|Home | About | Journals | Submit | Contact Us | Français|
The bacterium Burkholderia cenocepacia is pathogenic for sufferers from cystic fibrosis (CF) and certain immunocompromised conditions. The B. cenocepacia strain most frequently isolated from CF patients, and which serves as the reference for CF epidemiology, is J2315. The J2315 genome is split into three chromosomes and one plasmid. The strain was sequenced several years ago, and its annotation has been released recently. This information should allow genetic experimentation with J2315, but two major impediments appear: the poor potential of J2315 to act as a recipient in transformation and conjugation and the high level of resistance it mounts to nearly all antibiotics. Here, we describe modifications to the standard electroporation procedure that allow routine transformation of J2315 by DNA. In addition, we show that deletion of an efflux pump gene and addition of spermine to the medium enhance the sensitivity of J2315 to certain commonly used antibiotics and so allow a wider range of antibiotic resistance genes to be used for selection.
Burkholderia cenocepacia is part of the Burkholderia cepacia complex (Bcc), a group of closely related bacteria of soil, water, and roots (41) recently updated to at least 15 related species (42). Bcc displays many interesting features (see reference 27 for a review). Originally discovered as responsible for soft onion rot (3), Bcc species also interact beneficently with plants (see reference 34 for a review) and may degrade pollutants such as phthalate or the herbicide 2,4,5-trichlorophenoxyacetic acid (2,4,5,-T) (25, 33). But it is the emergence of Bcc as an opportunistic pathogen of people suffering from cystic fibrosis (CF) (19) and immunocompromizing conditions that has drawn most attention to these bacteria. Among Bcc species, Burkholderia multivorans and B. cenocepacia are the most prevalent in the epidemiology of CF. In particular, strains of the ET12 lineage of B. cenocepacia were responsible for a major transcontinental epidemic among CF patients in the 1990s (20), an outbreak aggravated by the high levels of resistance to nearly all antibiotics that characterizes Bcc. Species of the Bcc have large genomes (7 to 9 Mb) composed of two or three chromosomes and one or more plasmids, an unusual genomic organization among bacteria. The first Bcc genome to be sequenced was that of B. cenocepacia J2315 (also known as LMG16656), the type strain of the ET12 lineage and the reference strain for CF epidemiology; the sequence was completed and made available by the Wellcome Trust Sanger Institute in 2003. It revealed three chromosomes of 3.9, 3.2, and 0.9 Mb and a plasmid of 93 kb. The annotation of this genome was released recently (15).
The pathogenicity and multipartite genome of B. cenocepacia make it an important subject for both practical and fundamental study. Genetic modification is essential to the success of many such investigations. Unfortunately, J2315 throws up major barriers to genetic manipulation. Standard electrotransformation techniques are ineffective with this strain, as also found elsewhere (26). Conjugal introduction of DNA has proved unreliable despite adaptations (7) that have enabled occasional successes with B. cenocepacia species (9, 40) including J2315 (39) (see also Results below). Besides, the natural resistance of J2315 to antibiotics, high even on the scale of the generally extensive resistance of B. cenocepacia species (31), severely restricts the use of antibiotic resistance in genetic selections. Circumventing these problems by resorting to a proxy strain, B. cenocepacia K56-2, that has not been sequenced and is more permissive to gene transfer (26, 17, 32, 9) runs the risk that results will be of uncertain relevance to J2315.
In the context of our general aim to decipher the role of the four replicon-specific ParABS systems of J2315 (6), we have sought to overcome these obstacles. We report here the reproducible electrotransformation of J2315, and we analyze factors that improve its efficiency. We report also our isolation of a J2315 derivative with reduced antibiotic resistance and the broadened selection possibilities this offers. Detailed protocols are provided which should facilitate studies of this pathogen.
Plasmids and strains are described in Table Table1.1. Luria Bertani (LB) medium was used for routine growth of both Escherichia coli and B. cenocepacia, and super optimal broth (SOB) medium was used for preparation of electrocompetent cells.
