Diverse Burkholderia Species Isolated from Soils in the Southern United States with No Evidence of B. pseudomallei

The global distribution of the soil-dwelling bacterium Burkholderia pseudomallei, causative agent of melioidosis, is poorly understood. We used established culturing methods developed for B. pseudomallei to isolate Burkholderia species from soil collected at 18 sampling sites in three states in the southern United States (Arizona (n = 4), Florida (n = 7), and Louisiana (n = 7)). Using multi-locus sequence typing (MLST) of seven genes, we identified 35 Burkholderia isolates from these soil samples. All species belonged to the B. cepacia complex (Bcc), including B. cenocepacia, B. cepacia, B. contaminans, B. diffusa, B. metallica, B. seminalis, B. vietnamiensis and two unnamed members of the Bcc. The MLST analysis provided a high level of resolution among and within these species. Despite previous clinical cases within the U.S. involving B. pseudomallei and its close phylogenetic relatives, we did not isolate any of these taxa. The Bcc contains a number of opportunistic pathogens that cause infections in cystic fibrosis patients. Interestingly, we found that B. vietnamiensis was present in soil from all three states, suggesting it may be a common component in southern U.S. soils. Most of the Burkholderia isolates collected in this study were from Florida (30/35; 86%), which may be due to the combination of relatively moist, sandy, and acidic soils found there compared to the other two states. We also investigated one MLST gene, recA, for its ability to identify species within Burkholderia. A 365bp fragment of recA recovered nearly the same species-level identification as MLST, thus demonstrating its cost effective utility when conducting environmental surveys for Burkholderia. Although we did not find B. pseudomallei, our findings document that other diverse Burkholderia species are present in soils in the southern United States.


