The Role of Flagella in Clostridium difficile Pathogenesis: Comparison between a Non-Epidemic and an Epidemic Strain

Clostridium difficile is a major cause of healthcare-associated infection and inflicts a considerable financial burden on healthcare systems worldwide. Disease symptoms range from self-limiting diarrhoea to fatal pseudomembranous colitis. Whilst C. difficile has two major virulence factors, toxin A and B, it is generally accepted that other virulence components of the bacterium contribute to disease. C. difficile colonises the gut of humans and animals and hence the processes of adherence and colonisation are essential for disease onset. Previously it has been suggested that flagella might be implicated in colonisation. Here we tested this hypothesis by comparing flagellated parental strains to strains in which flagella genes were inactivated using ClosTron technology. Our focus was on a UK-outbreak, PCR-ribotype 027 (B1/NAP1) strain, R20291. We compared the flagellated wild-type to a mutant with a paralyzed flagellum and also to mutants (fliC, fliD and flgE) that no longer produce flagella in vitro and in vivo. Our results with R20291 provide the first strong evidence that by disabling the motor of the flagellum, the structural components of the flagellum rather than active motility, is needed for adherence and colonisation of the intestinal epithelium during infection. Comparison to published data on 630Δerm and our own data on that strain revealed major differences between the strains: the R20291 flagellar mutants adhered less than the parental strain in vitro, whereas we saw the opposite in 630Δerm. We also showed that flagella and motility are not needed for successful colonisation in vivo using strain 630Δerm. Finally we demonstrated that in strain R20291, flagella do play a role in colonisation and adherence and that there are striking differences between C. difficile strains. The latter emphasises the overriding need to characterize more than just one strain before drawing general conclusions concerning specific mechanisms of pathogenesis.


Introduction
Clostridium difficile is the principle cause of hospital acquired antibiotic associated diarrhoea in North America and Europe. The morbidity and mortality rates of nosocomial Clostridium difficile infection (CDI) continue to rise, particularly following the global emergence of epidemic C. difficile strains (027/BI/NAP1) [1,2]. The two main virulence factors of CDI are the large clostridial toxins A and B [3][4][5][6]. However, other factors undoubtedly contribute to disease. Gut colonisation is a prerequisite for CDI, yet little is known of the mechanisms involved. The mucosal surface carpeting the intestinal epithelium is the main site of hostpathogen interaction, in which this organism must both evade the immune response and interact with enterocytes and adhere to specific surface molecules. C. difficile possesses multiple putative surface adhesins, potentially functioning as colonisation factors, including cell surface-associated proteins (S-layer and SLPs), fibronectin-binding protein FbpA, proteases such as Cwp84, hydrolytic enzymes, heat-shock proteins such as GroEl [7][8][9] and flagellar cap FliD and flagellin FliC structural components [10]. FliD and FliC are both components of the bacterial flagellum, an important multi-purpose structure that has diverse biological functions to favour bacterial survival and host colonisation [11].
For most gastrointestinal pathogens, flagella and flagellummediated motility are recognised as essential virulence factors, rendering the pathogen more capable of moving towards the site of colonisation. For instance, the intestinal enteric pathogens Listeria monocytogenes [12] and Vibrio anguillarum [13] strictly require functional flagellum-mediated motility to invade their hosts and establish successful colonisation. Flagellum-mediated motility is also an essential virulence factor required by Helicobacter pylori to colonise the stomach [14]. Pathogen survival in vivo can be enhanced through the formation of complex communities known as biofilms, and flagella have been shown to play a role in the formation and development of biofilms in a number of pathogens [15,16], most recently C. difficile [17].
A role for flagellum components as adhesins mediating bacterial attachment to host cell surfaces noted above is not restricted to C. difficile. For example, the flagellum of enteropathogenic Escherichia coli contributes to adherence to epithelial cells independent of flagellum-mediated motility [18]. Moreover, Flagellin (FliC) and the flagellar cap protein (FliD) of Pseudomonas aeruginosa, are associated with adherence and colonisation of the respiratory tract through recognition of mucin (Muc1), an abundant component protein of airway mucus [19]. In the past, the absence of effective genetic tools led to the deployment of restricted and indirect techniques in attempting to define the potential role of flagella in adherence and colonisation of C. difficile. Tasteyre et al. [20] observed that the presence of flagella increases the capacity of C. difficile to associate with the intestinal epithelial tissue. The flagellated, motile C. difficile attach more efficiently to the caecal wall of axenic mice than non-flagellated strains of the same serogroup. Moreover, in a separate analysis purified recombinant flagellar cap (FliD) and flagellin (FliC) proteins were shown to attach to tissue culture cells [10]. These studies led to the conclusion that flagellin and the flagellar cap may serve as one of the multiple cell-surface adhesins of C. difficile. This contradicted an earlier study [21] of the flagella of clinical C. difficile strains which concluded flagella played no role in adherence, since the antiserum that was raised against the purified recombinant flagellin did not inhibit adherence to cultured cells.
Recently a paper was published by Dingle et al. [22] investigating the importance of flagella in the virulence of C. difficile strain 630Derm. For the first time they were able to construct isogenic mutants in fliC and fliD using ClosTron technology [23]. Interestingly they found that the flagella mutants adhered more strongly to Caco2 cells in vitro and showed increased toxicity in vitro and in vivo tested in the hamster model of infection. Whilst the majority of C. difficile isolates appear to produce flagella, a high degree of variation of flagella-related gene content is evident [10,24]. It is therefore of value to extend these studies to further strains before drawing any general conclusions as to the involvement of flagella and motility in the virulence of this bacterium. In particular, investigation of the role of flagella in more relevant epidemic strains is required. In the present study we have focussed on the epidemic 027/BI/NAP1 strain, R20291. Our aim was to elucidate the mechanism by which flagella contribute to C. difficile adhesion to human intestinal epithelial cells and the potential role of motility in colonisation of the intestinal tract in mice through the comparative analysis of directed mutants in specific flagellar genes in both C. difficile 630Derm and R20291.

