A Burkholderia cepacia complex non-ribosomal peptide-synthesized toxin is hemolytic and required for full virulence.

Members of the Burkholderia cepacia complex (Bcc) have recently gained notoriety as significant bacterial pathogens due to their extreme levels of antibiotic resistance, their transmissibility in clinics, their persistence in bacteriostatic solutions, and their intracellular survival capabilities. As pathogens, the Bcc are known to elaborate a number of virulence factors including proteases, lipases and other exoproducts, as well as a number of secretion system associated effectors. Through random and directed mutagenesis studies, we have identified a Bcc gene cluster capable of expressing a toxin that is both hemolytic and required for full Bcc virulence. The Bcc toxin is synthesized via a non-ribosomal peptide synthetase mechanism, and appears to be related to the previously identified antifungal compound burkholdine or occidiofungin. Further testing shows mutations to this gene cluster cause a significant reduction in both hemolysis and Galleria mellonella mortality. Mutation to a glycosyltransferase gene putatively responsible for a structural-functional toxin variant causes only partial reduction in hemolysis. Molecular screening identifies the Bcc species containing this gene cluster, of which several strains produce hemolytic activity.

Introduction

Members of the Burkholderia cepacia complex (Bcc) are Gram-negative bacilli ubiquitous in soil and water sources. These bacteria embody a diverse array of metabolic capabilities, among them environmental contaminant degradation,, fungal phytopathogen suppression- and nitrogen fixation. For these reasons, the Bcc has been identified as a powerful class of bioremediation and plant growth promoting agents. However, the potential usefulness of the Bcc has been compromised by their opportunistic pathogenicity and concerns surrounding their extensive antibiotic and biocide resistance.- For immune-compromised patients, such as those with the heritable genetic disease cystic fibrosis (CF), the Bcc represent a singular threat because of their adept transmissibility in clinical settings,, their extremely high antibiotic resistance, their persistence in disinfectant solutions- and their ability to survive in vivo.- Upon gaining access to the lung, members of the Bcc produce a number of virulence factors toxic to the host and instigate immune system stimulation that ultimately severely reduces lung function. As well, the Bcc occasionally create what has been termed “cepacia syndrome,” a disease characterized by an acute and severe necrotizing pneumonia, sepsis and death.

Toxic factors produced by the Bcc include proteases, lipases and type III secretion system effectors. Early reports suggested that only approximately 4% of Bcc isolates exhibit β-hemolytic activity when sheep erythrocytes were tested. However, later studies showed that up to 39% of clinical Bcc isolates possessed hemolytic activity when using sheep erythrocytes. One source of this hemolytic activity may arise from a β-hemolysin, although unlike Pseudomonas aeruginosa, Bcc phospholipase activity does not correlate with the presence of this hemolytic activity. In 1994, the isolation and purification of two related hemolysins with antifungal properties was published by Abe and Nakazawa. The hemolytic activity of these “cepalysins” was inhibited by sterols, suggesting that they require an interaction with cholesterol in the erythrocyte membrane to produce a biological effect. Hutchison et al. showed that a lipopeptide toxin produced by Burkholderia cenocepacia could cause hemolysis of erythrocytes, apoptosis of human neutrophils, and an increase in neutrophil degranulation. Fehlner-Gardiner et al. identified components of the Bcc general secretory pathway that were involved in the secretion of hemolytic and phospholipase C activities, but that were not necessary for Bcc intracellular survival within Acanthamoeba polyphaga. More recently, Bevivino et al., using ram erythrocytes, found that almost all Burkholderia ambifaria isolates were hemolytic and that the percentage of hemolytic environmental B. cenocepacia isolates was markedly higher than the percentage of hemolytic clinical isolates. Similarly, Carvalho et al. indicated that almost all of the 59 Bcc clinical isolates from a reference CF center in Rio de Janeiro, Brazil produced several different exoproducts except for hemolysin (as tested against human erythrocytes). From these results, it is evident that some Bcc members produce hemolytic toxins, although their importance toward pathogenicity is unclear.

Several secondary metabolite compounds with biological effects are also produced by the Bcc, either through polyketide synthetase (PKS) or non-ribosomal peptide synthetase (NRPS) enzyme complexes. Known secondary metabolite toxins produced by the Bcc include the broad-spectrum antifungal agent “pyrrolnitrin,” a group of antifungal and antitumor agents deemed “cepafungins,” hemolytic peptides named “cepalysins,” related antifungal compounds “cepacidine A1 and A2”, and the recently described antifungal cyclic peptide “occidiofungin” or “burkholdine,”- which may be similar to molecules previously isolated from Bcc strains as “xylocandins.”, While these latter compounds show a broad range of activity against fungi, in this manuscript, we identify through random and directed mutagenesis a gene cluster in Burkholderia vietnamiensis strain DBO1 that is homologous to the gene cluster that synthesizes occidofungin/burkholdine, and demonstrate that this gene product is also strikingly β-hemolytic. This study also examines the prevalence of this gene cluster within the Bcc and the lethality of this compound in an invertebrate infection model.

