Cyclic Di-GMP-Mediated Repression of Swarming Motility by Pseudomonas aeruginosa PA14 Requires the MotAB Stator

Abstract

The second messenger cyclic diguanylate (c-di-GMP) plays a critical role in the regulation of motility. In Pseudomonas aeruginosa PA14, c-di-GMP inversely controls biofilm formation and surface swarming motility, with high levels of this dinucleotide signal stimulating biofilm formation and repressing swarming. P. aeruginosa encodes two stator complexes, MotAB and MotCD, that participate in the function of its single polar flagellum. Here we show that the repression of swarming motility requires a functional MotAB stator complex. Mutating the motAB genes restores swarming motility to a strain with artificially elevated levels of c-di-GMP as well as stimulates swarming in the wild-type strain, while overexpression of MotA from a plasmid represses swarming motility. Using point mutations in MotA and the FliG rotor protein of the motor supports the conclusion that MotA-FliG interactions are critical for c-di-GMP-mediated swarming inhibition. Finally, we show that high c-di-GMP levels affect the localization of a green fluorescent protein (GFP)-MotD fusion, indicating a mechanism whereby this second messenger has an impact on MotCD function. We propose that when c-di-GMP level is high, the MotAB stator can displace MotCD from the motor, thereby affecting motor function. Our data suggest a newly identified means of c-di-GMP-mediated control of surface motility, perhaps conserved among Pseudomonas, Xanthomonas, and other organisms that encode two stator systems.

INTRODUCTION

Since its discovery in 1987 as an allosteric activator of bacterial cellulose synthesis (1), cyclic diguanylate (c-di-GMP) has been shown to be a remarkably important signaling molecule across diverse bacterial species, controlling a multitude of behaviors and processes, including biofilm formation, motility, virulence, cell cycle progression, and differentiation (2–4). An important feature of c-di-GMP regulation is the ability of this signal to control critical lifestyle transitions, such as motile-sessile transitions (e.g., planktonic to biofilm), which are undertaken by many bacterial species (3, 5, 6). Generally speaking, elevated levels of c-di-GMP promote sessile lifestyles such as biofilm formation; in contrast, low levels of c-di-GMP are associated with motility (3, 6). Intracellular levels of this dinucleotide are controlled by opposing activities of enzymes that synthesize c-di-GMP (GGDEF domain-containing diguanylate cyclases [DGCs]) and those that cleave this signaling molecule (EAL- or HD-GYP domain-containing phosphodiesterases [PDEs]) (3, 4, 7–11).

More recently, studies focused on how cells respond to changing c-di-GMP levels, indicated that this signaling network relies upon proteins or RNA molecules, known as c-di-GMP effectors (or receptors), which bind c-di-GMP and whose output functions are altered due to c-di-GMP-mediated structural changes (3, 4, 12). A number of distinct effectors have been identified and classified based on their c-di-GMP-binding motif. The PilZ class of c-di-GMP effector proteins is one of the best-studied classes to date, with homologs across multiple bacterial species (2, 13, 14). PilZ domain proteins, such as YcgR, play a critical role in the c-di-GMP-dependent control of swimming motility in Escherichia coli and Salmonella enterica (14–18), and the recently identified FlgZ participates in the regulation of swimming motility by Pseudomonas fluorescens F113 and Pseudomonas putida KT2440 (19). While the precise mechanism of PilZ-dependent control of motility is still controversial, the models presented to date suggest that the PilZ–c-di-GMP complex impacts flagellar motor function (2, 15, 16, 18, 20). Motility can also be impacted by production of exopolysaccharides or through transcriptional control of the flagellar biosynthesis machinery (21–24).

Here we have uncovered a new c-di-GMP-dependent mechanism of motility control in P. aeruginosa, a microbe with two stator sets and a single rotor available to mediate function of its single polar flagellum. This dinucleotide controls an apparent “stator swapping” mechanism. MotCD is the stator responsible for driving swarming motility in P. aeruginosa, while the MotAB stator is unable to support swarming. Under conditions of high c-di-GMP, the MotAB stators are involved in the repression of swarming. Deleting motAB or introducing point mutations into motA that are predicted to disrupt key interactions with FliG, the rotor component of the motor, relieves this c-di-GMP-mediated repression of swarming. Furthermore, under conditions of high c-di-GMP, we observed altered localization of a GFP-MotD fusion. Taking these data together, we propose a model wherein, under conditions of high c-di-GMP levels, MotAB participates in repression of swarming motility, perhaps by replacing or displacing the MotCD stators in the flagellar motor.

MATERIALS AND METHODS

Strains and media.

Strains used in this study are listed in Table S1 in the supplemental material. P. aeruginosa PA14, E. coli DH5α, S17-1 λpir, BTH101, JM109, and XL-Blue were routinely cultured in lysogeny broth (LB) medium, solidified with 1.5% agar when necessary. Gentamicin (Gm) was used at 25 μg/ml for P. aeruginosa and at 10 μg/ml for E. coli. Ampicillin (Ap) was used at 150 μg/ml, nalidixic acid (NA) at 20 μg/ml, streptomycin at 100 μg/ml, kanamycin at 50 μg/ml, and carbenicillin at 100 μg/ml for E. coli. For phenotypic assays with P. aeruginosa, either M63 or M8 minimal salts medium (as indicated) were supplemented with MgSO₄ (1 mM), glucose (0.2%), and Casamino Acids (CAA; 0.5%). For expression plasmids harboring the P_BAD promoter, arabinose was added to cultures at a 0.2% final concentration, unless noted otherwise. For plating of bacterial two-hybrid assays, X-Gal (5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside; 40 μg/ml) and IPTG (0.5 mM) were added to selective plates to visualize β-galactosidase reporter expression.

Saccharomyces cerevisiae strain InvSc1 (Invitrogen), used for plasmid construction via in vivo homologous recombination, was grown with yeast extract-peptone-dextrose (1% Bacto yeast extract, 2% Bacto peptone, and 2% dextrose), as reported previously (25). Selections with InvSc1 were performed using synthetic defined agar-uracil (4813-065; Qbiogene).

