Molecular Analysis of Shiga Toxin-Producing Escherichia coli Strains Isolated from Hemolytic-Uremic Syndrome Patients and Dairy Samples in France

Abstract

Shiga toxin-producing Escherichia coli (STEC) has been associated with food-borne diseases ranging from uncomplicated diarrhea to hemolytic-uremic syndrome (HUS). While most outbreaks are associated with E. coli O157:H7, about half of the sporadic cases may be due to non-O157:H7 serotypes. To assess the pathogenicity of STEC isolated from dairy foods in France, 40 strains isolated from 1,130 raw-milk and cheese samples were compared with 15 STEC strains isolated from patients suffering from severe disease. The presence of genes encoding Shiga toxins (stx₁, stx₂, and variants), intimin (eae and variants), adhesins (bfp, efa1), enterohemolysin (ehxA), serine protease (espP), and catalase-peroxidase (katP) was determined by PCR and/or hybridization. Plasmid profiling, ribotyping, and pulsed-field gel electrophoresis (PFGE) were used to further compare the strains at the molecular level. A new stx₂ variant, stx_2-CH013, associated with an O91:H10 clinical isolate was identified. The presence of the stx₂, eae, and katP genes, together with a combination of several stx₂ variants, was clearly associated with human-pathogenic strains. In contrast, dairy food STEC strains were characterized by a predominance of stx₁, with a minority of isolates harboring eae, espP, and/or katP. These associations may help to differentiate less virulent STEC strains from those more likely to cause disease in humans. Only one dairy O5 isolate had a virulence gene panel identical to that of an HUS-associated strain. However, the ribotype and PFGE profiles were not identical. In conclusion, most STEC strains isolated from dairy products in France showed characteristics different from those of strains isolated from patients.

Shiga toxin-producing Escherichia coli (STEC) causes human diseases, ranging from uncomplicated diarrhea to hemorrhagic colitis (HC) and life-threatening complications such as the hemolytic-uremic syndrome (HUS). STEC food-borne infections appear worldwide, with bovine feces being the main source of food contamination. While most outbreaks are associated with E. coli O157:H7, about half of the sporadic cases may be due to non-O157:H7 serotypes. Several studies have shown a high prevalence of STEC, belonging to a wide range of serotypes, in animals and food products (5, 34, 37). However, only a limited number of serotypes have been associated with human disease. In France, the first identified nationwide outbreak, consisting of 69 cases related to O157:H7 in ground meat, occurred in 2005 (A. Mailles, P. Mariani-Kurkdjian, C. Vernozy-Rozand, F. Grimont, N. Pihier, E. Bingen, B. Horen, V. Doireau, B. Llanas, E. Espié, and V. Vaillant, poster presented at the 6th International Symposium on Shiga Toxin (Verocytotoxin)-Producing Escherichia coli Infections, Melbourne, Australia, 2006, http://www.invs.sante.fr/surveillance/shu/poster_melbourne_o157.pdf).

The major characteristic of STEC linked to virulence is the production of one or more Shiga toxins (Stx1, Stx2, and variants of Stx2) (31). Epidemiological studies, together with in vivo and in vitro experiments, have shown that Stx2 is the most common virulence factor associated with severe human disease. Three stx₁ (stx_1-933J, stx_1c, stx_1d) and at least 12 stx₂ gene variants have been described, including stx_2-EDL933, stx_2c (stx_2vh-a, stx_2vh-b), stx_2d (stx_2d-Ount, stx_2d-OX3a), stx_2e, stx_2f, and stx_2-NV206 (3, 10, 24, 33, 44, 45). Some of the variants, which are associated with STEC strains isolated from specific hosts such as sheep (stx_2d) and pig (stx_2e), are probably less pathogenic for humans (18, 38). The type of variant could thus reflect both the origin and the clonality of STEC as well as its pathogenicity (18, 31).

Several other determinants have been involved in STEC virulence. The locus of enterocyte effacement (LEE) contains genes encoding proteins responsible for the “attaching and effacing” lesions in epithelial cells. One of these proteins, intimin, an outer membrane protein encoded by eae, is involved in the binding of bacteria to target cells. Several intimin types that may indicate the phylogenetic origin of STEC and determine the host tropism have been identified (1, 29). The LEE constitutes a pathogenicity island, a region of the bacterial chromosome transmitted by horizontal transfer, usually inserted adjacent to either the selC or the pheU tRNA locus (22, 27, 43). Large plasmids of STEC encode determinants that are thought to be additional virulence factors. These include enterohemolysin, which is encoded by the ehxA gene and acts as a pore-forming cytolysin on eukaryotic cells, an extracellular serine protease (EspP), and a catalase-peroxidase (KatP) (8, 9, 41). Phylogenetic studies have indicated that E. coli is composed of four main phylogenetic groups, A, B1, B2, and D, with STEC and enterohemorrhagic E. coli strains falling into phylogenetic groups A, B1, and D (14, 20). According to their incidence and association with HUS and outbreaks, STEC strains have also been classified into five seropathotypes (A to E) by Karmali et al. (25).

One of the more common sources of human infection is the consumption of undercooked ground beef, and many outbreaks described in the United States have been linked to this source. Dairy products have also been involved in outbreaks, although less frequently (39). In France, several clusters of HUS cases linked to dairy products have been identified and investigated (15), and there have been two recent reports of O157:H7- and O26-associated outbreaks (16; E. Espié, P. Mariani-Kurkdjian, F. Grimont, N. Pihier, V. Vaillant, S. Francart, H. de Valk, and C. Vernozy-Rozand, poster presented at the 6th International Symposium on Shiga Toxin (Verocytotoxin)-Producing Escherichia coli Infections, Melbourne, Australia, 2006, http://www.invs.sante.fr/surveillance/shu/poster_melbourne_o26.pdf). In the present study, we performed a detailed analysis of the molecular characteristics of STEC strains isolated from HUS patients and from dairy samples during prospective studies. Our aims were (i) to identify the characteristics of the strains isolated from HUS patients, (ii) to analyze the factors previously described as potentially associated with STEC virulence in a collection of dairy strains, (iii) to compare strains at the molecular level using powerful techniques such as ribotyping and pulsed-field gel electrophoresis (PFGE), and (iv) to assess the virulence of strains isolated from dairy products.

