Multiple Copies of flhDC in Paraburkholderia unamae Regulate Flagellar Gene Expression, Motility, and Biofilm Formation

ABSTRACT

FlhDC is a heterohexameric complex that acts as a master regulator of flagellar biosynthesis genes in numerous bacteria. Previous studies have identified a single flhDC operon encoding this complex. However, we found that two flhDC loci are present throughout Paraburkholderia, and two additional flhC copies are also present in Paraburkholderia unamae. Systematic deletion analysis in P. unamae of the different flhDC copies showed that one of the operons, flhDC1, plays the predominant role, with deletion of its genes resulting in a severe inhibition of motility and biofilm formation. Expression analysis using promoter-lacZ fusions and real-time quantitative PCR support the primary role of flhDC1 in flagellar gene regulation, with flhDC2 a secondary contributor. Phylogenetic analysis shows the presence of the flhDC1 and flhDC2 operons throughout Paraburkholderia. In contrast, Burkholderia and other bacteria only carry the copy syntenous with flhDC2. The variations in impact each copy of flhDC has on downstream processes indicate that regulation of FlhDC in P. unamae, and likely other Paraburkholderia species, is regulated at least in part by the presence of multiple copies of these genes.

IMPORTANCE Motility is important in the colonization of plant roots by beneficial and pathogenic bacteria, with flagella playing essential roles in host cell adhesion, entrance, and biofilm formation. Flagellar biosynthesis is energetically expensive. Its complex regulation by the FlhDC master regulator is well studied in peritrichous flagella expressing enterics. We report the unique presence throughout Paraburkholderia of multiple copies of flhDC. In P. unamae, the flhDC1 copy showed higher expression and a greater effect on swim motility, flagellar development, and regulation of downstream genes, than the flhDC2 copy that is syntenous to flhDC in Escherichia coli and pathogenic Burkholderia spp. The flhDC genes have evolved differently in these plant-growth-promoting bacteria, giving an additional layer of complexity in gene regulation by FlhDC.

INTRODUCTION

The bacterial flagellum is a complex rotary engine required by many bacteria for swimming motility. Motility plays a key role in the colonization of plant roots, by both beneficial and pathogenic bacteria, with the flagellum allowing for chemotaxis toward root exudates (1–3). The flagellum can also be important for adhesion to host cells, host cell entrance, and biofilm formation (4–6). Thus, the flagellum is often a critical component facilitating bacterial association with the host, in both beneficial and pathogenic associations.

In Escherichia coli, more than 50 genes are required for the biosynthesis and function of the flagellum (7–9). Thus, it is not surprising that motility is an energetically expensive process (10) and is regulated by a hierarchical level of gene expression. At the top of this hierarchy is the class I flhDC master operon, whose gene products form a heterohexameric complex (FlhD₄C₂) that binds upstream of class II promoters and recruits RNA polymerase to promote transcription (11–13). The class II genes encode the basal body and flagellar export system as well as FliA (σ²⁸). FliA is a transcriptional regulator that controls expression of the class III genes, such as fliM, which encodes the flagellar motor switch protein (14, 15). FlhDC is also implicated in the regulation of other pathways, including anaerobic metabolism and the Entner-Doudoroff pathway (8, 16).

The flhDC operon itself is transcriptionally controlled by numerous transcription factors that allow regulation by environmental cues, including quorum sensing (QseBC), temperature, osmolarity (OmpR) (17, 18), catabolite repression (CRP), and pH (19–21). Additionally, in E. coli and Erwinia amylovora, it was found that there is posttranscriptional regulation by small RNAs (sRNAs) (22–25) and RNA-binding proteins (9, 26, 27). FlhD₄C₂ binding to the DNA is also regulated by anti-FlhD₄C₂ factors through their interaction with either the FlhD or FlhC subunit (28–30).

The FlhD₄C₂ activator is generally thought to control the regulation of lateral (peritrichous) flagella, and has been characterized in enterics, such as E. coli and Salmonella enterica (31), as well as some Betaproteobacteria, such as Ralstonia (32). Only one copy of the flhDC operon is typically present in the organisms in which these genes have been studied and published. Bacteria with a polar flagellum, such as Pseudomonas and Vibrio, generally use the σ⁵⁴-dependent NtrC family of transcriptional activators (33, 34). However, Burkholderia glumae, a plant pathogen that is polarly flagellated, controls flagellar genes through an FlhDC system (35). B. glumae mutants in flhD are nonmotile and avirulent on rice (35), again demonstrating the importance of motility for the successful colonization of the plant.

Burkholderia sensu lato encompasses a wide range of pathogenic and environmental bacteria, including Paraburkholderia spp., which are plant-associative beneficial and environmental (PBE) bacteria that were recently split from the pathogenic Burkholderia group (36–38). Paraburkholderia unamae is a soil bacterium found in the rhizosphere of diverse plant species, including maize, coffee, sugarcane, and tomato (39–42). It is interesting agriculturally for its potential as a bioinoculant, as it can reside endophytically and promote plant growth, with plant-growth-promoting properties including N₂ fixation, ACC deaminase activity, siderophore production, and phosphate solubilization (40, 43). In addition, P. unamae can utilize and degrade aromatic compounds, so may be useful for bioremediation of organic pollutants (40).

