Enterohemorrhagic Escherichia coli in Human Infection: In Vivo Evolution of a Bacterial Pathogen

MAJOR ARTICLE Enterohemorrhagic Escherichia coli in Human Infection: In Vivo Evolution of a Bacterial Pathogen Alexander Mellmann, 1,a Martina Bielaszewska, 1,a Lothar B. Zimmerhackl, 4 Rita Prager, 3 Dag Harmsen, 2 Helmut Tschäpe, 3 and Helge Karch 1 1 Institute for Hygiene, National Consulting Laboratory on Hemolytic Uremic Syndrome and IZKF Münster, and 2 Department for Periodontology, University Hospital Münster, Münster, and 3 National Reference Center for Salmonella and Other Enteric Pathogens, Robert Koch Institute, Branch Wernigerode, Wernigerode, Germany; and 4 Department of Pediatrics, Innsbruck University Hospital, Innsbruck, Austria (See the editorial commentary by Robins-Browne on pages 793 4) Background. Enterohemorrhagic Escherichia coli (EHEC) cause most cases of the hemolytic uremic syndrome (HUS) worldwide. To investigate genetic changes in EHEC during the course of human infection, we analyzed consecutive stool samples and shed isolates from patients with HUS, focusing on the genes encoding Shiga toxin (stx) and intimin (eae). Methods. Sequential stool samples from 210 patients with HUS were investigated for the persistence of E. coli strains harboring stx and/or eae. Initial stool samples were collected during the acute phase of HUS, and subsequent stool samples were collected 3 16 days later (median interval, 8 days). Results. Organisms that were stx and eae positive (stx+/eae+ strains; n p 137) or stx negative and eae positive (stx /eae+ strains; n p 5) were detected in the initial stool samples from 142 patients. Subsequently, the proportion of those who shed stx+/eae+ strains decreased to 13 of 210 patients, whereas the proportion of those who shed strains that were stx /eae+ increased to 12 of 210 patients. Seven patients who initially excreted strains that were stx+/eae+ shed, at second analysis, stx /eae+ strains of the same serotypes; they had no free fecal Shiga toxin at follow-up. Comparison of the initial and follow-up isolates from these patients with use of molecular-epidemiological methods revealed loss of stx genes and genomic rearrangement. Conclusions. We demonstrate the loss of a critical bacterial virulence factor from pathogens during very brief intervals in the human host. These genetic changes have evolutionary, diagnostic, and clinical implications. Generation of stx mutants might contribute to subclonal evolution and evolutionary success. Gastrointestinal infections with enterohemorrhagic Escherichia coli (EHEC) usually manifest as acute, afebrile, painful, bloody diarrhea. EHEC are the leading precipitants of hemolytic uremic syndrome (HUS), an important cause of childhood acute renal failure [1]. E. coli O157:H7 is the most common EHEC serotype associated with HUS, but several non-o157:h7 serotypes, such as O26:H11, O103:H2, O111:H8, O145: H28, and O157:NM (nonmotile), have emerged as causes of this disease [2 8]. Major HUS-associated EHEC produce Shiga toxin Received 25 February 2005; accepted 26 April 2005; electronically published 4 August 2005. a A.M. and M.B. contributed equally to this article. Reprints or correspondence: Dr. Alexander Mellmann, Institut für Hygiene, Universitätsklinikum Münster, Robert Koch Str. 41, 48149 Münster, Germany (mellmann@uni-muenster.de). Clinical Infectious Diseases 2005; 41:785 92 2005 by the Infectious Diseases Society of America. All rights reserved. 