A combinational approach of multilocus sequence typing and other molecular typing methods in unravelling the epidemiology of Erysipelothrix rhusiopathiae strains from poultry and mammals

Erysipelothrix rhusiopathiae infections re-emerged as a matter of great concern particularly in the poultry industry. In contrast to porcine isolates, molecular epidemiological traits of avian E. rhusiopathiae isolates are less well known. Thus, we aimed to (i) develop a multilocus sequence typing (MLST) scheme for E. rhusiopathiae, (ii) study the congruence of strain grouping based on pulsed-field gel electrophoresis (PFGE) and MLST, (iii) determine the diversity of the dominant immunogenic protein SpaA, and (iv) examine the distribution of genes putatively linked with virulence among field isolates from poultry (120), swine (24) and other hosts (21), including humans (3). Using seven housekeeping genes for MLST analysis we determined 72 sequence types (STs) among 165 isolates. This indicated an overall high diversity, though 34.5% of all isolates belonged to a single predominant ST-complex, STC9, which grouped strains from birds and mammals, including humans, together. PFGE revealed 58 different clusters and congruence with the sequence-based MLST-method was not common. Based on polymorphisms in the N-terminal hyper-variable region of SpaA the isolates were classified into five groups, which followed the phylogenetic background of the strains. More than 90% of the isolates harboured all 16 putative virulence genes tested and only intI, encoding an internalin-like protein, showed infrequent distribution. MLST data determined E. rhusiopathiae as weakly clonal species with limited host specificity. A common evolutionary origin of isolates as well as shared SpaA variants and virulence genotypes obtained from avian and mammalian hosts indicates common reservoirs, pathogenic pathways and immunogenic properties of the pathogen. Electronic supplementary material The online version of this article (doi:10.1186/s13567-015-0216-x) contains supplementary material, which is available to authorized users.

E. rhusiopathiae was first recognized as a human pathogen causing localized cutaneous lesions, called erysipeloid, and sporadic cases of generalized cutaneous forms or septicaemia, often associated with endocarditis [2,5]. Erysipelas in swine occurs in different forms characterized by septicaemia often resulting in sudden death (acute form), cutaneous lesions (sub-acute form), polyarthritis and endocarditis (chronic form) [1,2,6]. In recent years, Erysipelas in poultry re-emerged and here, turkeys are most seriously affected and suffer from cyanotic skin to haemorrhages and petechiae in the breast and leg muscles. Following the change from conventional (battery) cage system to alternative housing systems, problems due to E. rhusiopathiae also increased in laying hen [2,[6][7][8][9][10][11][12].
While attenuated live vaccines and bacterins are commercially available to protect pigs and sheep, in Germany there is no licensed vaccine to prevent erysipelas in turkeys and laying hen and this is why the application of autogenous vaccines and/or antimicrobial therapy is the ultimate way to combat the disease [1,2,9]. One of the most promising vaccine candidates is SpaA, a surface protein with high immunogenic properties [13][14][15][16]. Based on amino acid sequence similarities the Spa proteins can be classified into 3 molecular variants, termed SpaA, SpaB, and SpaC and the occurrence of these variants has been found to be basically associated with the serotype of a strain [14]. Particularly the N-terminal half of the hypervariable region of the Spa protein is important for specific immunity, and strains with different spaA gene variants obviously differ in pathogenicity [17].
E. rhusiopathiae strains vary considerably in virulence but only little is known about pathogenic mechanisms and the presence and distribution of virulence factors [2]. Different factors, including neuraminidase, capsule, hemolysin, hyaluronidase, adhesins and the cell wall lipoprotein EwlA have been suggested to be involved in the pathogenesis of the disease [2]. The availability of the genome sequence of E. rhusiopathiae strain Fujisawa (Acc.-No. AP012027.1) offers the opportunity for the in silico-identification of additional genes that have been linked with virulence in other bacterial species before. The coding sequence ERH_1472 was for example annotated as internalin gene intI, which in case of Listeria monocytogenes contributes to the invasion of the bacteria into epithelial cells [18].