Results

Identification of an NRPS cluster through plasposon mutagenesis

The hemolytic patterns of the Bcc species B. vietnamiensis DBO1 and B. ambifaria AMMD, which form β-hemolysis (yellow clearing) on sheep blood agar, were investigated by randomly mutagenizing DBO1 with plasposon pTnMod-OTp′. Mutants were screened on sheep blood agar + 100 µg/mL Tp, and those deficient in hemolysis were carried forward for analysis, which revealed plasposon insertion sites in an uncharacterized gene cluster containing NRPS-encoding genes, as well as regulator and transporter-encoding genes. Further insertion mutation analysis identified genes essential to the hemolytic activity, and allowed delineation of the ends of the gene cluster. DNA sequence analysis of the B. vietnamiensis DBO1 gene cluster demonstrated 99.9% identity to a nearly identical gene cluster in the sequenced genome of B. ambifaria strain AMMD. As shown in Figure 1, and with functional activity for each gene shown in Table 1, the core Bcc NRPS gene cluster is approximately 50.5 kb in length and is comprised of 13 ORFs arranged identically in both strains. In B. ambifaria AMMD, the NRPS cluster of open reading frames (ORFs) is located on chromosome 3 and extends from basepair 1,101,558 to 1,152,073 for a total of 50,515 base pairs. BLASTN analysis identified only one other closely related gene cluster in the GenBank database, a gene cluster of Bcc member B. contaminans strain MS14 encoding occidofungin biosynthesis proteins. Not only are these 13 genes in MS14 similarly arranged (syntenic) to the NRPS genes in AMMD/DBO1, each ORF in the AMMD/DBO1 gene cluster has as its most closely related database entry an ORF in the MS14 occidofungin biosynthetic cluster. For example, the B. ambifaria AMMD 3176 amino acid adenylation protein Bamb_4672 is 89% identical (2812/3176) and 93% similar (2944/3176) to the same protein in Burkholderia contaminans MS14, along the entire sequence (12/3176 gaps). For comparison, the next closest database relative is an uncharacterized protein identified in Acidobacterium capsulatum ATCC 51196, that is only 52% identical (1629/3138) and 68% similar (2119/3138) to Bamb_4672 across 3138/3176 amino acids, with 72/3138 gaps. This analysis indicates that the gene clusters identified to be responsible for the hemolytic activity observed in strain DBO1 (and by homology AMMD), are closely related to the occidofungin biosynthesis gene cluster identified in B. contaminans MS14, and to few other genes in the known bacterial genetic complement.

Figure 1. The NRPS cluster giving rise to toxic activity in B. vietnamiensis DBO1. The cluster appears similar to that described in B. contaminans MS14 by Gu et al. except that an additional two genes, homologs of 6477 and 6478 in B. ambifaria AMMD, are present upstream of the described MS14 cluster that are required for virulence in DBO1. Filled vertical arrows indicate genes disrupted by plasposon mutagenesis; unfilled vertical arrows indicate genes disrupted through targeted mutagenesis. Genes are named according to the sequenced B. ambifaria AMMD genome: 6465, encoding an FAD-linked dioxygenase homolog; 6466, encoding a LuxR homolog; 6467, encoding a hypothetical protein with a DNA binding motif; 6468, encoding a LuxR homolog; 6469, encoding a cyclic peptide transporter; 6470, encoding a hypothetical protein; 6471, encoding a glycosyltransferase; 6472, encoding an adenylation domain-containing protein (ADCP); 6473, encoding an ADCP; 6474, encoding an ADCP; 6475, encoding a β-lactamase; 6476, encoding an ADCP-polyketide synthetase (PKS) hybrid; 6477, encoding a PKS; 6478, encoding a PKS; 6479, encoding a taurine dioxygenase. Only genes 6466 through 6478 appear to be involved in toxin production.

Table 1. Bacterial strains and plasmids constructed for this study

Bacterial strains	Genotype or relevant phenotype	Hemolytic activity	Source
B. vietnamiensis
DBO1	Parent strain	+	37
6465::TpR	DBO1 with targeted oriTp’ insertion in gene 6465, encoding an FAD-linked dioxygenase	+	This study
6466::TpR	DBO1 with targeted oriTp’ insertion in gene 6466, encoding a putative LuxR regulator	−	This study
6468::TpR	DBO1 with oriTp’ plasposon insertion in gene 6468, encoding a putative LuxR regulator	−	This study
6469::TpR	DBO1 with oriTp’ plasposon insertion in gene 6469, encoding a putative peptide transporter	−	This study
6470::TpR	DBO1 with targeted oriTp’ insertion in gene 6470, encoding a hypothetical protein	+	This study
6471::TpR	DBO1 with targeted oriTp’ insertion in gene 6471, encoding a putative glycosyltransferase within the NRPS gene cluster	+/−	This study
6472::TpR	DBO1 with oriTp’ plasposon insertion in gene 6472, encoding a putative adenylation domain-containing protein	−	This study
6473::TpR	DBO1 with oriTp’ plasposon insertion in gene 6473, encoding a putative adenylation domain-containing protein	−	This study
6474::TpR	DBO1 with oriTp’ plasposon insertion in gene 6474, encoding a putative adenylation domain-containing protein	−	This study
6476::TpR	DBO1 with oriTp’ plasposon insertion in gene 6476, encoding a putative adenylation domain-containing protein and polyketide synethetase	−	This study
6477::TpR	DBO1 with targeted oriTp’ insertion in gene 6477, encoding a putative polyketide synthetase	−	This study
6478::TpR	DBO1 with oriTp’ plasposon insertion in gene 6478, encoding a putative polyketide synthetase within the NRPS gene cluster	−	This study
6479::TpR	DBO1 with targeted oriTp’ insertion in gene 6479, encoding a putative taurine metabolism protein	+	This study
DBO1/pSCRhaTc	DBO1 carrying pSCRhaTc	+
6466::TpR/pSCRhaTc	DBO1 mutant 6466::TpR carrying pSCRhaTc	−
6466::TpR/p6466	6466::TpR carrying pSCRha-6466His	+	This study
E. coli
DH5α	Cloning host strain	NA	Invitrogen
Plasmids
pJET1.2/blunt	Cloning vector, AmpR	NA	Fermentas
pTnMod-OTp’	Plasposon used for random mutagenesis, TpR	NA	38
pSCRha-6466His	Rhamnose-inducible plasmid pSCRhaB2 modified with Tc resistance cassette and 6x histidine-tagged gene 6466 for genetic complementation	NA	This study
pSCRhaTc	pSCRha-6466His carrying no 6466His insert	NA	43

+, full clearing; +/−, partial clearing; −, no clearing; NA, not applicable.

