Flagellum-Mediated Mechanosensing and RflP Control Motility State of Pathogenic Escherichia coli.

Flagella and motility are widespread virulence factors among pathogenic bacteria. Motility enhances the initial host colonization, but the flagellum is a major antigen targeted by the host immune system. Here, we demonstrate that pathogenic E. coli strains employ a mechanosensory function of the flagellar motor to activate flagellar expression under high loads, while repressing it in liquid culture. We hypothesize that this mechanism allows pathogenic E. coli to regulate its motility dependent on the stage of infection, activating flagellar expression upon initial contact with the host epithelium, when motility is beneficial, but reducing it within the host to delay the immune response.

required for the expression of class III genes, which include those for the outer part of flagella and components of the chemotaxis pathway.
Flagella and motility have several functions in bacterial virulence (3,4). In the early phase of infection, motility is important for moving through the mucus layer and establishing initial contact with the host epithelium, and flagella can serve as adhesins. However, flagella are also highly immunogenic, being rapidly recognized by the innate immune system (5). Known strategies to avoid this immune response used by such pathogens as Salmonella, Campylobacter, and Helicobacter rely on stochastic variation in the expression of flagellar genes, due either to genetic changes (6)(7)(8)(9)(10) or to changes at the transcriptional level (11)(12)(13). The latter regulation in Salmonella is mediated by RflP (formerly called YdiV) (14), which represses fliA expression by sequestering the FlhDC complex and targeting it for degradation in a subpopulation of cells (15,16). Expression of RflP and hence of the fraction on nonmotile cells in the population depends on nutrient levels and cell envelope stress (15,17).
Although flagellar systems of E. coli and Salmonella and their regulation are highly similar, only a few E. coli strains exhibit genetic phase variation in flagellar expression (18,19). Furthermore, although the rflP gene is present in E. coli, it is inactive in laboratory strains because of attenuated translation (20). Thus, it remained unclear what mechanisms might be used by most pathogenic E. coli strains to control motility under conditions that resemble those encountered during infection.
In this work, we investigated the motility of several commensal and pathogenic strains of E. coli grown in liquid or in soft agar. We showed that pathogenic, but not commensal, E. coli strains activate motility only in a porous medium. The underlying regulation requires flagellar rotation as well as expression of RflP, and it is likely to represent a novel mechanism of motility-dependent mechanosensing.
Pathogenic isolates of E. coli downregulate flagellar gene expression in liquid. We investigated the motility of a number of pathogenic, as well as a few commensal, natural isolates of E. coli obtained from the Leibniz Institute's DSMZ collection (Table 1 and see Table S1 in the supplemental material). Surprisingly, while most tested strains (17 out of 19) appeared nonmotile or poorly motile when cultivated in liquid tryptone broth (TB), they became motile on soft (0.27% agar) TB (TBA 0.27%) motility plates, where E. coli cells can spread by swimming through agar pores (Table 1; Fig. 1A and B).  U5/41  UPEC  --DSM 50902  NA  ϩ  -QST 40139  Possibly commensal strain  --AMC 198 NA Newborn with meningitis Such dependence of motility on growth in agar was confirmed by reinoculation of liquid TB cultures from TBA 0.27% plates and vice versa. Interestingly, all E. coli strains following this motility pattern were implicated in pathogenicity, whereas both tested commensal strains from the DSMZ collection and the common laboratory strain MG1655 were motile under both conditions. For further analysis of the underlying regulatory mechanism, we transformed four pathogenic and two commensal E. coli strains with plasmids carrying green fluorescent protein (GFP) reporters for all three classes of promoters in a flagellar gene expression cascade. Whereas the activities of the flhD (class I) reporters were similar between cells grown in TBA 0.27% and in liquid TB for all strains (Fig. 1C to H and Fig. S1A to F), promoters of fliA (class II/class III) and fliC (flagellin, class III) were strongly downregulated in liquid TB in pathogenic strains (Fig. 1C to F and Fig. S1A to D). In contrast, no changes in flagellar gene expression were observed in commensal strains ( Fig. 1G and H and Fig. S1E and F), consistent with their unchanged motility in liquid culture. Together, these results indicate the existence of a lifestyle-specific motility control in E. coli acting at the level of class II flagellar gene expression.
Rotation of the flagellar motor under a load is required for activation of flagellar gene expression in pathogenic E. coli. Since the most apparent difference between growth in liquid and that in soft agar is the porous structure of agar, which interferes with flagellar rotation, we hypothesized that an increased mechanical load on the flagellar motor might signal flagellar gene expression in pathogenic E. coli. Such mechanosensing has indeed been described in several bacterial species (21,22), although its molecular mechanisms remain poorly understood. In E. coli, the load is known to cause posttranslational remodeling of the flagellar motor, increasing the number of force-generating stator units (23), but no gene-regulatory mechanosensing mediated by flagella has been reported so far.
