Phylogenetic study and multiplex PCR-based detection of Burkholderia plantarii, Burkholderia glumae and Burkholderia gladioli using gyrB and rpoD sequences

Bacterial seedling rot (Uematsu et al., 1976) and seedling blight (Azegami et al., 1987) caused by Burkholderia glumae and Burkholderia plantarii, respectively, are devastating diseases of rice seedling cultivation in nursery boxes. During germination, these pathogens infect seeds and invade plumules through stomata and wounds and proliferate in the intercellular spaces of the parenchyma (Azegami et al., 1988; Hikichi et al., 1995). Proliferation of B. glumae and B. plantarii in the plumules leads to production of the toxic materials toxoflavin (Sato et al., 1989; Yoneyama et al., 1998) and tropolone (Azegami et al., 1985), resulting in rot and blight in rice seedlings, respectively.

Within Burkholderia gladioli, three pathovars, all of which are recognized as phytopathogenic bacteria, have been delineated: B. gladioli pv. gladioli, which causes gladiolus rot (Hildebrand et al., 1973); B. gladioli pv. alliicola, which causes onion bulb rot (Young et al., 1978); and B. gladioli pv. agaricicola, which causes rapid soft rot of cultivated mushrooms (Lincoln et al., 1991). B. gladioli has also been isolated from rice seedlings showing bleaching symptoms, suggesting that this bacterium also infects rice seedlings (Kato et al., 1992). On the other hand, proliferation of B. glumae and B. plantarii is suppressed in rice seeds infected with B. gladioli (Miyagawa, 2000). Therefore, ecological studies on B. gladioli in rice plants are important to understand disease development caused not only by B. gladioli but also by B. glumae and B. plantarii. Such ecological studies require efficient detection and identification of these three rice-pathogenic Burkholderia species. Although the diversity of Burkholderia species has been analysed using 16S rRNA gene sequences (Hu et al., 2001; Salles et al., 2002), the discriminatory power of this gene is too restricted to reveal the detailed phylogenetic relationships among B. plantarii, B. glumae and B. gladioli. The 16S rRNA gene has been widely used for designing taxonomically meaningful, highly specific PCR primers, providing enough sequence information to allow the analysis of both close and distant phylogenic relationships among micro-organisms (Stackebrandt & Goebel, 1994). However, the degree of resolution obtained with 16S rRNA gene sequence analysis is not sufficiently discriminatory to permit resolution of intrageneric relationships among closely related micro-organisms, because of the extremely slow rate of evolution of the 16S rRNA gene (Yamamoto & Harayama, 1998). The genes encoding the β-subunit polypeptide of DNA gyrase (gyrB) and σ⁷⁰ factor (rpoD) are estimated to evolve much faster than the 16S rRNA gene (Yamamoto & Harayama, 1998).

In the present study, we analysed the phylogenetic diversity in the gyrB and rpoD gene sequences of these three rice-pathogenic Burkholderia species, in order to develop a specific and sensitive detection method.

Bacteria and DNA preparation.
Six Burkholderia species with validly published names, comprising a total of 108 strains, including 41 strains of B. glumae, 24 strains of B. plantarii and 37 strains of B. gladioli, were examined in this study (Table 1). Each bacterial isolate was grown aerobically in nutrient broth at 30 °C. Chromosomal DNA from the bacteria used as the PCR template was prepared using an AquaPure Genomic DNA Isolation kit (Bio-Rad), according to the supplier's instructions.

Table 1. Burkholderia strains analysed and classifications established according to nucleotide sequences of gyrB and rpoD ATCC, American Type Culture Collection; GTC, Gifu Type Culture Collection, Japan; KU, Kyushu University, Japan; KUCS, Kochi University, Japan; LMG, BCCM/LMG Bacteria Collection, Laboratorium voor Microbiologie, Gent, Belgium; MAFF, Micro-organisms Section of the MAFF Gene Bank.