Introduction
The Gram-negative genus Burkholderia is composed primarily of diverse soil-dwelling bacteria that play a variety of ecological roles as saprophytes, nitrogen-fixing mutualists, and pathogens. The genus includes plant pathogens, such as B. gladioli and B. glumae, and two species that are highly pathogenic to humans and other animals (B. pseudomallei and B. mallei). Burkholderia pseudomallei, the causative agent of melioidosis, is endemic to Southeastern Asia and Australasia [1,2]. However, melioidosis also is reported sporadically in other locations of the world, including the Americas, Africa, the Middle East, and various island communities [1]. As a result, the current known global distribution of B. pseudomallei is thought to be just "the tip of the iceberg" [3].
Seroreactivity to B. pseudomallei antigens has been observed in healthy U.S. individuals [4], possibly as a result of exposure to B. pseudomallei or genetic near neighbor species. Five naturally acquired human melioidosis cases [5,6] and four patients infected with genetic near neighbors of B. pseudomallei (B. oklahomensis and B. thailandensis) [7,8] have been described in the U.S. Despite the possible presence of B. pseudomallei and its close genetic near neighbors in North America, only B. oklahomensis and B. thailandensis have been cultured from environmental samples [5,7,8]. However, members of the more distantly related B. cepacia complex (Bcc), which contains numerous opportunistic human pathogens [9][10][11], are frequently isolated in North America.
The diverse taxa of the Bcc have received increased attention due to their importance to plants, agriculture, and human health. One species belonging to this group, B. vietnamiensis, has the ability to fix nitrogen, which allows it to form mutualistic relationships with rice plants [12]. A well-known strain of B. vietnamiensis (G4) is especially interesting due to its ability to degrade common organic pollutants. This strain was isolated from a wastewater treatment facility in Florida and is now used for bioremediation [13]. Other species within the Bcc (B. ambifaria, B. cenocepacia, B. cepacia) have been identified as significant pathogens to commercially valuable plants, such as onions and bananas [14,15]. Several members of the Bcc, as well as the more distantly related B. gladioli, have been described as opportunistic pathogens, particularly in cystic fibrosis (CF) patients [16][17][18]. The top three Bcc species responsible for infections of American CF patients are B. cenocepacia, B. multivorans, and B. vietnamiensis [18][19][20].
Based on clinical cases in North America involving individuals infected with B. pseudomallei or its close genetic near neighbors [21], we suspected these species might be present in North American environments that are similar to those in melioidosis-endemic regions of Asia and Australia. In particular, B. pseudomallei and its near neighbor species are found in Australia and Asia in sandy, acidic, moist soils that are well-oxygenated, and protected from UV exposure [22,23]. In endemic regions where melioidosis cases are common, B. pseudomallei, B. thailandensis, and B. oklahomensis can be readily isolated from soil and water samples using selective media developed for the isolation of B. pseudomallei, such as Ashdown's agar [24]. A recent consortium outlined effective methods and media for conducting surveys of B. pseudomallei in the environment [25]. However, few surveys for B. pseudomallei and its near neighbors have been conducted outside of endemic areas [26,27] despite the ongoing discovery of new species in this group [28]. To address this knowledge gap, we surveyed for B. pseudomallei and its genetic near neighbors in soils from three southern U.S. states using these well-established sampling and culturing methods. southern regions were selected because B. pseudomallei is largely endemic to tropical regions, such as Southeast Asia and northern Australia [1]. Due to their proximity to the hot and humid climate of the Gulf of Mexico, both Florida and Louisiana provide the closest approximation in the US to the environmental conditions where B. pseudomallei is endemic. Arizona was included due to a recent case of melioidosis described from the southern region of that state [5,29]. All collection sites were chosen for their close vicinity to surface freshwater with road access, but we avoided agricultural fields and human dwellings (Table 1). No specific permissions were required for our sampling activities, because they occurred along the right-ofway of public roads and in municipal parks without access restrictions. Also, the field sampling did not involve endangered species or protected habitats. Soil collection procedures followed those employed by the Menzies School of Health and Research [22] and are expanded versions of those used by the wider B. pseudomallei research community [25]. Soil was collected at multiple sites per state (AZ = 4, FL = 7, LA = 7), which are displayed on maps in S1 Fig. At each site five sampling holes were dug along linear 40m transects (10m between holes) using hand spades, with two samples collected at different depths (10cm and 30cm) from each hole for a total of 180 soil samples. All field equipment was rinsed with water and decontaminated with 70% ethanol between each sampling hole. Soil samples were collected in 50mL sterile conical tubes and stored in the dark at ambient temperature until they were shipped (also at ambient temperatures) to Northern Arizona University.
All culturing procedures were carried out at Northern Arizona University and followed methods previously described [25,30]. It is important to note that these methods were developed specifically to isolate B. pseudomallei and not all Burkholderia species. Briefly, 20g of each Diverse Burkholderia Species from Southern US collected soil sample was suspended in 20mL of sterile distilled water and incubated at 37°C while shaking (250 rpm) for 48 hours. Samples were allowed to settle for 1 hour and then 100μL of the water suspension was plated onto Ashdown's agar plates. Also, 10mL of water suspension was inoculated into 10mL of Ashdown's broth (containing 0.05 mg/mL colistin) [24]. The Ashdown's broth was then shaken at 37°C for seven days. We plated 10μL of the top layer  from the Ashdown's broth suspension onto Ashdown's agar plates (containing 4 mg/mL gentamycin) at Day 2 and Day 7 post broth inoculation. After 48 hours of incubation at 37°C, we sub-cultured single colonies from the Ashdown's agar plates (up to 5 colonies per plate). We were specifically interested in isolating members of the B. pseudomallei group and, thus, selected colonies that displayed a morphology similar to B. pseudomallei: lavender to purple colonies, dry, slightly textured, with a raised dome or fried-egg morphology, and dimpled/ wrinkled centers [31,32]. Because we used these selection criteria and methods developed specifically for the isolation of B. pseudomallei and not all Burkholderia species, it is likely that we missed some Burkholderia species present in the samples that are not closely related to B. pseudomallei [33].

Detection of B. pseudomallei
To quickly determine the presence of B. pseudomallei we screened DNA extractions using a real-time PCR assay that targets orf2 in the type three secretion system 1 (TTS1) cluster of B. pseudomallei. This target is highly specific to B. pseudomallei and is considered the gold standard for PCR-based detection of B. pseudomallei [34]. Although the consensus guidelines suggest using the latex agglutination assay, we used molecular identification because it is more accurate. DNA was extracted from all sub-cultured colonies using a 5% Chelex1-100 heat soak method [35,36]. All DNAs were screened using published conditions on ABI 7900 machines. We used DNA from a known positive control isolate of B. pseudomallei (K96243) and water was used for no-template controls (NTCs).