Uniquely structured toxin is a broad-specificity hemolysin

Bcc strains DBO1 and AMMD were grown on blood agar with hemocytes from various sources, including bovine, rabbit, sheep, horse and human, to determine whether hemolysis was limited to certain mammals. Both strains exhibited hemolytic activity on all types of blood agar (data not shown). Clearing on sheep blood agar produced large yellow zones of lysis, indicative of a β-hemolysis (Fig. 2), while cleared zones on other blood types appeared colorless. Mutants in DBO1 defective in producing zones of clearing on one type of blood agar were defective in producing zones of clearing on the other types of blood agar. In order to determine whether the diffusible hemolytic activity observed was active against bacterial cells, Bcc strains producing hemolytic activity were streaked on sheep blood plates and cross-streaked (either concurrently or after one day’s growth) with either a panel of Gram-negative and Gram-positive bacterial species including various pseudomonads, Escherichia coli, Staphylococcus and Bacillus species, or 25 different Bcc strains. In both conditions, the Bcc hemolytic activity was not inhibitory to the growth of any other bacteria tested, even though hemolytic activity was observable in the underlying blood agar.

Figure 2. Bcc hemolytic activity on TSA 5% sheep blood agar. B. vietnamiensis strain DBO1 and B. ambifaria strain AMMD (“V” streaked) produce large transparent yellow hemolytic zones of clearing on an inverted sheep blood plate following overnight growth at 30°C. B. cenocepacia strains C4455 and C6433 and B. multivorans strain C3430 produce no such hemolytic zones of clearing and are barely visible through the opaque growth media.

To quantify the effect of different NRPS mutations on hemolytic activity, we developed a high-throughput liquid human blood hemolysis assay using hemoglobin release (absorbance at wavelength 570 nm) as a marker. As shown in Figure 3, we first tested this assay against wild-type strain DBO1, and two constructed mutants including 6466::Tp, a putative luxR homolog mutant and 6477::Tp, a putative polyketide synthetase mutant, neither of which produce zones of clearing on sheep blood agar. Although hemocytes were inoculated with similar numbers of bacterial cells (Fig. 3B), only wild-type strain DBO1 was able to release substantial amounts of hemoglobin (Fig. 3A) or reduce the counted number of red blood cells (Fig. 3C).

Figure 3. (A) Absorbance of blood broth supernatants across a range of wavelengths after 24 h incubation with wt DBO1 and two NRPS cluster mutants. Dotted black line, wild-type DBO1; solid gray line, 6466::oriTpR (luxR mutant); dotted gray line, 6477::oriTpR (polyketide synthetase mutant). (B) Viable plate counts after 24 h. Black, wild-type DBO1; white, 6466::oriTpR; gray, 6477::oriTpR. (C) Erythrocyte counts. Black, wild-type DBO1; white, 6466::oriTpR; light gray, 6477::oriTpR; dark gray, blank control (no bacteria added).

As described previously,, the NRPS product occidiofungin/burkholdine is a cyclic lipopeptide comprising eight amino acids. At least two forms of the peptide exist. One form contains a xylose side chain (occidiofungin/burkholdine 1229), which is presumably added to the peptide by the predicted glycosyltransferase encoded within the NRPS gene cluster. The other form contains no xylose side chain (burkholdine 1097). While the xylose-containing peptide was shown to have antifungal activity, the peptide lacking xylose displayed little activity., To determine whether we would observe differences in hemolytic activity produced by the different B. vietnamiensis DBO1 NRPS cluster mutants, including the glycosyltransferase mutant 6471::TpR, DBO1 mutants were incubated with human erythrocytes for a period of two days, and the absorbance at 570 nm of the supernatants was taken each day. As shown in Figure 4, there are three potential hemolytic phenotypes demonstrated by the strains. First, there is a fully hemolytic phenotype characteristic of both wild-type DBO1 and strains containing mutations to genes outside of the NRPS gene cluster (6465 and 6479). In addition, genes 6469 and 6470, although by location included as a part of the NRPS gene cluster, are not essential to the production of hemolytic activity, as their mutation does not significantly alter wild-type hemolytic activity. Second, there is an intermediate hemolytic phenotype exhibited by mutant 6471::TpR, likely due to the lack of a xylose addition to the final product, as described for occidofungin/burkholdine., Whereas xylose addition to occidiofungin/burkholdine appears to be essential for antifungal activity, the absence of NRPS product glycosylation in DBO1 reduces but does not eliminate the hemolytic activity of the NRPS-derived compound in B. vietnamiensis DBO1. Finally, there is a complete abrogation of hemolysis in strains containing mutations to genes 6466 (encoding a LuxR regulator), 6469 (encoding a cyclic peptide transporter), 6472 [encoding an adenylation domain-containing protein (ADCP)], 6473 (encoding an ADCP), 6474 (encoding an ADCP), 6476 [encoding an ADCP-polyketide synthetase (PKS)], 6477 (encoding a PKS), and 6478 (encoding a PKS). To prove that the DBO1 genes identified by mutagenesis were indeed responsible for the defect in hemolytic activity observed, genetic complementation was achieved for cluster gene 6466 (Fig. 4). When gene 6466 is cloned behind a rhamnose inducible promoter in plasmid pSCRhaTc-6466 (forming plasmid p66), and introduced into DBO1 mutant 6466::TpR, the addition of 0.2% rhamnose results in full restoration of hemolytic activity as measured by hemoglobin release, whereas 6466::TpR/p66 without added rhamnose remains non-hemolytic.