Various studies have been performed to determine the clonal relatedness of E. rhusiopathiae strains [19][20][21][22][23][24][25]. Data from multilocus enzyme electrophoresis (MEE) [20], restriction fragment length polymorphism (RFLP) analyses [19], amplified fragment length polymorphism (AFLP) [21], randomly amplified polymorphic DNA methods (RAPD) [23,25], and pulsed-field gel electrophoresis (PFGE) [22,24] suggested that E. rhusiopathiae strains might be genetically diverse even among isolates belonging to the same serotype. Although the relevance of Erysipelas in poultry increased over the last years, intense epidemiological studies on E. rhusiopathiae from poultry combining established typing methods such as PFGE, Spa typing and virulence gene typing are limited [10,22] Moreover, there is currently no established scheme to study this bacterial species by Multilocus sequence typing (MLST), a widely applied method to determine the phylogeny of a bacterial population based on sequence analysis of representative genes of the bacterial core genome [26].
The overall objective of this study was to provide a deeper insight into the population structure of E. rhusiopathiae by developing an MLST scheme and determining sequence types of 165 E. rhusiopathiae strains predominantly isolated from poultry and also from mammalian hosts, including few human strains. The congruence of multilocus sequence types of individual strains with their grouping into PFGE clusters and their surface protective antigen (Spa) protein variant was evaluated and the distribution of 16 genes putatively linked with the pathogenic properties of the strains was investigated.

Bacterial strains and species designation
A total of 165 E. rhusiopathiae strains primarily from Germany (136) and from a few other countries (29) were included. The major focus was on isolates from avian sources (chicken, 71; turkey, 43; other birds, 6), while isolates from pigs (n = 36) and other animal species (n = 6) and humans (3) were included as well. Bacterial strains were isolated from organ and blood samples from diseased animals and from blood samples in case of the human strains. In case of outbreak situations, only one strain per flock was sampled. Strains obtained from recurrent disease events in identical flocks were only included if six months had passed since the previous event or if they revealed different banding patterns, as initially determined by pulsed-field gel electrophoresis (data not shown).
E. rhusiopathiae strains and the non-pathogenic species E. tonsillarum were differentiated by a species-specific PCR [27]. Reference strains E. rhusiopathiae ATCC 19414 T , E. tonsillarum DSM 14972 T (ATCC 43339 T ), and E. inopinata DSM 15511 T as well as two field strains of E. tonsillarum were also included. The reference strains were purchased from the Leibniz Institute DSMZ-German Collection of Microorganisms and Cell Cultures (Braunschweig, Germany). All strains were stored at −80°C in liquid medium (brain heart infusion broth) supplemented with 20% glycerol until used.

DNA purification
The total DNA of E. rhusiopathiae was purified using a protocol described previously [28]. Instead of using a 1.5 mL bacterial culture grown in BHI for 24 h, as recommended in the protocol, we took 20 mL of an overnight culture to obtain a higher amount of DNA.

Genes for MLST and nucleotide sequencing of gene fragments
To establish an MLST scheme, genes were selected based on published protein sequences of E. rhusiopathiae, located in the published contigs of the reference strain E. rhusiopathiae ATCC 19414 T (project ID: 38421), sequenced by the National Institute of Animal Health, Japan, in cooperation with the Dragon Genomics Center, Takara Bio Inc., Japan. Based on reversed translated protein sequences, primers were designed to amplify gene fragments of housekeeping genes gpsA (glycerol-3-phophate-dehydrogenase), recA (recombinase A), purA (adenylosuccinate synthetase), pta (phosphate acetyl-transferase), prsA (ribose-phosphatepyrophosphokinase), galK (galactokinase), and ldhA (D-lactate dehydrogenase) ( Table 1). Genes were chosen in line with MLST schemes for other bacteria taking into account (i) a gene size > 700 bp, (ii) a regular distribution of genes among the E. rhusiopathiae genome, and (iii) that predicted proteins should be under stabilizing selection.
PCRs were performed in a volume of 60 μL containing 60 ng DNA template, 10 pmol of each primer (Sigma Alderich, Munich, Germany), 10 mM dNTP mixture, 10 × PCR buffer including 20 mM MgCl 2 , and 1 U Taq polymerase (all Rapidozym, Berlin, Germany). The PCR conditions were initial denaturation at 95°C for 3 min followed by 30 cycles of denaturation at 95°C for 30 s, annealing at 55°C for 30 s (except for galK, for which an annealing temperature of 54°C was used), and extension at 72°C for 1 min, with a final extension at 72°C for 10 min. Amplicons were sequenced by LGC genomics (Berlin, Germany). Sequence analysis was performed with Ridom SeqSphere 1.0.1 [29].
