Development of a multilocus sequence typing scheme for intestinal spirochaetes within the genus Brachyspira

The fastidious anaerobic intestinal spirochaetes of the genus Brachyspira (formerly Serpulina) are commonly found colonizing the large intestines of a wide range of animals, including humans (Hampson & Stanton, 1997; Duhamel, 2001). Currently there are seven officially named species in the genus, and several provisionally named species. Some species are important causes of enteric disease, particularly in pigs and poultry. Brachyspira hyodysenteriae is the cause of swine dysentery (Taylor & Alexander, 1971) and Brachyspira pilosicoli causes intestinal spirochaetosis in pigs and other species (Taylor et al., 1980; Duhamel, 2001). In poultry, B. pilosicoli, Brachyspira intermedia and Brachyspira alvinipulli cause reduced egg production, wet faeces and dirty eggshells (Stephens & Hampson, 2001). Brachyspira innocens and Brachyspira murdochii have been found in various species of animals and birds, and are generally considered to be apathogenic. Brachyspira aalborgi appears to be largely confined to humans, and its potential role in enteric disease is controversial (Hovind-Hougen et al., 1982). Brachyspira canis (Duhamel et al., 1998), Brachyspira pulli (Stephens & Hampson, 2001) and Brachyspira suanatina (Råsbäck et al., 2007) are provisionally designated species; the last causes a disease in pigs similar to swine dysentery. Additionally, isolates exist that cannot be classified into presently recognized Brachyspira species because of atypical phenotypic or genotypic features or both (Atyeo et al., 1999b; Thomson et al., 2001; Jansson et al., 2004).

The routine identification and classification of Brachyspira species is normally based on culture and biochemical tests, often supported by species-specific PCR assays, which together make a very reliable combination for the clinically important porcine Brachyspira species. However, identification and classification of some Brachyspira isolates can be challenging, particularly when they have atypical features (Råsbäck et al., 2005, 2006). For many bacterial genera, species identification of isolates can be achieved by 16S rDNA sequence analysis (Woese, 1987), but for the Brachyspira genus, phylogenetic division based on 16S rDNA sequences is only reliable for the species that are not too closely related. For example, in a phylogenetic tree based on 16S rDNA sequences, B. hyodysenteriae and certain B. intermedia isolates clustered together, as did B. innocens and B. murdochii isolates (Pettersson et al., 1996). This is particularly bothersome because the two closely related species B. hyodysenteriae and B. intermedia have different veterinary medical significances in pigs. The latter is generally regarded as non-pathogenic, whereas B. hyodysenteriae can cause severe mucohaemorrhagic diarrhoea.

Multilocus enzyme electrophoresis (MLEE) has been an influential technique for identifying potentially new Brachyspira species (Lee et al., 1993; McLaren et al., 1997; Duhamel et al., 1998). Unfortunately, MLEE is a time-consuming and tedious technique that is not available in most diagnostic microbiology laboratories. Furthermore, MLEE is often not sufficiently discriminatory at the strain level for use in detailed molecular epidemiological studies on Brachyspira species. More discriminatory techniques that have been used include pulsed-field gel electrophoresis (PFGE) (Atyeo et al., 1996, 1999a) and random amplified polymorphic DNA (RAPD) analysis (Dugourd et al., 1996; Jansson et al., 2004). It should be acknowledged, however, that sometimes the results obtained by these techniques can be difficult to interpret or reproduce or both (Maiden et al., 1998), and improved methods are needed. In the case of many Brachyspira species, the molecular epidemiology may be much more complex than is currently appreciated. For example, in swine dysentery, carrier pigs are a major cause of transmission, but dogs, birds, rats and mice are also potential vectors (Songer et al., 1978; Joens, 1980; Hampson et al., 1991; Trott et al., 1996a; Duhamel, 2001; Fellström & Holmgren, 2005; Råsbäck et al., 2007). Such epidemiological connections could be more easily established if simple and improved strain-typing techniques were available for all Brachyspira species.

Multilocus sequence typing (MLST) is a robust, consistent and portable technique that now has largely replaced MLEE for analysis of bacterial population structure, determining relatedness of species, and as a molecular epidemiological tool (Maiden et al., 1998; Urwin & Maiden, 2003). The present study had two main objectives: first, to determine whether MLST could be used to help delineate and define species more widely across the genus, including examination of isolates with atypical characteristics for which the species affiliation is currently uncertain; and second, to investigate whether the system was sufficiently discriminatory for use as a tool for molecular epidemiological studies. MLST was applied to 66 carefully selected Brachyspira isolates and strains, and the results were compared with current classifications based on culture and biochemical tests, 16S rDNA sequences, and strain-typing results from RAPD analysis.

