Reclassification of Pasteurella gallinarum, [Haemophilus] paragallinarum, Pasteurella avium and Pasteurella volantium as Avibacterium gallinarum gen. nov., comb. nov., Avibacterium paragallinarum comb. nov., Avibacterium avium comb. nov. and Avibacterium volantium comb. nov.

Members of the family Pasteurellaceae are commonly isolated from birds – both wild and domestic (Bisgaard, 1993). The use of 16S rRNA gene sequencing has revealed that many of these avian members of the Pasteurellaceae form phylogenetically related clusters (Dewhirst et al., 1993; Olsen et al., 2004), with Dewhirst et al. (1993) recognizing avian clusters 3A, 3D and 7. More recently, the avian 3A and 3D clusters of Dewhirst et al. (1993) have been retermed avian 16S rRNA cluster 18 (Olsen et al., 2004). Cluster 7 of Dewhirst et al. (1993) is now termed 16S rRNA cluster 21, also known as the ‘Testudinis' cluster (Olsen et al., 2004).

The most recent taxonomic study has indicated that avian 16S rRNA cluster 18 of Olsen et al. (2004) consists of at least 11 species or species-like taxa (Christensen et al., 2003b). Within avian 16S rRNA cluster 18, the new genus Gallibacterium, which incorporates organisms once known as avian [Pasteurella haemolytica], ‘Actinobacillus salpingitidis’ and Pasteurella anatis, has recently been proposed (Christensen et al., 2003a). Similarly, Bisgaard taxon 33 has been reclassified as Volucribacter psittacicida and V. amazonae (Christensen et al., 2004b).

The other major subcluster within avian 16S rRNA cluster 18 consists of a collection of four named species that are well known to veterinary bacteriologists – [Haemophilus] paragallinarum, Pasteurella gallinarum, Pasteurella avium, Pasteurella volantium – as well as the unnamed taxon Pasteurella sp. A. While all of the named species are routinely encountered in the investigation of upper respiratory tract disease in birds, only [Haemophilus] paragallinarum is widely regarded as a primary pathogen, being the causative agent of infectious coryza, an economically important disease of chickens (Blackall et al., 1997). While not universally accepted as a primary pathogen, there are a number of publications reporting on the role of Pasteurella gallinarum as a pathogen in infections of chickens (Bock et al., 1977; Droual et al., 1992a, b; Mohan et al., 2000; Mushin et al., 1977; Terzolo et al., 1980; Yadav et al., 1977) and turkeys (Bisgaard et al., 2005). The other species within this subgrouping of 16S rRNA cluster 18, Pasteurella avium and Pasteurella volantium, are regarded as part of the normal microbiota of the upper respiratory tract of chickens (Bisgaard, 1993; Mutters et al., 1985a). In the current study, based on phenotypic properties as well as the results of the sequencing of 16S rRNA genes, we propose a revised taxonomy that accommodates these four species within a single genus.

The bacteria used in this study consisted of the type strain of [Haemophilus] paragallinarum, NCTC 11296^T, plus the three reference strains for the Page serotyping scheme for [Haemophilus] paragallinarum, strains 0083, 0222 and Modesto, respectively representing serovars A, B and C (Page, 1962). In addition, eight isolates of the NAD-independent form of [Haemophilus] paragallinarum (Mouahid et al., 1992), termed [Haemophilus] paragallinarum biovar 2 as suggested by Mutters et al. (1989), were included.

Phenotypic characterization of the type strain of [Haemophilus] paragallinarum and the eight strains of [Haemophilus] paragallinarum biovar 2 was performed according to Bisgaard et al. (1991). For the type strain of [Haemophilus] paragallinarum, all media were supplemented with 1 % chicken serum and 0·0025 % NAD (reduced form).

16S rRNA gene sequencing of [Haemophilus] paragallinarum biovar 2 strain SA7177 was performed as reported previously (Angen et al., 1998). For the type strain of [Haemophilus] paragallinarum and the three Page serovar reference strains of [Haemophilus] paragallinarum, the 16S rRNA gene was sequenced as described previously (Blackall et al., 2001).