We report here that multiple copies of the flhD and flhC genes are present in P. unamae as well as other beneficial plant-associative Paraburkholderia spp. We investigate the biological role of the flhD and flhC genes in P. unamae to establish a general framework to understand why these gene duplications may be useful. We show through systematic mutant analysis that the different copies of flhD and flhC impact swimming motility as well as play a role in biofilm formation. Furthermore, we show that the level of expression of the different copies correlates with how these mutations affect motility.

RESULTS

Multiple copies of flcD and flhC are present in Paraburkholderia species.

We had previously identified a P. unamae transposon mutant of flhC, MO384, that is greatly reduced, though not completely inhibited, in swimming motility, and determined the transposon to reside within the flhC gene (M. Onofre and M. R. Lum, unpublished data). The flhC gene in other organisms is required for motility and resides in an operon downstream of flhD. Intrigued by the reduction versus absence of motility in the MO384 mutant, we performed a BLASTP search of the P. unamae MTI-641^T genome using Integrated Microbial Genomes (IMG) tools and found four copies of flhC. Two of the P. unamae flhC copies reside in operons with flhD, which was supported by Operon-Mapper (44). We designated the copy with the transposon insertion as flhC1 and the operon it resides in as flhDC1. The other operon was designated flhDC2. The flhDC1 operon is in a locus with a manganese transporter and multidrug efflux pump (Fig. 1A). The flhDC2 operon resides upstream of the motA and motB genes (Fig. 1A), which is similar in location to the single flhDC operon in E. coli and B. glumae. The third copy of flhC, flhC3, resides alone. A fourth putative copy, flhC4, is in an operon with hypothetical proteins (Fig. 1A), with the predicted protein having an E value to E. coli FlhC just under e−05 (Table 1), suggesting it could be homologous. There are only two copies of flhD in the P. unamae genome.

FIG 1.

(A) Genetic maps of the P. unamae flhDC regions, showing two flhDC operons and two additional flhC genes. (B and C) Alignment of the different P. unamae (Pu) FlhC and FlhD proteins and those of B. glumae (Bg) and E. coli (Ec). Black boxes indicate identities, and gray boxes indicate similarities. Asterisks and numbers below the alignment designate residues in E. coli known to be involved in dimerization and binding, with numbers corresponding to the amino acids that interact between FlhD and FlhC. Positions in red boxes designate the cysteines in E. coli FlhC that are involved in zinc binding.

TABLE 1.

Percentages of identity and similarity of FlhC and FlhD proteins

	% of identity and similarity to protein showna
Protein	PuFlhD1	PuFlhD2	BgFlhD	EcFlhD
PuFlhD1	100 [2e−73]
PuFlhD2	58.09 (64.76) [1e−38]	100 [7e−75]
BgFlhD	56.6 (65.09) [4e−42]	70.47 (81.9) [9e−52]	100 [2e−81]
EcFlhD	44.33 (56.6) [4e−29]	41.9 (55.23) [1e−25]	46.08 (54.78) [8e−34]	100 [9e−84]
	PuFlhC1	PuFlhC2	PuFlhC3	PuFlhC4	BgFlhC	EcFlhC
PuFlhC1	100 [7e−152]
PuFlhC2	58.33 (71.11) [2e−82]	100 [8e−136]
PuFlhC3	49.47 (59.47) [2e−63]	44.44 (55.0) [1e−53]	100 [3e−139]
PuFlhC4	21.5 (29.5) [2e−05]	22.22 (33.88) [2e−05]	16.31 (26.31) [0.002]	100 [0.0]
BgFlhC	56.91 (65.42) [2e−77]	76.66 (83.88) [2e−102]	43.61 (53.19) [7e−54]	20.75 (32.44) [4e−08]	100 [3e−143]
EcFlhC	47.87 (57.81) [2e−65]	56.11 (65.55) [8e−77]	38.94 (49.47) [7e−49]	20.31 (27.6) [6e−06]	55.31 (63.82) [6e−74]	100 [6e−145]

^aThe percentages of identity of the different P. unamae (Pu) FlhC and FlhD copies are given relative to each other and relative to B. glumae (Bg) and E. coli (Ec) FlhC and FlhD, and the percentages of similarity are indicated in parentheses. E values from a BLAST comparison are shown in brackets.

Protein sequence comparison between FlhD1 and FlhD2 shows that they share 58.09% identity and 64.76% similarity to each other (Fig. 1B and Table 1). FlhC1 shows the greatest protein sequence identity to FlhC2 (58.33%), followed by FlhC3 (Fig. 1C and Table 1). flhC4 encodes a longer protein (406 amino acids [aa]), in contrast to the 180- to 200-aa proteins encoded by the other flhC copies, and it is the C-terminus half of the sequence that aligns with the other FlhC sequences. However, the identity is much less than that seen between other copies, with FlhC4 showing the greatest identity to FlhC2 (22.22% identity and 33.88% similarity) (Fig. 1C; Table 1). The single FlhDC operon of the betaproteobacterium B. glumae has greatest identity to FlhDC2 (Fig. 1B and C; Table 1). Thus, it is FlhDC2 that shows the greatest similarity to the FlhDC previously identified in other organisms.