Results and Discussion
Construction and phenotypic characterizations of flagellar-associated mutants of C. difficile 630Derm and R20291 In order to analyse the role of flagella in the pathogenesis of C. difficile three different mutants were made in both R20291 and 630Derm. Using ClosTron technology [25] we insertionally inactivated fliD, coding for the flagella cap, fliC, encoding flagellin and flgE, encoding the hook protein. The resultant mutants CRG3357 (Cdi630Derm-fliC515a::CT), CDR3356 (Cdi630Derm-fliD560a::CT). CRG3355 (Cdi630Derm-flgE310s::CT), CRG3351 (CdiR20291-fliC430s::CT), CRG3350 (CdiR20291-fliD121s::CT) and CRG3349 (CdiR20291-flgE311s::CT) were verified by PCR using a gene specific and an intron specific primer (Table S1) to amplify across the junction (Fig. S1) PCR products were also sequenced. Southern blot analysis using an intron specific probe confirmed that only one insertion had occurred in each mutant (Fig. 1). A Western blot probing against FliC with a polyclonal antibody revealed that none of the mutants produced flagellin, while it was readily detected in the parental strains (data not shown). Motility of the mutants was tested on swimming and swarming plates and compared to the parental strains. Whereas both 630Derm and R20291 showed motility on plates (Fig. 2a, extreme left handside) all three mutants in the respective strains were completely immobilized (Fig. 2a). The respective complementation plasmids pMTLSB-1 (fliC), pMTLSB-2 (fliD) and pMTLSB-3 (flgE) were introduced into each mutant. In each plasmid, the cloned gene had been placed under the transcriptional control of the promoter of the fliC gene (Materials and Methods). Plasmid pMTLSB-1 was introduced into the fliC mutants CRG3357 and CRG3351 and successfully complemented the motility phenotype. Likewise pMTLSB-2 introduced into CRG3356 and CRG3350, the fliD mutants, partially complemented that phenotype back to wild type. However complementation of CRG3355 and CRG3349 with pMTLSB-3 was only partially successful. This was most likely due to the fact that the fliC promoter controls gene expression in class three flagella genes, whereas flgE is transcribed as a class two gene. Disrupting the highly organised flagella assembly by inappropriate transcription signals might be an explanation why no full complementation was achieved with these constructs (Fig. 2a). The parental strains and derived mutants and complemented mutants, were all examined under the transmission electron microscope (TEM). Interestingly we observed a great difference between 630Derm and R20291. Whereas 630Derm was peritrichously flagellated, R20291 was monotrichously flagellated with only a single flagellum present on its surface. This most likely explains why R20291 displayed less motility than 630Derm in the swimming and swarming assays. No flagella were observed for any of the flagella mutants (Fig. 2b).