In addition to the five sequences determined in the present study, GenBank was searched for highly similar 16S rRNA gene sequences. The selected sequences included 11 taxa of the Pasteurellaceae that are isolated from birds and which fall within the avian 16S rRNA cluster 18 of Olsen et al. (2004) as recently extended by Christensen et al. (2003b). A further seven taxa of the Pasteurellaceae that are either isolated from birds or represent the major 16S rRNA clusters within the family were included (Olsen et al., 2004). The strains of [Haemophilus] paragallinarum and other members of the avian 16S rRNA cluster 18 used in the 16S rRNA gene sequencing are listed in the supplementary material in IJSEM Online.

Pairwise comparisons for similarity were performed by bestfit (GCG, Wisconsin Sequence Analysis Package). The alignment was constructed by pileup (GCG) and included the region between Escherichia coli positions 64 and 1391 of the rrnB gene, with 1246 positions left after removal of ambiguous positions and 218 distinct data patterns analysed. Maximum-likelihood analysis including bootstrap analysis was performed by fastDNAmL (Olsen et al., 1994) run on a Linux 7.2-compatible server. The analysis was run with a transition/transversion ratio of 1·5 and rate heterogeneities with 35 rates were introduced into the fastDNAmL analysis by DNArates 1.1 (). Parsimony and neighbour-joining analysis were performed by phylip (Felsenstein, 1995) on a Digital UNIX computer.

All eight strains of [Haemophilus] paragallinarum biovar 2 as well as the type strain, representing biovar 1, were non-motile, Gram-negative rods that were catalase-negative, oxidase-positive and fermented glucose without gas production. Strain CCUG 12835^T showed satellitic growth, while the biovar 2 strains did not. All nine strains did not cause haemolysis of bovine red blood cells and were negative in the Simmons' citrate, acid mucate, malonate, methyl red and Voges–Proskauer tests. The strains were alanine aminopeptidase-positive and reduced nitrate, without gas production, and did not possess urease, arginine dihydrolase, lysine decarboxylase, ornithine decarboxylase, phenylalanine deaminase or gelatinase. Strain CCUG 12835^T did not show phosphatase activity, while the eight biovar 2 strains did. All nine strains did not hydrolyse Tween 20 or Tween 80 and all failed to grow on MacConkey agar. The strains all produced acid from d-mannitol, d-sorbitol, d-fructose, d-mannose and sucrose. No acid was produced by any strain from glycerol, meso-erythritol, adonitol, d-arabitol, xylitol, l- or d-arabinose, l-xylose, dulcitol, meso-inositol, d-fucose, d-galactose, l-rhamnose, l-sorbose, cellobiose, lactose, d-melibiose, trehalose, d-melezitose, raffinose, d-glycogen, inulin, aesculin, amygdalin, arbutin, gentiobiose, salicin, d-turanose or β-N-CH₃-glucosamid. The biovar 2 strains produced acid from d-ribose, d-xylose and l-fucose, while the biovar 1 strain CCUG 12835^T did not. Strain CCUG 12835^T produced acid from maltose and dextrin, while the eight biovar 2 strains did not. None of the strains were positive for β-galactosidase, β-glucosidase, α-fucosidase, α-galactosidase, α-glucosidase, β-glucuronidase, α-mannosidase or β-xylosidase activity.

16S rRNA gene sequences were obtained in the region covering E. coli positions 28–1491 for four strains of [Haemophilus] paragallinarum biovar 1 (NCTC 11296^T, 0083, 0222 and Modesto) and one strain of [Haemophilus] paragallinarum biovar 2 (SA7177). The 16S rRNA gene sequences of the three strains NCTC 11296^T, 0083 and 0222 were identical when ambiguous positions were removed.

The similarity of 16S rRNA gene sequences of the type strain and strains 0222 and 0083 of [Haemophilus] paragallinarum to strain Modesto was 98·8 %, while the similarities between these strains of biovar 1 and strain SA7177 of biovar 2 were between 98·9 and 97·7 %. The highest similarities between strains of [Haemophilus] paragallinarum and other taxa of the Pasteurellaceae were obtained to the species Pasteurella gallinarum, Pasteurella volantium and Pasteurella avium and to Pasteurella sp. A, with between 96·8 and 98·1 % similarity. Similarities between strains of [Haemophilus] paragallinarum and strains of Gallibacterium ranged between 92·7 and 94·8 % and similarities of between 92·2 and 93·8 % were obtained to Volucribacter and the unnamed taxa 2, 3 and 34 of Bisgaard. The highest similarities to other members of the Avian group were 94·8 % to strain IPDH 697/78 of Gallibacterium anatis and 94·9 % to strain 101 of V. psittacicida. Outside the Avian group (which includes the genera Avibacterium, Gallibacterium and Volucribacter and taxa 2, 3 and 34 of Bisgaard), the highest similarity obtained was between the type strain of [Haemophilus] paragallinarum and the type strain of Actinobacillus capsulatus, at 94·7 %.