FlhDC has been characterized in numerous organisms because of its critical role in motility and other processes. However, in all these organisms, only a single copy of these genes is present. Intrigued by the multiple copies present in P. unamae, we investigated whether multiple copies are present in other species. We carried out a BLASTP analysis of FlhC or FlhD against genomes in IMG. We found multiple copies of both flhD and flhC in all species of Paraburkholderia and Caballeronia (see Table S3 in the supplemental material) (data not shown), which are both genera that reside within Burkholderia sensu lato and were recently separated from Burkholderia (45). In contrast, few species of Burkholderia contain multiple copies of flhD or flhC (Table S3). A single copy of flhDC is typically present in genera outside of Burkholderia sensu lato, although duplications occur on occasion—for instance, in Cuprividus taiwanensis and specific strains of Erwinia amylovora (Table S3).

We used the IMG gene neighborhood tool to compare the loci for the different flhD and flhC gene copies. We found that the locus containing the flhDC2 operon upstream of the motA and motB genes is syntenous not only in E. coli and B. glumae but also in other organisms with flhDC. However, there are exceptions, such as with Paraburkholderia sacchari, in which flhC2 is deleted (see Fig. S1 and Table S3 in the supplemental material). All Paraburkholderia and Caballeronia species we analyzed additionally contained a genetically syntenous flhDC1 operon (Fig. S1; Table S3). Using IMG tools, as well as tBLASTn against the NCBI nucleotide database, we could not find any organism outside of Paraburkholderia and Caballeronia containing the flhDC1 locus. The flhC3 and flhC4 genes were in loci syntenous in other genomes, but in a small subset of Paraburkholderia (Fig. S1). We also identified flhC4 copies in some strains of Burkholderia as well as in C. taiwanensis. Additional copies of flhD or flhC, including those in Burkholderia and Cupriavidus, appear to be unique, and the neighboring regions lack synteny with each other or that of flhDC1 or flhDC2 (Fig. S1).

We then carried out phylogenetic analyses to determine the evolutionary relationship between flhD and flhC in different species. As the numbers of copies of flhD and flhC are often different, we generated a phylogram for each flhD and flhC. High bootstrap values support the distinct clades that are formed for each of the gene copies, which correspond to the distinct genetic loci of the various gene copies. The flhD copies grouped into two major clades, with all members of Paraburkholderia represented in each of the two groups (Fig. 2A). The flhD1 group contains only Paraburkholderia and Caballeronia. In contrast, the flhD2 group includes all organisms that contain flhDC. The flhC phylogram shows four distinct clades corresponding to flhC1, flhC2, flhC3, and flhC4, which again correspond to groups of genetically syntenous loci (Fig. 2B). The flhC1 and flhC2 clades were largely congruent with those found for flhD1 and flhD2. The flhC3 clade is evolutionarily closer to flhC1. The analysis also shows that the additional copies of flhD or flhC found in some species of Paraburkholderia were most similar to flhDC1. Overall, the phylograms confirm that distinct lineages of flhD and flhC exist for flhDC1, flhDC2, flhC3, and flhC4. Although some Burkholderia species have multiple copies of the genes, none of the copies shared a neighborhood with flhDC1.

FIG 2.

Phylogenetic analysis of flhD and flhC DNA sequences by maximum likelihood analysis. (A) Analysis based on the flhD gene. (B) Analysis based on the flhC gene. Bootstrap values greater than 70% are shown at the nodes. Sequences highlighted in the same color share genetic synteny. Letters following the species name indicate additional copies of the gene, with those labeled with b and c not having genetic synteny. IMG accession numbers precede the species name.

FlhD and FlhC retain conserved amino acids.

Alanine scanning mutagenesis of FlhD in E. coli has identified a number of critical residues for binding to FlhC (46), and likewise, residues of FlhC have been determined that interact with FlhD (13). In addition, in FlhC, four cysteine residues make up a zinc-binding site and may be important for FlhC binding to DNA (13). We used this information to extrapolate the potential functional importance of certain residues in FlhD and FlhC in P. unamae. Alignment of FlhD protein sequences revealed that key amino acid residues important for FlhD function in E. coli are conserved in Paraburkholderia copies, including P. unamae (Fig. 1B) (data not shown). Specifically, Cys-65 is important for the FlhD dimerization, and Asp-28, Phe-34, Arg-35, His-91, and Ile-94 are important for FlhD to form a complex with FlhC in E. coli (Fig. 1B). These residues are present in both FlhD1 and FlhD2 of all species analyzed (Fig. 1B) (data not shown). Interestingly, the alignment also showed that at certain positions, FlhD1 and FlhD2 have different conserved residues. Asn-61 (E. coli) is present in FlhD2, but it is Asp in all FlhD1 sequences we analyzed (Fig. 1B) (data not shown). Similarly, at the residue that aligned with Thr-92 in E. coli, it is either a Thr or Ser. However, in all species analyzed, it is an Ala in FlhD1. At Leu-96, it is either predominantly Met or Leu in FlhD2, but it is Leu in FlhD1 (Fig. 1B) (data not shown). Asn-61, Thr-92, and Leu-96 are also important in forming the FlhDC complex. Residues in FlhC known to interact with FlhD in E. coli are well conserved in FlhC1, FlhC2, and FlhC3 but not in FlhC4 (Fig. 1C). However, all copies of FlhC, including FlhC4, have the four cysteines involved in zinc binding. Overall, it appears that the various FlhC and FlhD copies retain many of the critical residues found in E. coli that are important for forming the FlhDC complex and binding to DNA.

