Variation in Flagellin Genes and Proteins of Burkholderia cepacia

Abstract

The majority of isolates of Burkholderia cepacia, an important opportunistic pathogen associated with cystic fibrosis, can be classified into two types on the basis of flagellin protein size. Electron microscopic analysis indicates that the flagella of strains with the larger flagellin type (type I) are wider in diameter. Flagellin genes representative of both types were cloned and sequenced to design oligonucleotide primers for PCR amplification of the central variable domain of B. cepacia flagellin genes. PCR-restriction fragment length polymorphism analysis of amplified B. cepacia flagellin gene products from 16 strains enabled flagellin type classification on the basis of product size and revealed considerable differences in sequence, indicating that the flagellin gene is a useful biomarker for epidemiological and phylogenetic studies of this organism.

Burkholderia cepacia (formerly Pseudomonas cepacia; a member of the rRNA group II pseudomonads) has emerged as an increasingly important opportunistic pathogen, particularly in relation to patients suffering from cystic fibrosis (CF) (15). Acquisition of B. cepacia, often occurring after lengthy colonization with Pseudomonas aeruginosa, can lead to the rapid deterioration or death of CF patients, and this organism appears to be transmissible between patients (14). There is considerable evidence that some strains of B. cepacia are more virulent than others and that the outcome of colonization by a particular strain can vary from rapidly fatal septicemia to maintenance of stable respiratory function (16). A number of factors have been implicated in the greater virulence of some strains. These include adhesion to respiratory mucin (31, 32) and the presence of cable pili (33).

Motility in B. cepacia is by means of polar flagella. Flagella, each consisting of a flagellin filament, hook, and basal body, have been implicated as invasive virulence factors for a number of bacteria (28), including P. aeruginosa (11). Unlike P. aeruginosa, which appears to sit in microcolonies in the viscid mucus, leading to progressive lung damage with episodes of acute debilitating exacerbation, some strains of B. cepacia cause rapidly fatal pneumonia in CF patients (15), suggesting that they may have the ability to move through the mucus. Because of their location on the outside of bacterial cells, flagellins have been targeted in vaccine design. Brett et al. (4) demonstrated that flagellin-specific antisera were capable of protecting diabetic rats from challenge with strains of Burkholderia pseudomallei (another member of rRNA group II). In a recent study, an O-polysaccharide moiety of B. pseudomallei was covalently linked to the flagellin protein from the same strain. O-polysaccharide–flagellin conjugates elicited a high-level immunoglobulin G response capable of protecting diabetic rats from challenge with a heterologous strain of B. pseudomallei (5).

Two distinct flagellin protein molecular mass groups in B. cepacia have been reported by Montie and Stover (23). In this previous study, type I flagellins were reported as having a molecular mass of 31 kDa while the molecular mass of type II flagellins was reported as 44 to 46 kDa. This early study, using a limited number of isolates, suggested that with regard to flagellin, B. cepacia is analogous to another CF pathogen, P. aeruginosa, in which two flagellin antigenic types distinguishable by protein or gene size are found (43). Several representatives of the heterologous a-type and homologous b-type fliC loci of P. aeruginosa (encoding flagellins) have been sequenced (37). In addition, PCR amplification of flagellin genes coupled with restriction fragment length polymorphism (RFLP) analysis can be used as a method for differentiating between clinical isolates of P. aeruginosa (7, 43). In this paper we report the development of a similar approach to the study of populations of B. cepacia and discuss the divergence of a highly variable gene, the flagellin gene (fliC), within populations of B. cepacia.

MATERIALS AND METHODS

Bacterial strains.

The bacterial strains used in this study, listed in Table 1, are clinical or botanical isolates. All clinical isolates of B. cepacia were initially isolated from CF patients by the use of selective media (Mast Laboratories, Bootle, United Kingdom). Presumptive identification was accomplished with the API-20NE system (BioMérieux) and confirmed by the Central Public Health Laboratory Service, Colindale, London, United Kingdom, using a range of biochemical and molecular tests, including 16S rRNA sequencing. Several isolates were identified as the United Kingdom CF epidemic strain (ET12 lineage [16]) by electrophoretic typing, whole-protein sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE), and pyrolysis mass spectroscopy (8). Environmental isolates of B. cepacia (6) were kindly supplied by J. Govan, University of Edinburgh. Their identity had been established by comparisons of 16S rRNA sequences. Strains were maintained on nutrient agar and grown at 30°C.

