Peptidoglycan maturation enzymes affect flagellar functionality in bacteria

Authors


For correspondence. E-mail bonecai@pasteur.fr; Tel. (+33) 1 4438 9516; Fax (+33) 1 4061 3640.

Summary

The flagellar machinery is a highly complex organelle composed of a free rotating flagellum and a fixed stator that converts energy into movement. The assembly of the flagella and the stator requires interactions with the peptidoglycan layer through which the organelle has to pass for externalization. Lytic transglycosylases are peptidoglycan degrading enzymes that cleave the sugar backbone of peptidoglycan layer. We show that an endogenous lytic transglycosylase is required for full motility of Helicobacter pylori and colonization of the gastric mucosa. Deficiency of motility resulted from a paralysed phenotype implying an altered ability to generate flagellar rotation. Similarly, another Gram-negative pathogen Salmonella typhimurium and the Gram-positive pathogen Listeria monocytogenes required the activity of lytic transglycosylases, Slt or MltC, and a glucosaminidase (Auto), respectively, for full motility. Furthermore, we show that in absence of the appropriate lytic transglycosylase, the flagellar motor protein MotB from H. pylori does not localize properly to the bacterial pole. We present a new model involving the maturation of the surrounding peptidoglycan for the proper anchoring and functionality of the flagellar motor.

Introduction

Motility is an important virulence property for many pathogenic bacteria. It allows them to reach their specific niche, escape from hostile conditions and access nutrients. A good example of this is Helicobacter pylori. This Gram-negative bacterium is the etiological agent of peptic ulcer, gastric adenocarcinoma and gastric mucosa-associated lymphoid tissue lymphoma. It colonizes 50% of the world population, and inhabits one of the most hostile niches, the human gastric mucosa. To survive in such an environment, H. pylori is highly motile due to its 2–6 unipolar flagella, making motility an essential factor for H. pylori colonization in its niche (Eaton et al., 1989; 1996).

Bacterial flagella are surface appendages that extend from the cytoplasm to the extracellular space, consequently crossing two membranes and the periplasmic space containing the peptidoglycan (PG) layer. The field of flagellar biology has made important advances regarding the processes involved in assembly of the flagellar apparatus, gene hierarchy and regulation. However, few studies have addressed the role of the PG layer on the functionality of the apparatus. In Enterobacteria, type III secretion systems have been known to carry a dedicated lytic transglycosylase (LTG) of the FlgJ family. In Salmonella enterica serovar typhimurium, FlgJ has been shown to be necessary for the assembly of the flagellar complex across the PG (Nambu et al., 1999; Hirano et al., 2001). This protein is bimodular and consists of a chaperonin N-terminal domain and a C-terminal domain carrying a muramidase activity responsible for making a hole in the PG layer by degrading the sugar backbone composed of glycan chains (Hirano et al., 2001). Interestingly, the FlgJ homologue of H. pylori, HP0245, only carries the N-terminal chaperonin domain and completely lacks the muramidase-like domain (Nambu et al., 2006). From genome analyses, H. pylori does not seem to have a dedicated PG LTG for flagellum assembly (Tomb et al., 1997). In contrast, H. pylori has two ubiquitous PG cleaving enzymes carrying muramidase-like activity, the LTGs Slt and MltD (Chaput et al., 2007), raising the question whether these could fulfil important functions in flagellar motility. LTGs are proteins responsible for the cleavage of glycan chains, resulting in the generation of anhydro residues at the end of ‘newly-matured’ glycan chains (Vollmer et al., 2008). They can be either soluble or membrane-bound, and have an exo- or an endo-activity. Their number varies depending on the bacterial species: there are 2 in H. pylori, while so far 7 (Slt and MltA to F) have been described in Escherichia coli. Blackburn and Clarke in 2001 defined 4 families, based on consensus sequences, which could suggest some differences in function and substrate recognition (Blackburn and Clarke, 2001). Their roles are variable, from PG metabolism (bacterial elongation, cell division, cell turnover, etc.) to the insertion of secretion systems (flagella, pili, conjugation systems, etc., reviewed in Koraimann, 2003). Their potentially lethal activities for the host cell require a tight regulation of their activities. These can be regulated by specific localization and incorporation in PG assembly complexes (Vollmer and Bertsche, 2008), modifications of the chemical structure of the PG (O-acetylation; Moynihan and Clarke, 2011), or, by small inhibitory proteins such as the Ivy family (Clarke et al., 2010).