FlhD and FlhC copies have different impacts on motility.

The multiple copies of flhD and flhC suggest there may be some functional redundancy in the duplicate copies. In-frame deletion mutants were generated for each copy as well as in combination. We performed swimming motility assays on soft agar with all the mutants and measured swim diameters in order to compare motilities between mutants and the wild type. Mutants lacking flhD1 and/or flhC1 showed severely impaired motility (Fig. 3A and B). The flhD1 deletion mutant showed a more severe motility defect than the flhC1 mutant. Deletion of flhD2, flhC2, or flhDC2 also reduced motility, but to a lesser extent (Fig. 3A and B). Therefore, both operons regulate motility, but the flhDC1 operon plays the primary role under the conditions tested. Deletion of both flhD genes completely abolished swimming, consistent with the known requirement of flhD in other organisms for motility. An flhC1C2 double deletion mutant still visually retained some motility compared to the completely nonmotile fliM mutant (Fig. 3A and B), suggesting a role for flhC3 and/or flhC4. However, motility did not seem to be affected in flhC3 and flhC4 mutants, although it was completely absent in the flhC1C2C3 and flhC1C2C4 triple mutants (Fig. 3A and B). Therefore, both flhC3 and flhC4 play a role, albeit minor, in motility. As expected, mutants lacking both flhDC operons or all flhC copies lacked motility, similar to the fliM mutant (Fig. 3A and B).

FIG 3.

Swimming motility phenotype of the wild type, an fliM mutant, and flhD and flhC mutants of P. unamae. (A) Average swim halo diameters on motility agar after 48 h. Error bars represent standard deviation of results from at least three replicates. Organisms are noted as fully motile (+), partially motile (*), or nonmotile (−) based on swimming motility assay and visual observation. Different letters indicate a statistically significant difference as determined by one-way ANOVA and post hoc Tukey’s test (P < 0.05). (B) Swimming halos on motility agar after 48 h.

Scanning electron microscopy (SEM) analysis was performed to determine if mutant strains with defective motility are still able to produce flagella. Wild-type P. unamae strains are diplobacilli and lophotrichous, with multiple flagella expressed at one pole (Fig. 4). We were not able to detect the flagellar structure in mutant strains with the flhD1 gene removed. These strains demonstrated severely impaired motility (ΔflhD1 and ΔflhDC1) or were nonmotile (ΔflhD1D2 and ΔflhDC1DC2C3C4) in the swimming motility assay (Fig. 3). In contrast, mutant strains with observable swim halos expressed polar flagella. Multiple flagella were detected in flhC3 and flhC4 mutants, which displayed swim halos similar to that of the wild type. The flhC1 mutants, which had shown over 50% reduction in swim diameter, had flagella that appeared shorter and reduced in number compared to those of other mutants and the wild type (Fig. 4).

FIG 4.

SEM depicting cellular morphology of P. unamae strains. The cells depicted represent typical morphology observed from biological replicates. Arrows indicate the presence of flagella in strains that are fully (+) or partially (*) motile in the swimming motility assay; − indicates nonmotile strains.

FlhD and FlhC are involved in biofilm formation.

Previous studies have shown that in some organisms, the flagella and FlhDC can be involved in biofilm formation (47–49). To determine if mutations in flhD or flhC affect the ability of P. unamae to form biofilms on abiotic surfaces, biofilm assays on polyvinyl chloride (PVC) plates were performed. FliM mutants of Listeria monocytogenes lack flagella and are nonmotile (50). When we looked at a nonmotile P. unamae fliM mutant, we found no decrease in biofilm formation compared to the wild type (Fig. 5A and B). This indicates that motility is not required for biofilm formation in P. unamae. However, any mutant with deletion in flhC1 or flhD1 showed a severe inhibition of biofilm formation. Deletion of other flhC or flhD copies did not cause a decrease (Fig. 5A and B). Therefore, the flhDC1 operon in P. unamae plays a significant role in promoting biofilm formation on abiotic surfaces, but this appears independent of flagellar biosynthesis.

FIG 5.

Effect of flhD and flhC deletions in P. unamae on biofilm formation. (A) Mean crystal violet absorbance in a 48-h biofilm. (B) Crystal violet staining of biofilm in 96-well plates. Error bars represent the standard deviation of results from three independent experiments, and different letters indicate a statistically significant difference between means as found by analysis with one-way ANOVA and post hoc Tukey’s test (P < 0.05).

flhDC1 is expressed at higher levels than other copies.

The different phenotypes of the flhC and flhD mutants, especially with regard to motility and biofilm formation, suggest that although there is some redundancy in the function of the genes, they are regulated differently. To address this question, transcriptional promoter-lacZ fusions were made to each of the four flhC genes, and β-galactosidase activity from each promoter was quantified after the organisms were grown in TY (tryptone yeast extract) medium to mid-exponential phase. Of the different copies, the flhDC1 operon is expressed the highest, followed by the flhDC2 operon. flhC3 and flhC4 showed little expression, if any (Fig. 6A). In case we had missed regulatory elements in the promoters fused to lacZ, we also carried out quantitative PCR (qPCR) for each of the loci. The qPCR results were consistent with that of the lacZ fusions, with flhDC1 showing the greatest expression (Fig. 6B). As observed with the lacZ fusions, flhDC2 expression was less than that of flhDC1, and flhC3 and flhC4 showed very little expression. The different levels of expression observed under this growth condition correlate with the varying degrees of effect on motility and biofilm formation when the respective genes were deleted.