The phylogenetic analysis was performed with 24 sequences, including three from the present investigation; [Haemophilus] paragallinarum strains NCTC 11296^T, Modesto and SA7177. Maximum-parsimony analysis resulted in 25 trees with identical numbers of steps. The phylogenetic analysis shown in Fig. 1⇓ shows that the clade containing [Haemophilus] paragallinarum and the taxa Pasteurella gallinarum, Pasteurella volantium, Pasteurella avium and Pasteurella sp. A was monophyletic.

The family Pasteurellaceae has been studied extensively by molecular methods such as DNA–DNA hybridization, rRNA–DNA hybridization and 16S rRNA gene sequence analysis in recent years (Angen et al., 1999, 2003; Christensen et al., 2003a; De Ley et al., 1990; Dewhirst et al., 1993; Mutters et al., 1989). These studies have clearly shown that the original three genera of Haemophilus, Actinobacillus and Pasteurella were polyphyletic and have resulted in the establishment of six new genera: Gallibacterium, Histophilus, Mannheimia, Lonepinella, Phocoenobacter and Volucribacter (Angen et al., 1999, 2003; Christensen et al., 2003a, 2004b; Foster et al., 2000; Osawa et al., 1995). The studies on these new genera, particularly Gallibacterium (Christensen et al., 2003a) and Volucribacter (Christensen et al., 2004b), have highlighted that the other members of the avian 16S rRNA cluster 18 within the Pasteurellaceae represent another group of taxa that needs reclassification.

The strains represented by the three serovars and two biovars of [Haemophilus] paragallinarum analysed by 16S rRNA gene sequence comparison formed a single subgroup within the Avian 16S rRNA-group of Pasteurellaceae together with the species Pasteurella gallinarum, Pasteurella avium, Pasteurella volantium and the unnamed species A of Pasteurella, showing between 96·8 and 98·1 % similarity. The monophyly of the species [Haemophilus] paragallinarum as well as Pasteurella gallinarum and Pasteurella sp. A was unsupported by bootstrap, parsimony and neighbour-joining analysis. The species Pasteurella gallinarum was even polyphyletic by maximum-likelihood analysis. The explanation for this is probably that 16S rRNA gene sequence comparison is insufficient to resolve all species. This issue is well known for all species (Stackebrandt & Goebel, 1994; Christensen et al., 2002a). A similar comb-like topology has previously been shown for some taxa of Mannheimia (Blackall et al., 2001) and Actinobacillus equuli subsp. equuli and subsp. haemolyticus (Christensen et al., 2002a).

These results have confirmed earlier findings that the species [Haemophilus] paragallinarum, Pasteurella gallinarum, Pasteurella volantium and Pasteurella avium comprise a monophyletic group within the family Pasteurellaceae (Christensen et al., 2003b; Dewhirst et al., 1992, 1993). While the 16S rRNA gene phylogenetic tree shown in Fig. 1⇑ was generated by maximum-likelihood analysis, the monophyletic group that comprises the new genus Avibacterium was also supported by the phylograms generated by neighbour-joining and maximum-parsimony analysis (not shown). This support by all three analytical methods indicates a very high likelihood for the grouping. It has been proposed that new genera should be capable of being recognized from existing genera on the basis of both 16S rRNA gene sequence analysis and phenotypic data (Murray et al., 1990). As there is clear support by 16S rRNA gene sequence analysis and an existing knowledge of distinguishing phenotypic characters (see Tables 1 and 2⇓⇓), we propose that the four species should be housed within a separate genus for which we propose the name Avibacterium gen. nov.

