Genetic Analysis of Enteropathogenic and Enterohemorrhagic Escherichia coli Serogroup O103 Strains by Molecular Typing of Virulence and Hous
http://www.100md.com
微生物临床杂志 2005年第4期
Division of Microbial Toxins, Department of Biological Safety, Robert Koch Institute, Berlin
Institut für Medizinische Mikrobiologie und Hygiene, Medizinische Fakultt Carl Gustav Carus, Technische Universitt Dresden, Dresden, Germany
Laboratoire Associe INRA/ENVT de Microbiologie Moleculaire, Ecole Nationale Veterinaire, Toulouse, France
ABSTRACT
We investigated the genetic relationships of 54 Escherichia coli O103 strains from humans, animals, and meat by molecular typing of housekeeping and virulence genes and by pulsed-field gel electrophoresis (PFGE). Multilocus sequence typing (MLST) of seven housekeeping genes revealed seven profiles, I through VII. MLST profiles I plus III cover 45 Shiga toxin-producing E. coli (STEC) O103:H2 strains from Australia, Canada, France, Germany, and Northern Ireland that are characterized by the intimin (eae) epsilon gene and carry enterohemorrhagic E. coli (EHEC) virulence plasmids. MLST profile II groups five human and animal enteropathogenic E. coli (EPEC) O103:H2 strains that were positive for intimin (eae) beta. Although strains belonging to MLST groups II and I plus III are closely related to each other (92.6% identity), major differences were found in the housekeeping icdA gene and in the virulence-associated genes eae and escD. E. coli O103 strains with MLST patterns IV to VII are genetically distant from MLST I, II, and III strains, as are the non-O103 E. coli strains EDL933 (O157), MG1655 (K-12), and CFT073 (O6). Comparison of MLST results with those of PFGE and virulence typing demonstrated that E. coli O103 STEC and EPEC have recently acquired different virulence genes and DNA rearrangements, causing alterations in their PFGE patterns. PFGE typing was very useful for identification of genetically closely related subgroups among MLST I strains, such as Stx2-producing STEC O103 strains from patients with hemolytic uremic syndrome. Analysis of virulence genes contributed to grouping of E. coli O103 strains into EPEC and STEC. Novel virulence markers, such as efa (EHEC factor for adherence), paa (porcine adherence factor), and cif (cell cycle-inhibiting factor), were found widely associated with E. coli O103 EPEC and STEC strains.
INTRODUCTION
Enteropathogenic Escherichia coli (EPEC) strains of serogroup O103 have been recognized since 1988 as the causative agents of diarrheal disease in rabbits, goats, and chickens (19, 44, 54). The pathogenic mechanisms of EPEC O103 were investigated, and progress has been made in unraveling the structure of the locus of enterocyte effacement (LEE) of rabbit and bovine EPEC O103:H2 strains (31, 56).
A second group of pathogens is Shiga toxin (Stx)-producing E. coli (STEC) O103, whose members were first described as diarrheagenic agents in calves in 1988 (12). In 1992, STEC O103 strains were identified as causative agents of hemolytic-uremic syndrome (HUS) in humans, and it was suggested that EPEC O103 from rabbits and STEC O103 from humans have a common origin but different mechanisms of adaptation to their hosts (39). In the following years, STEC O103 strains were isolated from cattle and sheep and from diseased humans in many countries and on different continents (7, 34, 43, 49, 55). STEC O103 represents one of the most frequently identified serovars among STEC types isolated in Germany and other European countries (17, 23, 46, 60).
It has been suggested that epidemiologically unrelated STEC O103 strains are clonally related on the basis of randomly amplified polymorphic DNA patterns and virulence gene typing (49). STEC O103 has been shown to contain some differences from other enterohemorrhagic E. coli (EHEC) types. Although most STEC O103 strains carry the enterohemolysin operon, which is located on the EHEC virulence plasmid, the hemolytic phenotype resembles that of alpha-hemolysin (45, 49). Among the other EHEC virulence plasmid genes, espP is not present, and only a portion of STEC O103 strains carry the katP gene (45, 49).
Prager et al. (45) reported a considerable type diversity among 145 human STEC O103 isolates, which were investigated by pulsed-field gel electrophoresis (PFGE), ribotyping, P-gene profiling, and electrotyping. However, only a little is known about the relationship between E. coli O103 EPEC and STEC strains. A common trait of both groups of strains is the H2 antigen, but differences in the LEE-associated intimin (eae) gene have been found. STEC O103:H2 strains are predominantly positive for eae-, whereas EPEC O103:H2 strains are positive for eae- (42).
The aim of our study was to compare E. coli O103 EPEC and STEC strains from different sources and geographical origins to explore their phylogenetic relationships and to identify the potential infection sources for humans. The phylogenetic analysis was performed by combining multilocus sequencing typing (MLST) and PFGE.
MATERIALS AND METHODS
Bacterial strains and growth conditions. The relevant characteristics of the 54 E. coli O103 strains used in this study are listed in Table 1. The strains originated from six different countries (France, Germany, United Kingdom, Australia, Canada, and the United States) and were isolated between 1993 and 2000. Forty strains were from human patients, and 14 strains were from animals (cattle, chicken, rabbits, pigs, a dog, and a goat). Strains F22S5-2, P94-6011, CA92-7164, 92-1695, FH303, and FH356 were from John Fairbrother, Montreal, Canada, and strain HBI/01 was provided by Hywel Ball, Belfast, Northern Ireland. Strain CB3349 was provided by Karl A. Bettelheim, Melbourne, Australia. The German O103 strains were isolated in Berlin (L. Beutin, Robert Koch Institute) and in Würzburg (H. Karch and H. Schmidt, University of Würzburg) by routine methods (11, 49). Strain PMK5 was from France, provided by Patricia Mariani-Kurkdjian, Department of Microbiology, Hpital Robert Debre, Paris, France, and strain UTI was donated by Phil I. Tarr, Washington University School of Medicine, St. Louis, Mo. The animal E. coli O103 strains E22, C124, 889, 890, and 891 were from Eric Oswald (INRA, Toulouse, France). E. coli H515b (O103:H8) was used as a reference for O seroytping (41). E. coli strains Bi7455-41 (O43:H2), U5-41 (O1:K1:H7), and Su4321-41(O13:K11:H11) were used as controls for fliC PCR (41). The E. coli K-12 strain C600 was used as a non-O103 control in different experiments (47).
Phenotypic and genetic examination of virulence markers and other traits of E. coli O103 strains. The detection of virulence genes was performed by PCR as summarized in Table 2. Subtyping of eae and stx genes was conducted by PstI and HincII/AccI restriction endonuclease digestion of PCR products, respectively, and subsequent analysis of restriction fragment patterns as described previously (5, 33, 64).
The hemolytic phenotype was investigated on washed sheep blood agar (entero-hemolysin agar; Oxoid, Wesel, Germany). Production of Shiga (Vero) toxins was investigated by the reverse passive latex agglutination assay for detection of Stx (VTEC-RPLA assay; Denka Seiken, Tokyo, Japan), and the cytotoxic activity of bacteria was examined with the Vero cell assay (10). Typing of O (lipopolysaccaride) and H (flagellar) antigens was performed as described earlier (41). Genotyping of the flagellar (fliC) genes was conducted by PCR and analysis of HhaI-digested PCR products (35). Briefly, the oligonucleotide primers fliC-1 [5' CAA GTC ATT AAT AC(A/C) AAC AGC C 3'] and fliC-2 [5' GAC AT(A/G) TT(A/G) GA(G/A/C) ACT TC(G/C) GT 3']were used for amplification of internal parts of fliC genes present in the E. coli reference strains as described previously (35). The PCR was performed for 25 cycles at 94°C for 60 s, 55°C for 60 s, and 72°C for 120 s. PCR products between 950 and 2,500 bp were obtained with E. coli reference strains for 53 different H types (41). Amplified DNA was digested with HhaI, and the resulting restriction fragments were compared on 2% agarose gels (Fig. 1). The gel images were stored digitally and analyzed with BioNumerics software, version 2.5 (Applied Maths, Kortrijk, Belgium) for similarity (Dice; complete linkage). PCR amplification of different parts of cif and cif-flanking regions was performed as listed in Table 3 and described previously (42). The efa/lif gene was detected with primers Efa1 fwd, Efa1 rev, Efa1 3'fwd, and Efa1 3'rev according to the method of Morabito et al. (40). Oligonucleotides were obtained from TIB MolBiol, Berlin, Germany.
Restriction enzyme digestion of total genomic DNA, PFGE, and calculation of similarity indices. Preparation of total genomic DNA for PFGE was performed as previously described (8). The samples were digested with XbaI and analyzed by PFGE with the CHEF-DR II system (Bio-Rad Laboratories, Munich, Germany). Lambda concatemers (New England Biolabs, Frankfurt am Main, Germany) were used as molecular weight standards. Digested DNA was separated on 1% agarose gels with 200 V for 25 h at 14°C with pulse times increasing from 5 to 40 s. The gels were stained with ethidium bromide for 1 h and digitalized. Evaluation of PFGE profiles for similarity was performed with bionumerics software with Dice similarity indices (complete linkage; optimization, 1%; position tolerance, 1.3%) as described previously (9).
Amplification of target genes for MLST. For MLST analysis, fresh single colonies, which were grown overnight on Luria-Bertani agar plates, were suspended in 50 μl of a 0.9% NaCl solution using sterile toothpicks and subsequently diluted 1:6 in 0.9% NaCl solution. PCR samples were prepared in a total volume of 50 μl containing 5 μl of this bacterial suspension, 5 μl of 10-fold-concentrated polymerase reaction buffer containing 15 mM MgCl2 (Promega, Mannheim, Germany), 200 μM (each) deoxynucleoside triphosphate, 30 pmol of each primer, and 1.5 U of Taq DNA polymerase (Promega). PCR was performed in a GeneAmp PCR System 2700 (Applied Biosystems Applera, Weiterstadt, Germany). PCR conditions depended on the different target genes and primer sequences used and are depicted in Table 2. Eight microliters of the PCR products was analyzed for purity on 0.7 to 1% Tris-borate-EDTA agarose gels using a maximum voltage of 5 V/cm.
Purification of PCR products and DNA sequencing. For removal of PCR primers, five microliters of the PCR product was mixed with two microliters of an Exo-Sap reaction mixture containing equal numbers of units of exonuclease I (20 U/μl) (New England Biolabs) and shrimp alkaline phosphatase (USB Biochemicals) (1 U/μl). Exonuclease I was diluted to 1 U/μl in 50 mM Tris-HCl, pH 7.5, and incubated at 37°C for 30 min and subsequently for 15 min at 80°C to inhibit further enzymatic activity.
DNA sequencing was performed using the BigDye terminator version 1.1 cycle-sequencing RR-100 premix (Applied Biosystems Applera). Briefly, 2 μl of the Exo-SAP-treated PCR product was mixed with 1.5 μl of primer (30 pmol/μl), 2.0 μl of premix, and 4.5 μl of distilled water and subjected to PCR under the same conditions described above and as depicted in Table 2.
After completion of PCR, the sequencing products were purified with Autoseq G-50 columns (Amersham Biosciences Europe GmbH, Freiburg, Germany) as recommended by the manufacturer. Approximately 10 μl was eluted from the columns and dried in a MaxiDryLyo microconcentrator (Heto Holten, Dreieich, Germany). After that, 50 μl of formamide and 25 μl of sequencing buffer were mixed, and 2 μl of this mixture was added to the dried sequencing product. The mixture was incubated for 2 min at 90°C and subsequently cooled on ice.
Software for DNA analysis. Following double-strand sequencing, the sequences were edited and aligned. DNA fragments between 410 and 725 bp long from each sequence were used for sequence analysis (Table 2). Raw DNA sequence data were analyzed with ABI 377 software. DNA sequences were edited and aligned with BioEdit, version 4.8.10 (http://jwbrown.mbio.ncsu.edu/BioEdit/bioedit.html) (27), converted into FASTA files, and loaded into S.T.A.R.T. (Sequence Type Analysis and Recombinational Tests) (http://outbreak.ceid.ox.ac.uk/software.htm). The dendrograms were compiled with S.T.A.R.T., using the unweighted pair group method with arithmetic mean (UPGMA). This software constructs a phylogenetic tree on the basis of allele numbers. Split decomposition analysis (4, 28) was performed with Splitstree version 2.0 (http://bibiserv.techfak.uni-bielefeld.de/splits/).
Nucleotide sequence accession numbers. The nucleotide sequences obtained by sequencing of the PCR products of each allele of all genes have been entered into the GenBank-EMBL databases under continuous accession numbers from AJ629457 to AJ629510 for adk, AJ629511 to AJ629564 for arcA, AJ629565 to AJ629618 for fumC, AJ629619 to AJ629672 for icdA, AJ629673 to AJ629726 for mdh, AJ629727 to AJ629780 for mtlD, AJ629781 to AJ629834 for pgi, AJ629916 to AJ629965 for eae, and AJ629966 to AJ630015 for escD.