During flagellum assembly, in particular, of the basal body, it is believed that extensive modifications of the PG layer occur to accommodate the passage of the secretion apparatus. However, once the flagellum is assembled, remodelling of the PG should not affect its structure. In contrast, the stator, composed of MotA and MotB proteins, is also predicted to interact with the PG layer. MotA and B constitute the stator of the flagella, a nanomachine able to transmit energy to the switch structure, and allow the rotation of the flagellum. MotB has been proposed to provide the anchoring of the stator to the PG layer (De Mot and Vanderleyden, 1994; Koebnik, 1995) through its C-terminal domain. A recent crystal structure of this C-terminal domain of MotB of H. pylori revealed the presence of a sugar in the binding site suggesting that MotB could directly bind to PG (Roujeinikova, 2008). However, the exact purpose for MotB anchoring to the PG layer remains mysterious. It has been shown that the MotA/MotB complex switches from an inactive (and apparently plugged) to an active (and unplugged) state (Hosking et al., 2006; Kojima et al., 2009). Thus, MotB binding to the PG layer would facilitate the transition from an inactive to an active MotA/MotB complex. Very recently, a study suggested that FliI, a protein composing the P-ring that localizes to the periplasmic space and interacts with the PG layer, also provides some non-covalent interaction for the correct MotB positioning relative to the flagellum (Hizukuri et al., 2010). Alternatively, since the MotA/MotB complex is in close proximity to the integral membrane MS ring, during assembly of the stator, the membrane environment of the MS ring could be drastically changed with MotA replacing the phospholipids. The MotB anchoring to the PG layer could fulfil a stabilizing role favouring this transition from a lipid to a protein environment. However, it remains to be determined what structure MotB recognizes and binds to in the PG layer.

The present study addresses the question of whether LTGs might play a role in the assembly of the flagellum of H. pylori. We have combined gene inactivation, site-directed mutagenesis, reporter gene expression and electron microscopy to address this issue. We extended our observations in H. pylori to other bacterial models such as S. typhimurium and Listeria monocytogenes. We show that Slt and MltC, and Auto, respectively, played similar roles in S. typhimurium and L. monocytogenes as MltD in H. pylori, suggesting a new universal role for PG cleaving enzymes in the motility of bacteria. Finally, this work highlights the importance of PG remodelling for flagella functionality.

Results

Inactivation mltD leads to a deficiency in motility of H. pylori

For our study, H. pylori strain B128 was chosen for its ability to be motile in vitro and to colonize the mouse gastric mucosa. We first inactivated slt (hp0645) and mltD (hp1572) genes in a non-polar fashion in this strain using a kanamycin resistance cassette, and tested the motility of four independent clones of each mutant by soft-agar motility assays. Figure 1A and B illustrates representative examples of these experiments. We observed that mltD mutants were highly affected in their ability to move, whereas slt mutants exhibited normal motility (Fig. 1A and B). We confirmed that defects in motility were not due to differences in growth and viability (Fig. S1).

Figure 1.

Inactivation of mltD leads to a loss of H. pylori motility.

A. Soft-agar motility testing of strain B128 (a) and its isogenic mutants B128 mltDΩKm (b and c) and sltΩKm (d).

B. The area of the motility halo of each parental strain B128 and its mutants was measured (compilation of 2 independent experiments for each 4 different clones tested).

C. Soft-agar motility testing of strain B128-Cm strain (a), its mutant B128 mltDΔKm (d), the complemented mutant B128 mltD-Cm (c) and the catalytic active site mutant B128 mltD* (b).

D. The area of the motility halo of the control parental strain B128-Cm, the complemented strain B128 mltD-Cm, which showed a restored motility, and the B128 mltDΔKm mutant confirming that our phenotype is directly due to the absence of the targeted LTG (compilation of 2 independent experiments).

E. B128 mltD* exhibited an altered motility compared with the control parental strain B128-Cm, suggesting that the activity of MltD is required for full motility in the bacteria. For all those graphs, each dot represents the area of the motility halo performed per each strain, per independent experiment. Bars represent the median of each group. Asterisks indicate the level of the P-value: * < 0.05, ** < 0.001 and *** < 0.0001.

F. Mouse stomach colonization levels of B128 (F–G) and its mltDΩKm and sltΩKm mutants, at 3 (F), 15 (G) and 30 (H) days post-infection. Loss of motility of B128 mltDΩKm was correlated with an impaired ability to colonize the mouse gastric mucosa. The dashed line indicates the lowest level of detection of colony-forming units (CFU) per gram of stomach. Statistical analysis was performed for all assays using the Mann–Whitney test. Each dot represents the number of bacteria colonizing the stomach per gram of stomach. Bars represent the median of each group. Asterisks indicate the level of the P-value: * < 0.05, ** < 0.001 and *** < 0.0001.

Next, we analysed the PG composition of the parental strain and its respective mutants (Fig. S2 and Table S1) so as to confirm that in this strain MltD and Slt exhibit the same role as described in a previous study (Chaput et al., 2007). Inactivation of slt had the same impact described previously, with a characteristic increase of the muropeptide GlcNAc-MurNAc-tripeptide and an increase of the average glycan chain length (see Chaput et al., 2007 and Table S1). The inactivation of mltD had a different impact on PG composition to that of slt inactivation, and presented a more important increase of the average glycan chain length (see Chaput et al., 2007 and Table S1). These findings suggest that the distinct biochemical activities exhibited by Slt and MltD are maintained in B128 strain, while only MltD exhibited a major impact on bacterial motility.

The phenotype can be restored by reintroduction of an active LTG

In order to confirm that our phenotype was directly related to the inactivation of the mltD gene, we performed genetic complementation assays. Attempts to reintroduce in trans the mltD gene either on an inducible plasmid or the rdxA chromosomal locus were unsuccessful. A wild-type copy of mltD was thus reintroduced into its own locus in the B128 mltD mutant by natural transformation. To select for allelic replacement, we used a construct consisting of a non-polar chloramphenicol acetyltransferase cassette downstream of mltD. As a control, the same construct was introduced in the wild-type strain B128 (named B128-Cm). As illustrated in Fig. 1C (panel c), we observed that the reintroduction in cis of a wild-type copy of mltD in the mutant restored the motility in all three complemented mutants tested, as presented in Fig. 1D.

