ABSTRACT
Individual cells of the bacterium Proteus mirabilis can elongate up to 40-fold on surfaces before engaging in a cooperative surface-based motility termed swarming. How cells regulate this dramatic morphological remodeling remains an open question. In this paper, we move forward the understanding of this regulation by demonstrating that P. mirabilis requires the gene rffG for swarmer cell elongation and subsequent swarm motility. The rffG gene encodes a protein homologous to the dTDP-glucose 4,6-dehydratase protein of Escherichia coli, which contributes to enterobacterial common antigen biosynthesis. Here, we characterize the rffG gene in P. mirabilis, demonstrating that it is required for the production of large lipopolysaccharide-linked moieties necessary for wild-type cell envelope integrity. We show that the absence of the rffG gene induces several stress response pathways, including those controlled by the transcriptional regulators RpoS, CaiF, and RcsB. We further show that in rffG-deficient cells, the suppression of the Rcs phosphorelay, via loss of RcsB, is sufficient to induce cell elongation and swarm motility. However, the loss of RcsB does not rescue cell envelope integrity defects and instead results in abnormally shaped cells, including cells producing more than two poles. We conclude that an RcsB-mediated response acts to suppress the emergence of shape defects in cell envelope-compromised cells, suggesting an additional role for RcsB in maintaining cell morphology under stress conditions. We further propose that the composition of the cell envelope acts as a checkpoint before cells initiate swarmer cell elongation and motility.
IMPORTANCE Proteus mirabilis swarm motility has been implicated in pathogenesis. We have found that cells deploy multiple uncharacterized strategies to handle cell envelope stress beyond the Rcs phosphorelay when attempting to engage in swarm motility. While RcsB is known to directly inhibit the master transcriptional regulator for swarming, we have shown an additional role for RcsB in protecting cell morphology. These data support a growing appreciation that the Rcs phosphorelay is a multifunctional regulator of cell morphology in addition to its role in microbial stress responses. These data also strengthen the paradigm that outer membrane composition is a crucial checkpoint for modulating entry into swarm motility. Furthermore, the rffG-dependent moieties provide a novel attractive target for potential antimicrobials.
INTRODUCTION
Bacteria can migrate across a surface using a cooperative group motility termed swarming. For Proteus mirabilis, a Gram-negative opportunistic pathogen, rapid surface-based swarm motility likely contributes to its pathogenesis during catheter-associated urinary tract infections (1, 2). On hard agar (1.5% to 2%) surfaces, cells elongate from ∼2-μm rods into hyperflagellated, snake-like “swarmer” cells that carry multiple chromosomes and range in length from 10 to 80 μm (3–5). Cell elongation and enhanced flagellar gene expression are considered genetically linked and occur upon growth on a hard agar surface (6–9). Multiple swarmer cells closely associate into rafts that collectively move across a surface (3, 5, 10). After a defined period of motility, swarmer cells divide into short (1 to 2 μm) nonmotile rod-shaped cells (11). Iterative rounds of swarmer cell elongation, group motility, and cell division comprise the swarmer cell developmental cycle and result in the rapid occupation of centimeter-scale surfaces in a stereotypical concentric ring pattern (11).
Swarmer cell elongation entails a broad range of physiological changes in addition to the dramatic morphological remodeling. Dozens of genes experience drastic changes in expression; for example, genes for flagellum production become upregulated on surfaces (12–16). Elongation into swarmer cells is also coordinated with changes to the cell envelope that minimally include alterations of the outer membrane. For example, in the outer membrane, the lipopolysaccharide (LPS) structure is modified, fluidity is increased, and areas of phospholipid bilayer arise (17–19). Cells also transition via unknown mechanisms from being rigid to being flexible (3–5).
In many bacteria, a bidirectional relationship exists between swarmer cell motility and cell envelope structure. There are several cell envelope biosynthesis pathways in Enterobacteriaceae, such as LPS and enterobacterial common antigen (ECA) biosynthesis for the outer membrane and peptidoglycan biosynthesis for the cell wall. Interrogating the specific contribution of each pathway to P. mirabilis swarmer cell development has proven challenging, partly because these three pathways share a pool of substrates (20, 21). For example, in Escherichia coli, genetic modifications to each of these biosynthetic pathways can dramatically alter cell shape and motility due to perturbations in the balance of the shared cell envelope substrate, undecaprenyl phosphate (20, 21). Moreover, the disruption of cell envelope-associated genes inhibits swarmer cell development and motility of P. mirabilis through several mechanisms. For example, the loss of the LPS biosynthesis gene waaL (22, 23) inhibits swarmer cell elongation and motility through the activation of the Rcs phosphorelay (23), while the stress-associated sigma factor RpoE (24, 25) responds to disruptions of the LPS biosynthesis gene ugd (25). Less is known about the role of ECA biosynthesis in P. mirabilis.
Cell envelope structure and stress sensing also appear to play broadly conserved roles in the swarm regulation of many bacterial species, including P. mirabilis, E. coli, and Serratia marcescens (20–22, 24, 26–30). In the aforementioned organisms, the Rcs (regulator of capsule synthesis) phosphorelay, which is a complex cell envelope stress-sensing signal transduction pathway, plays a key role in swarm motility inhibition (22, 26, 31). The Rcs phosphorelay, through the transcriptional regulator RcsB, directly represses the flhDC genes, which themselves encode the master transcriptional regulator of swarming, FlhD4C2 (27, 29). The current paradigm is that cell envelope stress or outer membrane defects activate membrane-localized Rcs proteins, which then phosphorylate and activate the response regulator RcsB (22, 26, 27, 31; see also references 32 and 33 for review). Decreased levels of flhDC result in reduced flagellum production and the failure of cells to elongate, thus inhibiting swarm motility. RcsB directly activates the expression of the cell division-related genes, minCDE; however, the molecular mechanisms of this regulation remain unclear (6, 7). RcsB also induces the production of several fimbrial genes, including paralogous genes of the fimbrial transcriptional regulator MrpJ. Together, RcsB and MrpJ modulate broad transcriptional and behavioral changes to promote cell adherence and biofilm formation and to repress swarm motility (7, 34).