RESULTS
Flagellar types of E. coli O103 strains. The O types of the 54 E. coli O103 strains investigated in this study were confirmed by serotyping using strain H515b (O103:H8) as a positive control (41). Forty-eight of the O103 strains were motile; 47 of these expressed flagellar antigen H2, and 1 strain (92-1695) was O103:H11 (Table 1). Six strains were nonmotile (NM), and the H types of these were determined by analysis of HhaI-digested fliC PCR products as described above (Fig. 1). Five of the O103:NM strains showed the fliC genotype H2, which was also present in all motile O103:H2 strains (data not shown). The remaining O103:NM strain, CA92-7164, was genotyped as O103:[H7] (Fig. 1).
Multilocus sequence analysis. Nucleotide sequence analysis of unlinked housekeeping genes (multilocus sequencing typing) is widely used for evolutionary and population analysis, and also for epidemiological investigations (36, 62). Since differences in the sequences of essential housekeeping genes are thought to display long-term genetic changes, we used this technique to investigate the phylogenetic relationships of E. coli O103 strains and to assess the similarity of strains belonging to phylogenetic groups by PFGE and to compare their virulence genes. DNA sequences of corresponding housekeeping genes from the E. coli K-12 strain MG1655 (accession no. NC_000913), the O157:H7 strain EDL933 (accession no. NC_002655), and the uropathogenic E. coli O6 strain CFT073 (accession no. NC_004431) were compiled from the respective published genome sequences.
In addition to the housekeeping genes adk, arcA, fumC, icdA, mdh, mtlD, and pgi, we used the LEE-harbored virulence genes eae and escD (Table 2) for all strains that were positive with the universal eae primers SK1 and SK2. The genes chosen for MLST were distributed equally throughout the chromosome to minimize the influence of possible chromosomal hot spots of mutation and recombination.
Sequences were obtained from all 54 E. coli O103 strains for all housekeeping loci investigated and could be aligned without gaps. The corresponding DNA sequences of E. coli strains MG1655 (E. coli K-12), EDL933 (O157:H7), and CFT073 (O6:K:H1) were included in our scheme. Each allele of the seven genes analyzed was numbered randomly in ascending order (Table 4). Numbers and types of base substitutions were not taken into account for allele designation (63). The combinations of allele numbers for all isolates are shown in Table 5. Each unique combination of allele numbers represents one sequence type (profile). Ten MLST profiles, I to X, were established (Table 5), and seven MLST profiles (I to VII) were attributed to the E. coli O103 strains. Forty-three (79.6%) of these showed MLST profile I, five strains (8.8%) showed MLST profile II, and the remaining six strains fell into MLST profiles III to VII.
Figure 2 shows a dendrogram calculated with UPGMA using the S.T.A.R.T software. Most of the strains belonging to MLST profile I are STEC from humans or animals from different geographical origins (Table 1). MLST profile II covers all EPEC O103:H2 strains from animals and a human, which originated from France and Germany. MLST profile III is comprised of two STEC strains from Germany (CB7320) and from the United Kingdom (HBI/01). Profiles IV to VII are represented by single O103 strains (C124, P94-6011, CA92-7164, or 92-1695) whose virulence attributes or H types were different from those of all other E. coli O103 strains that were investigated (Fig. 2 and Table 1). Single profiles called VII, IX, and X were attributed to the non-O103 strains MG1655, EDL933, and CFT073, the last being most distantly related to the E. coli O103 strain group.
Sequences of housekeeping genes were analyzed with Splitstree (28) for evidence of recombination. The algorithm used in this software is able to display conflicting results in the phylogenetic descent of sequences. A treelike structure is created when the descent is clonal, but an interconnecting network with a number of splits or a bushlike structure will appear when recombination plays a role in the phylogenetic history of genes. The Splitstree graph calculated with MLST data using a distance matrix and all 57 alleles of seven loci is a combination of treelike and bushlike structures (Fig. 3). Since Splitstree creates a cladogram, the lengths of the branches depict the relative phylogenetic distances. On the basis of Splitstree and UPGMA tree analysis data, we conclude that STEC and EPEC O103 strains from profiles I, II, and III are closely related, whereas the four strains with profiles IV to VII are phylogenetically more distant. Interestingly, three of the latter strains did not carry an eae gene, and two of these strains (CA92-7164 [O103:H7; MLST profile VI] and 92-1695 [O103:H11; MLST profile VII]) were different in flagellar antigen type from all other E. coli O103:H2 strains. The group of E. coli O103:H2 strains was found to be more closely related to EDL933 (O157:H7) than to E. coli K-12 or the uropathogenic E. coli O6 strain. Only some of the seven housekeeping genes analyzed with Splitstree showed discriminatory differences between strains of the O103 group. The clear distinction between clusters of O103 strains (Fig. 4) is mainly achieved from analysis of the icdA gene. The splits seen with fumC, icdA, mdh, and pgi occur mainly between E. coli O103 and non-O103 strains.
We were interested to see if the inclusion of the chromosomally (LEE) encoded virulence-associated genes eae and escD would increase the resolution of MLST within the E. coli O103 strains, and this analysis was performed with 51 eae-positive E. coli O103 strains (Table 1). As a result, the profiles II, III, and VII are represented by the same strains as in the primary housekeeping analysis. MLST profile I is split into two related profiles, IA and IB (data not shown). Group IB comprises strains CB8198, 3386/98, 2905/96, and RW2198, and group IA comprises all other strains of MLST profile I.
PFGE. A dendrogram based on similarities of XbaI-digested DNA PFGE patterns among E. coli O103 strains was created with the Bionumerics software using Dice similarity indices as described above. PFGE patterns were arranged in five clusters, A to E, which show >55% similarity in their banding patterns (Fig. 5). One strain (P94-6011) was not typeable by PFGE because it repeatedly produced only a smear of degraded DNA after PFGE (Fig. 5).
The 43 MLST I strains were subdivided by PFGE into four clusters, A to D. PFGE group A (64% similarity) consisted of 22 STEC strains, which were from diarrheic humans (n = 20) and healthy cattle (n = 2) from Germany. All group A strains were stx1, eae-, E-hly, etpD, and efa positive. Twenty of the group A strains carried the plasmid-encoded katP gene, and 21 strains were porcine adherence factor gene (paa) positive.
PFGE cluster B (59% similarity) is formed by seven MLST I strains (STEC types from cattle, humans, and a goat) and by the MLST II group strain CB8432, which belongs to the EPEC group. These strains originated in Australia (n = 1), the United States (1), France (1), and Germany (5). From its virulence profile, this group was similar to PFGE cluster A STEC strains, except for the EPEC strain CB8432, which showed close similarity to the animal EPEC MLST II profile strains in a different PFGE cluster, E (see below and Table 1).
Eleven STEC strains were grouped into PFGE cluster C (60% similarity), which could be further divided in two subclusters, C1 and C2. Cluster C1 (70% similarity) is formed by five O103:H2 strains belonging to MLST groups I and III, which originated from Ireland (meat), Germany (human), and Canada (two humans and a pig). All strains were positive for eae-, for the E-hly gene, and for the aerobactin receptor gene iutA; the last was present in only nine (16.6%) of all E. coli O103 isolates from this study (Table 6). Four of the strains produced Stx1.
All STEC O103 strains that were positive for stx2 were gathered in PFGE cluster C2 (80% similarity). The group of six stx2-positive STEC strains was closely associated with HUS cases, and the strains were isolated from patients in different parts of Germany between 1997 and 2000. All strains belonged to MLST profile I.
PFGE cluster D (55% similarity) is formed by seven strains belonging to MLST profiles I, IV, and VII (Fig. 5). The strains originated from humans (n = 5), a rabbit, and cattle. Cluster D strains were heterogeneous for all virulence markers investigated in this study (Table 1).
Cluster E (70% similarity) is formed by three EPEC O103:H2 strains, which were isolated from chickens and from a rabbit in France. All cluster E strains belong to the MLST II group, together with two other EPEC strains from a human (CB8432; cluster B) and a chicken (891; a single PFGE pattern).
Two other E. coli O103 strains from animals could not be assigned to any of the PFGE clusters defined here. These were P94-6011 (MLST profile V) and CA92-7164 (MLST profile VI). These strains were negative for Shiga toxins and for intimin.
Virulence gene typing of E. coli O103. A detailed genetic analysis of virulence markers known to be associated with EPEC, EHEC, and extraintestinal pathogenic E. coli was performed with all 54 E. coli O103 strains (Tables 1 and 6). Fifty-one (94.4%) of the strains were positive for an intimin gene (eae), and 44 (81.5%) of the strains produced Stx1 (stx1) or Stx2 (stx2). Three O103 strains from animals belonging to MLST profiles IV to VI (C124, P94-6011, and CA92-7164) (Table 1 and Fig. 2) were negative for both Stx and intimin. The intimin epsilon gene (eae-) was associated with all O103 MLST I (n = 43) and MLST III (n = 2) strains and was not found in strains belonging to other MLST groups. The intimin epsilon-positive strains were also positive for Shiga toxins, except two Stx-negative strains, which came from a pig (F22S5-2) and from a human (CB3349).
The eae- gene was present in all MLST type II strains, encompassing five EPEC O103:H2 strains, which were from chickens, a rabbit, and a human. In addition, an eae- gene was found in a single human STEC O103:H11 strain (92-1695; MLST VII), whose virulence markers, MLST type, and H type were different from those of all other O103 strains (Table 1 and Fig. 2).
None of the 54 O103 strains was positive for the bfpA (bundle-forming pilus) gene, which is associated with typical EPEC, or for the lpfA (long polar fimbria) gene, which is present in STEC O157 and in some LEE-negative, non-O157 STEC strains (22, 58) (Table 6).
In contrast, 51 (94.4%) of the O103 strains were positive for the EHEC factor for adherence (efa) gene, which is widely distributed in EHEC and EPEC strains (3). Efa/LifA were absent in only two eae- and stx-negative isolates from animals (C124 and CA92-7164) and in a human EPEC O103:H2 strain (CB8432) (Table 1). Another adherence-associated gene, paa (3), was found in 44 O103 strains, including EPEC and STEC types (Table 1).
The analysis of virulence genes in STEC and EPEC O103 strains revealed patterns which corresponded to different pathotypes. The presence of the stx2 gene in STEC O103 strains was closely associated with HUS cases, whereas most of the stx1-positive O103 strains from humans were associated with diarrhea. EHEC virulence plasmid genes encoding EHEC hemolysin (E-hlyA), type II secretion pathway protein D (etpD), and catalase-peroxidase (katP) were frequently associated with STEC O103 strains but were absent in EPEC O103 from animals and in eae-negative, stx-negative O103 strains. The plasmid-encoded serine protease gene (espP) was present in only two stx1-positive O103 strains from humans. The STEC O103 group could be subdivided in regard to the stx genes and to the presence of the plasmid-borne katP gene: E-hlyA, etpD, and katP genes were present in 22 stx1-positive strains from humans and cattle; the katP gene was not found in a second group of 11 stx1-positive strains from humans, cattle, and a goat; and a third group consisted of six stx2-positive human O103 strains from patients with HUS (Table 1). Only one STEC O103 strain from cattle (RW2198) was negative for all EHEC virulence plasmid-associated genes.
Genes encoding the aerobactin receptor (iutA), cytolethal distending toxin 1 (cdt-1), and cytotoxic necrotizing factor 1 (cnf) were associated with E. coli causing extraintestinal infections (21, 29, 37). None of the O103 strains from our study was positive for cnf and cdt-I genes, but nine O103 strains (STEC, EPEC, and others) carried the aerobactin receptor gene iutA. Interestingly, the iutA gene was found to dominate in O103 strains from Canada and the United States (Table 1).
Presence and characterization of the genetic background of the Cif gene. Cif (cell cycle-inhibiting factor) was recently described as a translocated effector molecule that blocks cell cycle G2/M transition (38). Interestingly, 51 (94.4%) of the 54 E. coli O103 strains investigated in our study were positive with primers cif-int-s and cif-int-as, which are complementary to the middle portion of the recently described cif gene (38). The three cif-negative strains C124, P94-6011, and CA92-7164, were also negative for eae, stx, and most other virulence markers (Table 1). Based on this high prevalence of cif, we were interested in the presence of a complete cif gene in these isolates. PCR with primers cif-rbs-s and cif-as, covering the entire cif gene, revealed a 950-bp PCR product with 45 O103 strains, indicating that a full-length cif gene was present (Fig. 6). However, in six strains, the PCR product was 2.2 kb in length, giving evidence for additional DNA in the cif region (Fig. 6, lanes 5 and 9). A third PCR with primers phage-s and ybhB, which are complementary to a bacteriophage gene in the neighborhood of the cif gene and the chromosomal ybhB gene close to the prophage-chromosome border, confirmed this finding (38). A 5-kb PCR product was found in the same 45 O103 strains, which showed a 950-bp product with primers cif-rbs-s and cif-as, and a 6.2-kb product was found in the same 6 strains, which showed a 2.2-kb fragment with the cif internal primers cif-rbs-s and cif-as (Fig. 6). These findings demonstrate that the cif genes found in the O103 strains are located in the genome of prophages and that these prophages are located in the same orientation in the same integration site.