FIG 6.

Levels of flhD and flhC gene expression in wild-type P. unamae. (A) β-Galactosidase activity (Miller units) produced by lacZ transcriptional fusions to promoters of the different copies of flhDC or flhC. Error bars represent the standard deviation of results from at least three independent replicates. (B) Levels of flhD and flhC transcripts relative to rpoB transcripts as determined by quantitative PCR. Error bars represent the standard error of results from three biological replicates, and different letters indicate a statistically significant difference found by analysis with one-way ANOVA and post hoc Tukey’s test (P < 0.05).

flhDC regulates the expression of flagellar class II and class III genes.

fliA is a class II gene that is regulated by the master regulator of flagellar biosynthesis, FlhD₄C₂. It codes for a sigma factor (σ²⁸) and regulates the expression of fliC, a class III gene in the flagellum biosynthesis hierarchy that codes for the structural component, flagellin, to form the filament of the flagellum. To determine if the flhD and flhC genes in P. unamae behave the same as in E. coli and if they have different effects on the class II and III genes fliA and fliC, respectively, reverse transcription-quantitative PCR (RT-qPCR) was performed on mid-exponential growth cultures. Compared to wild-type P. unamae, fliA and fliC levels were reduced in a ΔflhD2 mutant and barely detectable in a ΔflhD1 mutant. The same decrease was observed when both flhD copies were deleted (Fig. 7A). Removal of the flhC1 gene also greatly reduced fliA and fliC expression. The ΔflhC2 mutant showed a trend in reduced expression of these genes, although the trend was not statistically significant, whereas ΔflhC3 and ΔflhC4 mutants did not show a change (Fig. 7B). As expected, removal of the flhDC1 or both flhDC operons severely reduced flagellar gene expression and resulted in the same phenotype as ΔflhD1, ΔflhD1D2, and ΔflhC1C2C3C4 mutants (Fig. 7C). Deletion of the flhDC2 operon resulted in some reduction, with a greater impact on fliC expression. Consistent with the motility and SEM results, mutants with a motility defect and no detection of the flagellum in SEM also showed significant reduction in fliA and fliC expression. Thus, the flhDC1 operon plays a significant role and the major role in regulating the expression of class II and III genes.

FIG 7.

Expression levels of fliA and fliC in P. unamae mutant strains compared to the wild type as determined by RT-qPCR. Shown is the fold change relative to wild-type expression by (A) flhD deletion mutants, (B) flhC deletion mutants, and (C) flhDC operon mutants. Results were calculated from biological triplicates, and error bars represent standard errors. rpoB was used as the internal control. Different letters indicate a statistically significant difference using one-way ANOVA and post hoc Tukey’s test (P < 0.05). Light gray bars show fliA expression, and dark gray bars show fliC expression.

DISCUSSION

Regulation of the flhD and flhC genes and their gene products has been well studied in numerous bacteria because of the critical role the FlhD₄C₂ master regulator plays in flagellar biosynthesis and as a global regulator of many nonflagellar genes (51, 52). This regulation is known to occur at the transcriptional level as well as numerous ways posttranscriptionally, and involves only a single copy of flhDC in organisms where it has been studied. We report here the presence of an flhDC gene family in P. unamae and other Paraburkholderia species and show that these copies have various roles in regulating gene expression, motility, and biofilm formation in P. unamae. Our findings indicate that in Paraburkholderia, an additional layer of flhDC regulation occurs by the use of multiple flhDC loci.

The presence of an flhDC gene family in P. unamae is in marked contrast to the single copy of these genes typically found outside the genus, including well-studied organisms such as E. coli and S. enterica. The presence of the flhDC gene family throughout Paraburkholderia suggests that these genes have evolved with the genus and confer some evolutionary adaptation. Gene duplication resulting in the expansion of protein families is associated with increased genome size (53). The expansion of protein families can result in novel functions, novel regulatory structures, and overall system robustness (54–56). Gene duplication can have an evolutionary advantage in the short term, allowing adaptation of organisms to various environmental conditions, such as thermal stress or starvation (57, 58). Prokaryotes cope with proteins that may no longer function appropriately (e.g., do not fold properly) in a changing environment by having multiple copies of genes that can function under different environmental conditions (59). Members of Burkholderia sensu lato are known for their large genome sizes (3.75 to 11.5 Mbp) and genomic plasticity, which is associated with their metabolic diversity and ability to thrive in a wide range of environments (60, 61). Members of Burkholderia display increased numbers of genes related to virulence, including multiple copies of secretion systems, which are not found in Paraburkholderia (62). Our analysis of the flhDC operon shows that the genes have undergone frequent duplication and rearrangements, especially in Paraburkholderia, which has a minimum of two conserved flhDC loci, flhDC1 and flhDC2, throughout the genus. The presence of the flhDC1 locus in only Paraburkholderia and Caballeronia further differentiates these genera from the pathogenic Burkholderia and other members of Burkholderia sensu lato.

