Diversities in Virulence, Antifungal Activity, Pigmentation and DNA Fingerprint among Strains of Burkholderia glumae

Burkholderia glumae is the primary causal agent of bacterial panicle blight of rice. In this study, 11 naturally avirulent and nine virulent strains of B. glumae native to the southern United States were characterized in terms of virulence in rice and onion, toxofalvin production, antifungal activity, pigmentation and genomic structure. Virulence of B. glumae strains on rice panicles was highly correlated to virulence on onion bulb scales, suggesting that onion bulb can be a convenient alternative host system to efficiently determine the virulence of B. glumae strains. Production of toxoflavin, the phytotoxin that functions as a major virulence factor, was closely associated with the virulence phenotypes of B. glumae strains in rice. Some strains of B. glumae showed various levels of antifungal activity against Rhizoctonia solani, the causal agent of sheath blight, and pigmentation phenotypes on casamino acid-peptone-glucose (CPG) agar plates regardless of their virulence traits. Purple and yellow-green pigments were partially purified from a pigmenting strain of B. glumae, 411gr-6, and the purple pigment fraction showed a strong antifungal activity against Collectotrichum orbiculare. Genetic variations were detected among the B. glumae strains from DNA fingerprinting analyses by repetitive element sequence-based PCR (rep-PCR) for BOX-A1R-based repetitive extragenic palindromic (BOX) or enterobacterial repetitive intergenic consensus (ERIC) sequences of bacteria; and close genetic relatedness among virulent but pigment-deficient strains were revealed by clustering analyses of DNA fingerprints from BOX-and ERIC-PCR.


Introduction
Burkholderia glumae is a seed-born rice pathogen that causes bacterial panicle blight (BPB), which is an emerging major disease problem in many rice-producing areas around the world, including the southeastern United States and South and Central American countries [1]. In particular, significant yield losses from BPB were experienced in rice-producing areas of the southeastern United States, including Louisiana, Texas and Arkansas in 1996,1997,2000, and most recently, in 2010 [2]. Burkholderia gladioli also causes BPB but tends to be less virulent and less prevalent than B. glumae [3]. Prolonged high night temperatures and frequent rainfalls at the rice heading stage are thought to be important environmental predispositions for outbreaks of this disease [4,5].
BPB is a problematic disease not only because it causes severe economic damages, but also because there are few effective methods to control this disease. Oxolinic acid used as a seed treatment or foliar application is the only chemical that can control BPB, but it is not commercially available in the United States [2]. Additionally, the occurrence of oxolinic acid-resistant strains of B. glumae can limit the use of this chemical [1,6]. Low nitrogen usage, which may reduce disease severity, has not been successful, and early planting, which may avoid high temperatures during the heading stage, can be useless if high temperatures are reached early in the season [2]. Growing disease resistant varieties would be the best option, but only partially resistant varieties that lack desired commercial characteristics are available [1,7,8].
Multiple virulence factors are involved in the bacterial pathogenesis of B. glumae. Molecular genetic studies performed by several research groups identified major pathogenic determinants of B. glumae, including the phytotoxin, toxoflavin [9,10], and lipase [11]. Bacterial motility mediated by flagella is also required for the pathogenicity of B. glumae [12]. Toxoflavin and lipase production as well as the bacterial motility mediated by flagella are controlled by a quorum-sensing system composed of a LuxI-family acyl-homoserine lactone (AHL) synthase, TofI, and a LuxR-family AHL receptor, TofR [9,11,12]. Additional virulence factors known to contribute to the full virulence of B. glumae include the PehA and PehB polygalacturonases [13], the KatG catalase [14], and the Hrp type III secretion system (T3SS) [15].
More than 300 B. glumae strains were previously isolated from rice plants with symptoms of BPB growing in rice fields in Louisiana and other states in the southeastern United States, including Texas and Arkansas [3]. In this previous study, some of the isolated strains showed asymptomatic or hypovirulent phenotypes on both rice sheaths and rice panicles in greenhouse tests [3]. In this study, 11of the 19 strains that showed asymptomatic phenotypes in preliminary virulence tests were confirmed to be naturally avirulent after a series of rigorous virulence assays in the greenhouse and in the field. These eleven naturally avirulent strains were tested with nine virulent strains of B. glumae for various phenotypes including virulence on onion bulb scales, toxoflavin production, pigmentation on CPG agar media, and antifungal activity, as well as for genetic variations revealed by BOX-and ERIC-PCR analyses.