Construction of mutant strains and plasmids.

Table S2 in the supplemental material lists all plasmids used in this study. Primers used in plasmid construction and in mutant construction and confirmation are listed in Table S3 in the supplemental material. In-frame gene deletions, chromosomal “knock-in” procedures for epitope tagging, mutagenesis, and complementation were performed via allelic exchange, as reported previously (26, 27). Plasmids for these purposes were constructed via cloning by homologous recombination of relevant PCR products into the pMQ30 vector in yeast, as reported (25).

Motility assays.

Swim (0.3% agar) and swarm (0.5% agar) motility plates contained M8 medium supplemented with glucose, MgSO₄, CAA, and arabinose, where indicated. Swim assays were performed as previously described (28). Swarm assays were performed as previously described (29). Quantification of swim and swarm zones was performed using ImageJ software.

Biofilm formation assay.

Biofilm formation in 96-well microtiter plates was assayed and quantified as previously described (30, 31). All biofilm assays were performed using M63 minimal medium supplemented with glucose, MgSO₄, CAA, and arabinose where indicated.

CR binding assays.

Congo red (CR) binding assays were performed as reported previously (32–34).

Protein expression and cellular localization experiments.

Strains were cultured on semisolid agar (0.5%) swarm plates and cells were allowed to swarm for 16 h at 37°C, followed by harvesting by gentle scraping with an ethanol-washed plastic coverslip into microcentrifuge tubes and samples were centrifuged for 1 min at room temperature (RT). Supernatants were removed, and cell pellets were frozen at −80°C until further processing was performed. For generation of whole-cell (WC) lysates, cell pellets were then resuspended in cell lysis buffer (200 mM Tris-HCl [pH 7.5], 1 mM EDTA, 2 mM MgCl₂, complete protease inhibitors [Roche Diagnostics Corp., Indianapolis, IN]), and Benzonase nuclease (Novagen, San Diego, CA) was added to a final concentration of ∼50 units/ml. Bacterial cells were lysed in a French pressure cell, and samples were centrifuged at 9,300 × g for 10 min at 4°C to remove unbroken cells. Supernatants were collected as whole-cell lysates, and, where indicated, samples were further fractionated to yield cytoplasmic, inner membrane, and outer membrane fractions, as described previously (34–36).

For Western blotting, WC lysates and fractions were prepared as described previously (37) with the exception that samples were resolved by SDS-PAGE using 12% polyacrylamide gels. Proteins transferred to a nitrocellulose membrane were probed with either an anti-penta-His antibody (Qiagen, Valencia, CA) or an antihemagglutinin (anti-HA) antibody (Covance, Princeton, NJ). Detection of proteins and quantification of protein levels were performed as previously described (37).

Bacterial two-hybrid analysis.

The bacterial adenylate cyclase two-hybrid (BACTH) system developed by Euromedex (Souffelweyersheim, France), based on published work (38), was used in this study. Experiments were performed according to the manufacturer's instructions.

Fluorescence microscopy.

Cells were grown on 0.5% agar swarm plates at 37°C for 14 to 18 h, as described previously (34). Cells were picked from the edges of the swarm branches just before imaging and embedded in 1.2% agarose on a microscope slide. A home-built inverted microscope with a 473-nm laser (50 mW, 473-50-COL-002; Laser 2000) was used, as described previously (39). Laser epifluorescence illumination was used for all fluorescence imaging of the motor spots with laser intensity of ∼0.1 μW · μm⁻². Fluorescence emissions of the GFP-labeled motor spots were imaged in frame transfer mode at 50 nm/pixel and 25 Hz using a 128- by 128-pixel, cooled, back-thinned electron-multiplying charge-coupled device camera (iXon DV860-BI; Andor Technology). For each field of view, 300 frames were recorded, and the number of cells with a polar spot was determined manually using ImageJ.

RESULTS

MotAB flagellar stator proteins participate in c-di-GMP-mediated swarming inhibition.

Previously, we reported the use of a genetic screen to identify factors required to repress swarming motility via the intracellular second messenger cyclic diguanylate (c-di-GMP) (34). For this screen, we exploited the swarming defect of a mutant strain bearing a deletion of the bifA gene (PA4367), which encodes a c-di-GMP-degrading PDE. We have shown that loss of the BifA PDE leads to an accumulation of intracellular c-di-GMP levels relative to the wild-type strain, resulting in failure of the ΔbifA mutant to swarm (34, 35) (Fig. 1A, top). We performed mutagenesis of the ΔbifA mutant using the mariner transposon and screened for restoration of swarming motility.

FIG 1.

Phenotypic analyses of flagellar stator mutant strains. (A) Assessment of motA and motAB mutation on suppression of ΔbifA mutant phenotypes. The strain genotype for each image in this panel is located along the x axis of the graph. The top row shows representative swarm images for the wild type and the ΔbifA mutant, followed by a representative ΔbifA motA::Mar suppressor strain (66B5), the ΔbifA ΔmotA double mutant, the ΔbifA ΔmotA::motA complemented strain, the ΔbifA ΔmotAB triple mutant, the single ΔmotA deletion mutant, and its complemented counterpart strain ΔmotA::motA. Swarm plates were incubated at 37°C for 16 h. The second row shows CR binding for each strain. CR assay plates were incubated at 37°C for 24 h, followed by 48 h at room temperature. The bottom row shows images of wells from a 96-well biofilm assay plate. Strains were grown in M63 medium supplemented with glucose, MgSO₄, and CAA for 24 h prior to crystal violet (CV) staining. The graph depicts quantification of CV-stained biofilms. CV was solubilized in 30% glacial acetic acid, and the absorbance was read at 550 nm. Error bars indicate standard deviations of the average of three experiments with four replicates per experiment. Data were analyzed by one-way analysis of variance (ANOVA) followed by Tukey's posttest comparison. ns, not significantly different; *, P < 0.05; ***, P < 0.001 (all relative to the wild type). (B and C) Representative swarm (B) and 0.3% soft-agar (C) motility assays of the strains indicated. The ΔflgK mutant does not make a flagellum and serves as a negative control in both assays. Swarm and soft-agar motility plates were incubated at 37°C for 16 h. (D) Impact of motA overexpression on swarming motility. The image is of a representative plate showing repression of swarming motility by overexpression of the multicopy pMotA-His plasmid (right) relative to the vector control (left). Arabinose was added to a final concentration of 0.2%, and plates were incubated at 37°C for 16 h.