TABLE 1.

Description of B. cepacia strains used in this study

Strain	PCR product size (kb)	Flagellin type	RFLP group	Source	Description
E241a	1.0	II	I	J. Govan, Edinburgh	CF epidemic strain
E242	1.4	I	II	J. Govan, Edinburgh	CF isolate, nonepidemic
E243 (J2543)b	1.0	II	III	J. Govan, Edinburgh	Botanical strain
E244 (J2534)b	1.0	II	IV	J. Govan, Edinburgh	Botanical strain
E245 (J2552)b	1.4	I	V	J. Govan, Edinburgh	Botanical strain
E195	1.4	I	VI	Sheffield	CF isolate
E196	1.0	II	VII	Sheffield	CF isolate
E197	1.6	NTc	VIII	Sheffield	CF isolate
PC3	1.0	II	I	Alder Hey Hospital, Liverpool	CF isolate, nonepidemic
PC7	1.0	II	I	Alder Hey Hospital, Liverpool	CF isolate, nonepidemic
PC9L	1.0	II	I	Alder Hey Hospital, Liverpool	CF isolate, nonepidemic
PC9S	1.0	II	I	Alder Hey Hospital, Liverpool	CF isolate, nonepidemic
PCI2	1.0	II	X	Alder Hey Hospital, Liverpool	CF isolate, nonepidemic
PCI4a	1.0	II	I	Alder Hey Hospital, Liverpool	CF epidemic strain
MI7a	1.0	II	I	Manchester	CF epidemic strain
NCIB 9085	1.0	II	IX	Trinidad	Type strain, soil isolate

^aUnited Kingdom CF epidemic strain confirmed by SDS-PAGE of whole protein and by pyrolysis mass spectroscopy (8). 

^bStrain is further described by Butler et al. (6). 

^cNT, neither type I nor type II is applicable to this strain, which has an unusually large flagellin gene and protein. 

Flagellin protein isolation and N-terminal sequencing.

Flagellin proteins were isolated by the procedure described by Brett et al. (4), using 15% (wt/vol) ammonium sulfate to precipitate the protein, and analyzed by SDS-PAGE on 10% (wt/vol) polyacrylamide gels. Bio-Rad low-range standards were used as molecular mass markers. N-terminal sequencing of B. cepacia E243 flagellin (MLGINSNINSLVAQQNLNGS) and two trypsin digestion-generated products (IGGGLVQKGQTVGTVT and NQVLQQAGI) was performed by Mark Wilkinson (University of Liverpool, Liverpool, United Kingdom).

Cloning of B. cepacia E243 and B. cepacia E242 flagellin genes.

Genomic DNA was extracted from B. cepacia strains as described previously (42). The N-terminal sequences of the B. cepacia E243 flagellin and the longer of the two internal amino acid sequences were used to design degenerate primers, obtained from Genosys, for PCR amplification of a region of fliC. A combination of the sense primer BC1 (GTIGCICARCARAAYCTIAAYGG) and the antisense primer BCR2 (CCNACSGTCTGRCCCTTCTG) was employed to generate an amplified product of approximately 450 bp. One-microliter aliquots of various dilutions of B. cepacia E243 genomic DNA were used directly in 100-μl volumes containing 2 U of Taq polymerase (Gibco BRL), 200 nM each primer (BC1 and BCR2), 1× Taq polymerase buffer, 100 μM each deoxynucleoside triphosphate (dATP, dCTP, dGTP, and dTTP), and 1.5 mM MgCl₂. Amplifications were carried out in a MiniCycler (Genetic Research Instrumentation Ltd.) for 30 cycles consisting of denaturation at 95°C (1 min), annealing at 45°C (1.5 min), and extension at 72°C (1.5 min) with an additional extension period at 72°C (10 min) following completion of the 30 cycles. At the end of the amplification, 20-μl samples were subjected to electrophoresis on a standard 1.0% (wt/vol) agarose gel to confirm the presence of an amplified product. The 450-bp amplified product was purified by the use of a gel extraction kit (Qiagen) and cloned by using the LigATor kit (R & D Systems Europe Ltd.). Both strands of three separate clones were sequenced by the University of Liverpool DNA Sequencing Service, using vector and internal sequencing primers. A comparison of this sequence information with sequences in the database confirmed that the PCR product was part of the flagellin gene. The 450-bp amplified product was labeled with digoxigenin–11-2′-dUTP (DIG) (Boehringer Mannheim) by repeating the PCR with 60 μM DIG in a total reaction volume of 50 μl. The labeled product was used as a probe to identify flagellin gene-containing clones from B. cepacia E243 and B. cepacia E242 gene libraries constructed from genomic DNA of these strains by using the SuperCos 1 Cosmid Vector Kit (Stratagene) under the conditions recommended by the supplier. The presence of DIG on colony blots was detected by using anti-DIG-alkaline phosphatase Fab fragments and the chemiluminescent substrate CDP-Star (Boehringer Mannheim) in the procedure recommended by the supplier. Smaller flagellin gene-containing fragments, identified by Southern blot analysis of digested cosmid clones, were subcloned into the plasmid vector pUC19 (Life Technologies Ltd.).