The medium for overnight growth was SOB containing 10 mM MgSO4 and 10 mM MgCl2, supplemented with 10 μg/ml gentamicin to minimize growth of contaminants. Five milliliters of this medium was inoculated with 50 μl of a culture of B. cenocepacia J2315 (optical density at 600 nm [OD600] of ~1; freshly prepared from a frozen stock) and incubated overnight with aeration at 37°C. The overnight culture reached an OD600 of 3. (Excessively long incubation, recognized by browning of the medium due to production of melanin, should be avoided because it results in slow and unpredictable growth.) The overnight culture was then diluted to an OD600 of 0.01 in 50 ml of SOB medium containing 10 mM MgSO4, 10 mM MgCl2, and 10 μg/ml gentamicin and supplemented with 0.8% glycine from a filter-sterilized 20% glycine stock solution. Next, the culture was incubated with aeration for 2 to 3 generations to an OD600 of 0.04 to 0.08, which takes 4 to 6 h. The culture was then chilled for 5 min and centrifuged at 4°C for 8 min at 5,000 × g. The cells were washed twice by gentle resuspension in 20 ml of ice-cold 0.5 M sucrose, followed by centrifugation as above. The washed cells were resuspended in 0.5 ml of 0.5 M sucrose-10% glycerol and either used directly for electrotransformation or aliquoted and frozen at −80°C. The loss of electrotransformation frequency after freezing and thawing was minor, and for convenience we usually employed frozen cells. Electrotransformation followed the standard procedure, with the following specific measures: DNA was added as ~0.1 μg of replicative plasmid or ~5 μg of nonreplicating DNA (for chromosomal recombination) in 1 to 15 μl of water (a volume variation with no effect on standard transformation frequency [TF]) to 55 μl of thawed competent cells. The DNA-bacteria mixture was subjected to a pulse of 2,500 V and 200 μF (using an Eppendorf 2510 electroporator); 950 μl of SOC medium (SOB medium supplemented with MgSO4 [10 mM] and glucose [2%]) was added, and the cells were incubated, standing, at 37°C for at least 4 h before being spread on selective LB or SOB medium, followed by incubation at 37°C for two (SOB) or three (LB) days.
The mating procedure was adapted from Engledow et al. (7). Overnight cultures of the donor, helper, and recipient strains were grown in LB medium supplemented with appropriate antibiotics (kanamycin for the E. coli helper carrying pRK2013; chloramphenicol, trimethoprim, or tetracycline for the E. coli donor; gentamicin for B. cenocepacia). Fresh overnight cultures of the E. coli strains and J2315 were diluted 200-fold and 100-fold, respectively, into LB medium without antibiotics, and the cultures were grown to an OD600 of ~0.5; E. coli cultures that reached this density early were maintained at 37°C without shaking. A total of 400 μl of each of the three cultures was mixed in 10-ml plastic tubes and left to stand at 37°C without shaking for 1 h. One milliliter of the mixture was poured onto a positively charged nylon membrane (Membrane Biodyne B; 0.45-μm pore size, nylon >0; VWR), placed on an LB agar plate, and incubated for 15 h at 37°C. Bacteria were then scraped from the membrane and suspended in 1 ml of LB medium. Samples of 100 μl and 900 μl were spread on LB agar medium containing antibiotics to select for the donor plasmid and counter-select (gentamicin) for E. coli, and samples were incubated 2 to 3 days at 37°C.
Antibiotic discs for the antibiogram diffusion test were from Sanofi Pasteur Diagnostics. Spermine and all antibiotics except nalidixic acid were from Sigma Aldrich. Nalidixic acid was from Calbiochem. Stock solutions of spermine (1 M) and of ampicillin, gentamicin, kanamycin, streptomycin, and spectinomycin (100 mg/ml) were made in distilled water and filter sterilized. Stocks of trimethoprim (100 mg/ml) and tetracycline (200 mg/ml) were made in dimethyl sulfoxide (Merck). Stocks of erythromycin and chloramphenicol were made at 100 mg/ml in absolute ethyl alcohol (VWR). A stock of nalidixic acid was made at 20 mg/ml in 0.1 M NaOH and filter sterilized.
