Burkholderia thailandensis harbors two identical rhl gene clusters responsible for the biosynthesis of rhamnolipids

Background Rhamnolipids are surface active molecules composed of rhamnose and β-hydroxydecanoic acid. These biosurfactants are produced mainly by Pseudomonas aeruginosa and have been thoroughly investigated since their early discovery. Recently, they have attracted renewed attention because of their involvement in various multicellular behaviors. Despite this high interest, only very few studies have focused on the production of rhamnolipids by Burkholderia species. Results Orthologs of rhlA, rhlB and rhlC, which are responsible for the biosynthesis of rhamnolipids in P. aeruginosa, have been found in the non-infectious Burkholderia thailandensis, as well as in the genetically similar important pathogen B. pseudomallei. In contrast to P. aeruginosa, both Burkholderia species contain these three genes necessary for rhamnolipid production within a single gene cluster. Furthermore, two identical, paralogous copies of this gene cluster are found on the second chromosome of these bacteria. Both Burkholderia spp. produce rhamnolipids containing 3-hydroxy fatty acid moieties with longer side chains than those described for P. aeruginosa. Additionally, the rhamnolipids produced by B. thailandensis contain a much larger proportion of dirhamnolipids versus monorhamnolipids when compared to P. aeruginosa. The rhamnolipids produced by B. thailandensis reduce the surface tension of water to 42 mN/m while displaying a critical micelle concentration value of 225 mg/L. Separate mutations in both rhlA alleles, which are responsible for the synthesis of the rhamnolipid precursor 3-(3-hydroxyalkanoyloxy)alkanoic acid, prove that both copies of the rhl gene cluster are functional, but one contributes more to the total production than the other. Finally, a double ΔrhlA mutant that is completely devoid of rhamnolipid production is incapable of swarming motility, showing that both gene clusters contribute to this phenotype. Conclusions Collectively, these results add another Burkholderia species to the list of bacteria able to produce rhamnolipids and this, by the means of two identical functional gene clusters. Our results also demonstrate the very impressive tensio-active properties these long-chain rhamnolipids possess in comparison to the well-studied short-chain ones from P. aeruginosa.


Background
Rhamnolipids are surface-active compounds that have been extensively studied since their early identification in Pseudomonas aeruginosa cultures in the late 1940s [1]. However, it was only in the mid 1960s that the structure of a rhamnolipid molecule was first reported [2]. Due to their excellent tensioactive properties, low toxicity and high biodegradability, these biosurfactants are promising candidates for a variety of industrial applications as well as bioremediation processes [3,4]. Furthermore, rhamnolipids have recently received renewed attention because of their involvement in P. aeruginosa multicellular behavior, such as biofilm development and swarming motility [5][6][7]. Rhamnolipids are also considered virulence factors as they interfere with the normal functioning of the tracheal ciliary system and are found in sputa of cystic fibrosis (CF) patients infected by P. aeruginosa [8][9][10]. Moreover, rhamnolipids inhibit the phagocytic response of macrophages and are known as the heat-stable extracellular hemolysin produced by P. aeruginosa [11,12].
Previously assigned to the Pseudomonas genus, Burkholderia spp. are attracting increasing interest because of their involvement in human infections. Burkholderia is best known for its pathogenic members like B. pseudomallei, the causative agent of melioidosis, as well as the opportunistic pathogens belonging to the B. cepacia complex [20,21]. Two studies have reported evidence of the production of a single dirhamnolipid by B. pseudomallei as well as by another member of the same genus, B. plantarii [22,23]. Here, we investigate the production of rhamnolipids by B. thailandensis, a non-infectious Burkholderia species closely related to B. pseudomallei [24], and by B. pseudomallei itself. In contrast to the mandated B. pseudomallei guidelines, an advantage to studying B. thailandensis is that it does not require biosafety level 3 conditions, and there is no restriction on the use of antibiotic-resistance markers for its genetic manipulation. In addition, numerous studies have shown to what extreme level these two Burkholderia species are closely related from a genetic point of view and that B. thailandensis can serve as a surrogate for studying many different traits, including physiological characteristics as well as pathogenic factors in regards to B. pseudomallei [25,26].

