Phylogenetic analysis of the genera Proteus, Morganella and Providencia by comparison of rpoB gene sequences of type and clinical strains suggests the reclassification of Proteus myxofaciens in a new genus, Cosenzaea gen. nov., as Cosenzaea myxofaciens comb. nov.

Members of the three genera Proteus, Morganella and Providencia are components of the normal bacterial flora of the intestinal tracts of humans and animals and are widespread in the environment. Some species belonging to these genera have been recognized as opportunistic pathogens, causing primary and secondary infections (O’Hara et al., 2000); these are mainly hospital-acquired infections (Burke et al., 1971; Krajden et al., 1987) with a high mortality rate directly related to bacteraemia (Kim et al., 2003). These bacteria are frequently involved in urinary tract infections (UTI) and species of the genus Proteus account for approximately 6 % of nosocomially acquired UTI (Johansen et al., 2006).

DNA relatedness studies have shown that the members of the tribe Proteeae, containing the three genera Proteus, Morganella and Providencia, are only distantly related to other species in the family Enterobacteriaceae and their taxonomic position is at the periphery of this family (Brenner et al., 1978). At the time of writing, the genus Proteus included five species: Proteus mirabilis, Proteus vulgaris, Proteus penneri, Proteus myxofaciens and Proteus hauseri (O’Hara et al., 2000a, b; Penner, 1992). The genus was first described by Hauser in 1885 and was successively separated into two species, Proteus vulgaris and Proteus mirabilis, on the basis of their ability to ferment maltose. Strains of Proteus vulgaris formed three biogroups characterized by different biochemical reactions for indole production, salicin fermentation and aesculin hydrolysis. Based on DNA–DNA hybridization studies, biogroup 1 was shown to be a separate genetic species and was named Proteus penneri (Hickman et al., 1982), biogroup 2 retained the name Proteus vulgaris and biogroup 3 was separated into four distinct genomospecies (Brenner et al., 1995; Penner, 2005a). The species name Proteus hauseri was proposed for Proteus vulgaris genomospecies 3, while genomospecies 4, 5 and 6 remained unnamed pending better phenotypic differentiation (O’Hara et al., 2000b). Proteus myxofaciens was included in the genus based on hybridization studies, since labelled Proteus myxofaciens DNA showed relatedness values of 44–46 % to both Proteus mirabilis and Proteus vulgaris strain PR1 but only 12–26 % to DNA from strains of members of the genera Morganella and Providencia and only about 10 % to other members of the family Enterobacteriaceae (Brenner et al., 1978).

The genus Providencia initially comprised five species, Providencia alcalifaciens, Providencia stuartii, Providencia rettgeri, Providencia rustigianii and Providencia heimbachae, as described by Penner (1992). More recently, three additional species of the genus Providencia have been described: Providencia burhodogranariea, Providencia sneebia and Providencia vermicola (Juneja & Lazzaro, 2009; Somvanshi et al., 2006). In biochemical characterization studies on Providencia alcalifaciens, four biogroups were initially recognized (Ewing et al., 1972). After DNA relatedness tests, strains belonging to biogroup 3 were classified as a novel species, Providencia rustigianii (Hickman-Brenner et al., 1983). Strains of biogroup 4 were found to be biochemically closer to Providencia stuartii and were therefore transferred to this species (Brenner et al., 1978). Five biogroups were recognized in the biochemical characterization of Providencia rettgeri (Penner et al., 1975). On the basis of DNA relatedness, biogroup 5 was found to be inseparable from Providencia stuartii (Brenner et al., 1978) but not all biogroups of Providencia rettgeri have yet been examined for relatedness. Although DNA–DNA relatedness studies resolved the taxonomic position of many species and allowed the classification, without controversy, of most of the bacteria belonging to the Proteus–Providencia group, the present taxonomy of species with validly published names does not account for all the biogroups and genomospecies described within the group (Brenner et al., 1978; Ewing et al., 1972; O’Hara et al., 2000b; Penner et al., 1975). Morganella morganii is the sole species in the genus Morganella (Penner, 1992).

In a previous study, type and clinical strains of 11 different species of the genera Proteus, Morganella and Providencia were characterized by rRNA gene restriction pattern analysis after digestion by EcoRV and HincII (Pignato et al., 1999). Clinical isolates of Proteus mirabilis, Proteus penneri, M. morganii, Providencia rustigianii and Providencia heimbachae had identical or very similar ribotyping patterns to those of the respective type strains; they could, therefore, be grouped into five homogeneous and discrete ribogroups corresponding to the five species. On the other hand, distinct ribogroups were detected within clinical strains of the species Proteus vulgaris (two ribogroups), Providencia alcalifaciens (two ribogroups), Providencia rettgeri (four ribogroups) and Providencia stuartii (two ribogroups). These results raised concerns about how the current taxonomy corresponded with the genetic evolution of the respective genera.