MATERIALS AND METHODS

Bacterial strains and culture conditions.

A total of 55 STEC isolates collected in France were examined in the study, of which 40 were isolated from 1,130 dairy samples, which included 205 raw-milk (166 from cows and 39 from ewes), 828 cheese (781 from cows, 32 from ewes, and 15 from goats), and 97 “dairy environment” (87 from cows and 10 from ewes) samples taken during previous studies (17, 37). The samples were randomly collected over a 2-year period from several regions in France. They were purchased from local retail stores, provided by the producers, or obtained from the Laboratoires Départementaux d'Analyses Vétérinaires et Biologiques. The different types of cheeses and milk were not epidemiologically related and were representative of French production as a whole. A total of 15 STEC samples were isolated from patients in central France admitted to hospitals for HUS or HC during the same period. The strains are representative of STEC isolated from patients in a French hospital and include strains of serogroups O157, O103, and O91. Serotype determination of HUS/HC-associated strains was performed by Flemming Scheutz at the International Escherichia and Klebsiella Reference Centre (WHO) in Copenhagen, Denmark. For the dairy STEC sample provided by Fach et al. (17), only the O group had been determined. E. coli EDL933 (ATCC 43895) serotype O157:H7 was used as an enterohemorrhagic E. coli reference strain, and E. coli K-12 was used as a negative control for PCR experiments.

Detection of genes associated with virulence.

This detection was done by PCR using the primers listed in Table 1. The DNA to be amplified was released from whole organisms by boiling the samples. The PCR cycles included denaturation for 60 s at 94°C, primer annealing for 60 s at 45 to 68°C (primer-specific annealing temperatures are shown in Table 1), and an extension for 60 or 120 s (according to the length of the fragment) at 72°C (30 cycles). The amplification reactions were performed on a Perkin-Elmer GeneAmp 2400 PCR system. The presence of three genes (espP, ehxA, katP) was also determined by colony blot hybridization, using probes labeled with [α-³²P]dCTP, as described previously (36). The probes used for the detection of target genes were obtained by PCR from strain EDL933, using the primer pairs espA-espB for espP, RH35-RH37 for ehxA, and wkatb-wkatf for katP (Table 1). The probe sizes were 1,830 bp, 321 bp, and 2,125 bp, respectively.

TABLE 1.

Target genes and primers used in PCR studies

Target gene(s) or region	Primer	Oligonucleotide sequence (5′-3′)	Annealing temp (°C)	Reference
stx1	VT1c	ACC CTG TAA CGA AGT TTG CG	55	35
	VT1d	ATC TCA TGC GAC TAC TTG AC	55	35
stx1-933J, stx1c, stx1d	Gannon-F	ACA CTG GAT GAT CTC AGT GG	60	19
	Gannon-R	CTG AAT CCC CCT CCA TTA TG	60	19
stx1c	Lin-up	GAA CGA AAT AAT TTA TAT GT	48	26
	AOX3	CTC ATT AGG TAC AAT TCT	48	26
	StxV1F1	TCG CAT GAG ATC TGA CC	50	26
	StxV1B1	AAC TGA CTG AAT TGA GATG	50	26
stx2	LP43	ATC CTA TTC CCG GGA GTT TAC G	55	12
	LP44	GCG TCA TCG TAT ACA CAG GAG C	55	12
stx2-EDL933, stx2vh-a, stx2vh-b	VT2c	AAG AAG ATG TTT ATG GCG GT	55	33
stx2-NV206	VT2d	CAC GAA TCA GGT TAT GCC TC	55	33
stx2d	VT2cm	AAG AAG ATA TTT GTA GCG G	55	33
	VT2f	TAA ACT GCA CTT CAG CAA AT	55	33
stx2d-Ount, stx2d-OX3	VT2e	AAT ACA TTA TGG GAA AGT AAT A	55	33
	VT2f	TAA ACT GCA CTT CAG CAA AT	55	33
stx2e	VTea	CCT TAA CTA AAA GGA ATA TA	45	33
	VTeb	CTG GTG GTG TAT GAT TAA TA	45	33
eae	eae F3	GAA CGG CAG AGG TTA ATC TGC	55	This study
	eae R3	TCA ATG AAG ACG TTA TAG CCC	55	This study
eae variants	B73	TAC TGA GAT TAA GGC TGC TAA	50	13
eae α	B138	GAC CAG AAG AAG ATC CA	50	13
eae β	B137	TGT ATG TCG CAC TCT GAT T	50	13
eae γ	B74	AGG AAG AGG GTT TTG TGT T	50	13
eae δ	Intδ	TAC GGA TTT TGG GGC AT	55	1
	Int-Ru	TTT ATT TGC AGC CCC CCA T	45	1
eae ɛ	SK1	CCC GAA TTC GGC ACA AGC ATA AGC	52	29
	LP5	AGC TCA CTC GTA GAT GAC GGC AAG CG	52	29
efa1	88AT	AAG GTG TTA CAG AGA TTA	51	28
	88TN	TGA GGC GGC AGG ATA GTT	51	28
ehxA	HlyAF	GCA TCA TCA AGC GTA CGT TCC	60	30
	HlyAR	AAT GAG CCA AGC TGG TTA AGC T	60	30
ehxA	RH35	CAC ACG GAG CTT ATA ATA TTC TGT CA	68	23
	RH37	AAT GTT ATC CCA TTG ACA TCA TTT GAC T	68	23
astA	east11a	CCA TCA ACA CAG TAT ATC CGA	45	47
	east11b	GGT CGC GAG TGA CGG CTT TGT	45	47
bfp	EP1	AAT GGT GCT TGC GCT TGC TGC	53	21
	EP2	GCC GCT TTA TCC AAC CTG GTA	53	21
katP	wkatb	CTT CCT GTT CTG ATT CTT CTG G	56	8
	wkatf	AAC TTA TTT CTC GCA TCA TCC	56	8
espP	espA	AAA CAG CAG GCA CTT GAA CG	56	This study
	espB	GGA GTC GTC AGT CAG TAG AT	56	This study
Right junction of LEE inserted into selC	K260	GAG CGA ATA TTC CGA TAT CTG GTT	62	27
	K255	GGT TGA GTC GAT TGA TCT CTG G	62	27
	K261	CCT GCA AAT AAA CAC GGC GCA T	62	27
Left junction of LEE inserted into selC	K295	CGC CGA TTT TTC TTA GCC CA	62	27
	K296	CAT TCT GAA ACA AAC TGC TC	62	27
Insertion into pheU	K913	CAT CGG CTG GCG GAA GAT AT	52	43
	K914	CGC TTA AAT CGT GGC GTC	52	43

