N.F. Burton, School of Applied Sciences, University of Wales Institute Cardiff, Western Avenue, Cardiff CF5 2YB, Wales (e-mail: firstname.lastname@example.org). *Present address: Institute for Animal Health, Compton Laboratory, Compton, Newbury RG20 7NN, UK.
Burkholderia cepacia is found in soils and waters, it can be used in biocontrol and bioremediation but is also a human pathogen. It is not yet clear what differentiates pathogenic from non-pathogenic strains of the organism. In this study the multiple replicon structure was investigated in 28 strains of B. cepacia by pulsed field gel electrophoresis. All strains examined, whether of clinical, environmental or plant pathogenic origin, were found to have two, three or four large (> 500 kbp) replicons. Many strains also contained small replicons. Clinical strains were more likely to have three or four large replicons than non-clinical strains. Multiple replicon structure was also demonstrated in B. gladioli and Alcaligenes eutrophus.
The Gram-negative bacterium Burkholderia cepacia was first described as a phytopathogen of onion in 1950 ( Burkholder 1950). In recent years B. cepacia has been widely investigated as a potential agent of biocontrol against fungal diseases of crops and as an agent of bioremediation in the breakdown of recalcitrant herbicides and pesticides. Burkholderia cepacia has also emerged as a human pathogen particularly in cystic fibrosis (CF) and chronic granulomatous disease patients ( Holmes et al. 1998 ). The B. cepacia complex has been found to have an unusual genome both in terms of its size of between 4 and 8 Mbp ( Lessie et al. 1996 ; Holmes et al. 1998 ), which is up to twice the size of the Escherichia coli genome ( Blattner et al. 1997 ), and in its division into one to five circular replicons ( Lessie et al. 1996 ; Holmes et al. 1998 ). The genome is also rich in insertion sequences (IS) ( Lessie et al. 1996 ).
Multiple replicon organization in strains of the B. cepacia complex, principally of environmental origin, has been demonstrated by mapping and the use of pulsed field gel electrophoresis (PFGE; Cheng and Lessie 1994; Rodley et al. 1995 ; Lessie et al. 1996 ; Hübner et al. 1998 ). Cheng and Lessie (1994) mapped strain ATCC 17616 into three replicons of 3·4, 2·5 and 0·9 Mbp, with rrn genes and auxotrophic markers mapped to each replicon. The species type stain ATCC 25416 (NCPPB 2993) was mapped to three large circular replicons of 3·5, 3·1 and 1·1 Mbp along with a cryptic plasmid of around 200 kbp ( Rodley et al. 1995 ). The six sets of rrn genes in this strain were mapped onto the three large replicons. Burkholderia cepacia ATCC 53867 (AC1100), a laboratory strain isolated from a mixed biostat culture with the ability to degrade the chlorinated aromatic compound 2,4,5-trichlorophenoxyacetic acid (2,4,5-T), has five replicons, although only two, of 4 and 2·7 Mbp, are of megabase size ( Hübner et al. 1998 ). Southern blot hybridization was used to map tft genes involved in the degradation of 2,4,5-T to two of the smaller replicons, where they are strongly associated with IS elements ( Hübner et al. 1998 ). It has been proposed that the development of degradative pathways in B. cepacia is related to the plasticity of its genome due to both the replicon structure and the frequent occurrence of IS elements allowing the recruitment of foreign genes ( Lessie et al. 1996 ). High levels of homology between tft genes in AC1100 and degradative genes from E. coli, klebsiella and Ralstonia pickettii, along with their close association with IS elements, add strength to this theory ( Hübner et al. 1998 ). Rearrangements in the chromosome have been found to occur ( Lessie et al. 1996 ) and it is thought that high levels of homologous and illegitimate recombination along with recruitment of foreign genes may lead to spontaneous evolutionary bursts in B. cepacia ( Holmes et al. 1998 ). Such bursts may help to explain the emergence of human pathogenic forms of B. cepacia from an apparently harmless saprophyte.
Multireplicon structures have been found in other Burkholderia species and in closely related β-2 proteobacteria, including B. glumae, B. glathei, R. picketti and R. eutropha (Alcaligenes eutrophus) ( Rodley et al. 1995 ), although not in the phytopathogenic B. gladioli. Multiple replicons have also been described in unrelated species including Brucella, Rhizobium and Agrobacterium ( Allardet-Servent et al. 1993 ; Honeycutt et al. 1993 ; Michaux et al. 1993 ).
