Uropathogenic Escherichia coli AL511 requires flagellum to enter renal collecting duct cells


*E-mail clb@pasteur.fr; Tel. (+33) 1 40 61 32 80; Fax (+33) 1 40 61 36 40.


Escherichia coli is the leading cause of urinary tract infections, but the mechanisms governing renal colonization by this bacterium remain poorly understood. We investigated the ability of 13 E. coli strains isolated from the urine of patients with pyelonephritis and cystitis and normal stools to invade collecting duct cells, which constitute the first epithelium encountered by bacteria ascending from the bladder. The AL511 clinical isolate adhered to mouse collecting duct mpkCCDcl4 cells, used as a model of renal cell invasion, and was able to enter and persist within these cells. Previous studies have shown that bacterial flagella play an important role in host urinary tract colonization, but the role of flagella in the interaction of E. coli with renal epithelial cells remains unclear. An analysis of the ability of E. coli AL511 mutants to invade renal cells showed that flagellin played a key role in bacterial entry. Both flagellum filament assembly and the motor proteins MotA and MotB appeared to be required for E. coli AL511 uptake into collecting duct cells. These findings indicate that pyelonephritis-associated E. coli strains may invade renal collecting duct cells and that flagellin may act as an invasin in this process.


Uropathogenic Escherichia coli (UPEC) are responsible for 80–90% of urinary tract infections (UTI) in humans (Foxman and Brown, 2003; Zhang and Foxman, 2003; Bagshaw and Laupland, 2006). Many UTI are asymptomatic but some UPEC strains cause significant clinical symptoms, ranging from pain in uncomplicated cases of cystitis to sepsis in cases of pyelonephritis, and antibiotic treatment may be difficult if the strains concerned are multiresistant (Khouri et al., 2006). The severity of infection may depend on both bacterial virulence and the efficiency of the host immune response (Svanborg et al., 2001).

Uropathogenic E. coli strains isolated from patients with cystitis or pyelonephritis have several pathogenic features in common, including type 1 fimbriae and flagella. Flagella have recently been reported to be involved in the interaction of various pathogenic E. coli strains with epithelial cells. They consist principally of flagellin, a protein for which 56 different variants defining the H antigens have been identified. Flagellins of the H6 and H7 subtypes have been shown to act as adhesins, mediating the interaction of enteropathogenic and enterohaemorrhagic E. coli strains with epithelial cells of the colon (Erdem et al., 2007). Flagella have also been reported to be essential for meningitis-associated E. coli, for the invasion of brain microvascular epithelium cells (Parthasarathy et al., 2007). The mechanism by which UPEC strains colonize the urinary tract has been correlated with the presence of the flagellum, the organelle responsible for motility (Lane et al., 2005; 2007; Wright et al., 2005), but the precise role of the flagellum in UTI development remains unknown. Type 1 fimbriae have clearly been implicated in the entry of UPEC strains into the bladder. Bacteria establish intracellular communities in bladder cells, which may protect them against antibiotic treatment and the host immune response, facilitating the development of persistent infections (Bergsten et al., 2005). The expression of type 1 fimbriae by UPEC strains has been reported to be repressed in the kidneys, and no role for these adhesins in the development of pyelonephritis has been documented.

The bacteria causing pyelonephritis frequently express additional adhesin-encoding operons and toxin genes (Kaper et al., 2004). P fimbriae encoded by the pap (pyelonephritis-associated pili) operons are found in most E. coli strains isolated from the urine of patients with pyelonephritis. The cytotoxic necrotizing factor CNF1 is produced by one-third of all pyelonephritis strains and may be involved in kidney invasion. In vitro, this protein is secreted by E. coli and stimulates actin stress fibre and membrane ruffle formation in a Rho GTPase-dependent manner, resulting in the entry of the bacterium into cells. However, its effects on invasion processes during pyelonephritis remain unclear and are a matter of debate (Bower et al., 2005). There are several lines of evidence to suggest that adhesins of the Afa (afimbrial) family expressed in UPEC strains are involved in UTI (Le Bouguénec, 2005). UPEC strains expressing these adhesins have a unique renal tissue tropism (Nowicki et al., 1988; Lalioui and Le Bouguénec, 2001). Clinical and experimental findings suggest that E. coli strains with Afa adhesins have properties potentially favouring the establishment of chronic and/or recurrent infection (Foxman et al., 1995; Le Bouguénec, 2005).

The collecting duct (CD) is the last segment of the renal tubule and the first tissue encountered by bacteria ascending from the bladder to the kidneys. We recently showed, in a mouse model of ascending pyelonephritis, that UPEC strains adhere to CD cells, thereby activating inflammatory responses mediated principally by TLR4, which recognizes the lipopolysaccharide of Gram-negative bacteria (Chassin et al., 2006). Intracellular bacteria have been observed in CD cells in renal biopsy specimens from patients with pyelonephritis (Iványi et al., 1988). These observations led to the hypothesis that bacteria may enter renal epithelial cells in the CD, favouring persistence within the kidney. We investigated the molecular basis of this interaction, by investigating the ability of various E. coli strains to interact with CD epithelial cells, using the mpkCCDcl4 cell line, an established murine CD cell line (Bens et al., 1999). These cells have the same principal properties as the parental CD cells from which they were derived, including the expression of a functional TLR4 (Chassin et al., 2006).

