qsmR encoding an IclR-family transcriptional factor is a core pathogenic determinant of Burkholderia glumae beyond the acyl-homoserine lactone-mediated quorum-sensing system

The whole genome sequence information of the B. glumae strains 336gr-1 (virulent, non-pigmenting), 257sh-1 (avirulent, pigmenting), and 411gr-6 (hyper-virulent, pigmenting) was announced previously [16]. Their phenotypic traits are presented in Fig 1. Each strain contains two chromosomes and two plasmids except for the 336gr-1 strain which has three plasmids. The strains also have genome sizes that range from 6,691,685 to 6,879,338 bp (Table A in S1 Text). In this study, we conducted a phylogenetic analysis to investigate the relatedness of the three B. glumae strains along with eight other strains based on the single-copy genes from each strain, which revealed that B. glumae strains 257sh-1 and 411gr-6 clustered together, while 336gr-1 was in a different clade along with other strains from different geographic origins, such as BGR1 from Korea (Fig A in S1 Text). This is consistent with our previous study, which reported the closer genetic relatedness between 257sh-1 and 411gr-6 based on rep-PCR and their pigmentation phenotypes on CPG medium [15]. The pairwise comparison among the B. glumae strains 336gr-1, 411gr-6, and 257sh-1 also revealed that the strains 257sh-1 and 411gr-1 share the highest average nucleotide identity (ANI) among other pairwise comparisons (Table B in S1 Text).

Fig 1. The phenotypes of the three Burkholderia glumae wild type strains, 257sh-1, 336gr-1, and 411gr-6.

(A) The differential colony morphology of B. glumae strains on the LB agar medium (upper panel) and the CPG agar medium (lower panel). (B) The differential virulence phenotypes of B. glumae strains on rice panicles. The B. glumae strains on LB agar and CPG agar were photographed 24 h and 48 h after incubation at 37°C, respectively, for panel A. The rice panicle pictures in panel B were taken 11 days after inoculation.

https://doi.org/10.1371/journal.ppat.1011862.g001

We further characterized the three genomes of B. glumae to identify possible genetic elements associated with their pathogenic traits, especially for the natural avirulence of 257sh-1. With the hypothesis that the different virulence phenotypes of the three strains are due to mutation or sequence variation in a gene or genes critical for pathogenic function, the comparative analysis was primarily focused on genes for known and potential virulence factors as well as their regulatory and secretory components. Total of 85 genes were selected to identify amino acid sequence variations specific to the avirulent strain 257sh-1 based on their known or potential virulence-related functions, which included toxoflavin biosynthesis and transport, flagellum-mediated motility, lipase (lipA), extracellular metalloprotease (prtA), TofIR QS (tofI/tofM/tofR), the PidS/PidR two component regulatory system (pidS/pidR), transcriptional regulators of virulence factors (qsmR, toxR, toxJ, flhDC, tepR, cidR, marR), and protein secretion systems (type II, III, and VI) (Table C in S1 Text). Pairwise comparisons of the 85 genes revealed that 29 genes had amino acid sequence variations in at least one of the three strains, among which only qsmR had a variation specific to the avirulent strain 257sh-1 (Table C in S1 Text).

In the previous studies with the virulent strains 336gr-1 and BGR1, the TofIR QS system was essential for the virulence functions of B. glumae, including toxoflavin production [10] [6] and extracellular protease activity [11]. In this study, we generated a ΔtofI-tofR (deletion of the tofI/tofM/tofR cluster) derivative of 411gr-6, 411ΔtofI-R, to examine if TofIR QS also exerts the same regulatory function for virulence in this hypervirulent strain (Fig B in S1 Text). The culture extract from both wild types, 336gr-1 and 411gr-6, induced violacein production by the biosensor Chromobacterium violaceum CV026 [19], while that from their ΔtofI-tofR derivatives, 336ΔtofI-R and 411ΔtofI-R, did not induce observable violacein, indicating their loss of function in AHL synthesis (Fig C in S1 Text).

Remarkably, the ΔtofI-tofR derivative of 411gr-6 retained the ability to produce toxoflavin, which was visualized as the yellow color of the bacterial colony on the LB agar medium and quantified after chloroform extraction (Fig 2A and 2B). Although this mutant deficient in TofIR QS produced toxoflavin at a lower level compared to its parent strain 411gr-6, its capacity to produce toxin was comparable to the less virulent strain 336gr-1 (Fig 2A and 2B). In contrast, the ΔtofI-tofR derivative of 336gr-1 almost completely lost its ability to produce toxoflavin (Fig 2A and 2B). A similar pattern was observed with the same set of B. glumae strains for the phenotypes in extracellular protease activity, which was visualized on the NA agar medium containing 1% skim milk and quantified using azocasein (Fig 2C and 2D).

Fig 2. The differential functions of TofIR QS in 336gr-1 and 411gr-6 in toxoflavin production and extracellular protease activity.