Bacterial Strains and Culture Conditions
Bacterial strains and plasmids used in this study are listed in Table 1. B. glumae and Escherichia coli strains were routinely grown and maintained in Luria Bertani (LB) agar or broth medium [16] at 30 to 37uC.

Recombinant DNA Techniques
Procedures for routine DNA cloning and amplification were conducted according to Sambrook et al. (2001).

Virulence Assay on Rice Plants
The rice variety Trenasse, which is highly susceptible to BPB, was used for testing the virulence of B. glumae strains on rice panicles. Overnight cultures of B. glumae strains on LB agar plates were resuspended in sterile tap water at a concentration of ca. 1610 8 CFU/ml (OD 600 = 0.1). Rice plants at the 20 to 30% heading stage were inoculated twice with a two-day interval between sprays (, 2 ml/plant). Disease symptoms were evaluated 14 d after the first inoculation. The experiments were conducted in the greenhouse and in the field during the 2009 growing season. For field experiments at the LSU AgCenter Rice Research Station (Crowley, Louisiana, USA), the susceptible variety Trenasse was grown in rows (12 to 15 plants per row) with ca. one-foot intervals between rows. A row of the partially resistant variety Jupiter was grown between every four rows of Trenasse. Virulence was scored using a 0-9 scale in which 0 indicated no symptoms and 9 indicated more than 80% discolored panicles. Overall disease severity of an entire row was scored with four replications for each treatment. For greenhouse experiments, rice plants were grown in plastic pots (15 cm diameter by 20 cm height) containing a soil mixture of clay, Jiffy MixH (Ferrry-Morse Seed Co, Fulton, KY, USA) and sand in a 3:1:1 ratio. Rice plants were inoculated with the same method used for field experiments. Greenhouse tests for determining the virulence phenotypes of B. glumae strains were repeated three times. In each test, disease severity of each rice plant was scored for each treatment with four replications. Yield reduction caused by B. glumae infection was determined with the yield data obtained from field experiments. Rice grains from a single row infested with each strain or mock-inoculated with water were collected, dried to 13% moisture, and weighed. Four replications were implemented in the yield data.

Onion Assay
Yellow onions used in this study were purchased from a local market. The virulence of each B. glumae strain on onion bulb scales was tested following a previously developed method with minor modifications [18]. Briefly, 5 ml of bacterial suspension containing ca. 5610 5 CFU in 10 mM MgCl 2 was applied to a ca. 2 mmwound on the inner surface of an onion bulb scale made with micropipette tip. Inoculated onion scales were incubated in a wetchamber at 30uC and virulence was determined by measuring the macerated area on each onion bulb scale after 48 h.

Determination of Toxoflavin Production and Pigmentation
To determine the production of toxoflavin by each B. glumae strain, B. glumae cells were streaked on a King's B (KB) agar plate [16] and incubated for 24 h at 37uC. Toxoflavin production was determined based on the presence of a yellow pigment diffused from the bacterial colonies into the surrounding agar medium. To determine the pigmentation phenotype of each B. glumae strain, bacterial cells were streaked on a CPG agar plate [16] and incubated for 48 h at 30uC.

Measurement of Antifungal Activity Against Rhizoctonia solani
Antifungal activities of B. glumae strains against R. solani were measured following a previously reported method [19] with some modifications. Briefly, one ml of an overnight culture of each B. glumae strain was centrifuged, washed twice with fresh LB broth, and resuspended in 100 ml of fresh LB broth. Ten-microliter aliquots of each suspension were pipetted onto three locations around the center of a potato dextrose agar (PDA) plate. Inoculated PDA plates were incubated overnight (, 16 h) at 37uC. Mycelial plugs 5 mm in diameter were cut from R. solani cultures grown on PDA at 30uC and placed in the center of each PDA plate containing three spots of B. glumae. The length of the inhibition zone between B. glumae and R. solani was measured for each bacterial spot 48 h after incubation at 25uC. Nine replications were performed for each strain of B. glumae.