We first analysed and compared the capacities of 13 E. coli strains of various origins to interact with mpkCCDcl4 cells, with the aim of defining a study model for characterizing the interaction of E. coli strains with CD cells. Electron microscopy and association and invasion assays showed that some clinical isolates from pyelonephritis patients not only adhered to renal CD cells, but also entered these cells. Isolate AL511 was found to be the most invasive of the isolates studied. We then characterized the process by which this isolate entered mpkCCDcl4 cells, focusing particularly on the role of the flagella. We showed that AL511 was able to survive within cells and that the uptake of this isolate was dependent on the flagella of the bacterium.


Gentamicin protection assays show that clinical UPEC strains invade renal CD cells

We analysed the interactions of E. coli strains with mpkCCDcl4 cells (Table 1). As the interaction of E. coli strains with uroepithelial cells has been reported to be mediated by type 1 pili and mannose-resistant adhesins, we selected a set of 13 strains expressing the genes encoding these adhesins in the conditions used for bacterial culture (Table S1). This set of strains included 11 ExPEC strains, eight of which were associated with pyelonephritis and were isolated from the urine of patients admitted to five different French hospitals between 1985 and 1989. All these strains belong to phylogenetic group B2 or A and carry various virulence genes. We also used the well characterized J96 pyelonephritis isolate and two cystitis isolates previously shown to enter bladder cells (Mulvey et al., 2001; Wright et al., 2005). Two commensal strains, including a faecal isolate carrying ExPEC virulence genes and the well-characterized MG1655 strain, were also used. Preliminary results for Giemsa staining indicated that nine strains adhered to mpkCCDcl4 cells without damaging them. These strains comprised five pyelonephritis strains, two cystitis strains and two commensal strains (data not shown).

Table 1. E. coli strains and plasmids used in this study.
NameIsoaPgbScRelevant characteristicsdMotilityeReference
  • a. 

    Human clinical isolate (Iso): pyelonephritis (Pyl.), cystitis (Cys.), commensal (Com.) from healthy individuals, or laboratory strain (LS).

  • b. 

    Phylogenetic group (Pg).

  • c. 

    Known serotype (S).

  • d. 

    Detection by PCR of f17A, afaC-8, papC, fimH, hly, cnf genes, used for detection of operons encoding F17, Afa-VIII, type P and type 1 adhesins, haemolysin α and cell necrotizing factor 1 toxins respectively (Le Bouguénec et al., 1992; 2001; Lalioui et al., 1999; Usein et al., 2001; Girardeau et al., 2003).

  • e. 

    Percentage motility for each strain relative to E. coli strain AL511, used as the reference (100%).

  • N.D., not determined.

E. coli strains
 AL511Pyl.AO9:H12:F11f17, afa-8, pap, fim100%Girardeau et al. (2003)
 AL511 ΔfimH   fimH::zeo (Zeor)34%This study
 AL511 ΔafaC-8   afaC-8::zeo (Zeor)67%This study
 AL511 ΔpapC   papC::zeo (Zeor)40%This study
 AL511 Δf17C   f17C::zeo (Zeor)89%This study
 AL511 ΔfliD   fliD::zeo (Zeor)8%This study
 AL511 ΔfliC   fliC::zeo (Zeor)5%This study
 AL511 ΔfliC   fliC::zeo, pZeKmfliC (H12) (Kmr, Zeor)28%This study
 AL511 ΔfliC   fliC::zeo, pZeKmfliC (H48) (Kmr, Zeor)95%This study
 AL511 ΔfliC   pZEKmGFP (Kmr)5%This study
 AL511 ΔmotAB   motAmotB::zeo7%This study
 AL10Pyl.AO101:H4afa-8, pap, fim155%Girardeau et al. (2003)
 AL11Pyl.B2N.D.pap, fim74%Laboratory Collection
 AL46Pyl.AN.D.pap, fimN.D.Laboratory Collection
 AL48Pyl.B2H1pap, fim10%Laboratory Collection
 AL213Pyl.AO+:H9f17, afa-8, pap, fim10%Girardeau et al. (2003)
 AL500Pyl.B2N.D.pap, fimN.D.Laboratory Collection
 AL645Pyl.B2N.D.f17, fim, hlyN.D.Laboratory Collection
 J96Pyl.B2O4:H5pap, fim, hly, cnfN.D.Svenson et al. (1996)
 NU14Cys.B2O18:H7f17, pap, fim, hly, cnf9%Hultgren et al. (1986)
 UTI89Cys.B2O18:H7pap, fim, hly, cnf170%Chen et al. (2006)
 183Com.AO101:H4afa-8, pap, fim55%Laboratory Collection
 MG1655Com.AO16:H48fim85%Laboratory Collection
 MG1655 ΔfliC   fliC::FRT6%Laboratory Collection
 MG1655 ΔfliC   fliC::FRT, pZEKmfliC (H12) (Kmr)50%Laboratory Collection
 MC1061LS.    Laboratory Collection
 pKOBEG-Apra   (Aprar) Link et al. (1997)
 pZE21-MCS2   (Kmr) Lutz et al. (1997)
 pZEKmfliC (H12)   fliC gene from AL511 inserted into pZE21-MCS2 This study
 pZEKmfliC (H48)   fliC gene from MG1655 inserted into pZE21-MCS2 This study
 pZEKmGFP   gfp gene inserted into pZE21-MCS2 Gift from Sandra Da Ré