Presence of rhlABC homologs in B. thailandensis and B. pseudomallei
Following a nucleotide and protein similarity search using algorithms blastn and blastp with standard parameters http://blast.ncbi.nlm.nih.gov/Blast.cgi, respectively, in sequenced B. thailandensis and B. pseudomallei genome sequences, close orthologs of the P. aeruginosa rhamnolipid-biosynthesis genes rhlA, rhlB and rhlC were found in all associated strains as gene clusters. Interestingly, both species possess two 100% identical rhl gene clusters on their second chromosome ( Figure 1). A search in the partially sequenced genome of B. pseudomallei 1026b (Genomes OnLine Database; http://www.genomeson line.org), the strain used in this study, indicated the same arrangement as in the completely sequenced strain 1710b. The rhlA/rhlB/rhlC orthologs of these two Burkholderia species are highly similar to one another with nucleotide identity ranging from 89% to 96%. Furthermore, the protein encoded by these genes share almost 50% identity with those of P. aeruginosa PAO1, which possesses a single copy of these genes on its genome. Another interesting observation is that for P. aeruginosa, rhlA and rhlB are found in one operon whereas rhlC is found in a different bicistronic operon ( Figure 1). Finally, Table 1 shows that the remaining ORFs present in the rhl gene clusters, including the one adjacent to rhlC in P. aeruginosa, all seem to have functions related to transport or efflux.

Rhamnolipid production by B. thailandensis and B. pseudomallei
Due to the high similarity between the rhlA/rhlB/rhlC genes found in P. aeruginosa and their homologs in B. thailandensis, the latter was tested for the production of rhamnolipids. Using B. thailandensis in various rhamnolipid production growth conditions, the initial results from liquid chromatography/mass spectrometry (LC/MS) analysis revealed a dominant peak in the total-ion chromatograph (TIC). This peak presented a pseudomolecular ion of m/z 761 in negative-ion mode, a value that is compatible with a compound consisting of two L-rhamnose molecules as well as two β-hydroxytetradecanoic acids. A corresponding rhamnolipid, 2-O-α-L-rhamnopyranosyl-α-L-rhamnopyranosyl-β-hydroxytetradecanoyl-βhydroxytetradecanoate (Rha-Rha-C 14 -C 14 ), with a molecular weight of 762 Da, has been previously reported from B. pseudomallei and B. plantarii cultures [22,23,27]. Further experimentation was done to investigate the effects of various growth media and carbon sources on the production of rhamnolipids by B. thailandensis. Mineral and rich culture media were assayed: tested substrates included carbohydrates such as mannitol, dextrose, sucrose, glycerol and fructose along with various vegetable oils such as canola oil, olive oil, palm oil and sunflower oil, all at a final concentration of 4% (data not shown). Several studies using Genetic arrangement of rhlA, rhlB and rhlC in the genomes Figure 1 Genetic arrangement of rhlA, rhlB and rhlC in the genomes. Schematic representation of the bicistronic P. aeruginosa PAO1 http://www.pseudomonas.com regions containing the rhlAB and rhlC genes as well as the two identical gene clusters containing the homologous rhlA, rhlB and rhlC genes in B. thailandensis E264 and B. pseudomallei 1710b http://www.burkholderia.com.