To test our hypothesis, we made E. coli Z36 knockout strains that lack flagellar filaments and therefore have largely reduced motor loads (ΔfliC mutant) or have paralyzed flagella due to the absence of the MotA stator protein (ΔmotA mutant). These strains were locked in the motility-off state at the level of gene expression in both liquid and agar ( Fig. 2A to C; Fig. S2A to C), suggesting that in pathogenic E. coli, rotation of the flagellar motor under a load is indeed perceived as a signal activating the expression of class II and class III flagellar genes. These knockouts could be complemented for both motility and gene expression when respective proteins were produced from a plasmid (Fig. S3).
Besides swimming in porous (such as agar) or viscous media, the mechanical load on the flagellar motor is also expected to increase when bacteria swim toward or attach to a surface. Such surface sensing may indeed be the major function of flagellummediated mechanosensing (21,24). Consistently, we observed that the expression of fliC became elevated when wild-type (WT) E. coli Z36 cells were either forced to the surface by centrifugation (Fig. 2H) or grown on the surface of a semisolid (0.5%) agar (Fig. S4). No fliC gene induction was observed in ΔmotA cells under these conditions, suggesting that induction is mediated by the same mechanosensing mechanism. Notably, when cells were grown on a 0.5% agar surface, the flhD promoter activity of ΔmotA cells was the same as that of WT E. coli Z36 cells (Fig. S4). This confirms that motA deletion has no direct effect on flhD activity and suggests that the lower expression of flhD in nonmotile knockout strains grown on TBA 0.27% than in TB ( Fig. 2C and D) was rather due to a high cell density and a corresponding nutrient depletion within the nonspreading colony.

Anti-FlhC 2 FlhD 4 factor RflP is involved in mechanosensing.
Since fliA and fliC are known to be repressed by RflP in Salmonella, we tested the effect of rflP deletion on the observed regulation. Unlike in the laboratory E. coli strains where RflP had no effect on flagellar gene expression (20), rflP deletion in E. coli Z36 resulted in strong derepression of fliA and fliC expression, particularly apparent in liquid TB medium ( Fig. 2D and Fig. S2D). Consequently, most E. coli Z36 ΔrflP cells remained motile in liquid (Fig. 2G). This effect could be complemented by expressing RflP from the plasmid (Fig. S5). Deletion of rflP also derepressed the flagellar gene expression in the ΔmotA and ΔfliC backgrounds, both in liquid and on TBA 0.27% (Fig. 2E and F and Fig. S2E and F). Although fliA and particularly fliC expression was still weakly upregulated on soft agar even in E. coli Z36 ΔrflP, it remains unclear whether this residual RflP-independent regulation is due to mechanosensing, since to some extent it was also observed in ΔrflP ΔmotA and ΔrflP ΔfliC double knockouts ( Fig. 2E and F and Fig. S2E and F).
The interplay between RflP-mediated repression and mechanosensing remains to be elucidated (see below). Nevertheless, we observed that the activity of the rflP promoter was independent of the cultivation conditions (Fig. S6), suggesting that RflP is not regulated by mechanosensing at the transcriptional level. This is in contrast to the transcriptional regulation of rflP by cell envelope stress in Salmonella (17).
Importantly, our results also indicate that the determinants of mechanosensing other than RflP are present in E. coli MG1655. The overexpression of rflP not only led to the expected dose-dependent inhibition of fliC expression in MG1655 (Fig. 3A) but also apparently restored mechanosensing in this laboratory strain, with elevated fliC expression in soft agar compared to that in liquid (Fig. 3B). Since the inhibition of fliC expression by RflP was only partial in MG1655, this increase was particularly pronounced at the level of mean reporter activity rather than as a fraction of positive cells.
Proteome regulation by mechanosensing. In order to identify nonflagellar proteins that could be regulated by mechanosensing, we performed comparative wholeproteome analysis of WT E. coli Z36 and ΔmotA cells grown either on 0.5% agar or in liquid TB medium. Apart from flagellar motor proteins, several metabolic enzymes and uncharacterized proteins were upregulated in a motA-dependent fashion on the agar surface, whereas several transporters and metabolic enzymes, as well as the sensor kinase QseC, were downregulated (Table 2). Since RflP could not be detected in these samples, its regulation at the protein level could not be assessed. Moreover, the expression of many other proteins, some of them involved in the acid stress response, was upregulated on the surface in a motA-independent manner (Table S2). Since QseC, a member of the QseB QseC two-component regulatory system, was among the most strongly downregulated proteins and it was previously shown to play a role in the regulation of flagellar expression (25,26), we tested whether this system might have a regulatory role in mechanosensing. However, no effect of qseC deletion on fliC expression could be observed in E. coli Z36, grown either on soft agar or in TB (Fig. S7), suggesting that under our conditions and in this strain, this two-component system is inactive. Thus, QseC does not seem to have a mechanosensory role. Nevertheless, its mechanosensitive regulation by pathogenic E. coli might be important during infection, since QseC was shown to be important for virulence (26,27).