We quantified association with and entry into mpkCCDcl4 cells by bacteria, by carrying out gentamicin protection assays. These assays were used to assess association with cells after 3 h of infection (binding to the cell surface or present within the cell) and invasion (intracellular bacteria only) after incubation with gentamicin for another 2 h to kill adherent bacteria. The data obtained in these experiments are expressed as a percentage of the values obtained for the AL511 strain, the most invasive strain tested (Fig. 1). Numeric data are available in Table S2. The percentage values obtained for association and invasion were considered significantly different from those for AL511 if they were at least 90% lower. Based on these criteria, the AL10 pyelonephritis strain was considered to be a poorly adherent, non-invasive E. coli strain. The two commensal strains (183 and MG1655), the two pyelonephritis isolates (AL11, AL213), and the cystitis strain NU14 were found to associate with cells but to be non-invasive. Only the pyelonephritis strain AL48 and the cystitis isolate UTI89 showed low-level, but nonetheless significant levels of invasion (Fig. 1). However, the standard deviation was also large for the UTI89 isolate, suggesting that this strain may not be highly invasive. Most of the UPEC strains interacted with mpkCCDcl4 cells, but considerable differences in the ability to invade cells were observed between strains.

Figure 1.

Interaction of E. coli strains with mpkCCDcl4 cells. Relative association (white bars) and invasion (black bars) values and their respective standard deviations, obtained from the values presented in Table S1 (AL511 as 100%).

We then analysed the ability of the invading E. coli AL511 strain to persist in mpkCCDcl4 cells. The number of viable intracellular bacteria remained constant until 48 h, suggesting that AL511 was able to survive within cortical CD cells (Fig. 2). However, experiments over longer periods of time revealed a decrease in the number of viable internalized bacteria between 48 and 72 h, probably due to the death of mpkCCDcl4 cells, as observed under a light microscope.

Figure 2.

Intracellular persistence of E. coli AL511 in mpkCCDcl4 cells in vitro. Percentage of bacteria internalized at various times after infection, expressed with respect to the value obtained for 2 h of gentamicin treatment (100%).

Ultrastructural analyses of E. coli AL511 adhesion to and uptake into mpkCCDcl4 cells

Confluent mpkCCDcl4 cells formed a polarized monolayer of closely apposed cells and dome-shaped structures with short microprojections on their apical surface (Fig. 3A and B). CD cells are composed of principal and intercalated cells (Madsen and Tischer, 1986). As previously reported (Bens et al., 1999), the presence of both cell types in mpkCCDcl4 cell cultures was confirmed by the heterogeneous staining of the intercalated cells with FITC-conjugated Dolichos biflorus agglutinin (data not shown).

Figure 3.

Scanning (SEM) (A–D) and transmission (TEM) (E and F) electron micrographs of mpkCCDcl4 cells infected or not infected with the AL511 strain.
A. View of uninfected apical surface of mpkCCDcl4 cells. The monolayer displays a planar surface with microprojections (inset).
B. View of mpkCCDcl4 cells infected with E. coli AL511. Bacteria are homogeneously distributed along the epithelial cell surfaces and interact with the microprojections (inset).
C and D. Infected mpkCCDcl4 cells displayed ruffle-like membrane extensions (white arrows) detected only in the vicinity of adherent bacteria (white arrow).
D. Detail of interaction between ruffle-like membrane structures and bacteria (white arrow).
E. Intimate contact between bacteria and the surface of an mpkCCDcl4 cell (black arrow) with membrane extensions enclosing the bacteria (white arrow).
F. Internalized bacteria in membranous vacuole structures.

Wild-type AL511 was uniformly distributed over the apical surface of confluent mpkCCDcl4 cells (Fig. 3B). Infected cells displayed the association of membrane ruffle-like structures with some, but not all adherent bacteria (Fig. 3C and D); no such structures were observed in uninfected cells. In many cases, the bacteria were in intimate contact with these ruffle-like structures, which enclosed the bacteria until they were completely surrounded (Fig. 3D). These different phases of interaction between bacteria and cells may be interpreted as inducing the uptake of bacteria into mpkCCDcl4 cells.

Transmission electron microscopy analysis of the interactions of wild-type AL511 with mpkCCDcl4 cells also revealed intimate contact between the bacteria and extensions of the eukaryotic cell membrane (Fig. 3E) and showed that bacteria were present within the membranous vacuole, close to the nucleus (Fig. 3E and F). The occurrence of membrane ruffling suggested that the entry of AL511 into mpkCCDcl4 cells resulted from the hijacking of host cytoskeleton components. Gentamicin assays in the presence of inhibitors of actin and microtubule filament assemblies clearly demonstrated the involvement of AL511 β-actin in the entry process (Supporting information, Fig. S1 and Table S3).