plant-derived oils have demonstrated that these inexpensive hydrophobic materials are excellent carbon substrates for biosurfactant production by P. aeruginosa [28,29]. Under our experimental conditions, glycerol and canola oil were the best carbohydrate and vegetable oil for rhamnolipid production, achieving concentrations of 419.10 mg/L and 1473.72 mg/L, respectively, after 13 days of culture (Table 2). In both cases, the dirhamnolipid Rha-Rha-C 14 -C 14 was the most abundant with values ranging from 70% to 77% relative to total rhamnolipids, while its precursor Rha-C 14 -C 14 dominates the monorhamnolipid category with 5.8 and 6.5% of total rhamnolipids. Detailed analysis of B. thailandensis cultures revealed a series of long chain rhamnolipids, as shown in Table 2. These rhamnolipids are predominately composed of a C 14 -C 14 chain length fatty acid moiety as well as others comprised of chains ranging from C 10 -C 12 to C 16 -C 16 chain length.
To confirm that the ions identified by LC/MS are indeed rhamnolipids, they were fragmented and analyzed by tandem mass spectrometry (LC/MS/MS). To allow for comparison with P. aeruginosa rhamnolipids, monorhamnolipids obtained from B. thailandensis were fragmented and the observed fragmentation pattern was similar to the one we observed for P. aeruginosa [13]. For an isomeric pair of rhamnolipid congeners bearing two 3hydroxy fatty acids of different chain lengths (for example Rha-C 12 -C 14 and Rha-C 14 -C 12 ), the relative abundance of the various congeners was studied. Because such isomers cannot be resolved chromatographically, the LC/MS/MS spectrum of their common pseudomolecular ions show fragment ions characteristic of both compounds, as seen in Figure 2. As we have previously reported for P. aeruginosa [14], within each isomeric pair, the rhamnolipid congener with the shortest chain adjacent to the sugar moiety is more abundant. To verify whether the rhamnolipids produced by B. thailandensis share this characteristic, they were subjected to an enzymatic hydrolysis of their rhamnose groups with naringinase [30] to produce the corresponding HAAs. The same stoichiometrical preference was confirmed.
With these results in hand, we investigated the potential of the highly genetically related species B. pseudomallei to produce a range of rhamnolipids other than the previously described Rha-Rha-C 14 -C 14 . Figure 3 shows the production of the most abundant rhamnolipids by this pathogen. The same long-chain bearing congeners found in B. thailandensis were also discovered in B. pseudomallei, including the dominant Rha-Rha-C 14 -C 14 .