A) The toxoflavin production phenotypes of the B. glumae strains on the LB agar. Bacterial cells were inoculated via streaking method onto LB agar media. Toxoflavin production is shown as yellow pigment diffused on solid media. The photo was taken 24 h after inoculation. B) Quantitative data of toxoflavin produced by the B. glumae strains. B. glumae strains were grown overnight in LB broth, and toxoflavin was extracted with chloroform. C) The extracellular protease phenotypes of the B. glumae strains on the indicator medium (NA agar with 1% skim milk). Bacterial strains were inoculated onto Nutrient Agar (NA) plate supplemented with 1% skim milk. The clear zones surrounding bacterial colonies indicate proteolytic activity. The photo was taken 48 h after inoculation. D) The extracellular protease activities of the B. glumae strains quantified using azocasein. B. glumae strains were grown overnight in LB broth and proteolytic activity of the cell-free supernatant was quantified using azocasein as the protease substrate. Error bars of each graph represent the standard deviation of three replications, and the letters on data columns indicate statistically significant differences at P < 0.05 based on Tukey’s post-hoc test. Three independent experiments yielded similar results.

https://doi.org/10.1371/journal.ppat.1011862.g002

It was reported previously that TofIR QS regulates the expression of major virulence factors in B. glumae, and disruption of this system led to a drastic reduction of disease severity caused by B. glumae BGR1 and 336gr-1 on rice panicles [6,10]. Since 411ΔtofI-R retained a high level of toxoflavin production and extracellular protease, we hypothesized that this mutant would be virulent to rice plants. To evaluate the virulence of 411ΔtofI-R, rice panicles were inoculated with the strain along with its parent 411gr-6 as well as 336gr-1 and its QS-deficient derivative 336ΔtofI-R. In the greenhouse tests for virulence, both wild type strains 336gr-1 and 411gr-6 were able to induce a high level of disease severity in the rice panicles with disease scores above 5.5 and 7.5 in a 0–9 scale disease severity index, respectively (Fig 3A and 3B). The QS mutants, 336ΔtofI-R and 411ΔtofI-R, were substantially less virulent than their parent strains but still able to cause disease, showing ca. 47% and 31% reduction of disease severity, respectively, compared to their parent wild types (Fig 3A and 3B). Remarkably, 411ΔtofI-R exhibited a similar level of virulence to 336gr-1 (Fig 3A and 3B).

Fig 3. The differential phenotypes of 336gr-1 and 411gr-6 and their TofIR QS-deficient (ΔtofI-R) derivatives in the symptom development on rice panicles.

A) Disease progress curves caused by B. glumae strains. Disease severity was scored 5 times in a 14-days period after inoculation, using a 0–9 scale, in which 0 indicated no symptoms and 9 indicated more than 80% discolored panicles. Error bar represents the standard deviation of five replications. B) Disease symptoms on rice panicles caused by the B. glumae strains. The photo was taken 14 days after inoculation. Greenhouse experiments were repeated three times, yielding similar results.

https://doi.org/10.1371/journal.ppat.1011862.g003

To validate the observed phenotypes of TofIR QS-deficient mutants in toxoflavin production, transcription patterns of the genes involved in toxoflavin production were examined using the quantitative reverse transcription (qRT)-PCR technique, which were toxA, toxJ, toxR, and qsmR. The toxA gene is the first gene of the toxABCDE operon, which encodes enzymes for biosynthesis of toxoflavin, and the toxJ and toxR genes are regulatory genes that directly control the genes for biosynthesis and transport of the phytotoxin [6]. The qsmR gene was first identified as a regulatory gene that controls the flagellum biogenesis, as well as the toxoflavin production, in B. glumae [8].

In the qRT-PCR tests of this study, both regulatory genes for toxoflavin production, toxJ and toxR, were almost completely repressed in transcription in both 336gr-1 and 411gr-6 via disruption of TofIR QS function (Fig 4). However, transcription of toxA, the first gene of the toxABCDE operon for toxoflavin biosynthesis, was only partially (but significantly) impaired in 411gr-6 by the disruption of TofIR QS, showing its transcription level comparable to 336gr-1 (Fig 4). These results were congruent with the phenotypes observed in toxoflavin production (Fig 2). They indicate that the expression of the toxABCDE operon, which is responsible for the biosynthesis of toxoflavin, is not solely dependent on TofIR QS in 411gr-6 (Fig 4). Interestingly, qsmR was not significantly affected by TofIR QS in both 336gr-1 and 411gr-6 (Fig 4), although this gene was first reported to be dependent on TofIR QS in the Korean strain BGR1[8].

Fig 4. The relative transcription levels of toxoflavin-related genes determined by qRT-PCR in the ΔtofI-tofR derivatives of 336gr-1(336ΔtofI-R) and 411gr-6 (411ΔtofI-R).

