rDNA sequence analyses of Streptococcus dysgalactiae subsp. equisimilis isolates from pigs

The species Streptococcus dysgalactiae is known to include a variety of strains that are biochemically and genetically very close but differ in serological grouping, type of haemolysis, host association and pathogenicity (Jones, 1980; Farrow & Collins, 1984; Kilpper-Bälz & Schleifer, 1984; Hommez et al., 1991; Efstratiou et al., 1994; Bert et al., 1997). Vandamme et al. (1996) showed that these bacteria could be divided into two subspecies (S. dysgalactiae subsp. dysgalactiae and S. dysgalactiae subsp. equisimilis) based on their whole-cell protein profiles and biochemical properties. Recently, Vieira et al. (1998) proposed a new classification for the organisms based on multilocus enzyme electrophoresis typing and genomic DNA relatedness. These analyses defined α-haemolytic or non-haemolytic strains of Lancefield group C as S. dysgalactiae subsp. dysgalactiae and β-haemolytic strains of group C, G or L antigen as S. dysgalactiae subsp. equisimilis. This is the traditional and most widely used method of classification to date. Furthermore, genealogical divergences between S. dysgalactiae isolates have also been detected by means of multilocus enzyme electrophoresis (Bert et al., 1995), random amplified polymorphic DNA (RAPD) (Bert et al., 1996) and 16S23S rRNA gene (rDNA) intergenic spacer analyses (Chanter et al., 1997).

The phylogenetic position of the species S. dysgalactiae within the genus Streptococcus has been estimated by sequencing analyses of 16S rRNA (Bentley et al., 1991; Kawamura et al., 1995a, b). However, subspecific information for sequences of S. dysgalactiae strains is limited and, therefore, the details of the phylogenetic relationship among strains of the organism remain unclear.

It is evident that the resolving power of 16S rDNA sequences is limited when closely related organisms are examined (Rogall et al., 1990; Amann et al., 1992; Fox et al., 1992), although analysis based on such sequences is a valuable phylogenetic marker, as is DNA reassociation (Stackebrandt & Goebel, 1994). On the other hand, 23S rRNA is about twice as long as 16S rRNA and contains more variable regions. Therefore, 23S rRNA is considered to be a more informative phylogenetic marker (Ludwig & Schleifer, 1999).

S. dysgalactiae subsp. equisimilis is classified into four subgroups of host-associated ecovars based on serogrouping and biotyping: human groups C and G, animal group C and group L (Devriese, 1991). Bacteria of the latter two groups are frequently isolated from pigs, in which the organisms cause septicaemia, arthritis or valvular endocarditis (Hommez et al., 1991). In this study, we used porcine S. dysgalactiae subsp. equisimilis isolates associated with arthritis, lymphadenitis or valvular endocarditis to examine subspecific relationships among S. dysgalactiae strains through a comparison of the nucleotide sequences of their 16S and 23S rDNA.

Bacteria.
Streptococcal isolates and strains used for 16S and 23S rDNA analyses are shown in Table 1. Eleven porcine isolates of S. dysgalactiae subsp. equisimilis, whose origins have been described previously (Katsumi et al., 1997), were examined along with strain ATCC 35666 of S. dysgalactiae subsp. equisimilis and two strains of S. dysgalactiae subsp. dysgalactiae. In addition, Streptococcus agalactiae ATCC 13813^T was used as an outgroup.

Table 1. Streptococci used for sequence analysis of 16S and 23S rDNA S. dysgalactiae subsp. equisimilis strains are identified as arthritis-associated (prefix A), lymphadenitis-associated (L) or valvular endocarditis-associated (V) strains. NT, Not typable.

Phenotypic characterization.
The Streptex agglutination procedure (Murex Diagnostics) was used for detection of the Lancefield group AD, F and G antigens and the group L antigen was examined by precipitation in capillary tubes with hyperimmune serum (Lancefield, 1938). Streptokinase activity was determined on fibrin-plasma plates as described previously (Vandamme et al., 1996).

DNADNA reassociation.
DNA preparation and microplate hybridization were performed using the method of Ezaki et al. (1989) at an optimal renaturation temperature of 30 °C in 50 % formamide.

Sequence determination of rDNA.
Genomic DNA was extracted from bacterial colonies with a DNA extraction kit (SMITEST; Sumitomo Kinzoku Kogyo) according to the manufacturer's instructions. rDNA sequences were determined directly from DNA fragments produced by PCR amplification (Hultman et al., 1989, 1991). Primers and conditions for 16S rDNA PCR were described previously (Takahashi et al., 1997). The 23S rDNA sequence was analysed in the same manner. Furthermore, sequence determination was carried out for DNAs cloned into a pZero-2 vector plasmid (Invitrogen) for precise analysis. Oligonucleotide sequences of PCR primers used for the 23S rDNA are listed in Table 2. These and most of the intergenic sequencing primers were designed on the basis of data from a streptococcal 23S rDNA multiple alignment and sequences from previous investigations (Lane et al., 1991; Ash et al., 1992; Ludwig et al., 1992; Van Camp et al., 1993; Takahashi et al., 1997).