1058-4838/2005/4106-0003$15.00 (Stx) 1, Stx2, or Stx2c, either singly or in combinations [2, 5, 6, 9, 10]. Stxs are believed to be the major precipitants of the microvascular thrombi that form the histopathological basis of HUS [11]. However, EHEC also produce factors other than Stx that can contribute to the pathogenesis of HUS [12]. Candidates for such non-stx putative virulence factors include the EHEC pore-forming hemolysin [13] and intimin, encoded by eae [14]. Intimin mediates intimate attachment to epithelial cells in vitro and in animal models [15, 16], and it is found in almost all EHEC strains from patients with HUS [5, 6]. Enteropathogenic E. coli (EPEC) share intimin with the most common varieties of EHEC but lack Stx [17]. Many of the potential pathogenicity factors of EHEC are encoded by mobile genetic elements, such as plasmids, pathogenicity islands, and bacteriophages [12]. The association between DNA transfer and virulence is well documented [18]. The stx genes in EHEC are encoded on temperate lambdoid bacterio- EHEC Evolution In Vivo CID 2005:41 (15 September) 785

phages [19, 20] and can be eliminated from such strains by inducing the stx-converting phages with sublethal doses of UV light, mitomycin C, or various antibiotics [19, 21, 22]. Little is known about genetic changes sustained by EHEC during human infection. Therefore, in this study, we investigated genetic changes of EHEC strains during the course of HUS. To accomplish this aim, we analyzed consecutive stool samples obtained from patients with HUS with use of molecular detection of stx and eae genes to identify strains that were stx positive and eae positive (stx+/eae+ strains) and strains that were stx negative and eae positive (stx /eae+ strains). We compared isolates from consecutive stool samples of the same patient with regard to extended virulence profiles, plasmid profiles, and PFGE patterns. Also, we investigated molecular and phenotypic characteristics that can be used for the laboratory diagnosis of both stx+ and stx E. coli strains associated with HUS. PATIENTS AND METHODS Patients. During routine diagnostic efforts conducted between 1996 and 2003, sequentially collected stool samples from 210 patients with HUS were investigated for the presence of E. coli strains harboring stx and/or eae genes. The patients were from different regions of Germany, and, except for 3 who were part of a small outbreak of HUS [23], none showed temporal or geographical linkage. HUS was defined as a case of microangiopathic hemolytic anemia (i.e., hematocrit!30% with peripheral evidence of intravascular hemolysis), thrombocytopenia (i.e., platelet count!150,000 platelets/mm 3 ), and renal insufficiency (i.e., serum creatinine concentration greater than the upper limit of the normal range for age) [24]. The initial stool samples were collected 5 14 days after the onset of prodromal diarrhea (median interval, 9 days), which corresponds to the early HUS phase of the illness. The follow-up stool samples were collected 3 16 days later (median interval, 8 days). Stool sample analysis. Stx in stool filtrates was detected using the Vero cell cytotoxicity assay [25]. Screening for and isolation of E. coli strains harboring target loci were performed as described elsewhere [5]. In brief, enriched stool samples were analyzed for E. coli O157 by immunomagnetic separation technique and culture of magnetically separated organisms on sorbitol MacConkey (SMAC) agar and cefixime-tellurite SMAC agar. To identify non-o157 stx+ or eae E. coli strains, 200 ml of the enrichment broth was cultured on SMAC agar and enterohemolysin agar [13]. The overnight bacterial growth was harvested into 1 ml of saline, and 10 6 cells were used in PCR with primer pairs KS7-KS8 [26], LP43-LP44 [27], and SK1- SK2 [26], which are specific for stx 1, stx 2, and eae, respectively. To isolate strains from samples with positive PCR results, bacterial suspensions were diluted and restreaked on SMAC plates to obtain 300 400 separated colonies. The colonies were transferred in parallel to 3 nylon membranes (Zeta-probe GT; Bio- Rad) that were