Construction of a paralyzed flagellum
To address the question as to whether flagella act as adhesins or function merely as a motility apparatus we generated mutants in the motor genes motA and B and also in the rotor gene fliG. After analysis of these mutants via TEM and motility plates we noticed that these mutants were not able to swim or swarm and the TEM revealed a complete lack of flagella for the motB and the fliGmutant and a 50% reduction for the motA-mutant (data not shown). This finding is in line with previously reported data in Listeria monocytogenes [12] which demonstrated that a complete disruption of motor genes leads to loss of flagella. We can thus conclude that these genes do not only play a role in the motor but also act as a structural or assembly component. A multiple alignment showed that the aspartate residue at position 23 was highly conserved. This residue has successfully been changed in Listeria to obtain a paralyzed flagellum. Using homologous recombination [26] we changed aspartate 23 to alanine in C. difficile R20291 and examined the mutant under the TEM and in motility assays (Fig. S2). The microscopy revealed a single flagellum, like the wild-type, but no motility was observed. This strain, R20291 motB:D23A, was used alongside the structural flagella mutants in in vitro adherence assays.
The flagellum of R20291 is used as an adhesin to enhance bacterial association to human intestinal epithelial cells in vitro The current literature describing the role of the C. difficile flagellum in adhesion is contradictory. In keeping with a previous report investing the adherence of the non-epidemic C. difficile strain 630Derm to Caco-2 cell [22], we found that both a fliC and a fliD mutant adhered more strongly than the wild-type. In contrast, the flgE mutant of 630Derm adhered less strongly than the wildtype, albeit the latter was not statistically significant (Fig. S3). We extended this investigation of adherence by examining the role of flagella in the epidemic strain R20291 for the first time. The adherence of the flagella mutants, paralyzed mutant, wild-type and complemented strains to Caco-2 cells was assayed (Fig. 3). All flagella mutants were less effective in adherence than the wild-type. In particular the fliC, fliD and flgE-mutant all adhered less than the paralyzed motB-mutant, which still adhered significantly less than the wild-type (p,0.05). These results indicate an important role for the flagella of R20291 in adherence and, moreover, indicate that the intact flagellar structure possesses adhesive properties that are independent of flagellum-mediated motility. These findings are in contrast to results obtained with strain 630Derm (this study and [22]), which adheres less effectively than the corresponding fliC and fliD flagella mutants. It is important to note that we did not observe any difference in growth kinetics between wild-type, mutants and complemented strains in vitro (data not shown).

Assessment of importance of the flagellum of R20291 in intestinal colonisation of gnotobiotic mice
As flagella in R20291 seem to play a role in adherence we tested the wild-type, the fliC-mutant (CRG3351) and the paralyzed flagellum mutant (CRG1987) in vivo using the monoxenic mouse model for a single infection study and thereafter the dixenic mouse model in a competition experiment. We chose not to put the complemented strains through our animal models as plasmids are known to be unstable in C. difficile if antibiotic selection is not maintained. Results would hence be doubtful and as such these experiments were deemed unethical.
For our individual mouse infection study, we used a monoxenic mouse model. All 6 mice infected with the wild-type R20291 survived the entire experimental duration (7 days) without succumbing to C. difficile disease. Colonisation was monitored through faecal shedding kinetics and at the end of the experiment adherence to the caecum was measured. Surprisingly, 70% of mice infected with the fliC-mutant CRG3351 succumbed to C. difficile disease by four days (Fig. 4A). A comparison between the wild-type and the paralyzed mutant (motB(D23A)-) revealed no difference in faecal shedding (Fig. 4B) and significant difference in adherence to the caeca (Fig. 4C). These findings suggest that flagella-associated motility is not needed for successful colonisation and adherence in mice, although the structure of the flagellum itself may be important as an adhesin. Further investigation is necessary to understand the increased virulence of the fliC-mutant.
To confirm the results obtained in single infection, and to further investigate the advantage for R20291 to possess flagella, we co-challenged mice with a 1:1 inoculum of wild-type and fliCmutant. The wild-type clearly out-competed the non-motile ) and adherence to the caeca, albeit the latter was not significant (Fig. 5B). We then cochallenged mice with the fliC-mutant and the paralyzed mutant (1:1). The paralyzed mutant out-competed the fliC-mutant in colonisation and adherence ( Fig. 5C and 5D). These in vivo experiments confirm the role of flagella as an adhesin during colonisation and demonstrate that flagellum-mediated motility is neither important nor essential for virulence of C. difficile R20291.
Taken together these in vitro adherence and in vivo intestinal colonisation studies provide the first strong evidence that by disabling the motor of the flagellum, the structural components of the flagellum of C. difficile R20291 rather than active motility, is needed for adherence and colonisation of the intestinal epithelium during infection.