Bacterial isolates and strains.
Forty-six well-described Brachyspira isolates and strains and 20 isolates that were either provisionally named, had atypical phenotypic characteristics, or could not be designated into any described Brachyspira species were included in the study (Table 1). The majority were recovered from pigs, and the selection was based on results from culture and biochemical tests (Fellström & Gunnarsson, 1995; Fellström et al., 1999), in some cases supported by the results of PCR or 16S rDNA sequence analysis (Johansson et al., 2004), or both. The selection of isolates for the study included B. hyodysenteriae and B. pilosicoli isolates with low or high MICs for tiamulin and tylosin. Isolates and strains were obtained from the National Veterinary Institute's strain collection (Uppsala, Sweden), except for the type strains of B. pilosicoli (P43/6/78^T), B. intermedia (PWS/A^T), B. innocens (B256^T) and B. alvinipulli (C1^T), which came from the collection held at the Reference Centre for Intestinal Spirochaetes at Murdoch University. All the type and reference strains used were originally obtained from the ATCC Bacteriology Collection. Supplementary information for some of the isolates is outlined below.

Table 1. Strain and isolate designation, species name at selection and origin of the 66 Brachyspira species used in this study, and their allelic assignment (ST/AAT), sequence type (ST) and amino acid type (AAT) for each locus

The B. hyodysenteriae isolates T20 (A91507-6x/01), T4 [A84193-2x/99 (CCUG 47386)] and A5677/96 were indole-negative (Fellström et al., 1999; Karlsson et al., 2004; Pringle et al., 2004; Råsbäck et al., 2005), P134/99 had an atypical 23S rDNA sequence (Thomson et al., 2001) and E2 was included because it was tiamulin resistant (Karlsson et al., 2004). Eight isolates (A5677/96 and Be45; AN1082/90 and AN3379/98; AN3730/96 and AN613/98; AN360/03 and AN551/03) consisted pairwise of four identical PFGE types (Fellström et al., 1999; Fellström & Holmgren, 2005). Five of the B. pilosicoli isolates (AN738/02, AN953/02, AN991/02, AN1085/02 and AN984/03) were tiamulin resistant and some were also tylosin resistant, of which two (AN1085/02 and AN984/03) originated from the same farm. The six B. intermedia field isolates were all weakly haemolytic, spot indole-positive, hippurate and α-galactosidase-negative, and β-glucosidase-positive. Two of the B. suanatina isolates (AN4859/03 and AN2384/04) were shown to represent one RAPD type (Råsbäck et al., 2007). Biochemical characteristics for the avian isolates described in this study are listed in Table 2. Characterization of these isolates will be further described elsewhere (D. S. Jansson, unpublished data).

Table 2. Characteristics of 10 atypical avian Brachyspira isolates with rare phenotypes used in MLST A comprehensive description of these isolates will be published elsewhere (D. S. Jansson).

16S rDNA sequence analysis.
Twenty-two porcine isolates and 10 avian isolates were subjected to 16S rRNA gene sequencing (Table 1; this study), as previously described (Pettersson et al., 1996; Johansson et al., 2004). The corresponding sequences of the type strains of Borrelia burgdorferi (GenBank accession number X98228), and Treponema denticola (GenBank accession number AF139203), were used as out-group when constructing a phylogenetic tree for all 66 isolates and strains used in this study.

RAPD analysis.
To identify possible clones, two isolates of tiamulin-resistant B. pilosicoli recovered from a single farm, and the eight B. hyodysenteriae isolates previously recognized as four PFGE types were analysed by RAPD. The 26 isolates that showed identical sequence type (ST) or closely related STs or amino acid types (AATs) in MLST analysis (see Figs 2 and 3; isolates with coloured designations) were also analysed by RAPD for comparison of data. Two primers, 5'-ACG CGC CCT-3' (P73) (Quednau et al., 1998) and 5'-CCG CAG CCA A-3' (P1254) (Torriani et al., 1999),. were used separately for RAPD. The PCR was performed under standard conditions in a 50 µl reaction mixture with the Taq DNA polymerase (Biotech International). The PCR programme consisted of four cycles of 94 °C for 45 s, 30 °C for 2 min, and 72 °C for 1 min, followed by 26 cycles of 94 °C for 5 s, 36 °C for 30 s and 72 °C for 30 s. A final extension period of 10 min at 72 °C was included before cooling to 14 °C.