Each unique sequence of a gene fragment was assigned an allele number. These allele numbers were consecutively numbered and combined in order to establish an allelic profile, which defined the sequence type (ST). Using the eBURST program [30] the MLST data set was divided into groups of related STs (clonal complexes resp. ST complexes). We used the most stringent definition of ST complexes, which were defined as groups when six out of seven alleles from the nearest neighbour were the same [31].
The program START version 2 [32], was used to determine the number of variable nucleotide sites and their effects on the amino acid sequence changes. This was done by calculating the ratio of non-synonymous substitutions (d N ) to the number of synonymous substitutions (d S ) (d N /d S ratio) [33]. The d N /d S ratio quantifies the evolutionary pressure on proteins [34] and thus indicates the presence or absence of a selective force on the locus [35].
Simpson's index of Diversity (D) was calculated on the basis of the molecular pattern of the seven loci [36][37][38]. A Simpson's index of diversity with 95% confidence intervals (CI 95% ) was calculated for a set of 165 strains. A value close to 1 indicates high diversity, and a value close to 0 reflects little diversity.

Pulsed-field gel electrophoresis
Pulsed-field gel electrophoresis (PFGE) followed a previously published protocol [24] and was performed to group E. rhusiopathiae isolates according to their macrorestriction profiles. PFGE was also applied to discuss the discriminatory power and interpretative value of this band-based method compared with the sequence-based MLST analysis. The chromosomal DNA was digested with SmaI (Fermentas Life Science, Germany). Electrophoresis was carried out in a contour-clamped homogeneous electric field (CHEF-DRIII; Biorad, Munich, Germany). MidRange PFG Marker I (Bio-Labs, #N3551S) was used as DNA size standard. Band patterns were analysed with the software BioNumerics (Version 6.6, Applied Maths, Belgium). Dendrograms were created by an unweighted pair group matching by an arithmetic averages algorithm (UPGMA) and the Dice coefficient. Optimization was set at 0.9%, band position tolerance at 2.0% [39]. To compare the discriminatory power of MLST and PFGE the Simpson's index of Diversity (D) was used [36][37][38]. An index of diversity close to 1 indicated a high diversity of the strain collection, whereas a value close to 0 suggested little diversity. We calculated the Simpson's index of diversity (D) with 95% confidence intervals (CI 95% ).

Detection of virulence-associated genes by Multiplex PCR
Two multiplex PCR protocols were established to screen for genes putatively linked with the virulence of E. rhusiopathiae. The selection of the genes was based on (i) previous studies, linking the gene products with a certain pathogenic mechanism [2,[40][41][42][43][44] and (ii) publicly available annotation data of coding sequences of E. rhusiopathiae (Acc.-No. AP012027.1), determined by whole genome analysis [45] (see Table 2). The sequences of oligonucleotide primers, either obtained from recent publications [14,46] or designed in the present study are given in Table 2. PCRs were performed in a T300 Thermocycler (Biometra, Göttingen, Germany). A 25 μL aliquot con-  Primers were designed in this study except for the primers to amplify the cap locus [46]. a In case Ogawa et al. [45] is given as a reference, the predicted function of the gene product has been deduced from genome annotation data of strain Fujisawa. Here, in vitro or in vivo studies would be mandatory to verify the linkage of the genes with the pathogenesis of Erysipelas in different hosts.
conditions: an initial denaturation step at 94°C for 10 min followed by 30 cycles of denaturation at 94°C for 1 min, annealing at 53°C for 2 min, extension at 72°C for 7 min, and finally one cycle of extension at 72°C for 10 min. PCR products were separated by gel electrophoresis in a 1.5% agarose gel containing ethidiumbromid for 90 min at 120 V and analysed by gel documentation system (Easy doc, Herolab, Wiesloch, Germany).