An antibiogram diffusion test (Sanofi Pasteur Diagnostics) was carried out to determine the approximate sensitivities of J2315 and its mex1 derivative. Colonies grown on Mueller-Hinton (MH) medium were suspended in physiological saline at ~106 bacteria per ml. Petri dishes containing solid MH medium were flooded with 3 ml of suspension; excess liquid was discarded, and the plates were allowed to dry for 15 min at room temperature. Antibiotic-carrying discs were applied, and the plates were incubated at 37°C for 20 h. The diameter of the complete growth inhibition zone, including the central antibiotic-carrying disc and the clear halo surrounding it, was measured with a graduated ruler put directly on the petri dish. The growth inhibition was defined as the width of the clear halo, which was calculated by subtracting the disc diameter (7 mm) from the measured zone and dividing by two. The values obtained allowed us to classify the bacteria as resistant, intermediate, or sensitive, according to supplier and clinical recommendations.
We also tested antibiotic sensitivity by broth dilution. Briefly, fresh LB cultures of B. cenocepacia grown to an OD600 of 0.1 to 0.3 were diluted in LB broth to ~5 × 106 bacteria/ml. Aliquots of the dilute suspension were added to 5-ml sterile capped tubes. Antibiotic was added to one tube, and the mixture was transferred from this into the others to make serial 2-fold dilutions. Tubes were incubated 18 h at 37°C with shaking. The MIC in LB medium (MICLB) is the lowest concentration of each antibiotic that completely inhibits growth under these conditions.
Plasmids were extracted from E. coli and B. cenocepacia using a miniprep kit (Qiagen). PCRs were carried out with Phusion high-fidelity DNA polymerase (Finnzymes). Oligonucleotides were made by Sigma and Eurogentec. Oligonucleotide sequences are given in Table Table2.2. Restriction enzymes were from New England Biolabs; 2-log and 1-kb DNA ladders were from New England Biolabs; a 1-kb Plus DNA ladder was from Gibco BRL. To delete the mex1 locus (6.6 kb) two flanking amplicons of 1,152 bp (primers mexA and mexB) and 1,720 bp (primers mexC and mexD), sharing 34 bp of homology on their respective B and C ends, were mixed and used as the template for a third PCR primed with oligonucleotides A and D. The resulting amplicon of 2.8 kb, corresponding to the fusion of the left and right flanks of the mex1 locus, was digested by HindIII and XbaI and inserted between the corresponding sites of pEX18Tc. The resulting plasmid was introduced into J2315 by triparental mating. Tetracycline resistant (Tetr) exconjugants were selected, and integration of the plasmid was confirmed by PCR. Plasmid excision was detected by replica plate screening for loss of Tetr following nonselective growth. Tets clones were analyzed by PCR to identify those from which mex1 had been deleted.
The plasmid pCM351-cat was constructed by PCR amplification (with oligonucleotides Fop106 and Fop-107) of all but the Genr cassette of pCM351, followed by XhoI-PstI digestion and ligation of the resulting amplicon to the chloramphenicol acetyltransferase (CAT) cassette amplified from pBR325 (oligonucleotides cat-up and cat-dwn) and similarly digested. DNA flanking the sequence to be deleted can be cloned on both sides of CAT, allowing the selection of exchange events that replace the sequence with CAT.
The pBL2 plasmid was obtained by cloning into pBR325 (digested with EcoRV and BamHI) a 1.4-kb fragment spanning BCAL0028 amplified with oligonucleotides ctg-E5 and ctg-B1. Plasmid pDAG824 is a pCM351-cat derivative, carrying both the parS locus of phage P1 and a 2.4-kb fragment of B. cenocepacia chromosome 3 (c3) origin (obtained by PCR with oligonucleotides tic-A and tic-B and then digested with XmnI and AatII and inserted into pCM351-cat). Plasmid pRF91 is a pCM351-cat derivative carrying on each side of CAT the left and right 1.3-kb flanks of parA chromosome 1 ([c1] amplified with dpa-1/dpa-2 and dpa-3/dpa-4, respectively) (Table (Table22).