Since there was evidence for additional DNA in the cif genes of six strains, we sequenced one of the 2.2-kb PCR products obtained with primers cif-rbs-s and cif-as and found a sequence that was identical to the insertion element IS2, which is inserted from positions 228 to1554 in the cif sequence in all six strains (data not shown). All strains carrying that IS2 element in the cif gene were stx2-positive STEC O103:H2 strains that were isolated from patients with HUS in Germany (Table 1).
DISCUSSION
Molecular epidemiology is largely descriptive and characterizes bacteria based on their natural genetic variation. A number of molecular methods have been employed to determine this variation, which may be due to mutation or horizontal gene transfer events. Basically, two branches of molecular epidemiology can be distinguished, classification and typing. Classification can illustrate evolutionary relationships and subdivides species into clonal groupings, whereas typing is used more often for differentiation of clinical or environmental isolates. Long-term epidemiology may require classification, whereas short-term local epidemiology is more often performed by typing methods. While the most prominent methods for molecular classification are MLST and multilocus enzyme electrophoresis (36, 18), PFGE or PCR-based methods, such as randomly amplified polymorphic DNA-PCR, AFLP, or virulence gene characterization, have been used frequently for typing approaches (for a review, see reference 1).
In our study, we combined MLST and PFGE with PCR detection of virulence genes for classification of pathotypes and for phylogenetic analysis of E. coli O103 strains, which constitute a group of human and animal pathogenic E. coli. In order to include the prominent pathotypes occurring among strains of the E. coli O103 group, we investigated strains from different places and different sources, which were isolated over a 7-year period. The strains were not known to be epidemiologically linked and were not associated with outbreaks. Analysis of virulence genes of E. coli O103 strains resulted in the detection of two major pathotypes, EPEC and STEC, which are present in both humans and animals.
MLST with housekeeping genes of E. coli O103 strains has demonstrated seven profiles, and profiles I and III encompass a group of phylogenetically closely related O103:H2 strains. Profiles I and III occurred most frequently, being present in 83.3% of the strains which were isolated in Australia, Canada, France, Germany, Ireland, and the United States. The leading markers of profile I and III strains are production of Stx and intimin and the presence of EHEC plasmid genes, characterizing them as typical STEC. Only two intimin -positive strains were found to be negative for Stx, but they were grouped with other STEC strains by PFGE and by MLST. It is therefore possible that these strains have lost the phage-borne stx genes upon subculture, as reported for other STEC (32, 53).
MLST profile II was associated with five EPEC O103:H2 strains from animals and a human. Data from phylogenetic analysis suggest that profile I and III STEC O103:H2 strains have evolved from MLST profile II strains. Major differences between profile II and profile I and III strains were found in only one of the seven housekeeping genes (icdA) that were analyzed and in the eae and escD sequences, whereas the eae and escD MLST patterns were not identical (data not shown). Analysis of chromosomally inherited virulence genes, such as eae, was found to be helpful for establishing phylogenetic groups among closely related EPEC and STEC O103:H2 strains.
MLST group IV to VII O103 strains were found to be phylogenetically far distant from MLST group I, II, and III strains, as were the non-O103 strains EDL933, MG1655, and CFT073. This finding indicates that strains belonging to the same O serogroup may have evolved separately from each other and that the O serogroup as such cannot be taken as an indicator of genetic relationships among E. coli strains.
By PFGE, MLST profile I and III strains were divided into four clusters, which may have been generated by intra- and intergenomic recombination events. A few MLST profile I strains showed very similar PFGE patterns, which could indicate that they are epidemiologically related (57). PFGE typing was found to be very useful for the identification of genetically closely related groups of highly pathogenic Stx2-producing STEC O103:H2 (C1 cluster; 80% similarity) among strains belonging to the same MLST profile I. This finding shows that PFGE is highly useful to detect groups of strains which have probably evolved very recently and which are therefore not detectable by MLST, which indicates differences that have evolved more slowly by mutation and recombination. On the other hand, PFGE typing alone is not sufficient to define groups of strains which are phylogenetically closely related, as was the case with MLST II group EPEC strains, which were separated into three different unrelated PFGE profiles. The combination of MLST, PFGE, and virulence gene typing allowed us to define EPEC and STEC O103 strains as two separate groups of pathogens that not only differ from each other in their virulence genes but represent two clonal lineages with different MLST profiles. Similarly, by MLST, Stx- and eae-negative E. coli O103 strains were identified as evolutionarily distant form STEC and EPEC O103, excluding the possibility that they were EPEC or STEC derivative strains that have simply lost their virulence markers.
Earlier studies showed that the gene compositions of the large virulence plasmids vary among STEC strains, but virulence gene patterns appeared to be serotype associated (14, 49). This is reflected in our experiments showing a close association of E-hlyA (93.1%) and etp (90.1%) genes with STEC O103 strains. In contrast, the katP gene was found in only 21 (47.7%) STEC O103 strains but is closely associated (90.1%) with STEC strains belonging to PFGE group A. This finding indicates that most of the katP-positive STEC strains belong to a genetically closely related subgroup among STEC O103 strains. The presence of a plasmid-borne espP gene in only two strains indicates that the ancestral STEC O103 strains lacked espP, which might have been acquired more recently in the evolution of STEC O103.
E. coli O103 EPEC strains were isolated from rabbits, chickens, and a human. These strains were classified as atypical EPEC, because none was positive for the bfpA gene, which is located on the EAF plasmid in typical EPEC strains (61). EPEC O103 strains were different from other intimin-positive, Stx-negative O103 strains (see below) in their intimin type (1) and their MLST profile.
Interestingly, the novel adhesion factors encoded by paa and efa and the cytotoxin encoded by the cif gene were very frequent in STEC and EPEC O103 strains and are probably important for their virulence (3, 6, 38). However, we cannot conclude that all cif genes were functional in our strains, since it had been shown earlier that cif in strain PMK5 contains an internal stop codon (42).The efa gene has been shown to be associated with the LEE in several E. coli serotypes (40).
The virulence genes of the group of six Stx2-producing STEC O103 strains (PFGE cluster C2; 80% similarity) were very similar. Interestingly, these strains differed from all 51 other cif-positive O103 strains by having an IS2 insertion located in their cif genes. None of these strains was related to an outbreak. Together with the high similarity found by PFGE, the IS2 insertion in cif indicates that these strains represent an evolutionarily young clone that might have evolved around 1996 and spread throughout Germany. The presence of the stx2 gene could be important for the capacity of these strains to cause HUS, since it was previously reported that Stx2, in contrast to Stx1 and Stx2c, is related to severe disease and HUS in E. coli O157 and other STEC strains (9, 13, 25). The detection of stx2-positive STEC O103 strains indicates that highly virulent strains may evolve from already existing human pathogenic STEC O103 strains, probably by horizontal transfer of phage-encoded stx2 genes. Further research is necessary to identify the evolutionary factors involved in the generation of highly virulent STEC O103 clones.
Only one (UTI) of the O103 strains from our study was associated with extraintestinal infections of humans. Among certain virulence markers that were associated with extraintestinal pathogenic E. coli, only the aerobactin receptor gene (iutA) was found to be present in 9 of the 54 O103 strains. Interestingly, the iutA gene was present in 5 of 8 strains isolated in North America but in only 4 of 46 strains from other places. The iutA gene is known to be located on plasmids in E. coli (20), which could contribute to the spread of this marker to strains that were unrelated in their sources and PFGE and MLST types.
E. coli strains of serogroup O103 are of particular interest, since they are widespread as commensals and pathogens in animals and humans. Although the clonal diversity and some pathogenic properties of E. coli O103 strains have been published, the present study extends our knowledge about the virulence spectrum and the genetic relationships of E. coli O103 strains from different sources and geographical origins. Phylogenetic analysis has demonstrated that EPEC and STEC strains of serotype O103:H2 are genetically closely related despite their different virulence spectra. This points to a recent emergence of pathogenic E. coli O103 strains and indicates that this serotype might possess a very successful combination of traits. STEC and EPEC O103 represent two different lineages, and the risk of human disease may be increased by infection with STEC O103 from the bovine and/or goat reservoir. This should be elucidated in more detail in future research approaches.
ACKNOWLEDGMENTS
We are grateful to John Fairbrother, Montreal, Canada; Hywel Ball, Belfast, Northern Ireland; Patricia Marjiani-Kurkdjian, Paris, France; Karl A. Bettelheim, Melbourne, Australia; and Phil Tarr, St. Louis, Mo., for donating E. coli O103 strains from their laboratory collections. We also thank Kerstin Rydzewski and Manuela Brandt for skillful and highly motivated technical assistance.
This work was supported by grants from the European Union (EU project QLK2-2000-0060).
The authors are solely responsible for the work described in this paper, and their opinions are not necessarily those of the EU.
REFERENCES
Achtman, M. 2001. A phylogenetic perspective on molecular epidemiology, p. 485-509. In M. Sussmann (ed.), Molecular medical microbiology. Academic Press, San Diego, Calif.
An, H., J. M. Fairbrother, C. Desautels, and J. Harel. 1999. Distribution of a novel locus called Paa (porcine attaching and effacing associated) among enteric Escherichia coli. Adv. Exp. Med. Biol. 473:179-184.
Badea, L., S. Doughty, L. Nicholls, J. Sloan, R. M. Robins-Browne, and E. L. Hartland. 2003. Contribution of Efa1/LifA to the adherence of enteropathogenic Escherichia coli to epithelial cells. Microb. Pathog. 34:205-215.
Bandelt, H. J., and A. W. Dress. 1992. Split decomposition: a new and useful approach to phylogenetic analysis of distance data. Mol. Phylogenet. Evol. 1:242-252.
Bastian, S. N., I. Carle, and F. Grimont. 1998. Comparison of 14 PCR systems for the detection and subtyping of stx genes in Shiga-toxin-producing Escherichia coli. Res. Microbiol. 149:457-472.
Batisson, I., M. P. Guimond, F. Girard, H. An, C. Zhu, E. Oswald, J. M. Fairbrother, M. Jacques, and J. Harel. 2003. Characterization of the novel factor Paa involved in the early steps of the adhesion mechanism of attaching and effacing Escherichia coli. Infect. Immun. 71:4516-4525.
Bettelheim, K. A., M. A. Hornitzky, and S. P. Djordjevic. 2001. First bovine and ovine isolations of Shiga toxin-producing Escherichia coli O103:H2 in Australia. Aust. Vet. J. 79:289-290.
Beutin, L., D. Geier, S. Zimmermann, S. Aleksic, H. A. Gillespie, and T. S. Whittam. 1997. Epidemiological relatedness and clonal types of natural populations of Escherichia coli strains producing Shiga toxins in separate populations of cattle and sheep. Appl. Environ. Microbiol. 63:2175-2180.
Beutin, L., S. Kaulfuss, T. Cheasty, B. Brandenburg, S. Zimmermann, K. Gleier, G. A. Willshaw, and H. R. Smith. 2002. Characteristics and association with disease of two major subclones of Shiga toxin (Verocytotoxin)-producing strains of Escherichia coli (STEC) O157 that are present among isolates from patients in Germany. Diagn. Microbiol. Infect. Dis. 44:337-346.
Beutin, L., S. Zimmermann, and K. Gleier. 1996. Rapid detection and isolation of Shiga-like toxin (verocytotoxin)-producing Escherichia coli by direct testing of individual enterohemolytic colonies from washed sheep blood agar plates in the VTEC-RPLA assay. J. Clin. Microbiol. 34:2812-2814.
Beutin, L., S. Zimmermann, and K. Gleier. 1998. Human infections with Shiga toxin-producing Escherichia coli other than serogroup O157 in Germany. Emerg. Infect. Dis. 4:635-639.
Blanco, J., E. A. Gonzalez, S. Garcia, M. Blanco, B. Regueiro, and I. Bernardez. 1988. Production of toxins by Escherichia coli strains isolated from calves with diarrhoea in Galicia (north-western Spain). Vet. Microbiol. 18:297-311.
Boerlin, P., S. A. McEwen, F. Boerlin-Petzold, J. B. Wilson, R. P. Johnson, and C. L. Gyles. 1999. Associations between virulence factors of Shiga toxin-producing Escherichia coli and disease in humans. J. Clin. Microbiol. 37:497-503.
Brunder, W., H. Schmidt, M. Frosch, and H. Karch. 1999. The large plasmids of Shiga-toxin-producing Escherichia coli (STEC) are highly variable genetic elements. Microbiology 145:1005-1014.
Brunder, W., H. Schmidt, and H. Karch. 1996. KatP, a novel catalase-peroxidase encoded by the large plasmid of enterohaemorrhagic Escherichia coli O157:H7. Microbiology 142:3305-3315.
Brunder, W., H. Schmidt, and H. Karch. 1997. EspP, a novel extracellular serine protease of enterohaemorrhagic Escherichia coli O157:H7 cleaves human coagulation factor V. Mol. Microbiol. 24:767-778.