16S rRNA gene sequence analysis is by far the most widely used method for the molecular identification and differentiation of bacterial species (Woese et al., 1990). However, members of the family Enterobacteriaceae have not been subjected to extensive phylogenetic 16S rRNA gene sequence analysis because the high degree of conservation in closely related species leaves many taxonomic problems unresolved (Brenner, 1992; Spröer et al., 1999). In particular, the genetic relationships between closely related species Proteus vulgaris and Proteus penneri could not be clearly resolved by this method (Cao et al., 2009). In the family Enterobacteriaceae, the use of rRNA for bacterial typing has also been questioned because of its allelic heterogeneity. Sequence divergence between several gene copies (seven in Escherichia coli and species of the genus Salmonella) makes direct sequencing of PCR products seldom suitable for achieving a representative sequence (Cilia et al., 1996; Hill & Harnish, 1981).

Among the core bacterial genes, the rpoB gene has been identified as one of the few potential candidates suitable for bacterial phylogenetic analyses and has been proposed as a powerful tool for universal bacterial genetic identification (Adékambi et al., 2009; Mollet et al., 1997). The rpoB gene, encoding the RNA polymerase β-subunit, is a highly conserved housekeeping gene and one copy is present in all bacteria because of its essential role in cellular metabolism. Primers and probes targeting the rpoB gene have been used for the specific detection and phylogenetic analysis of several bacterial groups belonging to the family Enterobacteriaceae (Drancourt et al., 1998; Hoffmann & Roggenkamp, 2003; Kwon et al., 2001; Merhej et al., 2008; Stoop et al., 2009). Partial rpoB gene sequencing and analysis has been proven to be more sensitive than using 16S rRNA gene sequences when used to differentiate type and clinical strains of 14 species of the family Enterobacteriaceae, including type strains of Proteus mirabilis and Providencia stuartii (Mollet et al., 1997). However, members of the tribe Proteeae have not yet been subjected to extensive phylogenetic analysis based on rpoB gene sequences.

In this study, partial rpoB gene sequences were obtained from type strains of 11 species of the genera Proteus, Morganella and Providencia: Proteus hauseri ATCC 13315^T, Proteus mirabilis ATCC 29906^T, Proteus myxofaciens ATCC 19692^T, Proteus penneri ATCC 33519^T, Proteus vulgaris ATCC 29905^T, Morganella morganii ATCC 25830^T, Providencia alcalifaciens ATCC 9886^T, Providencia heimbachae ATCC 35613^T, Providencia rettgeri ATCC 29944^T, Providencia rustigianii ATCC 33673^T and Providencia stuartii ATCC 29914^T. The phylogenetic structure of the three genera was evaluated, including in this evaluation the analysis of conserved sequences of 21 clinical strains belonging to different 16S rRNA gene-fingerprinting ribogroups and biogroups previously determined for these species (Supplementary Table S1, available in IJSEM Online) (Pignato et al., 1999). Genomic DNA was extracted using a commercial nucleic acid extraction kit (IsoQuick; InterBiotech ORCA Research, Othell, WA) according to the manufacturer’s instructions. A portion of the coding region of the rpoB gene was amplified with primers CM7 (5′-AACCAGTTCCGCGTTGGCCTGG-3′) and CM31b (5′-CCTGAACAACACGCTCGGA-3′), according to the technique described by Mollet et al. (1997). The PCR amplicons generated were sequenced with primers CM81 (5′-CAGTTCCGCGTTGGCCTG-3′) and CM31b(seq) (5′-TGAACAACACGCTCGG-3′) (Mollet et al., 1997) to obtain partial rpoB gene sequences. Sequence alignment and phylogenetic analysis was carried out on 930 nt fragments using the clustal w algorithm (Thompson et al., 1994) and mega software version 4.1 (Kumar et al., 2004). To ensure the stability and reliability of phylogenetic relationships among strains used in this study, phylogenetic trees were reconstructed by using the neighbour-joining and maximum-parsimony methods in mega 4.1 package, with the option of complete deletion of gaps, and the maximum-likelihood method in the phylip version 3.66 software package (Felsenstein, 2006). The Kimura two-parameter model was chosen as a substitution model for the neighbour-joining tree reconstruction. The statistical significance of the phylogenies inferred was estimated by bootstrap analysis with 1000 pseudoreplicate datasets. Accession numbers for the rpoB gene sequences of the strains used in this study are shown in Fig. 1.