The standard E. coli electrotransformation procedure yielded no transformants when it was applied to B. cenocepacia. Because Gram-positive bacteria are more refractory to electrotransformation than Gram-negative bacteria, we adapted a protocol designed for the Gram-positive Leuconostoc carnosum 4010 (13). The main differences of this protocol from the E. coli procedure are as follows: (i) inoculation of the culture at a lower cell density, (ii) fewer generations of growth before harvesting, (iii) addition of glycine to the culture medium, (iv) gentle centrifugation and resuspension during cell washing, and (iv) longer phenotypic expression. Using the protocol described in Materials and Methods, we were able to transform J2315 with plasmid DNA extracted from wild-type E. coli. As with L. carnosum (13), the duration of phenotypic expression before selection is important. Few or no transformants were obtained if electroporated cells were incubated for times shorter than 2 h; the number of transformants per viable cell was 2.5- to 5-fold higher at 4 h than at 2 h, and extending incubation to 6 h or more did not further increase this ratio. By routinely allowing 4 to 5 h of phenotypic expression, we obtained up to 104 transformants per μg of plasmid DNA.
To improve TF further, we examined additional factors. In each case the data (Fig. (Fig.1)1) are presented as ensembles of all the independent experiments for which valid comparisons of TF can be made.
Inclusion of glycine in the growth medium of Gram-positive bacteria has been observed to weaken the thick cell wall and to increase transformation efficiency (16). We therefore tested its effect on the TF of J2315. Glycine inhibited growth in LB and SOB media modestly at 0.8% but strongly at higher concentrations. Cells grown with 0.8% glycine were transformed with plasmid DNA 2- to 6-fold (and up to 25-fold) more efficiently than cells grown in parallel without it (Fig. (Fig.1A).1A). The improvement in TF conferred by glycine suggests that cell wall structure could contribute to the poor competence of J2315.
Burns and Hedin (4) reported that strain 249-2 of Pseudomonas cepacia (the original name of the Bcc) could be electrotransformed by plasmid DNA if it had been extracted from an E. coli dam dcm strain or from Pseudomonas aeruginosa and was thus unmethylated on GATC sites. We found electrotransformation of J2315 with plasmid DNA extracted from the dam dcm strain SCS110 to be 10 to 20 times more efficient than with methylated DNA (Fig. (Fig.1B).1B). These data suggest that J2315 possesses a restriction system specific for methylated GATC sites.
The bacteriophage T7 protein, Ocr, acts as a decoy to inhibit an attack by a type I restriction endonuclease on entering DNA by mimicking DNA structure (28, 1). Commercially available Ocr (Type One Restriction Inhibitor; Epicentre) may improve up to 100-fold the TF of restricting bacteria by unmodified DNA. We tested the effect of including 5 μg of Ocr protein with the DNA of our tester plasmid pMMB206 on electrotransformation of J2315. Ocr improved the TF by up to 20-fold but on average only 3-fold (Fig. (Fig.1C).1C). This is far from the 100-fold factor expected for inhibition of a type I endonuclease. Increasing the amount of Ocr did not help; ≥10 μg impairs transformation in E. coli or Salmonella enterica serovar Typhimurium according to the supplier, and we observed it to reduce (by ~0.2 ms) the time constant of the electric pulse. Interestingly, we found the same 3-fold Ocr-mediated increase whether DNA was extracted from DH10B (dam+ dcm+), SCS110 (dam dcm) or J2315 (Fig. (Fig.1C),1C), suggesting that Ocr can act beyond the type I restriction system to block nonspecific DNase activity. Our data suggest that homologues of type I restriction/modification systems identified in J2315 (e.g., BCAL0418 and BCAL0420 on chromosome 1) do not act on our tester plasmid DNA.
Even though the Ocr effect is weaker than expected, it adds to the effect produced by demethylation of DNA. The highest TF we have observed, 1.2 ×105 per μg, was with cells from cultures grown with glycine and transformed with unmethylated pMMB206 DNA in the presence of Ocr.