Sequencing of spaA genes
In order to distinguish the three different molecular variants of the Spa protein [14] we performed restriction length polymorphism (RFLP) analysis and single locus

Multilocus sequence typing
Allele sizes for the genes included for MLST analysis of 165 E. rhusiopathiae strains ranged between 531 bp for ldhA to 640 bp for recA ( Table 1). The mean GC content varied from 36.6% (pta) to 40.3% (ldhA). Among the 165 investigated isolates with the corresponding 3.995 bp of the concatenated sequences of seven loci, 75 (1.9%) variable sites were confirmed. The number of alleles ranged from six for prsA to 14 for purA. The nucleotide diversity within the single loci varied between four (0.7%) for prsA and 14 (2.6%) for ldhA, the diversity index (D) for each allele between 0.579 (pta) and 0.775 (recA). The d N /d S ratio for all seven loci demonstrated to be < 1 ( Table 1). The combination of allele numbers and the corresponding sequence types are shown in Table 3. A list of all strains with their sequence types is available as additional file (see Additional file 1). Of 72 sequence types (STs) assigned to 165 E. rhusiopathiae isolates, more than half (58.3%) were singletons. The remaining 29 STs were represented by two or more strains. ST9 was by far the most common sequence type, which grouped 18 isolates (10.9%) from poultry (n = 12), pigs (n = 3), sheep (n = 1), and also from a case of wound infection and a case of endocarditis in humans together. Sequence type 4 (n = 11 poultry isolates) and ST5 (n = 10 poultry and one pig isolate) were the second most frequent STs. Overall, eBURST analysis based on all allelic profiles revealed a highly divergent population structure among the E. rhusiopathiae strains under study (Figure 1).
Only one sequence type complex, i.e. a group of phylogenetic related strains was identified. ST complex 9 (STC9), named after its predicted founder, included 22 STs and accounted for 34.5% of the total collection of MLST-analysed strains. As depicted in Figure 2A, 40 avian and 17 mammalian E. rhusiopathiae strains were among the STC9 strains. In general, no host-specific or host-restricted sequence type was observed, as isolates from different hosts scattered around the various STs identified (Figure 2A). In addition, different countries shared isolates with identical STs, e.g. ST3 (isolates from Austria, Germany, Sweden), ST9 (Austria, Germany, USA), and ST19 (Estonia, Germany, USA), not suggesting any association between certain STs with distinct countries. However, due to a sample bias towards isolates from Germany, this should be verified by including a larger number of isolates from various countries.

Pulsed-field gel electrophoresis and comparison of sequence types with PFGE clusters
Based on an 85% cut-off-value we identified 54 different PFGE clusters containing one (n = 17 singletons) to 13 strains among 165 E. rhusiopathiae isolates (Additional file 2). Identical macrorestriction profiles were only rarely observed which is consistent with our criteria for strain selection, i.e. exclusion of copy strains from the analysis. In contrast to PFGE, MLST revealed 72 different sequence types and, as expected, the grouping achieved by both methods revealed only partial congruence. The most homogenous macrorestriction profile was observed for strains assigned to ST4. All eleven ST4 strains, isolated from 1999 to 2011 from chickens and turkeys in Germany clustered exclusively together at a similarity value of 90.9% ( Figure 3A). In contrast, strains of the most prominent sequence type ST9 were distributed among 12 different PFGE clusters, including three singletons, with an overall similarity of 64.5% ( Figure 3B). These PFGE clusters frequently harboured strains of other sequence types as well, which is also the case for most of the other non-singleton PFGE clusters.
A calculation of the Simpsons index of discriminatory (D) with 95% confidence interval (CI 95% ) revealed that the ability of MLST to discriminate E. rhusiopathiae isolates (D = 0.973; CL 95% [0.963 to 0.983]) was similar to that of PFGE (D = 0.976; CL 95% [0.970 to 0.982]). However, as exemplified by the dendrograms shown in Figure 3 and in the Additional file 2, strain groups established by the band-based method PFGE and the sequence-based method MLST frequently differ in their composition.

