Type 1 pili-mediated adherence of Escherichia coli strain LF82 isolated from Crohn's disease is involved in bacterial invasion of intestinal epithelial cells

Authors


Abstract

We previously characterized the invasive ability of Escherichia coli strain LF82, isolated from an ileal biopsy of a patient with Crohn's disease. In the present study, we performed TnphoA insertion mutagenesis to identify genes involved in LF82 invasion of intestinal epithelial cells. Most of the non-invasive mutants had an insertion mutation within the type 1 pili-encoding operon. Two non-invasive fim mutants, which harboured an insertion within the fimI and fimF genes, still adhered but had lost the ability to induce host cell membrane elongations at the sites of contact with the epithelial cells. Transcomplementation experiments with a fim operon cloned from E. coli K-12 restored both invasive ability and the ability to induce host cell membrane elongations. Expression of the cloned LF82 or K-12 fim operon into the non-invasive laboratory strain JM109 did not confer invasive properties. Thus, these findings showed that: (i) type 1 pili-mediated adherence is involved in LF82-induced perturbation of host cell signalling responsible for membrane elongations; (ii) native shafts are required for type 1 pilus-mediated induction of membrane elongations; (iii) this active phenomenon is a key step in the establishment of the invasive process; and (iv) type 1 pili alone are not sufficient to trigger bacterial internalization.

Introduction

Escherichia coli is abnormally predominant (between 50% and 100% of the total number of aerobes and anaerobes) in early and chronic ileal lesions of Crohn's disease (CD), an inflammatory bowel disease of unknown aetiology (Lederman et al., 1997). Most of the E. coli strains isolated from Crohn's ileal mucosa adhere strongly in vitro to intestinal epithelial cultured cells, a property that could enable them to colonize the intestinal mucosa. However, most of them do not harbour any of the known virulence factors expressed by pathogenic E. coli involved in gastrointestinal outbreaks or in urinary tract infections (Darfeuille-Michaud et al., 1998).

We recently characterized the invasive ability of E. coli strain LF82 isolated from a chronic ileal lesion of a patient with CD (Boudeau et al., 1999). Qualitative and quantitative analysis of its invasive properties revealed that in vitro it efficiently invades a wide range of epithelial cell lines (including the epithelial intestinal cells Intestine-407, Caco-2 and HCT-8 and the epithelial HEp-2 cells), its uptake is dependent upon both functioning host cell actin microfilaments and microtubules, and it survives and replicates in the host cell cytoplasm after lysis of the endocytic vacuole. Therefore, strain LF82 can be considered a true invasive pathogen. Its invasive process was found to be original, in that it possessed none of the known genetic invasive determinants of E. coli, Shigella or Salmonella. Recent data revealed a higher prevalence of such invasive strains associated with the ileal mucosa of patients with CD than in control subjects, which supports a putative role of E. coli invasiveness in the pathogenesis of CD (J. Boudeau et al., submitted). These strains belong to a new potentially pathogenic group of invasive E. coli, which we designated AIEC for adherent-invasive E.coli (Boudeau et al., 1999).

Type 1 pili are the most common filamentous bacterial appendages of E. coli and many Gram-negative bacteria and promote bacterial adhesion to various types of eukaryotic cells. The chromosomal gene cluster responsible for type 1 piliation is composed of nine genes. The type 1 pilus is a heteropolymer comprising the major subunit FimA (Orndorff and Falkow, 1985) and three minor subunits, FimF, FimG and FimH (Orndorff and Falkow, 1984; Klemm and Christiansen, 1987; Maurer and Orndorff, 1987). Assembly of type 1 pili depends on the periplasmic chaperone FimC and the outer membrane usher protein FimD, which are not part of the final shaft (Klemm and Christiansen, 1990; Jones et al., 1993). Two genes, fimB and fimE, encode products that regulate the transcription of fimA (Klemm, 1986). The 30 kDa FimH adhesin is located at the fimbrial tip and interspersed along the shaft of the pili (Krogfelt et al., 1990; Jones et al., 1995). FimH is responsible for adherence ability and binds to terminally located d-mannose moieties on cell-bound and secreted glycoproteins or to non-glycosylated peptide epitopes, allowing bacterial adhesion to a wide range of cells, including epithelial and immune cells (Ofek et al., 1977; Giampapa et al., 1988; Wold et al., 1990; Tewari et al., 1993; Sokurenko et al., 1994).

The pathogenic significance of type 1 pili has generally been dismissed, owing to their ubiquity. Despite the adherence-conferring properties of type 1 pili, their role as a virulence factor of E. coli in gastrointestinal tract infections remains unclear. On the other hand, it is well documented that type 1 pili expression enhances uropathogenic E. coli virulence for the urinary tract in humans (Connell et al., 1996; Langermann et al., 1997; Thankavel et al., 1997; Mulvey et al., 1998; Sauer et al., 2000). Recently, Martinez et al. (2000) demonstrated that the type 1 pilus adhesin FimH can mediate uropathogenic E. coli invasion of human bladder epithelial cells by triggering host cell signalling cascades. Moreover, it has been shown that type 1 pili play a role in Salmonella typhimurium invasion of cultured epithelial cells (Ernst et al., 1990; Horiuchi et al., 1992; Bäumler et al., 1996). Additionally, type 1 pili have been implicated in the development of inflammatory and potentially harmful reactions in the host after bacterial infections (Steadman et al., 1988; Matsumoto et al., 1990; Malaviya et al., 1994).

In order to identify genetic determinants involved in the invasive process of AIEC strain LF82, non-invasive mutants were generated using TnphoA insertion mutagenesis. Interestingly, most of the non-invasive mutants had an insertion mutation within the type 1 pili-encoding operon, which suggested that type 1 pili-mediated adherence plays an essential part in the invasive ability of AIEC strain LF82.