Flagella are involved in the invasion of mpkCCDcl4 cells by E. coli AL511

Four adhesin-encoding operons (fim, pap, afa-8 and f17) have been identified in the AL511 strain (Girardeau et al., 2003). Southern blot experiments showed that each of these operons was present as a single copy in the AL511 genome. Agglutination, immunofluorescence and motility assays and reverse transcription polymerase chain reaction (RT-PCR) experiments demonstrated that these four operons were expressed in AL511 bacteria cultured overnight in Luria–Bertani (LB) broth (Table S4). Isogenic derivatives of AL511 for the fimH, papC, afaC-8 and f17C genes were constructed to assess the role of these adhesin-encoding operons in the interactions of AL511 with mpkCCDcl4 cells. These mutants did not express the mutated adhesin operon, but systematically expressed the other adhesin operons (Table S4). It has been suggested that the flagellum plays a critical role in urinary tract colonization (Lane et al., 2007). We therefore constructed AL511 mutants in which fliC and fliD, encoding flagellin and the flagellum capping protein respectively, were deleted. These two mutants were neither flagellated nor motile (Table S4). All the AL511 mutants were compared with the parental AL511 strain in gentamicin protection assays, and data were interpreted according to the criteria given above. The adhesion properties of the derivatives were similar, and were not affected by any of the mutations (Fig. 4A; numeric data available in Table S5). By contrast, differences in invasion level were observed. The mutations in fimH, papC, afaC-8 and f17C had no effect on invasion levels, whereas relative inhibition of invasion of 94% and 82% was obtained for the AL511 ΔfliC and AL511 ΔfliD mutants respectively. We further investigated the role of the flagellum in the entry of AL511 into mpkCCDcl4 cells, by inhibition and complementation approaches. Bacteria were incubated with an anti-H12 flagellin serum before being mixed with cells in the gentamicin protection assay. Adhesion levels were similar to those observed without prior incubation of the cells with anti-H12 flagellin serum, but the rate of cell invasion was significantly lower (Fig. 4A). Complementation of the AL511 ΔfliC mutant for flagellin production and motility was obtained by introducing pZEKmfliC (H12), carrying the fliC gene from AL511 under the control of its natural promoter and terminator into the mutant (Table 1). The complemented mutant was only moderately motile (28%), but displayed significantly higher levels of invasion than the mutant, approaching those of the wild-type AL511 strain (Fig. 4A). As a control, the introduction of pZEKmGFP alone into AL511 ΔfliC did not restore invasive capacity (data not shown).

Figure 4.

Gentamicin protection assay results with mpkCCDcl4 and HEK293 cells.
A. Interaction of mutants of E. coli strains AL511 with mpkCCDcl4 cells.
B. Interaction of E. coli AL511 and its mutants with HEK293 cells. Relative association (white bars) and invasion (black bars) values and their respective standard deviations, as obtained from the values presented in Table S2. Asterisks indicate significant results. H12 designates the E. coli AL511 ΔfliC+ pZEKmfliC (H12) strain and H48 the E. coli AL511 ΔfliC+ pZEKmfliC (H48) strain.

There are 56 different flagellin serotypes. We therefore investigated the potential link between H12 flagellar serotype and the invasive capacity of the AL511 strain by means of two complementation assays. First, the gene encoding the H48 flagellin from MG1655 was inserted into pZE21-MCS2, and the resulting plasmid was introduced into AL511 ΔfliC. Despite the almost complete restoration of motility (95%), complementation of the deficient invasive phenotype was less efficient than that with the H12 fliC gene (Fig. 4A). We then introduced the H12 flagellin-encoding gene from AL511 [pZEKmfliC (H12) plasmid] into MG1655 ΔfliC. The resulting complemented MG1655 strain was motile (Table 1) but non-invasive (invasion level similar to that of MG1655; data not shown). The discrepancies observed between complementation experiments cannot result exclusively from plasmid expression systems not fully restoring the original phenotypes. These findings strongly suggest that the presence of a fully constructed flagellum is required to restore the invasive capacity of AL511, which may be influenced by flagellin serotype.

We also demonstrated the involvement of the flagellum in the entry of the AL511 strain into human kidney cells, in gentamicin protection assays with the human embryonic kidney cell line HEK293. The AL511 strain adhered to and invaded HEK293 cells (Fig. 4B, numeric data available in Table S5). The level of AL511 binding was similar to that observed with mpkCCDcl4 cells, but the invasion rate was higher for these undifferentiated cells. A comparison of the invasive capacities of the AL511, AL511 ΔfliC and H12 flagellin-complemented AL511 ΔfliC strains with HEK293 cells clearly indicated that flagella were also required for the entry of AL511 into human embryonic kidney cells (Fig. 4B).