Virulence gene typing
Highly homogenous patterns of putative virulence genes were observed among the 165 E. rhusiopathiae isolates with 136 (82.4%) of them possessing all 16 investigated genes ( Table 2). Another 24 (14.5%) isolates differed only by the absence of internalin gene intI. Five isolates lacked either algI (n = 4), or dnaB (n = 1), which might be associated with colonization and resistance to phagocytosis and with bacterial proliferation and adhesion, respectively [45]. None of the 16 genes investigated were present in the E. inopinata reference strain DSM 15511 T , whereas nanH and ewlA were detected in the E. tonsillarum strain DSM 14972 T as well as in both field strains of E. tonsillarum.   Table 4). Sequence types (STs) are indicated by numbers and sequence type complexes are highlighted by grey shade.  IMT17855  IMT9209  IMT17170  IMT20382  IMT23647  IMT22111  IMT17113  IMT6254  IMT24173  ATCC 19414  IMT11439  IMT9137  IMT27630  IMT11437  IMT8949  IMT11442  IMT22109  IMT23384  IMT27632  IMT28355  IMT17122  IMT17123  IMT17137  IMT18128  IMT17133  IMT22110  IMT17146  IMT11438  IMT8704  IMT9827  IMT17117  IMT8948  IMT18360  IMT17168  IMT20385  IMT23641  IMT23643  IMT23646   IMT20384   IMT19234   IMT20390  IMT17135  IMT17136  IMT17134 9  9  3  49  49  9  26  26  8  9  37  9  37  3  3  3  9  54  7  7  9  9  42  9  9  8  45  9  10  10  10  10  9  9  19  19  19  19  I  I  III  III  III  I  I  I  I  I  I  I  I  III  III  III  I  I  I  I  I  I  I  I  I  I  I  I  I  I  I  I  I  I  I  I  I  I   I   I   I  I  I  I  I  --- IMT18129  IMT23984  IMT28565  IMT17142  IMT17143  IMT6251  IMT17148  IMT6236  IMT17164  IMT18130  IMT17140   Chicken  Chicken  Turkey  Turkey  Turkey  Chicken  Turkey  Chicken  Turkey  Chicken  Turkey   G  G  G  G  G  G  G  G  G  G  G   2008  2010  2011  2007  2007  1999  2007  2002  2007  2008  2007   II  II  II  II  II  II  II  II  II  II  II   10  Sequences of the N-terminal hypervariable region of the spaA gene of field isolates and of E. rhusiopathiae serotype reference strains were compared with that of the virulent wild-type reference Fujisawa serotype 1a strain (see Table 4). Based on patterns of amino acid changes at eight different positions, the field isolates were divided into five groups as follows: (i) 87 isolates with isoleucine at position 55 (Ile-55), Asn-70, Asp-178, Asn-195, Ile-257 and Gln-303 classified as group I, (ii) 36 isolates with Ser-101 and Ile-257 classified as group II, (iii) 37 isolates with Ile-257 termed group III, (iv) one isolate with Ile-257 and Gln-303 classified as group IV, and (v) 4 isolates with Met-203 and Ile-257 termed group V (group I-V as specified in Table 4).
Except for serotype 2 reference strain ATCC19414 T , which revealed a unique pattern (Ile-55, Asn-70 and Ile-257) all other reference strains included (see Table 4) corresponded to one of the SpaA groups observed among the field isolates. None of the SpaA groups correlated with a certain host type (mixed avian and mammalian). As far as serotypes were available (n = 32; serotype 1a and 1b (n = 20); serotype 2 (n = 9), serotype N (n = 3), they also did not correlate with certain SpaA groups. In group I, we observed both serotype 1a/b and serotype 2 strains and within group III, strains of serotypes 1a/b, 2 and N were present (data not shown).