The aim of this work was to assess the distribution of multiple replicons in B. cepacia of environmental, clinical and phytopathogenic origin. A phytopathogenic isolate of B. gladioli and other β-proteobacteria were also investigated. Two methods of determining multiple replicon content have been described ( Lessie et al. 1996 ). Macrorestriction digestion of total genomic DNA with rare cutting endonucleases followed by PFGE allows a physical map of the replicons to be produced. Alternatively, undigested DNA may be separated by PFGE into whole replicons after linearization of the replicons during preparation of the DNA for electrophoresis ( Lessie et al. 1996 ). This second method was chosen on the basis of its ease of use in determining the replicon arrangements of larger numbers of bacterial strains.
MATERIALS and METHODS
The majority of B. cepacia strains from clinical and environmental sources were previously described by Wigley and Burton (1999). Burkholderia gladioli pv. allicola NCPPB 2478 and the phytopathogenic strains of B. cepacia, NCPPB 2993 (type strain) and NCPPB 3480, were obtained from the National Collection of Plant Pathogenic Bacteria (Harpenden, UK). The environmental B. cepacia isolates NCIMB 9087 and NCIMB 9092, R. eutropha (A. eutrophus) NCIMB 11842, A. faecalis NCIMB 8156, A. xylosidans subsp. xylosidans NCIMB 12033, Janthinobacterium lividum NCIMB 9130 and Chromobacterium violaceum NCIMB 9131 were obtained from the National Collection of Industrial and Marine Bacteria (Aberdeen, Scotland). Stenotrophomonas maltophilia 4543 was provided by Mr Alan Paull (PHLS, Cardiff, Wales).
Preparation of agarose-embedded bacterial DNA
The method was based on that of Grouthes et al. (1988) with modifications. Cells were harvested from an overnight culture at 30 °C on Nutrient agar (Oxoid Unipath, Basingstoke, UK) using 5 ml SE buffer per plate ( Grouthes et al. 1988 ). The cell concentration was then adjusted to approximately 9 × 108 cells ml−1 with SE buffer. A volume (0·5 ml) of this suspension was mixed with an equal volume of molten 2% low gelling temperature agarose (Sigma, Poole, UK) at 40 °C. The samples were then pipetted into CHEF plug moulds (Bio Rad, Hemel Hempstead, UK) to form plugs of approximately 100 µl. The plugs were allowed to set and cool for 90 min before removal from the mould and were then incubated for 48 h at 56 °C in N-lauroyl sarcosinate lysing solution ( Grouthes et al. 1988 ) containing 500 µg ml−1 proteinase K (ICN Biomedicals, Thame, UK). The lysing solution was changed after 24 h. Following incubation, plugs were washed in TE buffer ( Sambrook et al. 1989 ) for 5 min at room temperature. This was repeated for three more washes. The plugs could then be stored at 4 °C in TE buffer for up to 3 weeks.
Pulsed-field gel electrophoresis of agarose-embedded DNA
A 0·8% (w/v) chromosomal grade agarose gel (Bio Rad) was prepared with 0·5 × TBE buffer ( Sambrook et al. 1989 ). Agarose-embedded DNA plugs were placed in the wells using an alcohol-flamed spatula. The wells were then sealed with molten 0·8% (w/v) chromosomal grade agarose and allowed to set and cool for 1 h. The gel was then placed into the electrophoresis cell of a CHEF DR II PFGE system (Bio Rad) filled with 0·5 × TBE maintained at 9 °C. Gels were typically run for 50 h at 3 V cm−1, initial pulse time 250 s rising to 900 s, although the run time and final pulse time can be extended to improve the separation of close bands. Gels were stained with ethidium bromide (0·5 µg ml−1, 30 min), destained for 30 min under running water and photographed under u.v. transillumination with a CU5 land camera (Polaroid, Sigma–Aldrich, Poole, UK) fitted with a Kodak Wratten filter on 667 film (Polaroid, Sigma–Aldrich).
Estimation of replicon sizes
Replicon sizes were estimated by comparison with Hansenula wingei CHEF size markers (Bio Rad) on the gel photographs using simple linear regression. The R2 value was in excess of 99% for all gels. Hansenula wingei markers were chosen as their size range (1·05–3·13 Mbp) is similar to that of known B. cepacia replicons ( Lessie et al. 1996 ).
