (Fig. 2), whereas the fourth tree, derived from a reanalysis of the Hauben et al. (1998) data (data set D), supports two independent origins, one for Erwinia and Pantoea and one for Brenneria and Pectobacterium (Fig. 3). Moreover, it is important to note that even though the same plant pathogen species are represented in all four trees, the strains used to obtain the 16S rDNA gene sequences in data sets A and B are different from those used by Kwon et al. (1997) or by Hauben et al. (1998). Nonetheless, these analyses indicate that 16S rDNA gene sequences may be limited in their ability to reconstruct the phylogeny of enterorbacterial plant pathogens at the genus level.

To identify the reasons that 16S rDNA gene trees are so different from one another, we examined the data set alignments, which consisted of large highly conserved regions interspersed by small highly variable regions (Fig. 4). From our 16S rDNA alignment of the 1282 bp in data set B, we were able to identify approximately six hypervariable domains. The conserved parts of the gene sequences were nearly identical for all the enterobacterial taxa, whereas the variable parts seemed to consist of a random array of nucleotides. The observed rate heterogeneity is likely to reflect evolutionary constraints related to rRNA function. Some of the stem-and-loop structures contributing to the formation of the ribosomal subunit interact with other components and these structures must be conserved. Therefore, as is the case for Streptomyces (Ueda et al. 1999), portions of the 16S rDNA gene are subject to different evolutionary pressures.

Fig. 4
figure 4

A portion of the 16S rDNA nucleotide sequence alignment for 40 enterobacterial species/strains analyzed here depicting two of the six hypervariable sequence regions (designated ‘V2’ and ‘V3’) and adjacent conserved nucleotide segments (designated ‘C’). The region of the alignment shown here was extracted from the full-length sequence alignment of the 16S rDNA data set used to construct Fig. 1b and exemplifies the conserved and variable regions scattered throughout the 16S rDNA sequence alignment. The nucleotide positions are numbered at the top of the alignment and correspond to the full-length alignment analyzed in Fig. 1b

Full size image

To evaluate the resolution provided by the conserved portions of the gene sequences, we edited data set B, eliminating all variable regions and analyzing only the conserved regions of the gene sequences. Of the original alignment comprising a total of 1282 bp, 67 bp was excluded from the phylogenetic analysis. These included positions 338–362, 523–533, 885–908, and 1017–1027. The resulting ML tree (Fig. 5) resolved minimal nodes differentiating the outgroup and ingroup taxa, as well as the Yersinia, the Pantoea, and some Erwinia species, while most terminal nodes were unresolved and unsupported, resulting in a widespread polytomy. For example, the largest portion of the tree containing Brenneria quercina 2, Pectobacterium cypripedii 1, Escherichia coli, Escherichia vulneris, Shigella dysenteriae, Shigella sonnei, and others failed to resolve relationships among this diverse group of plant and animal pathogen representatives. As a result, different regions of the 16S rDNA gene sequence provide different phylogenetic signals, and the topology of the trees changes according to the number of taxa included.

Fig. 5
figure 5

Maximum likelihood (ML) analysis of the conserved 16S rDNA sequence regions exclusively. The 49 taxa included here are the same as those analyzed in Fig. 1b. The likelihood tree shown here was produced in GARLI and viewed in PAUP* and illustrates the loss of phylogenetic resolution when hypervariable domains are omitted from the alignment. ML bootstrap nodal support was generated in PAUP* and subsequent values are reported for each node in parentheses

Full size image

It was observed that the deeper nodes in the four phylogenies (Figs. 1–3) were not well supported by either the ML bootstrap or the Bayesian values. In the first two trees, the deeper nodes are denoted by fairly low bootstrap values but high Bayesian values (Fig. 1), whereas in Figs. 2 and 3 most support values, including many of those more derived in the tree, are below 50 and 0.95 respectively, lending no support to the backbone of the two phylogenies. It is interesting that low bootstrap support values were also present in both the Kwon et al. (1997) and the Hauben et al. (1998) original published phylogenies. Thus, we cannot confidently determine whether the plant pathogen genera within the Enterobacteriaceae are monophyletic as in Fig. 2 or polyphyletic as in Figs. 1 and 3.

Discussion

Despite known variation between previously published 16S rDNA gene phylogenies of enterobacterial species (Hauben et al. 1998; Kwon et al. 1997; Martinez et al. 2004; Sproer et al. 1999; Young 2001), it was anticipated that use of the same methods for alignment and analysis of the four data sets would result in robust and consistent phylogenies. Unexpectedly, however, the striking inconsistency present in the 16S rDNA gene trees resulting from our analyses (Figs. 1–3) has signaled a potential difficulty in correctly revealing the taxonomic history of the enterobacterial plant pathogens for this gene.