The number of C-terminal tandem repeats of similar 20-amino-acid sequences in the SpaA proteins of field isolates ranged from 7 to 13, corresponding to the different spaA nucleotide sequence lengths observed. The majority of strains (89.7%) harboured a domain with 9 repeats similar to that present in strain Fujisawa and in all SpaA-positive serotype reference strains, except for serotype 9 strain Kaparek ( Table 4). The remaining field isolates showed either higher or lower number of repeats. One strain from a chicken with septicaemia was truncated by 40 amino acids at positions 502 and 534 and another seven strains lacked a 20 amino acids repeat either at position 534 or 558. Seven strains revealed 10 repeats and two porcine strains even showed 11 and 13 repetitions of similar 20-amino-acid sequences in the C-terminal part of their SpaA protein, respectively. A host-specific distribution of SpaA types was not observed. Portraying the different SpaA alleles against the phylogenetic background of the strains reveals a strong linkage of group I (amino acid changes Ile-55, Asn-70, Asp-178, Asn-195, Ile-257 and Gln-303 compared with Fujisawa strain) with ST complex 9 ( Figure 2B). Group II (Ser-101, Ile-257) and group III isolates (Ile-257) occur in several STs which are not related to this major ST complex. Only in one case, different SpaA types occur in the same sequence type, namely ST48, which is made up by porcine isolates (Figure 2A).

Nucleotide sequence accession numbers
Nucleotide sequence of spaA genes of 165 E. rhusiopathiae strains have been submitted to the GenBank database under accession nos. KR606141 -KR606305.

Discussion
Although there is an increasing notion of severe outbreaks of Erysipelas in poultry [10,22,47], possibly related to the ban of conventional cages within EU from 2012 and the large changes of housing systems due to welfare demands for laying hens [9,10,12], only little is known about the molecular epidemiology of avian E. rhusiopathiae isolates. To overcome this, we investigated isolates of E. rhusiopathiae, mainly from poultry but also from mammalian hosts by a newly developed speciesspecific MLST scheme and in addition by more established methods such as PFGE, virulence genotyping and Spa typing.
The novel MLST scheme for E. rhusiopathiae is based on the identification of nucleotide polymorphisms of internal fragments from seven housekeeping genes involved in metabolic pathways of the bacterial pathogen. The target genes were chosen with a d N /d S ratio of <1 to guarantee a very low degree of evolutionary pressure on the proteins and absence of a selective force on the respective gene locus. This ratio was calculated for other MLST schemes, recently developed e.g. for the bovine pathogen Mannheimia haemolytica [48] or for Enterococcus faecalis, which has been associated with amyloid arthropathy in chickens [49] as well. The lower number of alleles in the E. rhusiopathiae MLST scheme, which varies between 6 (pta) and 14 (purA) among 165 E. rhusiopathiae strains together with a diversity of index D of 0.69 (95% CI, 0.579 to 0.775), which is also lower than that observed for other schemes [48][49][50][51] might indicate a lower diversity as compared e.g. to the Gram negative pathogen E. coli or to the Gram positive species Streptococcus gallolyticus. However, this assumption should be treated cautiously as the MLST databases of these two species contain significantly more isolates, i.e. 7.500 isolates in case of Escherichia coli [52] and 292 isolates in case of Sc. gallolyticus [53]. Whether E. rhusiopathiae  would still appear as a bacterial species with low diversity after including more isolates from different sources remains to be determined. The distribution of the STs as presented in a minimum spanning tree (MST) shows accumulations of E. rhusiopathiae strains isolated from various animals in certain STs (Figure 2A). Even if there was a bias towards avian isolates and towards isolates from Germany in our sample material, there was no indication for host or geographic specificity. Strains from poultry and swine randomly scattered throughout the MSTree and commonly shared identical STs, such as ST9, ST8 and ST5. This might indicate that birds and pigs may be infected by phylogenetically similar strains either by a common reservoir or by direct or indirect transmission from one animal species to the other. A limited host specificity has also been suggested in another recent study, where marine E. rhusiopathiae isolates were shown to be capable of inducing classical skin lesions in pigs in a skin inoculation model [54]. However, looking deeper into the clonal nature of our strains by PFGE, those from poultry and pigs were commonly found in same PFGE clusters, but no identical pairs of isolates were observed between the two species. This supports a previous finding by Eriksson et al. who also didn't observe a 100% band profile identity among any of the pig and poultry isolates albeit the strains were not separated into two distinct populations by PFGE analysis [22].
Notably, the most prominent ST observed in the present study -ST9 -includes human and animal strains. The grouping of porcine strain ATCC 19414 T and human strain IMT9137 in an identical sequence type and in the same PFGE cluster (89.9% similarity in banding pattern) indicates that these strains may be closely related and that a zoonotic potential of animals strains should be taken into consideration. Sporadic diseases in humans have been reported and these often occurred in farmers and in other persons whose work is closely related with contaminated animals, underlining that E. rhusiopathiae infection in humans is occupationallyrelated [2,55,56].