Results

Role of type 1 pili in LF82 invasion of Intestine-407 cells

To assess the role of type 1 pili in the adhesion and invasion properties of LF82, quantitative adhesion and invasion assays were performed with Intestine-407 cells in the presence of either 2% d-mannose or 2% methyl-α-d-mannopyranoside. After a 3 h incubation period with the epithelial monolayers, cell-associated LF82 bacteria represented 135.8 ± 15.2% of the inoculum, as a result of bacterial multiplication. In the presence of d-mannose or methyl-α-d-mannopyranoside, cell-associated bacteria represented only 5.4 ± 2.2% and 2.9 ± 0.4% of the inoculum, respectively, indicating inhibition of LF82 adhesion by 96.1% and 97.8% respectively. LF82 demonstrated a mean invasion level of 1.80 ± 0.71% of the original inoculum. When the incubation media contained either 2% d-mannose or 2% methyl-α-d-mannopyranoside, the invasion level decreased to 0.005 ± 0.002% and 0.004 ± 0.001%, respectively, which corresponded to inhibitory effects of 99.7% and 99.8% respectively. Taken together, these results show that LF82 adhesion to intestinal epithelial cells could be mediated by type 1 pili and that a mannose-dependent interaction was required for effective invasion.

Construction of non-invasive mutants of strain LF82

TnphoA mutagenesis was performed using the mini-Tn5phoA donor plasmid, pUTKm1. A total of 5329 TnphoA insertion mutants was isolated by selecting for kanamycin resistance, of which 767 were PhoA+, indicating that functional fusions with alkaline phosphatase activity were generated with potentially outer membrane or periplasmic proteins. By screening the whole mutant library with a qualitative gentamicin invasion assay using Intestine-407 cells, 16 mutants showed a decreased ability to invade epithelial cells compared with the parental strain, of which nine were PhoA mutants and seven were PhoA+ mutants (Table 1). These 16 non-invasive mutants were examined in a quantitative invasion assay in comparison with the parental invasive strain LF82. The invasive ability of each mutant relative to that of the parental strain is presented in Table 1. After a 3 h infection period and 1 h of gentamicin exposure, wild-type LF82 demonstrated a mean invasion level of 1.10 ± 0.28% of the original inoculum. As a result of the insertion mutations, the invasive ability of the 16 derivatives only represented 3.4–9.4% of that of parental LF82, depending on the mutants, indicating that the transposon-inserted regions were important for E. coli LF82 invasiveness.

Table 1. Characteristics of the TnphoA mutants.
 Expression ofInsertion
site of mini-
Tn5phoA
Adherent bacteriaIntracellular bacteria without centrifugationInvasion
indexg
Intracellular bacteria with centrifugation
PhoAType 1 pilicfu/well × 105a% adhesionb (relative %)ccfu/well × 103d% invasione (relative %)fcfu/well × 103d% invasione (relative %)f
  • a . Mean number of cell-associated (adherent + intracellular) bacteria well −1 ± SE.

  • b

    . Mean percentage of inoculum associated with the epithelial cells ± SE (adherent + intracellular bacteria).

  • c

    . Percentage of cell-associated bacteria relative to that of wild-type LF82, defined as 100%.

  • d . Mean number of bacteria surviving gentamicin treatment well −1 ± SE.

  • e

    . Mean percentage of inoculum surviving gentamicin treatment ± SE.

  • f

    . Percentage of intracellular bacteria relative to that of wild-type LF82, defined as 100%.

  • g

    . Percentage of intracellular bacteria relative to cell-associated bacteria.

  • h

    . ND, not determined.

Mutants
 52D11+ fimA 5.37 ± 2.7613.43 ± 6.89 (6.7)2.44 ± 0.400.061 ± 0.010 (5.0)0.455.18 ± 1.100.130 ± 0.028 (1.8)
 2E5+ fimI 47.00 ± 28.28117.5 ± 70.7 (82.1)4.05 ± 1.950.102 ± 0.049 (9.4)0.0910.83 ± 5.890.274 ± 0.151 (3.3)
 4C1 fimI 4.28 ± 1.1610.70 ± 2.90 (7.0)2.30 ± 0.480.058 ± 0.012 (5.5)0.542.77 ± 0.470.070 ± 0.012 (0.9)
 ZG4 fimC 7.75 ± 1.2019.38 ± 3.01 (12.8)2.32 ± 0.200.058 ± 0.005 (5.5)0.304.04 ± 2.460.099 ± 0.058 (1.2)
 1A1+ fimC 3.38 ± 0.748.45 ± 1.84 (5.6)1.97 ± 0.290.049 ± 0.007 (4.8)0.58NDhND
 IC9 fimD 2.70 ± 0.546.75 ± 1.34 (3.6)1.90 ± 0.080.048 ± 0.002 (3.8)0.703.18 ± 0.880.080 ± 0.022 (1.0)
 RA12 fimD 5.33 ± 0.5813.33 ± 1.45 (8.9)2.53 ± 0.580.064 ± 0.014 (6.0)0.48NDND
 ZB10 fimD 2.73 ± 0.476.83 ± 1.17 (3.7)1.78 ± 0.170.045 ± 0.004 (3.6)0.65NDND
 23D9 fimD 2.19 ± 0.045.48 ± 0.11 (3.1)1.90 ± 0.140.048 ± 0.004 (3.8)0.87NDND
 2G3+± fimF 34.70 ± 16.5586.75 ± 41.37 (60.1)2.15 ± 0.670.054 ± 0.017 (5.3)0.066.12 ± 0.740.155 ± 0.021 (2.0)
 ZG2 fimH 4.11 ± 0.2110.28 ± 0.53 (6.9)2.28 ± 0.420.057 ± 0.010 (5.4)0.553.18 ± 1.160.080 ± 0.029 (1.0)
 OA8++ b2512 NDND1.75 ± 0.350.044 ± 0.008 (3.6)NDNDND
 2E4++ b2512 NDND1.66 ± 0.400.042 ± 0.010 (3.4)NDNDND
 51F8++ b2512 NDND1.92 ± 0.850.048 ± 0.021 (4.0)NDNDND
 1C5+ yibP NDND3.72 ± 0.730.093 ± 0.018 (8.6)NDNDND
 2D2+ yhbM NDND4.20 ± 2.860.107 ± 0.074 (9.1)NDNDND
Reference strains
 LF82+ 74.64 ± 25.58186.6 ± 63.9 (100)43.87 ± 11.101.10 ± 0.28 (100)0.59319.5 ± 119.57.99 ± 2.98 (100)
 JM109± 1.10 ± 0.982.75 ± 2.44 (1.5)0.08 ± 0.140.002 ± 0.003 (0.03)0.07NDND