subsequently hybridized with digoxigenin-labeled stxb 1, stxa 2,oreae probes [26]. The hybridizing colonies were subsequently isolated from the original SMAC plates. Detection of other enteric bacterial pathogens in stool samples. Salmonella species, Shigella species, Yersinia enterocolitica, and Campylobacter jejuni were sought using standard procedures. Serological investigation. Purification of lipopolysaccharides (LPSs) O26 and O157 and the detection of IgM antibodies against these LPSs in serum samples obtained during the acute phase of HUS were performed as described elsewhere [28]. Phenotyping methods. Isolates were serotyped using a microtiter method [29]. Biochemical identification was performed with API 20 E (biomérieux). In addition, sorbitol fermentation was detected on SMAC agar. Stx production was tested using a latex agglutination assay [30]. The enterohemolytic phenotype was sought on enterohemolysin agar [13, 31]. Genotyping methods. The PCR strategy to detect stx was described elsewhere [5, 32, 33]. Detection and characterization of eae was performed as described elsewhere [26, 34, 35]. The EHEC hemolysin gene (EHEC-hlyA) and the sfpa gene [36] were detected using primer pairs hlya1-hlya4 [13] and sfpa- U sfpa-l [37], respectively. The cytolethal distending toxin (CDT) V gene cluster was detected as described elsewhere [38]. E. coli O157:H7 strain EDL933 [39], E. coli O26:NM strain 5720/96 [10], and SF E. coli O157:NM strain 493/89 [38] were used as positive controls in PCRs. flic restriction fragment length polymorphism (RFLP) analysis was performed as described elsewhere [40]. Plasmid profiles and PFGE analysis. Plasmid profiles were determined as described elsewhere [9]. PFGE was performed using XbaI-digested genomic DNA from Salmonella braenderup strain H9812 (Centers for Disease Control and Prevention) as a size marker [41]. Restriction fragment patterns of genomic DNA were analyzed with BioNumerics software, version 4.0 (Applied Maths BVBA) and analyzed with the criteria of Tenover et al. [42]. Statistical analysis. Statistical analysis was performed using 2-tailed Fisher s exact test. P values!.05 were considered to be statistically significant. RESULTS Analysis of initial stool specimens from patients with HUS for stx+ and/or eae+ E. coli. Initial cultures of stool samples from 137 of 210 patients with HUS were positive for stx and eae genes (table 1). From the initial stool cultures for 2 patients that tested positive for stx but not eae, stx+/eae strains of serotypes O91:H21 and O113:H21 were isolated. Stool cultures from 5 additional patients were eae+ and stx. Fecal filtrates from 2 patients in which neither stx+ nor eae strains were 786 CID 2005:41 (15 September) Mellmann and Bielaszewska et al.

Table 1. Analysis of consecutive stool samples from 210 patients with hemolytic uremic syndrome by means of PCR screening and colony blot hybridization with probes complementary to the stx 1, stx 2, and eae genes and by the Vero cell assay. Findings, by initial stool specimen result Variable stx+ a /eae+/vero+ stx /eae+/vero stx+ b /eae /Vero+ stx /eae /Vero stx /eae /Vero+ No. (%) of patients 137 (65.2) 5 (2.4) 2 (1.0) 64 (30.4) 2 (1.0) Serotype (no. of isolates) O26:H11/NM (21), O26:H11/NM (3), O91:H21 (1), NA NA O103:H2 (6), O111:H8/NM (8), O145:H28 (11), O157:NM (30), c O157:H7 (61) O145:H28 (1), O157:NM (1) c O113:H21 (1) Second specimen result stx+/eae+/vero+ No. of patients 12 0 0 1 0 Serotype (no. of isolates) O26:H11/NM (3), NA NA O157:H7 (1) NA O111:H8/NM (2), O145:H28 (1), O157:NM (2), c O157:H7 (4) stx /eae /Vero No. of patients 118 0 2 63 2 Serotype (no. of isolates) NA NA NA NA NA stx /eae+/vero No. of patients 7 5 0 0 0 Serotype (no. of isolates) O26:H11/NM (6), O157:NM (1) c O26:H11/NM (3), O145:H28 (1), O157:NM (1) c NA NA NA NOTE. eae+, eae gene present; eae, eae gene absent; NA, not applicable; stx+, stx gene present; stx, stx gene absent; Vero+, positive Vero cell assay result; Vero, negative Vero cell assay result. a stx types included stx 1 (22 cultures positive by PCR and hybridization), stx 2 (110), and stx 1 and stx 2 (5). b Both isolates were stx 2. c Sorbitol fermenting. detected were found to contain Stx activity. In all other patients, the Vero cell assay and genetic analysis agreed (table 1). Stools from all patients were negative for Salmonella species, Shigella species, Yersinia enterocolitica, and Campylobacter jejuni. Follow-up stool analysis. To determine whether the pathogen population changes over time within individual patients, we analyzed follow-up stool samples. In 1 of the 64 subjects with neither stx- and/or eae-harboring E. coli nor detectable Stx activity in their initial stools, an stx-positive, eae-positive E. coli O157:H7 strain was subsequently identified (table 1), suggesting intermittent shedding. The 5 patients shedding stx / eae+ E. coli strains in their initial stools had identical strains at follow-up (table 1). Of the 137 patients who had stx- and eae-positive strains in their initial stools, only 12 (8.8%) shed the infecting strains in follow-up stools. In 118 patients (86.1%), the follow-up stool samples were negative for stx 1, stx 2 and eae genes by PCR screening and contained no free Stx, as detected by the Vero cell cytotoxicity assay (table 1). In the remaining 7 patients who shed stx+/eae+ strains of serotypes O26:H11/NM (6 patients) or O157:NM (1 patient) in their initial stool samples, the follow-up stool samples yielded stx / eae+ strains of the same serotype (tables 1 and 2). Together, this replacement of an stx+/eae+ organism by an stx /eae+ organism of the same serotype in a follow-up stool sample was observed in 6 (28.6%) of 21 patients infected with E. coli O26, 1 (3.3%) of 30 patients infected with SF E. coli O157:NM, but none of 86 patients infected with EHEC strains of other serotypes including O157:H7 (table 1). Thus, the conversion from stx+ to stx strains was significantly more frequent among E. coli O26 (6 of 21) than among EHEC of all other serogroups combined (1 of 116) ( P!.001, by 2-tailed Fisher s exact test), suggesting that the stx prophage may be less stable in EHEC O26 than in other EHEC strains. Of interest, this process occurred independently in all 3 patients who were affected during a small EHEC O26:H11 outbreak [23] (patients A, C, and E; table 2), raising the possibility that stx lability is a function of individual strains. There was free fecal Stx in the initial stool samples but not in the subsequent stool samples (table 2). Cultures of initial and subsequent stool samples from each of these 7 patients were negative for other bacterial enteric pathogens. Acute-phase serum samples, which were available from 6 of these patients, demonstrated IgM antibodies to cognate (but not heterologous) LPS (table 2). Genotypic and phenotypic characteristics of consecutive EHEC Evolution In Vivo CID 2005:41 (15 September) 787

Table 2. Replacement of enterohemorrhagic Escherichia coli strains by stx-negative, eae-positive variants of the same serotypes within individual patients with hemolytic uremic syndrome. Patient, stool specimen Age, months Specimen collection, days after onset a PCR result, by primer Colony blot hybridization result, by probe b stx 1 stx 2 eae stx 1 stx 2 eae Serotype of the isolate Free fecal Stx result c Serum anti-o26 IgM Serum anti-o157 IgM A 13 Initial 5 Neg Pos Pos NP Pos Pos O26:H11 Pos Present Absent Follow-up 12 Neg Neg Pos Neg Neg Pos O26:H11 Neg NS NS B 24 Initial 9 Neg Pos Pos NP Pos Pos O26:H11 Pos Present Absent Follow-up 16 Neg Neg Pos Neg Neg Pos O26:NM