PCRs for LEE insertions.

The insertion of the LEE adjacent to selC was tested using the four primer pairs described by McDaniel et al. (27): K260-K261 (intact selC, 527 bp), K255-K260 (right junction of the LEE inserted into selC, 418 bp), K295-K296 (left junction of the LEE inserted into selC, 405 bp), and K260-K295 (intact selC, 2,173 bp). The primer pair K913-K914 was used to amplify the pheU gene (intact pheU, 303 bp) (43).

stx₁ subtyping.

The three stx₁ gene variants (stx_1-933J, stx_1c, and stx_1d; GenBank accession numbers M19473, Z36901, and AY170851, respectively) were amplified using the GannonF-GannonR primer pair. PCR products were then subjected to restriction endonuclease digestion with BglI, HaeI, and RsaI (Roche Applied Science). In silico analysis revealed that the 603-bp amplified product from STEC strains possessing stx_1-933J (strain EDL933), stx_1c (strain DG131/3), and stx_1d (strain MHI813) generated fragments of 215 and 387 bp with BglI, 220 and 382 bp with RsaI, and 415 and 187 bp with HaeI, respectively. The presence of the stx_1c variant was confirmed by PCR amplification using the VT1AvarF-VT1AvarR and lin-up-A_OX3 primer pairs (26).

Detection of stx₂ variants by PCR.

Restriction fragment length polymorphism (RFLP)-PCR systems were used to distinguish stx₂ subtypes, as previously described (3, 33, 38, 44). The VT2c-VT2d primer pair was used to detect the stx_2-EDL933, stx_2vh-a, stx_2vh-b, and stx_2-NV206 genes. The 285-bp PCR product obtained was then subjected to restriction endonuclease digestion with HaeIII, RsaI, and NciI (Boehringer Mannheim, Meylan, France), as described by Piérard et al. (33). To identify STEC strains harboring the stx_2-NV206 gene, the HpaI enzyme was also used (Fig. 1) (3). The VT2cm-VTf primer pair was used to specifically detect the stx_2d genes; discrimination between stx_2d-Ount and stx_2d-OX3a was performed with the VT2e and VT2f primers, the 348-bp amplicon being digested with HaeIII and PvuII. The stx_2e gene variant was detected using the VT2ea-VT2eb primer pair. DNA from the E. coli strains EDL933 (O157:H7 stx_2-EDL933), B2F1 (O91:H21 stx_2vh-a, stx_2vh-b), NV206 (stx_2-NV206), Fac9 (stx_2e), an OX3:H11 (stx_2d) strain, and K12C600 (negative control) was included in each PCR run. Digestion products were separated on 2% agarose gels, and the standard-size 100-bp DNA ladder marker (Promega, France) was used as the referee.

FIG. 1.

Identification of the stx₂ gene variants by RFLP-PCR. (A) Predicted sizes of the fragments obtained with the RFLP-PCR method of Tyler et al. (44) after restriction with the HaeIII, RsaI, NciI, and HpaI endonucleases. (B) Representative profiles obtained for the stx_2-EDL933, stx_2vh-a, stx_2vh-b, and stx_2-NV206 variants. The profile obtained for strain CH013 is also shown.

Phylogenetic-group determination.

The main phylogenetic groups (A, B1, B2, and D) of the E. coli strains were determined by triplex PCR amplification as described by Clermont et al. (14).

Preparation of plasmid DNA.

Plasmid DNA was prepared from 1.5-ml overnight cultures of bacteria grown in Luria broth (Difco Laboratories, Detroit, MI) by the alkaline lysis method. Plasmid DNA patterns were obtained by electrophoresis on 0.7% agarose gels in Tris-acetate-EDTA buffer, stained with ethidium bromide, and observed under UV light.

Ribotyping.

Total genomic DNA of STEC isolates was prepared from 10-ml overnight cultures in Mueller-Hinton broth (Biokar Diagnostics, Beauvais, France) by the method described by Picard-Pasquier et al. (32). About 3 μg of genomic DNA was digested independently with two restriction endonucleases, EcoRI and HindIII (Boehringer Mannheim), according to the manufacturer's instructions. Digested DNA was separated, transferred to Hybond N+ nylon membranes (Amersham Pharmacia Biotech, Orsay, France), and probed with an rrnB7 probe as previously described (36). The RFLP Extension Ladder System (Life Technologies, Cergy Pontoise, France) was used as the molecular weight marker. The Dice similarity coefficient and the Diversity Database software (Bio-Rad) were used to compare profiles. The hierarchic unweighted pair group arithmetic average algorithm was used for cluster analysis, and dendrographic trees were constructed.

PFGE analysis.

Genomic DNA was prepared according to the protocol described by Böhm and Karch (7) using low-melting-point agarose (Gibco BRL). For restriction endonuclease digestion, 50 U of XbaI (Life Technologies, Cergy Pontoise, France) was used at 37°C for 18 h. PFGE was performed in 1% agarose (Invitrogen) gels using a contour-clamped homogeneous electric field DRIII apparatus (Bio-Rad, Ivry, France) in 0.5× Tris-borate-EDTA buffer at 14°C at 6 V/cm. The pulse time was increased from 5 to 50 s over 22 h. A lambda ladder (Bio-Rad) was used as the size marker.