(23K): Fig. 2. Dendrograms of the 66 Brachyspira isolates used in this study based on concatenated (a) DNA sequences, and (b) amino acid sequences obtained from two MLST loci (adh and pgm). Isolates without species designation are either genotypically deviating species or have not been fully characterized to species level. Isolates that were identified as being related by eBURST (Fig. 4) are indicated in red for B. hyodysenteriae, green for B. suanatina, lilac for B. intermedia and blue for B. pilosicoli. Isolates recovered from sources other than pigs are indicated in italics, and field isolates originating from outside Sweden are marked with an asterisk. Bootstrap values are shown for stable nodes in Fig. 2(a). The scale bars show the distance of (a) 20 substitutions and (b) the distance equivalent to two substitutions per 100 amino acid positions, corresponding to approximately 9 substitutions in the sequenced gene fragment.

(13K): Fig. 3. Dendrogram based on concatenated DNA sequences from the 44 Brachyspira isolates with all seven MLST loci successfully sequenced. Isolates indicated in colours are identified as being related by eBURST (Fig. 4), italicized designations indicate isolates from sources other than pigs, and designations marked with an asterisk represents field isolates originating from outside Sweden. Bootstrap values are shown for stable nodes. The length of the scale bar is equivalent to 50 substitutions in the sequenced gene fragment.

Multilocus sequence typing.
Chromosomal DNA was prepared by boiling the bacterial cells, as previously described (Råsbäck et al., 2006). In addition, DNA was prepared and purified by conventional protein K lysis and phenol/chloroform extraction as well as by robot extraction (BioRobot EZ1, EZ1 DNA Tissue kit; Qiagen) for a subset of 10 of the isolates for which all loci could not be amplified by PCR from DNA extracted by boiling. Five of the eight genes tested in the MLST scheme represented genes coding for enzymes of Brachyspira species previously used in MLEE analysis: alcohol dehydrogenase (ADH), alkaline phosphatase (ALP), esterase (EST), glutamate dehydrogenase (GDH) and phosphoglucomutase (PGM) (Lee et al., 1993). MLEE enzymes have previously been shown to be present in all named Brachyspira species as well as in the proposed species B. pulli and B. canis. Further included were three genes: glp [glucose kinase (glpK)], thi [acetyl-CoA acetyltransferase (yqi, also known as thiolase)] and mut [DNA mismatch repair protein (mutS)], used for the MLST systems of Staphylococcus aureus (glpK and yqi) and Streptococcus pyogenes (mutS) (http://pubmlst.org/). The likely existence of single copies for all genes was confirmed by examining the near-complete (∼90 %) genome sequences of the B. hyodysenteriae strain WA1 and B. pilosicoli strain 95/1000 genomes, obtained in a sequencing project conducted at Murdoch University (unpublished data). Primers for MLST were designed by using the Primer3 program (Rozen & Skaletsky, 2000). Previously a small set of Australian isolates had been analysed primarily with primers amplifying the complete genes at the five MLEE loci (unpublished data). These isolates were not used in the current study, but the available sequences for each locus were aligned and primers were designed targeting conserved regions. The annealing temperature of the primers was 45 °C for whole genes, and 50 °C for shorter fragments. Primers used for all isolates are listed in Table 3.