Nucleotide sequencing.

DNA was purified from putative flagellin gene clones by using Qiagen Mini or Midi kits (Qiagen). Both strands of the B. cepacia E243 and B. cepacia E242 flagellin genes were sequenced by primer walking at the University of Liverpool DNA Sequencing Service, using vector and internal oligonucleotide primers. The nucleotide sequence of the amplified B. cepacia E197 fliC product was obtained following cloning with a LigATor cloning kit (R & D Systems Europe Ltd.). The nucleotide sequences of both strands were obtained for three separate clones.

Computer analyses.

Nucleotide sequence alignments, percent G+C values, determination of amino acid composition, prediction of protein masses, and alignment of predicted flagellin proteins with each other or with other flagellins (retrieved from EMBL, GenBank, or SwissProt [27]) were carried out with GAP, COMPOSITION, PEPTIDESORT, PILEUP, CONSENSUS, and FASTA from the GCG sequence analysis software package (Genetics Computer Group, University of Wisconsin). Phylogenetic analysis was carried out by aligning the first 100 N-terminal flagellin residues into a multiple sequence file by using PILEUP. A tree was subsequently constructed by using the PHYLIP program. Bootstrap values, indicated on the tree (see Fig. 5), were derived by using 100 alternatives (values under 50% were omitted).

FIG. 5.

Relationships among bacterial N-terminal flagellin sequences. The 100 N-terminal amino acid residues from 27 bacterial flagellin proteins, including B. cepacia E243, were aligned to construct a sequence similarity dendrogram by using PHYLIP. The representative sequences chosen (with GenBank accession numbers in parentheses) were deduced flagellin protein sequences from Agrobacterium tumefaciens flaA (X80701), Azospirillum brasilense laf1 (U26679), Aquifex pyrophilus (U17575), Bacillus subtilis (X56049), Bacillus thuringiensis flaB (X67139), Bartonella bacilliformis (L20677), Bordetella bronchiseptica (L13034), Borrelia burgdorferi (X16833), B. cepacia E243, C. jejuni (J05635), Caulobacter crescentus (J01556), E. coli (M14358), H. pylori flaA (X60746), Legionella pneumophila (X83232), Listeria monocytogenes (X65624), Proteus mirabilis fliC1 (L07270), P. aeruginosa (M57501), Rhizobium meliloti flaA (M24526), Roseburia cecicola (M20983), S. typhimurium (D13689), Serpulina hyodysenteriae (X63513), Serratia marcescens (M27219), Shigella boydii (D26165), Treponema phagedenis (M94015), Vibrio parahaemolyticus flaA (U12816), Wolinella succinogenes (M82917), and Yersinia enterocolitica (L33467). Representatives of the beta subdivision of Proteobacteria, Bordetella bronchiseptica and B. cepacia, are highlighted.

PCR amplification of B. cepacia flagellin genes.