Plasmid isolation and size determination
The sizes of smaller replicons and plasmids (< 300 kbp) were determined by the methods previously described by Wigley and Burton (1999).
RESULTS and DISCUSSION
Two to four large (> 500 kbp) replicons were found in all 13 strains of B. cepacia of clinical origin, most being from CF patients, with total genome sizes ranging from 5·0 to 7·7 Mbp ( Table 1). Figure 1 shows typical gel photographs. Two or three large replicons were found in all the non-clinical strains of B. cepacia, with genome sizes ranging from 4·7 to 7·8 Mbp ( Table 1). In the other species examined multiple large replicons could only be detected in B. gladioli (3·2 and 3·0 Mbp) and R. eutropha (A. eutrophus) (3·3 and 3·0 Mbp).
Table 1. Replicon and total genomic size in strains of Burkholderia cepacia
On the basis of the results described above, and previous studies ( Rodley et al. 1995 ; Lessie et al. 1996 ), it appears that a genome consisting of several large circular replicons is a feature of both the B. cepacia species complex and other closely related β-2 proteobacteria.
Lessie et al. (1996) described the presence of two to four large replicons with a genomic size of 4·6–8·1 Mbp in eight representative strains of B. cepacia. The replicon structures in this study fall into that range, with two large replicons (> 500 kbp) being the most common arrangement amongst B. cepacia strains (20 of 28 isolates tested), whether of clinical or non-clinical origin. Three or four large replicons were found more frequently in clinical strains (46%) than in non-clinical strains (20%) and although it is interesting to speculate that this might be an evolutionary response to intensive antibiotic therapy the sample is too small to assign any real significance to this difference. However, the mapping of the tft genes involved in the degradation of 2,4,5-T to the smaller replicons in B. cepacia AC1100, their homology with other species and their association with IS elements ( Hübner et al. 1998 ) suggest that acquisition of genes to form either smaller chromosomes or megaplasmids does occur in Burkholderia spp.
In addition to the multiple chromosome organization in B. cepacia similar organization was found in B. gladioli and R. eutropha (A. eutrophus). Rodley et al. (1995) found multiple replicons in a number of phylogenetically closely related species including the B. cepacia type strain and two R. eutropha strains LMG 1199 and LMG 1201, although they were unable to detect any signal by PFGE with the B. gladioli strain NCPPB 644a. In this work we have demonstrated the presence of two large replicons in B. gladioli NCPPB 2478, indicating that this organization does occur within the species.
The sizes of the replicons determined in this study for the B. cepacia type strain agree closely with those previously reported ( Rodley et al. 1995 ; Lessie et al. 1996 ). The methods used here for the determination of the presence and estimation of size of multiple replicons are relatively simple and quick to perform in comparison with the macrorestriction digest methods used in mapping techniques, and are therefore more appropriate for the initial examination of a larger number of strains. Possible improvements include the use of a larger marker to allow more accurate sizing of larger replicons and irradiation of the sample ( Beverly 1989) or treatment with pronase E rather than proteinase K ( Lessie et al. 1996 ) to increase the amount of linearized DNA in the sample, so increasing the sensitivity of detection by PFGE. A drawback of the technique is that different replicons of the same length may not be separated.
On the basis that multiple replicon structure is common throughout the B. cepacia species complex, further analysis and mapping may help to determine the relationships between strains and further resolve the complex into different species ( Vandamme et al. 1997 ). This may also allow a greater understanding of the relationships between environmental and clinical strains of B. cepacia. This is of particular importance in considering the potential widespread use of B. cepacia as a biopesticide and the possible threat that this poses to human health, particularly to CF patients ( Govan and Vandamme 1998). The highly transmissible epidemic strain of B. cepacia has been found to contain insertion sequences from B. pseudomallei, the causative agent of the life-threatening tropical disease mellioidosis ( Holmes et al. 1998 ). This again suggests the potential for the acquisition of virulence and other genes from other species and shows that further genetic analysis is needed to give a more complete understanding of the development of B. cepacia as a pathogen.
The authors wish to thank Mr Alan Paull, PHLS, Cardiff and Prof. J.R.W Govan, University of Edinburgh for kindly providing clinical isolates of B. cepacia.