In our previous study, we screened approximately 5,500 transposon mutants for suppressor strains that regained the ability to swarm on semisolid agar (34). Given that our genetic screen had not achieved a full coverage of the Pseudomonas genome (∼6,000 genes), we continued with mariner mutagenesis of the ΔbifA mutant and screened an additional ∼10,500 transposon insertion mutants. From this screen, we isolated 23 mutant strains that regained the ability to swarm, and we mapped the transposon insertions in these mutants to 12 different genes (see the supplemental material for a complete list of suppressor strains, gene designations, and phenotypic summaries). This screen identified mutations in genes or operons that we, or others, have characterized previously (sadB, pilY1, pel, and gacS) (30, 32, 34, 35, 40–42), mutations that likely exerted their effects through polarity on downstream functions (pvrS, fimU, and cupA), and a mutation in one gene that encodes a predicted DGC-PDE dual domain protein (43). A more detailed description of the mutants identified is included in the supplemental material and in Table S4 in the supplemental material.

There were three additional strains that fully recovered the ability to swarm, and each strain harbored an insertion that mapped to the motA gene (PA14_65450) (see Table S4 in the supplemental material). Strains with mutations in the motA gene were not isolated in the previous screen (34), confirming that the earlier screen had not reached saturation. The swarm phenotype of a representative suppressor strain (ΔbifA motA::Mar) is shown in Fig. 1A (top panel, third column from the left). As shown, this double mutant regained the ability to form tendril-like projections from the point of inoculation, in contrast to the ΔbifA mutant alone, and demonstrated swarming comparable to that of the wild type (Fig. 1A, top panel, leftmost image).

The motA gene encodes a component of the flagellar stator complex that, with MotB, forms the peptidoglycan-bound proton translocating complex that functions to power rotation of the flagellar rotor (44–46). Thus, we next asked whether suppression by the motA mutation is specific to the swarming motility defects of the ΔbifA mutant or whether additional c-di-GMP-related phenotypes are also affected. We therefore tested whether mutations in motA might also impact swimming, Pel polysaccharide production, and/or biofilm formation.

In the “swim” assay, cells move within the liquid of the 0.3% soft agar of the plate, as opposed to swarming motility, which occurs on the surface of the semisolid agar plate (28, 29). This soft-agar-based swim assay reflects the functionality of the flagellum, as well as the chemotaxis machinery. Thus, in these experiments, we used this low-percentage-agar assay as a surrogate to judge the functionality of the flagellar machinery. The ΔbifA mutant exhibits reduced motility (∼50% of wild type) in soft-agar plates (0.3% agar) (Fig. 2A and B). We found that the bifA motility defect in 0.3% agar is not apparently affected by the motA mariner mutation (Fig. 2A), indicating that this motility defect of the ΔbifA mutant is not obviously suppressed by mutation of the motA gene. Furthermore, as shown in Fig. 1A (middle row), the ΔbifA motA::Mar mutant shows no difference in CR binding relative to the ΔbifA mutant, suggesting that mutating the motA gene does not noticeably alter polysaccharide production under these assay conditions. In contrast, mariner disruption of the motA gene diminishes the hyperbiofilm formation phenotype of the ΔbifA mutant (Fig. 1A, bottom row and graph), reducing biofilm formation to a level that is not significantly different from that of the wild-type strain. Together, these data indicate that mutation of motA leads to restoration of swarming motility by the ΔbifA mutant in a manner independent of changes in swimming motility, as assayed on 0.3% soft-agar plates, or in polysaccharide production.

FIG 2.

Soft-agar motility assays of ΔbifA and stator mutants. (A) Representative motility assay of the indicated strains. Soft-agar plates (0.3% agar) were incubated at 37°C for 16 h. (B and C) Motility zone measurements are shown for the 0.3% soft-agar assay with the indicated strains. Measurements were obtained using Image J, with units in pixels (×10⁴). Error bars indicate standard deviations of the averages from three experiments with six replicate plates per experiment. Data were analyzed by one-way analysis of variance followed by a Tukey's posttest comparison. ns, not significantly different; ***, P < 0.001. In panel B, the ΔbifA mutant is compared to the wild type and the ΔbifA mot mutants are compared to the ΔbifA mutant, and in panel C, statistical comparisons are relative to the wild type.

To further confirm these phenotypes and rule out the possibility of secondary mutations from the mariner insertion, we generated an in-frame deletion of the motA gene in the ΔbifA mutant background and observed the same suppression of the bifA mutant swarming and biofilm phenotypes as for the mariner mutants (Fig. 1A, compare third and fourth columns). As was the case for the transposon mutation, deletion of the motA gene does not suppress the soft-agar motility defect of the ΔbifA mutant (Fig. 2A and B).

Thus far, our data suggest that MotA participates in the repression of swarming in the ΔbifA mutant. In the P. aeruginosa PA14 genome, the motA gene resides in a predicted operon with the motB gene (PA14_65430), which in all bacterial species studied so far encodes a second component of the flagellar stator. Studies in E. coli have established that the flagellar stator is comprised of MotA-MotB complexes, with stoichiometry of 4MotA:2MotB and a maximum number of 11 such complexes per rotor, at high torque, forming a peptidoglycan anchored ring in the inner membrane around the core rotor proteins (44, 46–50).