Critical Micelle Concentration (CMC) and surface tension assays
To investigate the potential of the long-chain rhamnolipids produced by Burkholderia species for lowering surface tension, the critical micelle concentration of this mixture of rhamnolipid congeners was established (Figure 4). At the CMC of about 250 mg/L, the surface tension is lowered to 43 mN/m.

Both rhlA alleles are functional and necessary for maximal rhamnolipid production
The contribution to rhamnolipid production of the two identical rhl gene clusters found on the B. thailandensis genome was tested by creating single ΔrhlA mutants for each allele, as well as a double ΔrhlA mutant. These three mutants were then investigated for their ability to produce rhamnolipids ( Figure 5). Five sets of replicates confirmed that the B. thailandensis ΔrhlA1 mutant produces more rhamnolipids than the ΔrhlA2 mutant. The rhamnolipids produced by each of these mutants are composed of the same congeners in the same proportions as the wild type strain and only a quantitative difference is observed. Intriguingly, total production of both single mutants do not equal to the production of the wild type strain. Finally, the double ΔrhlA mutant does not produce any detectable rhamnolipids.

Swarming motility requires both rhlA alleles
In P. aeruginosa, production of rhamnolipids is essential for expression of the multicellular behaviour called swarming motility [31]. It was therefore of interest to assess whether rhamnolipids are also important for this type of motility in B. thailandensis. Furthermore, since both rhlA alleles are functional and contributing to the production of rhamnolipids in this species, we wondered if the amount of biosurfactants produced by the single mutants would be sufficient to permit the swarming phenotype. ΔrhlA1 and ΔrhlA2 mutants of B. thailandensis Congener analysis of rhamnolipids from B. thailandensis Figure 2 Congener analysis of rhamnolipids from B. thailandensis. A) Mass spectra of the fragmented m/z 587, 615 and 643 pseudomolecular ions of congeners Rha-C 12 -C 14 , Rha-C 14 -C 12 , Rha-C 14 -C 14 , Rha-C 14 -C 16 and Rha-C 16 -C 14 . B) Schematic representation of observed fragmentation patterns of a monorhamnolipid. C) Daughter ions generated by fragmentation of the specified congeners.
were thus tested for their ability to swarm. Figure 6A (Control column) shows the swarming phenotype of the wild type strain as well as the single ΔrhlA mutants and the double ΔrhlA mutant. We observe that the single mutants have hindered swarming motility whereas the double mutant is incapable of such motility. Thus, one functional copy of rhlA does not provide enough rhamnolipid production to allow normal surface translocation on a semisolid surface. Interestingly, the ΔrhlA1 mutant is capable of moving to a greater distance than the ΔrhlA2 mutant ( Figure 6A). This observation concurs with the above results showing the superior rhamnolipid production by the ΔrhlA1 mutant compared to the ΔrhlA2 mutant (Figure 5). Finally, as expected, the double ΔrhlA mutant is incapable of any swarming.
To test whether swarming phenotype restoration is possible with our ΔrhlA mutants, swarm assays were performed with the addition of increasing concentrations of exogenous rhamnolipids. We observed that the ΔrhlA1 mutant requires less exogenous rhamnolipids to regain complete swarming motility compared to the ΔrhlA2 mutant, consistent with the finding that this latter mutant produces less rhamnolipids. These results indicate that a critical concentration of biosurfactant is necessary to enable bacteria to swarm. Accordingly, the double mutant requires much more exogenous rhamnolipids to restore this phenotype. Cross-feeding experiments with both ΔrhlA mutants were also performed to verify whether swarming phenotype could be regained. Interestingly, when the two mutants are mixed before plating, swarming is restored ( Figure 6B, right), contrary to when mutants are simply spotted side-by-side (Figure 6B, left). Rhamnolipid production by single ΔrhlA mutants rying genes highly similar to the P. aeruginosa genes rhlA, rhlB and rhlC, which are involved in the biosynthesis of rhamnolipids. Interestingly, in the latter species these three genes are arranged in two physically distant operons, while in the two Burkholderia species, they are part of the same gene cluster. The results presented here demonstrate that the purpose of these genes in B. thailandensis, and more than likely in B. pseudomallei, is for the production of rhamnolipids.