Comparison between the epidemic strain R20291 and non-epidemic isolate 630Derm
Previous flagella studies in C. difficile focused on the nonepidemic strain 630Derm. As epidemic strains, particularly those from PCR-Ribotypes 027 and 078 are becoming increasingly more common, it is important to ascertain whether they behave the same as 630Derm. In this study we observed differences between the two parental strains used, which is consistent with the fact that differences in the flagella regions of R20291 and 630 have been described previously [24]. In contrast to the peritrichiously flagellated 630Derm, R20291 only displays a single flagellum and is comparatively less motile in motility assays. This can clearly be seen in figure 2A and B. When tested in adherence assays in vitro, 630 and 630Derm adhered significantly less to Caco-2 cells than R20291. Interestingly flagella mutants of R20291 showed decreased adherence whereas the same mutants in a 630 background adhered more than the wild-type ( Fig. 3 and S3).
In vivo colonisation of C. difficile 630Derm and the fliCmutant CRG3357 Reminiscent of our murine infections with strain R20291 and the derived fliC mutant, it has been reported that hamsters succumb to disease quicker when infected with a fliC or a fliDmutant of 630Derm [22]. Although, Dingle et al. [22] measured higher toxicity levels with the mutants than with the parental strain, they did not analyse colonisation over time within the hamster. We compared the parental strain 630Derm to the fliCmutant, CRG3357, in three different mouse models, which are traditionally used to investigate colonisation over time and adherence. Firstly we infected groups of six human-microbiotaassociated (HMA) mice with the respective strains. This model is closest to humans as the implanted microbiota is disturbed by antibiotics prior to the bacterial challenge. We did not find any difference between the colonisation kinetics or adherence to the caecum between the two groups (Fig. S4). The microbiota might mask some subtle differences between wild-type and mutant. We subsequently tested the strains in the monoxenic mouse model in a single infection study. Again the fliC-mutant colonised and adhered at the same level as the wild-type (Fig. S4) leading to the conclusion that flagella are not required for strain 630Derm to colonise mice. In order to detect any subtle differences between the strains we undertook a co-infection study, giving the mice a 1:1 inoculum of wild-type and fliC-mutant. Although the kinetics remained generally equivalent, there were minor differences at several time points (Fig. S4), indicating a fitness advantage of the wild-type. Furthermore the fliC-mutant adhered significantly less to the mouse caecum than the wild-type (Fig. S4). In conclusion, these results show that flagella and motility are not required for C. difficile 630 to colonise mice; however, flagellum motility may contribute to the fitness of the bacterium and allow the wild-type strain to outcompete the fliC-mutant. Interestingly, other bacteria exhibit a similar contribution of flagella to their infection pathway. For example, competition of motility mutants with wild-type H. pylori during mixed infections results in greater attenuation than observed in independent challenges in mice. Moreover, a fliC mutant of UPEC has previously been shown [27,28] to be at a competitive disadvantage during mixed infection with the parental strain in a dixenic mouse model, whilst the mutant and the wildtype colonised to levels that were not significantly different during an independent challenge.
Toxin production of 630Derm and flagella mutants in comparison to R20291 and flagella mutants Since increased toxicity was noted in the fliCand fliD-630Derm flagella mutants [22], we measured the toxicity of a flgE-mutant in 630Derm (Fig. S5). In contrast to the fliC and fliD-mutant, the flgEmutant showed reduced toxicity compared to the wild-type; this was however not statistically significant. Interestingly no difference in cytotoxicity was observed when comparing R20291 and its flagella mutants (data not shown). We performed glutamate dehydrogenase (GDH) assays in order to investigate whether increased cell lysis is responsible for more toxin being released in the fliC and fliD-mutants of 630Derm. GDH activity did not show any significant difference in any of the strains (data not shown) leading to the conclusion that cell lysis cannot explain the altered levels of toxins in the culture supernatant of the flagella mutants. In order to investigate whether the increased amount of toxicity in the mutants was due to altered levels of expressed toxin genes we performed qRT-PCR comparing the relative expression of tcdA (encoding for TcdA) in each flagellar mutant (fliD, fliC and flgE) to the parental strain 630Derm. Total cellular RNA was extracted during late exponential to early stationary phase of growth, after 16 h. The results of qRT-PCR analysis showed that the expression of the tcdA gene was strongly up-regulated in the fliC and fliDmutants. The expression of tcdA was on average 44.1-fold greater in the fliC mutant and 7.4-fold greater in the fliD-mutant than in the wild-type strain (Fig. 6). These results confirmed that the high levels of toxin in the fliD and fliC-mutants were indeed due to increased expression of the tcdA gene. In the case of the flgEmutant tcdA expression was on average 10-fold lower compared to the wild-type strain (Fig. 6).
Taken together, these observations provide experimental evidence that inactivation of flagellar cap (fliD), flagellin (fliC) and hook (flgE) genes have had an influence on the expression of tcdA. However, the mechanism underlying this regulatory relationship between the synthesis of flagella structure and regulation of toxin production is as yet unknown. The bacterial flagella system is tightly regulated in an ordered cascade, which uses a series of intermediate assembly stages as checkpoints. This tight regulatory control is required because the production of flagella and rotation are a metabolically costly undertaking for the cell, in which as many as 20,000 to 30,000 flagellin subunits are produced and their secretion outside of the cell and subsequent polymerization requires about 2% of the cell's energy expenditure. In flagella mutants, this metabolic energy is saved and can perhaps be used for other cell functions, such as production of TcdA and TcdB. On the other hand, we observed a decrease of toxin expression level in the flgE-mutant, which suggests a regulatory mechanism between flagella and toxin expression at least in 630Derm. Moreover, the GDH enzyme assay confirmed that the increase of toxin level in supernatants of the fliD and fliC-mutants was not due to an increase in cell lysis.
Recently a paper was published investigating the modulation of toxin transcription by the flagellar regulon [29]. The authors compared a fliC-mutant in 630Derm to the parental strain and a set of early flagella genes including fliF, fliG and fliM. As before they found that a fliC-mutant produces more toxin than the wild-type and in line with our findings the authors observed upregulation of tcdA expression. All the mutants in the early flagella genes showed a down-regulation of toxicity and tcdA expression comparable to our finding with the flgE-mutant. Furthermore they showed that these mutants exhibit less virulence in the hamster infection model. In addition to the transcription of tcdA they also examined tcdB, encoding TcdB, by qRT-PCR. Changes were more subtle, but might yet be important given the toxic effect of TcdB.

