Introduction

Recently, Burkholderia spp. have been described that are not only endosymbionts in Rhizopus microsporus, a saprotrophic fungi, but are also responsible for the production of the toxins rhizoxin and rhizonin which had been previously believed to have been produced solely by the fungi [1], [2], [3]. Rhizoxin is a important virulence factor for infection of plants by Rhizopus and has antimitotic activity [4], [5], [6]. Rhizonin is a cyclopeptide exhibiting fatal hepatotoxic effects [3], [5], [7], [8]. Burkholderia rhizoxinica produces rhizoxin and is now associated with the ability of Rhizopus to cause rice seedling blight. Burkholderia endofungorum was recently shown to produce rhizonin. Both bacteria appear to reside in the cytosol of the fungal cell [6], [9].

Clinical isolates H2199 (Ohio - 2002), H3620 (New Mexico - 2005), H500 (California - 1997), G8810 (North Carolina - 1993), G7344 (Oregon - 1992), H2592 (District of Columbia - 2003), H3977 (South Dakota - 2006), and G4101 (New York - 1989) were submitted to the CDC from 1989 to 2006 by various public health agencies in the U.S. and placed in our archive. All but one were isolated from blood specimens with the exception being H2592 which was from a wound. These isolates were obtained from five adult males and three adult females. Later enquiry on the clinical background of these isolates did not yield any further information. Biochemical characterization was not able to provide identification; however, cellular fatty acid analysis of H500, G7344, G8810 and G4101 indicated profiles similar to that for Burkholderia pseudomallei and/or the Burkholderia cepacia complex. Analysis of the 16S rRNA gene sequences also indicated that these isolates were members of the genus Burkholderia, but did not yield a specific identity.

Based on the new descriptions of B. rhizoxinica and B. endofungorum, we re-examined the isolates in our archive using standard biochemical testing, 16S rRNA gene sequencing and DNA-DNA hybridization. We now identify these clinically derived strains as either B. rhizoxinica or B. endofungorum. We also further characterize them by cellular fatty acid analysis (CFA) and/or and multiple locus sequence typing (MLST).

Materials and Methods

Biochemical testing was performed on all strains and cellular fatty acid (CFA) analysis was performed on strains H500, G7344, G8810 and G4101 as described by Weyant et. al. [10].

We performed DNA-DNA hybridization as previously described [11]. In brief: Cells were harvested and lysed, and the chromosomal DNA was isolated and purified. DNA from the type strains of B. rhizonica (HKI 454^T) and B. endofungorum (HKI 456^T) were labeled with [³²P]dCTP using a commercial nick translation kit (Invitrogen Life Technologies, Carlsbad, CA) and tested for reassociation to unlabeled DNA from the same strains (homologous reaction). Reassociation of B. rhizonica DNA was tested with DNA from H3977 and H2199 and that of B. endofungorum was tested with DNA from G4101. A reciprocal reaction using labeled G4101 DNA was also performed. Relative binding ratios and percent divergence were calculated as described previously [11].

DNA sequencing was performed using methods and primers as previously described. In brief: whole cell suspensions of bacteria were used for PCR. Bacteria were grown by plating one loop (1 µl) of stock cell suspension (heavy suspension of Burkholderia spp. in defibrinated rabbit blood, stored at −70°C until use) on trypticase soy agar with 5% defibrinated sheep blood (SBA) (BBL Microbiology systems, Cockeysville, MD) and incubating aerobically 1–2 days at 37°C. A single colony was suspended in 200 µl of 10 mM Tris, pH 8.0 in a 1.5 ml Millipore 0.22 µm filter unit (Millipore, Bedford, MA), heated at 95°C for 30 min, and centrifuged at 6000× g for 5 min. Each final PCR reaction (100 µl) contained 5 U of Expand DNA polymerase (Boehringer Mannheim, Mannheim, Germany), 2 µl of DNA solution in H₂O, 10 mM Tris-HCl (pH 8.0), 50 mM KCl, 1.5 mM MgCl₂, 200 µM (each) dATP, dCTP, dGTP, and dTTP, and 0.4 µM of each primer. Reactions were first incubated for 5 min at 95°C. Then, 35 cycles were performed as follows: 15 sec at 94°C, 15 sec at 60°C, and 1 min and 30 sec at 72°C. Reactions were then incubated at 72°C for an additional 5 min. PCR products were purified with Qiaquick PCR purification kit (Qiagen, Valencia, CA). Sequencing was performed using an Applied Biosystems (ABI) BigDye terminator cycle sequencing ver 3.1 kit as per the manufacturer's instructions, except 0.25 µl of BigDye were used instead of 8 µl (Applied BioSystems, Foster City, CA). Sequencing products were purified by using Centri-Sep spin columns (Princeton Separations, Adelphia, NJ), and were resolved using an Applied Biosystems model 3130xl automated DNA sequencing system (Applied Biosystems) [9], [12], [13], [14]. Analysis was performed using the Accelrys GCG package ver 10.3 (Accelrys, San Diego, CA) and MEGA 3 as previously described [14], [15].