Table 3. Primers used for MLST of Brachyspira species

PCR was performed under standard conditions in a 25 µl reaction mixture generally with Taq DNA polymerase (Biotech International). Each PCR reaction set included a positive control represented by either B. hyodysenteriae B78^T or WA1, or B. pilosicoli P43/6/76^T, and a negative control (double-distilled water). The conditions of the PCR programme were 95 °C for 3 min, followed by 33 cycles at 95 °C for 30 s, 50 °C for 30 s and 72 °C for 1 min if the fragment was shorter than 600 bp, and 1.5 min for longer fragments. A final extension period of 7 min at 72 °C was included before cooling to 14 °C. For those isolates and strains that were not amplified under the above-described conditions, the annealing temperature was lowered gradually by 2–3 °C to an annealing temperature of 43 °C. Pfu DNA polymerase (Promega) was used for isolates that could not be amplified with Taq DNA polymerase. Amplification conditions with Pfu were 96 °C for 2 min and 20 s, followed by 33 cycles of 96 °C for 30 s, 50 °C for 30 s and 72 °C for 3 min. The amplification was ended with an extension step of 72 °C for 5 min before cooling to 14 °C. The annealing temperature was regulated as described above for isolates that could not be amplified. For isolates for which only weak bands were obtained, the number of cycles was increased to 40. The PCR products were purified with the UltraClean PCR Clean-up kit according to the manufacturer's instructions (Mo Bio Laboratories). For weak bands, a QIAquick PCR Purification kit (Qiagen) was used to increase the amount of DNA in the eluate. For cycle sequencing, an annealing temperature of 43–45 °C was used with one-eighth of the amount of Big dye. Sequencing was performed with a 3730 DNA analyser (Applied Biosystems and Hitachi).

The sequences were edited and analysed manually by using NTI Vector 9.0 (). The highest peak was consistently selected for further analysis. Allelic numbers were assigned manually and a different allelic number was given if any nucleotide or amino acid differences were registered. Isolates with the same allelic numbers were assigned to the same ST or AAT. The core sequences for the MLST loci used for the B. hyodysenteriae strain WA1 and B. pilosicoli strain 95/1000 were deposited with GenBank under accession numbers EF488202–EF488215. Sequences of the MLST alleles for each locus were deposited at the PubMLST site at Oxford University (http://pubmlst.org/). Each DNA sequence was translated into amino acid sequences by using Vector NTI 10.0 (). The nucleotide sequences for the seven genes of each isolate were concatenated in the order adh, pgm, est, glp, gdh, thi and alp. All sequences (66 isolates) were placed into a single FASTA formatted file and aligned by using CLUSTAL W [from EMBL-EBI, European Bioinformatics Institute ()]. The file with the aligned sequences was converted to the MEGA format (). A phylogenetic tree for the aligned DNA sequences was constructed by using the number of difference model and the neighbour-joining tree in MEGA version 2.0 (Kumar et al., 2001). For the amino acid trees, the translated DNA sequences were aligned (CLUSTAL W) and a Poisson correction model was used to construct a neighbour-joining tree in MEGA. A population snapshot was obtained by using the program eBURST () by setting the group definition to 0/7 (Feil et al., 2004), assigning a zero for loci without sequence data.

16S rDNA sequence analysis
A phylogenetic tree based on almost complete 16S rDNA sequences of the 66 strains and isolates is shown in Fig. 1. Despite having been repeatedly subcultured, isolate AN652/02 was shown to be mixed as the 16S rDNA sequence contained ten ambiguities. However, none of the nucleotides of the ambiguities could be identified as representing any recognized Brachyspira species other than B. pilosicoli (data not shown).

(21K): Fig. 1. Phylogenetic tree based on almost complete sequences of 16S rDNA from the 66 Brachyspira strains/isolates used in this study. The negative similarity value for AN652/02 indicates that the sequence contains ambiguities. Isolates without species designation are either genotypically deviating species or have not been fully characterized to species level. The scale bar shows the distance equivalent to one substitution per 100 nucleotide positions, corresponding to approximately 14 substitutions in the sequenced gene fragment.

MLST analysis
In MLST, the mut gene was excluded from the analysis due to poor sequence quality and the presence of too many ambiguous positions. All the isolates were successfully sequenced at two loci (adh and pgm), and 44 isolates were successfully sequenced at seven loci, however with a few ambiguities (superimposed peaks in the raw data) in the sequences of the type strains of B. intermedia (PWS/A^T), B. innocens (B256^T) and B. alvinipulli (C1^T), and a field isolate of B. pilosicoli (AN652/02). The different combinations of the seven genes that were successfully sequenced for the remaining isolates are shown in Table 4. It was still not possible to amplify the loci in the 10 isolates examined after extraction and purification of DNA by other methods. The gene-sequence lengths after editing were 492, 783–788, 569–587, 585–588, 793–797, 909–913 and 810–822, for the adh, alp, est, gdh, glp, pgm and thi genes, respectively. For the adh gene, primers for the whole gene were used to obtain sequences for the type strains of B. alvinipulli (C1^T) and B. innocens (B256^T).