The intricacies of FlhDC regulation have not been studied in Paraburkholderia. However, analysis of numerous other bacterial species reveals that regulation of FlhDC is complex and occurs through a number of different mechanisms. Transcriptional regulation of flhDC is controlled by a variety of transcription factors in E. coli, conferring sensitivity of the operon to environmental factors (52). There are RNA binding proteins that bind to the 5′ untranslated region of flhDC mRNA, thus negatively regulating protein expression (23, 63). Additional posttranscriptional control by sRNAs and proteins occurs from direct interaction with FlhDC and affects downstream gene regulation (9, 22–24). Anti-FlhDC factors identified in S. enterica and E. coli inhibit FlhDC activity by direct protein-protein interactions in response to changing nutrient conditions, altering DNA binding or preventing recruitment of RNA polymerase (28–30, 64, 65). For example, the anti-FlhD₄C₂ factor FliT binds FlhC and prevents the FlhDC complex from binding to DNA (28). Furthermore, even between closely related organisms, there may be marked differences in regulation of FlhDC. For instance, flhDC expression in the enterobacteria E. coli and S. enterica shows contrasting responses to nutrient concentrations and temperature (30, 52). Complementation analyses of these organisms also showed that although FlhD is functionally similar, FlhC from E. coli is less effective at binding S. enterica FlhDC-regulated promoters (85). The intricacies of FlhDC regulation seem to reflect the critical role FlhDC has in mediating flagellum production as well as acting as a global regulator for numerous nonflagellar genes. Our finding of multiple functional copies of each flhC and flhD in P. unamae adds additional layers of regulatory complexity, as each copy may have its own transcriptional and posttranscriptional mechanisms in play.

Under the experimental conditions in this study, flhDC1 had the highest expression and strongest effects on motility and biofilm formation. Deletion of flhD1, flhC1, or flhDC1 in P. unamae dramatically reduced motility. Deletion of flhC2, flhD2, or flhDC2, the copies syntenous with E. coli flhDC, resulted in a less dramatic reduction in swim diameter. This is further supported by our expression analysis, which shows that flhDC1 is expressed at the highest levels, followed by flhDC2. Likewise, deletion of flhDC1 had the greatest impact on expression of the flagellar genes fliA and fliC. Thus, it appears that flhDC1, at least in P. unamae under the conditions we tested, has acquired the primary role in regulating flagellar production and, thus, swimming motility, although flhDC2 plays a significant role as well. Similarly, flhDC1 plays a greater role in biofilm formation in P. unamae, although we found motility itself is not a requirement in the formation of biofilms. Motility and biofilm formation are generally considered to be inversely regulated, as bacteria in the biofilm are nonmotile. In some organisms, flagella are important for the initiation of the biofilm, with flagellar mutants of some bacteria showing reduced biofilm production (66). FlhDC is known to influence biofilm formation (49). However, mutation of flhD or flhC resulted in an increase in biofilm formation in S. enterica (48), but a decrease in E. coli (47), indicating the varied role of flhDC in biofilm formation.

The role of flhC3 and flhC4 is more difficult to define. They also form distinct flhC phylogenetic clusters of orthologous loci, but are present in only a subset of Paraburkholderia, including P. unamae. Our expression assays using real-time qPCR and lacZ promoter fusions both showed little, if any, expression of either gene. In addition, deletion of either gene alone did not impact fliA or fliC expression under the conditions studied. However, it was only when either flhC3 or flhC4 was deleted in the flhC1C2 deletion background that motility was completely abolished, indicating that flhC3 and flhC4 must play some role in regulating motility. The flhC3 gene, which resides alone, is in a chromosomal region near ompR and envZ. Interestingly, in organisms such as E. coli and Yersinia enterocolitica, OmpR regulates flhDC expression (17, 18). The role of FlhC4 appears complex. FlhC4 is double the length of the typical FlhC protein, and it is the C-terminal half that aligns with other FlhC sequences. The C-terminal half shows only 22.22% identity and 33.88% similarity to FlhC2, and it is not clear how this larger protein would interact in an FlhDC complex. FlhC4 lacks the conserved residues known to interact with FlhD but retains the four cysteines that form the zinc-binding site involved in DNA binding within promoters.

Studies have shown that the FlhD₂ homodimer can bind to DNA by itself; however, it is FlhD₄ that gives the FlhD₄C₂ complex specificity and guides it to the FlhDC-regulated flagellar genes (67). In E. coli, several residues in FlhD were found important for interaction with FlhC, and site-directed mutagenesis of those positions to alanine resulted in defective motility. For example, Cys-65 is important for the FlhD dimerization and is highly conserved in other Gammaproteobacteria as well as Betaproteobacteria (46). Our analysis found that these residues were also well conserved in P. unamae. Furthermore, residues important in E. coli for FlhC binding to FlhD were well conserved in three FlhC copies in P. unamae, FlhC1, FlhC2 and FlhC3, suggesting they all interact with FlhD. Although FlhC4 did not contain these residues, it does contain the four cysteine amino acids important for zinc binding in E. coli (13). The zinc-binding site is important for FlhC binding to the DNA; thus, it is possible that FlhC4 in P. unamae may retain the ability to bind the DNA.