Here, we address the role of the cell envelope and stress-sensing pathways in the regulation of swarmer cell development, an early stage of swarm motility. We show that P. mirabilis cells require the rffG gene, which is predicted to encode the sugar-modifying enzyme dTDP-glucose 4,6-dehydratase, to produce an uncharacterized LPS-linked structural component of the cell envelope. As a homologous rffG gene and its conserved cluster of flanking genes are responsible for ECA production in E. coli (35), we posit that these structures may be ECA derived. We further show that cells lacking the rffG gene remain short on swarm-permissive surfaces and suffer from cell envelope integrity defects that make elongated cells more susceptible to rupturing. We found that rffG-dependent moieties were not physically required for swarmer cell elongation; instead, the loss of the rffG gene activated several swarm-inhibitory pathways, including the Rcs phosphorelay. Indeed, an RcsB-mediated response was sufficient to restrict swarmer cell elongation of rffG-deficient cells by inhibiting flhDC expression. We have also identified a novel role for RcsB in the maintenance of cell morphology during swarmer cell elongation. We found that RcsB was necessary to suppress cell morphology defects of rffG-deficient cells that were genetically forced to elongate into swarmer cells. We posit that the cell envelope composition is a crucial signaling checkpoint before entry into surface-based swarm motility. The Rcs phosphorelay response regulator not only mediates this signaling checkpoint but also serves an important role in maintaining a normal cell shape during swarmer cell elongation.
RESULTS
Cells require the rffG gene to complete swarmer cell elongation and initiate swarming.Previous research explored the role of LPS biosynthesis genes in the regulation of P. mirabilis swarm motility, but a role for ECA has not been described (23, 25). Here, we interrogated the role in swarming of a gene associated with ECA biosynthesis. We characterized a swarm-deficient mutant strain presumably incapable of producing ECA by generating a chromosomal deletion of the rffG gene in P. mirabilis strain BB2000, resulting in a ΔrffG strain. A colony of the wild-type strain occupied a standard-size petri dish by 24 h on swarm-permissive and nutrient-rich CM55 agar; however, colonies of the ΔrffG strain did not expand beyond the site of inoculation (Fig. 1A). We complemented the rffG deletion through in trans expression of the rffG gene under the control of a lac promoter for constitutive expression in P. mirabilis (23), resulting in the ΔrffG(prffG) strain. The wild-type and the ΔrffG strain each carried empty vectors (pBBR1-NheI) to enable growth on the same selective medium as the ΔrffG(prffG) strain. The swarm colonies of the ΔrffG(prffG) strain were attenuated in comparison to those of the wild-type strain and more expansive than those of the ΔrffG strain (Fig. 1A), indicating a partial rescue of swarm motility.
Loss of the rffG gene inhibits swarmer cell elongation and swarm motility. (A) The wild-type(pBBR1-NheI), ΔrffG(pBBR1-NheI), and ΔrffG(prffG) strains were grown on swarm-permissive medium containing kanamycin as well as Coomassie blue and Congo red dyes for imaging. Images are representative of at least three independent experiments for each strain. (B) Phase-contrast and epifluorescence microscopy of the wild-type(pBBR1-NheI), the ΔrffG(pBBR1-NheI), and the ΔrffG(prffG) strains after 2, 4, and 6 h on swarm-permissive medium containing kanamycin. All strains encode Venus immediately downstream of fliA. Fluorescence corresponding to fliA reporter expression is shown for the 6-h time point. Rolling ball background subtraction was performed using FIJI (73). Arrowheads highlight an elongating cell in the ΔrffG strain that is bulging. Frames from a time lapse of such cells bulging are in Fig. S1E in the supplemental material. At bottom are cartoon depictions of the morphological state of cells grown on swarm-permissive solid medium. On surfaces, cells elongate up to 40-fold before engaging in motility and dividing into short (1 to 2 μm) nonmotile cells. These morphological and behavioral changes coordinate with broad changes to the transcriptome. Images are representative of at least three independent experiments for each strain. Bars, 10 μm.
We next examined the swim motility of these strains to determine whether the loss of rffG broadly inhibits flagellum-based motility. We analyzed the motility of the wild-type, ΔrffG, and ΔrffG(prffG) strains through 0.3% LB agar, which permits swimming. The ΔrffG strain, and to a lesser extent the ΔrffG(prffG) strain, was delayed in the initiation of swimming compared to the wild-type strain. However, all strains occupied the full petri dish within 24 h (see Fig. S1A in the supplemental material). We measured cell viability both in liquid (Fig. S1B) and in swarms (Fig. S1C) and found that all populations grew to equivalent densities. Thus, the rffG gene was essential for surface-based swarm motility but not for liquid-based swimming motility or growth.