Southern blot hybridization of PstI-digested genomic and plasmid DNA from each of the 16 non-invasive mutants with a 500 bp intragenic polymerase chain reaction (PCR) product of the kanamycin resistance gene confirmed the presence of a single copy of the transposon in their chromosome (data not shown).

Cloning and characterization of TnphoA insertion regions

The transposon insertion sites were identified for the 16 non-invasive mutants by cloning the PstI restriction fragments containing the kanamycin resistance gene and the genomic region downstream of the transposon insertion site. Nucleotide sequences of the genomic regions were matched to sequences in the NCBI databases. All the sequences identified matched with sequences of the K-12 E. coli strain MG1655 complete genome (Table 1). Nucleotide sequences from mutants OA8, 2E4 and 51F8 mapped with a putative open reading frame (ORF) of E. coli K-12 whose deduced amino acid sequence shows similarities to those of both a serine–threonine protein kinase and a dehydrogenase. The DNA sequences cloned from mutants 1C5 and 2D2 mapped with putative ORFs of E. coli K-12 encoding a putative outer membrane protein and a putative regulator respectively. Finally, 11 nucleotide sequences mapped to genes of the fim operon encoding type 1 pili of E. coli.

Characteristics of the fim mutants

The 11 fim derivatives were investigated further to assess the role of type 1 pili in bacterial invasiveness. One insertion mutation was located within the fimA gene, two within fimI, two within fimC, four within fimD, one within fimF and one within fimH (Table 1). To determine whether type 1 pili derivatives had lost type 1 piliation, type 1 pili expression was assessed by colony immunoblotting using polyclonal antibodies raised against type 1 pili. Ten non-invasive fim mutants did not react with the type 1 pili antiserum, indicating that the insertion mutations in fimA, fimI, fimC, fimD and fimH interfered with type 1 pili synthesis and/or assembly. Loss of type 1 pili synthesis for the fimH mutant was confirmed by Western immunoblotting with the same antiserum (data not shown). One additional mutant (2G3), harbouring an insertion mutation within fimF, reacted weakly with the antiserum compared with wild-type LF82, which suggests that a decreased amount of type 1 pili was synthesized.

The 11 fim mutants were examined in a quantitative adhesion assay and compared with the parental strain LF82 (Table 1). Cell-associated LF82 bacteria represented 186.6 ± 63.9% of the original inoculum, as a result of bacterial multiplication during the 3 h incubation period. As expected, nine of the 10 non-piliated mutants were found to be markedly affected in their ability to adhere to Intestine-407 cells, as their adhesion levels represented 3.1–12.8% of that of LF82. The invasion defect of these mutants closely paralleled their adherence defect, as the percentages of invasive bacteria relative to adherent bacteria were almost identical to that of LF82 (Table 1). A centrifugation step was performed during invasion experiments to approximate the adherent-defective mutants and the epithelial monolayers. Nevertheless, the invasion levels of all fim derivatives tested were not significantly increased after centrifugation, as they represented 0.9–3.3% of that of strain LF82 (Table 1).

In contrast, the piliated fimF mutant 2G3 and the non-piliated fimI mutant 2E5 still demonstrated adherent ability that was 60.1% and 82.1% of that of parental LF82 respectively. This indicates that the adhesin FimH was expressed on the bacterial surface and thus that their respective insertion mutations allowed transcription of downstream genes. However, mutant 2E5 (fimI) did not react with antibodies raised against type 1 pili, indicating the absence of the major structural subunit FimA. Results expressed as percentages of invasive bacteria relative to adherent bacteria clearly show that bacterial adherence of mutants 2G3 and 2E5 to Intestine-407 cells did not enable the bacteria to be internalized and, hence, that the defect in invasive ability was independent of adhesion properties (Table 1). The absence of a fimbrial structure as a result of the transposon insertion within fimI, or the absence of FimF within the fimbrial shaft, did not therefore drastically modify adhesive properties but abolished invasive ability.