Fold change of each gene in qRT-PCR was calculated with the 2^−ΔΔCt method, and expression was normalized using the reference genes, gyrA and 16S rRNA. Experiment was repeated twice, and error bars represent the standard deviation of experiments with three replications. The numbers on the top of each pair of columns indicate the p-values based on two-tail t-test.

https://doi.org/10.1371/journal.ppat.1011862.g004

Since the expression level of qsmR was not significantly affected by TofIR QS in both 336gr-1 and 411gr-6 (Fig 4), we generated a qsmR-deficient mutant of each strain to investigate its regulatory function in toxoflavin production and extracellular protease activity and further its functional relationship with TofIR QS in both strains. Remarkably, deletion of qsmR abolished toxoflavin production and extracellular protease activity in both 336gr-1 and 411gr-6 (Fig 5). In the same experiment, the TofIR QS-deficient derivatives (ΔtofI-R) of both strains showed the same pattern as Fig 2, which showed only partial reduction of toxoflavin production and extracellular protease activity by ΔtofI-R mutation in 411gr-6 but almost complete loss of functions in 336gr-1 by the same mutation (Fig 5). These results indicated that qsmR exerted a master regulatory role for toxoflavin production and extracellular protease activity in both strains.

Fig 5. The phenotypes of qsmR deletion mutation in toxoflavin production and extracellular protease activity.

A) The toxoflavin production in LB broth by the B. glumae strains. B) The proteolytic activity of the cell-free supernatant from each cell culture of the B. glumae strains. C) The extracellular protease phenotypes of the B. glumae strains on the NA agar supplemented with 1% skim milk. Experiments were repeated three times and error bars represent the standard deviation of three biological replications. Bars with different letters indicate statistically significant differences among the data curves at P < 0.05 based on Tukey’s post-hoc test.

https://doi.org/10.1371/journal.ppat.1011862.g005

Symptoms of wild type strains (336gr-1 and 411gr-6) initially appeared as browning around the inoculation sites. They expanded to a long lesion extending to most rice sheath (Fig 6A). 336ΔtofI-R made small legions limited to the surrounding area of the inoculation sites, while 411ΔtofI-R showed longer lesions extended to other parts of the rice sheath like its parent strain 411gr-6 (Fig 6A). In contrast, qsmR mutants of 336gr-1 and 411gr-6, 336ΔqsmR and 411ΔqsmR, respectively, showed almost no observable symptom development in rice seedlings in this experimental condition (Fig 6A). Consistent with the observed pattern of symptom development, the population level of 411gr-6 was the highest among all strains tested, and 336gr-1 and 411ΔtofI-R showed a similar population level to each other (Fig 6B). However, basal levels of bacterial population were still detected for the qsmR-deficient mutants of both 336gr-1 and 411gr-6 (Fig 6B). The wild type and mutant strains showed a similar pattern in their symptom development ability in rice panicles (Fig 6C and 6D).

Fig 6. The ΔtofI-R and ΔqsmR phenotypes of Burkholderia glumae strains in rice plants.

A) The disease symptoms on the sheaths of rice seedlings. B) The bacterial population in rice plants inoculated with the B. glumae strains at the seedling stage. The bacterial population was determined by real-time PCR 14 days after inoculation. Error bar represents the data of five replications, and the letters on the top of columns indicate statistically significant differences at P < 0.05 based on Tukey’s post-hoc test. NIC means ‘no inoculation control.’ Two independent experiments conducted for panels A and B yielded similar results. C) The disease symptoms on rice grains caused by the B. glumae strains. The photo was taken 6 days after inoculation. D) The disease severity on rice panicles caused by the B. glumae strains. Disease severity was rated 6 days after inoculation. The experiment for panels C and D was conducted once with six replications.

https://doi.org/10.1371/journal.ppat.1011862.g006

To investigate the regulatory function of qsmR on TofIR QS, we conducted qRT-PCR experiments and measured the relative transcription levels of tofI and tofR in the ΔqsmR backgrounds compared to the corresponding wild type strains. The ΔtofI-R derivatives were also included, as well as the expression of toxA was also measured to affirm the reliability of the qRT-PCR data by determining the reproducibility of those for the ΔtofI-R derivatives shown in Fig 4. As presented in Fig 7A, the transcription level of tofR, but not that of tofI, was significantly lower in 336ΔqsmR compared to its parent strain 336gr-1, consistent with our previous study [11]. However, in the other virulent strain 411gr-6, it was tofI, instead of tofR, that was suppressed by qsmR deletion (Fig 7B). Expression of toxA was completely suppressed in both 336ΔqsmR and 336ΔtofI-R, while it was partially (but significantly) suppressed in 411ΔtofI-R but completely suppressed in 411ΔqsmR (Fig 7B). These qRT-PCR results of the two virulent strains and their ΔtofI-R and ΔqsmR derivatives were congruent with their toxoflavin production phenotypes shown in Figs 2 and 5. Moreover, the production of AHL signal was not abolished but reduced in 336ΔqsmR, which was consistent with the qRT-PCR data showing partial reduction of tofR expression in the same strain (Fig D in S1 Text). Remarkably, the avirulent strain 257sh-1 also produced the AHL signal, but its production was not reduced in the ΔqsmR derivative of 257sh-1, 257ΔqsmR (Fig D in S1 Text). We also observed that addition of C8-HSL did not recover the expression of toxA and prtA (the gene conferring the extracellular protease activity) in the ΔqsmR derivatives of 336gr-1 and 411gr-6 (Fig 7C and 7D). Based on these results together, we conclude that qsmR and TofIR QS are functionally independent, except that qsmR partially influences the TofIR QS system.