BOX-and ERIC-PCR and Cluster Analyses of the DNA Fingerprints
BOX-and ERIC-PCR were conducted following a previously established method [20]. PCR products were separated on a 1.8% agarose gel run at 60 V for 18 h and were visualized with a Kodak Gel Logic 1500 imaging system (Rochester, NY, USA). DNA fingerprints of individual strains generated from BOX-and ERIC-PCR were converted into a binary matrix by scoring amenable DNA bands as present or absent. Cluster analyses of the DNA fingerprints were performed with unweighted pair-group method of averages (UPGMA) using MEGA5 [21].

Generation of the toxA Toxoflavin Deficient Mutant, LSUPB223
An internal region of toxA was amplified using the primers, 59-TTTCGGGCGTGAAATCTATC- 39 and 59-AGCGGTA-GAAGCTGAACTGG-39. The amplified PCR product was cloned into the PCR cloning vector, pSC-A-amp/kan, using a StrataClone PCR Cloning Kit (Agilent Technologies, Santa Clara, CA, USA) following the manufacturer's instructions. The insert of the resultant PCR clone, pSC::ToxA-int, was digested with the KpnI and SacII restriction enzymes and then ligated to the KpnI/SacII-cut pKNOCK Gm suicide vector [22], to generate pKNOCK Gm ::ToxA-int (Table 1). Escherichia coli S17-1 lpir [23] was used to maintain the pKNOCK vectors and pKNOCK Gm ::-ToxA-int. E. coli HB101, which carries the helper plasmid, pRK2013 [24], was used in triparental mating to introduce pKNOCK Gm ::ToxA-int into a virulent strain of B. glumae, 336gr-1. toxA mutants resulting from homologous recombination with the vector containing the internal fragment of toxA were selected on LB agar containing nitrofurantoin and gentamycin. The toxoflavin producing phenotypes of these mutants were verified on KB agar plates.

Determination of Naturally Avirulent Strains of B. glumae Isolated from Rice Fields in the Southeastern United States
In a previous study, some strains of B. glumae isolated from rice fields in Louisiana and other southeastern states failed to produce symptoms on both rice sheaths and panicles [3]. In this study, 19 out of 24 strains that showed asymptomatic phenotypes in the previous tests were confirmed by diagnostic PCR with species specific primers that anneal to the ITS region Table 1 Bacterial strains used in this study.

Test of Onion Bulb Scales as an Alternative for Efficient and High-throughput Virulence Assays for Strains of B. glumae
Remarkably, all of the virulent strains of B. glumae that were tested caused maceration of the onion bulb scale tissue, but all of the avirulent strains of B. glumae except 237gr-5 did not cause maceration of the onion bulb scale tissue (Figures 2A and 2B). Moreover, the virulence of B. glumae on onion bulb scales was highly correlated with that on rice panicles (R 2 = 0.6732) ( Figure 2C).

Production of Toxoflavin
Toxoflavin is a bright yellow pigment that has antibiotic and phytotoxic activity and that is produced by and functions as a major virulence factor of B. glumae [9,10,25]. A virulent strain of B. glumae, 336gr-1, produced a yellow pigment in culture media, including LB and KB. In solid media, this yellow pigment is  Figure 3A). Mutation of toxA, a gene required for toxoflavin biosynthesis [9,10], abolished the production of this yellow pigment ( Figure 3B). In addition, the yellow pigment extracted from the culture media with chloroform showed absorbance maxima at 258 and 393 nm (data no shown) like a recent study on toxoflavin [26]. These results together indicate that the yellow pigment is toxoflavin and that the production of toxoflavin can be determined by the presence of the yellow pigment in media with cultures of B. glumae. As shown in Figure 2A, all of the B. glumae strains that were virulent to rice produced toxoflavin, whereas all of the strains that were avirulent to rice did not, indicating that toxoflavin production is closely associated with the virulence of B.
glumae to rice. However, one toxoflavin-deficient strain, 237gr-5, produced maceration symptoms on onion bulb scales, indicating that toxoflavin is dispensable for the maceration activity of B. glumae in onion.