Sequencing of the new stx₂ variant.

The PCR product for sequencing was prepared by using the QIAquick PCR purification kit (Qiagen S.A.). DNA sequencing was carried out on an automated sequencer by the fluorescent-dye termination method, and the sequence was edited using the manufacturer's software (GENOME Express, Grenoble, France). BLASTN and BLASTX sequence homology analyses were performed using the National Center for Biotechnology Information BLAST network service (2).

Statistical analysis.

The data were analyzed with Epi-Info version 6.02 software by the χ² test, except for the variable needing the two-tailed Fisher exact test. A P value of <0.05 was considered statistically significant.

Nucleotide sequence accession number.

The sequence of the new stx₂ variant has been deposited in the GenBank database under accession no. EU184879.

RESULTS

Detection of genes associated with virulence and significant associations.

Forty STEC strains from 1,130 dairy products or dairy environments and 15 pathogenic STEC strains were analyzed. Data regarding the virulence factor-encoding genes stx₁, stx₂, eae, efa1, ehxA, espP, katP, and astA and their distribution according to the origin of the STEC strains (dairy products versus HUS patients) are given in Table 2. The characteristics of each of the 55 STEC isolates tested are shown in Table 3. The stx₁ gene was found, alone or in association with stx₂, in 25 out of 40 dairy STEC strains (62%) and in 4 out of 15 HUS-associated strains (27%) (P = 0.018). In contrast, the stx₂ gene was present in 20 (50%) dairy isolates and in 12 (80%) HUS-associated strains and was therefore significantly associated with pathogenic strains (P = 0.045). The stx₁-stx₂ association was not frequent and was almost equally distributed in both categories (six dairy strains [15%] and one HUS patient isolate [7%]). Only 5 dairy strains out of 40 (12%) were positive for eae; they belonged to the same serogroup (O5) and shared the same characteristics (Table 3). Interestingly, 8 of the 15 HUS strains (53%) harbored the eae gene, indicating a strong association with pathogenic isolates (P = 0.003). As expected, similar results were obtained with the efa1 gene (28), which was always associated with eae, except in the CH071 strain (Table 3). The enterohemolysin-encoding gene ehxA, found in 15 dairy STEC isolates (37%) and in 9 HUS patient isolates (60%), was not clearly associated with any category. The espP gene was found more frequently in strains from patients (53%) than in dairy isolates (22%) (P = 0.032). PCR and Southern blot hybridization demonstrated that the katP gene was rarely present in dairy STEC strains (only one strain, PL475, harbored this gene) but was found in 40% of the HUS isolates (P = 0.001). The enteroaggregative heat-stable enterotoxin-encoding gene astA was present in nine dairy STEC strains but was not found in any patient isolates. Finally, the bfp gene, encoding the pili involved in the adhesion of enteropathogenic E. coli strains to intestinal cells, was absent from all the strains tested, which indicates that all tested isolates belong to the STEC group of diarrheogenic E. coli organisms.

TABLE 2.

Distribution of virulence factor-encoding genes according to the origins of the STEC isolates^a

Genec	No. (%) of isolates of indicated originb		P valuee
	Dairy product or environmentd	HUS/HC patient
stx1	25 (62)	4 (27)	0.018
stx2	20 (50)	12 (80)	0.045
eae	5 (12)	8 (53)	0.003
efa1	5 (12)	8 (53)	0.003
ehxA	15 (37)	9 (60)	NS
espP	9 (22)	8 (53)	0.032
katP	1 (2)	6 (40)	0.001
astA	9 (22)	0	NS

^aThe bfp gene was absent from all strains tested.

^bForty STEC strains from 1,130 dairy products or dairy environments and 15 pathogenic STEC strains were analyzed. The numbers in parentheses indicate percentages of isolates positive for the indicated genes.

^cThe presence of the following genes was determined by PCR: stx₁, stx₂, eae, efa1, ehxA, katP, and astA. The presence of the following genes was determined by colony blot hybridization: ehxA, espP, and katP.

^dThe five dairy isolates harboring the eae and efa1 genes belong to the same serogroup and share identical virulence gene patterns.

^eA P value of <0.05 was considered statistically significant. P values for the genes eae, efa1, espP, and katP were determined by the Fisher exact test. NS, not significant.

TABLE 3.