Conclusions
This study compares the effect of flagella and flagellummediated motility between the non-epidemic C. difficile strain 630Derm and the UK outbreak strain R20291. R20291 produces  only a single flagellum whereas 630Derm is peritrichously flagellated. It has been shown previously that flagella mutants in 630Derm [22] adhere more strongly to Caco-2 cells than the parental strain. In contrast, here we show that flagella mutants of R20291 adhere less strongly in vitro than the parental strain. Mice experiments using 630Derm wild-type and flagellar mutants revealed that flagella are not required for adherence and colonisation in this strain, but flagella-mediated motility might contribute to an overall fitness of the bacteria. An increase in toxicity was noted in strains in which the late flagellar genes (fliC and fliD) of 630Derm were mutated, a phenotype which was not apparently due to increased cell lysis. We confirmed by qRT-PCR that expression of the tcdA gene was up-regulated in these mutants. The flgE-mutant of 630Derm showed lower cytotoxicity and also a decreased expression of tcdA. These results were recently confirmed in an independent study [29] which analysed a fliCmutant, as well as mutants in early flagellar genes. In the case of R20291, however, we did not see any change in toxicity between the wild-type and the flagellar mutants. Mice experiments showed that flagellum-mediated motility is not necessary for colonisation and adherence; however a paralyzed flagellum mutant showed that the filament structure acts as an adhesion. More work is needed to understand the involvement of the flagella regulon of R20291 in toxin expression. Our study shows that there are significant differences between different strains of C. difficile and that it is important to study phenomena in several strains before drawing general conclusions. As previously mentioned, differences in glycosylation might play an important role and go some way to explaining differences seen between aspects of motility, adherence and colonisation of different strains of C. difficile. It is also important to note that there are toxigenic, non-flagellated strains of C. difficile, for example PCR-ribotype 078, an epidemic strain with increasing importance. Interestingly these strains still possess the late flagellar genes, which might play a role in regulation.

Bacterial strains and growth conditions
All bacterial strains and plasmids used in this study are described in Table 1. C. difficile was routinely cultured on brain

General molecular biology techniques
DNA manipulations were carried out according to standard techniques [33]. For DNA cloning and PCR analysis, C. difficile genomic DNA was prepared using quick Chelex resin-based or phenol-chloroform extraction techniques. DNA was purified from agarose gels using the QIAquick gel extraction kit (Qiagen, UK). Plasmids were isolated from E. coli using the plasmid mini-prep kit (Qiagen, UK) according to the manufacturer's instructions. PCR amplification was performed using Failsafe high fidelity DNA polymerase (Merck, UK) or Taq polymerase (Sigma) in accordance with the manufacturers' protocols.
Construction of flagellar-associated mutants in C. difficile 630Derm and R20291 Flagellar target genes were insertionally inactivated in C. difficile 630Derm and R20291 using the ClosTron gene knock-out system as described previously [23,25]. The mutants were verified by PCR screening and subsequent DNA sequencing using the primers shown in Table S1.