In this study, we demonstrate that the expression of flagellar genes in a number of pathogenic E. coli strains is inhibited in liquid but that their expression is induced when bacteria are grown either in porous medium or on a surface. Because this regulation was not observed in commensal or laboratory strains of E. coli, it likely represents a specific mechanism to avoid the adaptive immune response of the host, akin to genetic or transcriptional variation in the expression of flagellar genes observed in other  (6)(7)(8)(9)(10)(11)(12)28). However, unlike with the stochastic antigenic variation or transcriptional switching observed in these bacteria, the control of flagellar gene expression in pathogenic E. coli apparently relies on sensing the mechanical load on the flagellar motor.
Although surface sensing-mediated control of gene expression, including control of bacterial flagellation, was reported before, most notably in Vibrio, Caulobacter, and Bacillus species (24,(29)(30)(31)(32)(33)(34)(35), this is, to our knowledge, the first example of such regulation in E. coli. It was previously shown that enterohemorrhagic E. coli (EHEC) can regulate the expression of EHEC-specific virulence genes upon sensing host attachment and fluid shear (36), but this surface sensing is not known to rely on flagella or to regulate motility. Moreover, this gene-regulatory mechanosensing controls motility in a different way than was previously described for posttranslational remodeling of the E. coli flagellar motor under a mechanical load (23,37).
We hypothesize that such motility-dependent mechanosensing allows pathogenic E. coli to express flagella upon contact with epithelial cells or in mucus when motility is beneficial for the initial attachment to the host but that pathogenic E. coli shuts down the expression of flagellar genes upon entry into the host in order to delay detection by the immune system. In line with apparent trade-offs of flagellin expression in pathogenic E. coli, we also observed bimodal flagellar expression in these strains, with a subpopulation of cells remaining in the flagellum-off state even in soft agar. A similar bimodal expression was previously observed in other bacteria, including Salmonella  16) and Bacillus subtilis (38), but whereas in Salmonella bimodal fliC expression is known to depend on RflP activity (16,17), deletion of rflP in E. coli Z36 did not abolish the bimodality of fliC expression, suggesting a different origin of bimodality. Though the underlying mechanism of signal transduction from the flagellar motor to transcriptional regulation remains to be established, it apparently requires flagellar rotation under a load. The motility-off state of gene expression was observed not only for the filament-deficient ΔfliC strain of E. coli that cannot efficiently sense the load but also for the ΔmotA strain, which has immobilized flagella. This contrasts with observations of B. subtilis and Vibrio parahaemolyticus; these organisms with deletions of motor proteins mimic strains under high loads (30,32), but their phenotype is similar to the lock-off mechanosensing phenotype of mot mutants of Caulobacter crescentus and Pseudomonas aeruginosa (24,39). a Proteins were selected from the whole-proteome data based on the following criteria: differential expression between the WT and the ΔmotA mutant (fold change, Ն3) on TBA 0.5%, differential expression between the WT on TBA 0.5% (WT surf ) and in liquid TB (WT liq ), and little or no difference (fold change, Ͻ3) between the ΔmotA mutant on TBA 0.5% (ΔmotA surf ) and in liquid TB (ΔmotA liq ). The E. coli CFT073 protein annotation database was used as a reference. CoA, coenzyme A; HTH, helix turn helix; FAD, flavin adenine dinucleotide.
Our results demonstrate that mechanosensing occurs in pathogenic E. coli strains due to the activity of the motility repressor RflP, which in these strains inhibits the expression of flagellar genes in liquid. Consistently, artificial upregulation of RflP in E. coli K-12 unmasks its mechanosensing response, demonstrating that other sensory determinants are present in the laboratory strain. While at this stage it cannot be ruled out that RflP-stimulated proteolysis of the FlhDC complex is required only to generally lower the basal expression of flagellar genes, its involvement in mechanosensing might also be direct, as occurred with a previously reported surface contact-dependent relief of proteolysis of a master flagellar regulator in B. subtilis (31).
Bacterial strains and culture conditions. All strains and plasmids used in this study are listed in Table S1. E. coli strains were grown in liquid tryptone broth (TB) medium (10 g tryptone and 5 g NaCl per liter) or in lysogeny broth (LB) medium (10 g tryptone, 10 g NaCl, and 5 g yeast extract per liter) supplemented with antibiotics where necessary.