Virulence factors of the 55 STEC strains tested

Strain	Origin	O serogroupa	H type	Presence of indicated genec								Phylogenetic group	Seropathotype
				stx1 (variant)	stx2 (variant)	eae (variant)	efa1	ehxA	espP	katP	astA
14	Cheese	NT	NDb	−	+ (stx2vh-b)	−	−	−	−	−	−	B1	ND
15	Cheese	NT	ND	−	+ (stx2vh-b)	−	−	−	−	−	−	B1	ND
40	Cheese	NT	ND	−	+ (stx2vh-b)	−	−	−	−	−	−	B1	ND
1429	Cheese (goat)	O76	ND	+ (stx1c)	−	−	−	+	−	−	−	B1	E
2206	Cheese (goat)	O5	H−	+ (stx1c)	−	−	−	+	−	−	−	B1	C
ECA15	Cheese	O79	ND	+ (stx1-933J)	+ (stx2-EDL933)	−	−	+	+	−	−	D	E
ECA34	Cheese	O5	H−	+ (stx1-933J)	−	+ (β)	+	+	+	−	−	B1	ND
ECA36	Cheese	O5	H−	+ (stx1-933J)	−	+ (β)	+	+	+	−	−	B1	ND
ECA37	Cheese	O5	H−	+ (stx1-933J)	−	+ (β)	+	+	+	−	−	B1	ND
ECA38	Cheese	O5	H−	+ (stx1-933J)	−	+ (β)	+	+	+	−	−	B1	ND
ECA43	Cheese	O5	H−	+ (stx1-933J)	−	+ (β)	+	+	+	−	−	B1	ND
ECA64	Cheese	O77	ND	−	+ (stx2vh-b)	−	−	+	+	−	−	D	D
ECA89	Cheese	O117	ND	+ (stx1c)	−	−	−	−	−	−	−	B1	E
ECA95	Cheese	O91	ND	−	+ (stx2vh-a)	−	−	−	−	−	−	B1	C
ECA97	Cheese	O113	ND	−	+ (stx2-NV206)	−	−	−	−	−	−	A	E
PL449	Raw milk	O3	ND	+ (stx1-933J)	−	−	−	+	−	−	+	A	E
PL473	Raw milk	O3	ND	+ (stx1-933J)	−	−	−	+	−	−	+	A	E
PL479	Raw milk	O6	ND	−	+ (stx2-NV206)	−	−	−	−	−	−	A	E
PL599	Raw milk	O6	ND	−	+ (stx2vh-b + stx2-NV206)	−	−	−	−	−	−	A	E
PL600	Cheese	O6	ND	−	+ (stx2vh-b + stx2-NV206)	−	−	−	−	−	−	A	E
PL602	Cheese	O6	ND	−	+ (stx2vh-b + stx2-NV206)	−	−	−	−	−	−	A	E
PL603	Cheese	O6	ND	−	+ (stx2vh-b + stx2-NV206)	−	−	−	−	−	−	A	E
PL650	Cheese (ewe)	O110	ND	+ (stx1c)	−	−	−	−	−	−	+	A	D
PL652	Cheese (ewe)	O136	ND	+ (stx1c)	−	−	−	−	−	−	+	A	E
PL657	Cheese	O136	ND	+ (stx1-933J)	−	−	−	−	−	−	+	A	E
PL915	Raw milk	O117	ND	−	+ (stx2vh-a)	−	−	−	−	−	−	B1	E
PL27	Raw milk (ewe)	O76/O22	ND	+ (stx1c)	−	−	−	+	−	−	−	B1	ND
PL230	Environment	O6	ND	+ (stx1c)	−	−	−	−	−	−	−	A	E
PL292	Milk	O76/O22	ND	−	−	−	−	−	−	−	−	B1	ND
PL324	Milk	NT	ND	−	+ (stx2vh-b)	−	−	−	−	−	−	B1	ND
PL325	Milk	O77	ND	−	+ (stx2vh-b)	−	−	−	+	−	−	D	ND
PL475	Milk (ewe)	O110	ND	+ (stx1c)	−	−	−	−	−	+	+	A	D
PL4790	Raw milk (ewe)	O91	ND	+ (stx1-933J)	+ (stx2d-Ount)	−	−	−	−	−	−	B1	C
PL487	Raw milk	NT	ND	+ (stx1-933J)	−	−	−	−	−	−	+	A	ND
NV63	Cheese	O103	H14	+ (stx1-933J)	−	−	−	−	−	−	−	B1	E
NV64	Cheese	O174	H2	+ (stx1-933J)	+ (stx2-EDL933)	−	−	+	+	−	−	B1	C
NV72	Cheese	O15	H45	+ (stx1-933J)	+ (stx2vh-b)	−	−	−	−	−	+	D	E
NV74	Cheese	O91	H21	+ (stx1-933J)	+ (stx2vh-b)	−	−	+	−	−	−	B1	C
NV79	Cheese	O8	H9	+ (stx1-933J)	−	−	−	−	−	−	−	A	D
NV81	Cheese (ewe)	O113	H+	+ (stx1c)	+ (stx2d-Ount)	−	−	+	−	−	+	A	D
CHVi-1	HUS child	O157	H7	−	+ (stx2vh-a)	+ (γ)	+	+	+	+	−	D	A
CH1898	HUS child	O157	H7	−	+ (stx2-EDL933 + stx2vh-a)	+ (γ)	+	+	+	+	−	D	A
CH013	HUS adult	O91	H10	−	+ (stx2-EDL933 + stx2-CH013)	−	−	−	−	−	−	B1	C
CH014	HUS adult	O91	H21	−	+ (stx2-EDL933 + stx2vh-a + stx2vh-b)	−	−	+	−	−	−	B1	C
CH015	HUS adult	OR	H16	−	+ (stx2-EDL933)	−	−	+	+	−	−	B1	C
CH016	HUS adult	O174	H−	−	+ (stx2vh-a)	−	−	−	−	−	−	B1	C
CH017	HUS adult	O+	H−	−	+ (stx2vh-a)	−	−	−	−	−	−	B1	C
CH023	HUS child	O+	H−	−	+ (stx2-EDL933 + stx2vh-a)	+ (ɛ)	+	+	+	+	−	B1	C
CH071	HUS child	O157	H26	+ (stx1-933J)	+ (stx2-EDL933)	+ (β)	−	−	−	−	−	B1	C
CH075	HUS child	O157	H7	−	+ (stx2-EDL933)	+ (γ)	+	+	+	+	−	D	A
CH085	HUS adult	O91	H10	−	+ (stx2-EDL933 + stx2vh-b)	−	−	−	−	−	−	B1	C
CH087	HUS child	O103	H2	+ (stx1-933J)	−	+ (ɛ)	+	+	+	+	−	B1	B
CH089	HC adult	O103	H2	+ (stx1-933J)	−	+ (ɛ)	+	+	+	+	−	B1	B
CH105	HC adult	O8	H2	−	+ (stx2-EDL933 + stx2vh-b)	−	+	−	−	−	−	B1	C
CH123	HC child	O5	H−	+ (stx1-933J)	−	+ (β)	+	+	+	−	−	B1	C
EDL933	ATCC (reference strain)	O157	H7	+ (stx1-933J)	+ (stx2-EDL933)	+ (γ)	+	+	+	+	−	D	A

^aNT, nontypeable intimin.

^bND, not determined.

^cThe presence and absence of genes are noted as + and −, respectively.

Analysis of the stx₁ gene variants.