Production of Yellow-green and Purple Pigments in CPG Agar Medium
Nine of the 20 strains tested produced unknown pigments when grown on CPG agar plates (Figures 2A and 3C). These pigments were easily distinguished from toxoflavin, which is bright yellow and diffused into the agar medium from bacterial colonies. Six of the nine virulent strains and three of the 11 avirulent strains produced pigments (Figures 2A and 3C). Partial purification of the pigments produced by the pigmenting strain, 411gr-6, yielded three fractions with different colors; yellowgreen, purple, and brown ( Figure S1A). The brown fraction ( Figure S1A) is likely a mixture of multiple unknown substances, not a single material (data not shown). Thus, the observed pigmentation may be due to at least two different pigments, which were yellow-green and purple ( Figures 3C and S1A). The purple pigment was restricted to the bacterial colonies, whereas the yellow-green pigment was diffused into the agar medium ( Figure 3C). The yellow-green pigment purified from agar extract exhibited strong green fluorescence when illuminated with UV light with UV absorption maxima at 202, 238, 301, 448 and 464 nm. The purple pigment exhibited a strong antifungal activity ( Figure S1B). In terms of color and antifungal activity, the purple pigment resembles the virulence factor of the opportunistic human pathogen Pseudomonas aeruginosa, pyocyanin, which is also known as ''blue phenazine'' [27]. However UV absorbance spectra and spectroscopic data including 1 H-NMR indicated that the purple pigment is different from pyocyanin (data not shown). It is noteworthy that a diverse range of phenotypic variation in pigmentation was observed among the pigment-producing strains ( Figure 3C). Especially, 257sh-1 produced excessive amounts of the purple pigment compared to the other pigment-producing strains ( Figure 3C). In addition, three virulent strains, 191sh-1, 411gr-6 and 11sh2-2-a, produced larger amounts of the diffusible yellow-green pigment compared to other B. glumae strains ( Figure 3C). Unlike toxoflavin, the pigmentation did not appear to be related to bacterial pathogenesis in onion (Figure 2A).

Antifungal Activities of Naturally Avirulent Strains of B. glumae Against Rhizoctonia solani
All of the virulent strains of B. glumae showed observable antifungal activities, whereas only seven of the 11 avirulent strains showed observable antifungal activities (Figures 4A and  4B). The highly virulent and toxoflavin-and pigment-producing strain, 411gr-6, showed the highest antifungal activity among all of the strains tested ( Figure 4B). An avirulent strain, 379gr-1-b, showed antifungal activity despite its toxofalvin-and pigmentdeficient phenotype, indicating that this strain possesses at least one antifungal system that is independent of toxoflavin and pigment production in CPG agar medium ( Figure 4B).

Comparison of Genomic Structures between Virulent and Avirulent Strains of B. glumae by BOX-and ERIC-PCR
The observed phenotypic variations in virulence-related traits (virulence on rice and onion and toxoflavin production), antifungal activity and pigment production among the strains of B. glumae prompted us to investigate the genetic relatedness among strains. BOX-and ERIC-PCR revealed genomic variation among the 23 tested strains of B. glumae, which included the 20 U.S. strains as well as the type strain ATCC33617 T [28], a Korean strain, BGR1 [29], and a clinical strain, AU6208 [11] (Figures 5 and 6). Because only DNA samples were available for BGR1 and AU6208, phenotypes of these strains could not be determined in our laboratory. However, the virulent Korean strain, BGR1, did not produce the dark pigments on CPG agar medium [29], whereas AU6208 was shown to be virulent to rice [11] and to produce pigments on CPG medium (J. LiPuma, personal communication). The genomic DNA of the B. gladioli strain ATCC51989 was included as a control and used as an outgroup in phylogenetic analyses.
Various BOX-PCR fingerprints were observed among strains of B. glumae (data not shown). Twenty-six band classes ranging from 330 bp to 3,500 bp that were reproducible in two independent reactions were scored as previously described and used to generate a dendogram ( Figure 5). The dendogram revealed two major phyletic groups. The dendogram indicated that major phyletic groups were not separated on the basis of virulence or pigmentation ( Figure 5). Group I was a polyphyletic group that contained the majority of the B. glumae strains, including all the naturally avirulent strains and all the pigment-producing strains. Group II was a polyphyletic group that contained virulent, pigment-deficient strains and the type strain ATCC33617 T ( Figure 5).
Genetic differences were also observed in ERIC-PCR fingerprints (data not shown). Twenty-one band classes between 130 bp and 3500 bp that were amenable and reproducible in two independent reactions were scored as previously described and used to create a dendogram ( Figure 6). This dendogram revealed three major phyletic groups. Group I contained all of the virulent, pigment-producing strains; Group II contained all of the virulent, non-pigmenting strains; and Group III contained only avirulent strains ( Figure 6).
Although differences were observed between the BOX and ERIC dendograms, all of the strains that grouped together in Group II in the BOX analysis also grouped together in the ERIC analysis ( Figures 5 and 6).