Figure 4. Lysis of human erythrocytes by B. vietnamiensis DBO1 requires the intact NRPS cluster, including biosynthetic genes, Lux regulators and transporter. Supernatant absorbance (570 nm) was taken of each strain incubated in TSB + 5% human erythrocytes after shaking at 30°C for 48 h. Wt = wild type; 6466::TpR/p66 –rha and 6466::TpR/p66 +rha = complemented mutant 6466::TpR/pSCRhaTc-6466 incubated with and without 0.2% rhamnose, respectively. Error bars denote standard deviation around the mean from three independent trials performed in triplicate. Means labeled with different digits (1–3) are statistically different (p < 0.05, n = 3).

Galleria mellonella infection model

B. vietnamiensis DBO1 strains (wild-type, 6466::Tp and 6466::TpR/p66) were grown overnight and inoculated at 5 × 106 cfu into G. mellonella (greater wax moth) larvae. Larval death counts were taken every 24 h. The results indicate that the DBO1 NRPS product is toxic to the G. mellonella moth larvae, whereas a gene 6466 LuxR knockout mutant lacking the ability to produce the hemolysin has greatly reduced virulence. Genetic complementation of this gene knockout, as described above for hemoglobin release, showed similar restoration of the toxic effects toward G. mellonella upon the addition of rhamnose. Since moth larvae do not contain red blood cells, the results illustrated in Figure 5 suggest that the hemolytic NRPS product is not simply a hemolysin sensu stricto, but rather a cytotoxin, possibly active against several cell types including fungi (as demonstrated for B. contaminans MS14 occidofungin/burkholdine), insects, amoebae and higher organisms. To further this hypothesis, Bcc strains exhibiting NRPS hemolytic activity, as well as isogenic mutants defective in hemolysin production, were tested in a Dictyostelium discoideum feeding infection model. We were unable to detect differences in virulence toward grazing D. discoideum in Bcc strains either producing or not producing the NRPS toxin/hemolysin. We also observed that constructed mutant 6471::Tp, which is unable to produce the NRPS toxin glycosyltransferase, and that produced intermediate levels of hemoglobin release in the liquid hemolysis assay, did not produce intermediate levels of death in the G. mellonella infection model as compared with wild-type DBO1 and NRPS structural mutants such as 6477::Tp (data not shown). Instead, 6471::Tp behaved similarly to wild-type NRPS toxin-producing strains and produced wild-type levels of G. mellonella death.

Figure 5. DBO1 requires a LuxR homolog regulator for virulence against Galleria mellonella larvae. Five microliters of bacterial suspension, corresponding to 5 × 106 colony forming units, was inoculated into 10 larvae in n independent trials. The infected larvae were incubated at 30°C and viability counts were taken at 24 h intervals. [66TpR = strain 6466::TpR; p66 = 6466::TpR carrying complementation vector pSCRha-6466]. All results are means shown +/− SD. Means labeled with different digits (1–3) are statistically different (p < 0.01).

Prevalence of the NRPS gene cluster in the Bcc species

In order to determine the prevalence of occidiofungin/burkholdine-like compounds across members of the Bcc, we screened all available Bcc isolates in our library using PCR primers designed from homologous regions between known sequences of B. contaminans MS14 and B. ambifaria AMMD. The primers were also designed to amplify the regions within these genes showing no homology to other sequenced genes. Of 54 isolates screened, 13 yielded PCR products for at least two of the primer sets, and of those 13, ten yielded products in all three primer sets (Table 2). Using the same assay as described above, hemolysis of human erythrocytes by each of the 13 strains was investigated. Of the ten Bcc strains positive for all three PCR products, seven of them demonstrated high levels of hemolytic activity against human erythrocytes. Of the three Bcc strains positive for only two of the three PCR products, only one of these exhibited any hemolytic activity, and this was low level of activity (Fig. 6). In addition, analysis of other Burkholderia genomes currently available in the GenBank database indicate that no additional Bcc genomes (n = 8) contain NRPS genes or gene clusters similar to those identified in AMMD, DBO1 or MS14 (Table 2). It is important to note that the apparent presence of these NRPS genes is required for hemolytic activity but does not guarantee its production, possibly due to small, acquired mutations in the NRPS gene cluster in some Bcc strains. Overall, hemolytic or toxic activity produced by an NRPS-derived occidiofungin/burkholdine-like compound appears to be limited to the Bcc species B. ambifaria, B. contaminans,, Burkholderia pyrrocinia and B. vietnamiensis.