Three main stx₁ variants, stx_1-933J (identified in the EDL933 reference strain), stx_1-OX3 (also called stx_1c), and stx_1d, have been described in the literature (10, 26). Using the lin-up-lin-OX3 primers, we showed that 9 strains (all isolated from dairy products) possessed stx_1c and 20 strains possessed stx_1-933J (the 4 stx₁-positive HUS isolates and 16 out of 25 stx₁-positive dairy strains). No stx_1d variant was detected. No more than one variant was observed.

Analysis of the stx₂ gene variants.

To further assess the pathogenicity of the STEC isolates, the stx₂ gene subtypes were identified by PCR and RFLP-PCR. Four restriction enzymes were used in this study: HaeIII, RsaI, NciI, and HpaI. Representative patterns are shown in Fig. 1, and the results of this analysis are summarized in Table 3. The 32 stx₂-positive isolates were found to contain stx_2-EDL933, stx_2vh-a, stx_2vh-b, stx_2-NV206, or stx_2d, but none was found to possess the stx_2e gene. The strains contained either one stx₂ gene type or a combination of two or three. The CH013 strain showed an atypical RsaI profile (Fig. 1). Cloning and sequencing experiments revealed that it contained two variants: stx_2-EDL933 and a new stx₂ variant, stx_2-CH013 (GenBank accession number EU184879), whose sequence showed only 92% identity with the sequence of stx_2-NV206. The distribution of the stx₂ variants according to the origin of STEC (dairy products versus HUS patients) is given in Table 4. Of the 12 stx_2-positive HUS strains, 9 (75%) harbored the stx_2-EDL933 variant, indicating a strong association with pathogenic isolates (P = 0.0003). The stx_2vh-a variant was more frequent in the HUS isolates (P = 0.018), and the stx_2vh-b and stx_2-NV206 genes were more frequent among dairy strains but not to a significant degree. The stx_2d variant, which was recently associated with ovine isolates (38), was absent from the HUS strains and was found in only two dairy isolates. Interestingly, both of the cheese samples from which the last-named strains were isolated were made from ewe's milk. Finally, the stx_2e variant was absent from all the strains tested, which was not surprising, since the variant is clearly associated with STEC strains isolated from pigs. Six out of 12 (50%) HUS patient strains harbored several stx₂ variants (Table 3), while only 4 dairy isolates harbored a combination of two genes (stx_2vh-b and stx_2-NV206).

TABLE 4.

Distribution of the stx₂ gene variants according to the origins of the STEC isolates

stx varianta	No. (%) of isolates of indicated originb		P valuec
	Dairy product or environment	HUS/HC patient
stx2-EDL933	2 (10)	9 (75)	0.0003
stx2vh-a	2 (10)	6 (50)	0.018
stx2vh-b	12 (60)	3 (25)	NS
stx2-NV206	6 (30)	0	NS
stx2d-Ount	2 (10)	0	NS

^aThe stx_2e gene was absent from all strains tested.

^bTwenty STEC strains from dairy products or dairy environments and 12 pathogenic STEC strains were analyzed. The numbers in parentheses indicate percentages of isolates positive for the indicated genes.

^cA P value of <0.05 was considered statistically significant. P values for the stx_2-EDL933 and stx_2vh-a variants were determined by the Fisher exact test. NS, not significant.

Analysis of the LEE.

The eae gene variants were identified from the 13 eae-positive isolates (Table 3). β Intimin was present in the five dairy strains of serogroup O5 (ECA34 to ECA43). The strains isolated from patients were more heterogeneous: two harbored the β variant (the O157:H26 isolate CH071 and the O5:H⁻ isolate CH123), three harbored the ɛ variant (including the two strains of serotype O103:H2, CH087 and CH089), and three possessed the γ variant (all of serotype O157:H7). The sites of insertion of the LEE were investigated using PCR systems previously described (27, 43), and the results of this analysis are shown in Table 5. In the five dairy STEC isolates belonging to the O5 serogroup (ECA34 to ECA43), both the selC and pheU sites were not intact. However, the signals indicating the insertion of the LEE into selC (i.e., a 418-bp fragment with K255 and K260 and a 405-bp fragment with K295 and K296) were missing. Therefore, if the LEE is inserted into selC or pheU, it is “modified” in the region where the primers were chosen. Identical results were obtained for two isolates from HUS patients, CH123 (which belongs to serotype O5:H⁻ and carries a β intimin) and CH023 (O⁺:H⁻, ɛ intimin). The LEE was inserted into selC in the four pathogenic isolates belonging to the O157 serogroup (CH071, CH075, CHVi-1, CH1898). For three of them, the LEE was modified either in the left junction (CH071) or in the right junction (CHVi-1, CH1898). An intact selC region was observed in three isolates. For two of them, strains CH087 and CH089, which belong to the O103:H2 serotype and carry the ɛ intimin, the LEE might be inserted into pheU.

TABLE 5.

Analysis of the insertion of the LEE into selC and pheU

Strain	O group	H typea	Intimin type	Presence of a signal of the size expected for the LEE with the indicated primer pair inb:
				Intact selC (K260-K261 [527 bp])	Right junction of selC (K255-K260 [418 bp])	Intact selC (K260-K295 [2,173 bp])	Left junction of selC (K295-K296 [405 bp])	Intact pheU (K913-K914 [303 bp])
ECA34	O5	ND	β	−	−	−	−	−
ECA36	O5	ND	β	−	−	−	−	−
ECA37	O5	ND	β	−	−	−	−	−
ECA38	O5	ND	β	−	−	−	−	−
ECA43	O5	ND	β	−	−	−	−	−
CH123	O5	H−	β	−	−	−	−	−
CH023	O+	H+	ɛ	−	−	−	−	−
CH087	O103	H2	ɛ	+	−	+	−	−
CH089	O103	H2	ɛ	+	−	+	−	−
CH071	O157	H26	β	−	+	−	−	+
CH075	O157	H7	γ	−	+	−	+	+
CHVi-1	O157	H7	γ	−	−	−	+	+
CH1898	O157	H7	γ	−	−	−	+	+
EDL933	O157	H7	γ	−	+	−	+	+

^aND, not determined.