Given that cells required the rffG gene to engage in surface-based motility, we hypothesized that cells of the ΔrffG strain might fail to progress through stages of swarming, such as increased expression of flhDC-regulated genes, elongation into swarmer cells, or migration across a surface (Fig. 1B). Therefore, we independently assessed the cell morphology of the wild-type, ΔrffG, and ΔrffG(prffG) strains using epifluorescence microscopy under swarm-permissive conditions. To visualize flagellar gene expression, a Venus fluorescent protein reporter was introduced on the chromosomes of each strain background downstream of fliA. The fliA gene encodes the flagellar sigma factor (σ28) and is both directly regulated by FlhD4C2 and highly expressed in swarming cells (36, 37). By 4 h after inoculation onto CM55 agar at 37°C, populations of the wild-type strain contained many short nonmotile and few elongated motile cells (Fig. 1B). After 6 h, elongated cells expressing the fluorescent fliA reporter dominated the inoculum edge and were apparent at the leading edge of the swarms (Fig. 1B). In contrast, most cells of the ΔrffG strain were short and nonmotile at 4 and 6 h; cells appeared modestly shorter than nonelongated wild-type cells (Fig. 1B). Cells of the ΔrffG strain largely did not exhibit fliA-associated fluorescence (Fig. 1B). We confirmed the reduction of flagella by visualizing cells harvested from a swarm using transmission electron microscopy. Cells of the ΔrffG strain were uniformly short and lacked the extended structures present on the wild-type elongated swarmer cells (see Fig. S1D). The general loss of flagellar gene expression and flagellar assembly on surfaces likely reflect differences in gene expression during liquid and surface growth, as the ΔrffG strain grown in liquid can engage in flagellum-dependent swimming motility (Fig. S1A). Notably, some cells of the ΔrffG strain did initiate elongation but often ruptured or divided into short cells before completing elongation (Fig. S1E), indicating a failure to complete swarmer cell elongation. In contrast, cells of the ΔrffG(prffG) strain formed elongated motile cells displaying fliA-associated fluorescence by 6 h (Fig. 1B); this progression was delayed compared to that of the wild-type strain, which is consistent with a partial rescue. Therefore, as cells of the ΔrffG strain failed to increase the expression of flagellar genes and to elongate into swarmer cells upon surface contact, we concluded that the rffG gene was necessary and sufficient for cells to initiate swarmer cell elongation.
The rffG gene is essential for the production of LPS-associated moieties necessary for cell envelope integrity.We considered that the lack of swarmer cell elongation in the ΔrffG strain might be caused either by a physical constraint such as a lack of membrane integrity or by activation of swarm-inhibitory signaling pathways. Therefore, we first examined the membrane composition and integrity of this strain. In P. mirabilis, ECA can exist in many forms: linked to other lipids, in a circularized and soluble form, or surface-exposed and linked to the LPS core in the outer membrane (35, 38–42). Repeated efforts to confirm the presence of ECA via Western blotting with E. coli O14 serum (SSI Diagnostica, Hillerød, Denmark), which is reactive against E. coli-derived ECA (43, 44), were unsuccessful. We instead characterized the overall LPS composition and cell envelope sensitivity to antibiotics. We extracted LPS from surface-grown colonies of the wild-type, ΔrffG, and ΔrffG(prffG) strains and then visualized the LPS-associated moieties using silver stain (45). The observed banding patterns of the wild-type and ΔrffG(prffG) strains were nearly equivalent (Fig. 2A). The banding pattern of the ΔrffG strain, however, lacked a high-molecular-weight smear, and the bands within the putative O-antigen ladder formed double bands instead of a single band (Fig. 2A). We concluded that the rffG gene was essential for the production of full-length and wild-type LPS, specifically, the O antigen and the high-molecular-weight components.
Loss of the rffG gene affects outer membrane structures and cell envelope integrity. (A) LPS was extracted from surface-grown cells of the wild-type, ΔrffG, and ΔrffG(prffG) strains. Samples were run on a 12% SDS-PAGE gel and detected via silver stain. Predicted LPS-associated moieties are labeled on the left. HMW, high molecular weight. Image is representative of at least three independent experiments for each strain. (B) The wild-type and ΔrffG strains were spread onto swarm-permissive agar pads containing 0 or 10 μg/ml carbenicillin (Cb). Shown are cells after 1, 2, or 4 h of incubation on a surface. Images are representative of three independent experiments. Bars,10 μm.
We reasoned that these perturbations to the LPS components might cause broader cell envelope damage in cells of the ΔrffG strain. To target the outer membrane, we measured the resistances to polymyxin B, bile salts, and sodium dodecyl sulfate (SDS). Polymyxin B is thought to bind LPS, and the disruption of LPS biosynthesis genes causes polymyxin B sensitivity in P. mirabilis (25, 46). Bile salts (47) and SDS broadly target membranes through detergent-like effects. For Salmonella enterica, ECA is likely involved in the resistance to bile salts (48); however, a similar role for ECA of P. mirabilis has not been explored. Populations of the wild-type and ΔrffG strains were resistant to fully saturated solutions (50 mg/ml) of polymyxin B (Table 1; see also Fig. S2A) and exhibited reduced growth on 0.5% bile salts (Table 1; Fig. S2B). However, the growth defects of the ΔrffG strain on 0.2% bile salts were more severe than those of the wild-type strain. The ΔrffG strain was reduced in growth and formed small and translucent colonies (Fig. S2B). The ΔrffG strain was also more sensitive to SDS than the wild-type strain; 0.5% SDS was permissive for the growth of the wild-type strain but not for the ΔrffG strain (Table 1; Fig. S2C). As controls, we measured the sensitivity to the non-membrane-targeting antibiotics gentamicin and kanamycin. We found no differences in growth between the wild-type and the ΔrffG strains when grown on gentamicin and kanamycin (Table 1). In sum, the rffG-deficient cells exhibited increased sensitivity to bile salts and SDS but not to polymyxin B, gentamicin, and kanamycin. Therefore, the outer membrane in rffG-deficient cells was compromised in a phenotypically distinct manner than previously studied LPS-deficient P. mirabilis strains.
Antibiotic sensitivity of the wild-type and ΔrffG strains on surfaces
To further interrogate cell envelope integrity, we analyzed cell morphology in response to a subinhibitory concentration (10 μg/ml) of the beta-lactam antibiotic carbenicillin. Within 1 h of growth on carbenicillin-containing CM55 agar at 37°C, we observed that wild-type cells lengthened to tens of microns while remaining uniform in diameter (Fig. 2B). By 4 h of growth, occasional bloating was apparent at midcell (Fig. 2B). By contrast, after 1 h of growth on carbenicillin-containing CM55 agar at 37°C, many cells of the ΔrffG strains remained short; the few elongated cells appeared wider or lemon shaped (Fig. 2B). After 2 h of growth, the population of the ΔrffG strain consisted of elongated bloated cells, including several with triangular protrusions (Fig. 2B). By 4 h of growth, most cells of the ΔrffG strain had ruptured; the remaining cells were several microns long (Fig. 2B). Equivalent results were attained with aztreonam, an inhibitor of the cell division protein FtsI (Fig. S2D). The rffG-deficient cells were therefore more susceptible to cell wall stress and/or cell membrane stress and membrane-targeting detergents, indicating that the composition of the outer membrane in rffG-deficient cells was compromised.