Caprioli, A., and A. E. Tozzi. 1996. Epidemiology of Shiga toxin-producing Escherichia coli infections in continental Europe, p. 38-48. In J. B. Kaper and A. D. O'Brien (ed.), Escherichia coli O157:H7 and other Shiga toxin-producing E. coli strains. ASM Press, Washington, D.C.
Caugant, D. A., K. Bovre, P. Gaustad, K. Bryn, E. Holten, E. A. Hoiby, and L. O. Froholm. 1986. Multilocus genotypes determined by enzyme electrophoresis of Neisseria meningitidis isolated from patients with systemic disease and from healthy carriers. J. Gen. Microbiol. 132:641-652.
Cid, D., M. Blanco, J. E. Blanco, J. A. Ruiz Santa Quiteira, R. de la Fuente, and J. Blanco. 1996. Serogroups, toxins and antibiotic resistance of Escherichia coli strains isolated form diarrhoeic goat kids in Spain. Vet. Microbiol. 53:349-354.
De Lorenzo, V., A. Bindereif, B. H. Paw, and J. B. Neilands. 1986. Aerobactin biosynthesis and transport genes of plasmid ColV-K30 in Escherichia coli K-12. J. Bacteriol. 165:570-578.
De Rycke, J., A. Milon, and E. Oswald. 1999. Necrotoxic Escherichia coli (NTEC): two emerging categories of human and animal pathogens. Vet. Res. 30:221-233.
Doughty, S., J. Sloan, V. Bennett-Wood, M. Robertson, R. M. Robins-Browne, and E. L. Hartland. 2002. Identification of a novel fimbrial gene cluster related to long polar fimbriae in locus of enterocyte effacement-negative strains of enterohemorrhagic Escherichia coli. Infect. Immun. 70:6761-6769.
Eklund, M., F. Scheutz, and A. Siitonen. 2001. Clinical isolates of non-O157 Shiga toxin-producing Escherichia coli: serotypes, virulence characteristics, and molecular profiles of strains of the same serotype. J. Clin. Microbiol. 39:2829-2834.
Elliott, S. J., L. A. Wainwright, T. K. McDaniel, K. G. Jarvis, Y. K. Deng, L. C. Lai, B. P. McNamara, M. S. Donnenberg, and J. B. Kaper. 1998. The complete sequence of the locus of enterocyte effacement (LEE) from enteropathogenic Escherichia coli E2348/69. Mol. Microbiol. 28:1-4.
Friedrich, A. W., M. Bielaszewska, W. L. Zhang, M. Pulz, T. Kuczius, A. Ammon, and H. Karch. 2002. Escherichia coli harboring Shiga toxin 2 gene variants: frequency and association with clinical symptoms. J. Infect. Dis. 185:74-84.
Gunzburg, S. T., N. G. Tornieporth, and L. W. Riley. 1995. Identification of enteropathogenic Escherichia coli by PCR-based detection of the bundle-forming pilus gene. J. Clin. Microbiol. 33:1375-1377.
Hall, T. A. 2004. BioEdit: a user-friendly biological sequence alignment editor and analysis program for Windows 95/98/NT. Nucleic Acids Symp. Ser. 41:95-98.
Huson, D. H. 1998. SplitsTree: analyzing and visualizing evolutionary data. Bioinformatics 14:68-73.
Johnson, J. R., T. T. O'Bryan, P. Delavari, M. Kuskowski, A. Stapleton, U. Carlino, and T. A. Russo. 2001. Clonal relationships and extended virulence genotypes among Escherichia coli isolates from women with a first or recurrent episode of cystitis. J. Infect. Dis. 183:1508-1517.
Johnson, J. R., A. E. Stapleton, T. A. Russo, F. Scheutz, J. J. Brown, and J. N. Maslow. 1997. Characteristics and prevalence within serogroup O4 of a J96-like clonal group of uropathogenic Escherichia coli O4:H5 containing the class I and class III alleles of papG. Infect. Immun. 65:2153-2159.
Jores, J., L. Rumer, S. Kiessling, J. B. Kaper, and L. H. Wieler. 2001. A novel locus of enterocyte effacement (LEE) pathogenicity island inserted at pheV in bovine Shiga toxin-producing Escherichia coli strain O103:H2. FEMS Microbiol. Lett. 204:75-79.
Karch, H., T. Meyer, H. Rüssmann, and J. Heesemann. 1992. Frequent loss of Shiga-like toxin genes in clinical isolates of Escherichia coli upon subcultivation. Infect. Immun. 60:3464-3467.
Lin, Z., H. Kurazono, S. Yamasaki, and Y. Takeda. 1993. Detection of various variant verotoxin genes in Escherichia coli by polymerase chain reaction. Microbiol. Immunol. 37:543-548.
Luzzi, I., A. E. Tozzi, G. Rizzoni, A. Niccolini, I. Benedetti, F. Minelli, and A. Caprioli. 1995. Detection of serum antibodies to the lipopolysaccharide of Escherichia coli O103 in patients with hemolytic-uremic syndrome. J. Infect. Dis. 171:514-515.
Machado, J., F. Grimont, and P. A. Grimont. 2000. Identification of Escherichia coli flagellar types by restriction of the amplified fliC gene. Res. Microbiol. 151:535-546.
Maiden, M. C., J. A. Bygraves, E. Feil, G. Morelli, J. E. Russell, R. Urwin, Q. Zhang, J. Zhou, K. Zurth, D. A. Caugant, I. M. Feavers, M. Achtman, and B. G. Spratt. 1998. Multilocus sequence typing: a portable approach to the identification of clones within populations of pathogenic microorganisms. Proc. Natl. Acad. Sci. USA 95:3140-3145.
Mainil, J. G., E. Jacquemin, and E. Oswald. 2003. Prevalence and identity of cdt-related sequences in necrotoxigenic Escherichia coli. Vet. Microbiol. 94:159-165.
Marches, O., T. N. Ledger, M. Boury, M. Ohara, X. Tu, F. Goffaux, J. Mainil, I. Rosenshine, M. Sugai, J. De Rycke, and E. Oswald. 2003. Enteropathogenic and enterohaemorrhagic Escherichia coli deliver a novel effector called Cif, which blocks cell cycle G2/M transition. Mol. Microbiol. 50:1553-1567.
Mariani-Kurkdjian, P., E. Denamur, A. Milon, B. Picard, H. Cave, N. Lambert-Zechovsky, C. Loirat, P. Goullet, P. J. Sansonetti, and J. Elion. 1993. Identification of a clone of Escherichia coli O103:H2 as a potential agent of hemolytic-uremic syndrome in France. J. Clin. Microbiol. 31:296-301.
Morabito, S., R. Tozzoli, E. Oswald, and A. Caprioli. 2003. A mosaic pathogenicity island made up of the locus of enterocyte effacement and a pathogenicity island of Escherichia coli O157:H7 is frequently present in attaching and effacing E. coli. Infect. Immun. 71:3343-3348.
Orskov, F., and I. Orskov. 1984. Serotyping of Escherichia coli. Methods Microbiol. 14:43-112.
Oswald, E., H. Schmidt, S. Morabito, H. Karch, O. Marches, and A. Caprioli. 2000. Typing of intimin genes in human and animal enterohemorrhagic and enteropathogenic Escherichia coli: characterization of a new intimin variant. Infect. Immun. 68:64-71.
Parma, A. E., M. E. Sanz, J. E. Blanco, J. Blanco, M. R. Vinas, M. Blanco, N. L. Padola, and A. I. Etcheverria. 2000. Virulence genotypes and serotypes of verotoxigenic Escherichia coli isolated from cattle and foods in Argentina. Importance in public health. Eur. J. Epidemiol. 16:757-762.
Peeters, J. E., R. Geeroms, and F. Orskov. 1988. Biotype, serotype, and pathogenicity of attaching and effacing enteropathogenic Escherichia coli strains isolated from diarrheic commercial rabbits. Infect. Immun. 56:1442-1448.
Prager, R., A. Liesegang, W. Voigt, W. Rabsch, A. Fruth, and H. Tschpe. 2002. Clonal diversity of Shiga toxin-producing Escherichia coli O103:H2/H– in Germany. Infect. Genet. Evol. 1:265-275.
Robert Koch Institut. 2004. Bakterielle Gastroenteritiden in Deutschland 2001. Epidemiol. Bull. 50:417-422.
Sambrook, J., E. F. Fritsch, and T. Maniatis. 1989. Molecular cloning: a laboratory manual, 2nd ed. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, N.Y.
Schmidt, H., L. Beutin, and H. Karch. 1995. Molecular analysis of the plasmid-encoded hemolysin of Escherichia coli O157:H7 strain EDL 933. Infect. Immun. 63:1055-1061.
Schmidt, H., C. Geitz, P. I. Tarr, M. Frosch, and H. Karch. 1999. Non-O157:H7 pathogenic Shiga toxin-producing Escherichia coli: phenotypic and genetic profiling of virulence traits and evidence for clonality. J. Infect. Dis. 179:115-123.
Schmidt, H., B. Henkel, and H. Karch. 1997. A gene cluster closely related to type II secretion pathway operons of gram-negative bacteria is located on the large plasmid of enterohemorrhagic Escherichia coli O157 strains. FEMS Microbiol. Lett. 148:265-272.
Schmidt, H., H. Rüssmann, and H. Karch. 1993. Virulence determinants in nontoxinogenic Escherichia coli O157 strains that cause infantile diarrhea. Infect. Immun. 61:4894-4898.
Schmidt, H., H. Rüssmann, A. Schwarzkopf, S. Aleksic, J. Heesemann, and H. Karch. 1994. Prevalence of attaching and effacing Escherichia coli in stool samples from patients and controls. Zentbl. Bakteriol. 281:201-213.
Schmidt, H., J. Scheef, H. I. Huppertz, M. Frosch, and H. Karch. 1999. Escherichia coli O157:H7 and O157:H– strains that do not produce Shiga toxin: phenotypic and genetic characterization of isolates associated with diarrhea and hemolytic-uremic syndrome. J. Clin. Microbiol. 37:3491-3496.
Sueyoshi, M., H. Fukui, S. Tanaka, M. Nakazawa, and K. Ito. 1996. A new adherent form of an attaching and effacing Escherichia coli (eaeA+, bfp–) to the intestinal epithelial cells of chicks. J. Vet. Med. Sci. 58:1145-1147.
Tarr, P. I., L. S. Fouser, A. E. Stapleton, R. A. Wilson, H. H. Kim, J. C. Vary, Jr., and C. R. Clausen. 1996. Hemolytic-uremic syndrome in a six-year-old girl after a urinary tract infection with Shiga-toxin-producing Escherichia coli O103:H2. N. Engl. J. Med. 335:635-638.
Tauschek, M., R. A. Strugnell, and R. M. Robins-Browne. 2002. Characterization and evidence of mobilization of the LEE pathogenicity island of rabbit-specific strains of enteropathogenic Escherichia coli. Mol. Microbiol. 44:1533-1550.
Tenover, F. C., R. D. Arbeit, R. V. Goering, P. A. Mickelsen, B. E. Murray, D. H. Persing, and B. Swaminathan. 1995. Interpreting chromosomal DNA restriction patterns produced by pulsed-field gel electrophoresis: criteria for bacterial strain typing. J. Clin. Microbiol. 33:2233-2239.
Torres, A. G., J. A. Giron, N. T. Perna, V. Burland, F. R. Blattner, F. Avelino-Flores, and J. B. Kaper. 2002. Identification and characterization of lpfABCC'DE, a fimbrial operon of enterohemorrhagic Escherichia coli O157:H7. Infect. Immun. 70:5416-5427.
Toth, I., F. Herault, L. Beutin, and E. Oswald. 2003. Production of cytolethal distending toxins by pathogenic Escherichia coli strains isolated from human and animal sources: establishment of the existence of a new cdt variant (type IV). J. Clin. Microbiol. 41:4285-4291.
Tozzi, A. E., A. Caprioli, F. Minelli, A. Gianviti, L. De Petris, A. Edefonti, G. Montini, A. Ferretti, T. De Palo, M. Gaido, and G. Rizzoni. 2003. Shiga toxin-producing Escherichia coli infections associated with hemolytic uremic syndrome, Italy, 1988-2000. Emerg. Infect. Dis. 9:106-108.
Trabulsi, L. R., R. Keller, and T. A. Tardelli Gomes. 2002. Typical and atypical enteropathogenic Escherichia coli. Emerg. Infect. Dis. 8:508-513.
Urwin, R., and M. C. Maiden. 2003. Multi-locus sequence typing: a tool for global epidemiology. Trends Microbiol. 11:479-487.
Wagner, P. L., D. W. Acheson, and M. K. Waldor. 2001. Human neutrophils and their products induce Shiga toxin production by enterohemorrhagic Escherichia coli. Infect. Immun. 69:1934-1937.