Table 2. Primers used for amplification of 23S rDNA Primer positions are based on the E. coli numbering system (Brosius et al., 1981).

Denatured PCR fragments or recombinant plasmids were submitted to a sequencing reaction using an Auto-Read sequencing kit and an ALF Express II DNA sequencer (Amersham Pharmacia Biotech).

Sequence alignment, calculation of similarity values and phylogenetic tree construction.
Nucleotide sequences of the 16S and 23S rDNA determined in this study were assembled and aligned manually using Gene Jockey II version 2.1 (Biosoft). Evolutionary distance values were estimated by Kimura's two-parameter method (Kimura, 1980) using the BioResearch SINCA program package (Fujitsu). The neighbour-joining method of Saitou & Nei (1987) was employed to construct a phylogenetic tree using the BioResearch SINCA program. The topology of the tree was evaluated by a bootstrapping method (Felsenstein, 1985).

The results of Lancefield grouping and streptokinase activity on human plasminogen of the 11 porcine isolates of S. dysgalactiae subsp. equisimilis are summarized in Table 1. Each of the isolates reacted with antibodies against the C or L antigen, which are considered to be the group antigens possessed by S. dysgalactiae subsp. equisimilis of animal origin. On the other hand, an atypical reaction was observed for streptokinase activity. Both Vieira et al. (1998) and Vandamme et al. (1996) have reported that a positive reaction for streptokinase activity on human plasminogen is a key characteristic for distinguishing subspecies equisimilis from subspecies dysgalactiae. However, the porcine isolates in this study did not exhibit streptokinase activity. These results were consistent with the results reported by Vandamme et al. (1996), in which most animal isolates of S. dysgalactiae lacked streptokinase activity, but were inconsistent with the classification proposed by Vieira et al. (1998). Streptokinase activity on human plasminogen may not be useful for identifying porcine isolates at the subspecies level as defined by Vieira et al. (1998).

Vieira et al. (1998) have reported that the levels of DNA relatedness between the two subspecies are 79 % or less. However, the porcine isolates in this study showed high levels of DNA relatedness (8588 %) against both S. dysgalactiae subsp. dysgalactiae and S. dysgalactiae subsp. equisimilis strains (Table 3). In this study, we tentatively designated the isolates as S. dysgalactiae subsp. equisimilis because of their β-haemolysis, which is a possible key factor for distinguishing one subspecies from the other based on the results reported by Vieira et al. (1998).

Table 3. DNA relatedness of porcine S. dysgalactiae strains against strains of the two subspecies of S. dysgalactiae

Alignment data from nucleotide residues 85 to 1500 (based on the Escherichia coli numbering system; Brosius et al., 1981) of the 16S rDNA were used for comparative analyses of the streptococcal strains. Of the 1429 bases of the 16S rDNA, nucleotide changes within the species S. dysgalactiae were found at 24 residues. Thirteen of the 24 residues were located in variable region V3 (residues 179220) designated by Raué et al. (1990) and the others were scattered in and around the other variable regions.

Fig. 1 shows part of the sequence alignment from residues 181 to 194 of the 16S rDNA. This substantial domain provided a criterion for subdividing strains of S. dysgalactiae subsp. equisimilis into two subgroups. One consisted of strains V26, L21, A1, A24 and L2, which all have sequences similar to that of S. dysgalactiae subsp. dysgalactiae. The other group consisted of the rest of the organisms, which were similar to each other but distinct from the other strains. We hypothetically designated the former and the latter as subgroups 1 and 2 (Fig. 1). The predicted secondary structures of rRNA estimated from these sequences were hairpin loop motifs. Although the primary sequences were significantly different at residues 181194, the hairpin loop motifs were very similar. All the motifs consisted of helices, 10 or 11 bp in length, and a 4-base loop. Furthermore, the numbers of canonical and non-canonical base pairs were nearly identical in each.

(27K): Fig. 1. Sequence alignment of residues 181194 of the 16S rDNA of strains of S. dysgalactiae subsp. dysgalactiae (ATCC 43078T and ATCC 27957) and S. dysgalactiae subsp. equisimilis (A1 to V21). Dots indicates bases identical to the residue of strain ATCC 43078T. Numbers 1 and 2 on the right correspond to subgroups designated in this study. Boxed sequences may form a helix structure with a minimum of free energy. Numbering of positions follows the convention for E. coli (Brosius et al., 1981).