Fig 7. Effects of ΔqsmR and ΔtofI-R mutations on the transcription of qsmR, tofI, tofR, and toxA in Burkholderia glumae strains 336gr-1 and 411gr-6.

A) Effects of ΔqsmR and ΔtofI-R mutations in 336gr-1. B) Effects of ΔqsmR and ΔtofI-R mutations in 411gr-6. C) Effects of ΔqsmR and ΔtofI-R mutations in 336gr-1 in the presence of 1 μM C8-HSL. D) Effects of ΔqsmR and ΔtofI-R mutations in 411gr-6 in the presence of 1 μM C8-HSL. The relative transcription levels of the tofI, tofR, toxA, qsmR, and prtA genes were determined by qRT-PCR. Each gene’s fold change was determined using the 2^−ΔΔCt method, and the reference genes gyrA and 16S rRNA were used to normalize the expression level. Bars with an asterisk on the top indicate significant differences compared to the expression levels of the corresponding parent strain, 336gr-1 or 411gr-6, at P < 0.05 based on the two-tail t-test. Experiments for this figure were repeated twice, and similar results were obtained.

https://doi.org/10.1371/journal.ppat.1011862.g007

Based on the observed pivotal function of qsmR in virulence for both 336gr-1 and 411gr-6 (Figs 5–7) and the amino acid sequence variation specific to the virulence phenotype (Table C in S1 Text), we compared the putative amino acid sequences of qsmR between the virulent (336gr-1 and 411gr-1) and the naturally avirulent (257sh-1) strains. From this sequence comparison, we identified one variation in the amino acid sequence of the qsmR gene specific to the avirulent strain 257sh-1, which was the T50K substitution predicted in the qsmR coding sequence (Fig 8A). Further analysis using the AlphFold2 software predicted the effects of this variation in the protein structure and functions, in which the single amino acid variation resulted in a remarkable change in the structure at the N-terminal region (Fig 8B and E in S1 Text). Specifically, the QsmR structure of the virulent strains has only loops and turns in its N-terminal region, which becomes an alpha helix structure (from Lys 6 to Aps 19) in that of the avirulent strain A257 (Fig 8B and E in S1 Text).

Fig 8. Comparison of predicted amino acid sequences of the QsmR protein from different strains of Burkholderia glumae.

A) An alignment of QsmR sequences and two major domains of the protein predicted using MUSCLE in Unipro UGENE software [20]. The T50K substitution in 257sh-1 is indicated by a red box. B) The tertiary structures of the QsmR proteins from the virulent (336gr-1, 411gr-6, and BGR1) and avirulent (257sh-1) strains. The protein structures were predicted using Alphafold2 [21]. Visualization and silico analysis of the protein were performed using UCSF ChimeraX [22]. The highlighted area in the protein structures denotes the T50K substitution.

https://doi.org/10.1371/journal.ppat.1011862.g008

Further, we examined if the T50K substitution in 257sh-1 contributes to the avirulence phenotype of 257sh-1. For this, a qsmR clone (pqsmR-1) derived from the virulent strain B. glumae 336gr-1 was introduced to B. glumae 257sh-1 and tested for its ability to restore the virulence of the avirulent strain. Remarkably, B. glumae 257sh-1 carrying the qsmR clone, pqsmR-1, exhibited toxoflavin production (Fig 9A) and gain of virulence in the surrogate virulence assays using onion bulb scales (Fig 9B). Another qsmR clone, pqsmR-2, which was constructed independently, also restored the virulence of 257sh-1 (Fig 9B). B. glumae 257sh-1 carrying pqsmR-1 also showed a strong virulence in rice, like the virulence strains B. glumae 336gr-1 and 411gr-6 (Fig 9C and 9D). Either the same plasmid carrying qsmR_257sh-1, pqsmR-A1, or the vector itself, pBBR1MCS-5, did not restore the virulence of B. glumae 257sh-1, ruling out the possibility that the restoration of virulence by pqsmR-1 or pqsmR-2 is due to the overexpression of qsmR or the vector (Fig 9A and 9B). Protein expression from the native and cloned qsmR gene of each strain was confirmed by the Western blot analysis using anti-QsmR polyclonal rabbit antibodies (Fig 9E).