The rffG deficiency induces stress response pathways in cells.The loss of the rffG gene resulted in altered cell envelope structure (Fig. 2A) and stability (Fig. 2B; Fig. S2D), both of which would likely induce stress response pathways. Interestingly, the rffG-deficient cells transiently elongated when artificially driven to expand in length using carbenicillin; therefore, the inhibited elongation in these cells was not purely due to disrupted physical structures of the cell envelope. Such cell envelope stress in P. mirabilis can activate several swarm-inhibitory pathways, such as those controlled by RcsB, RpoE, and RppA/B (22, 24, 25, 30). To identify whether one or more of these pathways was induced in rffG-deficient cells, we performed transcriptome sequencing (RNA-Seq) analysis on cells of the wild-type and ΔrffG strains harvested under swarm-permissive conditions. Two biological replicates were acquired and examined. Short nonmotile cells harvested from swarms of the wild-type strain were used as the control for these experiments.
Three hundred forty-three genes of ∼3,455 protein-coding genes in the P. mirabilis BB2000 genome were expressed at least 4-fold differently (see Fig. S3). One hundred thirty-six genes were decreased in the ΔrffG strain, of which approximately 18% were related to ribosome structure and translation and 20% were directly related to flagellum assembly or chemotaxis (see Table S1). Additional factors known to regulate swarmer cell development and motility also had decreased expression, including umoD at 0.12-fold, umoA at 0.19-fold, and ccm at 0.12-fold (Table 2). Cell envelope-associated genes were also downregulated, e.g., the penicillin-binding protein gene pbpC and the membrane lipid modifying gene ddg at 0.25-fold and 0.22-fold, respectively (Table S1). In contrast, 207 genes were increased, including the virulence-associated MR/P fimbria (200-fold-increased expression of mrpA) and Proteus P-like pili (55.5-fold-increased expression of pmpA) (Table 3; see Table S2). In addition, several genes related to carnitine metabolism were increased, including the transcriptional regulator caiF at 63.6-fold and caiA, fixC, and fixX at 4-fold (Tables 3 and S2). Carnitine can be metabolized, particularly under anaerobic conditions (49–51), and can act as a stress protectant for several bacterial species (52, 53; also reviewed in reference 54). Likewise, ompW was increased ∼33.7-fold along with dcuB, an anaerobic C4-dicarboxylate transporter, at 29.6-fold (Tables 3 and S2). In E. coli, maximal ompW expression is tied to survival in the transition from aerobic to anaerobic growth (55). Therefore, genes for fimbrial production and for metabolism under anaerobic or microaerobic environments were more highly expressed in the rffG-deficient cells; by contrast, genes promoting swarm motility were decreased.
Top 20 genes downregulated in the ΔrffG strain relative to the wild-type strain
Top 20 genes upregulated in the ΔrffG strain relative to the wild-type strain
Notably, three major regulators were expressed much higher in the ΔrffG strain (Table 2): mrpJ at 19.1-fold, rpoS at 8.7-fold, and the RcsB cofactor rcsA at 9.9-fold. Both mrpJ and rcsA were previously shown to contribute to swarm inhibition (6, 7, 34). We observed a partial overlap between differentially regulated genes and the characterized MrpJ regulon (34). However, we found a larger overlap between the genes differentially regulated in the ΔrffG populations and the genes recently characterized as regulated by RcsB in P. mirabilis (Tables 2 and 3) (6, 7). The overlap was especially striking among the downregulated genes. While approximately 8% of upregulated genes overlapped with the Rcs regulon, 24% of downregulated genes overlapped with the Rcs regulon (Fig. S3). RcsB directly represses flhDC, which in turn regulates flagellar and chemotaxis genes, positively regulates paralogous genes of the swarm-inhibitory mrpJ gene (6, 7), and regulates the minCDE genes (6). Therefore, we hypothesized that the Rcs phosphorelay was likely activated in the rffG-deficient cells.
RcsB inhibits swarmer cell elongation and morphology defects of rffG-deficient cells.We hypothesized that RcsB-mediated repression of the swarm transcriptional regulator genes, flhDC, was the primary cause for the loss of swarmer cell elongation and swarm motility of the ΔrffG strain. Therefore, we independently constructed a chromosomal deletion of the rcsB gene in the wild-type and the ΔrffG strains, resulting in the ΔrcsB and ΔrffG ΔrcsB strains, respectively. We also constructed a plasmid for constitutive and increased expression of flhDC (under the lac promoter) and introduced this in trans in the wild-type and the ΔrffG strains, resulting in the wild-type(pflhDC) strain and the ΔrffG(pflhDC) strain, respectively. As RcsB is an upstream repressor of flhDC, we predicted that swarm motility and swarmer cell elongation would be rescued in both the ΔrffG ΔrcsB and the ΔrffG(pflhDC) strains. The wild-type(pflhDC) strain and the ΔrcsB strain were predicted to contain constitutively swarming cells on the basis of equivalent constructs in other P. mirabilis wild-type backgrounds (9, 29). We inoculated all strains on separate swarm-permissive agar plates and analyzed colony expansion over 16 h of growth at 37°C (Fig. 3). As predicted, the wild-type(pflhDC) strain and the ΔrcsB strain expanded across the petri dish (swarm radii of 45 ± 0 mm) by 16 h (Fig. 3). However, neither the ΔrffG ΔrcsB nor ΔrffG(pflhDC) strain fully occupied the plate by 16 h (Fig. 3A). The ΔrffG(pflhDC) strain did reach the edge of the plate by 24 h (swarm radius of 45 ± 0 mm) (see Fig. S4), but the ΔrffG ΔrcsB strain remained constrained toward the center, with a swarm radius of 20.1 mm ± standard deviation of 3.4 mm (see Fig. S4). Extracted LPSs of these strains grown in liquid broth and on surfaces were analyzed. We found that the banding pattern of the ΔrcsB strain was equivalent to that of the wild-type strain (see Fig. S5). Likewise, the banding patterns of the ΔrffG ΔrcsB and the ΔrffG(pflhDC) strains were equivalent to that of the ΔrffG strain (Fig. S5). Thus, neither RcsB nor FlhD4C2 contributed to the production of the rffG-dependent LPS-associated moieties. However, increased expression of flhDC or deletion of rcsB was sufficient to increase swarm motility of rffG-deficient cells, indicating that RcsB-mediated repression of flhDC was sufficient for the swarm inhibition of rffG-deficient cells.