The striking difference in sequence conservation observed among various segments of the 16S rDNA sequences analyzed here warrants further discussion. It was apparent that conserved regions of the gene offer little taxonomic resolution below the family level, because they are not variable among closely related enterobacterial taxa. These same regions do, however, differ among more distantly related taxa, such as the Enterobacteriaceae and Pseudomonadaceae (Toth et al. 2006), and have been useful in defining higher-level bacterial groupings (Olsen and Woese 1993). The variable 16S rDNA domains, on the other hand, appear potentially unique to species, and have been used to discriminate among closely related taxa (Louws et al. 1999). Even among species-level analyses, however, these variable regions may accrue substantial background noise (homoplasy), as additional taxa are included in analyses. Consequently, in our analyses few stable characters can be found supporting the different groups of taxa, resulting in inconsistent and poorly supported phylogenies. This was confirmed through our analysis of the conserved 16S rDNA sequence regions (Fig. 5).

The potentially confounding phylogenetic signal emanating from the variable portions of 16S rDNA warrants a further cautionary note when undertaking short repetitive sequencing and genotyping of populations. This is particularly true for methods such as pyrosequencing and primer extension assays where only a few nucleotide substitutions are culled for diagnostic identification or diversity assessment (Ahmadian et al. 2000; Jordan et al. 2005; Monstein et al. 2001). For assays targeting 16S rDNA sequences, the end result of any short sequencing approach may be confounded, particularly if the informative SNP(s) resides within the variable domains noted here. Identification or diversity assessments that rely on such targets, especially in the absence of any surrounding sequence information from the 16S rDNA gene itself or other conserved genes, may do more to obscure our understanding of strain relationships than to accurately delineate evolutionary relatedness.

Substantial topological differences were noted between the trees reported here. While observed differences between the two sets of trees (Figs. 1a and 2, Figs. 1b and 3) could result from homoplasy in the highly variable regions of the 16S rDNA gene sequences, intraspecific variation may have also been a contributor to these differences. Here, we provide support for the presence of both homoplasy and intraspecific variation from our analyses. The trees presented in Fig. 1 consist of the same plant pathogen sequences, and only vary in the number of animal pathogen and outgroup taxa. As a result, observed differences between Fig. 1a and 1b cannot be attributed to intraspecific variation among Erwinia, Brenneria, Pectobacterium, and Pantoea strains, but are most probably the result of increased character conflict introduced by the additional sequences. Alternatively, in Figs. 1a and 2, we kept the number of animal pathogen representatives and outgroup taxa constant, and used sequences from different plant pathogen accessions, still representing most of the same species. Hence, the differences between these two trees are not due to an increase in animal pathogen representatives but, rather, appear to result from intraspecific variation (e.g., homoplasy between members of the same species in the hypervariable regions of the gene). This suggests that the phylogenetic positions of species might be affected by sequence variation among strains. As a result, caution should be taken when drawing conclusions about taxonomic groups from reconstructions that fail to account for intraspecific variation.

In general, the 16S rDNA data failed to recover the basal and intermediate nodes. Subsequently, our analyses were unable either to characterize the relationships among the plant pathogen species or to position the genera within the Enterobacteriaceae. As noted previously, many of the deep nodes are poorly supported and unstable. Poorly supported nodes were evident in the previously published phylogenies of enterobacterial taxa as well (Hauben et al. 1998; Kwon et al. 1997; Moran et al. 2005; Sproer et al. 1999), yet genera, species, and subspecies are revised continually, largely on the basis of these unstable nodes resulting from only a few phylogenetically informative sequence changes.

Taken together, these data indicate that current enterobacterial 16S rDNA phylogenies of the plant pathogen genera Erwinia, Brenneria, Pectobacterium, and Pantoea may be confounded, thus weakening their utility as taxonomic anchoring points for additional studies such as ascertaining the lateral gene transfer pathways of virulence. The observed inconsistencies among these 16S rDNA phylogenies result from a change in the number of homoplasious sites present, which could be explained by one of two mechanisms. The variable and conserved domains of the 16S rDNA gene are subject to different functional constraints, and as a result, a more relaxed tolerance of mutations in the hypervariable regions would result in an increased number of homoplasious sites. Alternatively, horizontal gene transfer of short segments of DNA into these regions, which do not affect gene function, would also serve to confound phylogenetic signal. These pitfalls in part have prompted a move toward combined sequence analyses of multiple loci when undertaking taxonomic studies. It should be stressed that while 16S rDNA analysis is useful for identifying bacterial strains and reconstructing relationships among large and disparate taxonomic groups, it may retain less utility when attempting reconstructions among more closely related bacterial taxa. Additionally, the observed tree differences, attributed to intraspecific variation, reinforce the notion that using a single bacterial strain to represent a species may not be practical when reconstructing microbial phylogenies, particularly where taxonomical inferences are drawn and species or genera are emended. The quest for a robust and informative DNA-sequence-based phylogeny of Erwinia, Brenneria, Pectobacterium, and Pantoea will likely require examination and analysis of other less universally conserved housekeeping genes.