Analysis of the flagellum during the infection of mpkCCDcl4 cells by E. coli

mpkCCDcl4 cells were infected with AL511 or its ΔfliC mutants and analysed by transmission electron microscopy to elucidate the role of the flagellum in bacterial invasion. AL511 bacteria produced a network of long filaments (Fig. 5A–C). These filaments were similar in diameter to flagella (∼20/23 nm), and immunogold labelling with an anti-H12 flagellin serum confirmed that these filaments were indeed flagella (Fig. 5A). The ends of flagella frequently bound to ruffle-like structures at the site of interaction between adherent bacteria and renal cells (Fig. 5B and C, white arrows). Ultrastructural analysis of the interaction of AL511 ΔfliC with mpkCCDcl4 cells confirmed the adhesion of mutant bacteria and the absence of flagellum production. Additionally, no production of ruffle-like membrane extensions close to the adherent bacteria was observed (Fig. 5D). These findings suggest that contact between flagella and cell membranes is required to induce actin polymerization (see Supporting information), resulting in the production of ruffle-like extensions.

Figure 5.

Identification of E. coli AL511 flagella at the surface of the bacterium during interaction with mpkCCDcl4.
A. Immunogold labelling of AL511 flagella with anti-H12 flagellin serum.
B and C. The ends of flagella from AL511 interact with the ruffle-like membrane extensions (white arrows), which surround the bacteria.
D. The AL511 ΔfliC strain had no filament network and mpkCCDcl4 cell ruffles were absent.
E. The non-motile AL511 ΔmotAB mutant was flagellated (white arrow), but mpkCCDcl4 cells ruffles were not formed with this mutant.
F. Centrifugation of the AL511 ΔmotAB isolate induced the development of ruffle-like membrane extensions associated with the flagella of this strain.

Role of motility in the uptake of AL511 into mpkCCDcl4 cells

Flagella render E. coli motile. We evaluated the potential link between motility and bacterial uptake into mpkCCDcl4 cells, by carrying out gentamicin protection assays including an additional centrifugation step at t0 during infection, rapidly bringing the bacteria into contact with the cells. This additional centrifugation step had no effect on the levels of adhesion and invasion of the AL511 isolate and did not affect the interaction of AL511 ΔfliC with cells (Fig. 4A, and Table S5). These data provide further evidence for the requirement of a flagellum for bacterial invasion.

We evaluated the potential link between motility and bacterial uptake into mpkCCDcl4, by constructing an AL511 derivative lacking flagellum motor activity by allelic replacement of the motA and motB genes (Table 1). The AL511 ΔmotAB strain is flagellated (Fig. 5E and F) but non-motile (Table 1). Like AL511 ΔfliC, AL511 ΔmotAB adhered to but did not enter mpkCCDcl4 cells under classical gentamicin protection assay conditions (Fig. 4A). However, by contrast to what was observed with AL511 ΔfliC, the entry of AL511 ΔmotAB into cells was partially and significantly restored by the centrifugation of cells and bacteria at t0 (Fig. 4A). Furthermore, the AL511 ΔmotAB and AL511 ΔfliC mutants did not induce the formation of membrane ruffle-like structures in the absence of centrifugation at t0 (Fig. 5D and E). By contrast, with centrifugation at t0, the mpkCCDcl4 cells displayed some membrane ruffle-like structures associated with adherent AL511 ΔmotAB bacteria. As observed with the wild-type strain, the ends of the flagella bound these membrane extensions (Fig. 5F). These observations suggest that both an intact flagellum structure and a torque-generating function are required to induce close contact between the flagellum and the mpkCCDcl4 cell membrane, although bacteria adhered to cell surfaces, resulting in signal transduction, actin polymerization and the initiation of bacterial uptake.


Despite tremendous advances in our understanding of the genetic bases of pathogenicity and of the evolutionary diversity of E. coli strains over the last 10 years, the mechanisms by which pyelonephritis-associated E. coli reaches and persists in the kidney remain poorly understood. We investigated the interaction of UPEC strains with the epithelia of the CD, the first kidney structures encountered by bacteria ascending from the bladder. We demonstrate here that the AL511 UPEC strain can enter and persist in CD mpkCCDcl4 cells, through an interaction process requiring flagellum filaments.

Most UPEC strains belong to the B2/D phylogenetic group and express specialized virulence factors, such as P fimbriae, α-haemolysin and CNF1 toxin. We previously characterized a distinct clonal group consisting of UPEC E. coli strains expressing the Afa-VIII adhesin and O9 and O101 antigens, most of which belonged to phylogenetic group A (Girardeau et al., 2003; and J.P. Girardeau, pers. Comm.). We found that the adhesion and internalization capacities of the pyelonephritic and cystitis-associated UPEC strains and of the faecal strain analysed were independent of the O serogroup in phylogenetic groups B2 and A. Our findings suggest that only a few UPEC strains can invade mpkCCDcl4 cells. These strains include AL511, originally isolated from the urine of a woman suffering from pyelonephritis in France (Archambaud et al., 1988) and highly invasive. This strain belongs to phylogenetic group A, expresses UPEC virulence factors, interacts with mouse renal CD epithelial cells and induces potent inflammatory responses (Chassin et al., 2006). All the characteristics of the AL511 strain and the findings of our study suggest that UPEC group A isolates may be involved in the development of pyelonephritis.