Site-directed mutagenesis of the active-site of MltD leads to similar impact on motility

Next, we wished to determine whether the effect of LTG deletion could be attributed to a loss of activity rather than to the absence of the protein. For this, site-directed mutagenesis was used to target the active site of MltD of the B128 strain. According to the prediction of the location of the active site (http://www.ncbi.nlm.nih.gov/Structure/cdd/cdd.shtml using as query LT_GEWL), the glutamate residue at position 103 of MltD was substituted by an alanine. Inactivation of the predicted active site of MltD (B128 mltD*) led to a strongly reduced motility in vitro, as compared with the parental strain (Fig. 1C and E), suggesting that inactivation of lytic transglycosylation activity was implicated in the impaired motility of the mutants. Comparison of the PG composition of B128 mltD* with its parental strain B128 and the B128 mltD mutant (see Table S1) showed that the inactivation of the glutamate residue had the same impact on PG composition as the mutant B128 mltD, specifically the increase of average glycan chain length, as well as an overall decrease of the cross-linking.

Loss of motility of mltD mutant is correlated with a mouse colonization defect

As bacterial motility is essential for in vivo colonization, we performed mouse colonization experiments in order to understand if the reduced motility of slt and mltD mutants could affect their ability to colonize the stomach mucosa. For this, we oro-gastrically challenged C57Bl/6J mice each with 2 × 108 bacteria and followed the ability of the parental strain, and corresponding slt or mltD mutants to colonize the mouse gastric mucosa at different time points post-infection: 3, 15 and 30 days (Fig. 1F–H). H. pylori B128 mltD mutants, which were strongly affected in motility in vitro, showed significant colonization defects from 3 days post-infection. In contrast, B128 slt mutants colonized mice as well as its parental strain (Fig. 1F–H). To ensure that the in vivo phenotype was directly related to reduced motility, we analysed the adherence properties of the strains to human gastric epithelial AGS cells in vitro. We measured the adhesion of B128 and its mutants, B128 slt and B128 mltD by flow cytometry (Fig. S3). The mutants adhered to AGS cells to the same extent as the parental strain.

mltD inactivation does not modify flagellin expression nor the presence of flagella

Flagella are complex protein structures whose assembly is regulated at different check points (Chevance and Hughes, 2008). The expression of the flagellin subunit FlaA will only occur if all previous protein components of the basal body of the flagella are correctly assembled through the cell wall and cell envelope. Hence, we used flaA gene expression as a reporter for the correct assembly of flagella in our mutants. Plasmid pSB13 carries the GFP protein, whose expression is under the control of the flaA gene promoter (Backert et al., 2005). Plasmid pSB13 was introduced by conjugation into B128 and its isogenic slt and mltD mutants. As a control, pSB13 was introduced in B128 1032ΩGm, in which flaA is not expressed (hp1032 encodes the H. pylori sigma factor 28). GFP expression was assayed at the single-cell level by fluorescent microscopy (Fig. 2A) and at the population level by flow-cytometry (Fig. 2B and Fig. S4). Indeed, B128 bacteria harbouring pSB13 were fluorescent, compared with those without the plasmid. Interestingly, slt and mltD mutants carrying pSB13 were as fluorescent as the parental strain, whereas B128 1032ΔGm pSB13 was not. Thus, the flaA gene was expressed normally in the slt and mltD mutants.

Figure 2.

H. pylori slt and mltD mutants express normal levels of flaA.

A. Phase-contrast and fluorescence microscopy of the different strains used for FACS analysis. Although the B128 strain and isogenic mutants mltDΩKm and sltΩKm when transformed with plasmid pSB13 showed similar fluorescence, our negative controls, strain B128 without the plasmid and its isogenic sigma 28 mutant, B128 1032ΔGm, did not show any fluorescence.

B. FACS analysis of the expression of GFP as a reporter of flaA expression in B128 and isogenic mltDΩKm and sltΩKm mutant strains harbouring the plasmid pSB13, showing that the two mutants expressed GFP in a similar fashion to the parental B128 strain.

C. Western blot analysis and colloidal blue staining of total extracts of the parental strain B128 and the different isogenic mutants. B128 mltDΩKm expressed FlaA at similar level as their parental strain, whereas 26695 (due to a frame shift of the fliE gene (Josenhans et al., 2000) and B128 1032ΔGm did not. UreA was used as a loading control.

D. Scanning electronic microscopy of the slt and mltD mutants, showing the presence of intact unipolar flagella as in the respective parental strain.

E. Enumeration of flagella per bacterium expressed as percentage. The number of flagella per bacterium is similar for the different mutants and their parental strain. The data presented is the average of three independent experiments based on the enumeration of flagella of a minimum of 50 bacteria per strain per experiment.

However, since gene expression and protein production do not necessarily correlate directly, we next addressed by Western blotting whether the protein FlaA was correctly produced in H. pylori B128 and its respective slt and mltD mutants. The results showed that FlaA was produced at the same levels in the parental strain and in all mutants (Fig. 2C), suggesting that the flagella are likely to be assembled normally.