Flagellin gene oligonucleotide primers BC4 (CTGGTCGCACAGCAGAACCTGAAC; N terminal) and BCR12 (ACAG/TGTTCGCGGTTTCCTG; C terminal) were obtained from Genosys (Cambridge, United Kingdom). Cells taken from a nutrient agar plate were suspended in sterile distilled water and boiled for 5 min. This lysed suspension (2.5 μl) was used directly in a standard amplification mixture. Amplifications were carried out in 50-μl volumes containing 2 U of DynaZyme (Flowgen Instruments Ltd., Sittingbourne, Kent, United Kingdom), 200 nM each primer (BC4 and BCR12), 1× DynaZyme buffer, and 100 μM each deoxynucleoside triphosphate (dATP, dCTP, dGTP, and dTTP) for 30 cycles consisting of denaturation at 95°C (1 min), annealing at 60°C (1 min), and extension at 72°C (2 min).

Restriction digestion of amplified products.

Amplified product samples (10 μl) were digested with the restriction enzymes HaeIII and MspI under the conditions recommended by the supplier (Life Technologies Ltd.). These digests were then subjected to electrophoresis on 3% (wt/vol) MetaPhor agarose gels (Flowgen) alongside a PCR size marker (R&D Systems Europe Ltd.; fragment sizes, 50, 150, 300, 500, 750, 1,000, 1,500, and 2,000 bp).

Electron microscopy.

Electron microscopic analysis of flagella was carried out by procedures described previously (43). For flagellar width measurements, the widths of three flagella from one area of a grid were measured three times each. This was repeated in two additional areas of the grid. Mean width and standard error (n = 27) were calculated.

Nucleotide sequence accession numbers.

The flagellin gene sequences for B. cepacia E243, E242, and E197 have been given GenBank accession no. AF011370, AF011371, and AF011372, respectively.

RESULTS

Flagellin protein variation in B. cepacia.

SDS-PAGE of flagellin proteins isolated from different strains of B. cepacia confirmed the presence of two major groups, based on approximate flagellin sizes: 45-kDa flagellins (type II, including E243) and 55-kDa flagellins (type I, including E242) (Fig. 1). Flagellin proteins were isolated from several type I strains and from three type II strains. One strain, E197, produces a larger flagellin protein (approximately 70 kDa). No significant size variation within the major groups was observed.

FIG. 1.

PAGE of B. cepacia flagellins. The figure shows a 10% (wt/vol) SDS-PAGE gel of flagellin proteins isolated from strains E197 (lane 2), E195 (lane 3), E245 (lane 4), E244 (lane 5), E243 (lane 6), E242 (lane 7), and E241 (lane 8). Lanes 1 and 9 contain Bio-Rad low-range SDS-PAGE standards; their molecular masses are indicated on the left.

Electron microscopy.

Electron microscopic analysis was used to obtain width measurements for flagella from four strains of B. cepacia. The results obtained for E241 (14.6 ± 0.3 nm), E242 (18.0 ± 0.5 nm), and E243 (14.5 ± 0.3 nm) indicate that the type I flagellum (E242) is significantly wider than the type II flagellum (E241 and E243) (Fig. 2). The flagellar width observed in strain E197 (19.4 ± 0.5 nm) was the largest.

FIG. 2.

Electron microscopic analysis of B. cepacia type I (a) and type II (b) flagellar preparations. Magnification, ×158,000.

DNA sequence analysis of B. cepacia flagellin genes.

The DIG-labeled flagellin gene probe was used to identify several hybridizing flagellin gene cosmid clones from clone banks of both B. cepacia E242 and B. cepacia E243. A 5-kb SstI fragment of E243 genomic DNA and a 3-kb SalI fragment of E242 genomic DNA, both identified as containing the flagellin gene, were subcloned from cosmid clones into pUC19, and the complete nucleotide sequences of the B. cepacia E242 and B. cepacia E243 flagellin genes were obtained. The putative transcriptional start site was identified by alignment with other flagellin gene sequences and, following computer-assisted translation to derive a polypeptide sequence, with the known N-terminal sequence (MLGINSNINSLVAQQNLNGS). The region immediately upstream from fliC in B. cepacia E243 was also sequenced (GenBank accession no. AF011370). The sequence TAAAGTTN₁₂GCCGAAAT is located upstream from the transcriptional start site of the B. cepacia E243 flagellin gene, in the same position as an identical sequence identified as a putative ς^F-type promoter in B. pseudomallei (10).

The percent G+C values obtained for the flagellin gene coding sequences of E242 (63.4%) and E242 (63.8%) are similar to the value obtained for the pilin gene (cblA) of B. cepacia cable pili (62% [33]) and the value reported for B. cepacia genomic DNA (67% [26]).