The uptake of the AL511 strain into cells involved a β-actin-dependent mechanism, as shown by experiments with cytochalasin D, an inhibitor of cytoskeleton rearrangements. This uptake was also sensitive to ciprofloxacin, an antibiotic taken up by epithelial cells (data not shown). Transmission electron microscopy demonstrated the presence of bacteria within invaginations of the cell membrane and within cytoplasmic vacuoles. Thus, renal tubule mpkCCDcl4 cells are a valuable in vitro model for studying the invasive properties of E. coli strains. These findings are consistent with the results of a previous study (Iványi et al., 1988), in which bacteria adhering to the apical surface or enveloped by a vacuolar membrane in the cytoplasm of CD cells surrounded by inflammatory cells were observed in biopsy specimens from the renal cortex of patients with pyelonephritis. These findings suggest that the uptake of UPEC isolates by CD cells depends on both the strain and host response capacity.

The adhesion of AL511 to renal CD cells seems to be multifactorial. No significant difference was observed in the level of adhesion of AL511 derivatives harbouring mutations in genes encoding proteins involved in the production of type 1, P, Afa-VIII and F17 adhesins. Similarly, none of the bacterial mutants analysed displayed significantly lower than normal levels of internalization, suggesting that the mechanism of entry of AL511 into CD cells differs from that of the UPEC UTI89 and NU14 strains in bladder cells, in which type 1 adhesins have been shown to play a central role. P fimbriae, which are produced by most pyelonephritis isolates (Bergsten et al., 2005) and recognize receptors located on the luminal surface of renal tubules (Korhonen et al., 1986; Connell et al., 2000), do not seem to be involved in the uptake of AL511 into renal epithelial cells. UPEC strains expressing Afa adhesins also display a unique renal tissue tropism (Nowicki et al., 1988; Lalioui and Le Bouguénec, 2001; Le Bouguénec, 2005). Experiments in vitro have suggested that the production of Afa molecules confers a predisposition to the establishment of intracellular reservoirs of bacteria (Le Bouguénec, 2005), but Afa-VIII molecules do not seem to interfere with the invasion of CD cells by AL511.

The flagellum is a protein complex responsible for motility and chemotaxis. It is composed of a type III-like secretion apparatus, which delivers an extracellular filament by an ATP-dependent mechanism (Berg, 2003). Bacterial motility has been shown to be essential for the colonization of the upper urinary tract by E. coli strains in vivo (Lane et al., 2007). Flagellins of the H6 and H7 subtypes have been shown to act as adhesins, mediating the interaction of enteropathogenic and enterohaemorrhagic E. coli strains with epithelial cells in the colon (Erdem et al., 2007). The flagellum has also been reported to be essential in meningitis-associated E. coli, for the invasion of brain microvascular epithelium cells (Parthasarathy et al., 2007). The flagellum filament consists principally of flagellin (FliC) molecules, and is capped by FliD, a protein essential for the initiation of FliC subunit polymerization (Yonekura et al., 2000). We found that non-flagellated derivatives of strain AL511 lacking FliC or FliD were significantly less invasive than the wild-type strain. The criteria for significance used in this study may be too restrictive, but they nonetheless highlight the importance of flagella for the uptake of UPEC strains. The detection of close interactions between the ends of the flagella and epithelial cell surface membrane ruffles suggests that FliD may mediate cell contact, thereby triggering subsequent cell signalling and actin polymerization. In this case, the non-flagellated ΔfliC mutant producing FliD at the bacterial surface would be expected to be invasive. This was not the case, suggesting that an intact filament structure is required for uptake. So, why are the flagella of AL511 required for entry into CD cells? FliD is strongly conserved in E. coli strains (100% identity between the FliD from AL511 and MG1655), but flagellin is a highly variable protein and the AL511 flagellin is of the H12 type. Complementation experiments showed that the entry of the bacteria did not depend on flagellin subtype, as shown by the trans production of H12 or H48 flagellins, which have very low levels of amino-acid sequence similarity, by the invasion-deficient AL511 ΔfliC mutant. Conversely, production of flagellin of the H12 type did not permit the uptake of the MG1655 ΔfliC mutant.

These results suggested two hypotheses, both proposed in previous studies. First, it is possible that post-transcriptional modifications of FliC or FliD (or both) are responsible for specific flagellum-mediated AL511 uptake into cells (Power and Jennings, 2003). Alternatively, the similarity between the type III secretion system and the flagellum secretion apparatus, including the presence of a central channel of about 25 Å in diameter in the flagellum filament (Yonekura et al., 2000), suggests that the flagellum could be an injectisome structure, delivering putative AL511 effectors to mpkCCDcl4 cells, as reported for Campylobacter jejuni Cia-secreted proteins involved in bacterial invasion (Konkel et al., 2004).

Unlike the type III secretion system, the flagellum apparatus is associated with the MotA and MotB proteins, which form the stator, a membrane pore channel essential for creation of the protonmotive force required to generate flagellum movement (Berg, 2003). We found that both the flagellum filament and motility were required to achieve significant levels of AL511 entry into mpkCCDcl4 cells. Centrifugation at t0 in the infection process partly restored the invasive capacity of the AL511 ΔmotAB mutant. Thus, bacterial motility seems to be required to promote the contact between host cells and the flagellum filament required to induce the entry of the bacteria into cells via the recognition of specific receptors and/or the injection of effectors, cell signalling and actin polymerization.