Discussion
Genes that share similarities with efflux pumps and transporters are also present within the rhl gene clusters. There is at least one instance of an efflux system implicated in the transport of a biosurfactant. In the Gram-positive species Bacillus subtilis, YerP, a homolog to the resistancenodulation-cell division (RND) family efflux pumps, was found to be implicated in surfactin resistance [32]. We propose that the other genes present within the rhl gene clusters are involved in the transport of rhamnolipids outside the cell; we are currently investigating this hypothesis.
Under our experimental conditions, B. thailandensis is capable of producing rhamnolipids with 3-hydroxy fatty acid moieties that are comprised of chains varying from C 10 -C 12 to C 16 -C 16 . Such long lengths have not been reported for rhamnolipids produced by bacteria other than those of the Burkholderia species, with the exception of one publication reporting trace amounts of Rha-Rha-C 10 -C 14:1 produced by P. aeruginosa 57RP and another describing the production of a C 14 -C 10 form by P. chlororaphis B-30761 [13,33]. Interestingly, the rhamnolipids produced by B. thailandensis are predominantly composed of dirhamnolipids, whereas monorhamnolipids and HAAs are only found in much smaller concentrations. Although the latter two are produced in smaller quantities by the bacteria, they are nevertheless comprised mostly of the corresponding molecule in the C 14 -C 14 chain lengths. The dirhamnolipid versus monorhamnolipid ratio found in this species is approximately 13, whereas we observe a factor of only 4 in P. aeruginosa. One possible explanation is that, unlike P. aeruginosa which harbors rhlA and rhlB in one operon and rhlC in another, Burkholderia species code for the three enzymes from the same gene cluster, predicted to be an operon. We hypothesize this favors the simultaneous production of all the enzymes of the biosynthetic pathway; hence, RhlC would be present simultaneously and in the same stoichiometric ratio as RhlB, therefore favoring the immediate addition of the second L-rhamnose unto the monorhamnolipids. Our result adds B. thailandensis to the few bacterial species able to produce rhamnolipids, and shows that rhamnolipids produced by Burkholderias are more likely to contain longer side chains than those by Pseudomonas species, which are predominantly of the C 10 -C 10 chain length.
The above mentioned facts are also true for the rhamnolipids produced by B. pseudomallei. More specifically, fatty acyl chains with carbon lengths of 12, 14 and 16 were observed in B. pseudomallei rhamnolipids, although only dirhamnolipids were detected. While production levels achieve 30 mg/L for B. pseudomallei, B. thailandensis can reach 80 mg/L under the same conditions (data not shown). Results of the present study further demonstrate that rhamnolipid congeners other than the previously described Rha-Rha-C 14 -C 14 are also produced by this pathogen.
Inactivation of each of the two rhlA alleles confirmed that both rhl gene clusters contribute to the synthesis of rhamnolipids. Rhamnolipid production is observed even when one of the two alleles is not functional, suggesting that one copy does not depend on the other. However, the production levels attained by each of the ΔrhlA mutants show that the gene cluster containing the rhlA2 allele contributes about two and half more rhamnolipids than the rhlA1 allele cluster ( Figure 5). Since the promoter sequences of the two rhl gene clusters only share approximately 270 bp directly upstream of both of the rhlA ATGs and therefore seem to have diverged, these results suggest that each cluster possesses its unique, differently control-Swarming phenotype restoration within the ΔrhlA mutants Figure 6 Swarming phenotype restoration within the ΔrhlA mutants. Swarm plates were incubated for 18 h at 30°C with B. thailandensis E264 wild type strain, both single ΔrhlA mutants as well as the double ΔrhlA mutant. Under these experimental conditions swarming motility is normally favored, as observed with the wild type strain. Experiments were done in triplicate. A) Swarming phenotype restoration of the ΔrhlA mutants with addition of 1, 5, 10 and 25 mg/L of exogenous purified rhamnolipids. B) Cross-feeding experimentation with both ΔrhlA single mutants. Left: mutants placed side-by-side; Right: mutants mixed before plating.
led promoter, which is apparently found upstream of this conserved region. The biphasic shape of the wild-type rhamnolipid production curve supports this conclusion. Furthermore, the addition of both levels of production by the two clusters does not reach the wild type production level. This could be explained by some sort of positive retroaction where rhamnolipids stimulate global production and that the gene clusters are in fact interconnected. Also, it must be considered that the different rhamnolipid production levels attained by the ΔrhlA single mutants could also be associated to polar effects on the downstream genes that could possibly interfere with rhamnolipid biosynthesis.
The presence of two paralogous gene clusters is interesting since gene duplication is normally not favored within genomes, as one copy is generally more susceptible to mutations and/or inactivation. However, a duplication event might be preserved if it is immediately beneficial to the organism because of protein dosage effects, e.g. in variable environments [34,35]. We therefore extrapolate that the Burkholderia species which harbor twin rhl gene clusters have conserved both copies because it must be advantageous for these bacteria to produce extra quantities of rhamnolipids. The requirement of both rhl gene clusters for normal swarming motility supports this model (see below). The presence of a transposase of the mutator family in close proximity of one of the gene clusters (BTH_II1082) can also be indicative that a past duplication of an original single copy occurred and positive selection throughout evolution of some bacterial lineages conserved the paralogs.