Molecular cloning of the type 1 pili-encoding gene cluster from strain LF82

A genomic DNA library from strain LF82 was constructed in recipient strain JM109 to clone the fim operon encoding type 1 pili. As recipient strain JM109 possesses its own type 1 pili-encoding genes, screening of the library was performed by hybridization of cosmid extracts with fimB and fimH nucleic probes under high-stringency conditions. Of the 250 cosmids tested, five hybridized with only the fimH probe and one (cosmid JE7) hybridized with both fimH and fimB probes. Endonuclease restriction analysis showed that cosmid JE7 harboured a 36 kb insert (data not shown). To investigate the presence of an entire fim operon, PCR experiments were performed with purified cosmid template JE7 using two pairs of primers designed on the basis of the K-12 fim operon sequence. The 5′ region of the fim operon was investigated using a primer designed in the 5′ region of the K-12 fim operon in duplex with a primer designed within fimB. The 3′ region of the operon was analysed with a couple of primers that anneal to sequences upstream and downstream of fimH. Amplification products were obtained with each pair of primers, indicating that the 36 kb insert cloned in cosmid JE7 carried the entire fim operon of strain LF82 (data not shown).

Nucleotide sequence of LF82 fimA, fimI, fimF and fimH genes

Nucleotide sequences of the genes encoding FimA, FimI, FimF and FimH subunits of LF82 type 1 pili were determined using cosmid template JE7 and compared with their previously published counterparts. Sequence reactions were performed using oligonucleotide primers designed in extragenic regions of each gene on the basis of the K-12 fim operon sequence. These sequence data have been submitted to the DDBJ/EMBL/GenBank databases under accession numbers AF286465 (fimA and fimI), AF288195 (fimF) and AF288194 (fimH).

Seven published sequences of E. coli were identified by the blast program that exhibited similarities to the LF82 fimA sequence. LF82 fimA sequence demonstrated 90% identity with that of E. coli K-12 (GenBank accession no. AE000502). Alignment of amino acid sequences revealed that the FimA subunit of AIEC LF82 differed from that of K-12 by several amino acid variations along the entire length of the sequence (Fig. 1A). LF82fimA sequence demonstrated 100% identity with that of E. coli strain MT78 of avian origin (GenBank accession no. Z37500) and between 89% and 97% identity with those sequenced from other E. coli strains (GenBank accession nos Y10902, D13186, M27603, X00981 and U20815).

Figure 1.

Multiple alignment of FimA (A) and FimH (B) products. The residues listed above are for the amino acids in the LF82 FimA sequence. Only polymorphic residues are shown, and the positions on the LF82 sequence are numbered vertically. Δ, deleted residues.

The LF82 fimI sequence demonstrated 98% and 100% identity with those of E. coli strains K-12 and MT78 respectively. The amino acid sequence alignment showed that LF82 and K-12 FimI sequences differed by two amino acid substitutions (data not shown).

The fimF sequence displayed 97% identity with that of E. coli strain K-12 and 99% identity with that of E. coli strain MT78. LF82 FimF differed from those of K-12 and MT78 by four and one amino acid substitutions respectively (data not shown).

The LF82 fimH sequence displayed strong homologies with nine fimH sequences of E. coli in the NCBI databases. 97% identity was found with the fimH sequences of E. coli K-12 (GenBank accession no. AE000502) and those of several allelic variants of type 1 pili of E. coli strain K-12 (GenBank accession nos AF154928, AF154927, AF154926 and AF154925). The multiple alignment of the amino acid sequences showed that the variations in the predicted LF82 FimH sequence compared with those of K-12 strains were point amino acid substitutions along the entire length of the sequence (Fig. 1B). The strongest DNA sequence identities (99%) were observed with the fimH sequences from the avian E. coli strain MT78 of O2:K1 serotype (GenBank accession no. AJ225176) and that of the meningitis isolate IHE3034 of O18:K1:H7 serotype (GenBank accession no. AF089840). The predicted amino acid sequence of the LF82 FimH subunit differed from that of MT78 only by the substitution Asn-54 for Lys-54 and from that of IHE3034 by the substitution Ser-83 for Ala-83. In addition, the FimH sequences from strains LF82, MT78 and IHE3034 all differed from those of K-12 and the K-12 derivatives by the common variations at residues Ala-48, Ser-91 and Asn-99 (Fig. 1B).

Transcomplementation of fim derivatives with the K-12 fim operon

Transcomplementation experiments were performed with a fim operon cloned from E. coli K-12. An 11.2 kb SalI fragment from plasmid pORN104 harbouring the entire fim operon of K-12 E. coli strain J96 (Orndorff and Falkow, 1984) was subcloned into plasmid vector pHSG575 and named pPBI01. Plasmid pPBI01 was transferred into the defective fim mutants ZG2 (fimH), ZG4 (fimC), 2E5 (fimI), 2G3 (fimF), 4C1 (fimI), IC9 (fimD) and 52D11 (fimA). Colony immunoblotting using a whole type 1 pili antiserum confirmed restoration of type 1 pili synthesis (data not shown). As a control, plasmid vector pHSG575 was transferred into the same mutants. All transformants were tested for their ability to adhere to and invade Intestine-407 cells in comparison with their corresponding mutants and the parental strain. Surprisingly, pPBI01 was able to complement both adhesion and invasion defects to the levels of wild-type LF82 for all the fim mutants tested (Fig. 2). Adhesion and invasion levels of the fim derivatives transcomplemented with pPBI01 ranged from 85.7% to 147.3% and from 103.6% to 223.1% of those of LF82. Transformation with plasmid vector pHSG575 modified neither the adhesion nor the invasion properties of any of the mutants tested except mutant 2E5 (fimI), which exhibited an unexplained decrease in adherent ability (Fig. 2).

Figure 2.

Transcomplementation of the adhesion (A) and invasion (B) defects of LF82 fim mutants with plasmid pPBI01 carrying the entire fim operon. Cell-associated bacteria were quantified after a 3 h infection period. Invasion was determined after gentamicin treatment for an additional hour. Results are expressed as cell-associated (adherent + intracellular) or intracellular bacteria relative to those obtained for wild-type LF82, taken as 100%. Each value is the mean of at least three separate experiments.