Table 2. Bacterial strains analyzed in this study

Species	Strain	PCR products	Hemolysis	Source/info
B. cepacia	ATCC25416T	ND	NT	Onion rot
B. cepacia	LMG18821	ND	None	CF isolate
B. cepacia	ATCC17759	ND	None	Soil
B. cepacia	CEP521	ND	None	CF isolate
B. multivorans	LMG13010T	ND	None	CF isolate
B. multivorans	ATCC17616	ND	None	Anthranilate enrichment
B. multivorans	PC249–2	ND	None	ATCC17616 mutant
B. multivorans	C5393	ND	None	CF isolate
B. multivorans	C3430	ND	None	CF isolate
B. multivorans	C5274	ND	None	CF isolate
B. multivorans	C5568	ND	NT	CF isolate
B. multivorans	CGD1	ND*	NT	CGD isolate
B. multivorans	CGD2	ND*	NT	CGD isolate
B. cenocepacia	J2315T	ND(*)	None	CF isolate
B. cenocepacia	K56–2	ND	None	CF isolate
B. cenocepacia	PC184	ND	NT	CF isolate
B. cenocepacia	715j	6472	None	CF isolate
B. cenocepacia	K63–3	6472	NT	CF isolate
B. cenocepacia	C1257	ND	None	CF isolate
B. cenocepacia	C4455	ND	None	CF isolate
B. cenocepacia	C5424	ND	NT	CF isolate
B. cenocepacia	C6433	ND	None	CF isolate
B. cenocepacia	BC7	ND	None	CF isolate
B. cenocepacia	CEP511	ND	None	CF isolate
B. cenocepacia	AU1045	ND*	NT	CF isolate
B. cenocepacia	HI2424	ND*	NT	Soil rhizosphere
B. cenocepacia	MC0–3	ND*	NT	Soil rhizosphere
B. cenocepacia	D1	ND	NT	Environmental isolate
B. stabilis	LMG14294T	6472, 6474, 6476	None	CF isolate
B. stabilis	LMG18870	ND	None	CF isolate
B. vietnamiensis	LMG10929T	6472, 6474, 6476	None	Soil rhizosphere
B. vietnamiensis	DB01	6472, 6474, 6476	High	Phthalate enrichment
B. vietnamiensis	LMG18835	6472, 6474, 6476	High	CF isolate
B. vietnamiensis	G4	ND(*)	NT	Trichloroethene enrichment
B. dolosa	LMG18943T	ND	None	CF isolate
B. dolosa	L06	ND	NT	CF isolate
B. dolosa	AU0645	ND	NT	CF isolate
B. dolosa	STM1441/LMG21443	ND	None	Soil rhizosphere
B. dolosa	CEP021	ND	NT	CF isolate
B. dolosa	E12	ND	NT	CF isolate
B. ambifaria	AMMDT	6472, 6474, 6476	High	Soil rhizosphere
B. ambifaria	CEP0996	6472, 6474, 6476	High	CF isolate
B. ambifaria	LMG17828	6472, 6474, 6476	None	Soil
B. ambifaria	M53	6472, 6474, 6476	High	Soil
B. pyrrocinia	LMG14191T	6472, 6474	Low	Soil rhizosphere
B. pyrrocinia	LMG21822	6472, 6474, 6476	Moderate	Soil
B. pyrrocinia	LMG21823	6472, 6474, 6476	None	Water
B. pyrrocinia	LMG21824	6472, 6474, 6476	High	CF isolate
B. anthina	W92T	ND	NT	Soil rhizosphere
B. anthina	J2552	ND	NT	Soil rhizosphere
B. anthina	C1765	ND	NT	CF isolate
B. anthina	AU1293	6472	NT	CF isolate
B. ubonensis	Bu	ND*	NT	Soil
B. contaminans	MS14	6472, 6474, 6476*	NT	Soil
B. lata	383	ND*	NT	Soil
Burkholderia sp	JS150	ND	NT	p-dichlorobenzene enrichment
R. pickettii	ATCC27511T	ND	NT	Patient isolate
R. pickettii	YH105	6472, 6474	None	p-nitrobenzoate enrichment

Superscript “T” following the strain name denotes the Type strain of each species. ND, not detected. NT, not tested. Asterisks indicate the result was obtained using genomic analysis. CF, cystic fibrosis; CGD, chronic granulomatous disease.

Figure 6. Human erythrocyte lysis generally requires the presence of the NRPS cluster in B. cepacia complex isolates. Forty-seven Bcc strains, one untyped Burkholderia species, and two R. picketii strains and were screened using PCR primers for three of the non-ribosomal peptide biosynthesis genes within the cluster. Of the 50 strains, 13 were positive for at least two of the genes, and 10 were positive for all three genes tested. White columns = strains positive for two of the genes, black columns = strains positive for all three genes tested. A570 of the supernatants were taken after 48 h incubation of the strains in TSB + 5% human erythrocytes in triplicate. Error bars = +/− SD of triplicates.

Discussion

Bcc species B. contaminans MS14 produces cellular toxins that are synthesized using a non-ribosomal peptide synthase pathway, which is characteristic of complex secondary metabolite compounds. These compounds, known as occidiofungins or burkholdines, have been previously shown to have antifungal activity,, but herein we show that these compounds, or compounds closely related to occidiofungins/burkholdines (based partially on biosynthetic gene cluster similarities), also possess high levels of hemolytic activity. Besides individual ORF homology and syntenic organization between the gene clusters in MS14 and AMMD/DBO1, the hemolysis results from the glycosyltransferase mutant in DBO1 suggest that the NRPS-derived compounds from the three Bcc strains are similar. Prior structural analysis of the MS14 NRPS compound suggests that the ring peptide can be glycosylated by xylose, and that without this glycosylation, the antifungal properties of occidiofungin/burkholdine are lost. However, genetic inactivation of B. vietnamiensis DBO1 NRPS gene cluster homolog 6471, encoding the putative glycosyltransferase, does not eliminate hemolytic activity completely, but does reduce hemolytic activity by approximately half, whereas in G. mellonella, the unglycosylated compound still exhibits wild-type bacterial toxicity. This suggests that the unglycosylated form of the occidofungin/burkholdine-like compound is fully toxic towards some organisms such as in insects, but only partially active against other cell types, including erythrocytes and fungal cells. Because constructed mutant 6471::Tp does not produce intermediate levels of killing in the G. mellonella infection model as compared with wild-type DBO1 and NRPS structural mutants, this suggests that the organismal assays used to test toxicity are less sensitive than the hemoglobin release cellular lysis assay.