^bThe presence and absence of a signal of the expected size with the indicated primer pairs are noted as + and −, respectively.

Plasmid content analysis.

The presence or absence of plasmids and their number and size were investigated in the 55 STEC strains. The plasmid patterns were heterogeneous: 11 isolates did not carry any visible plasmids, and 44 strains carried one to three plasmids that ranged from approximately 3 to more than 90 kb in size (data not shown). Some isolates belonging to the same serotypes presented distinct profiles, further evidence of the high heterogeneity of STEC isolates with regard to mobile elements such as plasmids. The numbers of plasmids according to the origins of the STEC isolates (dairy product versus HUS patients) are shown in Table 6. No significant association was observed.

TABLE 6.

Distribution of the plasmids observed after extraction and migration in a 0.8% agarose gel according to the origins of the STEC isolates

No. of plasmids	No. (%) of isolates of indicated origina		P valueb
	Dairy product or environment	HUS/HC patient
0	10 (25)	1 (7)	NS
1	19 (48)	9 (60)	NS
2	8 (20)	2 (13)	NS
3	3 (7)	3 (20)	NS

^aForty STEC strains from dairy products or dairy environments and 15 pathogenic STEC strains were analyzed. The numbers in parentheses indicate percentages of isolates positive for the indicated number of plasmids.

^bA P value of <0.05 was considered statistically significant. NS, not significant.

Analysis of the phylogenetic group.

Of the 40 STEC dairy isolates, 16 belonged to the A group, 20 to the B1 group, and 4 to the D group (Table 3). Most of the strains isolated from patients belonged to the B1 group (12 strains), and the three strains of serotype O157:H7 belonged to the D group, but none fell into the A group, confirming that the most virulent strains belong either to the D or to the B1 groups (20). Interestingly, a large number of isolates from dairy products belong to the A group (Fisher exact test, P = 0.002), which is associated with less virulent STEC strains.

Ribotyping profiles of the STEC isolates.

The 55 STEC isolates were analyzed by ribotyping using the rrnB7 probe. A total of 23 different patterns were obtained with HindIII-digested genomic DNA, but two isolates (NV63 and PL915) could not be typed. When the EcoRI enzyme was used, 28 different patterns were observed, but six isolates (CH016, PL27, 1429, PL325, PL915, ECA64) could not be typed. The HindIII ribotyping patterns showed between 7 and 10 fragments ranging from 1 to 26 kb in length. Figure 2 shows a dendrographic tree constructed from these HindIII profiles. The six O6 dairy isolates exhibited the same pattern, which was identical to those of O110 strains PL475 and PL650, O113 strain NV81, and O136 strain PL652. The four O157:H7 strains (including the reference strain EDL933) presented identical patterns, which were the same as those of the O3 and O15 isolates. Five O5 isolates (ECA34, ECA36, ECA37, ECA38, ECA43) out of seven shared the same profile. The O5 strain isolated from a patient (CH123) presented a slightly different pattern, and the last O5 dairy isolate, 2206, had a distinct ribotype profile, which indicates a more distant relationship. Four nontypeable isolates (14, 15, 40, PL324) out of five had identical ribotype profiles, as did two O77 isolates, three O91 isolates, three O76 isolates, two O103 isolates, a group of two isolates (one O117 and one O⁺ isolate), and a group of three isolates including one O174 strain and two O8 strains. The STEC strains isolated from patients were scattered along the tree and showed patterns similar to those of dairy STEC strains (Fig. 2). However, only six of the latter showed profiles identical to those of clinical isolates, and only one of them (ECA95, an O91 strain) also exhibited an EcoRI profile identical to that of clinical isolates (CH013 and CH085, both of which are O91:H10 strains) (data not shown). This may indicate that the great majority of dairy STEC strains are distantly related to pathogenic isolates.

FIG. 2.

Dendrogram analysis of STEC strains isolated from dairy samples and HUS patients in France. The strain names, serogroups, and origins are indicated. The tree was constructed using the hierarchic unweighted pair group arithmetic average algorithm on a matrix resulting from comparisons of HindIII ribotyping patterns.

PFGE analysis.

To further determine relatedness among human and dairy isolates showing similar ribotypes, we analyzed their XbaI patterns by PFGE. Ten to 19 fragments, ranging in size from about 50 kb to 580 kb, were observed. Strains showing identical ribotypes but belonging to different serotypes had only distantly related PFGE patterns, confirming the discriminatory power of PFGE. Most of the isolates were not genetically related, since their PFGE patterns showed more than six band differences. Of the strains with identical ribotypes, four O6 strains (PL599, -600, -602, -603) had identical PFGE patterns, suggesting that they belong to the same clone (named hereinafter the PL599 clone). The PFGE pattern of the O6 strain PL479 showed more than six band differences from the latter. Interestingly, this strain harbored only the stx_2-NV206 variant, whereas the PL599 clone harbored a combination of two genes (stx_2vh-b and stx_2-NV206). The PFGE pattern of PL230, the last O6 isolate, was completely different from the others, indicating a more distant relatedness. The five O5 isolates ECA34, ECA36, ECA37, ECA38, and ECA43 showed highly related restriction patterns, with just one additional band for ECA34 and two additional bands for ECA43. The profiles of the two other O5 isolates (including the HUS isolate CH123) were different and more distantly related. The profiles of the four O157:H7 isolates were related (11 bands in common) but differed by 3 to 5 bands. Among the O91 isolates, the dairy strain ECA95 had a profile close to that of the clinical O91:H10 strain CH013, but these two profiles differed by more than six bands.