Given the genomic arrangement of these genes and functional partnership of their encoded products, we sought to identify the role that each of these genes might play in the repression of swarming in the ΔbifA mutant. To ensure that the in-frame deletion of motA did not also disrupt motB function, we performed a complementation test of the motA defect in the ΔbifA ΔmotA double mutant by replacing the motA deletion allele with the wild-type motA allele in the ΔbifA ΔmotA strain via allelic exchange, generating the ΔbifA ΔmotA::motA strain. As shown in Fig. 1A, the ΔbifA ΔmotA::motA strain exhibits a swarming defect and a hyperbiofilm phenotype that are indistinguishable from those of the ΔbifA single mutant, indicating that introduction of the wild-type motA allele into the ΔbifA ΔmotA double mutant fully restores the ΔbifA single mutant phenotypes. These results confirm that MotA plays a role in swarming inhibition in the ΔbifA mutant.

We next tested whether MotB also played a role in swarming inhibition by generating an in-frame deletion of the motB gene followed by complementation with a His₆ epitope-tagged version of motB on a multicopy plasmid (pMotB-His). Results indicate that deletion of motB in the ΔbifA mutant background showed similar suppression of the ΔbifA mutant phenotypes as observed when motA was mutated (Fig. 3A). Furthermore, complementation with pMotB-His fully restored the ΔbifA single mutant phenotypes, indicating that MotB also participates in swarming inhibition in the ΔbifA mutant (Fig. 3A). As was the case for mutation of the motA gene, deletion of the motB gene does not suppress the motility defect of the ΔbifA mutant in the soft-agar assay (Fig. 2B and 3B). Examination of a ΔbifA ΔmotAB triple mutant showed that this strain has phenotypes similar to that of either the ΔbifA ΔmotA or ΔbifA ΔmotB double mutant (Fig. 1A, 2A and B, and 3) and confirmed that MotA and MotB participate together in repression of swarming in the ΔbifA mutant.

FIG 3.

Assessment of motB mutation on suppression of bifA mutant phenotypes. (A) The genotype for each strain tested in this panel is shown on the x axis of the graph. The top row shows representative images of swarm assays for the indicated strains. Swarm plates were incubated at 37°C for 16 h. The second row shows CR binding for each strain. CR plates were incubated at 37°C for 24 h followed by 48 h of incubation at room temperature. The bottom row shows images of wells from a 96-well biofilm assay plate. The graph depicts quantification of CV-stained biofilms. Error bars indicate standard deviations of the averages from three experiments with four replicates per experiment. Data were analyzed by one-way ANOVA followed by Tukey's posttest comparison. ns, not significantly different; ***, P < 0.001 (relative to the ΔbifA mutant). (B) Representative soft-agar motility assay for the indicated strains. Plates (0.3% agar) were incubated at 37°C for 16 h.

Finally, we generated single motA and motB deletion strains as well as a double motAB deletion strain in an otherwise wild-type background to further examine the role these stator components play in regulating motility. The ΔmotA, ΔmotB, and ΔmotAB strains all showed increased swarming motility. For example, the ΔmotA mutant showed an ∼2.5-fold increase in percent surface coverage of the swarm plate relative to the wild-type strain (Fig. 1A and B). Additionally, we showed that this hyperswarming phenotype can be complemented by allelic replacement of the motA deletion allele with a wild-type motA allele (Fig. 1A, compare the wild type [WT] to the ΔmotA::motA strain). Furthermore, each of these ΔmotA, ΔmotB, and ΔmotAB mutant strains was able to move in the standard soft-agar (0.3% agar) motility assay, albeit at approximately 60% the level of the wild type (Fig. 1C, 2C, and 3B). These results indicate that the MotAB stator likely plays a positive but relatively minor role in motility under the soft-agar assay conditions studied here; however, this stator has a measurable negative impact on swarming motility, even when c-di-GMP levels are relatively low in the WT strain (compared to elevated conditions observed for the ΔbifA mutant). Taken together, our data indicate that the role of MotAB in repressing swarming motility is not specific to the ΔbifA mutant, as we observe similar negative impacts on swarming in the WT.

Based on the observations that the MotAB stator negatively impacts swarming, we assessed whether overexpression of a His₆ epitope-tagged version of motA from a multicopy plasmid (pMotA-His) under the control of the arabinose-inducible pBAD promoter in the wild-type strain would lead to repression of swarming, and that is indeed what we observed (Fig. 1D). However, overexpression of the pMotB-His plasmid alone did not repress swarming in the wild type (data not shown). These data suggest that MotA may play a specific role in repression of swarming; possible explanations include that only MotA and not MotB can be incorporated into the stator when overexpressed or, alternatively, that MotB incorporation into the stator requires MotA and that assembly of MotB into the motor is necessary for its repressive function. Recent studies in Salmonella support the notion that overexpression of MotA alone can lead to its incorporation into the motor, with subsequent deleterious effects on motility (51).

P. aeruginosa has two stators sets with distinct and opposing roles in motility.

Despite the fact that P. aeruginosa builds a single polar flagellum, its genome encodes two flagellar stator sets, MotAB (as described above) and MotCD (PA14_45560/45540) (52–54). In contrast, E. coli and Salmonella enterica, both well-studied model organisms for flagellar structure and function, build multiple flagella per cell (i.e., they are peritrichous), but the genomes of these microbes encode only a single MotAB stator set. As outlined above, our data suggest that the MotAB stator has a negative impact on swarming motility under wild-type conditions as well as when c-di-GMP levels are highly elevated (i.e., in the ΔbifA mutant).

In contrast, the single ΔmotC and ΔmotCD double mutant strains exhibit complete loss of swarming motility and are indistinguishable from the ΔflgK nonmotile control (Fig. 1B; also, see Fig. S1A in the supplemental material), indicating that the MotCD stator is critical for swarming motility under these conditions. In soft-agar motility assays, the ΔmotC and ΔmotCD mutants exhibit a very small but detectable swim zone that is typically less than 5% of that of the wild type and that is absent in the ΔflgK mutant (Fig. 1C; also, see Fig. S1B in the supplemental material). These results are consistent with MotCD being the predominant stator required for swimming motility in addition to its critical role in swarming. However, taken together with the observation that ΔmotAB mutants also exhibit a modest decrease in swimming, our data suggest that swimming motility under these conditions is optimal when both MotAB and MotCD stators are present. Given that a stator is comprised of multiple Mot protein complexes, with up to 11 MotAB complexes in E. coli (49), we infer from these data that MotAB and MotCD complexes could both be present in the stator of a given flagellar motor.