Table 4. PCR amplification and successful sequencing for Brachyspira isolates for which fewer than seven MLST loci sequences were obtained The eighth locus, mut, was not used in the MLST analysis.

Deletions/insertions of segments of 2–92 nt were found in the sequenced fragment of the est gene. The deletion/insertion patterns showed intra-species similarities. The alp and glp genes had 7–10 point mutations/insertions/deletions, respectively. The sequences for the glp genes were sometimes of poor quality, mainly at the 3'-end of the fragment. In the translation to amino acids, the field isolate C378 showed a diverging 3'-end of 102 amino acids in the sequence when compared to other B. murdochii isolates. The sequences from the three isolates recovered from corvids (birds of the genus Corvus) (Table 2) showed a 3 nt insertion in the pgm gene. The B. pilosicoli isolates and the B. canis isolate had a 3 nt deletion in the gdh gene. One tiamulin-resistant B. pilosicoli (AN984/03) could not be sequenced with the reverse primer for the adh and est genes, respectively, which resulted in sequenced fragments that were 21 and 20 nt shorter, respectively.

A dendrogram based on concatenated DNA sequences of adh and pgm (approx. 1400 nt) for all 66 isolates is shown in Fig. 2(a), and the equivalent dendrogram based on amino acid sequences (approx. 460 positions) is shown in Fig. 2(b). A dendrogram based on concatenated DNA sequences of the 44 isolates for which all seven MLST loci (up to 4953 nt) were successfully sequenced is shown in Fig. 3. A large distance was observed between the cluster embracing the field isolates of B. murdochii/B. innocens and the B. hyodysenteriae cluster, particularly relative to the differences in their 16S rRNA gene sequences (Fig. 1). Also, the type strains of B. innocens (B256^T) and B. intermedia (PWS/A^T) did not cluster with other members of their respective species in the MLST dendrograms or the 16S rDNA-based phylogenetic tree. For some of the atypical isolates there were discrepancies between their positions in the dendrograms based on 16S rRNA gene sequences and those based on MLST data.

eBURST and RAPD analysis
An allelic assignment for each isolate is presented in Table 1. Allelic frequency ranged from 30 to 46 alleles per locus, with a mean of 38.9. In total, 58 STs were identified. For the amino acid sequences, an allelic range of between 14 and 47 was identified, with a mean of 32.6. The number of AATs identified (58) was the same as the number of STs. Population snapshots obtained by using STs and AATs are shown in Fig. 4. Isolates in the population snapshot that had a close evolutionary relationship originated from the same geographical areas. The results of RAPD analysis were concordant with those of MLST for isolates in the same or in closely related STs or AATs in the eBURST population snapshot, with one exception. Although the four indole-negative isolates of B. hyodysenteriae (T4, T20, Be45 and A5677/96) showed very similar banding patterns, two different patterns could be distinguished. The two German tiamulin-resistant isolates (T4 and T20) could be distinguished from the two tiamulin-susceptible isolates from Belgium and Germany (Be45 and A5677/96). For all other clones identified by eBURST and/or PFGE, the results agreed with those of RAPD analysis.

(11K): Fig. 4. Population snapshot of 66 Brachyspira isolates obtained by using eBURST. The data were derived from the allelic variation in seven MLST loci. Isolates identical in six of seven loci were identified as closely related in evolutionary descent, which is represented with a line between the relevant isolates. Isolate designations are indicated in red for B. hyodysenteriae, in green for B. suanatina, in lilac for B. intermedia and in blue for B. pilosicoli. Isolates recovered from sources other than pigs are indicated in italics, and field isolates originating from outside Sweden are marked with an asterisk (compare with Figs 2 and 3). The thick lines represent results obtained from DNA STs and AATs. Thin lines represent results obtained from AATs only. The positions and distance between the dots are incidental and do not include additional information. Missing sequence data or genes that were too variable limited the identification of closely related isolates; only the 44 isolates with complete sequence data for all loci can be connected if an epidemiological or evolutionary link has been predicted. Dots without isolate designations represent single isolates.

Presently, phylogeny based on 16S rDNA is considered to be the gold standard for the classification of bacteria (Woese, 1987; Ludwig & Klenk, 2001), but this technique is not completely satisfactory for all species in the genus Brachyspira (Pettersson et al., 1996; Stanton et al., 1996). Here we describe an MLST system designed for the entire Brachyspira genus that was developed to help understand relationships between Brachyspira species and strains, as well as to provide an easy and reliable method for identification and classification. Most other published MLST schemes only cover a single species, and this emphasizes that it is difficult to cover an entire genus with the same sets of primers. Similar difficulties have been reported with other MLST systems targeting multiple species (Miller et al., 2005).