In E. coli, which has a single copy of flhDC, the gene products homodimerize and form a heterohexameric complex (11, 13). We found flhDC in P. unamae to play a role in motility and gene expression similar to that seen in E. coli and other organisms; therefore, it seems likely an analogous FlhDC heteromeric complex is formed. Heteromeric transcription factors, in which the subunits are made up of different genes, are not common in bacteria. However, they account for as much as 10% of gene regulation in E. coli. Some, such as the integration host factor IHF-αβ and RcsBA are global regulators (68, 69). Others, such as RelBE, can act as a RelB homodimer that derepresses transcription or a heteromeric RelBE complex that acts as a repressor, with complex formation regulated by the concentration of the different subunits to meet a particular stochiometric ratio (70). With multiple copies of the flhD and flhC genes in Paraburkholderia, a question raised is whether the FlhD and FlhC subunits from different operons might be interchangeable and act as an additional regulatory mechanism. It will be of interest to determine how the FlhDC complex is formed in Paraburkholderia. In particular, is the complex formed from the assembly of the FlhD and FlhC proteins encoded from the same operon or is the complex formed from a combination of proteins encoded by the other gene copies? Perhaps stochiometric regulation driving the assembly between different FlhD and FlhC subunits may contribute to the regulation of this complex and downstream effects. Future studies on protein interactions of the different copies of the flhD and flhC gene products will help to further elucidate their function. Future analyses can also reveal how these different copies respond to environmental signals and influence downstream gene expression. The presence of multiple genetically distinct flhDC operons throughout Paraburkholderia indicates an additional layer of complexity to what is known about the regulation of gene expression by FlhDC.

MATERIALS AND METHODS

Bacterial strains and growth conditions.

The strains used in the study are listed in Table S1 in the supplemental material. E. coli strains were maintained on Luria-Bertani agar at 37°C with 50 mg liter⁻¹ kanamycin. A rifampin-resistant strain of P. unamae MTl-641^T (42) was selected for and used for all subsequent studies. P. unamae strains were grown at 30°C on tryptone yeast extract (TY) (71) or yeast mannitol agar (YMA) (72) with 20 mg liter⁻¹ rifampin and 50 mg liter⁻¹ kanamycin as needed.

Generation of mutant and transcriptional lacZ-fusion strains.

In-frame deletion mutants in each copy of flhD and flhC were made by allelic exchange using the vector pK18mobsacB (73). In brief, regions bordering the gene to be deleted were amplified and connected by fusion PCR, then cloned into the pK18mobsacB plasmid to generate the vector for allelic exchange. Constructs were sequenced to verify absence of mutation, then transformed into E. coli S17 using the TransformAid bacterial transformation kit (Thermo Scientific, Inc.) and selected on LB with kanamycin (50 mg liter⁻¹). The vector was introduced into P. unamae by a biparental mating strategy, selecting for kanamycin resistance for plasmid integration and then using sucrose sensitivity to select for subsequent excision of the plasmid. PCR on individual colonies was performed to identify strains where the gene of interest had been excised.

Multiple gene deletions were generated by using sequential rounds of this strategy. P. unamae mutants were conjugated with the S17 strain carrying the construct with the additional gene to be removed. The mutant strains generated and vectors used are listed in Table S1.

The lacZ-promoter fusions were generated by PCR amplification of the region encompassing the promoter for each flhDC or flhC copy using primers designed with 5′ end restriction sites. PCR products and the pVIK112 vector (74) were digested with the appropriate restriction enzymes and ligated together. Plasmids were transformed into E. coli DH5α λpir and confirmed by sequencing. The pVIK112-promoter-lacZ fusion constructs were introduced into P. unamae MTl-641^T by triparental mating with E. coli (pRK2013) (75). Selection for integration of the constructs by Campbell insertion was done using rifampin and kanamycin. All primers used are listed in Table S2 in the supplemental material.

Swimming motility assay.

We found that carrying out soft agar motility assays using bacteria from fresh colonies was more consistent than with an inoculum from liquid cultures. Therefore, cultures of bacteria were grown on TY agar at 30°C for 48 h and immediately used in motility assays. An inoculating needle was used to obtain bacteria from the center of the colony and stabbed into TY plates containing 0.25% agar (Sigma). Motility plates were incubated at 30°C, and halo diameters were measured after 48 h. Three independent trials were performed, with each trial analyzing at least three individual colonies of each strain.

Biofilm assay.

P. unamae strains were grown for 24 h in liquid LB medium without NaCl, as this medium was found to promote the best biofilm formation by P. unamae on 96-well polyvinyl chloride (PVC) plates (A. Escamilla and M. R. Lum, unpublished data). Cultures were then diluted to an optical density at 600 nm (OD₆₀₀) of 0.05, and 150 μl was placed into each well of a column in a 96-well PVC microplate, covered with breathable film (AeraSeal; Genesee Scientific), and incubated at 30°C for 48 h in a container with a wet paper towel to maintain humidity. Equivalent growth between wells was checked by measurement at OD₅₉₅ using a microplate reader. Medium was aspirated, and plates were gently rinsed with water, stained with 0.1% crystal violet for 20 min, rinsed three times with water, and dried, and the bound crystal violet was resuspended in 175 μl of 95% ethanol. Crystal violet was quantified using a microplate reader at OD₅₇₀. Average absorbance was obtained for at least six wells for each trial, and three independent trials were carried out. Wells were imaged prior to resuspension in ethanol for qualitative demonstration of biofilm formation.