Flagellin gene nucleotide sequences were used to derive predicted protein sequences for the E242 and E243 flagellins. The N-terminal sequences deduced from purified flagellin protein and trypsin-generated products could be identified within the predicted protein sequence of B. cepacia E243 flagellin (Fig. 3). The predicted flagellin protein molecular masses were 38.7 and 50.0 kDa for B. cepacia E243 and E242, respectively. As with B. pseudomallei flagellin (10), there are no histidine or cysteine residues found in the flagellin of B. cepacia E242 or E243. In common with other flagellins, there are few or no tyrosine (one and three residues in E242 and E243 flagellins, respectively) or tryptophan (no residues in either flagellin) residues (41), with the majority of the protein (90 to 91%) consisting of aliphatic uncharged amino acid residues. The proportions of aromatic (2%), acidic (3 to 4%), and basic (3%) residues are also consistent with those of previously reported flagellins.

FIG. 3.

Alignment of Burkholderia flagellins. The best computer-assisted alignment of B. cepacia E242, B. cepacia E243, and B. pseudomallei flagellins is shown. The consensus sequence indicates residues common to all three proteins. Determined E243 flagellin N-terminal sequences are indicated in boldface.

An alignment of the predicted flagellin protein sequences of B. cepacia E242, B. cepacia E243, and B. pseudomallei (10) is shown in Fig. 3. B. cepacia E243 flagellin has 83.8% peptide sequence similarity (77.5% identity) to the B. pseudomallei flagellin. In SwissProt, the non-Burkholderia sequences most closely related to the E243 flagellin sequence were the flagellins of Legionella micdadei (accession no. p53606; 61.4% similarity, 44.4% identity) and Salmonella enteritidis (accession no. q06972; 57.6% similarity, 42.7% identity). B. cepacia E242 flagellin exhibits 80.5% peptide sequence similarity (72.0% identity) to the B. pseudomallei flagellin. In SwissProt, the non-Burkholderia sequences most closely related to B. cepacia E242 flagellin were the flagellins of L. micdadei (55.0% similarity, 40.6% identity) and Proteus mirabilis (accession no. p42273; 57.4% similarity, 41.8% identity). A comparison of the peptide sequences of the B. cepacia E242 and E243 flagellins revealed a similarity of 82.3% (74.5% identity).

The amino acid sequence of the central region of the larger flagellin (B. cepacia E242 flagellin residues 240 to 400) was compared separately to the sequences present in SwissProt. The five best matches show no more than 30% identity in 64 to 149 amino acid residues.

The larger amplified product obtained from B. cepacia E197 was sequenced and found to exhibit 68% identity to the B. cepacia E242 flagellin gene. Although the E197 product does not include the entire gene sequence, by alignment with other flagellin gene sequences it was possible to identify a putative open reading frame and to determine a predicted peptide sequence of 540 residues with 78 to 80% similarity and 69 to 73% identity to the flagellins of B. pseudomallei, B. cepacia E242, and B. cepacia E243. A comparison of the variable central domain of the B. cepacia E197 flagellin gene or protein sequence with sequences in the GenBank and SwissProt databases revealed no significant homology.

PCR amplification and RFLP analysis of amplified flagellin gene products.

Oligonucleotide primers were designed by alignment of Burkholderia flagellin sequences and tested on the 16 B. cepacia isolates. With the exception of strain E197 (1.6-kb product), all strains yielded amplified products of either 1.0 kb (type II) or 1.4 kb (type I). Products could be further distinguished by digestion with restriction enzymes HaeIII and MspI (Fig. 4). The 16 strains of B. cepacia were subdivided into a total of 10 flagellin gene RFLP groups (designated I to X) by this approach (Fig. 4; Table 1).

FIG. 4.

RFLP patterns of PCR-amplified flagellin genes. All three gels feature a size marker (M), consisting of either λ DNA digested with HindIII and EcoRI (a) or a PCR marker (b), and amplified DNA from strains representing RFLP groups I to X (lanes 1 to 10, respectively). The PCR products were undigested (a) or digested (b) with HaeIII (upper gel) or MspI (lower gel).