Zhang, W. L., B. Khler, E. Oswald, L. Beutin, H. Karch, S. Morabito, A. Caprioli, S. Suerbaum, and H. Schmidt. 2002. Genetic diversity of intimin genes of attaching and effacing Escherichia coli strains. J. Clin. Microbiol. 40:4486-4492.(Lothar Beutin, Stefan Kau)
Institut für Medizinische Mikrobiologie und Hygiene, Medizinische Fakultt Carl Gustav Carus, Technische Universitt Dresden, Dresden, Germany
Laboratoire Associe INRA/ENVT de Microbiologie Moleculaire, Ecole Nationale Veterinaire, Toulouse, France
ABSTRACT
We investigated the genetic relationships of 54 Escherichia coli O103 strains from humans, animals, and meat by molecular typing of housekeeping and virulence genes and by pulsed-field gel electrophoresis (PFGE). Multilocus sequence typing (MLST) of seven housekeeping genes revealed seven profiles, I through VII. MLST profiles I plus III cover 45 Shiga toxin-producing E. coli (STEC) O103:H2 strains from Australia, Canada, France, Germany, and Northern Ireland that are characterized by the intimin (eae) epsilon gene and carry enterohemorrhagic E. coli (EHEC) virulence plasmids. MLST profile II groups five human and animal enteropathogenic E. coli (EPEC) O103:H2 strains that were positive for intimin (eae) beta. Although strains belonging to MLST groups II and I plus III are closely related to each other (92.6% identity), major differences were found in the housekeeping icdA gene and in the virulence-associated genes eae and escD. E. coli O103 strains with MLST patterns IV to VII are genetically distant from MLST I, II, and III strains, as are the non-O103 E. coli strains EDL933 (O157), MG1655 (K-12), and CFT073 (O6). Comparison of MLST results with those of PFGE and virulence typing demonstrated that E. coli O103 STEC and EPEC have recently acquired different virulence genes and DNA rearrangements, causing alterations in their PFGE patterns. PFGE typing was very useful for identification of genetically closely related subgroups among MLST I strains, such as Stx2-producing STEC O103 strains from patients with hemolytic uremic syndrome. Analysis of virulence genes contributed to grouping of E. coli O103 strains into EPEC and STEC. Novel virulence markers, such as efa (EHEC factor for adherence), paa (porcine adherence factor), and cif (cell cycle-inhibiting factor), were found widely associated with E. coli O103 EPEC and STEC strains.
INTRODUCTION
Enteropathogenic Escherichia coli (EPEC) strains of serogroup O103 have been recognized since 1988 as the causative agents of diarrheal disease in rabbits, goats, and chickens (19, 44, 54). The pathogenic mechanisms of EPEC O103 were investigated, and progress has been made in unraveling the structure of the locus of enterocyte effacement (LEE) of rabbit and bovine EPEC O103:H2 strains (31, 56).
A second group of pathogens is Shiga toxin (Stx)-producing E. coli (STEC) O103, whose members were first described as diarrheagenic agents in calves in 1988 (12). In 1992, STEC O103 strains were identified as causative agents of hemolytic-uremic syndrome (HUS) in humans, and it was suggested that EPEC O103 from rabbits and STEC O103 from humans have a common origin but different mechanisms of adaptation to their hosts (39). In the following years, STEC O103 strains were isolated from cattle and sheep and from diseased humans in many countries and on different continents (7, 34, 43, 49, 55). STEC O103 represents one of the most frequently identified serovars among STEC types isolated in Germany and other European countries (17, 23, 46, 60).
It has been suggested that epidemiologically unrelated STEC O103 strains are clonally related on the basis of randomly amplified polymorphic DNA patterns and virulence gene typing (49). STEC O103 has been shown to contain some differences from other enterohemorrhagic E. coli (EHEC) types. Although most STEC O103 strains carry the enterohemolysin operon, which is located on the EHEC virulence plasmid, the hemolytic phenotype resembles that of alpha-hemolysin (45, 49). Among the other EHEC virulence plasmid genes, espP is not present, and only a portion of STEC O103 strains carry the katP gene (45, 49).
Prager et al. (45) reported a considerable type diversity among 145 human STEC O103 isolates, which were investigated by pulsed-field gel electrophoresis (PFGE), ribotyping, P-gene profiling, and electrotyping. However, only a little is known about the relationship between E. coli O103 EPEC and STEC strains. A common trait of both groups of strains is the H2 antigen, but differences in the LEE-associated intimin (eae) gene have been found. STEC O103:H2 strains are predominantly positive for eae-, whereas EPEC O103:H2 strains are positive for eae- (42).
The aim of our study was to compare E. coli O103 EPEC and STEC strains from different sources and geographical origins to explore their phylogenetic relationships and to identify the potential infection sources for humans. The phylogenetic analysis was performed by combining multilocus sequencing typing (MLST) and PFGE.
MATERIALS AND METHODS
Bacterial strains and growth conditions. The relevant characteristics of the 54 E. coli O103 strains used in this study are listed in Table 1. The strains originated from six different countries (France, Germany, United Kingdom, Australia, Canada, and the United States) and were isolated between 1993 and 2000. Forty strains were from human patients, and 14 strains were from animals (cattle, chicken, rabbits, pigs, a dog, and a goat). Strains F22S5-2, P94-6011, CA92-7164, 92-1695, FH303, and FH356 were from John Fairbrother, Montreal, Canada, and strain HBI/01 was provided by Hywel Ball, Belfast, Northern Ireland. Strain CB3349 was provided by Karl A. Bettelheim, Melbourne, Australia. The German O103 strains were isolated in Berlin (L. Beutin, Robert Koch Institute) and in Würzburg (H. Karch and H. Schmidt, University of Würzburg) by routine methods (11, 49). Strain PMK5 was from France, provided by Patricia Mariani-Kurkdjian, Department of Microbiology, Hpital Robert Debre, Paris, France, and strain UTI was donated by Phil I. Tarr, Washington University School of Medicine, St. Louis, Mo. The animal E. coli O103 strains E22, C124, 889, 890, and 891 were from Eric Oswald (INRA, Toulouse, France). E. coli H515b (O103:H8) was used as a reference for O seroytping (41). E. coli strains Bi7455-41 (O43:H2), U5-41 (O1:K1:H7), and Su4321-41(O13:K11:H11) were used as controls for fliC PCR (41). The E. coli K-12 strain C600 was used as a non-O103 control in different experiments (47).
Phenotypic and genetic examination of virulence markers and other traits of E. coli O103 strains. The detection of virulence genes was performed by PCR as summarized in Table 2. Subtyping of eae and stx genes was conducted by PstI and HincII/AccI restriction endonuclease digestion of PCR products, respectively, and subsequent analysis of restriction fragment patterns as described previously (5, 33, 64).
The hemolytic phenotype was investigated on washed sheep blood agar (entero-hemolysin agar; Oxoid, Wesel, Germany). Production of Shiga (Vero) toxins was investigated by the reverse passive latex agglutination assay for detection of Stx (VTEC-RPLA assay; Denka Seiken, Tokyo, Japan), and the cytotoxic activity of bacteria was examined with the Vero cell assay (10). Typing of O (lipopolysaccaride) and H (flagellar) antigens was performed as described earlier (41). Genotyping of the flagellar (fliC) genes was conducted by PCR and analysis of HhaI-digested PCR products (35). Briefly, the oligonucleotide primers fliC-1 [5' CAA GTC ATT AAT AC(A/C) AAC AGC C 3'] and fliC-2 [5' GAC AT(A/G) TT(A/G) GA(G/A/C) ACT TC(G/C) GT 3']were used for amplification of internal parts of fliC genes present in the E. coli reference strains as described previously (35). The PCR was performed for 25 cycles at 94°C for 60 s, 55°C for 60 s, and 72°C for 120 s. PCR products between 950 and 2,500 bp were obtained with E. coli reference strains for 53 different H types (41). Amplified DNA was digested with HhaI, and the resulting restriction fragments were compared on 2% agarose gels (Fig. 1). The gel images were stored digitally and analyzed with BioNumerics software, version 2.5 (Applied Maths, Kortrijk, Belgium) for similarity (Dice; complete linkage). PCR amplification of different parts of cif and cif-flanking regions was performed as listed in Table 3 and described previously (42). The efa/lif gene was detected with primers Efa1 fwd, Efa1 rev, Efa1 3'fwd, and Efa1 3'rev according to the method of Morabito et al. (40). Oligonucleotides were obtained from TIB MolBiol, Berlin, Germany.
Restriction enzyme digestion of total genomic DNA, PFGE, and calculation of similarity indices. Preparation of total genomic DNA for PFGE was performed as previously described (8). The samples were digested with XbaI and analyzed by PFGE with the CHEF-DR II system (Bio-Rad Laboratories, Munich, Germany). Lambda concatemers (New England Biolabs, Frankfurt am Main, Germany) were used as molecular weight standards. Digested DNA was separated on 1% agarose gels with 200 V for 25 h at 14°C with pulse times increasing from 5 to 40 s. The gels were stained with ethidium bromide for 1 h and digitalized. Evaluation of PFGE profiles for similarity was performed with bionumerics software with Dice similarity indices (complete linkage; optimization, 1%; position tolerance, 1.3%) as described previously (9).
Amplification of target genes for MLST. For MLST analysis, fresh single colonies, which were grown overnight on Luria-Bertani agar plates, were suspended in 50 μl of a 0.9% NaCl solution using sterile toothpicks and subsequently diluted 1:6 in 0.9% NaCl solution. PCR samples were prepared in a total volume of 50 μl containing 5 μl of this bacterial suspension, 5 μl of 10-fold-concentrated polymerase reaction buffer containing 15 mM MgCl2 (Promega, Mannheim, Germany), 200 μM (each) deoxynucleoside triphosphate, 30 pmol of each primer, and 1.5 U of Taq DNA polymerase (Promega). PCR was performed in a GeneAmp PCR System 2700 (Applied Biosystems Applera, Weiterstadt, Germany). PCR conditions depended on the different target genes and primer sequences used and are depicted in Table 2. Eight microliters of the PCR products was analyzed for purity on 0.7 to 1% Tris-borate-EDTA agarose gels using a maximum voltage of 5 V/cm.
Purification of PCR products and DNA sequencing. For removal of PCR primers, five microliters of the PCR product was mixed with two microliters of an Exo-Sap reaction mixture containing equal numbers of units of exonuclease I (20 U/μl) (New England Biolabs) and shrimp alkaline phosphatase (USB Biochemicals) (1 U/μl). Exonuclease I was diluted to 1 U/μl in 50 mM Tris-HCl, pH 7.5, and incubated at 37°C for 30 min and subsequently for 15 min at 80°C to inhibit further enzymatic activity.
DNA sequencing was performed using the BigDye terminator version 1.1 cycle-sequencing RR-100 premix (Applied Biosystems Applera). Briefly, 2 μl of the Exo-SAP-treated PCR product was mixed with 1.5 μl of primer (30 pmol/μl), 2.0 μl of premix, and 4.5 μl of distilled water and subjected to PCR under the same conditions described above and as depicted in Table 2.
After completion of PCR, the sequencing products were purified with Autoseq G-50 columns (Amersham Biosciences Europe GmbH, Freiburg, Germany) as recommended by the manufacturer. Approximately 10 μl was eluted from the columns and dried in a MaxiDryLyo microconcentrator (Heto Holten, Dreieich, Germany). After that, 50 μl of formamide and 25 μl of sequencing buffer were mixed, and 2 μl of this mixture was added to the dried sequencing product. The mixture was incubated for 2 min at 90°C and subsequently cooled on ice.
Software for DNA analysis. Following double-strand sequencing, the sequences were edited and aligned. DNA fragments between 410 and 725 bp long from each sequence were used for sequence analysis (Table 2). Raw DNA sequence data were analyzed with ABI 377 software. DNA sequences were edited and aligned with BioEdit, version 4.8.10 (http://jwbrown.mbio.ncsu.edu/BioEdit/bioedit.html) (27), converted into FASTA files, and loaded into S.T.A.R.T. (Sequence Type Analysis and Recombinational Tests) (http://outbreak.ceid.ox.ac.uk/software.htm). The dendrograms were compiled with S.T.A.R.T., using the unweighted pair group method with arithmetic mean (UPGMA). This software constructs a phylogenetic tree on the basis of allele numbers. Split decomposition analysis (4, 28) was performed with Splitstree version 2.0 (http://bibiserv.techfak.uni-bielefeld.de/splits/).
Nucleotide sequence accession numbers. The nucleotide sequences obtained by sequencing of the PCR products of each allele of all genes have been entered into the GenBank-EMBL databases under continuous accession numbers from AJ629457 to AJ629510 for adk, AJ629511 to AJ629564 for arcA, AJ629565 to AJ629618 for fumC, AJ629619 to AJ629672 for icdA, AJ629673 to AJ629726 for mdh, AJ629727 to AJ629780 for mtlD, AJ629781 to AJ629834 for pgi, AJ629916 to AJ629965 for eae, and AJ629966 to AJ630015 for escD.