Molecular identification of Burkholderia spp
To identify any potential Burkholderia spp. from the Ashdown's medium, we sequenced a 365bp section of the recombinase A gene (recA) and, when necessary, the 16S rRNA gene. The RecA protein is essential for DNA recombination and repair and its nucleotide gene sequence exhibits mutations among Burkholderia species, making it valuable as a molecular target for species-level identification [37]. Since other soil-dwelling bacteria besides Burkholderia are known to grow on Ashdown's media (including Delftia, Pandoraea, Pseudomonas, and Ralstonia) [31], we used Burkholderia-specific primers BUR 3 [38] and BUR5 [39] to amplify a 365bp region of the recA gene for the next screening of all sub-cultured isolates. The 20μL PCR contained final concentrations of the following reagents: 1 x buffer, 1.5mM MgCl 2 , 0.25mM dNTPs, 0.2μM of each primer, 1.0 U of Platinum 1 Taq (Invitrogen, Grand Island, NY), 1.2M betaine, and 1μL of DNA template (diluted genomic DNA). A modified version of a previously described "slowdown PCR" designed for GC-rich template [40,41] was used for cycle conditions (SEQSLOWD). The modifications made to the slowdown PCR were as follows: initial denature of 95°C for 5 minutes was excluded, extension time lengthened from 40 seconds to 3 minutes, starting annealing temperature reduced from 70°C to 65°C, ending annealing temperature reduced from 53°C to 52°C, starting annealing temperature reduced from 58°C to 55°C for the last 15 cycles, and final extension of 72°C for 10 minutes was added. The complete PCR cycle conditions for SEQSLOWD are listed in S1 Text. The 365bp PCR product (4μL) was visualized on a 1.5% agarose gel using a 100bp ladder (Invitrogen, Grand Island, NY) for reference to the target size and estimation of the dilution needed for cycle sequencing PCR (below). The 16S rRNA gene was sequenced from isolates that yielded negative recA results to ensure that we did not exclude any Burkholderia spp. that may possess incompatible recA priming sites. We used modifications of universal 16S primers 27F.1G (forward, 5'-GAGRGTTT GATCMTGGCTCAG-3') and 1391R (reverse, 5'-TGRACACACCGCCCGTC-3') to amplify ã 1400bp region of the 16S rRNA gene [42]. The 20μL PCR contained final concentrations of the Both recA and 16S rRNA genes were sequenced using the Sanger method. To remove excess primers and dNTPs from the post-PCR product, 4μL of Exo-SAP-IT 1 (USB Corporation, Cleveland, OH) was added to each reaction and incubated for 15 min at 37°C, followed by enzyme deactivation for 15 min at 80°C. PCR dilutions were made depending on the band intensity from the gel electrophoresis. Faint bands were diluted 1:2 in water whereas bright bands were diluted 1:5. The diluted PCR product was used as template for sequencing using the BigDye1 Terminator v3.1 Cycle Sequencing Kit (Life Technologies, Grand Island, NY). The same primers used for the initial amplification of recA were used in two cycle sequencing reactions. An additional five internal primers (335F2, 5'-CTCCTACGGGAGGCAGCAG-3'; 926F, 5'-CTCCTACGGGAGGCAGCAG-3'; 926F, 5'-TTAAAACTCAAATGAATTGA CGGGG-3'; 1053F, 5'-GTGCTGCATGGCTGTCGTCAG-3'; 515R, 5'-ATTACCGCGGCT GCTGGCAC-3'; 787R, 5'-ATTAGATACCCRNGTAGTCC-3'; 1391R, 5'-TGRACACAC CGCCCGTC-3') [42] were used to achieve full coverage of the 1400bp 16S rRNA gene in seven cycle sequencing reactions. Cycle sequencing conditions for recA and 16S rRNA genes were the same except for the primer starting concentrations (recA, 0.8μM; 16S rRNA, 3.2μM). The components for the cycle sequencing were 2μL 5x sequencing buffer, 1μL BigDye 1 v3.1, 0.32μM of a single primer and 2μL of diluted PCR product producing a 10μL reaction. The cycle conditions for cycle sequencing consisted of 1 min, 96°C; (30 sec, 96°C; 10 sec, 50°C; 4 min, 60°C) x 30 cycles; held at 16°C. An EDTA/ethanol precipitation cleanup was performed on the products before they were sequenced on a 3130xl Sequencer (Applied Biosystems, Carlsbad, CA).
Once the recA and 16S rRNA fragments were sequenced they were assembled and edited by visual inspection with Sequencher 5.1 (Gene Codes, Ann Arbor, MI). Using NCBI BLAST (http://blast.ncbi.nlm.nih.gov/Blast.cgi), all recA and 16S rRNA amplicons were identified to genus and any isolates that were not Burkholderia were excluded. The Burkholderia recA sequences were then aligned with other Burkholderia recA sequences from NCBI GenBank, including B. xenovorans as an outgroup taxon, as previously established by Cesarini