Discussion
In this study, 11 strains of B. glumae indigenous to the southeastern U.S. were confirmed to be naturally avirulent through a series of virulence tests conducted in the field and greenhouse. These naturally avirulent strains did not produce any obvious symptoms of BPB or cause any significant yield reduction compared to the nine virulent strains tested (Figures 1A and B). At least 5% of the B. glumae strains isolated from rice fields in the southeastern U.S. showed avirulent phenotypes (data not shown). Because the avirulent strains studied were isolated from rice plants showing BPB symptoms, they are presumably derivatives of a virulent strain that lost their pathogenicityvia spontaneous mutation after host infection. It might also be possible that populations of B. glumae present in natural conditions typically contain avirulent cells. Although it is unknown how avirulent strains of B. glumae are generated in nature, this phenomenon may be explained by several possible mechanisms. First, major deletion(s) of genomic DNA could cause a loss of multiple virulence genes. It was recently reported that the cluster of hrp/hrc genes encoding the Hrp T3SS is absent in the genome of a naturally avirulent strain of Pseudomonas syringae, 508, probably due to a deletion event [30]. Second, a point mutation could occur in a core regulatory gene that globally controls multiple virulence factors or in a gene that encodes an important virulence factor. Indeed, the ATCC strain of B. glumae tested in this study, ATCC33617 T , was reported to be avirulent because of a frameshift mutation in tofR, which encodes the TofR AHL receptor that functions as a central regulatory element for the production of major virulence factors of B. glumae, including toxoflavin and lipase [11]. In addition, four different natural mutant alleles of avrBs2, the avirulence gene corresponding to Bs2 of pepper in gene-for-gene resistance, were observed in natural strains of Xanthomonas campestris that showed attenuated virulence phenotypes in pepper plants with a bs2/bs2 genetic background [31]. Finally, multiple mutations in a genome could also be Figure 5. Phylogenetic tree generated from UPGMA analysis of the BOX-PCR fingerprints of Burkholderia glumae strains. *: AU6208 was reported to be virulent to rice in a previous study [11] and tested for the pigmentation phenotype by another research group (J. J. LiPuma, personal communication). **: The phenotypes of BGR1 were indicated based on the previous studies reported by Kim et al. (2004) and Jeong et al. (2003). ***: ATCC33617 is the type strain originally isolated as the causal agent of bacterial panicle blight and its lost pathogenicity by a spontaneous mutation of tofR could be restored by the addition of a functional copy of tofR [11]. doi:10.1371/journal.pone.0045376.g005 responsible for the loss of pathogenicity in nature. To identify the cause(s) of the avirulent phenotypes of B. glumae strains from infected rice plants, the expression of known and potential virulence genes and the presence of the functional quorum-sensing system required for the pathogenicity of B. glumae in the 11 avirulent strains are currently being studied in our laboratory. Whole genome sequencing and comparison with virulent strains would also help to elucidate the genetic basis of naturallyoccurring, non-pathogenic strains of B. glumae.
An alternate assay system using onion bulb scales for determining the virulence of B. glumae was also developed in this study. This assay system was previously used by Jacobs et al. (2008) to test strains of B. cepacia and B. cenocepacia pathogenic to onion. In our study, virulent strains of B. glumae were able to cause symptoms on onion bulb scales and virulence on onion bulb scales was highly correlated with virulence on rice panicles ( Figure 2C). Strain 237gr-5 was the only strain of the 20 strains tested, that showed different virulence/avirulence features between two different host systems (avirulent to rice but virulent to onion) (Figures 1 and 2A). Virulence tests on rice are timeconsuming, labor-intensive, and require a lot of space. The high correlation observed between the virulence of B. glumae in rice Figure 6. Phylogenetic tree generated from UPGMA analysis of the ERIC-PCR fingerprints of Burkholderia glumae strains. *: AU6208 was reported to be virulent to rice in a previous study [11] and tested for the pigmentation phenotype by another research group (J. J. LiPuma, personal communication). **: The phenotypes of BGR1 were indicated based on the previous studies reported by Kim et al. (2004) and Jeong et al. (2003). ***: ATCC33617 is the type strain originally isolated as the causal agent of bacterial panicle blight and its lost pathogenicity by a spontaneous mutation of tofR could be restored by the addition of a functional copy of tofR [11]. doi:10.1371/journal.pone. 0045376.g006 and that in onion strongly suggests that onion bulb scales can be used as an excellent alternative assay for convenient and highthroughput virulence tests of B. glumae strains and would serve as a powerful tool for large-scale functional genomic studies of this bacterium. Nevertheless, caution should be made when onion bulb scales are used as a surrogate system for determining the virulence of B. glumae in rice because virulence factors responsible for maceration of onion tissue may be different from those responsible for panicle blighting of rice.