DISCUSSION

The N-terminal sequence of the flagellin protein isolated from B. cepacia E243 was identical to the first 17 residues reported for the N-terminal sequence of B. pseudomallei flagellin (4), confirming that the protein is flagellin and indicating a close relationship between the B. cepacia and B. pseudomallei flagellins. This was subsequently confirmed by comparing the predicted flagellin peptide sequences of two B. cepacia strains with the published sequence of the B. pseudomallei flagellin (10). The three Burkholderia flagellin sequences exhibit high levels of homology in the conserved terminal regions but differ considerably in the central region. This is a common feature of flagellin proteins, which are believed to fold into a hairpin-like conformation, with the terminal domains being responsible for defining the basic filament structure lying on the inner surface and the central, variable region being surface exposed (41). The flagellins of B. cepacia E243 (385 residues) and B. pseudomallei (388 residues) are very similar in length, but the N-terminal methionine residue at position 1 in the purified B. cepacia E243 protein is not present in the B. pseudomallei flagellin (10). The flagellins of strains E242 (505 residues) and E197 include extensive areas in the central region that are not represented in the other two Burkholderia flagellins. A comparison of these central regions with other sequences in the database indicated that they show no significant homology to any previously sequenced protein.

A comparison of the B. cepacia E242 and E243 flagellins with database sequences revealed that the most closely related non-Burkholderia flagellin sequences were representatives of the gamma subclass of the class Proteobacteria. This is in agreement with the phylogenetic analysis of B. pseudomallei reported by DeShazer et al. (10). A dendrogram constructed following alignment of the 100 N-terminal amino acid residues of 27 bacterial flagellin proteins, including B. cepacia E243 and representing 26 different genera, is shown in Fig. 5. The figure shows output typical of several approaches, including both distance-matrix and parsimony analysis, taken to construct a dendrogram. The actual position of B. cepacia is not well supported by bootstrap analysis. However, regardless of the approach taken, B. cepacia does not cluster closely to representatives of any other genera. B. cepacia has been assigned to the beta subdivision of the Proteobacteria (24). The only other flagellin sequence available for a representative of the beta subdivision, Bordetella bronchiseptica, is far more closely related to flagellins obtained from gamma-subdivision proteobacteria (1). More flagellin sequence information for other representatives of the beta subdivision is required to resolve this apparent anomaly, but it may be that Burkholderia spp. flagellins are more representative of beta-subdivision proteobacteria as a whole.

Our analyses of SDS-PAGE gels of flagellin proteins suggest that either the previously reported type I flagellin molecular mass of 31 kDa (23) is inaccurate or there are strains of a third major type with a considerably smaller flagellin protein. The latter explanation seems unlikely because of our inability to identify any such strains in this study and because of the tendency for flagellins of the same species to exhibit a higher degree of similarity than such a scenario would allow. Montie and Stover (23) reported the presence of double-banded patterns in the gels of 31-kDa flagellin protein preparations, suggesting that the 31-kDa protein band may be degraded protein or an artifact of the isolation procedure rather than complete flagellin.