Multiple locus sequence type (MLST) analysis
Isolates identified as Burkholderia spp. were streaked from a single colony to form a lawn and then stored at -80°C in Luria Bertani (LB) broth with 20% glycerol. Culture was grown on LBA plates and incubated at 37°C for 24-48 hours. High molecular weight DNA was extracted using the Qiagen1 DNeasy Blood and Tissue Kit (catalog no. 69504; Valencia, CA) in preparation for whole genome sequencing using Illumina HiSeq, MiSeq, or GAIIx (Illumina, Inc.; San Diego, CA) sequencing technology. Using approximately 2.7μg of gDNA, libraries were prepared for whole genome sequencing as previously described [48].
Raw reads were assembled using SPAdes v3.5.0 [49] (data not shown). A multiple locus sequence type (MLST) system specific for the Bcc was used [50]. First, alleles were called by aligning assemblies against MLST alleles from the PubMLST website (see http://pubmlst.org/ bcc/) with BLASTN [51]. For each case, the top BLAST hit for each allele was identified and the sequence type was reported for exact matches. For our 35 samples from this study, concatenated sequences of the seven MLST genes (atpD, gltB, gyrB, recA, lepA, phaC, and trpB) were generated by extracting the exact match from BLASTN alignments. Each sequence type (ST) was searched for in the PubMLST Bcc database (see http://pubmlst.org/bcc/) to determine its closest match. At least one representative was downloaded as an external reference in addition to the type strains for members of the Bcc and B. gladioli as a reference outside of the Bcc (40 reference sequences). All concatenated MLST sequences (n = 75) were aligned using MUS-CLE [49] and a maximum parsimony analysis was performed using MEGA v6 with a bootstrapping method [45,47,52]. Only bootstrap values of 50% are reported on the consensus tree (Fig 2).

Burkholderia species identified
Burkholderia pseudomallei was not identified in any of our soil collections and, to date, this species has yet to be isolated from environmental samples within the United States. Despite the collection of both clinical and environmental isolates in the U.S. of B. oklahomensis [8] and B. thailandensis [7], we were unable to identify these species or any other close genetic near neighbors within the B. pseudomallei group (Fig 2). All 176 isolates sub-cultured from the Ashdown's agar plates (AZ = 45, FL = 82, LA = 49) were negative for TTS1 but we identified 36 isolates that showed amplification using the Burkholderia recA assay. Of these 36 recA positive isolates, 35 were identified as Burkholderia spp., whereas the last isolate had a closest NCBI BLAST identity with Rubrivivax gelatinosus. The 35 Burkholderia isolates were cultured from multiple locations in all states, including 2/4 (50%) sampling sites in AZ, 3/7 (43%) sampling sites in Louisiana, and 6/7 (86%) sampling sites in Florida ( Table 2) (Table 2 and Fig 2).
The state that yielded the most Burkholderia isolates was Florida with a total of 30 isolates, followed by Louisiana with three isolates, and Arizona with two isolates (S1 Fig). Since B. pseudomallei prefers moist soil [22] it was not surprising that Arizona, with the lowest average annual rainfall among these three states (Table 1), yielded the smallest number of isolates. All Burkholderia were cultured from sandy soils except for one strain that was isolated from a clay site in LA (Table 1). Both sampling depths yielded about equal numbers of Bcc isolates. One of the Florida sites in particular (site 2, Fox Lake, FL) provided a significant proportion (10/35, or 28%) of the total isolates. One explanation as to why more Burkholderia were isolated from Florida than Louisiana could be due to the pH of the soil. Burkholderia has been shown to have a higher tolerance for acidic soil than other bacteria found in soil [23,53], and Florida has the greatest extent of low pH soil among the three states we sampled [54]. Although we did not collect pH data at our specific sampling locations, it is possible that the acidic soils in Florida could contribute to a higher abundance of Burkholderia compared to the other two states with higher soil pH. Of course, there will be local variations in soil pH within each of the states, which can influence the bacterial community on a fine scale.
A better understanding of U.S. soil microbial communities and their environmental conditions may provide important information about the presence and environmental preferences of opportunistic pathogens. A number of the Bcc species we recovered have been described as opportunistic pathogens, particularly in cystic fibrosis (CF) patients and other immunocompromised individuals [9,11,[16][17][18]. In Florida, we isolated the two of the top three Bcc species responsible for the greatest number of infections in CF patients (B. cenocepacia and B. vietnamiensis) [10,[18][19][20]. Interestingly, we did not recover any B. multivorans from our soil samples despite the greater occurrence of this species in CF patients compared to all other members of the Bcc [18]. Based on the large number of CF clinical cases caused by these three species throughout the U.S., these Burkholderia spp. could be present in a wide range of soils throughout the U.S.

