Burkholderia cepacia H111, which was isolated from a cystic fibrosis patient, effectively kills the nematode Caenorhabditis elegans. Depending on the medium used for growth of the bacterium two different killing modes were observed. On high-osmolarity medium the nematodes became paralysed and died within 24 h. Using filter assays we provide evidence that this killing mode involves the production of an extracellular toxin. On nematode growth medium killing occurs over the course of 2–3 days and involves the accumulation of bacteria in the intestinal lumen of C. elegans. We demonstrate that the cep quorum-sensing system of H111 is required for efficient killing of C. elegans under both killing conditions. Using the C. elegans phm-2 mutant that has a non-functional grinder evidence is provided that the cep system is required to enter the intestinal lumen but is dispensable for the colonization of the gut. Furthermore, we demonstrate that the type II secretion machinery is not essential for nematode killing.
The Gram-negative bacterium Burkholderia cepacia, which can be isolated from various habitats including soil, water and plant surfaces, was first described as a phytopathogen, associated with a soft rot of onion bulbs (Burkholder, 1950). More recently, B. cepacia has emerged as an opportunistic pathogen of humans, particularly those with cystic fibrosis (CF) (Isles et al., 1984). Burkholderia cepacia is usually acquired late in the course of the disease when the patients are already chronically colonized with the opportunistic pathogen Pseudomonas aeruginosa. The clinical outcome of this co-infection is variable and unpredictable, ranging from asymptomatic carriage to a fulminant and fatal pneumonia, the so-called ‘cepacia syndrome’ (Isles et al., 1984). In contrast to P. aeruginosa very little is known about the pathogenic mechanisms and virulence determinants of B. cepacia.
In P. aeruginosa as well as in many other Gram-negative bacteria expression of virulence factors is not constitutive but is regulated in a cell density-dependent manner. These regulatory systems ensure that pathogenic traits are only expressed when the bacterial population density is high enough to overwhelm the host before it is able to mount an efficient response. To monitor the size of the population P. aeruginosa utilizes a cell–cell communication system that relies on diffusible N-acyl homoserine lactone (AHL) signal molecules in a process known as quorum sensing (for recent reviews see Van Delden and Iglewski, 1998; Williams et al., 2000; de Kievit and Iglewski, 2000). Such communication systems depend on an AHL synthase, usually a member of the LuxI family of proteins, and an AHL receptor protein, which belongs to the LuxR family of transcriptional regulators. At low population densities cells produce a basal level of AHL via the activity of the AHL synthase. As the cell density increases, the diffusible AHL signal molecule accumulates in the growth medium. On reaching a critical threshold concentration, the AHL binds to the cognate LuxR-type receptor protein, which in turn leads to the induction/repression of target genes. The pivotal role of quorum sensing for the pathogenicity of P. aeruginosa has been demonstrated in a number of animal models including the neonatal mouse model of pneumonia (Tang et al., 1996), the burned mouse model (Rumbaugh et al., 1999), and a Caenorhabditis elegans model (Tan et al., 1999).
Recent work has identified a quorum-sensing system in B. cepacia consisting of the N-acyl homoserine lactone (AHL) synthase CepI, which directs the synthesis of N-octanoylhomoserine lactone (C8-HSL) and, as a minor product, N-hexanoylhomoserine lactone (C6-HSL) (Lewenza et al., 1999; Gotschlich et al., 2001), and CepR, which after binding of C8-HSL is thought to activate or repress transcription of target genes. The cep system positively regulates biofilm formation and extracellular proteolytic and chitinolytic activity, but represses synthesis of the siderophore ornibactin (Lewenza et al., 1999; Huber et al., 2001; 2002; Lewenza and Sokol, 2001). This study was initiated to investigate the importance of the cep quorum-sensing system for the pathogenicity of B. cepacia.