Search for invasive ability of JM109 transformed with cloned fim operons

Cosmid JE7 and plasmid pPBI01, harbouring the entire LF82 and K-12 fim operons, respectively, were transferred into the non-invasive laboratory strain JM109. The JM109/JE7 and JM109/pPBI01 recombinant strains expressed large amounts of type 1 pili, as detected by colony immunoblotting (data not shown). Nevertheless, neither JM109/JE7 nor JM109/pPBI01 invaded Intestine-407 cells, as their invasion levels were only 0.013 ± 0.010% and 0.118 ± 0.153%, respectively, compared with 0.002 ± 0.003% for the JM109 recipient strain. These results indicate that type 1 pili from strain LF82 and K-12 strain J96 are not sufficient to confer invasive properties to a non-invasive laboratory strain.

Electron microscope examination of infected Intestine-407 cells

Electron microscope examination of LF82-infected Intestine-407 cells revealed numerous bacteria interacting with the host cell surface and the elongation of membrane extensions at the sites of intimate contact between the entering bacteria and the epithelial cells (Fig. 3). The elongated microvilli enwrapped the bacteria, and numerous bacteria were observed within the cells as a result of bacterial uptake by a macropinocytosis-like mechanism. To confirm quantitative invasion results obtained with the gentamicin kill assay, intracellular bacteria were quantified by transmission electron microscope examination (TEM) of the infected epithelial cells. The mean number of intracellular bacteria cell−1 was determined by counting intracellular bacteria in 50 cells. A mean number of 1.48 intracellular bacteria cell−1 was found for wild-type LF82.

Figure 3.

Transmission electron micrograph of Intestine-407 cells infected with E. coli strain LF82. The invading bacteria induce the elongation of membrane extensions upon contact with the eukaryotic cell membranes. Bacteria are engulfed by the elongated microvilli and are internalized within endocytic vacuoles.

When epithelial monolayers were infected with the adherent fim derivatives 2G3 (fimF) and 2E5 (fimI), TEM examination showed numerous bacteria interacting with the epithelial cell membranes (Fig. 4A and C). However, unlike those with strain LF82, these interactions were not accompanied by morphological modifications of the host cell membranes. No intracellular bacteria were observed for mutant 2G3, and mutant 2E5 entered epithelial cells at a low rate (0.17 intracellular bacteria cell−1). On the other hand, when these mutants were transcomplemented with plasmid pPBI01 encoding the fim operon, their interactions with Intestine-407 cells led to the same morphological modifications of the host cell membranes as those observed with wild-type LF82, and numerous bacteria were internalized within the cells (Fig. 4B and D). Indeed, mutants 2G3 and 2E5 transformed with pPBI01 entered Intestine-407 cells with a mean number of 1.86 and 1.98 intracellular bacteria cell−1 respectively. The relative invasion levels of mutants 2G3 and 2E5 represented 0% and 11.5% of that of LF82, respectively, whereas those of mutants 2G3 and 2E5 transcomplemented with the fim operon represented 125.7% and 133.8%, respectively, which corroborated the results obtained with the gentamicin kill assay. Thus, restoration of type 1 piliation in LF82 was associated with ability to promote both host cell membrane morphological modifications and bacterial uptake, indicating that the type 1 pilus-mediated interaction is necessary to induce host cell alterations that are required for bacterial uptake.

Figure 4.

Transmission electron micrographs showing the interaction of the non-invasive fimF (A) and fimI (C) mutants and their counterparts transcomplemented with plasmid pPBI01 carrying the entire fim operon (B and D respectively) with Intestine-407 cells. Both mutants adhere to the cell membranes without inducing any morphological alteration in the host cells. Transcomplemented mutants induce host cell membrane elongations that result in bacterial internalization.

Discussion

We recently characterized the invasive ability of AIEC strain LF82 isolated from the ileal mucosa of a patient with CD. This strain was able to invade a wide range of epithelial cell lines. Electron microscopic examination of LF82-infected HEp-2 and Intestine-407 cells revealed a macropinocytosis-like process of entry, characterized by the elongation of membrane extensions, which surrounded the bacteria at the sites of contact between the entering bacteria and the epithelial cells (Boudeau et al., 1999; this study).

The binding of LF82 bacteria to epithelial cells via mannose-containing receptors appears to be a key step in the invasion process, to the extent that the presence of d-mannose or methyl-α-d-mannopyranoside during cell–bacteria interactions completely inhibited LF82 invasion of the epithelial cells. Strain LF82 adherence to host cells also decreased drastically. Nevertheless, some bacteria still adhered to the intestinal cells, which suggests that other adhesive components are produced by the strain and is also corroboration of findings from a previous study in which we used the intestinal epithelial Caco-2 cell line (Darfeuille-Michaud et al., 1998).

Nucleotide sequence analysis of fimA and fimH genes revealed that strain LF82 produces variant type 1 pili compared with those of E. coli K-12. On the other hand, nucleotide sequence comparisons showed that two E. coli fimH sequences were almost identical to that of LF82. One belonged to the avian E. coli strain MT78, which was isolated from a chicken with acute colisepticaemia and is therefore invasive in vivo (Dho-Moulin et al., 1990; Marc and Dho-Moulin, 1996), and the other corresponded to E. coli strain IHE3034, a member of the K-1 serogroup, which was isolated from a meningitis case (Pouttu et al., 1999). K-1 E. coli strains involved in meningitis are able to invade brain microvascular endothelial cells to cross the blood–brain barrier and reach the cerebrospinal fluid (Huang et al., 1995). Thus, interestingly, AIEC strain LF82 shares strong homologies in the FimH sequence with other invasive E. coli strains.