To examine whether other Bcc strains possess the NRPS gene cluster, PCR primer sets localized to unique sequences were used to probe for the presence of two NRPS genes. This analysis identified 13 of 55 Bcc strains potentially carrying the NRPS gene cluster. Further analysis with another unique PCR primer pair reduced this number to ten strains that carried three NRPS genes. Upon functional testing, only seven of these ten strains exhibited hemolysis of human erythrocytes, suggesting that two strains (B. pyrrocinia LMG 14294 and B. vietnamiensis LMG 10929) possessed undefined mutations to the hemolytic gene cluster. In addition, only two PCR products were amplified from B. ambifaria CEP0996; however, this strain exhibited hemolytic activity in this assay. This suggests that either one set of primer pairs was degenerate from the CEP0996 genomic sequence to allow amplification of the NRPS ortholog from this strain even though a functional protein was expressed, or that this strain produces a hemolytic toxin different from occidiofungin/burkholdine. In terms of prevalence, it is interesting to note that this NRPS gene cluster has not been identified in the two most clinically important Bcc species, Burkholderia multivorans and B. cenocepacia. Although these two species are found both as environmental and clinical isolates, and clinical Bcc strains can exhibit relatively little difference from strains found in the environment, the identification of this NRPS gene cluster in Bcc species better adapted to soil environments than in association with humans suggests that this gene cluster evolved to protect the Bcc from ecological niche predators such as fungi and amoeba rather than as a virulence factor to assist invasion of human tissue.

The literature on NRPS and PKS systems indicates that associated dedicated glycosyltransferases are not uncommon. Although the putative xylose moiety appears to enhance hemolytic activity of the Bcc NRPS-derived compound, the exact mechanism through which this is achieved awaits further investigation. Walsh et al. describes a possible function of such glycosyltransferases as having a role in imparting polarity or solubility to hydrophobic compounds, thus improving their access into target cell surfaces. The results presented in this and previous publications on occidiofungin/burkholdine, suggest that glycosylation is important for full toxicity of the compound.

Several key questions remain with respect to the function of some ORFs found within the NRPS cluster. For example, gene 6470 downstream of the glycosyltransferase 6471 encodes a hypothetical protein that upon mutagenesis, appears to have no effect on hemolytic or toxic activity. Gene 6467, overlapping the two luxR homolog ORFs, is a small gene encoding a potential product of only 110 amino acids, with no homology to any known protein. This may be a truncated ORF that was introduced along with the upstream luxR family transcriptional regulator gene 6468 during acquisition and assembly of the NRPS cluster. Finally, gene 6475, situated between two large genes encoding proteins with amino acid adenylation domains, encodes a 538 amino acid protein that contains a β-lactamase domain. Rather than providing cellular resistance to antibiotics, it is likely that this protein is somehow involved in modifying the cyclic peptide structure, as all homologs to this particular protein are associated with NRPS and PKS gene clusters. Because we were unable to mutagenize gene 6475, a definitive function for the associated protein remains to be assigned. All other genes in the strain DBO1 NRPS gene cluster are essential for hemolytic activity.

Given the interest in developing the Bcc NRPS-derived compound known as occidiofungin or burkholdine as an antifungal agent for agricultural (crop preservation and protection) or pharmaceutical use, it is important to note its toxic and hemolytic activities. Clearly, the development of biological-based fungicides should include extensive analysis of their toxic properties. A previous study has noted that occidiofungin/burkholdine produces aberrant cell membrane morphology in fungi similar to what has been described for the echinocandin class of antifungal compounds, which disrupt β-glucan polymerization. A recent study describes occidiofungin’s in vitro potency against Candida species, as well as its chemical stability in the presence of human serum. This finding suggests that if occidiofungin/burkholdine displays pathogenic traits against humans, the compound may resist degradation within the body. Furthermore, a toxicological evaluation of occidofungin is described where B6C3F1 mice were given a single dose of occidiofungin up to 20 mg/kg body weight or a daily dose for 5 d at 2 mg/kg body weight. Histological examination of treated mice did not identify organ toxicity; however, overall effects were a reduction in body and organ weights, and anemia or neutropenia were not tested. Our results suggest that the Bcc occidofungin/burkholdine-like cyclic peptide targets and disrupts components of the membranes of eukaryotic cells (but not prokaryotic cells), especially erythrocytes, potentially binding to cholesterol or another cell scaffolding carbohydrate component. Future experiments will be aimed at further investigation of Bcc NRPS-derived hemolysins and their mechanism of action against erythrocytes.