Neg Present Absent C 16 Initial 7 Neg Pos Pos NP Pos Pos O26:H11 Pos Present Absent Follow-up 14 Neg Neg Pos Neg Neg Pos O26:H11 Neg NS NS D 19 Initial 7 Neg Pos Pos NP Pos Pos O26:H11 Pos Present Absent Follow-up NA Neg Neg Pos Neg Neg Pos O26:H11 Neg NS NS E 17 Initial 5 Neg Pos Pos NP Pos Pos O26:H11 Pos Present Absent Follow-up 13 Neg Neg Pos Neg Neg Pos O26:H11 Neg NS NS F 27 Initial 9 Pos Pos Pos Pos Pos Pos O26:NM Pos NS NS Follow-up 13 Neg Neg Pos Neg Neg Pos O26:NM Neg NS NS G 22 Initial 8 Neg Pos Pos NP Pos Pos O157:NM Pos Absent Present Follow-up 11 Neg Neg Pos Neg Neg Pos O157:NM Neg NS NS NOTE. NA, not applicable; Neg, negative; NM, nonmotile; NP, not performed; NS, no serum was available; Pos, positive; Stx, Shiga toxin. a Interval between the onset of prodromal diarrhea and the collection of the stool specimen. b In total, 302 406 colonies from each stool specimen that was investigated were subjected to colony blot hybridization. Isolate was defined as positive for the specified gene if 11 360 colonies hybridized with the respective probe (the colonies that hybridized with the stx 1 and/or stx 2 probe also hybridized with the eae probe) and was defined as negative for the specified gene if none of the colonies hybridized. c Isolate was defined as positive for Stx if a stool filtrate diluted 1:50 displayed a cytotoxic effect on Vero cells and as negative for Stx if no cytotoxic effect on Vero cells was observed with a stool filtrate diluted 1:50 or 1:10. stx+ and stx E. coli isolates. One stx+/eae+ isolate from the initial stool sample and 1 stx /eae+ isolate from the followup stool sample for each of these 7 patients with HUS were further analyzed. HhaI RFLP analysis of flic PCR products demonstrated that the 9 O26:H11 isolates and 3 O26:NM isolates shared a flic restriction pattern that was identical to that of the H11 type strain Su4321-41 (data not shown). Also, each of these isolates possesses eae-encoding b-intimin (table 3). The stx containing isolates produced Stx1 and/or Stx2, in accordance with their stx genotypes, whereas no Stx was detected in culture supernatants of any stx strains (table 3). All 12 O26 isolates, whether or not they possessed stx, contained the EHEC-hlyA gene and displayed an enterohemolytic phenotype. In contrast, the stx+ and stx SF E. coli O157:NM isolates produced the same flic RFLP pattern H7 and possessed eae g and EHEC-hlyA genes but did not express the enterohemolytic phenotype. Also, both contained a plasmid-borne sfpa gene encoding a sorbitol-fermenting EHEC fimbriae and the cdta, cdtb, and cdtc genes encoding CDT-V. Plasmid profiles and PFGE analysis. The consecutive stx+/ eae+ and stx /eae+ E. coli O26 isolates from patients A, B, C, and E had identical plasmid profiles (both within the same patient and among the patients). In contrast, the consecutive isolates from patient F differed from each other by 2 plasmids (table 3). The stx+ and stx E. coli O26 isolates from patient D shared a 90-kb plasmid, but the stx isolate lacked the 75- kb plasmid that was present in the stx 2 -positive strain (table 3). PFGE demonstrated that, with a single exception (patient F), the intrapatient stx+/eae+ isolates and the stx+/eae isolates differed by only 2 or 3 bands. The first isolate and the followup isolate from patient F differed by 17 bands, possibly because of a coinfection. stx E. coli as the only pathogen in patients with HUS. Five of 210 patients with HUS had, in both initial and followup stool samples, E. coli strains that possessed eae but lacked stx 1 and stx 2 genes (table 1). The eae+ isolates from initial and follow-up stool samples from each of these 5 patients belonged to the same serotype (table 1) and shared identical genotypic 788 CID 2005:41 (15 September) Mellmann and Bielaszewska et al.