Discrepancy between findings here and in a previous study from our group.

In earlier work from our group, we reported that the ΔmotCD mutant was able to swim as well as the wild type and the ΔmotAB mutant on 0.3% agar plates and that the motCD mutant was also able to swarm on plates solidified with 0.45% agar but not on 0.5% agar plates (53). In this study, we observed very little swimming and no swarming on either 0.45% or 0.5% agar plates by the ΔmotC or ΔmotCD mutants (Fig. 1B and C; also, see Fig. S1 in the supplemental material), findings that are similar to those reported for P. aeruginosa strain PAO1 (54). The exact cause of the discrepancies between the earlier and current study is not clear, but one possibility is that the ΔmotCD mutant became contaminated with wild-type cells at some point during experimentation in the earlier study.

For the current study, the ΔmotCD deletion strain was reconstructed, and additionally, an individual ΔmotC deletion mutant was built for a more thorough phenotypic characterization of the ΔmotCD locus. Here we show that the newly generated ΔmotCD and ΔmotC deletion strains are phenotypically indistinguishable from one another, and we further show complementation of the ΔmotC strain using a His₆ epitope-tagged version of ΔmotC on a multicopy plasmid (pMotC-His) to restore swimming and partially restore swarming to this mutant (see Fig. S1 in the supplemental material). Thus, we feel confident that the current study provides a more accurate assessment of the role that MotCD plays in motility. Taken together, the data in the current study suggest that the stator sets of P. aeruginosa, MotAB and MotCD, have distinct roles in controlling motility and most notably, as described above, have opposite roles in impacting swarming motility in particular.

c-di-GMP levels do not impact MotA and -C protein levels or subcellular localization.

Given that the MotAB stator is required for robust repression of swarming when c-di-GMP levels are elevated, we hypothesized that c-di-GMP levels might influence the abundance and/or subcellular localization of MotAB or MotCD in cells. To address this possibility, we constructed epitope-tagged versions of either motA or motC at their native loci on the chromosome in either the wild-type or bifA mutant backgrounds (motA⁺-His, ΔbifA motA⁺-His, motC⁺-HA, and ΔbifA motC⁺-HA). It should be noted that the His₆ epitope tag had no impact on MotA function in either the wild-type or ΔbifA mutant strain (see below). However, given the critical role of MotC in swimming and swarming motility, we did observe that the HA epitope tag led to a moderate reduction in swimming by both the wild-type and ΔbifA mutant strains, as well as a strong reduction in swarming by the wild type (data not shown).

We next fractionated (swarm) surface-grown cells and probed for the presence of these proteins in various cellular fractions. In the wild-type strain backgrounds, we detected both the MotA-His and MotC-HA proteins predominantly in the total membrane (TM) and inner membrane (IM) fractions but not in the cytosolic or outer membrane fractions (Fig. 4), which is consistent with the cellular organization of the stator proteins in the flagellum from studies in E. coli (47, 48). We found no obvious changes in the localization pattern of either MotA-His or MotC-HA in the ΔbifA mutant cells relative to wild-type cells in any of the fractions, nor did we observe any obvious differences in the levels of these two proteins when comparing these strains.

FIG 4.

Subcellular fractionation and level of MotA-His and MotC-HA. Cellular fractions of swarm-grown cells of the motA⁺-His and ΔbifA motA⁺-His strains (top) and motC⁺-HA, and ΔbifA motC⁺-HA strains (bottom) were separated on 12% SDS-PAGE gels. Fractions are indicated as whole cell (WC), soluble cytoplasmic (Cyt), total membrane (TM), inner membrane (IM), and outer membrane (OM). Lanes are labeled according to the legend in the lower left corner. Western analysis was performed using the following antibodies, as indicated: anti-His (for MotA-His), anti-HA (for MotC-HA), anti-SecY, and anti-OprF. SecY and OprF serve as controls for integrity of inner and outer membrane fractions, respectively.

To confirm that MotA protein levels do not increase with c-di-GMP levels, we quantified MotA-His protein levels in whole-cell-lysate samples for each strain by Western blotting and found no significant difference between the two strains (measured as integrated pixel density × 10⁴: the WT value was 21.5 ± 1.9, whereas that for the bifA strain was 19.2 ± 3.8 [P = 0.29 by t test; three independent experiments, with three replicates per strain]). Together these data argue against a simple model wherein c-di-GMP levels alter motility by controlling expression levels or membrane localization of either MotA or MotC.

MotA-FliG interactions are important for repression of swarming.

While the precise mechanism by which the Mot proteins are assembled into the stator of the flagellum is not well understood, genetic studies in E. coli and S. enterica have indicated that specific interactions between MotA and the FliG rotor protein are important not only for rotation of the flagellum but also for assembly of MotA into the motor (51, 55–57). The rotor/stator interface involves a small number of highly conserved amino acids in each protein, the main feature of which is their charge, as these interactions are believed to be largely electrostatic (56–58).

Extensive mutational analyses have been used to define the MotA-FliG interface (56–58). In MotA, an arginine at position 90 (R90) and a glutamic acid at position 98 (E98) have been shown to be important in interacting with D289 and R281, respectively, of FliG. For example, mutation of MotA R90 to E (R90E) reverses the positive charge to negative and leads to full inhibition of motility, whereas compensatory mutation of FliG D289 to A (neutral) or K (positive) in the motA(R90E) strain partially suppresses this motility defect. Furthermore, additional studies from Morimoto et al. indicate that the MotA R90/FliG D289 interaction may play a more important role in stator assembly, while the MotA E98/FliG R281 interaction appears to be more influential in torque generation (55).