Overall, our MLST analysis distinguished 165 isolates into 72 sequence types (STs) and only partial congruence was observed to the grouping of strains into 58 PFGE clusters. MLST intends to portray the long term epidemiology and the evolutionary history of a bacterial pathogen based on a representative set of the bacterial core genome rather than to unravel disease outbreaks, at least not without additional typing methods [26]. In contrast, PFGE is a suitable tool to monitor short-term local outbreaks mainly by identifying clones [57].
In studies comparing MLST to PFGE e.g. among Enterococcus faecalis, Staphylococcus aureus, and Vibrio cholerae isolates, MLST has been found to have similar or greater discriminatory ability than PFGE [58][59][60][61]. In contrast, a study comparing MLST and PFGE data in Escherichia coli O157:H7, found that PFGE revealed a greater discriminatory ability than MLST did, which may be due to the highly clonal nature of this E. coli serotype [62]. In case of the Gram-negative pathogen Pseudomonas aeruginosa Johnson et al. found that MLST was better for detecting genetic relatedness, while PFGE was more discriminatory than MLST for determining genetic differences in this bacterial species and to confirm the a patientto-patient transmission [63]. Nemoy et al. observed a greater discriminatory ability and reproducibility of MLST than PFGE for ESBL-producing E. coli [58]. They suggested MLST is suitable to a priori define genetically related bacterial strains, which might be very helpful for surveillance.
Our MLST data for the first time allow determining E. rhusiopathiae as a weakly clonal species as 165 isolates are dispersed among 72 different sequence types. The accumulation of a number of strains in ST complex 9 may indicate that this group of genetically related strains represents a particularly successful subpopulation. If this holds true after including more strains from different countries is uncertain. Likewise, possible reasons for this success or at least high prevalence of STC9 strains (e.g. high virulence, environmental survival) also need to be studied.
Interestingly, although serotypes for only 32 of our isolates were available (serotype 1a and 1b (n = 20); serotype 2 (n = 9); serotype N (n = 3)), these serotypes were not strictly linked with a certain sequence type or ST complex (data not shown). In addition, different serotypes were observed in identical STs, indicating that the O antigen encoding genetic structure may have, at least partly, evolved independently from the phylogenetic background of the strains, which has likewise been reported for other bacterial species, such as E. coli and Bordetella spp. [64,65]. No correlation between serotype and PFGE pattern could also be observed in a recent study including E. rhusiopathiae isolates from poultry, pigs, emus, and the poultry red mite Dermanyssus gallinae [22]. Here, the authors found that serotypes were randomly scattered throughout the dendrogram based on PFGE patterns and that isolates with the same PFGE pattern were of different serotypes in some cases. They concluded that serotyping may be less suitable for studies on epidemiological relationships between flocks.
Concerning the putative virulence genes investigated in our study, there was no congruence between one of these genes with a certain sequence type or PFGE cluster, which is primarily attributable to the frequent detection of the genes. Almost all 16 markers selected and putatively involved in pathogenesis, except for intI (85.5%), the Alginate-O-acetyltransferase gene algI (97.6%), and the attachment protein gene dnaB (99.4%) were regularly present in our isolates. According to the annotation of whole genome sequenced E. rhusiopathiae strains Fujisawa (GenBank AP012027) and SY1027 (Gen-Bank CP005079) intI encodes an internalin-like protein.
We observed 15% intI-negative isolates that occurred in 12 different STs, mostly from poultry and two isolates from sheep. But, as the function of this internalin-like protein has been deduced from amino acid sequence comparisons only, further molecular genetic studies and in vitro assays would have to be performed to verify its putative role in E. rhusiopathiae pathogenesis. Although most of the other genes were regularly present in our isolates, their expression might vary due to different body compartments or environmental factors. For some of these factors, such as the neuraminidase, a correlation between the produced amount of this enzyme, which may serve the nutritional requirements and play a role in adherence and antiphagocytic activity of the bacteria, and the virulence of strains has been reported [2,66]. It was not unexpected to detect the capsule biosynthesis locus in all our isolates. In previous studies capsule-negative mutants or the loss of capsule in general resulted in avirulent strains which did not resist phagocytosis by murine polymorphonuclear leukocytes and lost the capability of intracellular survival [2,40]. Thus, such strains would not have been able to cause severe disease in the various hosts.