Table 3. Genotypic and phenotypic characteristics of stx-positive and stx-negative Escherichia coli isolates from consecutive stool specimens from patients with hemolytic uremic syndrome. Patient, stool specimen Serotype stx gene(s) present eae gene present Plasmid profile, kbp Stx titer a Stx1 Stx2 A Initial O26:H11 stx 2 eae b 120!2 1024 Follow-up O26:H11 None eae b 120!2!2 B Initial O26:H11 stx 2 eae b 120!2 2048 Follow-up O26:NM None eae b 120!2!2 C Initial O26:H11 stx 2 eae b 120!2 1024 Follow-up O26:H11 None eae b 120!2!2 D Initial O26:H11 stx 2 eae b 90, 75!2 128 Follow-up O26:H11 None eae b 90!2!2 E Initial O26:H11 stx 2 eae b 120!2 1024 Follow-up O26:H11 None eae b 120!2!2 F Initial O26:NM stx 1, stx 2 eae b 90, 6.9 256 512 Follow-up O26:NM None eae b 120, 90!2!2 G Initial O157:NM stx 2 eae g 120!2 1024 Follow-up O157:NM None eae g 120!2!2 NOTE. NM, nonmotile; Stx, Shiga toxin. a Reciprocals of the highest dilutions of culture supernatants that caused agglutination of latex particles sensitized with polyclonal antibodies to Stx1 or Stx2, determined with the VTEC-RPLA kit (Denka Seiken). and phenotypic characteristics (table 4). Notably, except for stx genes, the genotypic and phenotypic characteristics of these stx /eae+ isolates were identical to those identified in stx+ isolates of the corresponding serotypes. Culture supernatants of these strains were not toxic to Vero cells, suggesting that they did not contain stx genes that might have been undetectable with our PCR protocol. Accordingly, initial and followup stool samples from these patients contained no free fecal Stx, as demonstrated by the Vero cell assay. None of these patients were infected with Salmonella species, Shigella species, Y. enterocolitica, orc. jejuni. DISCUSSION Genomic alterations in an infecting pathogen, in the course of an acute infection, have multiple implications. From a diagnostic standpoint, 2 observations are particularly important. First, the timing of the stool sample collection may be critical for finding Stx-producing strains in patients with HUS. Tarr et al. [44] demonstrated that, if the stool samples of patients with HUS were cultured within 6 days after the onset of diarrhea for EHEC O157:H7, the recovery rate was nearly 100%. This rate decreased to 33.3% in stool samples collected 16 days after the onset of diarrhea. In our study, we analyzed stool samples obtained from patients with HUS that were usually collected 17 days after the onset of diarrhea, and we screened the samples for EHEC by the detection of Stxs or stx. At initial examination, 141 (96.6%) of 146 strains were detected using such methods, whereas only 13 (52.0%) of 25 were detected in follow-up stool samples. This confirms a previous report [44] that EHEC are difficult to identify in patients feces late in illness. It also demonstrates that, in such cases, stx- and Stx-independent procedures are required to detect strains that might have lost their stx genes. An efficient method for the detection of stx /eae+ strains is colony blot hybridization with a probe complementary to the eae conserved region. Although this procedure is labor intensive and time consuming, it is presently used to recover stx /eae+ strains in our laboratories. However, such an approach is not without pitfalls, because it would miss strains of serotypes O91:H21 and O113:H21, which were present in 2 patients in our study. Therefore, a combination of stx and eae probes is required for the colony blot hybridization to identify members of the known spectrum of pathogenic E. coli in patients with HUS. Second, the loss of stx genes in 6 of 21 EHEC O26 and in 1 of 30 SF EHEC O157:NM but in none of 61 EHEC O157:H7 strains suggests that this phenomenon is associated with particular non-o157:h7 serotypes, especially O26:H11/NM. In this context, the finding in a clinical microbiological laboratory of an stx /eae+ E. coli O26 strain in a patient with HUS should lead a treating physician to consider such a patient as potentially infected with an EHEC strain and to follow therapeutic procedures recommended for patients