The role of type 1 pili-mediated adherence in the invasion process was confirmed when non-invasive mutants of strain LF82 were generated using TnphoA insertion mutagenesis. Nucleotide sequence analysis of the TnphoA insertion regions revealed that 11 non-invasive mutants were inactivated in a gene of the fim operon encoding type 1 pili. Nine of these mutants, which harboured an insertion mutation within fimA, fimI, fimC, fimD or fimH, exhibited lower adherent ability than the parental strain and had lost type 1 piliation, as determined by immunoblotting experiments. This suggests that these mutants had undergone either polar insertions or insertions within genes that are required for type 1 pili synthesis and/or assembly. Abolition of type 1 pili synthesis for the fimA, fimC and fimD insertion mutants is consistent with our present knowledge about type 1 pilus biogenesis. The fimA gene, which encodes the major structural subunit of type 1 pili, is known to be required for type 1 piliation, and studies have shown that, in addition to FimA, fimC and fimD products are required for the assembly and presentation of type 1 pili on the bacterial cell surface (Orndorff and Falkow, 1984; Klemm et al., 1985; Klemm and Christiansen, 1990; Jones et al., 1993). On the other hand, loss of piliation for the fimH mutant was not expected, as previously published results have shown that deletion of the fimH gene does not affect fimbrial expression (Klemm and Christiansen, 1987). One possible explanation for the loss of type 1 pili synthesis is that a fusion protein is produced as a result of the transposon insertion, which subsequently interferes with the transport and/or assembly of the fimbrial subunits on the bacterial surface.

Two fim derivatives that harboured a transposon insertion within fimF and fimI conserved adherent ability. The fimF mutant appeared to produce type 1 pili, albeit at a lower level than the parental strain. This corroborated previously published results, which demonstrated that the fimF gene product is not required for type 1 pili structure, that the receptor-binding function of type 1 pili is independent of the fimF product and that a fimF mutant exhibited reduced numbers of pili cell−1 (Russel and Orndorff, 1992). Immunoblotting experiments failed to detect any type 1 piliation for the fimI mutant, which points out the absence of the major structural subunit FimA. No function has yet been assigned for fimI. We have shown in this study that a transposon insertion within fimI is not critical for the adhesin FimH to be located on the bacterial surface. We also speculate that the fimI gene product plays a role in the assembly of the fimbrial subunits, but this needs to be confirmed as the fimI gene had undergone a transposon insertion that may lead to the production of a fusion protein, which, in turn, may interfere with the transport and/or assembly of the type 1 pili subunits.

Loss of invasion was not only a result of the inability of the mutants to gain access to host cells. When invasion assays were preceded by a centrifugation step intended to approximate the bacteria and the epithelial cells, the internalization levels of all non-invasive mutants were not higher than those obtained without centrifugation. Moreover, the adherent ability of the fimF and fimI mutants did not enable the bacteria to be internalized. When non-invasive fim derivatives were transcomplemented with a recombinant plasmid containing the entire type 1 pili operon cloned from E. coli strain K-12, restoration of type 1 pili expression correlated with restoration of both adhesion and invasion properties to the level of wild-type LF82. Taken together, these results strongly support the hypothesis that type 1 pili-mediated binding to host cells is a crucial step in the establishment of the invasive process of strain LF82. They also demonstrate that FimH-mediated adherence by itself is not sufficient to achieve bacterial uptake. The absence of correlation between the adherent and invasive abilities of the fimF and fimI mutants might result from the absence of the minor subunits FimF or FimI, which could either induce differences in the exposure of the saccharide binding site on the surface of type 1 pili or impede shaft assembly respectively. This indicates that native type 1 fimbrial shafts are required for triggering bacterial internalization.

The adherent but non-invasive fimF and fimI mutants provided us with the opportunity to assess the role of type 1 pili in host cell morphological alterations. Although the adherent properties of the fimF and fimI derivatives enabled the bacteria to interact with the host cell surfaces, they failed to induce any morphological changes in the host cells, and the bacteria remained exclusively extracellular. When transcomplemented with the fim operon, these two mutants recovered the ability to both induce host cell membrane elongations and invade the epithelial cells as wild-type LF82. This indicates that the molecular basis of the type 1 pilus-mediated interaction is crucial for LF82 to induce membrane elongations of intestinal epithelial cells and that this early stage of interaction is required for invasion. Similar host cell alterations were reported in vivo for murine bladder epithelial cells and in vitro for human cultured bladder cells infected with type 1-piliated uropathogenic E. coli, in which cytoskeletal rearrangements were observed in areas surrounding bacterial internalization via a ‘zipper’ mechanism, subsequent to the activation of host cell signalling cascades (Mulvey et al., 1998; Martinez et al., 2000). LF82-induced host cell morphological changes were shown to be based on localized alterations in the host cell cytoskeleton, which may also be the result of LF82 activation of host signal transduction pathways (Boudeau et al., 1999).

Expression of variant type 1 pili from E. coli strain LF82 or type 1 pili from E. coli K-12 was not sufficient to confer invasive properties to a non-invasive laboratory strain. JM109 recombinant strains expressing type 1 pili cloned from either K-12 or LF82 did not exhibit effective invasion of Intestine-407 cells. Thus, our results demonstrate that K-12 type 1 pili and the LF82 variant of type 1 pili play a similar role in the invasion of intestinal epithelial cells, but only when expressed in the LF82 genetic background. This suggests that additional bacterial factors co-operate with type 1 pili to achieve bacterial internalization, a hypothesis supported in the present work by the isolation of non-invasive mutants that were inactivated in three additional uncharacterized loci. This feature distinguishes LF82 invasion of intestinal epithelial cells from uropathogenic E. coli invasion of human bladder epithelial cells, for which recent findings demonstrated the direct involvement of type 1 pili (Martinez et al., 2000).