DISCUSSION

In this study, we performed a detailed analysis of the molecular epidemiology of STEC, using a collection of 55 STEC isolates from sporadic cases of HUS or HC and from milk or other dairy products from France. The two groups of strains had clearly different characteristics. Regarding virulence genes, STEC strains associated with severe human disease were statistically associated with the stx₂ gene (particularly with the stx_2-EDL933 and stx_2vh-a variants and frequently with several stx₂ gene variants), but they were also associated with eae, espP, and katP. In contrast, dairy food STEC strains were characterized by a predominance of stx₁, with a minority of isolates harboring eae, espP, and/or katP. Recent studies of strains isolated from cheese and other dairy products evidenced the difficulties encountered in defining virulence markers of STEC (4, 11, 40, 46). Our findings supply further information. Although the ehxA gene had previously been suspected to be a marker for virulence (41), possibly because of a preferential association of ehxA with eae-positive isolates (6, 37), we did not find any association of ehxA with pathogenic isolates. In contrast, our data revealed a very strong association of katP (and to a lesser extent espP) with pathogenic strains. KatP is a catalase-peroxidase which is thought to enhance the resistance of strains to oxidative stress, but its role in virulence remains unclear (8). EspP is able to degrade coagulation factor V, which could increase hemorrhage into the gastrointestinal tract (9). In our study, the katP gene was found only in eae-positive strains, with one exception (the dairy O110 isolate PL475), and similar results were found when analyzing bovine STEC strains (37; unpublished data). In STEC strains of serogroup O103, O26, O111, and O157, the katP gene was also always present in eae-positive isolates (42). This raises the question of the preferential association of plasmid-borne genes with chromosomal genes such as eae.

Among bovine STEC strains isolated in central France, the most frequent stx₂ subtypes were stx_2vh-b (39.5% of the isolates), stx_2-EDL933 (39%), and stx_2vh-a (25.5%), followed by stx_2-NV206 (14.5%) and stx_2d (8.5%) (3). In dairy STEC strains, the stx_2vh-b variant predominated (60%) over stx_2-NV206 (30%), stx_2-EDL933 (10%), stx_2vh-a (10%), and stx_2d (10%). In contrast, the stx_2-EDL933 and stx_2vh-a variants were overrepresented in strains isolated from patients (75% and 50%, respectively), suggesting a major role in the genesis of severe disease in humans. The percentages of stx_2d-positive strains were similar in dairy STEC strains and in bovine strains. The stx_2d variants, which were recently associated with STEC strains isolated from sheep (38), are rarely found in STEC strains involved in severe disease, suggesting that stx_2d production is not a characteristic of human-pathogenic strains. This observation is in correlation with the results obtained by Vernozy-Rozand et al. (46). Finally, we identified a new stx₂ variant, stx_2-CH013 (GenBank accession number EU184879), in an O91:H10 HUS isolate and observed a combination of several stx₂ variants in 27% of the clinical isolates. We previously showed that HUS-associated strains were more cytotoxic than their bovine counterparts belonging to the same serotype (37). The number of stx₂ gene variants could affect the level of cytotoxicity and thus the level of pathogenicity of clinical isolates.

The most frequent serogroups isolated from cattle during a previous study were OX3/O174, O113, O22, O91, O172, O6, OX178, O171, O46, and O74. Strains belonging to these serogroups accounted for 70% of the total number of bovine isolates but for only 32% of the dairy STEC isolates (37). This, together with data regarding virulence factors and stx₂ variants, confirms that the STEC strains isolated from cheese samples and from infected patients represent only particular subsets of strains isolated from cattle. It is conceivable that STEC strains isolated from cheese possess special properties enabling them to survive during the cheese-making process and that the properties needed for full virulence in humans are different.

One of the molecular techniques used to assess relationships between the STEC strains was the determination of the LEE insertion site. Because of numerous modifications in the regions surrounding the tRNA genes, which can be considered “hot spots” of insertion, the PCR systems described to date are not efficient enough to determine into which site the LEE is inserted. Although these PCR systems provide important information on clonal relationships between isolates, new tools are needed to better define the locus into which the LEE is inserted. Ribotyping was useful for establishing relationships among STEC strains belonging to a wide range of serogroups, but it was not able to discriminate between STEC isolates of identical serotypes. Plasmid profile analysis has been a useful tool in molecular epidemiology for the interstrain differentiation of STEC isolates belonging to the same serotype (36). In the present study, this method provided confirmation of a common clonal origin of STEC strains of identical serotypes having identical virulence patterns, but no significant association was found between the presence and/or number of the plasmids and the pathogenicity of the strains.

The molecular techniques allowing analysis of clonal relationships (i.e., subtyping of intimin, determination of the LEE insertion site, plasmid profiling, ribotyping, and PFGE) revealed that two groups of four O6 and five O5 isolates corresponded to two clones. The O5 “ECA34 clone” exhibited characteristics very close to those of a strain isolated from a child suffering from HC in 2000 (CH123); both belong to the O5 serogroup, harbor the β intimin variant, and have identical profiles for LEE insertion (Table 5). However, the HindIII ribotypes and PFGE patterns were different, indicating that the O5 dairy clone was probably not the origin of this clinical strain. A dairy strain belonging to the O91 serogroup (ECA95) showed HindIII and EcoRI ribotypes identical to those of two STEC strains belonging to the O91:H10 serotype, isolated from HUS patients in March 1997 (CH013) and in December 1998 (CH085). The pulsotypes of the clinical strains CH013 and CH085 differed by more than six bands from that of ECA95. These strains harbored none of the virulence factors described previously except the stx₂ gene. Interestingly, the dairy isolate harbored only the stx_2vh-a variant, whereas the clinical isolates harbored two stx₂ variants, which may account for the differences in the pulsotypes and possibly in virulence power.

In conclusion, our results indicate that STEC strains isolated from dairy food samples in France possessed major characteristics different from those of strains isolated from patients. STEC strains involved in severe human disease were statistically associated with the presence of several stx₂ gene variants, mainly stx_2-EDL933 and stx_2vh-a, but also with eae, espP, and katP. In contrast, dairy food STEC strains were characterized by a predominance of stx₁ and belonged to phylogenetic A group. These findings may be helpful for the selective detection of strains that present a risk to public health.