Evidence that has accumulated over the past few years has established that the nematode Caenorhabditis elegans is a highly valuable model for the study of bacterial pathogenicity (for recent reviews see Aballay and Ausubel, 2002; and Ewbank, 2002). Initially this model was developed to identify genes in P. aeruginosa that are important for pathogenesis (Tan and Ausubel, 2000). Depending on the strain and the culture conditions P. aeruginosa kills C. elegans by at least three distinct mechanisms. When P. aeruginosa strain PA14 is grown on nematode growth (NG) medium cells colonize the nematode intestine and killing occurs over the course of 2–3 days (‘slow killing’). In contrast, PA14 grown on high-osmolarity medium (PGS) kills worms much faster (within 4–24 h –‘fast killing’) due to the production of low-molecular-weight toxins. Another P. aeruginosa strain, PAO1, was shown to kill the nematode by cyanide poisoning (‘paralytic killing’) when it is grown on brain–heart infusion (BHI) broth (Gallagher and Manoil, 2001). Recent work has shown that a number of pathogens including Salmonella typhimurium (Aballay et al., 2000; Labrousse et al., 2000), Enterococcus faecalis (Garsin et al., 2001), Serratia marcescens (Kurz and Ewbank, 2000), and several members of the genus Burkholderia kill C. elegans (O’Quinn et al., 2001; Gan et al., 2002). In this study we used the C. elegans model system to investigate the role of quorum sensing in pathogenesis of B. cepacia.
Results and discussion
Burkholderia cepacia H111 kills C. elegans
In accordance with the results reported by O’Quinn et al. (2001) we observed that when the usual food source (Escherichia coli OP50) was replaced by B. cepacia H111, a member of the genomovar III of the B. cepacia complex (Gotschlich et al., 2001), C. elegans had a significantly shorter life span (Fig. 1). When grown on NG medium worms died over a period of 1–3 days (Fig. 1A). To investigate the fate of ingested bacteria in better detail we constructed a H111 strain expressing the Aequorea victoria green fluorescent protein (GFP). Microscopic inspection after 24 h of feeding on GFP-tagged H111 revealed that green fluorescent cells were present throughout the lumen of worm intestines, which often appeared markedly distended (Fig. 2A and B). These data suggest that H111 is able to proliferate within the nematode gut and thus slow-killing resembles an infection-like process. In contrast, when the nematodes were fed with E. coli MT102 expressing GFP, green fluorescence was restricted to the pharynx and only very few cells were observed within the gut (Fig. 2E). When the strain was grown on the high-osmolarity PGS medium, killing of the worms occurred within 24 h (Fig. 1B) and no fluorescent cells were detectable in the intestines of the nematodes when the GFP-tagged wild-type strain was used as food source (data not shown), indicating that the worms may be poisoned by factors exported by B. cepacia H111.
These results are reminiscent of the slow and fast killing mode of pathogenesis observed with P. aeruginosa PA14 (Tan et al., 1999). In the case of PA14 it was shown that fast killing is mediated by the production of a number of diffusible toxins (Mahajan-Miklos et al., 1999). To test whether H111 is producing toxins that are responsible for the rapid kinetics of killing on PGS medium we performed filter killing assays (Fig. 3). To this end, 100 µl of an overnight culture of B. cepacia H111 were spread on nitrocellulose filters covering the surface of PGS plates. Following incubation for 24 h at 37°C the filters were removed and the worms were placed on the plates. Microscopic inspection revealed that the worms became paralysed on this conditioned medium and approximately 30% died after 24 h (data not shown). Thereafter, however, the surviving worms started moving again and eventually began to proliferate. Importantly, when filter assays were performed on NG medium instead of PGS medium no paralysis or death of the nematodes could be observed.
Recent work has shown that P. aeruginosa PAO1 provokes a neuromuscular paralysis of C. elegans, which causes the death of the nematodes (Darby et al., 1999; Gallagher and Manoil, 2001). The primary toxic factor responsible for this ‘paralytic killing’ was identified as hydrogen cyanide (Gallagher and Manoil, 2001). We therefore tested whether B. cepacia H111 is capable of producing hydrogen cyanide. However, we were unable to detect hydrogen cyanide production irrespective of the medium (NG, PGS and BHI) used for growing H111. By contrast, two P. aeruginosa strains, PAO1 and AC869, which were included as controls in these experiments, produced significant amounts of hydrogen cyanide when grown on BHI agar, but not on NG or PGS agar (data not shown).