In conclusion, we demonstrate here the involvement of the flagellum in bacterial entry into renal CD cells. Studies are currently underway to determine the molecular bases and mechanisms of AL511 entry and persistence in CD cells and their contribution to the pathogenicity of UPEC isolates and the recurrence of intrarenal bacterial infections.

Experimental procedures

Bacteria, plasmids and growth conditions

The clinical E. coli isolates from patients with or without UTI and the plasmids used in this study are listed in Table 1. These isolates had previously been assigned to phylogenetic groups and their adhesin-encoding operons (f17, afa-8, pap, fim) and toxin genes (hly, cnf) characterized by PCR (Girardeau et al., 2003; C. Le Bouguénec, pers. comm.). All E. coli strains were cultured overnight (∼16 h) in LB broth (10 g l−1 BactoTryptone, 5 g l−1 yeast extract, 5 g l−1 NaCl, pH adjusted to 7.2) at 37°C, without shaking, with the exception of strains carrying the thermosensitive plasmid pKOBEG-Apra, which were cultured at 30°C. When required for the selection of strains and plasmids, antibiotics were added to the medium at the following concentrations: kanamycin, 50 μg ml−1; zeocin, 60 μg ml−1; apramycin, 100 μg ml−1.

Construction of isogenic mutants

Various E. coli AL511 mutants were constructed by the allelic exchange recombination method, using the thermosensitive plasmid pKOBEG-Apra, which carries lambda red recombination genes (Chaveroche et al., 2000). Two approaches were used to replace genes from AL511 with a zeocin-resistance (zeo) cassette. The papC, afaC-8, f17C and fimH genes encoding the usher subunits of type P, F17 and Afa-VIII adhesive structures and the adhesin subunit of type-1 fimbriae, respectively, were exchanged by a three-step method (http://www.pasteur.fr/recherche/unites/Ggb/matmet.html). Strains with mutations in fliC, fliD and motAB, encoding flagellin, the flagellum capping protein and the stator components of the flagellum motor, respectively, were obtained using a similar one-step strategy in which the upstream and downstream DNA fragments were replaced by synthetic long primers (Table 2). All strains derived from AL511 were checked for susceptibility to apramycin, indicative of the loss of pKOBEG-Apra.

Table 2.  List of primers used for PCR amplification.
 NamePrimer sequence
  • a. 

    Underlined sequences correspond to XhoI (CTCGAG) and XbaI (TCTAGA) restriction enzyme recognition sequences.

Allelic exchange
Allelic exchange verification primer
 fliC (AL511)fliCH12.ext-55′-GTGACCCTGATGGTGTATTTC-3′
Detection by RT-PCR of gene expression

Plasmid construction

The fliC genes, with their σ28 promoter, from E. coli AL511 and MG1655 were inserted between the XbaI and XhoI sites of pZEKmGFP, as follows. DNA fragments were amplified by PCR from AL511 and MG1655, using the Expand High Fidelity PCR System kit (Roche Applied Science) with primers PrfliCH12F and fliCH12R (AL511) or PrfliCH48F and fliCH48R (MG1655) (Table 2), with 5% dimethylsulfoxide (DMSO, Sigma Aldrich, France) added to each reaction (4 min at 94°C, followed by 35 cycles of 94°C, for 30 s, 55°C for 30 s, 72°C for 90 s and a final extension phase at 72°C for 7 min). PCR products and pZEKmGFP were digested with 1 unit of each of XbaI and XhoI (Roche Applied Science), subjected to electrophoresis and the bands were purified from the gel with the QIAquick Gel Extraction Kit (Qiagen). Each PCR product was ligated to the vector in a reaction mixture containing 1 unit of T4 DNA ligase (Roche Applied Science). The recombinant plasmids pZEKmfliC (H12) (fliC gene from AL511 strain) and pZEKmfliC (H48) (fliC gene from MG1655 strain), respectively, were constructed and inserted into competent E. coli MC1061 cells, and were then used to transform the AL511 ΔfliC and MG1655 ΔfliC strains respectively. We checked for complementation, in the form of restored motility, as described below.

Motility test

Escherichia coli isolates were cultured in LB medium for 18 h at 37°C without shaking. A 2 μl sample was added to a thin layer of 0.35% agar LB medium in a Petri dish, which was then incubated overnight at 37°C. The diameter of the motility disk was then determined. For all strains, motility is expressed as a percentage of the diameter of the motility disk covered by the reference E. coli strain, AL511 (for which motility is, by definition, 100%).

Cell culture

We cultured mpkCCDcl4 cells (seeded at 2 × 105 cells/dish) in 2 ml of SVi-defined medium (Bens et al., 1999) in 3 cm collagen-coated Petri dishes, at 37°C, under an atmosphere containing 5% CO2. Cell differentiation and confluence were obtained after 7 days, with renewal of the culture medium every 2 days. Human embryonic kidney HEK293 cells (seeded at 2 × 105 cells/dish) were cultured to confluence (day 3 after seeding) in DMEM (Invitrogen) supplemented with 10% fetal calf serum (Eurobio), 0.45% d-glucose (Sigma-Aldrich), 1 mM sodium pyruvate (Sigma-Aldrich) and 2 mM glutamine (Invitrogen).