B. vietnamiensis clade
Our study suggests that B. vietnamiensis may be a common component of the soil bacterial community in the southern U.S. We isolated B. vietnamiensis in all three states and it was the most common species in our survey (n = 22/35). As mentioned above, B. vietnamiensis is an opportunistic pathogen particularly in CF patients. However, it is also a beneficial species that can fix nitrogen in association with rice plants [12], and at least one strain (G4) is capable of degrading a common organic pollutant, trichloroethylene [13,55]. The same sampling location that yielded B7020 (FL, site 2, sampling hole 3) also provided two other B. vietnamiensis strains with unique sequence types (STs) (B7019 FL-2 and B7021 FL-2), indicating that genetic diversity within a single soil collection hole can be quite high; a similar pattern has been found with B. pseudomallei [56]. Other members belonging to the Bcc could also be common components of the soil but we may have not sampled them due to our methods used for the preferential isolation of B. pseudomallei.

Diversity within the B. cepacia complex
The MLST maximum parsimony tree (Fig 2) provided species-level resolution and demonstrated a substantial amount of diversity among our soil samples. Of the 35 Burkholderia isolates, six were assigned to currently defined STs whereas the other 29 represented 28 previously uncharacterized STs (Table 2). We obtained species-level identification for all isolates except two (B6994 and B7043). It appears that B6994 may represent a new lineage within the PubMLST database because its closest match was an undescribed member of the Bcc (QLD039) that shared the same allele at only three of seven MLST loci. The other exception, B7043, also had a closest match with an unknown member of the Bcc (AU7004) that shared alleles at five loci (Table 2 and Fig 2).
A short fragment from one of the MLST loci (recA) was capable of providing the same species identification as MLST loci in nearly all cases (33/35;see S2 Fig). Sequencing this smaller, relatively variable fragment of recA provides a more rapid and inexpensive tool for species  identification of Burkholderia isolates compared to MLST. All SNPs within recA were synonymous mutations, as might be expected for a gene that is essential for DNA recombination and repair. Although MLST is a high resolution tool capable of recovering within-species diversity, recA by itself is able to provide adequate species-level resolution for molecular surveys of Burkholderia from environmental and clinical samples.
The isolates found in this study do not represent the overall community of Burkholderia species found in U.S. soils for two main reasons. First, we utilized a selective medium designed to isolate B. pseudomallei and it is likely that not all members of the Bcc grow on this medium. Second, the primary goal of this study was to attempt to isolate B. pseudomallei and, as a result, we preferentially selected colony morphologies similar to B. pseudomallei. During this selection process other members of the Bcc with less similar colony morphologies to B. pseudomallei could have been missed.

Conclusions
All the Burkholderia species isolated from this study (with the exception of B. metallica) have been described from multiple continents and have world-wide distributions (see http://pubmlst. org/bcc/) [51]. We expected to find more Burkholderia in Florida and Louisiana than in Arizona since both states receive more than five times the amount of annual rainfall than Arizona (Table 1). A surprising result was that Florida yielded ten times as many Burkholderia isolates as Louisiana, as well as greater species diversity, although sampling sizes were low. One possible explanation to the dissimilarity in the number of Burkholderia isolates between Florida and Louisiana may be primarily due to soil type. Most soils sampled in Louisiana were of a heavy clay composition and probably had a neutral pH, whereas the Florida soils were predominantly sandy and more likely to be acidic. Clay-based soils are more likely to result in anaerobic conditions that are not ideal for the growth of Burkholderia, whereas an increased fraction of sand may increase the available oxygen and favor the survival of Burkholderia. Some soil dwelling bacteria have difficulty inhabiting soil with a lower pH whereas Burkholderia can tolerate a wide range of soil pH [23]. This provides Burkholderia with an advantage to survive in soils where other soil bacteria cannot and may be a contributor to why the majority Burkholderia isolates in this study  (Fig 2).
were from Florida. Due to the lack of melioidosis cases in the U.S., we did not expect B. pseudomallei to be prevalent in our samples, especially considering our relatively small sample sizes. However, if B. pseudomallei is indeed present in isolated regions in the U.S., we hypothesize that Florida is the most likely of these three southern states to contain B. pseudomallei, based upon rainfall, soil type, and the results from this study.