To determine if the flagella were correctly produced and localized, scanning electron microscopy studies were performed. The slt and mltD mutants of H. pylori B128 exhibited unipolar flagella as expected (Fig. 2D), and their enumeration revealed a similar distribution of flagella per bacterium between parental strain and its isogenic mutants (Fig. 2E). Taken together, these findings suggest that although mltD inactivation led to a loss of motility, the flagella of these mutants were correctly assembled, had normal length, properly localized and present in normal numbers (1 to 5 flagella per bacterium).

mltD inactivation results in a paralysed phenotype for the bacteria in liquid media

During the primary infection of a new host, H. pylori encounters two distinct environments, first the gastric lumen, followed by the gastric mucus. In both these steps of the infectious process, motility is crucial for H. pylori, allowing it to escape from the gastric lumen and to move through the viscous mucus layer to reach the gastric epithelium. The viscosity of soft agar mimics the environment in the mucus layer. Therefore, in order to address the behaviour of the LTG mutants during the first step of colonization, we undertook video microscopy studies in liquid medium. For this, we analysed movies (1–2 min duration), with intervals of 100 ms between each frame (see supplementary movies). Both B128 and its slt mutant had similar speeds (∼ 13 μm s−1). It appeared that loss of motility for B128 mltD was associated with the bacterial population presenting a paralysed phenotype. Interestingly, H. pylori motB mutants had a very similar phenotype to our LTG mutants, i.e. the flagella are normally assembled but are impaired in their rotation and torque generation (Ottemann and Lowenthal, 2002).

Inactivation of slt and mltC leads to a reduced motility in S. typhimurium

As the motor is universally present in all bacteria that assemble flagella, we speculated whether the impact of LTGs on motility may be true for other motile bacteria such as S. typhimurium. This bacterium has several LTGs, including the proteins Slt and MltC. Interestingly, the inactivation of the corresponding genes led to a similar loss of motility as seen for H. pylori (Fig. 3A and B). As shown in Fig. 3C, the S. typhimurium slt and mltC mutants produced normal amounts of the flagellin, FliC. The less drastic effect of slt and mltC inactivation may be due to redundancy of function and/or to the different patterns of flagellar arrangement between H. pylori and S. typhimurium bacteria, i.e. unipolar and peritrichous respectively. As observed on Fig. S5, however, the loss of motility in the S. typhimurium slt and mltC mutants did not appear to be due to impaired assembly of the flagellum, a finding consistent with the situation H. pylori.

Figure 3.

The inactivation of LTGs homologues and glucosaminidase results in motility defects in S. typhimurium and L. monocytogenes respectively.

A. Motility assays testing of S. typhimurium SL1344 (a); a motA mutant (b) of the same parental strain, which assembles fully its flagella but lacks a functional stator to generate torque from the proton-motive force; and slt and mltC mutants, analogous to H. pylori sltΩKm and mltDΩKm respectively. Inactivation of slt (c and d) in S. typhimurium led to a loss of motility.

B. The area of the motility halo of the parental strain and its slt, mltC and motA mutants were measured and compared. (Compilation of 3 independent experiments). Statistical analysis was done using the Mann-Whitney test. Each dot represents the area of the motility halo performed per each strain, per independent experiment. Bars represent the median of each group. Asterisks indicate the level of the P-value: * < 0.05, ** < 0.001 and *** < 0.0001.

C. Western blotting and Coomassie blue staining of total protein extracts of the different Samonella strains. The amount of flagellin FliC was estimated using a commercially available antibody (Biolegends). Two independent clones of motA, as well as 4 clones of slt and of mltC were loaded and compared with the wild-type strain and a fliC/fljB double mutant. Total protein samples (20 μg) were loaded per well and visualized by Coomassie staining. Western blotting against FliC clearly shows that all 4 independent clones of slt or mltC produced normal amounts of flagellin as per the wild-type strain. FliC antibody specificity was confirmed using a fliC/fljB mutant.

D. Motility assay testings of L. monocytogenes EGDe and EGD600 and their mutants. Inactivation of aut led to a severe loss of motility, which could be restored by the reintroduction in trans of the wild-type copy of the gene. Inactivation of the ami gene did not lead to a similar reduction in motility.

E. The area of the motility halo of the parental strains EGDe and EGD600 and their mutants were measured and compared. As a negative control, we used a flagellin deficient mutant (compilation of 3 independent experiments). Statistical analysis was performed using the Mann-Whitney test. Each dot represents the area of the motility halo performed per each strain, per independent experiment. Bars represent the median of each group. Asterisks indicate the level of the P-value: * < 0.05, ** < 0.001 and *** < 0.0001.

F. Total protein extracts of the different Listeria strains (EGDe, EGDe Δfla, EGDe Δaut and EGDe Δaut+aut in lanes 1 to 4 respectively) were loaded and analysed by Western blotting using an anti-FlaA antibody and colloidal blue straining. FlaA was produced to similar levels in the aut mutant as in the parental strain.