The electrotransformation protocol established (see Materials and Methods) allows the routine introduction of replicative plasmids into J2315. We tested whether it was efficient enough to enable the isolation of J2315 genome recombinants. In one test, an amplicon of 1.4 kb spanning the locus BCAL0028, a putative citrate transporter gene (ctg) located downstream of the operon gidAB-parAB in chromosome 1, was inserted into the vector pBR325 (Cmr; nonreplicative in B. cenocepacia). The resulting plasmid, pBL2 (see Materials and Methods), was demethylated by passage through SCS110 and introduced by electrotransformation into J2315. Integration was confirmed by PCR using oligonucleotide pairs P1r/P1f and P8r/P8f (Table (Table22 and Fig. Fig.2A).2A). Both chromosome-plasmid junctions were detected as amplicons of 2.5 kb (Fig. (Fig.2A.).2A.). In a second test, pDAG824, a derivative of pCM351 carrying a 2.4-kb fragment of the replication origin region of chromosome 3 (see Materials and Methods), was introduced by electrotransformation into J2315 mex1 (see below). Integration of pDAG824 at the expected site of a Cmr Tetr transformant in chromosome 3 was confirmed by the generation of an amplicon of 2.89 kb using PCR with oligonucleotides vrfA/vrfB (Fig. (Fig.2B).2B). In a third trial, the parA gene of chromosome 1 was deleted. The pRF91, a pCM351-cat dervivative, carries two amplicons of 1.3 kb corresponding to the flanks of chromosome 1 parA (see Materials and Methods) on either side of cat. Each of three independent electrotransformations of a Δmex1 derivative of J2315 mex1 with pRF91 yielded ~10 Cmr Tetr transformants, with the plasmid integrated via a single crossover at the parA locus, as confirmed by PCR with oligonucleotides F34/F35 (Fig. (Fig.2C,2C, lanes 8 and 9). In addition, one of the transformations gave a Cmr Tets clone with parA replaced via a double crossover by cat, while another gave three such clones. Double crossover events were confirmed by PCR analysis with oligonucleotides F34/F35 (Fig. (Fig.2C,2C, lanes 2 to 5). Thus, double crossovers are sufficiently frequent, ~10% in this case, to be obtainable by electrotransformation of J2315 with plasmid DNA.
The high level of intrinsic resistance to most antibiotics is another factor that limits our ability to obtain transformants and exconjugants of J2315. It appeared likely that tripartite multidrug efflux pumps of the resistance-nodulation-division (RND) family (for reviews, see references 35 and 38) would be responsible for this since using the Mex/Opr peptides that perform this function in P. aeruginosa as in silico probes of the translated J2315 genome revealed 14 homologous systems: five on chromosome 1, six on chromosome 2, and three on chromosome 3. Most of them correspond to those found in a similar search by Guglierame et al. (11) and were recently reported in the genome analysis of J2315 (15). In an attempt to increase antibiotic sensitivity of J2315, we chose to delete the genes encoding the highest-scoring homologue of P. aeruginosa mexAB-oprM, a locus on chromosome 1 that we called mex1 (BCAL2820-2822 in reference 15), as described in Materials and Methods. The antibiotic sensitivity of the Δmex1 strain (Mex1 strain) was first assessed by an antibiogram diffusion test (Table (Table3).3). Although according to the clinical recommendation the Mex1 strain is still resistant to all the tested antibiotics, slight reductions in resistance to streptomycin, nalidixic acid, rifampin, and tetracycline were evident (Table (Table3).3). These indications were extended using serial dilution to determine MICs (Table (Table4).4). Deletion of mex1 causes significant increases in sensitivity to chloramphenicol, tetracycline, trimethoprim, kanamycin, gentamicin, and ampicillin and slightly higher sensitivity to rifampin, streptomycin, and spectinomycin. Deletion of other mex loci has been initiated with the aim of increasing the spectrum of sensitivity to antibiotics although it happens that this has been achieved in part by another approach, which we now describe.
The natural polyamines, putrescine, spermine, and spermidine, are cationic compounds essential for eukaryotic and prokaryotic cells and are involved in many cellular processes (for reviews, see references 12 and 18). Exogenous polyamines promote closure of the OmpF and OmpC porin channels in E. coli and possibly of OprD in P. aeruginosa (5, 23). They thus modify membrane permeability and, as a possible consequence, impair growth and/or enhance antibiotic sensitivity. In particular, clinical isolates of P. aeruginosa and strains of E. coli, S. Typhimurium, and the Gram-positive Staphylococcus aureus are sensitized to some antibiotics by spermine (23). Enhancement of antimicrobial susceptibility by chemical cationic compounds was reported for strains of B. cepacia while spermidine and cadaverine had no effect (spermine was not tested) (36). We have tested the effects of spermine on the growth and antibiotic sensitivity of J2315. Growth was unaffected at spermine concentrations below 2 mM, impaired at 2 to 3 mM, and effectively abolished at 3.5 mM (the MIC). The Mex1 strain is slightly more sensitive (MIC of 3 mM). It was at first surprising that the multidrug-resistant J2315 is inhibited by such low concentrations of spermine, especially as 16 mM spermine does not impair growth of P. aeruginosa (23).