Long chain rhamnolipids from Burkholderia: effects on the CMC
Considering the length of the carbon chains of the fatty acid moiety of rhamnolipids produced by Burkholderia species, it was compelling to determine their effect on lowering the surface tension of water. A total rhamnolipid extract from B. thailandensis lowers the surface tension to 42 mN/m, with a CMC value of 225 mg/L. These values are higher than those traditionally reported for rhamnolipids produced by Pseudomonas species (typically around 30 mN/m and CMC in the order of 20 to 200 mg/L) [36]; however, it is only recently that HAAs have been discovered, as well as their efficacious surface tension-lowering potential [16]. Thus, we assume that results pertaining to surface tension properties of rhamnolipids published prior to this report could have been biased by a contamination with easily co-purified HAAs. For the purpose of the present study, we compared our results with those we have published for purified rhamnolipids and HAAs produced by P. aeruginosa PG201 [16]. The purified rhamnolipids from this strain lower surface tension to 40 mN/m with a CMC value of approximately 600 mg/L, while the HAA mixtures displays values of 29 mN/m with a CMC of approximately 800 mg/L. Consequently, it is clear that the longer chain rhamnolipids produced by B. thailandensis start forming micelles at a much lower concentration than P. aeruginosa rhamnolipids, 225 mg/L versus 600 mg/L. These values can be compared as the rhamnolipid mixture from B. thailandensis used for our tests contained only traces of HAAs. The effect of alkyl ester chain length of sophorolipids, a class of biosurfactants produced by Candida bombicola, has been studied with regards to micellization. The study reported a direct effect of carbon chain length on decreasing the CMC. Additional CH 2 groups render the molecule more hydrophobic and thus facilitate micelle formation [37]. This might explain the lower CMC value obtained with the longer chain rhamnolipids produced by B. thailandensis in comparison to those obtained by P. aeruginosa.

Both rhlA alleles are necessary for normal swarming motility
Swarming motility always involves biosurfactants. For example, serrawettin W2, a wetting agent produced by Serratia liquefaciens, is required for swarming motility in a nonflagellated mutant [38,39]. In regards to P. aeruginosa, biosurfactants such as rhamnolipids and HAAs are essential for swarming motility [7,16,40]. Only a few studies have reported on swarming motility of Burkholderia species, which is at least in part attributed to the lack of knowledge available regarding wetting agents produced by members of this genus. The swarming motility of B. cepacia has been observed, and the authors hypothesized that biosurfactants are involved [41]. We have also recently reported conditions under which B. thailandensis can swarm [42].
The present study demonstrates that swarming motility of a B. thailandensis double ΔrhlA mutant is completely prevented. This is in agreement with previous studies showing that inactivation of rhlA inhibits swarming by P. aeruginosa [16,40]. Furthermore, a mutation in any of the two rhlA genes hinders swarming of B. thailandensis, suggesting that a critical concentration of rhamnolipids is required and that the levels reached when only one of the two gene clusters is functional are not sufficient to allow the bacteria to completely overcome surface tension. The complementation experiment with exogenous addition of increasing concentrations of rhamnolipids further corroborates that there is indeed a critical concentration of biosurfactant necessary for B. thailandensis to swarm, and that both rhl gene clusters contribute differently to the total concentration of rhamnolipids produced. The cross-feeding experiment suggests that rhamnolipids produced by B. thailandensis diffuse to only a short distance in the agar medium surrounding the colony.