Fig 9. Restoration of virulence in the natural avirulent strain of B. glumae, 257sh-1, by heterologous expression of qsmR from the virulent strain, 336gr-1.

A) Restoration of toxoflavin production by a qsmR clone carrying the virulent allele, pqsmR-1. A qsmR clone carrying the avirulent allele, pqsmR-A1, did not restore toxoflavin production. The photo was taken 24 h after inoculation and following incubation at 37°C. B) Restoration of virulence in the surrogate virulence assay system using onion scales by two independent qsmR clones carrying the virulent allele, pqsmR-1 and pqsmR-2. A qsmR clone carrying the avirulent allele, pqsmR-A1, did not restore the virulence phenotype. The photos were taken 48 h after inoculation and following incubation at 30°C. The experiments for A) and B) were conducted more than three times, consistently yielding similar results. C&D) Restoration of virulence in rice panicles by pqsmR-1. The photo and the disease severity data were obtained 6 days after inoculation. The experiment with rice plants for panels C and D was conducted once in the greenhouse with six replications. 257sh-1(EV) and 257sh-1(pqsmR-A1) indicate the 257sh-1 carrying the empty vector, pBBR1MCS-5, and a clone of the avirulent qsmR allele, respectively. ‘Mock’ indicates the inoculation with the buffer (10 mM MgCl₂) used for bacterial suspensions. E) A western blot image showing QsmR protein expression in the strains of B. glumae. The GST-tagged QsmR protein extracted from E. coli is included as the positive control. The Coomassie-stained polyacrylamide gel (left) indicates the amount of each protein sample transferred to the nitrocellulose membrane exhibiting the QsmR signal (right). The green arrows indicate the GST-tagged QsmR protein. The red and blue arrows point to the overexpressed QsmR from the qsmR clones and the native QsmR protein from the wild type B. glumae strains 336gr-1 and 257sh-1, respectively. The presented Western blot result represents three independent experiments that exhibited the same pattern.

https://doi.org/10.1371/journal.ppat.1011862.g009

Furthermore, we made an allelic exchange of the qsmR locus via double homologous recombination between the 336gr-1 genome and a qsmR_257sh-1 clone in the suicide vector pKKSacB, and vice versa (between the 257sh-1 genome and a qsmR_336gr-1 clone in the same vector). Eight and twelve recombinant derivatives of 257sh-1 and 336gr-1, respectively, were obtained from this allelic exchange trial, and they were examined for their toxoflavin phenotypes and qsmR sequences. As shown in Fig 10, all three recombinants of 257sh-1 that have the A to C change at the 149^th nucleotide position conferring T50 in QsmR gained the ability to produce toxoflavin, while those retained the A149 nucleotide sequence did not (Fig 10A). In the same way, the nine recombinants that have the C to A change at the 149^th position conferring the T50K substitution lost the ability to produce toxoflavin, while those retained the C149 nucleotide sequence did not (Fig 10B). The other 257sh-1 specific SNP at the 270^th position, which does not change amino acid residue, did not affect the toxoflavin phenotype of 336gr-1 (Fig 10B). Collectively, it is evident that the disabled qsmR function by the T50K substitution is the cause of the lost virulence of B. glumae 257sh-1.

Fig 10. Phenotypes of B. glumae 257sh-1 (A) and B. glumae 336gr-1 (B) in toxoflavin production dependent on allelic exchanges of the qsmR gene nucleotide sequence, C149A and A270G.

A) Derivatives of B. glumae 257sh-1 having the C149 allele from B. glumae 336gr-1 exhibit toxoflavin-positive phenotypes. B) Derivatives of B. glumae 336gr-1 having the A149 allele from B. glumae 257sh-1 exhibit toxoflavin-negative phenotypes. Experiments for this figure were conducted twice, consistently yielding similar results. Photos were taken 24 h after inoculation and following incubation at 37°C.