This study was mainly focussed on poultry isolates, as we intended to provide a rational typing method to assess the clonal relatedness of these strains, particularly with respect to the selection of epidemiological relevant strains to be included in future vaccines. In this respect we also aimed to determine, whether the N-terminal hyper-variable region of the surface protein SpaA, which apart from SpaC, represents one of the most promising vaccine candidates [13][14][15] might show a co-evolution with the genomic backbone of E. rhusiopathiae. The exclusive finding of SpaA and the absence of SpaB and SpaC in our sample collection might be due the prevalence of certain serotypes among our strains. Our partial serotype data confirm what has been reported in previous studies, namely that the spaA gene usually, but not exclusively, occurs in strains belonging to serotype 1a, 1b, 2,5,8,9,12,15,16,17 and type N, while spaB and spaC are mainly observed in E. rhusiopathiae serotypes 4, 6, 11, 19, and 21 and in serotype 18, respectively [3,13,14,17]. Ingebritson et al. reported that one strain may contain more than one Spa type and that these were not always confined to a certain serotype [3]. It is not likely that single isolates of our study harboured different Spa types as this would have become apparent due to ambiguous sequencing data.
Different authors have explored the variations in the N-terminal half of the spaA gene, particularly in the protective domain which is located at amino acid positions 30 to 413, and in the repeat domain which is located at amino acid positions 448 to 626 (in case of 8 to 10 tandem repeats) [14,67,68]. A broad collection of avian isolates and their spa gene has not been investigated so far, limiting our knowledge about the diversity of this major antigen considerably. By sequence analysis of the spaA gene of 165 clinical E. rhusiopathiae isolates, thereof 120 from birds, we found high sequence identity within the signal sequence region, located at the most N-terminal part of the Spa protein, corroborating the findings of To and Nagai and supporting the idea that Spa proteins may cross bacterial cell membranes and eventually be secreted from the bacterial cell by using a secretion machinery commonly present in E. rhusiopathiae strains [14]. Although most of our isolates (89.7%) revealed nine Cterminal tandem repeats in their SpaA protein, which is also common to most of the serotype reference strains investigated by To and Nagai and listed in Table 4 [14], the diversity in number was quite high with 7 to 13 tandem repeats and has not been reported before. As this repeat domain has been suggested to function as an anchor for binding Spa proteins tightly to the bacterial surface it remains to be determined, e.g. by in vitro adhesion experiments, if this binding might perhaps be influenced by the number of repeats [13,14,68]. Regarding the amino acid substitutions in the N-terminal protective part of the Spa protein we detected five different variants (group I-V) that occurred independently of the original host species and have been described at least in part in previous publications on porcine isolates as well. Among swine isolates from Japan, Uchiyama et al. observed three groups, looking at substitutions at amino acid positions 195, 203 and 257 [17]. These positions were also part of our investigations, while we additionally included another 6 amino acid substitutions resulting in a finer resolution of the groups. Thus, their group 2 resembles our groups II-IV; group 1 matches with our group V; and group 3 was not detected among our isolates at all. In that study, more than half of the 34 porcine strains showed the Met-203 variant in combination with Ile-257 (group 1 according to Uchiyama et al. [17]), corresponding to group 2 recently determined among another collection of isolates from Japan [69]. To et al. [68] found that 95% of their overall 80 pig isolates carried the Met-203 variant of the spaA, while the same variant accounted for 55.6% of the isolates included by Uchiyama et al. [17]. Notably, we observed this variant only very infrequently (2.4%) among our isolates. In the present strain collection we rather detected SpaA variants (group I, 52.7%; group II, 21.8%; group III, 22.4%) that have been described for different serotype reference strains before (Table 4) [14]. Of note, the distribution of spaA variants demonstrated to be irrespective of the original host of the strains, i.e. avian or mammalian. Whether the current geographic linkage of the Met-203 variant to Japan remains after studying more isolates from different