At the genus level, Avibacterium is clearly phenotypically distinguishable from all other genera in the family Pasteurellaceae except the genus Gallibacterium (Table 1⇑). All members of the genus Gallibacterium do not show symbiotic growth, while most members of the genus Avibacterium show symbiotic growth (Table 1⇑). Also, most isolates of the genus Gallibacterium show strong haemolysis on blood agar (Christensen et al., 2003a), allowing a further clear differentiation of these haemolytic forms of Gallibacterium from the non-haemolytic genus Avibacterium. However, the non-haemolytic forms of Gallibacterium, i.e. G. anatis biovar anatis, are not distinguishable at the genus level from the NAD-independent forms of Avibacterium (Table 1⇑). Inclusion of speciation-level identification tests (Table 2⇑) allows all members of the genus Avibacterium to be clearly distinguished from all members of the genus Gallibacterium. Table 2⇑ presents phenotypic properties established in this study as well as other studies in which strains have been confidently identified on the basis of genetic methods (Bisgaard et al., 2005; Christensen et al., 2002b, 2003b; Mutters et al., 1985a). Additional phenotypic data for Avibacterium paragallinarum comb. nov. has been included from the studies of Blackall & Reid (1982) and Blackall et al. (1989), as all the strains of Avibacterium ([Haemophilus]) paragallinarum in these studies have been subsequently reconfirmed as Avibacterium ([Haemophilus]) paragallinarum by PCR (Chen et al., 1996) and/or have been serotyped by the Page scheme (Page, 1962). In addition to the differences shown in Table 2⇑, differences in polyamine patterns can be used for separation between Gallibacterium and Avibacterium (Busse et al., 1997), just as crossed immunoelectrophoresis of antigens clearly separated the genera (Schmid et al., 1991).

The significance of V-factor dependency in the taxonomy of Haemophilus species and related organisms was discussed by Niven & O'Reilly (1990). A number of species within the Pasteurellaceae that contain both NAD-dependent and NAD-independent strains have been reported, e.g. Actinobacillus pleuropneumoniae, which causes pleuropneumonia in swine (Pohl et al., 1983), Avibacterium ([Haemophilus]) paragallinarum (Bragg et al., 1993; Mouahid et al., 1992), Haemophilus parainfluenzae, which can cause pneumonia and meningitis in humans (Gromkova & Koornhof, 1990), and Haemophilus ducreyi, which causes the sexually transmitted disease chancroid in humans (Windsor et al., 1993). To date, there have been no extensive studies on the variations, if any, between the 16S rRNA gene sequences of NAD-dependent and NAD-independent strains within these species. Hence, it is not possible to compare the level of sequence variation we found between the NAD-dependent and NAD-independent Avibacterium ([Haemophilus]) paragallinarum strains with comparable data from other species.

Martin et al. (2001) have recently shown that a single plasmid-encoded gene, nadV, from Haemophilus ducreyi is responsible for the occurrence of NAD independence in Haemophilus ducreyi and can confer NAD independence on NAD-dependent strains of both Actinobacillus pleuropneumoniae and Haemophilus influenzae. While the explanation for the NAD independence of Avibacterium ([Haemophilus]) paragallinarum biovar 2 strains has not been elucidated, the work of Martin et al. (2001) suggests that NAD independence may be a very minor component of the overall genome.

The available DNA–DNA hybridization data of Mouahid et al. (1992), who reported 89 % DNA binding between strain ATCC 29545^T and the biovar 2 strain SA 4461, indicate that the biovar 1 and 2 forms of Avibacterium ([Haemophilus]) paragallinarum can be accommodated in one species. The DNA fingerprinting work of Miflin et al. (1995) has provided evidence that the South African NAD-independent isolates of Avibacterium ([Haemophilus]) paragallinarum are clonal, indicating that the DNA–DNA hybridization data for strain SA 4461 are applicable for all of the biovar 2 strains in this study. As well, all South African biovar 2 isolates of Avibacterium ([Haemophilus]) paragallinarum tested to date have yielded a positive reaction for the [Haemophilus] paragallinarum-specific PCR (Chen et al., 1996; Miflin et al., 1999). Furthermore, all of the South African biovar 2 isolates that have been serotyped (Miflin et al., 1995) have been allocated to the existing Page serotyping scheme, which was originally created using typical biovar 1 strains (Page, 1962). Hence, there are a number of separate bodies of evidence – the DNA–DNA hybridization data of Mouahid et al. (1992), the clonality of the South African biovar 2 isolates as shown by Miflin et al. (1995), the positive reaction in the [Haemophilus] paragallinarum PCR (Chen et al., 1996, Miflin et al., 1999), the ability to assign the South African isolates of biovar 2 to the existing [Haemophilus] paragallinarum serotyping scheme (Miflin et al., 1995) and the 16S rRNA gene sequencing results of the current study – that all support the inclusion of the biovar 2 strains within the species Avibacterium ([Haemophilus]) paragallinarum.