https://doi.org/10.1371/journal.ppat.1011862.g010

To determine the differential regulatory functions between TofIR QS and qsmR in B. glumae, transcriptomic analysis was performed by comparing the transcriptome profiles of the wild type 336gr-1 with 336ΔtofI-R and 336ΔqsmR. The principal component analysis with the RNA-seq data showed that the variations in the transcriptome profiles among the bacterial samples were nicely clustered according to their genotypes and the growth stages (the exponential phase vs. the early stationary phage), indicating that the dataset was highly reliable (Fig 11A). A total of 844 and 846 DEGs were identified from the comparisons of 336gr-1 with its ΔqsmR and ΔtofI-R derivatives, respectively, at the cutoff value of FDR 0.05 (Fig 11B and S1 Table). Among the DEGs depending on each mutant derivative, the majority were specific to ΔqsmR or ΔtofI-R, where 367 and 96 genes were up- and down-regulated only in 336ΔqsmR, and 173 and 292 genes were in 336ΔtofI-R, respectively (Fig 11B). Whereas 331 genes showed the same up or down regulation patterns between the ΔqsmR and ΔtofI-R genotypes, among which 111 and 220 genes were up- and down-regulated in both 336ΔqsmR and 336ΔtofI-R (Fig 11B). Besides these DEGs, 50 genes showed opposite regulation patterns between the ΔqsmR and ΔtofI-R genotypes, among which 48 genes were up-regulated in 336ΔqsmR but down-regulated in 336ΔtofI-R, and 2 genes showed the reverse pattern (Fig 11B).

Further, we examined the expression patterns of known and potential virulence-related genes in each genotype at the early stationary phase (OD₆₀₀ = 1.0) to compare the regulatory functions of TofIR QS and qsmR for bacterial pathogenesis. In both 336ΔqsmR and 336ΔtofI-R, genes for toxoflavin production and the PrtA metalloprotease, which is responsible for the extracellular protease activity of B. glumae, were down-regulated compared to the wild type 336gr-1, which conformed to their phenotypes observed in this study (Fig 11C). We also conducted the same analysis with the three wild type strains, 336gr-1, 411gr-6 and 257sh-1. However, we observed a much lower depth of differential transcriptome pattern between 336gr-1 and 257sh-1 compared to between 336gr-1 and 336ΔqsmR or 336gr-1 and 336ΔtofI-R, despite the drastic difference between the two wild type strains in pathogenicity (Fig 11C and S1 Table). Between the two virulent strains, 336gr-1 and 411gr-6, the depth of differential expression of virulence-related genes was minimal (Fig 11C).

Fig 11. The differential transcriptome profiles of the wild type and mutant strains in this study.

A) The PCA plot showing the variations of transcriptome profiles depending on the genotypes and the growth stages of B. glumae 336gr-1. Gene expression data was first normalized using EdgeR [23], and PCA plot was generated using the iDEP.96 platform (http://bioinformatics.sdstate.edu/idep96/). B) The Venn diagram showing the numbers of differentially regulated genes by the TofIR QS and the qsmR gene in B. glumae 336gr-1. The plot was created using the online tool, Venny 2.1.0 (https://bioinfogp.cnb.csic.es/tools/venny/), and the list of DEGs from DeSeq2 analysis was used as the input file [24]. C) Differential expression of the known and predicted virulence-related genes in the mutant and wildtype strains compared with the reference strain 336gr-1. The gene functions were annotated based on the reference genome BGR1 [25]. The horizontal bars indicate log2-transformed fold-changes of genes regulated in each strain at the early stationary phase.

https://doi.org/10.1371/journal.ppat.1011862.g011

To examine if the differential regulatory functions of TofIR QS and qsmR in other strains of B. glumae, ΔtofI-R and ΔqsmR mutations were made with additional 20 strains of B. glumae isolated from the rice fields in Louisiana and adjacent states, and the individual ΔtofI-R and ΔqsmR derivatives of each strain were tested for their phenotypes in toxoflavin production and extracellular protease activity. Among the 20 strains tested, 11 and 4 strains retained the ability to produce toxoflavin and extracellular protease activity, respectively, in the ΔtofI-R background like 411gr-6 (Table 1). All 4 strains that were positive in protease activity were also positive in toxoflavin production. The pigmentation trait on CPG agar was not correlated with the differential phenotypes in toxoflavin and extracellular protease from the ΔtofI-R mutation (Table 1). Remarkably, all the 20 strains additionally tested showed complete loss of both toxoflavin production and extracellular protease activity in the ΔqsmR background, which was consistent with the other results for 336gr-1, 411gr-6 and 257sh-1, and confirmed the essential regulatory role of qsmR for the virulence-related functions across the strains of B. glumae (Table 1).

The complete genome sequences of the Burkholderia glumae strains 336gr-1, 411gr-6 and 257sh-1 were obtained and downloaded from the NCBI (https://www.ncbi.nlm.nih.gov). The annotation of the sequences was performed using Rapid Annotation using the Subsystem Technology (RAST) server (http://rast.nmpdr.org/) [28] [29]. Briefly, the annotated set of proteins of each strain was used to predict different subsystem categories using the same program based on the protein functions. Then, the obtained subsystem categories were used for further analysis and manual curation of genes associated with various secretion systems, regulatory factors, and the known and putative effectors of the B. glumae strains. Pairwise comparison of the amino acid sequence identity of these genes was performed using Blastp program in the National Center for Biotechnology Information (NCBI) Blast (https://blast.ncbi.nlm.nih.gov/Blast.cgi) [30]. Furthermore, known and putative type III effectors were predicted using Effectidor, an automated machine-learning-based web server to predict type-III secretion system effectors [31]. Moreover, a pairwise analysis of the average nucleotide identity (ANI) between the three strains was conducted [32]. The phylogenetic analysis of the B. glumae strains 336gr-1, 411gr-6, and 257sh-1 was performed to determine their relationships with other rice-associated B. glumae strains based on the aligned proteins and coding DNA from single-copy genes. The maximum likelihood was performed using RaxML v.8, with 1000 bootstrap replicates [33].