Loss of the rffG gene impacts swarm colony development. Loss of the rffG gene extended the lag phase before swarm colony expansion. Liquid-grown populations of the wild-type(pflhDC), the ΔrffG(pflhDC), the ΔrcsB(pBBR1-NheI), and the ΔrffG ΔrcsB(pBBR1-NheI) strains were normalized on the basis of the OD600 and inoculated onto swarm-permissive plates containing kanamycin for plasmid retention. Swarm radii of visible swarm colony expansions were measured at indicated times; each time point comprises separate plates. Three independent experiments were performed for each strain and time point.
Since the ΔrffG ΔrcsB strain did not have full recovery of swarm motility by 24 h, we hypothesized that the loss of the rcsB gene might affect a pathway separate from that of the FlhD4C2-regulated genes. We integrated the chromosomal fliA-Venus transcriptional reporter into each strain and observed the resultant cells using epifluorescence microscopy under swarm-permissive conditions. After 6 h of growth at 37°C, the wild-type(pflhDC)-derived and the ΔrffG(pflhDC)-derived strains consisted of motile elongated cells with fliA reporter-associated fluorescence (Fig. 4). Likewise, cells in the ΔrcsB strain-derived strain were generally elongated and motile with fliA reporter-associated fluorescence (Fig. 4). Surprisingly, cells of the ΔrffG ΔrcsB strain-derived strain exhibited severe cell shape defects: cells were bloated and uneven in width, forming spheres, tapering at the cell poles, or bulging at midcell (Fig. 4). In addition, several cells of the ΔrffG ΔrcsB strain were forked at the cell pole or branched at midcell, resulting in the formation of more than two cell poles. Similar shape defects were not observed in cells of the ΔrcsB strain-derived population nor have they been reported for previously described P. mirabilis (6, 29) or E. coli (20, 56) ΔrcsB mutant strains. Nonetheless, the elongated cells of the ΔrffG ΔrcsB strain-derived strain exhibited fliA reporter-associated fluorescence and were motile (Fig. 4). In sum, cells of the ΔrffG ΔrcsB strain-derived strain did not retain fidelity of a two-pole, rod-shaped swarmer cell morphology, even though they had increased fliA expression. Thus, RcsB contributed to the suppression of shape defects in rffG-deficient cells. As these defects only arose in the absence of RcsB, we posit this was achieved via an RcsB-dependent and FlhD4C2-independent pathway(s).
Swarm cell elongation of the ΔrffG strain is rescued by increased flhDC expression or deletion of the rcsB gene. Epifluorescence microscopy of the wild-type(pflhDC), the ΔrffG(pflhDC), the ΔrcsB, and the ΔrffG ΔrcsB strains on swarm-permissive agar pads. All strains encode Venus controlled by the promoter for fliA. For populations of the wild-type(pflhDC) strain and the ΔrffG(pflhDC) strain, the agar contained kanamycin for plasmid retention. Images were taken once swarm motility initiated (4 to 6 h on surface). Rolling ball background subtraction on fliA reporter images was conducted using FIJI (63). Arrowheads indicate a cell exhibiting shape and polarity defects. Cropped images are shown on the far right for emphasis on cell shape defects. Left, phase contrast; middle left, Venus expression; middle right, membrane stain; far right, cropped selection of membrane stain image at higher zoom. Images are representative of at least three independent experiments for each strain. Bars, 10 μm.
DISCUSSION
Here crucial insights were elucidated about the role of cell envelope structure and stress sensing in the development of P. mirabilis swarmer cells, specifically regarding the cell envelope biosynthesis gene rffG and the signaling pathways that respond to its absence (Fig. 5). We have shown that the rffG gene was essential for the assembly of a swarm-permissive cell envelope. The loss of the rffG gene resulted in the loss of LPS-associated moieties, the alteration of the O-antigen ladder, and increased sensitivity to antimicrobials that specifically target the cell envelope. On the basis of the RNA-Seq results, cells of the ΔrffG strain entered into a distinctive transcriptional state, resulting in the upregulation of several stress response pathways. While most of the pathways activated by the loss of rffG have yet to be characterized, RcsB-mediated inhibition of flhDC expression was a major regulatory factor in restricting elongation in the ΔrffG strain. The loss of rcsB or overexpression of flhDC rescued swarm motility in the ΔrffG strain. Moreover, an additional role for RcsB in the maintenance of cell shape and polarity during swarmer cell elongation was uncovered, as RcsB served to also maintain the two-pole rod shape of rffG-deficient cells.
Checkpoint model in which rffG deficiency induces multiple stress-associated pathways controlling swarmer cell elongation and cell shape. Loss of the rffG gene induced activity of RcsB, which in turn directly repressed the expression of the flhDC genes that encode the master regulator of flagellum production and swarm motility. We found that RcsB in P. mirabilis was also necessary to maintain cell shape and polarity through yet uncharacterized mechanisms, supporting a role for RcsB as a multifunctional regulator of swarmer cell development and motility. The composition of the cell envelope, including the rffG-dependent moieties, may serve as a developmental checkpoint before engaging in swarmer cell development. Cell elongation and increased flhDC expression are initial steps in swarmer cell development.