β-Galactosidase assay.

Preliminary tests indicated a higher level of expression of the flhDC genes during the mid-exponential phase (S. N.-M. Thai, unpublished data). P. unamae strains carrying the promoter-lacZ fusions were grown overnight at 30°C with shaking in liquid TY, diluted to an OD₆₀₀ of 0.02, and then grown to mid-exponential phase (OD₆₀₀ of 0.8 to 0.9). β-Galactosidase assays were performed, and Miller units were calculated as described by Miller et al. (76).

Scanning electron microscopy.

P. unamae strains were streaked onto TY agar with rifampin and incubated for 48 to 72 h at 30°C, and motility assays were then performed with the fresh cultures. A sterile spatula was used to excise a small piece of agar containing bacterial growth at the edge of the swim diameter from biological replicate samples and submerged in ultrafiltered water for 5 min to allow the cells to swim out of the agar. Cells were then fixed in 4% paraformaldehyde (PFA) and transferred to a cleaned silicon wafer (Ted Pella, Inc.) coated with marine mussel glue (Cell-Tak; BD Bioscience) to attach the cells. Preparation of wafers and imaging were performed as outlined by Billi et al. (77). The wafers were then washed with ultrafiltered water, mounted onto aluminum stubs, and allowed to air dry before imaging. The wafers were scanned for the presence of cells, and their morphology was then imaged with a field emission scanning electron microscope (Supra VP-40; Zeiss) at a voltage of 1 kV with a 3-mm working distance. Multiple images were taken from each sample that represents the typical cell morphology of the strains observed after scanning the majority of the wafer surface.

RNA extraction and RT-qPCR.

To extract total RNA from respective test organisms for RT-qPCR, overnight cell cultures were diluted to OD₆₀₀ of 0.01 and then grown for 10 h in TY with rifampin. Samples were prepared in biological triplicates. RNAprotect cell reagent (Qiagen, Inc.) was added to the cell cultures per the manufacturer’s recommendation to provide immediate stabilization of the RNA prior to cell collection. Approximately 10⁹ cells were pelleted and quickly frozen and stored at −80°C until RNA extraction was performed using the RNeasy minikit (Qiagen, Inc.) per the manufacturer’s suggestion. Purified RNA samples (6.5 μg each) were treated with the Turbo DNA-free kit (Invitrogen, Inc.), and 240 ng of the DNase-treated total RNA was synthesized to cDNA with the iScript cDNA synthesis kit (Bio-Rad, Inc). For the qPCR, this cDNA reaction mixture was used at 1/20 per reaction with iTaq Universal SYBR green Supermix (Bio-Rad, Inc.) on a 7500 fast real-time PCR system (Applied Biosystems, Inc.) for 40 cycles with an annealing temperature of 60°C. Each qPCR mixture contained 500 nM forward and reverse primers. The primers used in the qPCR are listed in Table S2 in the supplemental material. The melt curves showed a single peak for the first derivative of fluorescence versus temperature, indicating a single product. The primer efficiencies for targeted genes and housekeeping genes were similar. Analysis was done using the relative standard curve method (ABI User Bulletin 2). Serial dilutions of P. unamae genomic DNA were used to generate a standard curve for each primer pair to calculate the relative amount of transcript in each sample. Levels of gene expression were normalized to rpoB.

Sequence analysis and phylogenetic tree generation.

The E. coli FlhC1 and FlhD1 sequences were used to carry out a BLASTP search (78) in IMG (79) (to identify FlhC and FlhD sequences in P. unamae, using default parameters with the BLOSUM62 matrix and a cutoff of 1e−05. Each copy of P. unamae FlhC and FlhD was then used for BLASTP searches of the IMG bacterial genomes and NCBI databases. The corresponding DNA and protein sequences were downloaded for selected type strains. MUSCLE was used to construct multiple sequence alignments (80) and MEGAX (81) to generate phylogenetic trees using the maximum likelihood (ML) algorithm. Bootstrap analysis with 1,000 resamplings was used to provide statistical support for the trees (82). The best model to use for each tree was determined by MEGAX, with the flhC phylogeny done using the general time-reversible (GTR) model (83) with a gamma distribution (G) and the flhD phylogeny using the Tamura 3-parameter (T92) model (84) with gamma-distributed selection and invariant sites (G+I). A MUSCLE alignment of the protein sequences was used as input into the SIAS server to obtain the sequence identities and similarities (http://imed.med.ucm.es/Tools/sias.html). FlhC and FlhD alignments were shaded using Boxshade 3.21 (https://embnet.vital-it.ch). To determine whether the genes were in operons, DNA regions encompassing the P. unamae flhDC1 and flhDC2 regions were entered into the operon predicting tool Operon-Mapper (44).

Statistical analysis.

Statistical analyses were carried out using Excel or SPSS and one-way analysis of variance (ANOVA) with post hoc Tukey’s test.