Bacterial strains used in this study are listed in Table 2. All strains of Escherichia coli, Burkholderia glumae, and Chromabacterium violaceum were grown and maintained in Luria-Bertani (LB) broth or LB agar media at 30 °C or 37 °C. Antibiotics were supplemented as needed at the following concentrations: 100 μg/ml for ampicillin (Ap), 50 μg/ml for kanamycin (Km), 20 μg/ml for gentamycin (Gm), and 100 μg/ml for nitrofurantoin (Nit). LB agar plates containing 30% sucrose were used to select secondary homologous recombinants that lost the sucrose-sensitive gene, sacB.

Procedures for DNA cloning and amplification were conducted according to Sambrook et al. (2001) [38]. Genomic DNA was extracted using the GenElute Bacterial Genomic DNA Extraction kit (Sigma-Aldrich). PCR products were purified using the QuickClean 5M PCR Purification Kit (GenScript, Piscataway, NJ, USA). NanoDrop ND-100 Spectrophotometer (NanoDrop Technologies, Inc., Wilmington, DE, USA) was used to assess the quality and quantity of DNA samples. The DNA sequencing of PCR products and DNA clones was performed by Macrogen USA (Rockville, MD, USA).

Since QS genes (tofI, tofM, and tofR) and qsmR are conserved in both strains of B. glumae, the same constructs used to generate 336ΔtofI-R [10] and 336ΔqsmR [11] were used to generate 411ΔtofI-R and 411ΔqsmR, respectively. Plasmids containing the flanking sequences of the tofI/tofM/tofR gene cluster (pKKSacBΔtofI-R) and the qsmR gene (pKKSacBΔqsmR) were transformed into E. coli S17-1λpir through electroporation and then introduced into B. glumae via triparental mating with the help of E. coli HB101(pRK2013::Tn7). A single homologous recombinant of B. glumae was selected on an LB agar medium containing Km and Nit. Selected mutants were grown overnight at 30 C in LB broth without adding any antibiotics. LB agar plates containing 30% sucrose were used to select recombinants that lost sucrose-sensitive gene (sacB) through secondary homologous recombination. Sucrose-resistant colonies of B. glumae were spotted on LB agar and LB agar containing Km. Mutants that only grew in LB agar were selected for PCR confirmation. The deletion of the whole QS gene cluster (tofI, tofM, and tofR) was confirmed by PCR using primers TofI(H)F and TofR(H)R, correspondent to the flanking regions of tofI and tofR (Fig B in S1 Text) [10]. AHL production of the QS mutant derivatives of 336gr-1 and 411gr-6 was tested using the biosensor strain Chromabacterium violaceum CV026. C. violaceum produces a purple pigment named violacein in the presence of exogenous AHL compounds, including C6-HSL and C8-HSL [19].

The pectolytic activity of B. glumae wild type strains and mutants was first assessed on NA or LB agar plates supplemented with 1% skim milk, following Huber’s method [39] with some modifications. Briefly, bacterial suspension was obtained from an overnight culture of each B. glumae strain or mutant in LB broth at 37 °C. The suspension was washed twice and resuspended in fresh LB broth to a final OD₆₀₀ of 1.0. Five microliters of the bacterial suspension were spotted on an NA or LB agar plate amended with 1% skim milk, followed by incubation at 37 C for 48 h. Extracellular protease activity was determined based on the presence of a halo zone formed around each bacterial colony due to the degradation of skim milk.

Quantification of enzymatic activity was done using azocasein (Sigma, Saint-Louis, MO, USA) as the proteolytic substrate, following Chessa’s method [40]. Briefly,cell-free supernatants were obtained from an overnight bacterial culture grown in LB broth with a final OD₆₀₀ of 1.0 across samples. 100 μl of each bacterial suspension was added to 100 μl of 30 mg ml^-1 azocasein and 300 μl of 20 mM Tris/1 mM CaCl₂/pH 8. The reaction solutions were incubated for 1 hour at 37 C. A stop solution (500 μl of 100 mg ml^-1 trichloroacetic acid (TCA) was added to the reaction, following centrifugation at 13,000 X g for 2 min. Quantification was done by measuring the absorbance of each reaction at 366nm using a spectrophotometer (Biomate 3, Thermo Electron Corp.). A fresh medium without bacterial cells was used for the blank control.