PCR amplification and sequencing of gyrB and rpoD.
PCR of gyrB and rpoD was performed using primers UP-1E and AprU, and 70F2 and 70R2, respectively, as shown in Table 2. PCR was performed with a thermal cycler (TaKaRa) using PCR buffer (TaKaRa) containing 200 µM of each of the dNTPs, 0.5 µM of each primer, 0.2 µg template DNA and 2.5 U Ex-Taq polymerase (TaKaRa), in a total volume of 40 µl. A total of 35 amplification cycles were performed with template DNA denaturation at 94 °C for 1 min, primer annealing at 58 °C for 1 min and primer extension at 72 °C for 1 min. PCR products were electrophoresed on 1.0 % agarose gels and purified using Quantum Prep Freeze N Squeeze DNA Gel Extraction spin columns (Bio-Rad), following the manufacturer's instructions. The nucleotide sequences of gyrB and rpoD were determined directly from the PCR fragments using primers M13R (5'-CAGGAAACAGCTATGACC-3') and M13(21) (5'-TGTAAAACGACGGCCAGT-3'), and 70Fs (5'-ACGACTGACCCGGTACGCATGTA-3') and 70Rs (5'-ATAGAAATAACCAGACGTAAGTT-3'), respectively. Sequencing was carried out using an ABI Automated DNA Sequencer model 373 (Applied Biosystems) and analysed using DNASIS-Mac software (Hitachi Software Engineering).

Table 2. PCR and sequencing primers used in this study N, any; R, A or G; S, C or G; Y, C or T; M, A or C.

Data analysis.
The nucleotide sequences of gyrB (768801 bp) and rpoD (807834 bp) genes were aligned and a phylogenetic tree was constructed using CLUSTAL W (DNA database of Japan; ) with the neighbour-joining method (Saitou & Nei, 1987), using genetic distances computed with Kimura's two-parameter model (Kimura, 1980). The neighbour-joining phylogenetic tree was drawn using TreeView. The nucleotide sequences of gyrB and rpoD from Escherichia coli K-12 were used as the outgroup for phylogenetic tree reconstructions.

Analysis of the phylogenetic diversity of B. plantarii strains.
Aliquots (40 µl) of each PCR product from the genomic DNA of B. plantarii strains obtained using primers UP-1E and AprU were ethanol-precipitated and the pellets were dissolved in 10 µl distilled water. The DNA in the solutions was digested with SacI and HaeII (both from TaKaRa) at 37 °C for 6 h, loaded onto horizontal 1.2 % TAE agarose gels and stained with ethidium bromide after electrophoresis for detection of specific DNA fragments corresponding to the gyrB nucleotide sequences of the strains.

Multiplex PCR.
To detect and discriminate B. plantarii, B. glumae and B. gladioli using single-step PCR, primers (Table 2) developed for the gyrB gene of the bacteria were used for development of a multiplex-PCR protocol. PCR was performed with one cycle of 94 °C for 2 min and 35 cycles of 94 °C for 1 min, 63 °C for 1 min and 72 °C for 1 min. Aliquots (9 µl) of each PCR product were loaded onto horizontal 1.5 % TAE agarose gels and stained with ethidium bromide after electrophoresis for detection of 597, 529 and 479 bp DNA fragments corresponding to the gyrB nucleotide sequences of B. plantarii, B. glumae and B. gladioli, respectively.

Analysis of rice seeds infected with B. plantarii, B. glumae and B. gladioli.
To test whether infection of rice seeds with B. plantarii, B. glumae and B. gladioli could be detected using multiplex PCR, rice seeds naturally infected with B. plantarii and B. glumae were obtained from Dr H. Miyagawa, National Agricultural Research Center for Western Region, Japan. Flowering spikelets of rice plants cultivated in pots were inoculated with a suspension of B. plantarii MAFF 302391, MAFF 302907, MAFF 302924, MAFF 311030 or MAFF 301723^T, B. glumae MAFF 301169^T or B. gladioli MAFF 302386, MAFF 302543, MAFF 302918, MAFF 302919 or T-1 at 1.0x10⁸ c.f.u. ml¹ by spray application at 10 ml per plant, and the resultant rice seeds infected with each strain were used in this study. Non-infected rice seeds were used as a control. One gram of rice seeds was ground in a mortar with 1.0 ml distilled water and the suspension was then placed in a microtube and centrifuged at 1000 g for 1 min. The supernatant was held at 100 °C for 8 min and then centrifuged at 12 000 g for 3 min. Twenty-five microlitres of the supernatant was used as a template for the multiplex PCR.