In the complete 23S rDNA sequences (2904 bp), 21 positions varied between strains of the species S. dysgalactiae. However, no significant differences were apparent except at residues 294 and 652. In strains of S. dysgalactiae subsp. dysgalactiae, the bases at these positions were A and T, respectively, while, in most strains of S. dysgalactiae subsp. equisimilis, the bases were G and A.

It is apparent from the results of the treeing analyses based on 16S and 23S rDNA (Fig. 2) that the species S. dysgalactiae was distinct from the type strain of S. agalactiae with large evolutionary distance values (divergence 2·433·45 and 3·273·47 %, respectively). However, the branching order and the length of sublines within the S. dysgalactiae clusters were significantly different. S. dysgalactiae subsp. equisimilis strains were subdivided into two distinct clades in the 16S rDNA tree. The subgroups were recovered in 99 or 74 % of the bootstrapped trees. Furthermore, one of the subgroups (subgroup 1) formed a distinct clade with the strains of S. dysgalactiae subsp. dysgalactiae (bootstrap value of the group, 99 %). This conflicts with the degree of difference between the two subspecies of S. dysgalactiae demonstrated by DNADNA reassociation analysis (Vieira et al., 1998). Furthermore, the results based on 16S rDNA were inconsistent with those based on PFGE (Bert et al., 1997) and PAGE profiling of bacterial whole-cell proteins (Vandamme et al., 1996). These studies have shown no distinct groups within porcine isolates of S. dysgalactiae, although subtle differences, as in the biochemical properties of the isolates used in this study, were detected in individual strains.

(26K): Fig. 2. Unrooted neighbour-joining trees based on 16S and 23S rDNA sequences of 12 strains of Streptococcus dysgalactiae subsp. equisimilis and two strains of S. dysgalactiae subsp. dysgalactiae. Numbers at nodes are estimated confidence levels expressed as percentages and determined by bootstrap analysis with 1000 replications. Bar, evolutionary distance (Knuc) between sequences, determined by measuring the lengths of horizontal lines connecting two organisms.

In contrast, the subgroups produced by 16S rDNA treeing analysis were not distinguishable in the 23S rDNA tree. All the sublines of S. dysgalactiae strains branched at positions that were much closer to each other in the 23S rDNA tree (divergence, 0·39 % or less) than in the 16S rDNA tree (divergence, 1·51 % or less), although each of the clusters of S. dysgalactiae subsp. dysgalactiae and S. dysgalactiae subsp. equisimilis was monophyletic (bootstrap values 92 and 69 %, respectively). Subspecies-specific nucleotide differences were found at not more than three residues, though these could distinguish the two strains of S. dysgalactiae subsp. dysgalactiae from the other strains with a high bootstrap value in the treeing analysis. These findings were not inconsistent with the classification shown by Vieira et al. (1998).

It has been demonstrated that phylogenetic trees constructed from 16S and 23S rDNA sequence data are generally congruent with each other (Höpfl et al., 1989; Ludwig & Schleifer, 1994; Sallen et al., 1996; Ward et al., 2000). However, in this study, the phylogenetic relationship derived from 16S rDNA analysis was not supported by the result of the 23S rDNA analysis. This indicated that phylogenetic trees derived from 16S and 23S rDNAs are not always in agreement at the intraspecific level and poses the question of which phylogenetic marker reflects the true relatedness. A DNADNA reassociation study (Vieira et al., 1998) clearly subdivided S. dysgalactiae strains into two subspecies, one of which (β-haemolytic) reacts with Lancefield group C, G or L antigen and is positive for streptokinase activity on human plasminogen. The coincidence of these properties, except for streptokinase activity, in the porcine isolates used in this study suggests that the 23S rDNA analysis has an advantage over the 16S rDNA in the case of these taxa.

The strains used in our study carried either Lancefield C or L antigen. However, these serogroups were unrelated to the phylogenetic relationships deduced from 16S and 23S rDNA sequences. Hommez et al. (1991) reported that Lancefield group C and L S. dysgalactiae from pigs have almost identical cultural and biochemical traits. Furthermore, it is nearly impossible to distinguish these S. dysgalactiae serogroups on the basis of genealogical analyses such as DNADNA hybridization (Farrow & Collins, 1984), RAPD (Bert et al., 1996) and PFGE (Bert et al., 1997). These data indicate that distinctions based on Lancefield C or L serogrouping have no practical value for the subgrouping of S. dysgalactiae subsp. equisimilis.

The present phylogenetic analysis based on 16S rDNA sequences indicates that two groups exist within S. dysgalactiae subsp. equisimilis of porcine origin and that this divergence does not seem to be related to the phylogenetic relationships of 23S rDNA. Even so, the unique sequence found in the group-2 porcine isolates will be a useful marker for epidemiological studies.

Due to limitations in the number and source of strains examined, genealogical relationships between porcine strains of S. dysgalactiae subsp. equisimilis and other groups of strains, such as ecovars or group G strains, are still unclear.