Materials and Methods

Bacterial strains, plasmids, antibiotics and culture conditions

Luria-Bertani (LB) broth was prepared in distilled water to half concentration for Bcc strains and full concentration for Escherichia coli strains. Sheep, horse, bovine and rabbit blood broth (Dalynn Biologicals) and human blood broth (Canadian Blood Services) were prepared to 5% v/v blood in trypticase soy broth (TSB). Agar plates were prepared with 1.5% w/v agar. Antibiotics were included where necessary to the following concentrations: ampicillin (amp), 100 μg/mL; trimethoprim (Tp), 100 μg/mL; tetracycline (Tc), 100 μg/mL. All antibiotics were purchased from Difco through BD Canada. E. coli DH5α was used for plasmid manipulation. E. coli strains were grown at 37°C, while DBO1 was grown at 30°C. All liquid cultures were shaken at 225 rpm. High-copy number plasmid pJET (AmpR) was employed in cloning experiments. For random plasposon mutagenesis, pTnMod-OTp′ (TpR) was used. To complement gene 6466, pSCRhaB2 was used. A list of bacterial strains and plasmids used or constructed is shown in Table 1.

Random plasposon mutagenesis

A random plasposon insertion mutant library was created in B. vietnamiensis DBO1 using plasposon pTnMod-OTp′ as previously described. Cells containing integrated plasposons were selected on 1/2 LB + 100 μg/mL trimethoprim. Twenty thousand mutants were patched onto 5% sheep blood TSA agar, incubated at 30°C and observed for hemolysis after 48 h. Mutants exhibiting loss of hemolysis were carried forward and their plasposon insertion sites were isolated as previously described. To ensure that the random plasposon mutants obtained carried authentic mutations responsible for the observed phenotype, plasposon site of insertion plasmid clones were reintroduced into wild-type DBO1, and following recombination and selection on 1/2 LB + 100 μg/mL trimethoprim, and PCR analysis to ensure proper integration, the reconstructed mutants were again tested for hemolytic activity on 5% sheep blood TSA agar.

Targeted mutagenesis

Once the NRPS cluster had been identified through random plasposon mutagenesis, mutations in the remaining genes within the cluster were created through a targeted insertion mutagenesis strategy using homologous recombination. Primers were designed to amplify open reading frames (ORFs) and XbaI sites at either end. A full list of primers is shown in Table 3. PCR amplification was performed using TopTaq (Qiagen Inc.). In many cases, because of imperfect local homology between the AMMD and DBO1 sequences, techniques were used to enhance primer binding in order to obtain PCR products. DNA bands were extracted from agarose gel using GeneClean (Fermentas), ligated to pJET1.2 (Fermentas) and cloned in E. coli DH5α (Invitrogen Corp.), and DNA sequencing of the cloned fragments was performed to confirm the identity of the product. To generate plasmids containing insertions, the OriTpR region was extracted from pTnMod-OTp′ by digestion with BglII and XbaI, ligated to the PCR products using T4 ligase (Promega Corp.) and the three-way ligation product was cloned into E. coli DH5α, with selection of transformants on LB + 100 μg/ml Tp. OriTp-insert plasmids were subsequently extracted from their DH5α hosts, electroporated into B. vietnamienesis DBO1, and transformants were selected on ½ LB + 100 μg/ml Tp. Mutations were confirmed by PCR and sequencing using previously described methods.

Table 3. Oligonucleotides used in this study

Oligonucleotides	Sequence
Mutagenesis*
m6465–1	GGTTCGACATTCTGACGTT
m6465–2	CCTCTATGTGCCGAACAG
m6466–1	AGAGTCAGATGTTCGCGAAG
m6466–2	GTCTCGACCGTGCTTTCC
m6470–1	AAGCGGCGTTCGTCAAGTC
m6470–2	AGGTGGCTGAGTTCGACATTG
m6471–1	AAGGTCTGCATCAATCTGG
m6471–2	AGGGAATAGGTCAGCGGC
m6477–1	GCCGTTCTGCAACTACATCC
m6477–2	AGGCGGTCGGTCAGTTCG
m6479–1	ATGGTACCGGCGTCCTTCGAATC
m6479–2	ATAAGCTTGGCAGACGTCGGGTT
Complementation*
c6466–1	ATAATACATATGCATCATCACCATCACCACGTTCGCGAAGCTTG
c6466–2	ATTCTAGACTACGCCGCCGACGCGCAC
Sequencing
pJET-F	CGACTCACTATAGGGAGAGCGGC
pJET-R	AAGAACATCGATTTTCCATGGCAG
pBBR-F	AACAGCTATGACCATG
pBBR-R	AATACGACTCACTATAG
Bcc cluster screening
s6472–1	TACATGCTCGACGACGCGCT
s6472–2	ATGTTGTAGGTGGCCGACGGG
s6474–1	TCTCGACCAGCGGGCAATACC
s6474–2	TCCTCGATCATGAAGCGCAG
s6476–1	AAGGTCACGTGGTTCGGCTCG
s6476–2	ATTTCGCGGACCAGTTCGGC

Restriction sites shown in bold. 6x histidine tag underlined.