Phylogenetic diversity of B. plantarii, B. glumae and B. gladioli
The phylogeny of six plant-pathogenic Burkholderia species (Burkholderia andropogonis, Burkholderia caryophylli, Burkholderia cepacia, B. plantarii, B. glumae and B. gladioli) was analysed using the combined nucleotide sequences of the gyrB and rpoD genes. These species all represented different clusters in the resulting phylogenetic tree (Fig. 1). The combined gyrB and rpoD nucleotide sequences showed high similarity and the sequence similarity between B. glumae MAFF 301169^T and B. plantarii MAFF 301723^T was 96.2 %, indicating a close phylogenetic relationship between the two species. Furthermore, the B. plantarii, B. glumae and B. gladioli clusters were supported by high bootstrap probabilities, indicating that they each form a tight monophylogenetic branch (Fig. 1).

(11K): Fig. 1. Phylogenetic tree of Burkholderia species based on the combined nucleotide sequences of gyrB and rpoD genes constructed with CLUSTAL W (DNA Database of Japan; ) using the neighbour-joining method (Saitou & Nei, 1987). Bar, 0.1 substitutions per position. Numbers at nodes represent percentage bootstrap values of 1000 resamplings. Sequences from E. coli K-12 were used as the outgroup.

Phylogenetic relationships among B. plantarii strains
Twenty-three of 24 B. plantarii isolates from Japan were from rice seedlings and soil used for cultivating rice seedlings in Chiba, Miyagi, Iwate and Yamagata (located in the northern part of the main island of Japan) and in Hokkaido (Fig. 2). The remaining strain, LMG 16020, is the type strain of B. vandii that is now classified as a strain of B. plantarii (Coenye et al., 1999). Whereas the Vanda strain LMG 16020 occupied a distinct position in the tree, the 23 rice strains were divided into three subgroups (Fig. 1). Nucleotide sequences of the gyrB PCR products from strains in subgroup I showed one restriction site for both HaeII and SacI (data not shown). One restriction site for HaeII was located in the gyrB PCR products from strains in subgroup II. No HaeII and SacI restriction sites were present in the gyrB PCR products from strains in subgroup III. SacI and HaeII digestion of the gyrB PCR products allowed discrimination of strains among subgroups I, II and III (Fig. 2).

Table 1 and Fig. 1, respectively, for strain numbers and subgroups.

Phylogenetic relationships among B. glumae strains
Analysis of combined nucleotide sequences of the gyrB qnd rpoD genes showed no diversity among 41 strains of B. glumae. The nucleotide sequences of gyrB in all 41 strains were identical and those of rpoD in six Japanese strains and two Indonesian strains differed by only one and two nucleotides, respectively, compared with those of the other 33 strains (data not shown). These results indicate that the diversity of nucleotide sequences of gyrB and rpoD among B. glumae strains is very restricted.

Phylogenetic relationships among B. gladioli strains
B. gladioli R-20202, which was obtained from a specimen from a cystic fibrosis patient in France, occupied a unique position among the 37 B. gladioli strains examined (Fig. 1). All other isolates formed a rather loose assemblage without subdivision supported by high bootstrap values. The combined nucleotide sequences of gyrB and rpoD of two rice strains, MAFF 302544 and MAFF 302914, were identical to those of the mung bean strain MAFF 302409 and the Cymbidium strain MAFF 302424. Furthermore, those of strain T-1 were identical to those of the Tulip strain MAFF 302515. Moreover, the combined translated amino acid sequences of gyrB and rpoD of the rice strains MAFF 302543, E-14 and H-1 were identical not only to those of the corn flour strain GTC 1088, B. gladioli pv. agaricicola GTC 1730, the gladiolus strain ATCC 10248^T and soil strains MAFF 302533 and MAFF 302534, but also to the cystic fibrosis patient strains LMG 18157 (USA) and R-15279 (Germany). These results indicate that the pathovar, host plant and geographical origin of the strains correlate poorly with the phylogenetic diversity among the B. gladioli strains.