Inactivation of aut results in a loss of bacterial motility in Listeria monocytogenes

Our data on H. pylori and S. typhimurium suggested that PG maturation and generation of anhydro-muropeptides could constitute an important feature for the functionality of the flagellar motor. In Gram-positive bacteria, the glycan chains do not carry significant amounts of anhydro-N-acetylmuramic acid (Boneca et al., 2000; Hayhurst et al., 2008). Maturation of the glycan backbone occurs primarily by glucosaminidases (Boneca et al., 2000) instead of LTGs, thus generating reducing ends with an N-acetyl-glucosamine in place of anhydro-N-acetylmuramic acid. Hence, we hypothesized that in Gram-positive bacteria, glucosaminidases could fulfil the same role as LTGs in a motile Gram-positive pathogen. To test this hypothesis, we used L. monocytogenes. In this bacterium, the autolysin Auto is the most studied glucosaminidase (Cabanes et al., 2004; Bublitz et al., 2009), and its inactivation did not have any morphological effect on the bacterium. Hence, we tested whether Auto encoded by the aut gene could be involved in L. monocytogenes motility. In parallel, we tested another major hydrolase of L. monocytogenes, the amidase Ami as a control (McLaughlan and Foster, 1998). The aut-deficient strain was severely affected when tested in the soft-agar motility assay (Fig. 3D and E). The phenotype was however completely restored by providing aut in trans on a plasmid. The phenotype was specific to Auto as the ami mutant was as motile as its parental strain. We tested whether the aut mutant was able to correctly assemble its flagella. As with H. pylori, the loss of motility did not appear to be due to an impaired assembly of the flagellum as shown by electron microscopy (Fig. S6). Furthermore, the aut mutant produced similar amounts of flagellin A (FlaA) as the parental and complemented strains (Fig. 3F).

Localization of hpMotB requires the presence of active LTGs

We hypothesized that the absence of the LTG activities might impair hpMotB binding to the PG layer. Namely, that without the LTG, hpMotB would not be able to find the right location in the bacterial cell wall. To test this hypothesis, we constructed a plasmid harbouring the gene coding for hpMotB (MotB from H. pylori) tagged with the GFP protein. GFP protein was placed at the N-terminus of MotB so as to not interfere with the PG-binding motif present in the C-terminus of the protein. Moreover, we introduced between the GFP protein and MotB a linker, so as to prevent an alteration of the arrangement of the MotA/MotB complex. We introduced this plasmid in different strains of E. coli and induced the synthesis of the GFP–hpMotB (Fig. 4). We confirmed by Western blot analysis that GFP–hpMotB was expressed similarly in all the strains harbouring the plasmid (Fig. S7A). In the wild-type non-flagellated E. coli strain MC1061, we observed that H. pylori MotB localized predominantly at the pole, which is where the flagella and the MotA/MotB stator are localized in H. pylori. The same observation was made in the flagellated E. coli strain, MG1655 (Fig. 4). However, when one or several LTGs were inactivated, such as in MG1655 sltΩKm or MHD79 (where 6 LTGs are inactivated), respectively, we observed that GFP–hpMotB protein was distributed throughout the bacterial surface (Fig. 4). Inactivation of sltY in strain MG1655 also displayed an altered motility (Fig. S8) when compared with the wild-type strain or a non-flagellated mutant MG1655 fliC (Pichon et al., 2009). However, the localization of IcsA, a protein that exhibits a polar localization in Shigella flexneri (Charles et al., 2001), when fused to the same GFPmut2 protein, did not change in the absence of one or several LTGs in the same E. coli strain (Fig. S9). These data strongly suggested that LTGs are specifically implicated in the localization of hpMotB.

Figure 4.

Localization of hpMotB in E. coli strains. Fluorescence microscopy of the different E. coli strains harbouring plasmid pAM001. In wild-type strains MC1061 pAM001 and MG1655 pAM001, GFP–hpMotB was clearly localized at the bacterial pole, whereas in the LTG mutant strains, MHD79 pAM001 and MG1655 sltΩKm pAM001, the fluorescent protein was distributed over the entire bacterial surface. The graph represents the percentage of bacteria possessing polar patches with respect to the total fluorescent bacteria. For each strain, 100 fluorescent bacteria were observed.