From these experiments we conclude that B. cepacia H111 produces a diffusible toxin on PGS medium, which paralyses and kills the nematode. This toxin, which is not hydrogen cyanide, appears to be unstable, as upon prolonged incubation the nematodes resumed reproduction. Interestingly, killing of C. elegans by B. pseudomallei is also mediated, at least in part, by a diffusible toxin (Gan et al., 2002). However, as maximal killing efficiency required live bacteria it has been suggested that the toxin is unstable and requires continuous production.
Mutants of B. cepacia H111 defective in quorum sensing have an impaired ability to kill C. elegans
To investigate whether the cep quorum-sensing system of B. cepacia H111 is required for killing of C. elegans, we tested the cepI mutant H111-I and the cepR mutant H111-R in the two nematode-killing models. Both strains were attenuated in their virulence against the nematode under both slow-killing (Fig. 1A) and fast-killing (Fig. 1B) conditions. When 200 nM C8-HSL was added to the medium the killing ability of mutant H111-I was restored, albeit with a reduced fast-killing rate. One possible explanation for the different killing kinetics could be that efficient fast-killing requires a specific temporal expression pattern of nematocidal traits. Importantly, the presence or absence of 200 nM C8-HSL had no effect on the developmental cycle or life span of the nematode (data not shown). When mutant H111-R was complemented with plasmid pBAH27 (cepR+) killing kinetics were virtually indistinguishable from those of the wild type in both nematode killing models. As expected from their attenuated virulence phenotypes we did not observe any green fluorescent cells in the nematode guts when GFP-tagged derivatives of H111-I were used as food source on NG medium (Fig. 2C and D). As the grinder in the pharynx of the nematode is a major barrier to the entry of bacteria into the intestine we next investigated whether the attenuated pathogenicity of the cep mutants is uniquely a problem of penetration. Noteworthy in this context is the fact that the grinder is largely made of chitin and production of extracellular chitinolytic activity by B. cepacia is controlled by the cep quorum-sensing system (Huber et al., 2001). To address this issue we performed slow-killing experiments with the C. elegans phm-2 mutant that has a non-functional grinder (Avery, 1993). We observed that GFP-expressing H111-I cells accumulated in the intestinal lumen of the nematode mutant (Fig. 2F), indicating that the ability to pass the grinder is indeed dependent on a functional cep system. To investigate whether the cepI mutant was capable of stably colonizing the intestine of the nematode mutant we fed C. elegans phm-2 with GFP-tagged B. cepacia H111-I for 48 h and then transferred the nematodes to plates containing E. coli OP50. Twenty-four hours after the transfer, green fluorescent cells could still be found in the intestine suggesting that the cep system is not essential for the colonization of the nematode gut (data not shown). Most interestingly, however, the cepI mutant still gave a significantly lower rate of C. elegans phm-2 killing relative to the wild-type strain (Fig. 4). These data indicate that the ability to colonize of the nematode gut is not entirely sufficient for nematode killing and that expression of yet unknown quorum sensing-regulated function(s) is required for efficient slow-killing.
In summary, these results show that the cep quorum-sensing system of B. cepacia H111 plays a pivotal role in the regulation of nematocidal traits under both slow-killing and fast-killing conditions. Interestingly, the cepR mutant H111-R, although being clearly attenuated when compared with the wild type, is more pathogenic in both killing models than the cepI mutant H111-I. A recent analysis of the cep quorum-sensing regulon by proteomics showed that CepR functions not only as a transcriptional activator but also as a repressor in B. cepacia H111 (Riedel et al., 2003). We therefore speculate that a negatively regulated virulence factor is upregulated in H111-R, which partly compensates for the lack of positively regulated pathogenic traits. In fact, Lewenza and Sokol (2001) demonstrated that a cepR mutant of B. cepacia K56-2 hyperproduces ornibactin, a siderophore shown to be important in both chronic and acute models of respiratory infection (Sokol et al., 1999; 2000).