The transcriptional state of cells in the ΔrffG strain was characterized by the activation of pathways such as those controlled by the transcriptional regulators RpoS, CaiF, and RcsB. There was increased expression of several fimbrial gene clusters. There was also a notable increase in carnitine metabolism genes, which is associated with growth under anaerobic conditions (50, 51). Many of the identified genes are controlled by MrpJ and oxygen availability (57), raising the possibility that cells of the ΔrffG strain bias toward a more adherent low-oxygen lifestyle. Flagellar and chemotaxis genes had decreased expression in the ΔrffG strain. The disruption of the flagellar pathway was consistent with the loss of swarm motility in the ΔrffG strain. However, the potential mechanisms for inhibiting cell elongation and driving cell shape defects were less apparent in the RNA-Seq data. For example, while RcsB has been implicated in cell elongation via the regulation of minCDE (6, 7), differential regulation of these genes in the ΔrffG strain was not evident. Further research will need to be done to completely categorize the genes differentially regulated in rffG-deficient cells and to fully understand the physiological and behavioral implications of these altered expression levels.
Several questions remain regarding the mechanisms of activation, as well as the downstream activity, of RcsB and MrpJ in the ΔrffG strain. First, Bode et al. recently demonstrated that MrpJ acts as a regulator mediating the transition of cells between swarm motility (MrpJ repressed) and nonmotile adherence (MrpJ induced) similar to RcsB (34). MrpJ and RcsB may positively regulate each other and have overlapping regulons (6, 7, 34), making it difficult to genetically disentangle the contributions of each regulator. Additionally, the perturbation of outer membrane structures appears to be communicated to the Rcs phosphorelay through both RcsF-dependent and -independent pathways in P. mirabilis (22). Whether cell envelope stress of rffG-deficient cells is communicated through the outer membrane-localized RcsF, through the upregulation of RcsA, or through an uncharacterized additional pathway remains to be determined. Also unknown is whether the rffG-dependent LPS-associated moieties communicate to Rcs via the Umo system, as was previously shown for O antigen (22).
Previous research has elucidated how disrupting LPS induces stress response pathways and leads to swarm inhibition. For example, abrogation of the O-antigen structure through the deletion of the O-antigen ligase (waaL) or chain length determinant (wzz) inhibits the activation of flhDC upon surface contact (23). The loss of the sugar-modifying O-antigen biosynthesis genes ugd and galU inhibits swarmer cell elongation and motility (25). The aforementioned genes are implicated in LPS biosynthesis. Here, we propose that P. mirabilis cells require cell envelope structures in addition to LPS for the initiation of swarmer cell elongation (Fig. 5). The rffG-dependent high-molecular-weight LPS-associated moieties are not chemically characterized; we hypothesize that these might consist of LPS-associated ECA or ECA-derived moieties, since the E. coli rffG homologue is needed for the production of the broadly conserved ECA (35). Further research is needed to characterize the structural changes to the outer membrane in rffG-deficient cells, especially as these moieties contribute to the overall membrane integrity on surfaces.
The outer membrane structure also plays a crucial mechanical role in resisting turgor pressure fluctuations associated with cell wall stress, specifically, by beta-lactam drugs (58). Cells of the ΔrffG strain were sensitive to detergent-like membrane-targeting antimicrobials, altogether suggesting that the rffG-dependent moieties are crucial for the outer membrane composition and integrity. One explanation is that these rffG-associated cell envelope defects are caused by pleiotropic effects resulting from disrupting a cell envelope biosynthesis pathway that uses a shared pool of precursor molecules. We raise this possibility because the perturbation of ECA or LPS biosynthesis genes in E. coli causes the accumulation of dead-end intermediates that broadly impact cell envelope integrity (20, 21). However, we instead posit that the cell envelope defects in rffG-deficient cells might be sufficient to sensitize the cells to form defective cell shapes.
We propose that RcsB acts to suppress cell wall defects in rffG-deficient cells as well as potentially other cell-envelope compromised cells (Fig. 5). We observed that the absence of the rffG-dependent moieties did not mechanically restrict swarmer cell elongation or result in the formation of more than two poles in artificially elongated cells constitutively expressing flhDC from a nonnative promoter. Indeed, the populations of the ΔrffG(pflhDC) strain occupied a standard-sized petri dish in 24 h, though these populations exhibited a lower rate of swarm colony expansion than populations of the wild-type(pflhDC) strain. These results suggest that rffG-mediated cell defects may impact swarm motility independent of flhDC regulation. The cells lacking both rffG and rcsB exhibited growth from the midcell and the formation of more two cell poles in addition to other physical perturbations. Thus, although the deletion of rcsB rescues cell elongation and motility through the derepression of flhDC, the absence of RcsB also perturbed a yet uncharacterized morphology-generating pathway critical for the cell shape and integrity of rffG-deficient cells. Others have also proposed that the Rcs phosphorelay plays a conditional role in cell shape maintenance in other bacteria. L-form E. coli cells require the Rcs phosphorelay to recover a rod shape; cells lacking this response rupture (59). The authors of that study proposed that Rcs might function to maintain cell shape under conditions in which cells lose the cell wall through exposure to lysozyme or cationic antimicrobial peptides, including in several niches within a human host (59). Moreover, in E. coli and Agrobacterium tumefaciens, cell polarity defects similar to those of the ΔrffG ΔrcsB strain appear to arise from the formation of patches of inert peptidoglycan and mislocalized division planes (60–64). Further study is needed to mechanistically understand which aspects of the RcsB regulon are specific for cell shape maintenance and how additional poles emerge in cells lacking both RcsB and the rffG-dependent moieties.
It remains unclear how RcsB, which is presumably inactive in swarming cells, can play a role in swarmer cell shape maintenance. We propose two broad mechanisms that may resolve this contradiction. First, the role(s) of RcsB in cell shape maintenance may occur prior to the initiation of swarmer cell elongation. Elongation may exacerbate unrepaired envelope flaws that manifest in cell shape and polarity defects. As such, the Rcs phosphorelay would act as a developmental checkpoint to restrict swarmer cell development under conditions challenging to the cell envelope. Second, RcsB may have multiple states beyond simply “active” and “inactive” that may enable differential activity across time and cell states. The DNA-binding activity of RcsB has been shown to be modulated by both phosphorylation and the association with auxiliary transcription factors in E. coli (65). How RcsB activity is modulated downstream of phosphorylation and the association with potential auxiliary transcription factors remain unknown in P. mirabilis.