Complementation of the flagellar mutants
For complementation of the fliC mutant, a 973-bp fragment encompassing the fliC open reading frame (873 bp) with the 100-bp 59 noncoding region which likely encompasses the fliC promotor was amplified as a full-length DNA fragment with primers NotI-pfliC-630-F and XhoI-fliC-630-R and cloned using NotI and XhoI into the expression vector pMTL84151 to generate pMTL-SB1. The complementation primers were designed to allow cleavage of the fliC promoter and fliC gene as a full-length fragment. As the proteins encoded by fliC, fliD, flgE and motB are the same in 630Derm and R20291, only one plasmid (pMTL-SB1, pMTL-SB2, pMTL-SB3 and pMTL-SB4) was constructed for use in both mutants, respectively. To complement the fliD mutant, the fliC promoter was joined to the fliD gene by using splicing by overlap extension (SOEing) PCR [34] with the following primers: NotI-pfliC-R20291-F/NdeI-pfliC-fliD-R-SOE and NdeI-pfliC-fliD-F-SOE/XhoI-fliD-R to create two PCR products that were joined in a second step using the primers: NotI-pfliC-R20291-F/XhoI-fliD-R. The final product was cloned into the modular expression vector pMTL84151 resulting in pMTL-SB2. Likewise, SOEing PCR was used to construct a complementation plasmid with flgE (pMTL-SB3). For complementation of the motB (D23A) mutant, the 696bp motB gene was amplified by standard PCR and cloned directly into the modular vector pMTL84152 (pMTL-SB4). All complementation plasmids were transferred into C. difficile via conjugation and transconjugants were verified by PCR.

Southern blot
Mutants were verified by Southern blot using an intron specific probe. Genomic DNAs (2 mg) were digested with HindIII (NEB) overnight. The blots were carried out using a DIG high prime labelling and detection kit (Roche) according to the manufacturer's instructions.

Motility assays
Swarming and swimming motility behaviours of C. difficile were studied by performing swarming and swimming motility assays as described [35] with minor modifications. Briefly, cultures of C. difficile were grown to mid-exponential phase for 8 h in BHI broth under anaerobic condition at 37uC. Motility plates were prepared by adding 25 ml of agar media into each plate and let to set for 2 h for swarm agar plates and overnight for swim agar plates. Soft agar plates were then air-dried for 15 min and transferred into the anaerobic workstation and allowed to pre-reduce for 4 h prior to inoculation with C. difficile strains. The swimming agar was made of BHI broth medium (37 g L 21 ) containing 0.3% (w/v) Difco bacto-agar and the swarming agar with 0.4% agar. To assay swimming plates were stab-inoculated with 3 ml of a midexponential growth culture of each strain at the same cell density and then incubated at 37uC for 48 h. Swimming motility was quantitatively determined by measuring the radius. To assay swarming plates were spot-inoculated with 3 ml of a midexponential growth culture of each strain and incubated as described above. Swarming motility was quantitatively determined by measuring the radius of the swarming zone. Motility assays were performed in six replicates for each strain and repeated independently three times.

Transmission Electron Microscopy (TEM)
Electron microscopy was performed to examine the presence of flagellar structures on the surface of C. difficile. Cells were negatively stained using 0.5% Uranyl acetate (pH 4.5, Agar Scientific) on carbon formvar copper 200-mesh grids (Agar Scientific Ltd., UK). Briefly, cells from a mid-exponential C. difficile culture were adsorbed onto Formvar-coated copper grid for 5 min and the excess was carefully removed with blotting paper. This preparation was then fixed on the grid with 1% glutaralde-

Cell culture
The human intestinal epithelial cell line Caco-2 and Vero cells (derived from African Green monkey kidney) were obtained from the American Type Culture Collection (ATCC, via Health Protection Agency, UK). Caco-2 and Vero cells were maintained at 37uC in 5% CO 2 atmosphere in Dulbecco's modified Eagle's medium (DMEM) containing 10% (v/v) heat-inactivated foetal bovine serum (FBS), 1% (v/v) nonessential amino acids (NEAA) and 1% (v/v) antibiotic-antimycotic solution (containing penicillin [1 U ml 21 ], streptomycin [1 g ml 21 ], and amphotericin B as fungizone [2.5 mg ml 21 ]). For cell adherence assays, 12-well tissue-culture plates were seeded with 1610 6 Caco-2 cells per well and cells were grown for 9 days to obtain differentiated confluent monolayers. For cell cytotoxicity assay, 96-well tissue-culture plates was seeded at a density of approximately 0.25610 5 Vero cells per well and cells were grown for 48 h to obtain a confluent monolayer.