Type-II-pathway-secreted extracellular proteins are not required for killing of C. elegans
The fact that the cep system controls expression of extracellular hydrolytic enzymes including a protease (Lewenza et al., 1999; Huber et al., 2001), which is considered to represent an important virulence factor of B. cepacia, prompted us to investigate the role of this enzyme in killing of C. elegans. For this purpose we screened a collection of 3000 random insertion mutants of B. cepacia H111 for lack of protease activity. Genetic analyses showed that in four of the isolated mutants the transposon had inactivated genes encoding components of the type II secretion machinery (see Experimental procedures for details), which is part of a two step secretion pathway, known as the general or type II secretion pathway (Pugsley, 1993). The deduced amino acid sequences showed the highest homologies to proteins of the type II secretion machinery of Burkholderia pseudomallei, which utilizes the general pathway to secrete protease, lipase and phospholipase (DeShazer et al., 1999). Consistent with these data we found that the isolated mutants no longer produce extracellular proteolytic and lipolytic activity (Table 1). All isolated mutants killed C. elegans as efficiently as the wild type under both slow-killing and fast-killing conditions (Fig. 5). Hence, type-II-pathway-secreted extracellular proteins of B. cepacia, including protease and lipase, are not essential for killing of C. elegans. These results are in agreement with the work of O’Quinn et al. (2001) who showed that the general secretion machinery of B. pseudomallei is not required for nematode killing.
Table 1. . Phenotypic characteristics of B. cepacia wild-type H111 and mutants used in this study.
Production of AHLs was determined by testing culture supernatants for stimulation of the bioluminescent sensor plasmid pSB403. Synthesis of extracellular hydrolytic enzymes was assessed by streaking strains on appropriate indicator plates as described in Experimental procedures. +, activity exhibited by the wild type; <, significantly reduced activity; –, no detectable activity; NA, not applicable.
H111-I + 200 nM C8-HSL
The clinical B. cepacia isolate H111 (genomovar III) effectively kills the nematode C. elegans. By analogy to P. aeruginosa PA14 we defined two different killing modes, depending on the medium used for growth of B. cepacia H111. Fast killing occurs within 24 h when the strain is grown on the high-osmolarity medium PGS. This killing mode involves the production of an extracellular toxin, which paralyses and eventually kills the nematode. Similar to the diffusible toxin produced by B. pseudomallei (Gan et al., 2002) this exotoxin appears to be unstable and as a consequence maximum killing of C. elegans requires live bacteria. By contrast, slow killing occurs over the course of 2–3 days when the strain is grown on nematode growth medium NGM and involves the accumulation of bacteria in the intestinal lumen of C. elegans.
We have shown that the cep quorum-sensing system of B. cepacia H111 is required for killing of C. elegans under both slow-killing and fast-killing conditions, whereas the type II secretion machinery is not essential for nematode killing. This finding excludes the possibility that extracellular protease and lipase, which are both secreted via the type II pathway, contribute to the virulence of the bacterium against C. elegans. Whereas the importance of AHL-mediated cell–cell communication in the regulation of virulence factors has been demonstrated for P. aeruginosa in various pathogenesis models (Rumbaugh et al., 2000), this is the first report that provides evidence that expression of pathogenic traits in B. cepacia is quorum sensing-regulated. However, further work will be required to identify those cep-regulated factors that are directly involved in killing of C. elegans.
Organisms and culture conditions
Bacterial strains and plasmids used in this study are listed in Table 2. Unless otherwise stated, the strains were grown at 37°C in modified Luria–Bertani (LB) broth (Bertani, 1951) containing 5 g of NaCl l−1 instead of 10 g of NaCl l−1. Solid media were routinely solidified with 1.5% agar. Antibiotics were added as required at final concentrations of 20 µg ml−1 gentamicin, 50 µg ml−1 kanamycin, 10 µg ml−1 tetracycline and 10 µg ml−1 choramphenicol. Synthetic C8-HSL was purchased from Fluka and was used at a final concentration of 200 nM. Growth of liquid cultures was monitored spectrophotometrically by an Ultrospec Plus spectrophotometer (Pharmacia) by measurement of optical density at 600 nm (OD600). Killing assays were performed using Caenorhabditis elegans strain Bristol N2 or the C. elegans phm-2 mutant, which has a non-functional grinder. Both strains were obtained from the Caenorhabditis Genetics Center (University of Minnesota, St Paul, MN, USA). The nematode was maintained on NG agar (Tan et al., 1999) at 20–22°C with E. coli OP50 as food source (Brenner, 1974). Killing assays were either performed on NG medium or on high-osmolarity PGS medium (1% Bactopeptone, 1% NaCl, 1% glucose, 0.15 M sorbitol and 1.7% Bacto-Agar).