Gentamicin protection assay

All experiments were carried out on confluent mpkCCDcl4 or HEK293 cells. Association assays were performed as follows. Cells were rinsed twice with DMEM that had been pre-warmed to 37°C and were then infected with 40 μl (∼ 2 × 107 cfu of bacteria) of an overnight static culture in LB of E. coli in freshly prepared SVi medium without antibiotics. Cells were incubated with bacteria for 3 h, then rinsed twice with DMEM, lysed with 1% Triton X-100 and plated on LB agar for the counting of viable bacteria. We assessed the uptake of bacteria by carrying out the same experiment but with an additional treatment with 20 μl of 10 mg ml−1 gentamicin in freshly prepared SVi medium for 2 h before the two washes with DMEM, cell lysis and viable cell counting. All assays were carried out in duplicate and repeated at least three times. Association and invasion rates are expressed as a percentage of the inoculum. When required, the bacterial inoculum was incubated with an equal volume of undiluted rabbit serum against E. coli H12 flagellin (Statens Serum Institut, Copenhagen, Denmark) for 15 min before infection.

For persistence assays, cells were incubated with bacteria for 3 h and then treated with gentamicin for 2 h (control experiment), 24 h, 48 h or 72 h before cell lysis. For the longer incubations, the defined SVi medium was renewed daily.

Scanning electron microscopy

Confluent mpkCCDcl4 cells grown on collagen-coated coverslips were or were not infected by incubation with the E. coli strains for 3 h. The cells were then rinsed twice in 1× phosphate-buffered saline (PBS), fixed by incubation in 2.5% glutaraldehyde in 0.1 M cacodylate buffer (pH 7.2) overnight at 4°C or for 30 min at room temperature, and washed three times, for 5 min each, in 0.2 M cacodylate buffer. Cells were then post-fixed by incubation for 1 h in 1% (w/v) osmium tetroxide in 0.2 M cacodylate buffer and rinsed with distilled water. Samples were dehydrated through a graded series of ethanol solutions (25%, 50%, 75%, 95% and 100%), washed (5 min each) and dried to critical point with CO2 in a CPD BALTEC apparatus. The dried specimens were mounted on stubs with carbon tape and sputtered with 15 nm gold/palladium ions using GATAN 681 high-resolution ion-beam coating techniques. Secondary electron images and backscattered images (YAG detector) were obtained with a JSM 6700F JEOL microscope with a field emission gun operating at 5 kV.

For flagellum identification, E. coli-infected cells were rinsed in 1× PBS, fixed by incubation in 4% paraformaldehyde/0.1% glutaraldehyde for 15 min and washed again in 1× PBS. They were then blocked by incubation with 50 mM NH4Cl and washed again with 0.1% BSA in 1× PBS. Cells were incubated with a 1:100 dilution of a rabbit anti-E. coli H12 serum for 30 min, washed in 1× PBS, incubated with 10 nm gold bead-coupled protein A and processed as described for other scanning electron microscopy experiments. The cells were sputtered with a 15 nm layer of carbon, using a Gatan Ion Beam coater, and were observed with a backscattered electron detector (YAG) in the scanning electron microscope.

Transmission electron microscopy

Confluent mpkCCDcl4 cells were fixed by incubation for 1 h at room temperature with 2.5% glutaraldehyde in 0.1 M cacodylate buffer (pH 7.2) and washed three times, for 5 min each, in 0.1 M cacodylate buffer. They were then post-fixed by incubation for 1 h at room temperature with 1% osmium tetroxide in 0.1 M cacodylate buffer. Cells were treated with 1% uranyl acetate for 1 h at room temperature. Samples were dehydrated in a graded series of acetone solutions and embedded in Epon. Thin sections were cut and stained with 2% uranyl acetate and lead citrate. Sections were observed in a JEOL 1200 EX transmission electron microscope operating at 80 kV and images were recorded with a CCD camera (Megaview, Eloise Ltd).


This work was supported by Grant PTR165 from Institut Pasteur Programme Transversal de Recherche, and by a grant from ANR (Agence Nationale de la Recherche) for the ERA-NET Pathogenomics Project, ‘Deciphering the intersection of commensal and extraintestinal pathogenic E. coli’. C. Pichon and C. Héchard were supported by these grants. We thank T. Bruhns for providing the HEK293 cells, S. Da Ré and C. Beloin for technical assistance and material transfer, M.C. Prévost for expertise in electron microscopy, and I. Filliol for the H typing of our isolates. All these individuals work at the Pasteur Institute (Paris, France). We thank E. Sokurenko (Washington University) for providing us with E. coli strain NU14, E. Oswald (INRA ENVT, Toulouse, France) for E. coli strain UTI89 and D. Buzoni Gatel (INRA, France) for helpful discussions.