Production of toxoflavin, was closely related to the bacterial virulence phenotypes in rice. All nine virulent strains produced toxoflavin, whereas all 11 avirulent strains did not produce toxoflavin (Figures 1 and 2A). The ability of each bacterial strain to produce toxoflavin was determined by observing the production of the bright yellow pigment released into the culture medium ( Figure 3A). Directional mutation of toxA via homologous recombination, which resulted in the loss of ability to produce toxoflavin [10], caused the loss of yellow pigment production ( Figure 3B), indicating that the yellow pigment is toxoflavin. These facts strongly suggest that naturally avirulent strains of B. glumae can be readily determined by their disability to produce toxoflavin in the culture media. Even though the pathogenicity of the B. glumae strains in rice was tightly linked to their ability to produce toxoflavin in this study, toxoflavin may not fully account for the pathogenesis of B. glumae because mutation of a toxoflavin synthesis gene could not completely abolish bacterial virulence in previous studies [10](Shrestha and Ham, unpublished). Since all of the avirulent strains produced very weak observable symptoms ( Figure 1A), it is likely that these strains are defective in the production of multiple virulence factors that are collectively required for full virulence. In fact, it has been shown that B. glumae produces additional virulence factors, including lipase [11], the Hrp T3SS [15] and polygalacturonases [13], which contribute to the virulence of this pathogen to rice. Production of known and potential virulence factors of B. glumae by individual avirulent strains is currently being analyzed (Karki and Ham, unpublished).
It is intriguing that some strains of B. glumae produce unique pigments in CPG medium ( Figure 3C). Diverse variations in the pigmentation phenotype were observed within the pigmentproducing strains of B. glumae ( Figure 3C). Two different pigments (yellow-green and purple) were partially purified from a pigmenting strain, 411gr-6 ( Figure S1), and their chemical and physical properties were also partially determined (data not shown). They were easily distinguished from the bright yellow toxoflavin. To the best of our knowledge, this pigmentation phenomenon of B. glumae has not been reported elsewhere. It was reported that some strains of B. cenocepacia produced melanin-like pigments in tyrosineenriched media and that those pigments may act as scavengers of reactive oxygen species generated from oxidative burse responses of host cells [32]. Later, it was found that the melanin-like pigment produced by a clinical strain of B. cenocepacia was likely to be a pyomelanin synthesized from a homogentisate (HGA) through the action of 4-hydroxyphenylpyruvic acid dioxygenase (HppD), and that pigment production was abolished by disruption of hppD [33]. However, mutation of hppD in 411gr-6, a pigment-producing strain of B. glumae, did not change the pigmentation phenotype of this strain, suggesting that the pigments produced by B. glumae on CPG agar medium are different from the melanin-like pigments produced by B. cenocepacia (data not shown). We recently found that a two-component regulatory system, composed of the PidS sensor histidine kinase and the PidR response regulator, is essential for the production of all three pigments in B. glumae [34]. In addition, all of the pigment-deficient mutants screened from random Tn5-mutangenesis of the pigment-producing strain 411gr-6 were deficient in the production of all three pigments, implying that they may be synthesized via a common regulatory and biosynthetic pathway [34].