Table 2. . Bacterial strains and plasmids used in this study.
The hybrid transposon mini-Tn5 Km2-luxCDABE was randomly inserted into the chromosome of B. cepacia H111 by the triparental mating procedure described below. Transconjugants were selected on LB medium containing kanamycin and tetracycline. These random insertion mutants were picked and grown in 150 µl LB medium in the wells of polypropylene MicroWell dishes (Nunc). For storage 75 µl of 50% (v/v) glycerol were added and the dishes frozen at − 80°C.
Isolation of type II secretion mutants
A collection of 3000 random insertion mutants of B. cepacia H111 was screened for protease production on LB medium supplemented with 2% skim milk. 11 protease negative mutants were obtained in this screen. Southern blot analysis of the 11 mutants using a DIG-labelled 600 bp DNA fragment derived from the kanamycin resistance gene of mini-Tn5 Km2-luxCDABE as a probe showed that single copies of the hybrid transposon had inserted at different chromosomal locations (data not shown). As production of extracellular proteolytic activity is controlled by the cep system we tested all mutants for AHL production. Three mutants no longer produced significant amounts of AHL signal molecules and these mutants were not investigated further. DNA flanking the transposon insertion in each of the remaining mutants was amplified by arbitrary PCR (see Experimental procedures for details). The DNA sequences of the amplicons were analysed using the blastx program (Altschul et al., 1990), which translates the DNA sequences in all six reading frames and compares these predicted protein sequences with those in GenBank. This analysis showed that in four of the investigated mutants the transposon had inactivated genes encoding components of the type II secretion machinery.
Conjugative plasmid transfer
Plasmids were delivered to B. cepacia by triparental mating as described previously (de Lorenzo and Timmis, 1994). Briefly, donor and recipient strains as well as the helper strain E. coli HB101 (pRK600) were grown at 37°C overnight in 5 ml LB supplied with the appropriate antibiotics. Following subculturing to an OD600 of 0.9, the cells from 2 ml of culture were harvested, washed, and resuspended in 200 µl LB. Donor and helper cells (100 µl each) were mixed and incubated for 30 min at room temperature. Then 200 µl of recipient cells were added and the mixture was spot-inoculated onto the surface of prewarmed LB agar plates. After overnight incubation at 37°C, the cells were scraped off and were resuspended in 1 ml 0.9% NaCl. Serial dilutions were plated on LB medium containing antibiotics for counter-selection of donor, helper and untransformed recipient cells.
DNA manipulations and sequence analysis of Tn5 mutants
Cloning, restriction enzyme analysis and transformation of E. coli were performed essentially as described previously (Sambrook et al., 1989). Polymerase chain reaction (PCR) was performed using the TaKaRa rTaq DNA polymerase (TaKaRa Shuzo). Plasmid DNA was isolated with the QIAprep Spin Miniprep kit and chromosomal DNA from B. cepacia was purified with the DNeasy Tissue kit. DNA fragments were purified from agarose gels using the QIAquick Gel Extraction kit (all kits were from Qiagen).