Discussion

Previous studies reported that LTGs play a role in motility by allowing the assembly of the flagella through the PG layer, such as is the case for FlgJ in S. typhimurium (Nambu et al., 2006). Besides the dedicated FlgJ, other LTGs have been implicated in motility, such as PleA of Caulobacter crescentus, but again, this protein is implicated in the assembly of the flagella (Viollier and Shapiro, 2003). For the first time, we have shown that LTGs can also interfere with motility without impairing flagellar assembly. Indeed, inactivation of the LTG gene mltD led to a loss of motility for H. pylori strain B128 (Fig. 1A and B) without affecting flagellar assembly, localization or number (Fig. 2). The mltD mutant phenotype is similar to those exhibited by motB mutants in different bacterial species, i.e. a ‘paralysed'-like phenotype as observed by video microscopy. These data clearly suggested that the flagella were not rotating, despite being correctly assembled. Furthermore, it appeared that the muramidase activity of MltD was directly implicated in its new role. Indeed, the replacement of the glutamate residue, the predicted active site residue of MltD, recapitulated the phenotype of the mltD mutant (Fig. 1C and E, and Table S1). Thus, it appears that the remodelling of PG composition around the flagellum is a crucial component for full motility. It is likely that MltD generates specific residues or environment to which MotB is anchored. In the absence of the LTG, MotB would have a defective anchoring to the PG layer, resulting in an impaired transition from inactive/plugged to a fully active/unplugged status and thus preventing the proton flux through the channel defined by MotA and MotB. Alternatively, MotB would misalign with MotA thereby affecting the transmission of energy and the generation of flagellar rotation. Finally, the MotA/MotB complex has been shown to be dynamic, with MotB moving from the flagellum to a dispersed pool of MotB over the cell membrane (Leake et al., 2006). During this process, the MotA/MotB complex and the MS ring undergo major changes from a phospholipid environment to a protein one. MotB anchoring to the PG would favour the stabilization of the MotA/MotB complex around the MS ring. Indeed, we proposed that MltD plays an important role in the binding of MotB to the PG layer, allowing the correct location of MotB in the flagellar apparatus. This model is supported by our hpMotB localization studies (Fig. 4). Although hpMotB localizes specifically at the bacterial poles in wild-type E. coli bacteria, the proteins are distributed all over the bacterial surface in E. coli mutants for one or more LTGs. Yet, another polar protein, IcsA, did not show any variation of its localization in the same mutants (Fig. S9). This strongly suggests that without LTGs, hpMotB does not localize properly at the poles where flagella are assembled in H. pylori leading to a loss of motility, and that LTGs are specifically implicated in this correct localization. Unfortunately, conducting the same experiments in H. pylori, namely, following GFP–hpMotB, was not possible to do. At each attempt to express GFP–hpMotB in this bacterium, we could not detect either the fusion or cleaved GFP (Fig. S7B), although we detected GFP protein as a cytosolic reporter protein in bacteria harbouring the plasmid pSB13 (Fig. S7B). In E. coli, we can exclude the localization at the pole as a consequence of inclusion bodies. Indeed, GFP–hpMotB was expressed at similar levels in wild-type and LTG mutants. Also, GFP–hpMotB delocalized in the LTG mutants arguing against this possibility. Furthermore, IcsA exclusive localization at the pole insured that the specific polar environment had not change. Thus, we can conclude that the only changes occur exclusively at the PG level.

Due to the importance of tightly controlling each step of the motility process, it was crucial to investigate if the role of MltD, and more generally LTG, could be similar in Gram-negative or Gram-positive species. We showed that the inactivation of homologous LTGs in S. typhimurium, Slt and MltC, also lead to a loss of bacterial motility (Fig. 3A and B) but without impairing flagellin expression (Fig. 3C). Moreover, electron microscopy studies showed correctly assembled flagella in the LTGs mutants (Fig. S5), comforting our thesis that the loss of motility is related to a loss of functionality of the flagella, and not impaired assembly. Interestingly, all LTGs involved in the motility defect belong to the same sub-family 1 (or Slt-like; Blackburn and Clarke, 2001). As no LTG activity has been described in Gram-positive species, we took an interest on Auto, a glucosaminidase, which remodels the glycan chains in L. monocytogenes (Cabanes et al., 2004). Interestingly, its inactivation showed the same phenotype as observed for mltD in H. pylori: a major loss of motility for the bacteria (Fig. 3D and E), despite the correct expression of flagellin (Fig. 3F) and the presence of flagella (Fig. S6). Thus, as the phenotypes were similar in very different backgrounds, we postulate that MotB needs the remodelling of sugars chains for its proper anchoring and this is made possible through the activity of either LTGs in Gram-negative species, or glucosaminidases in Gram-positive species. LTGs specifically generate anhydro-N-acetyl-muramic acid residue, at the end of glycan chains. Indeed, our data suggest that in Gram-negative bacteria, the motor could preferentially bind to this residue at the end of glycan chains. We propose then a model in which the LTGs mediate the maturation of the PG by generating the anhydro-N-acetyl muramic acid moieties to which MotB binds and tethers the motor to the surrounding PG layer. In Gram-positive bacteria, which do not generate the anhydro-N-acetyl-muramic acid residues, the tethering would occur instead on N-acetyl glucosamine residues at the reducing end of the glycan chains, or N-acetyl muramic acid residues at the non-reducing end. Since the 1970s, several reports have implicated PG hydrolases in the motility of Bacillus subtilis (Fein, 1979). However, these reports described hydrolases that were required for flagellar assembly similarly to FlgJ in the Enterobactericiae. Only recently, Chen and colleagues observed in B. subtilis that inactivation of two PG hydrolases including the glucosaminidase LytD led to a similar phenotype to the one observed in this work (Chen et al., 2009), suggesting again a universal role for PG cleaving enzymes in the maturation of the PG and the proper functioning of the MotA/MotB stator. In Gram-negative bacteria, LTGs seem to be dedicated to this role whereas in Gram-positive organisms it seems to involve glucosaminidases. However, the recent crystallization of MotB from H. pylori (Roujeinikova, 2008) allowed the modelling of a MotB dimer binding to the PG layer. The model suggests that MotB would bind two glycan chains and accommodate in a grove between the two dimers the cross-linked stem peptides. Thus, the observed decrease of cross-linking in the mltD mutant (Table S1) could also affect MotB binding to the PG layer resulting in decreased motility.