Given that the R90 and E98 residues of MotA and D289 and R281 of FliG are conserved in the respective P. aeruginosa proteins, we tested whether MotA-FliG interactions might be important for the participation of MotA in repression of swarming. To this end, we generated a C-terminal His₆ epitope tag in the motA gene at its native locus on the chromosome in either the wild-type or the ΔbifA background. As mentioned above, the His₆ epitope tag had no impact on MotA function in either the wild-type or ΔbifA mutant strain (Fig. 5, compare to Fig. 1). We then introduced the substitutions to generate R90E or E98K variations in this chromosomal motA locus encoding MotA-His. First, we examined protein levels in cells cultured on a swarm agar surface and found that the MotA(R90E)-His and MotA(E98K)-His mutant proteins are present at levels comparable to that of wild-type MotA-His (Fig. 5 and Fig. S2 in the supplemental material, respectively).

FIG 5.

Assessment of the role MotA-FliG interactions play in swarming inhibition. The strain genotype for each strain tested in this panel is shown on the x axis of the graph. The top row shows representative swarm images for the indicated strains. Swarm plates were incubated at 37°C for 16 h. The second row shows CR binding for each strain. CR assay plates were incubated at 37°C for 24 h, followed by 48 h at room temperature. The bottom row shows images of wells from a representative 96-well biofilm assay plate. Strains were grown in M63 medium supplement with glucose, MgSO₄, and CAA for 24 h prior to crystal violet (CV) staining. The graph depicts quantification of CV-stained biofilms. CV was solubilized in 30% glacial acetic acid and the absorbance was read at 550 nm. Error bars indicate standard deviations of the average of three experiments with four replicates per experiment. Data were analyzed by one-way ANOVA followed by Tukey's posttest comparison. ns, not significantly different; *, P < 0.05; ***, P < 0.001 (all relative to the ΔbifA motA⁺-His strain, except where indicated). The bottom panel shows a representative Western blot assessing chromosomally expressed MotA-His and MotA(R90E)-His protein levels in strains with the indicated genotypes. Lanes: 1, motA⁺-His; 2, motA(R90E)-His; 3, ΔbifA motA⁺-His; 4, ΔbifA ΔmotA; 5, ΔbifA motA(R90E)-His; 6, ΔbifA motA(R90E)-His fliG(D289A); 7, ΔbifA motA⁺-His fliG(D289A). Equal amounts of total protein from whole-cell lysates prepared from each strain were resolved by SDS-PAGE on a 12% polyacrylamide gel. The His₆ epitope-tagged proteins were detected using an anti-penta-His antibody.

If the MotA-FliG interaction is important for repression of swarming, then we would predict that the R90E and E98K mutations would disrupt this interaction, thereby leading to increased swarming, particularly in the ΔbifA mutant background. Indeed, this is what we observed (Fig. 5; also, see Fig. S2 in the supplemental material). Both the ΔbifA motA(R90E)-His and ΔbifA motA(E98K)-His strains exhibited recovery of swarming motility comparable to that of the ΔbifA ΔmotA deletion strain, suggesting that these mutations largely disrupt function of MotA in repressing swarming.

Interestingly, however, the R90E substitution shows virtually no effect on the hyperbiofilm phenotype of the ΔbifA mutant, in contrast to the substantial reduction in biofilm formation observed for the ΔbifA motA(E98K)-His strain (see Fig. S2 in the supplemental material), which more closely resembles the ΔbifA ΔmotA strain in this regard. Thus, the R90E mutation appears to specifically impair the ability of MotA to participate in swarming repression but not in biofilm formation, thereby separating these as two distinct functions of MotA. This notion is further supported by the motA(R90E)-His single mutant strain, which exhibits increased swarming relative to the wild type {surface coverage = 20% ± 4% [WT] and 52% ± 8% [motA(R90E)]} but has no significant impact on biofilm formation (Fig. 5, first and second columns on the left). In contrast, the motA(E98K)-His single mutant strain shows increased swarming relative to the wild type but shows a significant reduction in biofilm formation (see Fig. S2 in the supplemental material). It is unclear why the R90E mutation has no impact on biofilm formation, but it seems unlikely that this distinction could be explained simply by differences in swimming motility or CR binding given that there are no obvious phenotypic differences between the motA(R90E)-His and motA(E98K)-His strains and the motA deletion strain in either the wild-type or ΔbifA mutant background in these assays (Fig. 5; also, see Fig. S2 and S3 in the supplemental material). Furthermore, the observation that the motA(R90E)-His and motA(E98K)-His mutations in the WT background also result in enhancement of swarming (Fig. 5; also, see Fig. S2, top; compare to the motA⁺-His strain) suggests that the observed effects of these motA point mutations on swarming are not specific to the ΔbifA mutant background.

To further probe whether the MotA-FliG interaction is important for swarming repression, we sought to introduce a compensatory fliG mutation into either the ΔbifA motA(R90E)-His or ΔbifA motA(E98K)-His strain; based on the studies in E. coli and Salmonella, this mutation would be expected to restore interaction of FliG with the mutant version of MotA. If our hypothesis is correct, then restoration of the interaction between these variant proteins in the ΔbifA mutant background should restore swarming inhibition. One additional but important consideration, in our case, was that the FliG-MotC interaction presumably must be maintained for swarming motility to occur based on the essential role for MotC particularly in swarming motility (see above). If this interaction was disrupted, then we could not properly evaluate the impact of a fliG mutation on the ΔbifA motA(R90E)-His or ΔbifA motA(E98K)-His strain.

Given that the R90 and E98 residues of MotA are also conserved in MotC, we surmised that interaction of MotC with FliG would occur via the same electrostatic interactions. Thus, we chose the fliG(D289A) substitution that in E. coli acted as an allele-specific suppressor of the motA(R90E) nonmotile mutant (restored ∼40% of wild-type motility) with little impact on motility in a motA⁺ background (exhibits ∼92% of wild-type motility). In contrast, fliG(R281V) or fliG(R281W) single mutations which act as allele-specific suppressors of the motA(E98K) mutant led to loss of motility and would be predicted to negatively impact the FliG/MotC interaction.

In P. aeruginosa, the motA⁺-His fliG(D289A) strain showed a modest decrease (∼20% reduction) in motility in the soft-agar assay relative to the motA⁺-His parent strain (Fig. 6A). However, this strain did not exhibit an obvious defect in swarming motility (Fig. 6B), nor was there a significant difference in biofilm formation (data not shown). It is unclear why the fliG(D289A) mutation has an impact on swimming but not swarming motility; however, based on these phenotypes, we concluded that MotCD function is not significantly impaired by this fliG mutation.