Genetic complementation

To restore the mutated hemolysis phenotype to DBO1 mutant 6466::Tp we generated a PCR product of B. ambifaria AMMD gene 6466, a putative LuxR regulator of the NRPS gene cluster, and cloned this into the rhamnose-inducible plasmid pSCRhaB2 modified with a tetracycline cassette from p34S-Tc with SmaI and ligated into the unique EcoRV site of pSCRhaB2. The 6466 PCR construct was amplified with an N-terminal 6x histidine tag and NdeI and XbaI sites at the 5′ and 3′ termini (Table 3), respectively, and inserted into the NdeI and XbaI sites of pSCRhaB2 following its NdeI + XbaI digestion and purification from a 0.8% agarose gel using GeneClean. Triparental mating was performed to transform mutant 6466::Tp with the plasmid; meanwhile, wild type DBO1 and 6466::Tp were similarly transformed with a blank version of the same TcR-carrying plasmid, produced by digestion with NdeI + XbaI, purification with GeneClean, digestion with Mung Bean exonuclease (Promega Corp.) and self-ligation, to control for the physiological effects of the vector. Fifty pSCRhaTc-6466 transformants were screened on blood agar with and without 0.2% rhamnose to identify transformants expressing hemolysin; of these, approximately half showed rhamnose-dependent hemolytic activity. Two complemented mutants were carried forward, and one of these was used in virulence experiments.

Liquid hemolysis assay

To quantitatively compare hemolytic activity produced by B. vietnamienesis DBO1 with that produced by mutants of the NRPS cluster constructed in DBO1 (and subsequently, other strains exhibiting positive PCR results for genes 6472, 6474 and 6476), a liquid hemolysis assay was developed. DBO1 and mutants were grown in 2 mL sheep blood broth for 48 h in duplicate. One milliliter of supernatant was assayed at 24 h and 48 h and absorbance was measured using an Ultrospec 3000 (Pharmacia Biotech) at a range of wavelengths from 350 nm to 700 nm (Fig. 3). Viable bacterial plate counts were obtained by serial dilution and plating on ½ LB and red blood cell counts were taken using a Bright-Line Hemacytometer (Hausser Scientific). This method measured released heme at the optimal wavelengths of 540 nm and 570 nm. A higher-throughput assay was then developed based on these results. The strains were grown for 24 h in four separate 200 μL TSB with appropriate antibiotics in a Costar® 3599 96 Well Culture Cluster (Corning Inc.). Complemented mutant 6466::Tp/pSCRhaTc-6466 was grown with and without 0.2% rhamnose. OD600 was taken of all cultures, and three cultures falling within 10% of OD600 = 1.1 were selected for each strain; cultures of mutant 6473::Tp did not grow within this cut-off, so this strain was not used in the experiment. Five microliters of each cell suspension was added to quadruplicates of 200 μL 5% sheep blood in TSB in 96-well plates and incubated with shaking at 30°C; to reduce evaporation from wells, plates were plastic-wrapped. After 48 h, 180–200 μL was removed from each well to a microfuge tube and centrifuged at 1,000x g for 15 min. Carefully, avoiding any cell pellet, 140 μL of each supernatant was removed into a 96-well plate and OD570 was measured. Averages of quadruplicates for each trial were taken and each quadruplicate average statistically represented 1 n. Student's t-tests were performed to determine statistical significance; to determine t-test type, an F-test was performed for each comparison.

Wax moth larva killing assay

To compare in vivo virulence of wild-type B. vietnamiensis DBO1 with NRPS cluster gene insertion mutants, G. mellonella (greater wax moth) larvae were infected. The standard protocol from Seed and Dennis was followed with several adjustments. Bacterial inoculums used were 5 × 106 c.f.u. and were OD600 standardized prior to injection. Ampicillin was omitted from the buffer due to its inhibitory activity against DBO1; to compensate for the lack of antibiotic, larvae were dipped in ethanol and dried on paper towel prior to injection. DBO1/pSCRhaTc, 6466::Tp/pSCRhaTc and 6466::Tp/pSCRhaTc-6466 were grown at 37°C for 24 h on LB agar + Tp + Tc + 0.2% rhamnose, and an additional 6466::Tp/pSCRhaTc-6466 culture was grown without rhamnose for reduced 6466 expression. All strains except 6466::Tp/pSCRhaTc-6466 grown without rhamnose were injected with 0.4% rhamnose added to the 10 mM MgSO4 suspension. The control injections included 10 mM MgSO4, Tp, Tc, and 0.4% rhamnose. Injections were spread over a period of 4 d. Separate cultures were grown, suspended in buffer, and diluted to OD595 = 0.170 (200 μL in a 96-well plate) for each set of 10 larvae to be injected, with each set of 10 larvae statistically representing 1 n. Infected larvae were then incubated at 30°C for 120 h. Mortality data were analyzed statistically by a Student's t-test; to determine t-test type, an F-test was performed for each comparison.

Prevalence of the hemolysin NRPS cluster in the Bcc

To determine the prevalence of the NRPS cluster among clinical and environmental Bcc isolates, 47 Bcc strains, one untyped Burkholderia species and two Ralstonia picketii strains were screened using primers (Table 3) based on regions of homology between the known sequences of B. contaminans and B. ambifaria AMMD but displaying no homology beyond those known sequences. Genes corresponding to B. ambifaria strain AMMD genes 6472, 6474 and 6476 were chosen for this screen because these appear to be three of the most significant NRPS biosynthetic genes of the cluster, and their larger size offered more options for primers fitting the above criteria. Colony PCR was performed on each isolate and the PCR products were separated on a 0.8% agarose gel. Strains exhibiting correctly sized products for at least two of the three primer pairs were analyzed further using the liquid hemolysis assay. An additional 10 strains were examined through bioinformatic comparison, although two of these strains had also been tested using PCR. Paired BLASTN sequence analysis was performed using each of AMMD genes 6472, 6474 and 6476 along with the GenBank deposited genome files, using default algorithmic settings. Overall, a total of 58 bacterial strains or isolates were tested for the presence of these NRPS genes, including 55 Bcc strains.