Cloning and knockout strain construction. In order to construct the rflP-gfp fluorescent reporter, the bp -253-to-bp ϩ90 promoter region of the rflP gene was amplified from the E. coli Z36 chromosome and cloned into the pUA66 vector using XhoI and BamHI restriction sites (40). For rflP-FLAG plasmid construction, the rflP gene sequence without a stop codon was amplified from the E. coli Z36 chromosome with a FLAG coding sequence, with the stop codon being added via a primer overhang. The resulting fusion construct was cloned into the pTrc99A vector using SacI and XbaI restriction sites (41).
All the knockout strains described in this work were obtained using the lambda red recombination system as described previously (42). Where necessary, kanamycin cassettes were flipped out using FLP-FLP recombination target (FRT) recombination (43).
Motility assay in soft agar. Single colonies of different E. coli strains were transferred onto the surfaces of soft agar plates (TB containing 0.27% agar), and the motile behavior of E. coli strains was assessed after 6 to 8 h of incubation at 37°C by analyzing the motility halo diameter. The cells from the edges of such halos were subsequently transferred into 3 ml of fresh TB medium and grown with shaking at 37°C to an optical density at 600 nm (OD 600 ) of 0.5 to 0.6. The motility of the cells was then analyzed using phase-contrast microscopy.
Single-bacterium movement tracking. E. coli MG1655, Z36, and Z36 ΔrflP overnight cultures were diluted in fresh TB and grown at 37°C with shaking to the OD 600 of 0.6. The samples were subsequently diluted to final OD 600 s of 0.05 to 0.1 in TB and loaded into ibidi channels (-Slide Chemotaxis 3D ; ibidi GmbH, Germany). Movies were recorded at ϫ10 magnification with a phase-contrast microscope at a rate of 20 frames per second using a complementary metal-oxide semiconductor (CMOS) camera (exposure time, 1 ms), normalized by dividing each frame by an image of the background illumination under the same conditions and inverted in order for the cells to appear bright. A projection of the temporal maximum of each pixel was computed to produce an image of the cell trajectories.
Flow cytometry. Promoter activities of the flagellar genes flhD, fliA, and fliC and the motility regulator rflP were assayed using plasmid-based reporters containing the respective promoter regions fused to gfp (Table S1). Promoter activities were analyzed in the cells collected from the edges of motility halos formed on 0.27% TBA plates and after subsequent cultivation in liquid TB medium. The samples were diluted in 2 ml tethering buffer (10 mM KH 2 PO 4 , 100 M EDTA, 1 M L-methionine, and 10 mM sodium lactate, pH 7.0), and fluorescence was measured with a BD LSRFortessa SORP cell analyzer (BD Biosciences, Germany).
In order to compare PfliC-gfp activity in attached and planktonic cells, 300 l of WT Z36 and ΔmotA mutant day cultures in TB (OD 600 ϭ 0.6, diluted in fresh TB to the final OD 600 of 0.05) were inoculated into the wells of a 96-well microtiter plate (ibidi, Germany), and the cells were centrifuged for 45 min (37°C, 4,000 rpm). The supernatant with nonattached planktonic cells was transferred into new wells. The wells with attached cells were washed three times with 500 l tethering buffer, and the presence of attached cells was verified with phase-contrast microscopy. Fresh TB (300 l) was subsequently added to the wells, and the plates were incubated at 37°C. The measurements of PfliC-gfp activity in attached and planktonic cells were performed with flow cytometry after 60 and 90 min of incubation.
Proteomic analysis. To compare the proteomes of WT E. coli Z36 and ΔmotA strains grown on the surfaces of soft agar plates and in the liquid medium, the cells were grown on the surface of 0.5% TBA for 4 h at 37°C. Flagellar gene expression was confirmed with flow cytometry using PfliC-gfp promoter fusion activity as a readout. The cells were then washed off the plates with 2 ml cold phosphate-buffered saline (PBS) (8 g NaCl, 0.2 g KCl, 1.42 g Na 2 HPO 4 , 0.27 g KH 2 PO 4 per liter), and the final sample size was adjusted to 5 ml with an OD 600 of 0.6. The cells were then washed twice in 5 ml cold PBS, and the pellets were stored at -20°C. Before the samples were frozen, 50 l of the suspension was reinoculated into 5 ml of liquid TB and grown at 37°C with shaking to an OD 600 of 0.6. The cells were then washed twice in 5 ml cold PBS, and the pellets were stored at -20°C. The collected samples were subsequently subjected to proteomic analysis using mass spectrometry and subsequent data analysis as described before (44,45). Although no reference database for E. coli Z36 is available, using the E. coli CFT073 protein database resulted in full coverage of the analyzed proteome.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.