RESULTS
Flagellar types of E. coli O103 strains. The O types of the 54 E. coli O103 strains investigated in this study were confirmed by serotyping using strain H515b (O103:H8) as a positive control (41). Forty-eight of the O103 strains were motile; 47 of these expressed flagellar antigen H2, and 1 strain (92-1695) was O103:H11 (Table 1). Six strains were nonmotile (NM), and the H types of these were determined by analysis of HhaI-digested fliC PCR products as described above (Fig. 1). Five of the O103:NM strains showed the fliC genotype H2, which was also present in all motile O103:H2 strains (data not shown). The remaining O103:NM strain, CA92-7164, was genotyped as O103:[H7] (Fig. 1).
Multilocus sequence analysis. Nucleotide sequence analysis of unlinked housekeeping genes (multilocus sequencing typing) is widely used for evolutionary and population analysis, and also for epidemiological investigations (36, 62). Since differences in the sequences of essential housekeeping genes are thought to display long-term genetic changes, we used this technique to investigate the phylogenetic relationships of E. coli O103 strains and to assess the similarity of strains belonging to phylogenetic groups by PFGE and to compare their virulence genes. DNA sequences of corresponding housekeeping genes from the E. coli K-12 strain MG1655 (accession no. NC_000913), the O157:H7 strain EDL933 (accession no. NC_002655), and the uropathogenic E. coli O6 strain CFT073 (accession no. NC_004431) were compiled from the respective published genome sequences.
In addition to the housekeeping genes adk, arcA, fumC, icdA, mdh, mtlD, and pgi, we used the LEE-harbored virulence genes eae and escD (Table 2) for all strains that were positive with the universal eae primers SK1 and SK2. The genes chosen for MLST were distributed equally throughout the chromosome to minimize the influence of possible chromosomal hot spots of mutation and recombination.
Sequences were obtained from all 54 E. coli O103 strains for all housekeeping loci investigated and could be aligned without gaps. The corresponding DNA sequences of E. coli strains MG1655 (E. coli K-12), EDL933 (O157:H7), and CFT073 (O6:K:H1) were included in our scheme. Each allele of the seven genes analyzed was numbered randomly in ascending order (Table 4). Numbers and types of base substitutions were not taken into account for allele designation (63). The combinations of allele numbers for all isolates are shown in Table 5. Each unique combination of allele numbers represents one sequence type (profile). Ten MLST profiles, I to X, were established (Table 5), and seven MLST profiles (I to VII) were attributed to the E. coli O103 strains. Forty-three (79.6%) of these showed MLST profile I, five strains (8.8%) showed MLST profile II, and the remaining six strains fell into MLST profiles III to VII.
Figure 2 shows a dendrogram calculated with UPGMA using the S.T.A.R.T software. Most of the strains belonging to MLST profile I are STEC from humans or animals from different geographical origins (Table 1). MLST profile II covers all EPEC O103:H2 strains from animals and a human, which originated from France and Germany. MLST profile III is comprised of two STEC strains from Germany (CB7320) and from the United Kingdom (HBI/01). Profiles IV to VII are represented by single O103 strains (C124, P94-6011, CA92-7164, or 92-1695) whose virulence attributes or H types were different from those of all other E. coli O103 strains that were investigated (Fig. 2 and Table 1). Single profiles called VII, IX, and X were attributed to the non-O103 strains MG1655, EDL933, and CFT073, the last being most distantly related to the E. coli O103 strain group.
Sequences of housekeeping genes were analyzed with Splitstree (28) for evidence of recombination. The algorithm used in this software is able to display conflicting results in the phylogenetic descent of sequences. A treelike structure is created when the descent is clonal, but an interconnecting network with a number of splits or a bushlike structure will appear when recombination plays a role in the phylogenetic history of genes. The Splitstree graph calculated with MLST data using a distance matrix and all 57 alleles of seven loci is a combination of treelike and bushlike structures (Fig. 3). Since Splitstree creates a cladogram, the lengths of the branches depict the relative phylogenetic distances. On the basis of Splitstree and UPGMA tree analysis data, we conclude that STEC and EPEC O103 strains from profiles I, II, and III are closely related, whereas the four strains with profiles IV to VII are phylogenetically more distant. Interestingly, three of the latter strains did not carry an eae gene, and two of these strains (CA92-7164 [O103:H7; MLST profile VI] and 92-1695 [O103:H11; MLST profile VII]) were different in flagellar antigen type from all other E. coli O103:H2 strains. The group of E. coli O103:H2 strains was found to be more closely related to EDL933 (O157:H7) than to E. coli K-12 or the uropathogenic E. coli O6 strain. Only some of the seven housekeeping genes analyzed with Splitstree showed discriminatory differences between strains of the O103 group. The clear distinction between clusters of O103 strains (Fig. 4) is mainly achieved from analysis of the icdA gene. The splits seen with fumC, icdA, mdh, and pgi occur mainly between E. coli O103 and non-O103 strains.
We were interested to see if the inclusion of the chromosomally (LEE) encoded virulence-associated genes eae and escD would increase the resolution of MLST within the E. coli O103 strains, and this analysis was performed with 51 eae-positive E. coli O103 strains (Table 1). As a result, the profiles II, III, and VII are represented by the same strains as in the primary housekeeping analysis. MLST profile I is split into two related profiles, IA and IB (data not shown). Group IB comprises strains CB8198, 3386/98, 2905/96, and RW2198, and group IA comprises all other strains of MLST profile I.
PFGE. A dendrogram based on similarities of XbaI-digested DNA PFGE patterns among E. coli O103 strains was created with the Bionumerics software using Dice similarity indices as described above. PFGE patterns were arranged in five clusters, A to E, which show >55% similarity in their banding patterns (Fig. 5). One strain (P94-6011) was not typeable by PFGE because it repeatedly produced only a smear of degraded DNA after PFGE (Fig. 5).
The 43 MLST I strains were subdivided by PFGE into four clusters, A to D. PFGE group A (64% similarity) consisted of 22 STEC strains, which were from diarrheic humans (n = 20) and healthy cattle (n = 2) from Germany. All group A strains were stx1, eae-, E-hly, etpD, and efa positive. Twenty of the group A strains carried the plasmid-encoded katP gene, and 21 strains were porcine adherence factor gene (paa) positive.
PFGE cluster B (59% similarity) is formed by seven MLST I strains (STEC types from cattle, humans, and a goat) and by the MLST II group strain CB8432, which belongs to the EPEC group. These strains originated in Australia (n = 1), the United States (1), France (1), and Germany (5). From its virulence profile, this group was similar to PFGE cluster A STEC strains, except for the EPEC strain CB8432, which showed close similarity to the animal EPEC MLST II profile strains in a different PFGE cluster, E (see below and Table 1).
Eleven STEC strains were grouped into PFGE cluster C (60% similarity), which could be further divided in two subclusters, C1 and C2. Cluster C1 (70% similarity) is formed by five O103:H2 strains belonging to MLST groups I and III, which originated from Ireland (meat), Germany (human), and Canada (two humans and a pig). All strains were positive for eae-, for the E-hly gene, and for the aerobactin receptor gene iutA; the last was present in only nine (16.6%) of all E. coli O103 isolates from this study (Table 6). Four of the strains produced Stx1.
All STEC O103 strains that were positive for stx2 were gathered in PFGE cluster C2 (80% similarity). The group of six stx2-positive STEC strains was closely associated with HUS cases, and the strains were isolated from patients in different parts of Germany between 1997 and 2000. All strains belonged to MLST profile I.
PFGE cluster D (55% similarity) is formed by seven strains belonging to MLST profiles I, IV, and VII (Fig. 5). The strains originated from humans (n = 5), a rabbit, and cattle. Cluster D strains were heterogeneous for all virulence markers investigated in this study (Table 1).
Cluster E (70% similarity) is formed by three EPEC O103:H2 strains, which were isolated from chickens and from a rabbit in France. All cluster E strains belong to the MLST II group, together with two other EPEC strains from a human (CB8432; cluster B) and a chicken (891; a single PFGE pattern).
Two other E. coli O103 strains from animals could not be assigned to any of the PFGE clusters defined here. These were P94-6011 (MLST profile V) and CA92-7164 (MLST profile VI). These strains were negative for Shiga toxins and for intimin.
Virulence gene typing of E. coli O103. A detailed genetic analysis of virulence markers known to be associated with EPEC, EHEC, and extraintestinal pathogenic E. coli was performed with all 54 E. coli O103 strains (Tables 1 and 6). Fifty-one (94.4%) of the strains were positive for an intimin gene (eae), and 44 (81.5%) of the strains produced Stx1 (stx1) or Stx2 (stx2). Three O103 strains from animals belonging to MLST profiles IV to VI (C124, P94-6011, and CA92-7164) (Table 1 and Fig. 2) were negative for both Stx and intimin. The intimin epsilon gene (eae-) was associated with all O103 MLST I (n = 43) and MLST III (n = 2) strains and was not found in strains belonging to other MLST groups. The intimin epsilon-positive strains were also positive for Shiga toxins, except two Stx-negative strains, which came from a pig (F22S5-2) and from a human (CB3349).
The eae- gene was present in all MLST type II strains, encompassing five EPEC O103:H2 strains, which were from chickens, a rabbit, and a human. In addition, an eae- gene was found in a single human STEC O103:H11 strain (92-1695; MLST VII), whose virulence markers, MLST type, and H type were different from those of all other O103 strains (Table 1 and Fig. 2).
None of the 54 O103 strains was positive for the bfpA (bundle-forming pilus) gene, which is associated with typical EPEC, or for the lpfA (long polar fimbria) gene, which is present in STEC O157 and in some LEE-negative, non-O157 STEC strains (22, 58) (Table 6).
In contrast, 51 (94.4%) of the O103 strains were positive for the EHEC factor for adherence (efa) gene, which is widely distributed in EHEC and EPEC strains (3). Efa/LifA were absent in only two eae- and stx-negative isolates from animals (C124 and CA92-7164) and in a human EPEC O103:H2 strain (CB8432) (Table 1). Another adherence-associated gene, paa (3), was found in 44 O103 strains, including EPEC and STEC types (Table 1).
The analysis of virulence genes in STEC and EPEC O103 strains revealed patterns which corresponded to different pathotypes. The presence of the stx2 gene in STEC O103 strains was closely associated with HUS cases, whereas most of the stx1-positive O103 strains from humans were associated with diarrhea. EHEC virulence plasmid genes encoding EHEC hemolysin (E-hlyA), type II secretion pathway protein D (etpD), and catalase-peroxidase (katP) were frequently associated with STEC O103 strains but were absent in EPEC O103 from animals and in eae-negative, stx-negative O103 strains. The plasmid-encoded serine protease gene (espP) was present in only two stx1-positive O103 strains from humans. The STEC O103 group could be subdivided in regard to the stx genes and to the presence of the plasmid-borne katP gene: E-hlyA, etpD, and katP genes were present in 22 stx1-positive strains from humans and cattle; the katP gene was not found in a second group of 11 stx1-positive strains from humans, cattle, and a goat; and a third group consisted of six stx2-positive human O103 strains from patients with HUS (Table 1). Only one STEC O103 strain from cattle (RW2198) was negative for all EHEC virulence plasmid-associated genes.
Genes encoding the aerobactin receptor (iutA), cytolethal distending toxin 1 (cdt-1), and cytotoxic necrotizing factor 1 (cnf) were associated with E. coli causing extraintestinal infections (21, 29, 37). None of the O103 strains from our study was positive for cnf and cdt-I genes, but nine O103 strains (STEC, EPEC, and others) carried the aerobactin receptor gene iutA. Interestingly, the iutA gene was found to dominate in O103 strains from Canada and the United States (Table 1).
Presence and characterization of the genetic background of the Cif gene. Cif (cell cycle-inhibiting factor) was recently described as a translocated effector molecule that blocks cell cycle G2/M transition (38). Interestingly, 51 (94.4%) of the 54 E. coli O103 strains investigated in our study were positive with primers cif-int-s and cif-int-as, which are complementary to the middle portion of the recently described cif gene (38). The three cif-negative strains C124, P94-6011, and CA92-7164, were also negative for eae, stx, and most other virulence markers (Table 1). Based on this high prevalence of cif, we were interested in the presence of a complete cif gene in these isolates. PCR with primers cif-rbs-s and cif-as, covering the entire cif gene, revealed a 950-bp PCR product with 45 O103 strains, indicating that a full-length cif gene was present (Fig. 6). However, in six strains, the PCR product was 2.2 kb in length, giving evidence for additional DNA in the cif region (Fig. 6, lanes 5 and 9). A third PCR with primers phage-s and ybhB, which are complementary to a bacteriophage gene in the neighborhood of the cif gene and the chromosomal ybhB gene close to the prophage-chromosome border, confirmed this finding (38). A 5-kb PCR product was found in the same 45 O103 strains, which showed a 950-bp product with primers cif-rbs-s and cif-as, and a 6.2-kb product was found in the same 6 strains, which showed a 2.2-kb fragment with the cif internal primers cif-rbs-s and cif-as (Fig. 6). These findings demonstrate that the cif genes found in the O103 strains are located in the genome of prophages and that these prophages are located in the same orientation in the same integration site.