FIG 6.

Motility assays to assess the impact of motA and fliG mutations. Images show representative motility in 0.3% soft-agar motility (A) and swarm (B) assays for strains with the indicated genotypes.

Introduction of the fliG(D289A) mutation into the ΔbifA motA⁺-His strain had little impact on the ΔbifA mutant phenotypes, save for a modest reduction in biofilm formation (Fig. 5). Swarming motility remained inhibited (Fig. 5), indicating that mutation of fliG alone did not disrupt the ability of MotA to repress swarming. However, introducing the fliG(D289A) mutation into the ΔbifA motA(R90E)-His strain led to a marked reduction in swarming motility relative to the robust swarming observed for the ΔbifA motA(R90E)-His strain alone; more specifically, the ΔbifA motA(R90E)-His fliG(D289A) strain showed severely reduced tendril formation (Fig. 5). Thus, while the triple mutant strain is not identical to the ΔbifA single mutant in terms of full swarming inhibition, the observed reduction in swarming motility suggests that the compensatory fliG(D289A) mutation does restore some measure of swarming inhibition to the ΔbifA motA(R90E)-His mutant, as predicted.

In contrast to the swarming phenotype, there is no negative impact on swimming motility of this triple ΔbifA motA(R90E)-His fliG(D289A) mutant relative to either of the double mutants alone, supporting the notion that these mutations together do not impair MotCD function (Fig. 6; also, see Fig. S3 in the supplemental material). Taken together, these data support the hypothesis that repression of swarming by MotA relies upon an interaction with FliG, which, based upon the studies in E. coli and Salmonella, likely influences assembly of MotA into the stator.

Elevated c-di-GMP levels impact polar localization of GFP-MotD but not GFP-MotB.

Thus far, our data suggest that the different stator complexes of P. aeruginosa, MotAB and MotCD, play distinct roles in regulating swarming motility. That is, the MotCD complex is required for swarming motility, while the MotAB stator complex contributes to the repression of swarming motility in the wild type as well as in the ΔbifA mutant, where c-di-GMP levels are elevated. Furthermore, the ability of MotAB to repress swarming depends upon interaction of MotA with FliG in the motor, suggesting that repression by MotAB may require incorporation of this stator complex into the motor.

Together these data lead us to hypothesize that c-di-GMP levels might impact stator composition at the flagellar motor. In support of this hypothesis, it is known that stators in functioning motors exchange with stators diffusing in the membrane (without requiring disassembly and reassembly of the transmembrane rotor components of the motor) and that such exchanges could occur rapidly enough to allow cells to respond to changing c-di-GMP levels (39, 59, 60).

To assess whether c-di-GMP levels affect localization of the stator components, we replaced motB and motD genes with gfp-motB and gfp-motD, respectively, at their native chromosomal loci so that expression would occur similarly to that in their wild-type counterparts. When they were assessed in swarming assays, we found that the gfp-motB and gfp-motD strains swarmed comparably to the wild-type parental strain (see Fig. S4A in the supplemental material), indicating that the gfp fusions did not noticeably impact function of either MotB or MotD. Furthermore, the gfp-motB fusion had little impact on the ΔbifA swarming-defective strain, indicating no significant loss of motB repressive function (see Fig. S4B in the supplemental material).

Using fluorescence microscopy, we analyzed the localization of the GFP-labeled stator units in bacteria grown on the surface under conditions identical to those used in our swarming motility assays. For each strain, mobile and static bright spots (or puncta) of fluorescence were observed, although with different patterns depending on the strain. For each field of view imaged, we counted the total number of cells and the number of cells showing a polar fluorescent spot to assess whether c-di-GMP level affected polar localization of either of the GFP-labeled stator units.

For GFP-MotB, we saw no differential polar localization of the GFP-MotB fusion protein in the wild type compared to any of the other mutants analyzed, including the ΔbifA mutant, the ΔbifA motA(R90E) or ΔbifA fliG(D289A) double mutant, or the ΔbifA motA(R90E) fliG(D289A) triple mutant (Fig. 7A). These data indicate that the polar localization of GFP-MotB is not noticeably affected by c-di-GMP level or by alterations in MotA-FliG interactions. Furthermore, there was also no difference in GFP-MotB polar localization in the ΔmotCD deletion mutant relative to the wild type (Fig. 7A), suggesting that polar localization of GFP-MotB is not detectably influenced by the presence or absence of MotCD. However, we cannot rule out the possibility that more subtle changes in MotB localization do occur.

FIG 7.

High c-di-GMP alters the localization of GFP-MotD, as assessed by fluorescence microscopy. (A) Percentage of cells showing polar puncta of GFP-MotB plotted for the strains indicated. There are no discernible differences among any of these strains tested (2 replicate experiments per strain tested). (B) Representative images of the WT and ΔbifA mutant strains expressing GFP-MotD from the chromosome. (C) The percentage of cells with GFP-MotD polar puncta is plotted for the wild type (WT) and the ΔbifA mutant. Three independent replicate experiments were performed per strain, with an average of at least 200 total cells counted per experiment. *, significantly different, as judged by a t test (P < 0.05).

In contrast, we did observe changes in GFP-MotD localization. Shown in Fig. 7B are representative images of cells expressing the GFP-MotD fusion protein. In the wild-type genetic background, the GFP-MotD puncta were observed at the pole as well as in other locations on the periphery of the cells, consistent with the known membrane localization of stator components. Interestingly, the percentage of cells with polar puncta for GFP-MotD (∼25%) was less than that observed for GFP-MotB (∼45%) (Fig. 7A and C). The significance of this difference is not clear.