Caco-2 cell adherence assay
The adherence of the C. difficile strains to differentiated human colonic enterocyte-like Caco-2 cells was assessed in the in vitro cell-culture adherence model under anaerobic condition using the protocol described recently by Barketi-Klai et al. [36] with the following modifications: 16 h cultures of C. difficile were harvested by centrifugation at 40006 g for 3 min. Cell pellets were resuspended gently in 1 m of the cell-line culture medium. Caco-2 cells were infected with 1610 8 cells ml 21 C. difficile by adding 1 ml of culture to each well of the 12-well tissue-culture plate and plates were then incubated for 1.5 h at 37uC under anaerobic condition. Initial bacterial counts were determined by plating onto selective BHI plates prior to infection. After incubation, monolayers were washed four times with sterile PBS to remove non-adherent bacteria. Then 1 ml of saponin 1% was added to each well, incubated for 10 min and attached bacteria were detached from cell monolayer by repeated pipetting. Bacterial counts at post-infection were determined by plating serial dilutions on selective BHI agar plates in triplicates. Colonies were counted after 24 h incubation. In parallel, uninfected monolayers (negative control) were collected by trypsinization and counted by trypan blue staining in order to express the adherence results as number of viable adherent CFU per one Caco-2 cell. Each adherence assay was performed in triplicate, and repeated at least three times in its entirety.

Vero cell cytotoxicity assay
The kinetic of toxin secretion was examined in culture supernatants of C. difficile strains at different time points during growth (determined from early exponential to late stationary phase of growth) using Vero cell cytotoxicity assay as described previously by Kuehne et al. [6].

Production of C. difficile culture supernatant filtrates
For the Vero cell cytotoxicity and GDH enzymatic activity determination assays, C. difficile strains were grown as follows. Overnight cultures in TY were used to set up a new starter culture using 1/100 inoculum. 2 ml of each C. difficile culture was collected at different time points during bacterial growth after recording OD 600 . Cells were harvested by centrifugation at 150006g for 10 min at 4uC and the supernatant was filter-sterilized through 0.2 mm syringe filters (Sartorius, Germany). Filtered supernatant samples were then immediately frozen at 220uC to be used later for these assays.

Preparation of C. difficile crude-cell extracts
For the glutamate dehydrogenase (GDH) enzymatic assay, whole cell protein extracts (soluble cytoplasmic fraction) were prepared at different time points during bacterial growth as follows: cell pellets from 2 ml C. difficile culture were harvested by centrifugation at 130006g for 10 min, and lysed by resuspending in 200 ml Bugbuster 106 protein extraction buffer (Novagen UK) containing 2 ml of Benzonase Nuclease reagent (25 U ml 21 , containing 50 mM tris-HCl, 20 mM NaCl, and 2 mM MgCl 2 , pH 8.0; Novagen UK) and 10 ml of Lysozyme solution (100 mg ml 21 ; Novagen UK). After 1 h incubation at room temperature with agitation, debris was pelleted by centrifugation at 130006g for 20 min at 4uC and the crude cell extracts transferred to a new microfuge tube and stored at 220uC for later use.
Glutamate dehydrogenase enzyme assay GDH assay was performed using a previously described methodology [37] with minor modifications. Briefly, the GDH enzyme activity in C. difficile culture supernatant and soluble fraction was measured at 340 nm at 25uC with reaction mixture containing 0.333 ml of 300 mM Potassium Phosphate Buffer (pH 8.0), 0.167 ml of 300 mM L-glutamic acid (pH 7.5), 0.15 ml of 1 mM b-NAD, 0.333 ml of deionized water, and 0.017 ml of test sample. GDH enzyme (Sigma Chemical Co., UK) was used as the positive control. The reagents were added in the following order: Potassium Phosphate Buffer; L-glutamic acid; b-NAD; deionized water and finally the reaction was started by adding 6 U ml 21 of GDH [EC 1.4.1.3] as positive control or tested sample. The GDH activity was measured spectrophotometrically by recording the increase in absorbance at 340 nm for 5 min and the total absorbance was obtained by using the maximum linear rate for both tested samples and controls. GDH activity was calculated using the molar extinction coefficient for NAD(P) as 6.22 mM 21 6cm 21 .
Enzyme specific activity was defined as one unit of GDH enzyme required to oxidate 1.0 mmole of L-glutamate to aketoglutarate per min using NAD(H) or NADP(H) as cofactors, at pH 8.0 and 25uC.