Recently, biovar 2 isolates of [Haemophilus] paragallinarum have been identified in Mexico (García et al., 2004). While these biovar 2 isolates were not available for the current study, the Mexican isolates have been shown to be positive in the [Haemophilus] paragallinarum-specific PCR and have been allocated to existing Page serovars (García et al., 2004). This is strong evidence that these new Mexican biovar 2 isolates of [Haemophilus] paragallinarum are also typical of the species Avibacterium paragallinarum.

In the initial description of Avibacterium (Pasteurella) avium, two biovars were recognized, with biovar 1 being NAD-dependent and associated with chickens while biovar 2 was NAD-independent and associated with cattle (Mutters et al., 1985a). There has been continuing confusion about Pasteurella avium biovar 2, with Miflin & Blackall (2001) reporting that these biovar 2 isolates gave a positive reaction with a Pasteurella multocida-specific PCR based on the 23S rRNA gene. In recent work, an extended genotypic study has shown conclusively that Pasteurella avium biovar 2 is, in fact, Pasteurella multocida (Christensen et al., 2004a). Hence, our study has only involved the avian, NAD-dependent forms of Avibacterium (Pasteurella) avium.

The proposed new genus contains four well-characterized species: Avibacterium gallinarum comb. nov., Avibacterium paragallinarum comb. nov., Avibacterium avium comb. nov. and Avibacterium volantium comb. nov. In addition, the taxon once called Pasteurella sp. A is clearly a member of the genus and is best termed Avibacterium sp. A. Phenotypic differentiation of the members of the new genus from Gallibacterium is possible but relies on a few key characteristics such as catalase, NAD dependence, haemolysis, ONPG reaction and fermentation of glycerol, d-mannitol, d-galactose and maltose (Table 2⇑). The differentiation of these organisms has always relied upon a few key characteristics, as has been highlighted in the traditional veterinary texts and laboratory manuals (Rhoades & Rimler, 1991; Rimler et al., 1998).

The existence of Avibacterium paragallinarum and Avibacterium gallinarum within a single genus does raise the possibility of some confusion. There has been a long tradition within the genus Haemophilus of the use of the prefix ‘para’ to highlight that such a species is like an existing species but differs in growth factor requirements. This was the case for avian haemophilic organisms – with many publications from the 1960s to the 1980s referring to two species as the cause of infectious coryza – ‘Haemophilus gallinarum’ and Avibacterium [Haemophilus] paragallinarum (Blackall & Yamamoto, 1989). These two species differ in growth factor requirements – ‘Haemophilus gallinarum’ being X- and V-factor dependent and [Haemophilus] paragallinarum being V-factor dependent (Blackall & Yamamoto, 1989). While no valid strains of ‘Haemophilus gallinarum’ exist, experimental work has shown that isolates of [Haemophilus] paragallinarum mistakenly appear to be X- and V-factor dependent when tested using the limited methodologies of the 1930s to 1950s (Blackall & Yamamoto, 1989). Hence, while it is not possible to be totally certain, it appears that the organism ‘Haemophilus gallinarum’ never actually existed but was a mischaracterized Avibacterium ([Haemophilus]) paragallinarum (Blackall & Yamamoto, 1989). There is a possibility that some bacteriologists may assume the Avibacterium gallinarum is a renamed ‘Haemophilus gallinarum’ or that species Avibacterium paragallinarum and Avibacterium gallinarum reflect the existence of species with V-factor and X- and V-factor dependency, respectively. We accept this risk of confusion as a reasonable compromise: the only alternative would be to propose a new species name for Avibacterium paragallinarum, since Avibacterium gallinarum was reported first and consequently has priority according to the rules of nomenclature. We do not feel that this renaming is justifiable.