Since there was evidence for additional DNA in the cif genes of six strains, we sequenced one of the 2.2-kb PCR products obtained with primers cif-rbs-s and cif-as and found a sequence that was identical to the insertion element IS2, which is inserted from positions 228 to1554 in the cif sequence in all six strains (data not shown). All strains carrying that IS2 element in the cif gene were stx2-positive STEC O103:H2 strains that were isolated from patients with HUS in Germany (Table 1).
DISCUSSION
Molecular epidemiology is largely descriptive and characterizes bacteria based on their natural genetic variation. A number of molecular methods have been employed to determine this variation, which may be due to mutation or horizontal gene transfer events. Basically, two branches of molecular epidemiology can be distinguished, classification and typing. Classification can illustrate evolutionary relationships and subdivides species into clonal groupings, whereas typing is used more often for differentiation of clinical or environmental isolates. Long-term epidemiology may require classification, whereas short-term local epidemiology is more often performed by typing methods. While the most prominent methods for molecular classification are MLST and multilocus enzyme electrophoresis (36, 18), PFGE or PCR-based methods, such as randomly amplified polymorphic DNA-PCR, AFLP, or virulence gene characterization, have been used frequently for typing approaches (for a review, see reference 1).
In our study, we combined MLST and PFGE with PCR detection of virulence genes for classification of pathotypes and for phylogenetic analysis of E. coli O103 strains, which constitute a group of human and animal pathogenic E. coli. In order to include the prominent pathotypes occurring among strains of the E. coli O103 group, we investigated strains from different places and different sources, which were isolated over a 7-year period. The strains were not known to be epidemiologically linked and were not associated with outbreaks. Analysis of virulence genes of E. coli O103 strains resulted in the detection of two major pathotypes, EPEC and STEC, which are present in both humans and animals.
MLST with housekeeping genes of E. coli O103 strains has demonstrated seven profiles, and profiles I and III encompass a group of phylogenetically closely related O103:H2 strains. Profiles I and III occurred most frequently, being present in 83.3% of the strains which were isolated in Australia, Canada, France, Germany, Ireland, and the United States. The leading markers of profile I and III strains are production of Stx and intimin and the presence of EHEC plasmid genes, characterizing them as typical STEC. Only two intimin -positive strains were found to be negative for Stx, but they were grouped with other STEC strains by PFGE and by MLST. It is therefore possible that these strains have lost the phage-borne stx genes upon subculture, as reported for other STEC (32, 53).
MLST profile II was associated with five EPEC O103:H2 strains from animals and a human. Data from phylogenetic analysis suggest that profile I and III STEC O103:H2 strains have evolved from MLST profile II strains. Major differences between profile II and profile I and III strains were found in only one of the seven housekeeping genes (icdA) that were analyzed and in the eae and escD sequences, whereas the eae and escD MLST patterns were not identical (data not shown). Analysis of chromosomally inherited virulence genes, such as eae, was found to be helpful for establishing phylogenetic groups among closely related EPEC and STEC O103:H2 strains.
MLST group IV to VII O103 strains were found to be phylogenetically far distant from MLST group I, II, and III strains, as were the non-O103 strains EDL933, MG1655, and CFT073. This finding indicates that strains belonging to the same O serogroup may have evolved separately from each other and that the O serogroup as such cannot be taken as an indicator of genetic relationships among E. coli strains.
By PFGE, MLST profile I and III strains were divided into four clusters, which may have been generated by intra- and intergenomic recombination events. A few MLST profile I strains showed very similar PFGE patterns, which could indicate that they are epidemiologically related (57). PFGE typing was found to be very useful for the identification of genetically closely related groups of highly pathogenic Stx2-producing STEC O103:H2 (C1 cluster; 80% similarity) among strains belonging to the same MLST profile I. This finding shows that PFGE is highly useful to detect groups of strains which have probably evolved very recently and which are therefore not detectable by MLST, which indicates differences that have evolved more slowly by mutation and recombination. On the other hand, PFGE typing alone is not sufficient to define groups of strains which are phylogenetically closely related, as was the case with MLST II group EPEC strains, which were separated into three different unrelated PFGE profiles. The combination of MLST, PFGE, and virulence gene typing allowed us to define EPEC and STEC O103 strains as two separate groups of pathogens that not only differ from each other in their virulence genes but represent two clonal lineages with different MLST profiles. Similarly, by MLST, Stx- and eae-negative E. coli O103 strains were identified as evolutionarily distant form STEC and EPEC O103, excluding the possibility that they were EPEC or STEC derivative strains that have simply lost their virulence markers.
Earlier studies showed that the gene compositions of the large virulence plasmids vary among STEC strains, but virulence gene patterns appeared to be serotype associated (14, 49). This is reflected in our experiments showing a close association of E-hlyA (93.1%) and etp (90.1%) genes with STEC O103 strains. In contrast, the katP gene was found in only 21 (47.7%) STEC O103 strains but is closely associated (90.1%) with STEC strains belonging to PFGE group A. This finding indicates that most of the katP-positive STEC strains belong to a genetically closely related subgroup among STEC O103 strains. The presence of a plasmid-borne espP gene in only two strains indicates that the ancestral STEC O103 strains lacked espP, which might have been acquired more recently in the evolution of STEC O103.
E. coli O103 EPEC strains were isolated from rabbits, chickens, and a human. These strains were classified as atypical EPEC, because none was positive for the bfpA gene, which is located on the EAF plasmid in typical EPEC strains (61). EPEC O103 strains were different from other intimin-positive, Stx-negative O103 strains (see below) in their intimin type (1) and their MLST profile.
Interestingly, the novel adhesion factors encoded by paa and efa and the cytotoxin encoded by the cif gene were very frequent in STEC and EPEC O103 strains and are probably important for their virulence (3, 6, 38). However, we cannot conclude that all cif genes were functional in our strains, since it had been shown earlier that cif in strain PMK5 contains an internal stop codon (42).The efa gene has been shown to be associated with the LEE in several E. coli serotypes (40).
The virulence genes of the group of six Stx2-producing STEC O103 strains (PFGE cluster C2; 80% similarity) were very similar. Interestingly, these strains differed from all 51 other cif-positive O103 strains by having an IS2 insertion located in their cif genes. None of these strains was related to an outbreak. Together with the high similarity found by PFGE, the IS2 insertion in cif indicates that these strains represent an evolutionarily young clone that might have evolved around 1996 and spread throughout Germany. The presence of the stx2 gene could be important for the capacity of these strains to cause HUS, since it was previously reported that Stx2, in contrast to Stx1 and Stx2c, is related to severe disease and HUS in E. coli O157 and other STEC strains (9, 13, 25). The detection of stx2-positive STEC O103 strains indicates that highly virulent strains may evolve from already existing human pathogenic STEC O103 strains, probably by horizontal transfer of phage-encoded stx2 genes. Further research is necessary to identify the evolutionary factors involved in the generation of highly virulent STEC O103 clones.
Only one (UTI) of the O103 strains from our study was associated with extraintestinal infections of humans. Among certain virulence markers that were associated with extraintestinal pathogenic E. coli, only the aerobactin receptor gene (iutA) was found to be present in 9 of the 54 O103 strains. Interestingly, the iutA gene was present in 5 of 8 strains isolated in North America but in only 4 of 46 strains from other places. The iutA gene is known to be located on plasmids in E. coli (20), which could contribute to the spread of this marker to strains that were unrelated in their sources and PFGE and MLST types.
E. coli strains of serogroup O103 are of particular interest, since they are widespread as commensals and pathogens in animals and humans. Although the clonal diversity and some pathogenic properties of E. coli O103 strains have been published, the present study extends our knowledge about the virulence spectrum and the genetic relationships of E. coli O103 strains from different sources and geographical origins. Phylogenetic analysis has demonstrated that EPEC and STEC strains of serotype O103:H2 are genetically closely related despite their different virulence spectra. This points to a recent emergence of pathogenic E. coli O103 strains and indicates that this serotype might possess a very successful combination of traits. STEC and EPEC O103 represent two different lineages, and the risk of human disease may be increased by infection with STEC O103 from the bovine and/or goat reservoir. This should be elucidated in more detail in future research approaches.
ACKNOWLEDGMENTS
We are grateful to John Fairbrother, Montreal, Canada; Hywel Ball, Belfast, Northern Ireland; Patricia Marjiani-Kurkdjian, Paris, France; Karl A. Bettelheim, Melbourne, Australia; and Phil Tarr, St. Louis, Mo., for donating E. coli O103 strains from their laboratory collections. We also thank Kerstin Rydzewski and Manuela Brandt for skillful and highly motivated technical assistance.
This work was supported by grants from the European Union (EU project QLK2-2000-0060).
The authors are solely responsible for the work described in this paper, and their opinions are not necessarily those of the EU.
REFERENCES
Achtman, M. 2001. A phylogenetic perspective on molecular epidemiology, p. 485-509. In M. Sussmann (ed.), Molecular medical microbiology. Academic Press, San Diego, Calif.
An, H., J. M. Fairbrother, C. Desautels, and J. Harel. 1999. Distribution of a novel locus called Paa (porcine attaching and effacing associated) among enteric Escherichia coli. Adv. Exp. Med. Biol. 473:179-184.
Badea, L., S. Doughty, L. Nicholls, J. Sloan, R. M. Robins-Browne, and E. L. Hartland. 2003. Contribution of Efa1/LifA to the adherence of enteropathogenic Escherichia coli to epithelial cells. Microb. Pathog. 34:205-215.
Bandelt, H. J., and A. W. Dress. 1992. Split decomposition: a new and useful approach to phylogenetic analysis of distance data. Mol. Phylogenet. Evol. 1:242-252.
Bastian, S. N., I. Carle, and F. Grimont. 1998. Comparison of 14 PCR systems for the detection and subtyping of stx genes in Shiga-toxin-producing Escherichia coli. Res. Microbiol. 149:457-472.
Batisson, I., M. P. Guimond, F. Girard, H. An, C. Zhu, E. Oswald, J. M. Fairbrother, M. Jacques, and J. Harel. 2003. Characterization of the novel factor Paa involved in the early steps of the adhesion mechanism of attaching and effacing Escherichia coli. Infect. Immun. 71:4516-4525.
Bettelheim, K. A., M. A. Hornitzky, and S. P. Djordjevic. 2001. First bovine and ovine isolations of Shiga toxin-producing Escherichia coli O103:H2 in Australia. Aust. Vet. J. 79:289-290.
Beutin, L., D. Geier, S. Zimmermann, S. Aleksic, H. A. Gillespie, and T. S. Whittam. 1997. Epidemiological relatedness and clonal types of natural populations of Escherichia coli strains producing Shiga toxins in separate populations of cattle and sheep. Appl. Environ. Microbiol. 63:2175-2180.
Beutin, L., S. Kaulfuss, T. Cheasty, B. Brandenburg, S. Zimmermann, K. Gleier, G. A. Willshaw, and H. R. Smith. 2002. Characteristics and association with disease of two major subclones of Shiga toxin (Verocytotoxin)-producing strains of Escherichia coli (STEC) O157 that are present among isolates from patients in Germany. Diagn. Microbiol. Infect. Dis. 44:337-346.
Beutin, L., S. Zimmermann, and K. Gleier. 1996. Rapid detection and isolation of Shiga-like toxin (verocytotoxin)-producing Escherichia coli by direct testing of individual enterohemolytic colonies from washed sheep blood agar plates in the VTEC-RPLA assay. J. Clin. Microbiol. 34:2812-2814.
Beutin, L., S. Zimmermann, and K. Gleier. 1998. Human infections with Shiga toxin-producing Escherichia coli other than serogroup O157 in Germany. Emerg. Infect. Dis. 4:635-639.
Blanco, J., E. A. Gonzalez, S. Garcia, M. Blanco, B. Regueiro, and I. Bernardez. 1988. Production of toxins by Escherichia coli strains isolated from calves with diarrhoea in Galicia (north-western Spain). Vet. Microbiol. 18:297-311.
Boerlin, P., S. A. McEwen, F. Boerlin-Petzold, J. B. Wilson, R. P. Johnson, and C. L. Gyles. 1999. Associations between virulence factors of Shiga toxin-producing Escherichia coli and disease in humans. J. Clin. Microbiol. 37:497-503.
Brunder, W., H. Schmidt, M. Frosch, and H. Karch. 1999. The large plasmids of Shiga-toxin-producing Escherichia coli (STEC) are highly variable genetic elements. Microbiology 145:1005-1014.
Brunder, W., H. Schmidt, and H. Karch. 1996. KatP, a novel catalase-peroxidase encoded by the large plasmid of enterohaemorrhagic Escherichia coli O157:H7. Microbiology 142:3305-3315.
Brunder, W., H. Schmidt, and H. Karch. 1997. EspP, a novel extracellular serine protease of enterohaemorrhagic Escherichia coli O157:H7 cleaves human coagulation factor V. Mol. Microbiol. 24:767-778.