The 16S rRNA gene sequences obtained for strains G7344, G8810, H500, H2199, H2592, H3620, H3977 and G4101 were deposited in GenBank (Supporting Information S1). Analysis of 16S rRNA sequences were performed as previously described [13]. A reconstruction of the phylogenetic relationship of the bacterial isolates by multi-locus sequence typing (MLST) based on six gene loci (gltB, gmhD, lepA, lipA, ndh, and rhiE) was performed as previously described [9] and sequences were deposited in GenBank (Supporting Information S1).

The obtained dataset consisted of 3160 nucleotides including 710 variable sites, 263 of them being parsimony informative. Phylogenetic analyses were carried out in a Bayesian and distance matrix framework. For Bayesian analyses we employed the GTR+I+G model of nucleotide evolution. Bayesian tree sampling was performed using the MrBAYES 3.1.1 program [16], [17]. Two analyses including 500,000 generations each were run in parallel. Four chains were run simultaneously. Trees were sampled every 100th generation. The first 3,000 generations were deleted as the “burn in” of the chain. The remaining trees were summarized as 50% majority rule consensus trees. Phylogenetic trees were visualized using the program Treeview [18]. Neighbor joining trees were constructed using MEGA version 3.1 [15]. Branch support for NJ trees was obtained by performing 10000 bootstrap replicates. The corresponding Burkholderia pseudomallei ST1 (BPS) sequence was used as an outgroup to root the tree. For the construction of phylogenetic networks with the neighbor-net method we used the program SplitsTree version 4.8 [19].

For secondary metabolite analysis the bacterial strains were cultured on nutrient agar plates at 30°C. After 3 days a single colony was used to inoculate 1 ml TSB. After 2 days cultivation at 30°C and 120 rpm, 1 ml fresh TSB was added. After another 2 days cultivation, the grown culture was transferred to 20 ml TSB and again incubated at 30°C (120 rpm) for 48 h. An aliquot of 5 ml was used to inoculate 100 ml of production medium (1% corn starch, 0.5% glycerol, 1% gluten meal, 1% dried yeast, 1% corn steep liquor, and 1% CaCO3, pH = 6.5). Cultivation took place at 30°C, 120 rpm, for 4 days.

Extraction and HPLC analysis was performed as described previously [20].

Results

These eight clinical isolates shared common phenotypic characteristics. They were nonmotile, gram-negative, coccobacilli. They produced oxidase but did not produce catalase, indole, urease, did not reduce nitrite, did not hydrolyze gelatin or esculin, and did not utilize citrate. Nitrate was reduced by only one isolate H3977. No acid production in the slant or butt of triple-sugar iron agar was noted. There was no growth on MacConkey agar, Salmonella Shigella agar or on cetrimide agar. No acid production was observered in King's oxidation-fermentation base from D-glucose, D-xylose, mannitol, lactose, sucrose, and maltose. Variable reactions were produced with litmus milk. We were unable to demonstrate catalase production and motility with neither of these clinical strains or with the type strains of B. rhizoxinica (HKI 454^T) and B. endofungorum (HKI 456^T) in our laboratory even though these characteristics were reported by Partida-Martinez and coworkers [6]. Partida-Martinez and coworkers noted that growth of these bacteria was poor in pure culture and did not allow for consistent biochemical characteristics [6].

Analysis of the 16S rRNA gene sequences indicates that H2199, H3620, H500, G8810, G7344, H2592, H3977 have >99.4% identity to the 16S rRNA gene sequence for the B. rhizoxinica type strain (HKI 454^T) and G4101 has an identity of 99.4% to the sequence for the B. endofungorum type strain (HKI 456^T) (Figure 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. MEGA 3 analysis of 16S rRNA gene sequences using neighbor-joining and Kimura 2-parameter settings with 1000 step bootstrap.

Phylogenetic tree of Burkholderia rhizoxinica and B. endofungorum based on comparisons with 16S rDNA sequences of related bacteria. Pseudomonas aeruginosa DSM 50071 (GenBank accession no. X06684) was used as an outgroup for this analysis. Bar, 2% sequence dissimilarity.

https://doi.org/10.1371/journal.pone.0015731.g001

Four strains were selected for cellular fatty acid analysis (CFA). H500, G7344, G8810 and G4101 shared a unique profile which is easily recognized by the presence of two cyclopropane acids (17:0cyc, 19:0cyc11-12), 16:0 and 18:1w7c as major acids (8–27%), and smaller amounts (1–5%) of six hydroxy acids (3-OH-14:0, 2-OH-16:1, 2-OH-16:0, 3-OH-16:0, 2-OH-18:1, 2-OH-19:0cyc). The CFA composition of H500, G7344, G8810 and G4101 is consistent with that of the type strains of B. rhizoxinica and B. endofungorum [6], and is most similar to that of the CFA group containing B. cepacia, B. gladioli, B. mallei and B. pseudomallei [10].