Altogether, we propose that cell envelope stress, such as the absence of rffG-dependent moieties, functions as a developmental checkpoint before swarmer cell elongation and increased flagellar gene expression (Fig. 5). Under swarm-permissive conditions in the presence of a wild-type cell envelope structure, the Rcs phosphorelay, along with other stress-sensing pathways, are inactive, thereby enabling the swarmer development to progress. When the cell envelope is perturbed, we posit that the activity of cell envelope stress response sensors culminates in the adaptation of an adherence-promoting lifestyle that may provide protection against external stressors. As the Rcs phosphorelay, flhDC, and rffG, among other discussed genes, are conserved among the Enterobacteriaceae family, we predict these factors may broadly serve to modulate bacterial swarm motility and potentially cell development.
MATERIALS AND METHODS
Growth conditions.Liquid cultures were grown in Lennox broth (LB). Colonies were grown in 0.3% LB agar for swimming motility assays, on low-swarm (LSW−) agar (66) for plating nonmotile colonies, and on CM55 blood agar base (Oxoid, Hampshire, UK) for swarming. The antibiotics for selection used throughout all assays were as follows: 15 μg/ml tetracycline (Amresco Biochemicals, Solon, OH), 25 μg/ml streptomycin (Sigma-Aldrich, St. Louis, MO), and 35 μg/ml kanamycin (Corning, Corning, NY). For swarm assays, overnight cultures were normalized to an optical density at 600 nm (OD600) of 1.0, and 1 μl of culture was inoculated with a needle onto swarm-permissive CM55 blood agar base (Oxoid, Hampshire, UK) plates containing 40 μg/ml Congo red, 20 μg/ml Coomassie blue, and kanamycin (Corning, Corning, NY) as needed. Plates were incubated at 37°C. When indicated, we used strains carrying an empty vector (pBBR1-NheI [67]) to confer kanamycin resistance. Images were taken with a Canon EOS 60D camera.
Strain construction.Strain construction was performed as described previously (68). The strains and plasmids are listed in Table 4. All plasmids were confirmed by Sanger sequencing (Genewiz, South Plainfield, NJ). For all strains, expression plasmids were introduced into P. mirabilis via E. coli SM10λpir as previously described (66). The resultant strains were confirmed by PCR of the targeted region. The ΔrffG strain was additionally confirmed through whole-genome sequencing as described in reference 69.
Strains and plasmids used in this study
For the construction of the ΔrcsB strain, a gBlock (Integrated DNA Technologies, Coralville, IA) containing the 452 bp upstream and downstream of rcsB (P. mirabilis BB2000, accession number CP004022; nucleotides [nt] 1972701 to 1973153 and 1973809 to 1974261) was generated and introduced to pKNG101 at SpeI and XmaI sites using SLiCE (70). Similarly, for the construction of the ΔrffG strain, a gBlock containing a chloramphenicol resistance cassette (amplified from pBAD33 [71]) flanked by 1,000 bp upstream of rffG and downstream of rffG (P. mirabilis BB2000, accession number CP004022; nt 3635400 to 3636399 and 3637473 to 3638473) (Integrated DNA Technologies, Coralville, IA) was introduced to pKNG101 (68) at the same sites. For the construction of fliA reporter strains, a gBlock encoding the last 500 bp of fliA (P. mirabilis BB2000, accession number CP004022; nt 1856328 to 1856828), ribosomal binding site (RBS; AGGAGG), a modified variant of Venus fluorescent protein (a gift from Enrique Balleza and Philippe Cluzel [72]), and 500 bp downstream of fliA (P. mirabilis BB2000, accession number CP004022; nt 1855828 to 1856328) (Integrated DNA Technologies, Coralville, IA) was inserted into pKNG101 (68) at the ApaI and XbaI sites. For the construction of the rffG expression strains, the nucleotide sequence for rffG was amplified via PCR from the P. mirabilis chromosome using oKL273 and oKL274 and inserted into expression vector pBBR1-NheI (67) using AgeI and NheI restriction enzyme sites. For the flhDC expression strains, the nucleotide sequence of flhDC (amplified using oKL277 and oKL278) was similarly inserted into pBBR1-NheI. A gBlock containing the lac promoter (Integrated DNA Technologies, Coralville, IA) was introduced upstream of the coding region using SLiCE (70).
LPS extraction and analysis.Cells were grown overnight at 37°C on swarm-permissive CM55 (Oxoid, Hampshire, UK) agar with antibiotics as needed and harvested with LB. LPS from cells was extracted using an LPS extraction kit according to the manufacturer's instructions (iNtRON Biotechnology Inc., Seongnam, Gyeonggi, South Korea). Extracts were resuspended in 10 mM Tris buffer (pH 8.0) and run on a 12% SDS-PAGE gel. The gels were stained with a modified silver stain protocol (45).
Halo assays (MIC determination).Cultures were top-spread on LSW− medium and allowed to sit on the benchtop until the surface appeared dry (a couple of hours). Six-millimeter sterile filter disks were placed on the plates and soaked with 10 μl of dilutions containing polymyxin B (Sigma-Aldrich, St. Louis, MO), gentamicin (Calbiochem, San Diego, CA), or kanamycin (Corning, Corning, NY) in water. A water-alone control was included. Once the filter disks dried (a couple of hours), the plates were incubated at 37°C overnight and imaged.
Resistance assays to bile salts and sodium dodecyl sulfate.Cultures were grown overnight, normalized to an OD of 1.0, and serially diluted 10-fold. One-microliter spots of 10−1 to 10−8 dilutions of each strain (in technical triplicates) were inoculated onto LSW− plates containing bile salts (Sigma-Aldrich, St. Louis, MO) or sodium dodecyl sulfate (Sigma-Aldrich, St. Louis, MO) (filter sterile, added to medium after autoclaving). The plates were incubated overnight at 37°C.