Detection of B. gladioli, B. glumae and B. plantarii in infected rice seeds by multiplex PCR
A multiplex PCR-based detection method for B. gladioli, B. glumae and B. plantarii was developed using the gyrB nucleotide sequences (Table 2). A 479 bp fragment specific for B. gladioli was amplified from the genomic DNA of all B. gladioli strains used in this study (Fig. 3). From the genomic DNA of all B. glumae strains tested, a 529 bp DNA fragment was specifically amplified in the multiplex PCR. A 597 bp DNA fragment was specifically amplified from the genomic DNA of all B. plantarii strains tested, including LMG 16020 from Vanda sp. No DNA fragments were amplified from the genomic DNA of B. cepacia ATCC 25416^T. Sequencing of the PCR products confirmed the specificity of the multiplex PCR (data not shown).

Table 1 for strain numbers. BC, B. cepacia ATCC 25416T.

To assess the usefulness of multiplex PCR in discriminatory detection of rice seeds infected with B. gladioli, B. glumae and B. plantarii, DNA isolated from pathogen-infected rice seeds was used as a template. A 479 bp fragment specific for B. gladioli was only amplified from DNA of rice seeds from plants inoculated with B. gladioli MAFF 302386, MAFF 302543, MAFF 302918, MAFF 302919 and T-1 (Fig. 4a). One fragment of 529 bp in length was amplified from DNA of rice seeds from plants inoculated with B. glumae MAFF 301169^T. A 597 bp DNA fragment was amplified from DNA of rice seeds from plants inoculated with B. plantarii MAFF 302391, MAFF 302907, MAFF 302924, MAFF 311030 and MAFF 301723^T. Furthermore, two fragments of 529 and 597 bp in length were amplified from DNA of rice seeds naturally infected with B. glumae and B. plantarii, respectively (Fig. 4b). Sequencing of the PCR products confirmed the specificity obtained by PCR (data not shown). No products were obtained from DNA from non-infected rice seeds (Fig. 4b). These results show that the multiplex-PCR protocol facilitates specific detection and discrimination of rice seeds infected with B. plantarii, B. glumae and B. gladioli.

(29K): Fig. 4. Detection of B. plantarii, B. glumae and B. gladioli in rice seeds by multiplex PCR. The expected sizes of partial fragments of gyrB amplified by PCR from DNA isolated from rice seeds infected with B. plantarii, B. glumae and B. gladioli were 597, 529 and 479 bp, respectively. (a) PCR products from DNA of rice seeds produced from rice plants inoculated with B. plantarii, B. glumae or B. gladioli. Lanes: 15, B. gladioli MAFF 302386, MAFF 302543, MAFF 302918, MAFF 302919 and T-1, respectively; 6, B. glumae MAFF 301169T; 711, B. plantarii MAFF 302391, MAFF 302907, MAFF 302924, MAFF 311030 and MAFF 301723T, respectively. (b) Lanes: 1 and 2, PCR products from DNA of rice seeds naturally infected with B. glumae and B. plantarii, respectively; 3 and 4, PCR products from genomic DNA of B. glumae MAFF 301169T and B. plantarii MAFF 301723T, respectively; 5, PCR products from DNA of non-infected rice seeds.

Many ecological studies on B. glumae and B. plantarii have been reported, leading to the development of systems to control the diseases caused by these two Burkholderia species. However, the ecology and pathogenicity mechanism of B. gladioli on rice plants remain unclear. Moreover, the antagonistic activity of B. gladioli against B. glumae and B. plantarii in rice plants (Miyagawa, 2000) shows that an understanding of the ecology of B. gladioli on rice plants would facilitate not only the development of control systems for B. gladioli-induced disease, but also an understanding of the ecology of B. glumae and B. plantarii on rice plants. In the present study, we have designed a multiplex PCR that can be used for the simultaneous detection of B. plantarii, B. glumae and B. gladioli in rice seeds infected with these Burkholderia species. Therefore, this assay might facilitate an understanding of the ecological significance of Burkholderia species, especially the interactions that control their survival fitness on rice plants, thus leading to the development of relevant control systems. This work was supported financially by a grant from Sumitomo Chemical Co. Ltd and a Grant-in Aid for Scientific Research (no. 16658020) from the Japanese Society for the Promotion of Sciences to Y. H. and by a Sasakawa Scientific Research Grant from the Japan Science Society to Y. M.