The DNA sequences flanking transposon mutants were determined by arbitrary PCR as described previously (O’Toole and Kolter, 1998), with a few modifications. Briefly, we performed two rounds of PCR amplification using arbitrary primers to prime from the chromosome and primers specific to the mini-Tn5 transposon. Primers used in the first round were ARB6 (5′-GGCCACGCGTCGACTAGTACNNNNNNNNNNACGCC-3′) and luxCext2 (5′-AGTCATTCAATATTGGCAGG-3′). First-round reaction conditions were (i) 5 min at 95°C; (ii) 6 × [30 s at 95°C, 30 s at 30°C, 1 min at 72°C]; (iii) 30 × (30 s at 95°C, 30 s at 45°C, 1 min at 72°C); (iv) 5 min at 72°C. The second round of PCR amplification used 5 µl purified first round PCR product and the following primers: ARB2 (5′-GGCCACGCGTCGACTAGTAC-3′) and luxCint2 (5′-GGATTGCACTAAATCATCAC-3′). Second-round reaction conditions were (i) 30 × (30 s at 95°C, 30 s at 45°C, 1 min at 72°C); (ii) 5 min at 72°C. The PCR products were purified from an agarose gel and ligated into the vector pCR 2.1-TOPO.
Sequencing was performed by the dideoxynucleotide chain determination method in a LI-COR 4200 DNA sequencer. The primer 5′-CACTTGTGTATAAGAGTCAG-3′, which binds to the O-end of the mini-Tn5 transposon, was used for determination of the transposon insertion point. DNA sequences were compared to other sequences in GenBank using the online blast search engine at the Center for Biotechnology Information (http:www.ncbi.nlm.nih.gov).
Nematode killing assays
Burkholderia cepacia test strains were grown in 5 ml LB liquid cultures for 4–5 h or overnight with shaking. The cultures were adjusted to a density of about 1.3–1.5 × 104 CFU ml−1 and 100 µl of the suspensions were plated on 6.0-cm diameter NG (slow killing) or PGS (fast killing) agar plates. After 24 h of incubation at 37°C the plates were allowed to cool to 23°C. Approximately 25 hypochlorite synchronized L4 larvae or adult worms were collected from M9 buffer (Stiernagle, 1999) and used to inoculate the plates. The actual number of worms per plate was determined using a Stemi SV 6 microscope (Zeiss) at × 50 magnification (time point 0). Plates were then incubated at 23°C and scored for live worms at the indicated time points. Nematodes were considered dead when they failed to respond to touch. All experiments were carried out three to five times and E. coli OP50 was used as a control in the assays.
Toxin filter diffusion assay
To test whether fast killing is caused by the production of diffusible toxins, 100 µl of an overnight LB culture was spread onto a sterilized nitrocellulose filter (0.025 µm pores) which was placed on PGS agar. Following incubation for 24 h at 37°C the filter together with the entire bacterial lawn was removed and 15–25 hypochlorite-synchronized C. elegans L4-stage larvae were spotted onto the conditioned agar. Movement and mortality of the worms was monitored at 4 and 24 h. E. coli OP50 was used in place of B. cepacia for negative controls.
Exoenzyme and cyanide production
Exoenzyme production was tested by streaking strains on appropriate indicator plates. Proteolytic activity was determined on LB medium supplemented with 2% skim milk and lipolytic activity on tributyrin agar base containing 1% glycerol tributyrate (both Merck). Clear halos around the colonies after incubation at 37°C overnight indicated exoenzyme activity.
Cyanide production was measured according to the procedure described by Gallagher and Manoil (2001). Strains were grown on 3 cm diameter BHI, NG or PGS agar plates for 24 h at 37°C. Then the agar plates were placed without lids into an 8 cm diameter Petri dish containing a reservoir of 1 ml of 4 M NaOH. The larger Petri dish was closed with a lid and sealed with parafilm. After 4 h of incubation at 30°C, the NaOH was collected and diluted to 0.09 M with double-distilled H2O. If necessary the samples were further diluted with 0.09 M NaOH to bring the cyanide concentration to within the linear range of the detection procedure (0–1 nM). The cyanide in the sample was quantified by comparison with standards of KCN dissolved in 0.09 M NaOH: 210 µl aliquots of the samples were mixed with 700 µl aliquots of a fresh 1 : 1 mixture of 0.1 M o-dinitrobenzene (Sigma) in ethylene-glycol monomethyl ether (Sigma) and 0.2 M p-nitrobenzaldehyde (Sigma) in ethylene-glycol monomethyl ether. After 30 min of incubation at 20°C the OD578 was measured.
This work was supported by grants from the BMBF and the Deutsche Forschungsgemeinschaft (EB 2051/1–3).