Conclusions
The discovery that B. thailandensis is capable of producing considerable amounts of long chain dirhamnolipids makes it an interesting candidate for the production of biodegradable biosurfactants with good tensioactive properties. Furthermore, that this bacterium is non-infectious makes it an ideal alternative to the use of the opportunistic pathogen P. aeruginosa for the large scale production of these compounds for industrial applications. Finally, identification of the same paralogous rhl gene clusters responsible of the production of long chain rhamnolipids in the closely-related species B. pseudomallei might shed some light on the virulence mechanisms utilized by this pathogen during the development of infections.
To follow the production of rhamnolipids by B. thailandensis and its ΔrhlA mutants, cultures were grown in 50 ml of NB supplemented with 2% glycerol in 500 ml Erlenmeyer flasks at 37°C with gyratory shaking (240 rpm). For B. pseudomallei, cultures were carried out in 25 ml of NB supplemented with 4% glycerol in 250 ml Erlenmeyer flasks at 34°C with gyratory shaking (200 rpm).

Rhamnolipid production and extraction
Cultures for high yield rhamnolipid production were grown in 200 ml of NB supplemented with 4% of glycerol or canola oil in 2 L Erlenmeyer flasks at 34°C with gyratory shaking (240 rpm). Extraction of total rhamnolipids was performed as described previously [16], with slight modifications. Briefly, cells were removed from the medium by centrifugation (13,000 × g, 15 min) and the supernatant acidified to pH 3-4 with concentrated HCl. The rhamnolipids were then extracted three times with 1/ 3 of the volume of ethyl acetate. The organic extract was then dried with anhydrous sodium sulfate and evaporated using a rotary evaporator. The oily residue was finally dissolved in methanol.

Construction of rhlA mutants
For the construction of single ΔrhlA mutants in B. thailandensis, a 464 bp fragment was amplified using primers rhlASVF and rhlASVR, containing XbaI and KpnI restriction sites, respectively ( Table 3). The PCR product was cloned by the means of its XbaI and KpnI sites into the suicide vector pKNOCK-Tc [45]. The construct was transformed into competent E. coli SM10 cells by the heat shock method. The plasmid was then mobilized into B. thailandensis by mating and transformants were selected on TSB agar plates containing 50 μg/ml gentamicin, 15 μg/ml polymyxin B and 150 μg/ml tetracycline. To verify in which of the two rhlA alleles the homologous recombination took place, diagnostic PCRs were conducted using promoter-specific forward primers, rhlA1PF and rhlA2PF, as well as a common reverse primer, rhlAR, located at the end of the 3' regions of both rhlAs. Rhamnolipid production of mutants was also quantified (see below) and compared to typical wild type production values.
To inactivate the second rhlA allele, targeted mutagenesis through natural transformation of PCR fragments was exploited [46]. Briefly, three fragments corresponding to the regions flanking the specific rhlA gene to be deleted and a trimethoprim resistance gene were joined by PCR. The 5' and 3' flanking regions of rhlA were amplified using primers rhlA5'2F and rhlA5'2R as well as rhlA3'3F and rhlA3'3R, respectively. The trimethoprim resistance marker was amplified from the pFTP1 plasmid (a gift from H. P. Schweizer, Colorado State University) using primers rhlATp1F and rhlATp1R [47,48]. Cells of the single ΔrhlA mutant were rendered competent using DM medium and then exposed to various concentrations of the mutagenic PCR fragment. Double ΔrhlA mutants were selected on TSB agar containing 150 μg/ml tetracycline and 100 μg/ml trimethoprim. The B. thailandensis ΔrhlA double mutant was confirmed by diagnostic PCR to verify proper recombination and insertion of the resistance marker. Absence of rhamnolipid production by LC/MS analysis also served as a confirmation.

Preparation of culture samples for LC/MS analysis
To prepare samples for LC/MS analysis, the culture samples were firstly centrifuged to remove cells (16,000 × g, 15 min). To the cell-free supernatant was then added either 16-hydroxyhexadecanoic acid or deuterium-labeled 4-hydroxy-2-heptylquinoline (HHQ-D4) [49] as internal standards used for quantitative measurements, both at a final concentration of 10 mg/L. For the highly pathogenic B. pseudomallei, cell-free supernatants were obtained by centrifugation (16,000 × g, 15 min) followed by filtration on a 0.22 μm filter. Twenty μl of samples were injected for LC/MS analysis. Quantification was performed by integration of the pseudomolecular and the proper fragment ions and the use of dose-response calibration curves using purified rhamnolipids.