The apparent absence of a dedicated LTG for the assembly of the flagellum suggests that the network defined by the PG is apparently large enough to allow the assembly of the flagella without specific remodelling. In contrast, it does not seem large enough to allow the right positioning of MotB next to the flagellum. A second hypothesis will be that the activity of MltD would be required to create enough space instead of generating specific moieties for MotB anchoring. Finally, we could also envision an alternate hypothesis involving a component of the flagellum apparatus ‘recruiting’ MltD to create space around the flagellum necessary for the assembly of the stator.

Furthermore, we could observe in Salmonella that Slt and MltC seem to have a redundant role in bacterial motility, while in H. pylori, we do not observe such things. We observe a minor impact of slt inactivation on H. pylori motility, in spite of some similarities with mltD inactivation in its impact on PG composition. This could be explained by the more important number of LTGs in Salmonella compare with only 2 in H. pylori. In H. pylori B128 strain, it appears that each LTG has specific role which can not be permutable. As both Slt and MltD generate anhydro-N-acetyl-muramic acid residue at the end of glycan chains, we could hypothesis that the fact that only MltD is implicated in B128 motility would be attributed to the fact that MltD would be located at the pole of the bacteria, while Slt is not, or that MltD is specifically recruited by some flagellar apparatus member.

Motility represents a major virulence property of pathogenic bacteria, allowing them to reach and colonize their environmental niche successfully. Accordingly, we tested the ability of our H. pylori mutants to colonize the mouse gastric mucosa. We showed that B128 mltD was affected in its ability to colonize the mouse gastric mucosa (Fig. 1F–H), while we demonstrated that it adhesion properties was not affected (Fig. S3). Hence, as LTGs appear to be involved in motility and successful in vivo colonization, these enzymes could constitute a new family of therapeutic targets that would allow the blocking of the virulence potential of pathogens. H. pylori is a major human pathogen colonizing around 50% of the human population. Unfortunately, clinical strains are becoming increasingly resistant to most available antibiotics and most vaccination strategies have been abandoned. Thus, it is one of the few pathogens for which a targeted therapeutic strategy is economically sound. Hence, we propose that the development of additional molecules inhibiting not only MltD but also other LTGs constitutes an alternative approach to the increasing problem of antibiotic resistance by targeting phenotypes essential for survival of pathogenic bacteria in their niche.

Experimental procedures

Bacterial strains and growth conditions

All strains used in this study are described in Table S2. Briefly, H. pylori strains were grown at 37°C in microaerophilic conditions (Campygen, Oxoid or Anoxomat), on Blood Agar (Oxoid) with 10% Defibrinated Horse Blood (Biomerieux) and an antimicrobial-antifungal mix (in-house made and called ABmix) or in liquid culture in Brain Heart Infusion (BHI, Oxoid) with 10% of decomplemented Fetal Calf Serum (FCS, Eurobio) and ABmix. ABmix is composed of polymyxin B (0.31 μg ml−1), amphotericin B (2.5 μg ml−1), vancomycin (12.5 μg ml−1) and trimethoprim (6.25 μg ml−1). Motility assays were performed in 0.35% agar Brucella Broth medium (Oxoid) complemented with 10% of decomplemented FCS and ABmix. When necessary, media were supplemented with kanamycin (20 μg ml−1), apramycin (5 μg ml−1) or chloramphenicol (3 μg ml−1). For mouse colonization trials, bacitracin (200 μg ml−1) and nalidixic acid (10 μg ml−1) were added to the plates.

Motility assays with S. enterica serovar typhimurium SL1344 were performed on the same media as described elsewhere (Kutsukake, 1997).

Listeria monocytogenes was grown in BHI media, either liquid or solid, at either room temperature or 30°C. For the EGDeΔaut + aut, 5 μg ml−1 of erythromycin (Sigma) was added to the medium. Motility assays were performed on 0.35% BHI agar plate.

Cultures of E. coli strains were performed in Luria–Bertani (LB) media, added with, when necessary, chloramphenicol (20 μg ml−1 except for strain MHD79 for which a 200 μg ml−1 was used for transformation event) or with apramycin (100 μg ml−1).

Construction of mutants

H. pylori mutants

Helicobacter pylori mutants were constructed by natural transformation using non-polar kanamycin (Skouloubris et al., 1998; Chaput et al., 2007) or non-polar gentamicin (Bury-Mone et al., 2003) cassettes, as described previously. More details on the strategies used to construct mutants and the primers used for the constructions are described in the supplementary materials and methods (Supporting information). Briefly, for complementation studies, a wild-type copy of mltD from B128 was introduced in the original locus in B128 mltDΔKm, with downstream of mltD a non-polar chloramphenicol cassette. B128 received the same construction, so as to obtain B128-Cm, which served as the control for further analyses. Site-directed mutagenesis was performed by reverse PCR as previously described (Bury-Mone et al., 2001) in the same construct used for the complementation assays. Allelic exchange was selected for using the chloramphenicol non-polar cassette, resulting in B128 mltD* mutant. Insertion of the mutation on the chromosome was checked by sequencing. Conjugation assays to transfer the pSB13 plasmid into H. pylori was performed as described previously (Backert et al., 2005).