The phylogenetic tree derived from partial rpoB gene sequences of type and clinical strains of 11 species of the genera Proteus, Morganella and Providencia (Fig. 1) showed genus-related branching patterns. The strains of the genus Providencia and almost all the strains of the genus Proteus grouped in two respective branches, while Morganella morganii was clearly separate in a third branch. However, Proteus myxofaciens constituted a well-defined fourth lineage. Bootstrap levels ≥57 % allowed the identification of rpoB clades that generally corresponded to the species level within the main genetic branches. No rpoB clade included strains belonging to different species and the majority of the species were confirmed as discrete and homogeneous, since they constituted a single monophyletic group. Within the Proteus branch, Proteus penneri and Proteus vulgaris strains were more closely related, while the type strain of Proteus hauseri stood apart as did the strains of Proteus mirabilis. Within the Providencia branch, two subclusters could be defined according to the phylogenetic relationships; one included the species Providencia alcalifaciens, Providencia rustigianii and Providencia rettgeri, while the second included Providencia stuartii and Providencia heimbachae. Clinical strains of the different species generally clustered in close proximity to their respective type strains but distinct rpoB clades could be recognized in Proteus vulgaris (VurpoB1–2), Providencia alcalifaciens (AlrpoB1–2) and Providencia rettgeri (RerpoB1–3). As shown in Supplementary Table S1, when the phylogenetic tree obtained using rpoB gene sequences was compared with the results of previously obtained ribotyping studies from the same type and clinical strains after genomic DNA digestion by EcoRV or HincII (Pignato et al., 1999), the rpoB clades generally corresponded to the ribogroups.

Percentage nucleotide identity and pairwise uncorrected (p-) distances were calculated for the rpoB gene sequences of type and clinical strains of the 11 different species examined (Supplementary Table S2). Sequence divergences of 2.3–18.9 % were observed in the rpoB gene region analysed when comparing strains of different species. On the basis of rpoB gene sequences, strains of M. morganii diverged by 14.1–16.1 % and 12.1–16.5 % from strains of the genera Providencia and Proteus, respectively, while the strains of the latter genera differed by 13.0–18.7 % from each other. The type strain of Proteus myxofaciens, however, was only distantly associated (12.1–12.2 % sequence divergence) with the closest taxon, M. morganii. Divergences between the rpoB gene sequences of the type strain of Proteus myxofaciens and other members of the genus (17.3–18.9 %) were similar to those calculated between Proteus myxofaciens and members of the genus Providencia (16.4–18.7 %), suggesting that a genetic distance exists at the genus-level between Proteus myxofaciens and the rest of the Proteus–Providencia group. The DNA G+C content of Proteus myxofaciens, determined by HPLC according to the method of Mesbah et al. (1989), was 36.5 mol%, while the type strains of other members of the genus Proteus ranged from 38 to 39.3 mol% (Penner, 2005a). Interestingly, sequence divergences of 2.4–1.9 % were observed when a 5′ 978 nt 16S rRNA gene fragment from the type strain of Proteus myxofaciens was compared to that of other type strains belonging to the same genus and 16S rRNA phylogenetic analysis could not clearly separate it from the other species of the genus Proteus (Fig. 2). However, the low degree of divergence in the 16S rRNA gene sequence was similar to that reported between closely related members of other genera within the family Enterobacteriaceae, such as between Escherichia coli and members of the genera Shigella and Salmonella (95.3–99.6 % similarity) or between members of the genera Enterobacter and Klebsiella (97.2–98.3 % similarity) (Fukushima et al., 2002). Among the other species of the genus Proteus, 5.1–7.6 % rpoB gene sequence divergences were observed, except when strains of Proteus penneri were compared to strains of Proteus vulgaris (2.3–2.7 % differences). The low rpoB gene sequence divergences between strains of Proteus penneri and Proteus vulgaris, being the two closest taxa among all those examined in this study, and the high sequence similarity observed among the rpoB gene sequences of Proteus penneri clinical strains suggests that this species, formerly Proteus vulgaris biogroup 1 (Hickman et al., 1982), could be the species that most recently diverged from other species within the genus Proteus. Accordingly, a single homogeneous restriction pattern had previously been observed in the EcoRV-digested rRNA gene fragments of all the strains of Proteus penneri analysed (Pignato et al., 1999). In the present study, the type strain of Proteus hauseri was confirmed as occupying a separate position with respect to Proteus vulgaris in the phylogenetic tree based on rpoB gene sequences; a sequence divergence of at least 5.1 % with respect to the rpoB gene sequence of other members of the genus Proteus further supports its position as a separate species within this genus. In fact, the present type strain Proteus hauseri ATCC 13315^T was the former type strain of Proteus vulgaris but it resided in a separate genogroup (genogroup 3) within biogroup 3, which was composed of only two strains that were not biochemically representative of the majority of the isolates identified as Proteus vulgaris (Brenner et al., 1995; O’Hara et al., 2000b). In addition, the five species of the genus Providencia that were analysed could be easily distinguished from one another based on rpoB gene sequence divergences, generally differing by 6.0–10.6 %, except for Providencia alcalifaciens and Providencia rustigianii, whose strains showed only 3.4–5.3 % rpoB gene sequence divergence.