Under conditions of high c-di-GMP in the ΔbifA mutant background, there were fewer distinct GFP-MotD puncta in these cells compared to the WT (Fig. 7B). When we quantified the percentage of cells showing a polar fluorescent spot, we found that there was a significant reduction of the GFP-MotD puncta at cell poles in the ΔbifA mutant relative to the wild type (Fig. 7C). These data suggest that an elevated c-di-GMP level does impact localization of MotD at the cell pole, with fewer GFP-MotD puncta in the ΔbifA mutant, a strain that lacks swarming motility.

DISCUSSION

Our work here is consistent with the identification of a new mechanism for c-di-GMP-mediated regulation. That is, we propose that c-di-GMP can impact regulation of swarming motility, likely via the differential utilization of the two sets of stators present in P. aeruginosa. It is clear that swimming motility in organisms with a single stator system (i.e., E. coli and S. enterica) is also c-di-GMP regulated, but for these organisms, control appears to be at the level of regulating the single stator function and/or its interaction with the motor (15, 16, 18). Here we propose that the two stators play discrete roles in the regulation of swarming motility. MotCD appears to be the major torque-generating stator, as mutating the motCD genes results in loss of swarming. In contrast, loss of MotAB results in a hyperswarming phenotype, while overexpression of MotA fully represses swarming.

We also showed using a genetic approach that interactions between MotA and FliG are critical for MotA-mediated suppression of swarming, indicating that this stator component functions in repression by occupying the motor in a mechanism analogous to that characterized in E. coli and S. enterica (51, 57). As a follow-up to these studies, using a bacterial two-hybrid analysis, we explored whether the mutations we generated in MotA and FliG do indeed impact MotA-FliG interactions. We were unable to detect any evidence for MotA-FliG interactions using a variety of different construct combinations (data not shown); however, these interactions have not been demonstrated for other species and are considered unlikely given the transient, electrostatic nature of the stator/rotor interaction.

Overall, our data suggest that differential occupancy of the motor by the MotAB versus MotCD stators may impact swarming motility. Such a model would suggest that stators must exchange in the motor in response to differential environmental signals. Two key studies in E. coli provide strong evidence for stator exchange. Both studies examined stator incorporation into functioning flagellar motors in real time and led to the discovery that stator assembly is a remarkably dynamic process. An early study showed that stator components of E. coli could be added to an otherwise assembled but nonrotating flagellum of a mot mutant (tethered by its flagellum) by inducing expression of the motAB genes on a plasmid, leading to discrete stepwise increases in the speed of flagellar rotation, with each step corresponding to the addition of a new stator complex (59). These results established not only that stator complexes could be added as the last component to the flagellum (i.e., flagellar assembly was not dependent upon stator incorporation) but also that stator complexes could be added while the flagellum was actually rotating. A more recent study used total internal reflection fluorescence (TIRF) microscopy to track the movement of GFP-MotB molecules into and out of the stator of E. coli cells tethered by their flagellar filaments (39). This study found that GFP-MotB (presumably in complex with its MotA partner) undergoes a process of stator exchange whereby this molecule shuttles between an inner membrane pool of ∼200 GFP-MotB molecules and the stator of the motor, with a given GFP-MotB molecule having an average residence time in the stator of only ∼30 s. Moreover, recent studies showed that the number of stators engaged in a motor at any one time also may change as the external force on the flagellum changes. For example, the number of stators engaged in a motor increases with increasing external forces, as would occur with increasing viscosity, such as on swarm agar or at a surface during biofilm formation (50, 61), suggesting a structural change within the rotor that retains stators and allows increased torque.

Additional evidence for stator exchange comes from studies in Shewanella, an organism with both proton-driven and sodium-driven stator sets powering a single polar flagellum. Shewanella oneidensis has been shown to preferentially use one stator set over the other depending on environmental conditions (60), thus providing an additional precedent for more than one stator functioning in the same motor.

At this point, the mechanism by which the cell employs c-di-GMP to control stator utilization is not clear. Our findings show that high c-di-GMP level is associated with a decrease in polar localization of GFP-MotD. Thus, it is possible that c-di-GMP drives relocalization of the MotCD stator away from the pole; however, we cannot rule out the possibility that additional mechanisms contribute to this decrease in polar-localized MotD. It is important to note that we used a mutant strain in these localization studies (the ΔbifA mutant), and thus, we may need to identify more physiologically relevant conditions for high c-di-GMP levels to better understand this localization phenomenon. We reported previously that c-di-GMP levels are elevated when P. aeruginosa is grown on a surface versus planktonically (37). However, comparing planktonically versus surface-grown cells to mimic a physiological increase in c-di-GMP levels is complicated by the fact that to image planktonically grown cells, one must place them on a surface. Future studies might be aided by using fixed cells grown under these different conditions. In contrast to the MotCD stator, the MotAB stator set, as judged by GFP-MotB localization, seems largely unresponsive to c-di-GMP levels or any other condition tested in this study regarding its polar localization. Thus, our current data are consistent with a model whereby c-di-GMP likely controls occupancy of the motor by the different stator sets primarily by shifting localization of MotCD.

Previous studies identified a role for PilZ-like proteins in the control of flagellar function, for example, the YcgR protein in E. coli, although the precise mechanism of this control is still controversial (2, 14–16, 18). P. aeruginosa PA14 has 8 PilZ-like proteins, including one protein with high sequence similarity to YcgR, and we are currently exploring which of these proteins might contribute to c-di-GMP-mediated swarming regulation via the stators.

In summary, our data are consistent with a model wherein P. aeruginosa uses two distinct stator sets to control surface motility in response to c-di-GMP. The novelty of these findings indicates that there are likely still other, as-yet-unidentified c-di-GMP outputs to uncover and highlights the complexity and adaptability of this signaling network. Furthermore, analyses of bacterial genomic sequence data indicate that in addition to P. aeruginosa, microbes with a single flagellar motor that possess two or more stator sets are surprisingly common and widespread across bacterial species (53, 62). Thus, it is possible that our findings may be extended to these other organisms with two stator sets, such as species of Xanthomonas and Burkholderia, as well as Yersinia pestis and Chromobacterium violaceum (53, 62); however, we cannot rule out the possibility that the dual stators play roles other than c-di-GMP-mediated motility control in these different organisms.