The results of DNA relatedness studies are given in Table 1. Isolates H3977 and H2199 exhibited greater than 78% relatedness (RBR) under both the optimum and stringent reassociation criteria and had divergence (D) of less than 5 to the labeled DNA from the B. rhizoxinica type strain. Isolate G4101 exhibited 76% relatedness under optimum criteria, but 69% under the more stringent conditions when matched with labeled DNA from the B. endofungorum type strain, however in the reciprocal reaction using labeled G4101 the relatedness is over 70% for both the optimum and stringent reassociation criteria.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. DNA/DNA hybridization.

https://doi.org/10.1371/journal.pone.0015731.t001

The results of the MLST indicate that strains G8810, H2190, H500, G7344, H2592, and H3620 are members of the B. rhizoxinica clade called the “Pacific group” (97.7% to 99.4% identity with B. rhizoxinica type strain), whereas isolate H3977 is more remotely related to this group (95.9% identity) (Figure 2). Strain H3977 is most likely not a member of the known subclades and appears to be the only known member of a new group within the complex. Alleles for isolate G4101 cluster closely with Burkholderia spp. from the “Eurasian branch” of endofungal symbionts (99.3% to 99.8% identity) and still has 95.7% identity with the B. endofungorum type strain from Mozambique.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. Phylogenetic network of Burkholderia spp from. clinical isolates together with fungal symbionts (B1–B8) based on MLST data.

B1: Burkholderia rhizoxinica type strain. B5: B. endofungorum type strain. BPS: B. pseudomallei (outgroup).

https://doi.org/10.1371/journal.pone.0015731.g002

To verify the production of toxins by the clinical isolates the metabolic profiles of two bacterial strains were investigated. Strain H2199 and G7344 proved to be culturable under the conditions optimized for rhizoxin formation whereas we were unable to grow G4101 for the toxin study. HPLC and MS analyses clearly showed that H2199 and G7344 produce high amounts of cytotoxic rhizoxin analogues (Figure 3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 3. HPLC profiles of culture extracts of strains H2199 (A) and G7344 (B) showing the production of various antimitotic rhizoxin derivatives.

https://doi.org/10.1371/journal.pone.0015731.g003

Discussion

Recently, the toxin rhizonin was shown to be produced by endosymbiont B. endofungorum and not by the host Rhizopus [3]. It is unknown if the clinical isolate (G4101) of B. endofungorum produces rhizonin, but the production of this toxin could be clinically significant since rhizonin is a known hepatotoxin. Further studies will determine if this strain produces rhizonin. However, in all hitherto examined cases there has been no evidence for the involvement of Burkholderia symbionts for the development of zygomycoses [21], [22].

The presence of an acyl transferase AT gene (rhiE) from the rhizoxin biosynthesis gene cluster suggests that the clinical isolates of B. rhizoxinica might have the capacity to produce cytotoxic polyketides [23]. By metabolic profiling of the bacterial cultures, we unequivocally showed that strains H2199 and G7344 produce significant amounts of rhizoxin derivatives. The production of rhizoxin could influence the course of human infection because the toxin has anti-mitotic activity in mammalian cells and has potential as an antitumor drug. Previous work has demonstrated that derivatives of rhizoxin vary in anti-mitotic activity [20]. Further studies will indicate whether these isolates produce the toxin or derivatives of the toxin.

We report for the first time that strains of B. rhizoxinica and B. endofungorum have been associated with human clinical specimens. Various Burkholderia spp. are known pathogens with B. pseudomallei causing melioidosis and B. mallei causing glanders [13], [24]. There are also cases of opportunistic infection by less pathogenic members of the Burkholderaea such as in the case of B. cepacia infections, especially among cystic fibrosis patients, and infection by B. thailandensis [25], [26].

Since members of the B. rhizoxinica complex are known to form tight associations with their host, fungal involvement is possible, however there was no known detection of fungal infections in these cases. Unfortunately the clinical records for these isolates is incomplete and further enquiries did not prove productive. The presence of the bacteria by contamination of the specimens cannot be excluded. Seven of the isolates were derived from blood and one from wound tissue. Since Rhizopus are ubiquitous and some are opportunistic human pathogens, it is possible that the patients were colonized by the fungi and that culturing was attempted only to detect bacterial growth, thus missing the presence of Rhizopus [27]. When B. rhizoxinica or B. endofungorum are detected in clinical specimens, clinicians may wish to check for the presence of fungal involvement.

Supporting Information

Acknowledgments

We thank Mary Brandt and Christina Scheel for thoughtful discussions about Rhizopus and fungal infections.

The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention/the Agency for Toxic Substances and Disease Registry. Names of vendors or manufacturers are provided as examples of available product sources; inclusion does not imply endorsement of the vendors, manufacturers, or products by the Centers for Disease Control and Prevention or the U.S. Department of Health and Human Services.