Toxoflavin extraction was conducted following the previously established method [6] with some modifications. Briefly, bacterial strains were grown in the LB broth for 24 h at 37 C in a shaking incubator rotating at 200 rpm. One ml of bacterial supernatant was obtained after centrifugation of bacterial cultures at 16,000 g for 1 min. Extraction of toxoflavin from 1ml of bacterial suspension was done by mixing 1:1 ratio (v:v) of chloroform and bacterial supernatant, following centrifugation at 12,000 g for 5 min to separate the chloroform phase. The chloroform phase was transferred to a new microtube and placed in a fume hood overnight to evaporate. Toxoflavin was resuspended in 1 ml of 80% methanol. Absorbance was measured at 393 nm (OD₃₉₃) using the BioMate 3 spectrophotometer (Thermo Electron Corp.). Instead of a bacterial culture, a fresh medium without bacterial cells was used for the blank control.

The virulence assay using onion bulb scales was conducted following a previously established method [41]. Briefly, B. glumae strains and mutants were grown overnight in LB broth at 37°C. Five μl of the bacterial suspension at OD₆₀₀ = 0.1 in 10 mM MgCl₂ was inoculated into an onion bulb scale with a pipette. The inoculated onion scale was placed in a wet chamber at 30°C. The virulence in the onion was determined by measuring the macerated zone around each inoculation site at 48 h after inoculation.

The real-time PCR (RT-PCR) was conducted following the previous method established by Nandakumar et al. (2009) [42]. A standard curve was generated by plotting Ct values and the log values of pre-determined DNA concentrations of B. glumae 336gr-1 at 10¹ to 10⁸ CFU/ml. A linear relationship was obtained between the values with a correlation coefficient (r²) of 0.987. Population quantification was then determined for the B. glumae-infected seedlings along with healthy seedlings to exclude possible false positive data.

The structure of the QsmR protein of each strain was initially predicted based on the conserved motifs and patterns within the protein sequence using the web tool version of Prosite (a database of protein domains, families, and functional sites) (https://prosite.expasy.org). Multiple amino acid sequence alignment was performed using MUSCLE [43] in Unipro UGENE software [20]. The tertiary structures of the QsmR protein from virulent and avirulent strains of B. glumae were analyzed using Alphafold2 (https://github.com/sokrypton/ColabFold) [21]. This software ranks the single chains predictions based on the prediction confidence measures (pLDDT and PAE). This analysis used the predicted tertiary structure with the highest per-residue confidence (pLDDT) (scale of 0–100) score of 91.2 for further analysis. Visualization and further in silico analysis of the QsmR protein were carried out using the next-generation molecular visualization program, UCSF ChimeraX [22].

The raw sequencing reads were pre-processed to remove adapter sequences, short reads, and low-quality sequences using the Trimmomatic tool [44], and the quality of reads was then evaluated using the FastQC tool [45]. High-quality reads were mapped to the genome of B. glumae BGR1 (ASM2264v1, http://www.ncbi.nlm.nih.gov/assembly/; Lim et al., 2009) [25], using the Burrows-Wheeler Aligner tool, and the featureCounts tool was used to determine the levels of gene expression [46,47]. The association of samples was assessed by conducting hierarchical clustering and principal component analyses. Briefly, genes with minimum counts per million of 0.5 in at least two libraries were kept for analysis. Count data transformation was performed using EdgeR’s log2(CPM + 4), while gene median was used for missing values imputation. The differential gene expression analysis was carried out using Deseq2 with the following threshold—FDR of 0.05 and a fold change of 2, to call a significant differentially expressed gene (DEG). Fold-change values from previous analyses were used for gene set enrichment analysis to compare the regulons in between samples using the KEGG database (https://www.genome.jp/) and the clusterProfiler package in R [48].

Production of AHLs was determined following the methods of Iqbal et al. [49] with some modifications. The strain A. tumefaciens KYC55 [50] was used as a β-galactosidase-based biosensor, which responds to the presence of AHLs by expressing lacZ. The overnight grown bacterial cultures were adjusted to OD₆₀₀ = 0.1 and the culture supernatants were collected by centrifugation. Ethyl acetate was then added to the collected supernatants at 1:1 (v/v) ratio in 2mL microcentrifuge tubes to extract the AHL. The upper layer of the suspension was collected and airdried overnight in a fume hood then resuspended in a 1% volume of sterile distilled deionized water. Ten μL of the AHL extract from each strain was added to culture tubes containing 5 ml AT broth supplemented with 40 μg/ml X-gal and inoculated with the overnight grown A. tumefaciens KYC55 (OD₆₀₀ = 0.2). The samples were incubated at 30°C for 24 h. Control tubes were supplemented with 0 or 5 μM C8-HSL. The development of blue color was read at 635 nm. The absorbance of the negative control was subtracted from the absorbance of each sample.