S. typhimurium and E. coli mutants

Salmonella typhimurium slt, mltC and motA genes were inactivated using a non-polar kanamycin resistant cassette, aphaA3, introduced in the presence of the pKOBEG plasmid (Derbise et al., 2003).

The E. coli sltΩKm mutant was constructed by inactivation of the E. coli sltY gene as described for the S. typhimurium mutants.

To generate E. coli transformants carrying hpMotB or IcsA, we first constructed plasmid pAM001, encoding the GFP fused to the N-terminus of H. pylori MotB in the shuttle vector pILL2150 (Boneca et al., 2008). Plasmid pMAB002 was constructed by introducing icsA507–620–gfpmut2 (Charles et al., 2001) in the same vector. Details of their construction are described in the supplementary material and methods. The resulting plasmids were used for transforming E. coli strains MC1061, MG1655, MHD79 and MG1655 sltΩKm.

Mice experiments

C57Bl/6J female mice, aged of 4–5 weeks (Charles River laboratories) were infected oro-gastrically by B128 strains (2 × 108 bacteria per mouse) (Ferrero et al., 1998). Colonization rates were determined after 3, 15 or 30 days by enumeration of colony forming units per gram of stomach.

Fluorescent microscopy and expression analysis by flow cytometry

Bacteria from exponential liquid cultures were fixed with 4% paraformaldehyde (PFA) and washed twice with phosphate buffer saline (PBS). Cover-slides were fixed on slides with Dako fluorescence mounting medium (Prolong Gold, Invitrogen). Images were acquired with an inverted microscope Axio Observer (Zeiss), by phase contrast and in GFP fluorescence microscopy using the Axiovision software. Exposure of samples was set at 200 ms.

Overnight cultures of bacteria were fixed in 4% PFA. Fluorescence of 10 000 events was analysed by flow cytometry on a FACSCAN (Becton Dickinson).

Generation of the polyclonal FlaA antibodies and analysis of flagellin production

Polyclonal antiserum (α-FlaA) was raised according to standard protocols (Biogenes, Berlin, Germany) by immunization of two rabbits with a conserved H. pylori flagellin FlaA-derived peptide (amino acids 93–106: KVKATQAAQDGQTT) conjugated to Limulus polyphemus haemocyanin carrier protein. The antiserum was affinity-purified and the specificity against the 60-kDa FlaA protein was confirmed by Western blotting.

Exponentially growing bacteria were pelleted then sonicated after PBS washing. Twenty or thirty micrograms of protein were loaded in 10% SDS-PAGE gel and either transferred onto a PVDF membrane or stained with Colloidal Blue (Neuhoff et al., 1988). For Western blotting analysis of H. pylori samples, we used polyclonal antiserum α-FlaA as a primary antibody, or against UreA as a control (Ferrero et al., 1994) and then polyclonal goat anti-rabbit antibody was used as a secondary antibody. Detection was performed using the ECL kit (Pierce). For L. monocytogenes and Salmonella samples, we used a polyclonal antiserum α-Fla (Peel et al., 1988) or α-FliC (Biolegends) respectively, and a polyclonal goat anti-rabbit IgG antibody as a secondary antibody.

Scanning electron microscopy of H. pylori

Exponential bacteria were stained in a fixative solution containing 2.5% glutaraldehyde in 0.1 M cacodylate buffer (pH 7.2). Dried samples were sputtered with 10 nm gold palladium, with a GATAN Ion Beam Coater and were examined and photographed with a JEOL JSM 6700F field emission scanning electron microscope operating at 5 kV. Images were acquired with the upper SE detector (SEI).

Video microscopy

Bacteria were cultured in BHI liquid medium untill mid-log phase growth. They were observed at room temperature on a 10× phase-contrast objective on a Zeiss microscope, with an AxioCam camera. Bacteria were recorded every 100 ms for a total period of 1–2 min, using the AxioVision Software (Zeiss).

PG extraction and analysis

The PG of H. pylori was extracted from 100 ml of exponentially growing cultures. The PG was purified, digested and analysed as described previously (Glauner, 1988).

Acknowledgements

Sophie Roure, Mathilde Bonis and Catherine Chaput were supported by a PhD fellowship (Ministère de l'Enseignement Supérieur et de la Recherche, France). Sophie Roure and Catherine Chaput were also supported by FRM (Fondation pour la Recherche Médicale) fellowships. Mathilde Bonis was also supported by a Pasteur-Weizmann fellowship. Austin Mattox was supported by the Pasteur Foundation. This study was supported by the Programme Transversal de Recherche (PTR) 153 (Institut Pasteur) and the ERC starting grant (PGNfromSHAPEtoVIR No. 202283). Richard Ferrero is a Senior Research Fellow of the National Health and Medical Research Council of Australia.

We would like to thank Pascale Cossart, Hélène Bierne and Olivier Dussurget for providing Listeria strains and anti-FlaA antibody, Michael Kolbe for a kindly gift of the Salmonella typhimurium fliC/fljB double mutant and Hilde De Reuse for providing an anti-UreA antibody. We would also thank Hélène Bierne and Christine Josenhans for the fruitful discussions and experimental suggestions. The authors have no conflict of interest to declare.

Ancillary