Caprioli, A., and A. E. Tozzi. 1996. Epidemiology of Shiga toxin-producing Escherichia coli infections in continental Europe, p. 38-48. In J. B. Kaper and A. D. O'Brien (ed.), Escherichia coli O157:H7 and other Shiga toxin-producing E. coli strains. ASM Press, Washington, D.C.
Caugant, D. A., K. Bovre, P. Gaustad, K. Bryn, E. Holten, E. A. Hoiby, and L. O. Froholm. 1986. Multilocus genotypes determined by enzyme electrophoresis of Neisseria meningitidis isolated from patients with systemic disease and from healthy carriers. J. Gen. Microbiol. 132:641-652.
Cid, D., M. Blanco, J. E. Blanco, J. A. Ruiz Santa Quiteira, R. de la Fuente, and J. Blanco. 1996. Serogroups, toxins and antibiotic resistance of Escherichia coli strains isolated form diarrhoeic goat kids in Spain. Vet. Microbiol. 53:349-354.
De Lorenzo, V., A. Bindereif, B. H. Paw, and J. B. Neilands. 1986. Aerobactin biosynthesis and transport genes of plasmid ColV-K30 in Escherichia coli K-12. J. Bacteriol. 165:570-578.
De Rycke, J., A. Milon, and E. Oswald. 1999. Necrotoxic Escherichia coli (NTEC): two emerging categories of human and animal pathogens. Vet. Res. 30:221-233.
Doughty, S., J. Sloan, V. Bennett-Wood, M. Robertson, R. M. Robins-Browne, and E. L. Hartland. 2002. Identification of a novel fimbrial gene cluster related to long polar fimbriae in locus of enterocyte effacement-negative strains of enterohemorrhagic Escherichia coli. Infect. Immun. 70:6761-6769.
Eklund, M., F. Scheutz, and A. Siitonen. 2001. Clinical isolates of non-O157 Shiga toxin-producing Escherichia coli: serotypes, virulence characteristics, and molecular profiles of strains of the same serotype. J. Clin. Microbiol. 39:2829-2834.
Elliott, S. J., L. A. Wainwright, T. K. McDaniel, K. G. Jarvis, Y. K. Deng, L. C. Lai, B. P. McNamara, M. S. Donnenberg, and J. B. Kaper. 1998. The complete sequence of the locus of enterocyte effacement (LEE) from enteropathogenic Escherichia coli E2348/69. Mol. Microbiol. 28:1-4.
Friedrich, A. W., M. Bielaszewska, W. L. Zhang, M. Pulz, T. Kuczius, A. Ammon, and H. Karch. 2002. Escherichia coli harboring Shiga toxin 2 gene variants: frequency and association with clinical symptoms. J. Infect. Dis. 185:74-84.
Gunzburg, S. T., N. G. Tornieporth, and L. W. Riley. 1995. Identification of enteropathogenic Escherichia coli by PCR-based detection of the bundle-forming pilus gene. J. Clin. Microbiol. 33:1375-1377.
Hall, T. A. 2004. BioEdit: a user-friendly biological sequence alignment editor and analysis program for Windows 95/98/NT. Nucleic Acids Symp. Ser. 41:95-98.
Huson, D. H. 1998. SplitsTree: analyzing and visualizing evolutionary data. Bioinformatics 14:68-73.
Johnson, J. R., T. T. O'Bryan, P. Delavari, M. Kuskowski, A. Stapleton, U. Carlino, and T. A. Russo. 2001. Clonal relationships and extended virulence genotypes among Escherichia coli isolates from women with a first or recurrent episode of cystitis. J. Infect. Dis. 183:1508-1517.
Johnson, J. R., A. E. Stapleton, T. A. Russo, F. Scheutz, J. J. Brown, and J. N. Maslow. 1997. Characteristics and prevalence within serogroup O4 of a J96-like clonal group of uropathogenic Escherichia coli O4:H5 containing the class I and class III alleles of papG. Infect. Immun. 65:2153-2159.
Jores, J., L. Rumer, S. Kiessling, J. B. Kaper, and L. H. Wieler. 2001. A novel locus of enterocyte effacement (LEE) pathogenicity island inserted at pheV in bovine Shiga toxin-producing Escherichia coli strain O103:H2. FEMS Microbiol. Lett. 204:75-79.
Karch, H., T. Meyer, H. Rüssmann, and J. Heesemann. 1992. Frequent loss of Shiga-like toxin genes in clinical isolates of Escherichia coli upon subcultivation. Infect. Immun. 60:3464-3467.
Lin, Z., H. Kurazono, S. Yamasaki, and Y. Takeda. 1993. Detection of various variant verotoxin genes in Escherichia coli by polymerase chain reaction. Microbiol. Immunol. 37:543-548.
Luzzi, I., A. E. Tozzi, G. Rizzoni, A. Niccolini, I. Benedetti, F. Minelli, and A. Caprioli. 1995. Detection of serum antibodies to the lipopolysaccharide of Escherichia coli O103 in patients with hemolytic-uremic syndrome. J. Infect. Dis. 171:514-515.
Machado, J., F. Grimont, and P. A. Grimont. 2000. Identification of Escherichia coli flagellar types by restriction of the amplified fliC gene. Res. Microbiol. 151:535-546.
Maiden, M. C., J. A. Bygraves, E. Feil, G. Morelli, J. E. Russell, R. Urwin, Q. Zhang, J. Zhou, K. Zurth, D. A. Caugant, I. M. Feavers, M. Achtman, and B. G. Spratt. 1998. Multilocus sequence typing: a portable approach to the identification of clones within populations of pathogenic microorganisms. Proc. Natl. Acad. Sci. USA 95:3140-3145.
Mainil, J. G., E. Jacquemin, and E. Oswald. 2003. Prevalence and identity of cdt-related sequences in necrotoxigenic Escherichia coli. Vet. Microbiol. 94:159-165.
Marches, O., T. N. Ledger, M. Boury, M. Ohara, X. Tu, F. Goffaux, J. Mainil, I. Rosenshine, M. Sugai, J. De Rycke, and E. Oswald. 2003. Enteropathogenic and enterohaemorrhagic Escherichia coli deliver a novel effector called Cif, which blocks cell cycle G2/M transition. Mol. Microbiol. 50:1553-1567.
Mariani-Kurkdjian, P., E. Denamur, A. Milon, B. Picard, H. Cave, N. Lambert-Zechovsky, C. Loirat, P. Goullet, P. J. Sansonetti, and J. Elion. 1993. Identification of a clone of Escherichia coli O103:H2 as a potential agent of hemolytic-uremic syndrome in France. J. Clin. Microbiol. 31:296-301.
Morabito, S., R. Tozzoli, E. Oswald, and A. Caprioli. 2003. A mosaic pathogenicity island made up of the locus of enterocyte effacement and a pathogenicity island of Escherichia coli O157:H7 is frequently present in attaching and effacing E. coli. Infect. Immun. 71:3343-3348.
Orskov, F., and I. Orskov. 1984. Serotyping of Escherichia coli. Methods Microbiol. 14:43-112.
Oswald, E., H. Schmidt, S. Morabito, H. Karch, O. Marches, and A. Caprioli. 2000. Typing of intimin genes in human and animal enterohemorrhagic and enteropathogenic Escherichia coli: characterization of a new intimin variant. Infect. Immun. 68:64-71.
Parma, A. E., M. E. Sanz, J. E. Blanco, J. Blanco, M. R. Vinas, M. Blanco, N. L. Padola, and A. I. Etcheverria. 2000. Virulence genotypes and serotypes of verotoxigenic Escherichia coli isolated from cattle and foods in Argentina. Importance in public health. Eur. J. Epidemiol. 16:757-762.
Peeters, J. E., R. Geeroms, and F. Orskov. 1988. Biotype, serotype, and pathogenicity of attaching and effacing enteropathogenic Escherichia coli strains isolated from diarrheic commercial rabbits. Infect. Immun. 56:1442-1448.
Prager, R., A. Liesegang, W. Voigt, W. Rabsch, A. Fruth, and H. Tschpe. 2002. Clonal diversity of Shiga toxin-producing Escherichia coli O103:H2/H– in Germany. Infect. Genet. Evol. 1:265-275.
Robert Koch Institut. 2004. Bakterielle Gastroenteritiden in Deutschland 2001. Epidemiol. Bull. 50:417-422.
Sambrook, J., E. F. Fritsch, and T. Maniatis. 1989. Molecular cloning: a laboratory manual, 2nd ed. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, N.Y.
Schmidt, H., L. Beutin, and H. Karch. 1995. Molecular analysis of the plasmid-encoded hemolysin of Escherichia coli O157:H7 strain EDL 933. Infect. Immun. 63:1055-1061.
Schmidt, H., C. Geitz, P. I. Tarr, M. Frosch, and H. Karch. 1999. Non-O157:H7 pathogenic Shiga toxin-producing Escherichia coli: phenotypic and genetic profiling of virulence traits and evidence for clonality. J. Infect. Dis. 179:115-123.
Schmidt, H., B. Henkel, and H. Karch. 1997. A gene cluster closely related to type II secretion pathway operons of gram-negative bacteria is located on the large plasmid of enterohemorrhagic Escherichia coli O157 strains. FEMS Microbiol. Lett. 148:265-272.
Schmidt, H., H. Rüssmann, and H. Karch. 1993. Virulence determinants in nontoxinogenic Escherichia coli O157 strains that cause infantile diarrhea. Infect. Immun. 61:4894-4898.
Schmidt, H., H. Rüssmann, A. Schwarzkopf, S. Aleksic, J. Heesemann, and H. Karch. 1994. Prevalence of attaching and effacing Escherichia coli in stool samples from patients and controls. Zentbl. Bakteriol. 281:201-213.
Schmidt, H., J. Scheef, H. I. Huppertz, M. Frosch, and H. Karch. 1999. Escherichia coli O157:H7 and O157:H– strains that do not produce Shiga toxin: phenotypic and genetic characterization of isolates associated with diarrhea and hemolytic-uremic syndrome. J. Clin. Microbiol. 37:3491-3496.
Sueyoshi, M., H. Fukui, S. Tanaka, M. Nakazawa, and K. Ito. 1996. A new adherent form of an attaching and effacing Escherichia coli (eaeA+, bfp–) to the intestinal epithelial cells of chicks. J. Vet. Med. Sci. 58:1145-1147.
Tarr, P. I., L. S. Fouser, A. E. Stapleton, R. A. Wilson, H. H. Kim, J. C. Vary, Jr., and C. R. Clausen. 1996. Hemolytic-uremic syndrome in a six-year-old girl after a urinary tract infection with Shiga-toxin-producing Escherichia coli O103:H2. N. Engl. J. Med. 335:635-638.
Tauschek, M., R. A. Strugnell, and R. M. Robins-Browne. 2002. Characterization and evidence of mobilization of the LEE pathogenicity island of rabbit-specific strains of enteropathogenic Escherichia coli. Mol. Microbiol. 44:1533-1550.
Tenover, F. C., R. D. Arbeit, R. V. Goering, P. A. Mickelsen, B. E. Murray, D. H. Persing, and B. Swaminathan. 1995. Interpreting chromosomal DNA restriction patterns produced by pulsed-field gel electrophoresis: criteria for bacterial strain typing. J. Clin. Microbiol. 33:2233-2239.
Torres, A. G., J. A. Giron, N. T. Perna, V. Burland, F. R. Blattner, F. Avelino-Flores, and J. B. Kaper. 2002. Identification and characterization of lpfABCC'DE, a fimbrial operon of enterohemorrhagic Escherichia coli O157:H7. Infect. Immun. 70:5416-5427.
Toth, I., F. Herault, L. Beutin, and E. Oswald. 2003. Production of cytolethal distending toxins by pathogenic Escherichia coli strains isolated from human and animal sources: establishment of the existence of a new cdt variant (type IV). J. Clin. Microbiol. 41:4285-4291.
Tozzi, A. E., A. Caprioli, F. Minelli, A. Gianviti, L. De Petris, A. Edefonti, G. Montini, A. Ferretti, T. De Palo, M. Gaido, and G. Rizzoni. 2003. Shiga toxin-producing Escherichia coli infections associated with hemolytic uremic syndrome, Italy, 1988-2000. Emerg. Infect. Dis. 9:106-108.
Trabulsi, L. R., R. Keller, and T. A. Tardelli Gomes. 2002. Typical and atypical enteropathogenic Escherichia coli. Emerg. Infect. Dis. 8:508-513.
Urwin, R., and M. C. Maiden. 2003. Multi-locus sequence typing: a tool for global epidemiology. Trends Microbiol. 11:479-487.
Wagner, P. L., D. W. Acheson, and M. K. Waldor. 2001. Human neutrophils and their products induce Shiga toxin production by enterohemorrhagic Escherichia coli. Infect. Immun. 69:1934-1937.
Zhang, W. L., B. Khler, E. Oswald, L. Beutin, H. Karch, S. Morabito, A. Caprioli, S. Suerbaum, and H. Schmidt. 2002. Genetic diversity of intimin genes of attaching and effacing Escherichia coli strains. J. Clin. Microbiol. 40:4486-4492.(Lothar Beutin, Stefan Kau)