The comparison of rpoB gene sequence divergences among clinical strains revealed that they were generally closely related to their respective type strain (Supplementary Table S2). However, as already shown by the phylogenetic analysis, distinct rpoB gene sequence types could be recognized in Proteus vulgaris, Providencia alcalifaciens and Providencia rettgeri. In particular, two clades, corresponding to the Proteus vulgaris rpoB clades observed in the phylogenetic tree, VurpoB1 (type strain ATCC 29905^T and clinical strain 27) and VurpoB2 (clinical strains 23 and 26), showed sequence divergences of 3.1–3.6 % from each other but both were equally distant from the other species of the genus Proteus. These two Proteus vulgaris rpoB clades, which diverged from each other more than they did from Proteus penneri sequences (2.3–2.7 %), also belonged to separate ribogroups (Pignato et al., 1999). Interestingly, strain 26, which belonged to the genetically heterogeneous biogroup 3, since it was positive for indole production and negative for salicin fermentation and aesculin hydrolysis, grouped with the same rpoB clade as the classical biogroup 2 Proteus vulgaris strain 23, further demonstrating the lack of accordance between biochemical types and genotypes, already demonstrated by the results of ribotyping. Of the two clinical strains of Providencia alcalifaciens analysed, one was identical to the type strain based on the rpoB gene fragment that was analysed (AlrpoB1 clade), while the other strain showed 3.9 % sequence divergence (AlrpoB2 clade). Three rpoB clades could be defined for Providencia rettgeri; one clade comprised the type strain and a single clinical strain (RerpoB1 clade); the second included four clinical strains, whose sequence divergence from the type strain was between 2.5 and 2.7 % of the fragment analysed (RerpoB2 clade), and the third was represented by a single clinical strain that differed from the first two sequence types by 5.9 and 5.7–5.8 %, respectively (RerpoB3 clade). Genetic diversity had already been observed by DNA–DNA hybridization studies with species of the genus Providencia. In particular, Providencia rettgeri contained at least two relatedness groups and a variety of biogroups and atypical strains that have been described but have yet to be examined for DNA-relatedness (Brenner et al., 1978). In fact, the rpoB clades observed in Providencia rettgeri and in Providencia alcalifaciens corresponded exactly to the ribogroups already described within these species (Pignato et al., 1999). On the contrary, the two ribotypes of Providencia stuartii grouped within a single rpoB clade.

Attempts have been made to define a cut-off for rpoB gene sequence-based identification of bacteria. Mollet et al. (1997) found an intraspecies similarity range of 98–100 % when they analysed a 512 bp fragment of the rpoB gene in clinical isolates of enteric strains. For 600–825 bp gene-fragments, a rpoB sequence similarity of at least 96–97 % seems to be the threshold for correct species identification (Adékambi et al., 2003; Khamis et al., 2003; La Scola et al., 2003). According to these limits, the ≥3.1 % sequence divergence detected between members of the two rpoB clades of Proteus vulgaris could be sufficient to suggest possible further taxonomic adjustments within this genus. Similarly, the detection of two Providencia alcalifaciens rpoB clades with 3.9 % sequence divergence from each other and the sequence divergence of the type strain of Providencia rettgeri from members of the two additional RerpoB clades (2.5–2.7 % and 5.9 %) provide further evidence of the lack of genetic homogeneity in strains of the two species of the genus Providencia, as demonstrated by 16S rRNA gene-fingerprinting.

Our results indicate that rpoB gene sequence comparison seems to be an appropriate method for inferring genetic relationships within the Proteus–Providencia group and suggest the need for taxonomic adjustments. Differentiation of strains presently belonging to the species Proteus vulgaris, Providencia alcalifaciens and Providencia rettgeri and the introduction of novel species according to the groups outlined by rpoB gene sequence analysis should be considered but larger studies, combining 16S rRNA gene sequence analysis with more discriminative methods like full rpoB gene sequence analysis and including the investigation of more than one gene (multi-locus sequence typing), are necessary in order to clarify the molecular relationships between strains of these species. The assignment of the type strain of Proteus myxofaciens to a separate genus was strongly supported by phylogenetic analysis based on rpoB gene sequences, showing low levels of sequence similarity to, and great evolutionary distances from, the other members of the tribe Proteeae. This was further supported by differences in the DNA G+C contents, since the 1.5–2.8 mol% differences between strains of Proteus myxofaciens and other species of the genus Proteus is comparable to the 0.3–5 mol% differences observed between Proteus myxofaciens and members of the genus Providencia, the closest related taxon (Penner, 2005a, b). Based on the evidence presented in this study, it is proposed that Proteus myxofaciens be classified as a new genus, Cosenzaea gen. nov., as Cosenzaea myxofaciens, comb. nov.