Sensitization to antibiotics was assessed by measuring MICs of J2315 and Mex1 strains with and without 1 mM spermine (Table (Table4).4). The results showed that spermine sensitizes B. cenocepacia to many antibiotics. Spermine and the Δmex1 mutation act in apparent synergy to reduce resistance to aminoglycosides and ampicillin. Sensitivity to rifampin, trimethoprim, and chloramphenicol is increased by Δmex1 but virtually unaffected by spermine while the reverse (affected by spermine, not by Δmex1) is seen for erythromycin. These data suggest that spermine inhibits at least one antibiotic resistance pathway other than Mex1 efflux. On the other hand, spermine might interact with the Mex1 pump: in the presence of spermine, resistance to nalidixic acid increases in J2315 but decreases in the Mex1 strain (Table (Table4).4). Possibly, spermine activates a slight efflux of nalidixate by the Mex1 pump, as has been proposed to explain polyamine-induced resistance to quinolones in P. aeruginosa (22).
Tetracycline resistance is a special case. Spermine has no effect, and Δmex1 has a weak one. Curiously, although the MIC of tetracycline measured by the serial dilution assay is 64 μg/ml, concentrations above 512 μg/ml allow growth (Table (Table4).4). Tetracycline at high levels appears to induce resistance to itself, at least in a fraction of the population. Such levels should be avoided in the selection of Tetr transformants.
The applicability of reduced antibiotic resistance levels to transformant selection was tested using plasmids with a variety of resistance genes. Plating of electroporated cells of the J2315 and Mex1 strains on LB agar plates with or without 1.5 mM spermine gave the results shown in Table Table5.5. For chloramphenicol and trimethoprim, antibiotics previously used in experiments with J2315, the concentrations required to select transformants were significantly reduced by addition of spermine and/or the Δmex1 allele. More importantly, spermine and Δmex1, alone or in combination, enabled selection of resistances that were not previously usable—to gentamicin, kanamycin, and streptomycin. Under the conditions used (Table (Table5)5) there is no or negligible background, as seen by lack of growth of cells electroporated with a plasmid carrying a different resistance gene and plated in parallel. In addition, transformation was verified by identifying plasmids extracted from resistant colonies. However, despite the strong reduction in the MIC of ampicillin achieved using spermine and Δmex1 (Table (Table4),4), we were unable to distinguish Ampr transformants from the heavy background that grows after 3 days.
Electrotransformation of B. cenocepacia strain J2315 at frequencies high enough to allow selection of single and double crossover recombination events is now a routine procedure. Genetic manipulation of J2315 can thus be performed directly. The main factors contributing to this improvement are additional glycine in the growth medium, demethylation of transforming DNA by extraction from an E. coli dam dcm strain, and, to a lesser but useful extent, inclusion of the Ocr protein in the transformation mixture.
The additive effects of spermine and the Δmex1 allele on sensitivity to a number of antibiotics have proven to be remarkably efficient in decreasing the MICs of ampicillin, kanamycin, gentamicin, and streptomycin. As a consequence, we can now use streptomycin, kanamycin, and gentamicin to select for electrotransformation of B. cenocepacia J2315 by plasmids carrying the corresponding resistance genes. This should greatly facilitate genetic experimentation with this multichromosomic pathogen.
We are grateful to Jerôme Rech for technical assistance, Laetitia Grillères for her help in the first steps of isolation of the Mex1 strain, and to all the members of the group Dynamique des Réplicons Bactériens for comments and advice.
This work was supported by grant ANR-06-BLAN-0280-01 from the Agence Nationale de la Recherche.
Published ahead of print on 18 December 2009.