In conclusion, type 1 pilus-mediated adherence is involved in LF82-induced alterations in epithelial cells, and this active phenomenon is a key step in the establishment of its invasive process. In addition, type 1 pili have to be expressed in the genetic background of E. coli strain LF82 to promote bacterial uptake, indicating that other bacterial factors are involved in the invasion process of AIEC LF82. In addition, type 1 pilus-dependent cytoskeletal rearrangements in host cells are not only dependent on the FimH adhesive structure, as native fimbrial shafts are essential to achieve bacterial uptake. Hence, a proper type 1 pilus-mediated binding is crucial in order for LF82 to activate host cell signal transduction pathways. Further investigations are in progress to identify other determinants that co-operate with type 1 pili in the bacterial invasion of intestinal epithelial cells.

Experimental procedures

Bacterial strains, plasmids and cell lines

E. coli strain LF82 was isolated from a chronic ileal lesion of a patient with CD. It belongs to serotype O83:H1 and is sensitive to most antibiotics but not to amoxycillin (Boudeau et al., 1999). E. coli strains DH5α and JM109 were used as host strains for cloning experiments. E. coli strain SM10 (λpir) was used for TnphoA mutagenesis as a source of the mini-Tn5phoA donor plasmid pUTKm1 (De Lorenzo et al., 1990). Recombinant K-12 E. coli strain P678-54/pORN104 (Orndorff and Falkow, 1984) was kindly provided by P. Orndorff. Plasmid pORN104 contains an 11.2 kb chromosomal DNA fragment carrying the entire fim operon of the K-12 E. coli clinical isolate J96. Plasmid vectors pHSG575 and pUC18 and cosmid vector pHC79 (Collins, 1979) were used in cloning experiments. Bacteria were grown routinely in Luria–Bertani (LB) broth or on LB or Mueller–Hinton agar plates (Institut Pasteur Production) overnight at 37°C. When required, antibiotics were added to media at the following concentrations: ampicillin (20 µg ml−1), kanamycin (100 µg ml−1) and chloramphenicol (25 µg ml−1).

HEp-2 cells derived from a human laryngeal carcinoma and the Intestine-407 cell line derived from human embryonic jejunum and ileum were purchased from Flow Laboratories. They were maintained in an atmosphere containing 5% CO2 at 37°C in modified Eagle medium (MEM; Seromed, Biochrom) supplemented with 10% (v/v) heat-inactivated fetal calf serum (FCS; Seromed), 1% non-essential amino acids (Life Technologies), 1% l-glutamine (Life Technologies), 100 000 U l−1 penicillin, 100 mg l−1 streptomycin, 25 µg l−1 amphotericin B and 1% MEM vitamins solution X-100 (Life Technologies).

Invasion and adhesion assays

Invasion assays were performed using the gentamicin protection assay as described previously (Boudeau et al., 1999). Briefly, epithelial cells seeded in 24-well tissue culture plates (Polylabo) with 4 × 105 cells well−1 and grown for 20 h were infected at a multiplicity of infection (MOI) of 10 bacteria cell−1. After a 3 h incubation period at 37°C, the monolayers were washed in phosphate-buffered saline (PBS; pH 7.2), incubated for 1 h in cell culture medium containing 100 µg ml−1 gentamicin (Sigma Chemical) and washed again with PBS. The epithelial cells were then lysed with 1% Triton X-100 (Sigma) in deionized water. Samples were diluted and plated onto Mueller–Hinton agar plates to determine the number of cfu recovered from the lysed monolayers.

To determine the total number of cell-associated bacteria corresponding to adherent and intracellular bacteria, the eukaryotic cells were lysed after the 3 h infection period, and the bacteria were quantified as described above.

For inhibition experiments of bacterial adhesion and invasion, 2% (w/v) d-mannose or 2% (w/v) methyl-α-d-mannopyranoside were added to the cell culture medium before infection and were present throughout the 3 h infection period.

When needed, the infected monolayers were centrifuged for 10 min at 2000 g before the 3 h infection period.

A qualitative invasion assay was developed to screen the mutant library using 96-well tissue culture plates. The invasion procedure was the same as that described above. Lysis of the epithelial monolayers was performed with 100 µl of 1% Triton X-100, and 10 µl was spotted onto Mueller–Hinton agar plates. The absence of bacterial growth or the presence of a few individual colonies within the spot corresponded to percentages of invasion of < 0.2% of the original inoculum.

Extraction and purification of type 1 pili

Bacterial surface proteins were extracted as described by Dodd and Eisenstein (1982), with some modifications. Bacteria grown statically overnight in LB were harvested in 0.1 M PBS by centrifugation at 6000 g for 10 min. The bacterial surface proteins were recovered from the bacterial cells by heating the suspension at 60°C for 20 min with vigorous shaking. Bacterial cells and debris were sedimented at 10 000 g for 20 min. The supernatant was brought to pH 4.0 and stored overnight at 4°C. The precipitated proteins were collected by centrifugation at 20 000 g for 30 min and suspended in 0.1 M PBS. This suspension constituted the crude extract of bacterial surface proteins. As fimbriae are resistant to SDS disaggregation, surface protein extracts were subjected to HCl hydrolysis before SDS–PAGE analysis in order for the subunits to enter the gel. Samples were acidified to pH < 2 with HCl, heated to 100°C for 5 min, cooled and neutralized with NaOH.

Immunoblotting

The rabbit antiserum raised against purified type 1 pili preparations was a generous gift from Karen Krogfelt (Krogfelt and Klemm, 1988).