RNA extraction and reverse transcription
These procedures were carried out essentially as described previously by Lyras et al. [5] with the following modifications. Total RNA was extracted from C. difficile grown to early stationary phase (16 h) in TY broth in four independent replicates for each strain. RNA stabilization was achieved by using RNAprotect Bacteria Reagent and an RNeasy Mini Kit (Qiagen) as described by the manufacturer's instructions. RNA isolation was then carried out using the FastRNA pro Blue extraction Kit (MP Biosciences), followed by phenol-chloroformisoamyl alcohol (25:24:1) treatment and precipitated with ethanol according to the manufacturer's instructions. The total-RNA was dissolved in 50 ml of RNase-free water (DEPC-treated water) and DNase I digestion was then carried out to remove contaminating DNA in RNA samples by adding TURBO DNase buffer and TURBO DNase (Ambion) and incubated at 37uC for 1 h, according to the manufacturer's instructions. The RNA was purified using the RNeasy spin column purification kit (QIA-GEN) according to the manufacturer's instructions. RNA samples were eluted from the spin column in two volumes of 20 ml of DEPC-treated water. The concentration of RNA samples was measured using the nanodrop spectrophotometer and stored at 280uC. The integrity and quality of extracted C. difficile RNA was determined using the Prokaryote Total RNA Pico assay by the Agilent 2100 bioanalyzer according to the manufacturer's instructions before use in RT-PCR. The absence of contaminating DNA in extracted RNA samples was confirmed by PCR with fliC gene specific primers. First-stranded cDNA synthesis was carried out by reverse transcription on 2 mg template RNA using Omniscript RT kit (Qiagen, UK) with random hexamer primers (Promega, UK) according to manufacturer's specifications. The cDNA samples were then purified using Qiaquick columns (Qiagen, UK).

qRT-PCR (reverse transcriptase) analysis of tcdA gene expression
Real-time quantitative reverse transcriptase PCR (qRT-PCR) experiments were performed using the AB7500 cycler real-time PCR instrument. The qRT-PCR primers were designed and reactions were carried out using the SYBR Green Master Mix (Ap Biosciences, UK) with cDNA as template, as described by Lyras et al. [5]. PCR were performed on duplicate cultures and results were evaluated using rpoA as the endogenous control.

Gnotobiotic mouse models
Animal care and animal experiments were carried out in strict accordance with the Committee for Research and ethical Issues of the International Association for the study of pain (IASP). The animal experimentation protocol was approved by the Animal Welfare Committee of the Paris Sud University and animal experiments were performed according to the University Paris Sud guidelines for the husbandry of laboratory animals. C3H/HeN germ-free (6-8 weeks old) and human microbiota-associated mice, obtained from INRA of Jouy-en-Josas (ANAXEM, France), were housed in sterile isolators provided with sterilized bedding and received standard nutrients sterilized by irradiation, and water sterilized by autoclaving. Prior to infection with C. difficile strains, sterility of axenic mice was confirmed by testing faecal culture of all tested animals. Faecal pellets were collected from mice, homogenised and serial dilutions were plated on BHI agar plates. Following 24 h incubation, plates showing no growth of bacteria indicated that tested animals were considered germ-free. Cultures of C. difficile for challenge were prepared for each mouse model according to the method described recently by Barketi-Klai et al. [36].
For the independent challenges (monoxenic mouse model), single infections of 1610 8 C. difficile were performed by oral gavage route on groups of 6 axenic mice. For the co-challenge (competition colonisation assay) (dixenic mouse model), mice were co-infected by oral gavage with 1610 8 bacterial CFU comprised of 1:1 ratio of wild-type and mutant. Human Microbiota Associated mice (HMA) were dosed with 3 mg of amoxicillin/clavulanic acid (8/1: v/v) by oral gavage on 8 consecutive days to disrupt the intestinal microbiota. 24 h after the last antibiotic administration, each mouse was challenged with 1610 8 bacterial CFU. In these mouse models of colonisation, two parameters were studied as a surrogate for determining the level of C. difficile intestinal colonisation, these were: (1) Intestinal implantation (C. difficile faecal shedding).
The concentration of C. difficile CFU recovered in faeces during the course of infection (kinetic of faecal shedding) was determined by homogenising and plating on C. difficile selective agar medium (supplemented with antibiotics, where appropriate). (2) Caecal adherence. The C. difficile caecal colonisation was studied by determining the level of C. difficile associated to the caecal wall of mice. Seven days after C. difficile challenge, mice were euthanized, dissected to excise the entire caecum of each mouse. Each caecum was rinsed in sterile PBS and weighed. The caecum of each mouse was harvested by homogenising using the ultra-Turrax apparatus (IKA-Labortechnik, Staufen, Germany) for 1 min at 120006 g using and diluted in PBS to obtain a final concentration of 10 mg ml 21 . Ten-fold serial dilutions were then cultured on BHI agar plates (supplemented with antibiotics, where appropriate). Colonies were counted after 48 h incubation at 37uC under anaerobic condition.

Statistical analysis
Statistical analysis was performed using StatEL and GraphPad Prism software. Data were analysed by Mann-Whitney test (mice experiments), Student's t-test, one-way analysis of variance ANOVA, followed by Dunnett's multiple comparison test (cytotoxicity and GDH assays). A statistically significant difference was considered to be p values of ,0.05.