Antifungal activities were detected from most of the strains tested in this study ( Figure 4B). In particular, all of the virulent strains showed various antifungal activities ( Figure 4B). It is probable that the observed antifungal activities of the virulent strains may be conferred in part, if not fully, by toxoflavin, which is known to have a broad toxic effect on prokaryotes and eukaryotes [35]. However, some avirulent strains that do not produce toxoflavin also showed antifungal activities, indicating that these avirulent strains produce additional antifungal compounds. Interestingly, the purple pigment from 411gr-6 showed a strong antifungal activity against C. orbiculare ( Figure S1B) and all the pigmenting avirulent strains also showed high levels of antifungal activities against R. solani ( Figure 4B), suggesting that the production of this pigment may contribute to the antifungal activities of the pigmenting B. glumae strains. The chemical structures of the yellow-green and purple pigments are currently being characterized. Meanwhile, the antifungal activities shown by non-pigmenting avirulent strains including 379-gr-1-b ( Figure 4B) indicate the presence of additional antifungal compound(s) produced by some B. glumae strains. The avirulent strains with antifungal activities against R. solani may be useful tools for biological control of sheath blight and possibly other fungal rice diseases, including blast caused by Magnaporthe grisea. We are currently testing the antifungal activities of these strains against other fungal pathogens. Antifungal activities of Burkholderia spp. other than B. glumae and their application as biological control agents have previously been reported [36,37]. Nevertheless, naturally avirulent B. glumae strains could be a more useful tool for the biological control of phylloplane diseases of rice, including sheath blight and blast, because they could inhibit pathogens in the rice phyllosphere better than other Burkholderia spp. that typically persist in the soil and plant rhizophere.
BOX-and ERIC-PCR analyses revealed variations in genome structure among strains of B. glumae showing various phenotypic traits and originating from diverse geographic locations ( Figures 5  and 6). All of the virulent strains showing pigment-deficient phenotypes were grouped in a single major polyphyletic group in both BOX-and ERIC-PCR phylograms (Group II in both BOXand ERIC-PCR phylograms) regardless of geographic origin ( Figures 5 and 6). Even though the strain ATCC33617 T , previously known to be an avirulent strain, was grouped together with virulent strains in Group II in both rep-PCR analyses, it would not be unacceptable to consider this strain as a virulent one since it was originally isolated as the causal agent of the rice disease [28] and later lost its ability to perceive the quorum-sensing signal due to a spontaneous point mutation in tofR encoding the cognate receptor for the quorum-sensing signal of B. glumae and since the production of virulence factors was restored by the introduction of a functional tofR clone [11]. In our independent study, all of the avirulent U.S. strains with the exception of 237gr-5 that was virulent to onion (Figure 2A), also showed deficiency in quorumsensing (Karki and Ham, unpublished). However, unlike ATCC33617 T , pathogenicity could not be restored by the introduction of a DNA clone carrying both tofI and tofR genes in these strains, suggesting that mutation on the tofI/tofR locus is not the only cause of the avirulent phenotype of the naturally avirulent strains (Karki and Ham, unpublished). It is also noteworthy that all of the virulent pigment-producing strains belonged to Group I in both BOX-PCR and ERIC-PCR analyses, and that Group III from the ERIC-PCR analysis only contained avirulent strains ( Figure 6). The clinical strain AU6208, which was previously reported to be virulent to rice [11], was grouped together with the virulent strains producing pigments (Figures 5 and 6). According to the tests by Dr. LiPuma's research group in University of Michigan, AU6208 also produces pigments on CPG agar plates (J. J. LiPuma, personal communication). These cluster analyses from BOX-and ERIC-PCR data and the phenotypic characteristics associated with major polyphyletic groups suggest that multiple lineages of B. glumae may exist.
Conclusively, significant phenotypic variations were observed among strains of B. glumae, including variation in virulence, pigmentation and antifungal activities. In addition, some phyletic groups based on BOX-and ERIC-PCR fingerprints were associated with virulence and pigmentation phenotypes. Nevertheless, genetic backgrounds of the observed phenotypes, including natural avirulence and pigmentation are still unknown. Comparative genomics approaches with whole genome sequence information, which can now be readily obtained by high-throughput sequencing, would provide important clues to identify the causes of these phenotypic traits. Additional comprehensive population genetics studies should also be conducted to elucidate the genetic lineages of this pathogenic bacterium. Figure S1 Partially purified pigments of Burkholderia glumae 411gr-6 (A) and antifungal activity of the partially purified purple pigment (B). Peak 4 from high pressure liquid chromatography showed purple color and an antifungal activity against Collectotrichum orbiculare. The photo of the antifungal activity was taken 48 h after incubation at 28uC. (TIF)