Rhamnolipid analysis (LC/MS)
All rhamnolipid quantifications and analyses were performed using a Quattro II (Waters, Mississauga, Ontario, Canada) triple-quadrupole mass spectrometer in negative electrospray ionization mode coupled to an HP 1100 (Agilent Technologies, Saint Laurent, Quebec, Canada) high-performance liquid chromatograph (HPLC) equipped with a 4.6 × 50 mm 300SB-C3 Zorbax 5 μm (Agilent) reverse-phase column. The HPLC flow rate was set at 400 μl/min and was split to 10% by the means of a Valco Tee prior to being introduced into the mass spectrometer. An acetonitrile-water gradient containing 2 mM of ammonium acetate was used starting with 25% acetonitrile during the first 5 min, raised to 50% by 18 min and 100% by 19 min. This concentration was held until 22 min, where the initial concentration was resumed and kept until 26 min. Voltage of the capillary was set to 3.5 kV and cone voltage to 30 V. The temperature of the source block was kept at 120°C. Scan mass range was set from 130 to 940 Da. A calibration curve was performed to determine the long chain rhamnolipid response factor.
During LC/MS/MS experimentation, fragmentation of the molecules were induced with argon serving as the collision gas at 2 × 10 -3 mTorr.

Enzymatic hydrolysis of rhamnolipids -Naringinase
To study the rhamnolipid congeners produced by B. thailandensis, a modified protocol using enzymatic hydrolysis was employed, as described previously [30]. Briefly, 100 mg of extracted and purified rhamnolipids were suspended in 5 ml of 50 mM sodium acetate buffer, pH 4.1.
To this solution was added 100 mg of naringinase from Penicillum decumbens (Sigma). The mixture was then kept at 50°C for 2 h with gyratory shaking (240 rpm), at which point 20 ml of buffer were added. After 24 h, another 150 mg of naringinase were added as well as 25 ml of buffer. The reaction was kept under these conditions for 8 days. A final 50 mg of naringinase in 20 ml of buffer were added to the mixture and was left for another 24 h. Thereafter, the solution was acidified to pH 3-4 using concentrated HCl and extracted three times with ethyl acetate. The fatty acid moieties generated by naringinase cleavage were then analyzed by LC/MS after the extract had been dried and evaporated.

CMC -Surface tension assay
Critical micelle concentration and surface tension were measured by the du Noüy ring method [50] using a surface tensiometer (Fisher). The instrument was calibrated against water and assays were performed in triplicate at room temperature.

Swarming motility
For swarming assays, cultures were grown overnight, diluted in fresh medium and subcultured until OD 600~6 .0 was reached. Swarm plates were prepared as follows: freshly autoclaved medium consisting of NB supplemented with 0.5% dextrose (Fisher) and 0.5% Bacto-agar (Difco) was poured into standard Petri dishes and dried under laminar flow for 30 min, as before [42]. Immediately following the drying period, plates were inoculated at their center with 5 μl of bacterial culture and placed at 30°C.
For swarming phenotype restoration, 1, 5, 10 and 25 mg/ L of purified B. thailandensis E264 rhamnolipids were deposited (10 μl) at the center of respective plates and left to dry for 15 minutes before spot inoculation with swarming-deficient ΔrhlA mutant strains. For cross-feeding experiments, either equal parts of the cultures were mixed before being plated at the center on the swarm plate, or cultures were simply spotted side-by-side.

Authors' contributions
ED and DD designed the experiments. DD carried out all experimental procedures and analyzed the data. FL pro-vided critical knowledge in LC/MS experimentation. DEW provided B. pseudomallei samples for LC/MS analysis. DD wrote the manuscript. FL and ED corrected the manuscript. All authors read and approved the final manuscript.