Carbenicillin sensitivity assay.Cells were harvested from swarm-permissive medium using LB and spread onto 1-mm CM55 (Oxoid, Hampshire, UK) agar pads containing 10 μg/ml carbenicillin (Corning, Corning, NY). The cells were imaged after 1, 2, or 4 h on the surface. Three biological replicates were analyzed.
Microscopy.Microscopy was performed as previously described (69). Briefly, CM55 (Oxoid, Hampshire, UK) agar pads, supplemented as needed with 35 μg/ml kanamycin for plasmid retention, were inoculated from overnight stationary cultures and incubated at 37°C in a modified humidity chamber. The pads were imaged using a Leica DM5500B (Leica Microsystems, Buffalo Grove, IL) and a CoolSnap HQ2 cooled charge-coupled-device (CCD) camera (Photometrics, Tucson, AZ). MetaMorph version 7.8.0.0 (Molecular Devices, Sunnyvale, CA) was used for image acquisition. The images were analyzed using FIJI (73) (National Institutes of Health, USA); where indicated, the images were subjected to background subtraction equally across the entire image. Where indicated, the cells were stained with 25 μM 1-(4-trimethylammoniumphenyl)-6-phenyl-1,3,5-hexatriene p-toluenesulfonate (TMA-DPH; Invitrogen, Carlsbad, CA) (maximum excitation, 355 nm; maximum emission, 430 nm) imaged in the DAPI (4′,6-diamidino-2-phenylindole) channel using an A4 filter cube (excitation, 360/40 nm; emission, 470/40 nm) (Leica Microsystems, Buffalo Grove, IL). Venus (maximum excitation, 515 nm; maximum emission, 528 nm) was visualized in the green fluorescent protein (GFP) channel using a GFP ET filter cube (excitation, 470/40 nm; emission, 525/50 nm [Leica Microsystems, Buffalo Grove, IL]). The fluorescence intensity and exposure time for each fluorescence channel were equivalent across all fluorescence microscopy experiments. Fluorescence due to Venus was not quantified and was visible due to the overlapping excitation and emission spectra with the GFP ET filter cube.
Transcriptional analysis.Strains were grown on CM55 (Oxoid, Hampshire, UK) plates at 37°C. For wild-type samples, the colonies were inoculated on CM55 (Oxoid, Hampshire) agar and incubated overnight for swarm development. The presence of short nonmotile cells in consolidation phase was confirmed by light microscopy. Wild-type cells from the swarm edge were then harvested by scraping with a plastic loop into 1 ml of RNA Protect solution (Qiagen, Venlo, Netherlands). Samples of the ΔrffG strain were harvested after an overnight incubation by scraping whole colonies into 1 ml RNA Protect solution. Total RNA was isolated using an RNeasy minikit (Qiagen, Venlo, Netherlands) according to the manufacturer's instructions. RNA purity was measured using an Agilent 2200 Tapestation (Agilent, Santa Clara, CA). To enrich mRNA, rRNA was digested using terminator 5′ phosphate-dependent exonuclease (Illumina, San Diego, CA) according to the manufacturer's instructions and purified by phenol-chloroform extraction (74). cDNA libraries were prepared from mRNA-enriched RNA samples using an NEBNext Ultra RNA library preparation kit (New England BioLabs, Ipswich, MA) according to the manufacturer's instructions. The libraries were sequenced on an Illumina NextSeq 2500 instrument with 250-base pair single-end reads at the Harvard University Bauer Core. Sequences were matched to the BB2000 reference genome (accession number CP004022) using TopHat 2 (75). Differential expression data were generated using the Cufflinks RNA-Seq analysis suite (76) run on the Harvard Odyssey cluster, courtesy of the Harvard University Research Computing Group. The data were analyzed using the CummeRbund package for R and Microsoft Excel (76). Bioinformatics information was derived from KEGG (77). The data in this paper represent the combined analyses of two independent biological repeats.
Accession number(s).Data for this study have been deposited in the GEO database under accession no. GSE116587. Accession numbers for individual data sets are GSM3243501, GSM2343502, GSM3243503, and GSM3243504.
ACKNOWLEDGMENTS
We thank Enrique Balleza and Philippe Cluzel for the kind gift of the Venus construct and members of the Gibbs, D'Souza, Gaudet, and Losick labs for thoughtful discussion and feedback. Doug Richardson, Sven Terclavers, and Sebastian Gliem assisted with imaging on the Cell Observer. We thank the staff of the Bauer Core at Harvard University for assistance with RNA-Seq and Maria Ericsson of the Harvard Medical School Electron Microscopy facility for assistance with transmission electron microscopy (supplemental material).
This research was funded by a Smith Family Graduate Fellowship in Science and Engineering, a Simmons Family Award for the Harvard Center for Biological Imaging, the David and Lucile Packard Foundation, The George W. Merck Fund, and Harvard University.
We declare no conflicts of interest.
K.L. and K.A.G. conceived and coordinated the study and wrote the paper. K.L. performed all experiments, except the RNA-Seq, which was performed by M.J.T. All authors edited the paper.
FOOTNOTES
- Received 21 April 2018.
- Accepted 27 June 2018.
- Accepted manuscript posted online 2 July 2018.
- Address correspondence to Karine A. Gibbs, kagibbs{at}mcb.harvard.edu.
Citation Little K, Tipping MJ, Gibbs KA. 2018. Swarmer cell development of the bacterium Proteus mirabilis requires the conserved enterobacterial common antigen biosynthesis gene rffG. J Bacteriol 200:e00230-18. https://doi.org/10.1128/JB.00230-18.
Supplemental material for this article may be found at https://doi.org/10.1128/JB.00230-18.
REFERENCES
- Copyright © 2018 American Society for Microbiology.