The clusters obtained in the MLST dendrograms (Figs 2 and 3) were generally in concordance with the identification of Brachyspira species based on culture and biochemical tests. This points to the potential of MLST as a tool for establishing the species affiliation and differentiation of Brachyspira strains, as previously described for other MLST systems (Diavatopoulos et al., 2005; Ventura et al., 2006). The sequence ambiguities observed could be associated with repeated gene copies, which previously have been suggested to occur in B. hyodysenteriae (Zuerner & Stanton, 1994). Two isolates of B. intermedia were separated from the main B. intermedia cluster, which is consistent with a previous study that indicated that there is considerable overall genetic diversity amongst isolates with the phenotype of B. intermedia (Suriyaarachchi et al., 2000). Some atypical isolates could not be identified to a species level by MLST.

Phylogenetic trees or dendrograms are useful for obtaining an overview of evolutionary relations. However, dendrograms provide almost no information on the evolutionary descent of isolates within a clonal complex (Feil et al., 2004). Therefore, eBURST analysis was used to identify isolates with an epidemiological connection and isolates with a closely related evolutionary history (Fig. 4). The results support the previous conclusion that MLST data can be useful for epidemiological studies (Urwin & Maiden, 2003), and that eBURST analysis gives a more accurate epidemiological identification for some of the isolates than the dendrograms (Figs 1, 2, 3, 4). Limitations of eBURST may arise from too high a degree of allelic variations observed in the DNA sequences. Nevertheless, the relationships identified in this study by eBURST are highly likely to be correct because they were supported by independent epidemiological data. eBURST analysis of MLST data has been claimed to show only a fifth of the evolutionary relationships present within a cluster in a dendrogram (Didelot & Falush, 2006). However, AATs instead of DNA STs might give a clearer representation of closely related isolates, which in turn will result in a more accurate picture of the evolutionary relationships in the bacterial population. In this study, constructing a dendrogram based on amino acid sequences substantially unified the respective major Brachyspira species (Fig. 2b). A close evolutionary connection between mallard and porcine isolates of B. hyodysenteriae and B. suanatina (Fig. 4; thin lines) was identified by eBURST analysis with the same data. The epidemiological relationship between mallard and porcine isolates was strengthened by their similarity in RAPD banding patterns, and the known migration patterns of wild mallards (Fransson & Pettersson, 2001) that enhance opportunities for transmission to farmed pigs.

In conclusion, with few exceptions each of the Brachyspira species clustered separately in the MLST dendrograms, and the majority of the isolates pathogenic to pigs could be delineated and defined. This observation demonstrates the utility of the MLST scheme that was developed, although it could be improved further. A high level of genetic variability was observed amongst members of the genus. The MLST data were also shown to be useful for molecular epidemiological studies, and in particular the eBURST analysis was shown to be more discriminatory than the use of dendrograms for determining relationships. The use of amino acid sequence data revealed evolutionary connections between isolates within the same geographical area and between isolates from pigs and mallards.

This study was a joint project conducted between the Swedish University of Agricultural Sciences (SLU), Uppsala, Sweden and Murdoch University, Perth, Australia. Internationalization grants from SLU allowed the recipient T. R. to undertake the MLST analysis at Murdoch University. M. Y. A. participated as a Visiting Research Associate at Murdoch University, supported by a scholarship from the Ministry of Health and Medical Education, the Islamic Republic of Iran. We thank Bjarne Bergsjø (NVI, Oslo, Norway), Josef Hommez (Regional Veterinary Investigation Centre, Torhout, Belgium), Branko Kokotovic (DFVF, Copenhagen, Denmark), Kevin Perry (VLA, Winchester, UK), Jill Thomson (SAC, Edinburgh, UK), and Judith Rohde (Institute für Mikrobiologie und Tierseuchen, Hannover, Germany) for supplying isolates. We further acknowledge Novartis Animal Vaccines, Murdoch University and the Swedish Farmers' Foundation for Agricultural Research (SLF) for funding, and Professor Matthew Bellgard (CCG, Murdoch University) for helpful discussion.

Edited by: A. Fouet