Western immunoblotting was performed according to the procedure of Towbin et al. (1979), with minor modifications. HCl-treated surface protein extracts were heated for 5 min in SDS sample buffer. Proteins were resolved by SDS–PAGE using 10% polyacrylamide gels (Laemmli, 1970) and electroblotted onto nitrocellulose membranes (Amersham International). The membranes were then dried and blocked with 2% (w/v) bovine serum albumin (BSA; Sigma) in Tris-buffered saline–Tween 0.05% (TBST) at room temperature for 2 h. The membranes were reacted with the type 1 pili antiserum diluted in 1% (w/v) BSA in TBST at room temperature for 2 h. Immunoreactants were then detected using a secondary anti-rabbit antibody conjugated with alkaline phosphatase. A substrate composed of 5-bromo-4-chloro-3-indolyl phosphate (X-P) and 4-nitroblue tetrazolium chloride (NBT) in a detection buffer (100 mM Tris, 100 mM NaCl, 5 mM MgCl2, pH 9.5) was used to visualize reaction products.

For colony immunoblotting, bacteria grown overnight in LB were harvested in PBS and spotted onto nitrocellulose membranes. The membranes were then dried and processed as described above.

Cosmid cloning

Total genomic DNA from strain LF82 was prepared by a sarkosyl–proteinase K lysis procedure including CsCl purification and partially digested with Sau3A (Boehringer Mannheim). DNA fragments were size fractionated by centrifugation through a 10–40% sucrose density gradient and then ligated to dephosphorylated, BamHI-linearized cosmid vector pHC79. Cosmids were packaged in vitro into phage lambda particles (Gigapack III XL packaging extract; Stratagene). These particles were used to infect E. coli JM109. Ampicillin-resistant transductants were selected on to Mueller–Hinton agar plates containing ampicillin and dispensed into individual wells of microtitre plates.

Transposon mutagenesis and molecular cloning

Random insertional mutations were generated using suicide plasmid pUTKm1 carrying the mini-Tn5phoA transposon. Plasmid pUTKm1 was isolated from E. coli strain SM10 (λpir) using the Qiagen plasmid midi kit and transferred into E. coli strain LF82 by electroporation. Transposition mutants were selected onto agar plates containing kanamycin and 50 µg ml−1 X-P and dispensed into microtitre plates containing kanamycin. Loss of the suicide vector was checked by colony blot hybridization with a specific nucleic probe generated by PCR.

The unique PstI restriction enzyme site, located upstream of the kanamycin resistance gene in the mini-Tn5phoA, was exploited for cloning of the genomic region downstream of the transposon insertion. Total DNA from each mutant was extracted and digested with PstI (Boehringer Mannheim). Southern hybridization was performed to confirm the presence of single insertions. PstI restriction fragments were separated on 1% agarose gels and transferred to nylon membranes (Schleicher and Schuell) according to the Southern procedure (Sambrook et al., 1989). Fragments carrying the mini-Tn5phoA were detected by hybridization with an intragenic kanamycin probe generated by PCR. For cloning, Pst1 restriction fragments of total DNA containing TnphoA insertions from each mutant were ligated into linearized pUC18. The ligation products were used to transform E. coli DH5α, and transformants were selected on agar plates for the kanamycin resistance phenotype.

DNA sequencing and sequence analysis

Cosmid and plasmid DNA templates were prepared using the Qiagen plasmid midi kit. Single-stranded DNA was sequenced by the Euro Sequence Gene Service (Cybergene) according to the dideoxynucleotide chain termination method. Cloned insertion regions of the mutants were sequenced using an oligonucleotide primer that anneals to the insertion sequence (IS) of the transposon and directs the synthesis of the region immediately downstream of the transposon. Sequences of the fim genes were determined using extragenic oligonucleotide primers. DNA sequences were translated and analysed using gene jockey II software. DNA and amino acid sequence comparisons were performed with blast programs available from the National Center for Biotechnology Information at http://www.ncbi.nlm.nhi.gov.

DNA manipulations, hybridization and PCR experiments

Standard DNA procedures, PCR conditions and hybridization experiments were performed as described elsewhere (Sambrook et al., 1989). Minor modifications, if necessary, are detailed in the Results section. All PCR primer sequences will be made available upon request.

Transmission electron microscopy

To perform cross-sections of infected cultured cells, the cells were fixed with 3% glutaraldehyde in 0.2 M cacodylate buffer (pH 7.4) at 4°C for 2 h and post-fixed in 1% OsO4 in cacodylate buffer at 4°C for 1 h. After dehydration in a graded series of ethanol, the cultures were embedded in a 2-mm-thick Epon coating in the tissue culture well and polymerized for 3 days at 60°C. Suitable areas were reoriented parallel to the cell layer surface on Epon blocks with an Epon mixture. Ultrasections were contrasted with uranyl acetate and lead citrate.

Acknowledgements

This study was supported by grants from Association F. Aupetit, Société Nationale Française de Gastroentérologie (SNFGE), Institut de Recherche des Maladies de l'Appareil Digestif (IRMAD, Laboratoires Astra France) and the Ministère de la Recherche et de la Technologie (EA2148). J.B. was supported by a grant from IRMAD. We gratefully acknowledge Karen Krogfelt for providing the type 1 pili antiserum, Chantal Le Bouguénec for helpful discussions about the cosmid library, and Paul Orndorff for providing plasmid pORN104. The technical assistance of Chantal Rich, Marie-Agnès Bringer and Nathalie Rolhion is gratefully acknowledged. We thank Josiane Payen, Annie Fraisse and Monique Oron from the Electron Microscopy Department of Michel Bourges for technical assistance. We also thank Anne-Lise Glasser for helpful discussions.

Ancillary