The presence of two major flagellar types within populations of B. cepacia suggests analogy with P. aeruginosa, whose type a and type b flagellins can be distinguished by flagellin protein and gene size (43). Southern blot analysis of chromosomal DNA from E242 and E243, digested with several different restriction enzymes and hybridized with flagellin gene probes, leads to single hybridizing bands except for those enzymes for which there is a recognition site within the flagellin gene (data not shown). This suggests that, as in P. aeruginosa, individual strains possess the genetic information required for the production of only one flagellin type. There is no evidence for the kind of phase variation exhibited by strains of Salmonella typhimurium, for which switching between flagellar types can be observed in a single strain (35). Flagella are highly immunogenic and have been shown to undergo recombination to generate antigenic variation in a number of bacteria, including Salmonella spp. (36) and Campylobacter spp. (2), which contain multiple flagellin genes within individual strains. In Campylobacter jejuni, recombinational events following the uptake of exogenous DNA by naturally competent cells has been demonstrated (39). Further antigenic variation may be generated by posttranslational modification (18). Such modifications have also been found in P. aeruginosa flagellins in which phosphorylated tyrosines have been observed (19, 20). P. aeruginosa type a flagellins, despite having conserved gene sizes, are heterogeneous in molecular weight, a characteristic that has been attributed in part to posttranslational modification. We found no evidence for similar variation in either type I or type II flagellins of B. cepacia, although the observed discrepancy between predicted molecular weights and those estimated from SDS-PAGE gels suggests that there may be a limited role for posttranslational modification. However, since there are only one to three tyrosine residues evident in the B. cepacia flagellins, any contribution of tyrosine phosphorylation to the molecular weight would be limited. One B. cepacia isolate, E197, produces an unusually large flagellin protein. A similar finding was reported in a study of P. aeruginosa isolates (43), in which one isolate not conforming to the type a or type b flagellin sizes was observed. It is known that considerable variation in the size of the flagellin central domain can occur without detrimental consequences for the function of flagella. In Escherichia coli K-12, the minimally sized functional flagellin comprises only 310 of the normal 497 residues; the remainder can be deleted or altered without loss of function (21). Although the present study included only 3 representative type I strains and 16 strains in all, we have screened approximately 50 other clinical isolates of B. cepacia for flagellin gene size without successfully identifying a strain with a flagellin similar to that found in B. cepacia E197. This suggests that the fliC gene of strain E197 is exceptional. Our screening indicates that the vast majority of clinical isolates of B. cepacia contain type II flagellins, although the significance of this finding is not clear since there are exceptions to this general rule.

Electron microscopic analysis of representative B. cepacia strains indicates that the flagellar widths of the two major flagellin types differ. The difference in width between type I and type II B. cepacia flagella was greater than the flagellar width difference observed between type a and type b flagella in P. aeruginosa (43).

Although amplification of the flagellin gene by PCR was successful for all 16 B. cepacia strains tested, the amount of product obtained varied. The primers employed (BC4 and BCR12) were designed by comparing the flagellin gene sequences of B. cepacia E243 and B. pseudomallei. There is some degeneracy evident when these primer sequences are compared to the flagellin gene sequence of B. cepacia E242. This may account for the differences in concentrations of the flagellin gene amplified products obtained. It is possible that further fliC sequence data could lead to improved primer design, although the variation exhibited by B. cepacia flagellins may make this impossible. The extent of flagellin sequence variation can be seen from the separation of the 16 B. cepacia isolates tested into 10 RFLP groups on the basis of digests generated by two restriction enzymes. Although seven of the isolates analyzed could not be distinguished, three of them correspond to isolates identified previously as being the major United Kingdom epidemic strain or multilocus enzyme electrophoresis type 12, (ET12 [16]), also known as the Edinburgh/ Toronto lineage (38). We are currently investigating the possibility that the remaining isolates could be further differentiated by employing more enzymes. Butler et al. (6) reported the isolation of 12 B. cepacia environmental isolates, none of which displayed the phenotypic properties of a multiresistant CF epidemic strain with which they were compared. Although current evidence suggests that the environment may pose only a low risk as a source of B. cepacia for CF patients, a more detailed analysis of B. cepacia populations is required.

The flagellin gene is a widely applicable and useful genetic marker for studying variation within populations of closely related bacteria (44). Flagellin gene sequences have been used to study diversity in a number of bacteria, including E. coli (34), Salmonella spp. (22), Campylobacter spp. (3), and Helicobacter pylori (12). Flagellin gene sequence comparisons have been used for phylogenetic analysis of bacterial pathogens such as Borrelia burgdorferi (13) and Listeria monocytogenes (30). It has been suggested that B. cepacia may represent at least three distinct species, and the clarification of Burkholderia taxonomy has been identified as an important step toward the ultimate goal of identifying the pathogenic potential of environmental isolates (16). Flagellin gene RFLP analysis, in conjunction with other methods, may provide a rapid and reliable means to distinguish B. cepacia strains with a view to achieving this goal. A study of larger numbers of isolates, including environmental isolates, and the correlation of flagellin gene RFLP groups with B. cepacia genomovars are required to better assess this possibility.

Flagellin gene sequences have also been shown to serve as sensitive and specific targets for PCR detection or identification of Campylobacter coli and C. jejuni (25), Borrelia spp. (29), Salmonella spp. (40), Listeria spp. (17), and Pseudomonas fluorescens (9). By comparing additional flagellin gene sequences, it may be possible to design similar species-, group-, or strain-specific probes for use with B. cepacia.