infected with EHEC (e.g., to avoid antibiotic therapy). Of interest, in our study, the tendency of the stx gene loss was correlated with EHEC serogroup rather than with the median time interval between the collection of initial and follow-up stool samples. Indeed, this time interval was shorter (median interval, 6 days) for patients infected with E. coli O26 and O157:NM strains that lost their stx genes (table 2) than it was for all other patients (median interval, 8 days). Although less likely, an alternative to the loss of stx genes by the 6 EHEC O26 and 1 SF O157:NM strains during infection might be that these patients were infected with a mixed population of stx+ and stx strains from the same source and that the stx+ strains overwhelmed the stx strains during the acute phase of the illness but did not persist for as long as the stx strains. Except for the 2 SF E. coli O157:NM strains, all stx /eae+ E. coli strains could be distinguished on enterohemolysin agar. The enterohemolytic phenotype is produced by EHEC hemolysin (EHEC-Hly) [13]. Its structural gene, EHEC-hlyA [13], was present in each of the stx /eae+ E. coli strains. However, the mechanism underlying its nonexpression in SF E. coli O157: NM is unknown. Therefore, our data demonstrate that EHEC- Hly production is independent of Stx production in stx /eae+ E. coli strains O26:H11 and O145:H28, even though the pro- EHEC Evolution In Vivo CID 2005:41 (15 September) 789

Table 4. Characteristics of stx-negative, eae-positive strains of Escherichia coli isolated as the only pathogens from patients with hemolytic uremic syndrome. Patient Age, months Serotype flic- RFLP stx gene present eae gene present EHEC -hlya EHEC -Hly a Sorbitol fermentation b c Stx titer Stx1 Stx2 Vero cell assay result d H 32 O26:H11 H11 None eae b Pos Pos Pos!2!2 Neg I 21 O26:NM H11 None eae b Pos Pos Pos!2!2 Neg J 18 O26:NM H11 None eae b Pos Pos Pos!2!2 Neg K 34 O145:NM H28 None eae g Pos Pos Pos!2!2 Neg L 16 O157:NM H7 None eae g Pos Neg Pos!2!2 Neg NOTE. Only 1 isolate from each patient is shown, because the isolates from the initial and follow-up stool specimens shared identical characteristics. EHEC, enterohemorrhagic E. coli; Neg, negative; NM, nonmotile; Pos, positive; RFLP, restriction fragment length polymorphism. a Isolate was defined as positive for EHEC-Hly if a turbid zone was detected on enterohemolysin agar. b As determined by API 20 E system (biomérieux) and by growth on sorbitol MacConkey agar. c As determined by the VTEC-RPLA assay (Denka Seiken) in culture supernatants. d Isolate was defined as negative for Stx if no cytotoxic effect was observed with culture supernatants diluted 1:10. duction of EHEC-Hly is a useful marker for the detection of EHEC, as proposed by Beutin et al. [31]. From the standpoint of pathogenesis, it is particularly interesting that, in 5 patients with HUS, stx /eae+ E. coli strains were the only pathogens identified. It is possible that these strains could cause disease by a yet-to-be determined Stx-independent mechanism. Alternatively, these organisms might have colonized these patients without being involved in the pathogenesis of the underlying HUS. In some studies, stx / eae+ E. coli strains have been isolated from healthy infants [26, 43], but strains of serotype O157:H7/NM were not found, and O26:H11 strains were extremely rare [43]. In view of the instability of the stx genotype that we have demonstrated, it also seems plausible that the stx /eae+ strains might have directly descended from Stx-producing E. coli strains, which had been present in these patients earlier in the infection. In this scenario, they would have lost these loci before our analysis and, as such, would represent isogenic mutants of Stx-producing strains that have replaced the original EHEC strains. Our data offer an insight into mechanisms of pathogen evolution. Maintaining the phage encoding stx may be lethal to the bacterial host cell. Survival might be favored by loss of the phage, because such stx progeny of stx+ progenitors are less prone to lysis. Mammalian host signals, such as those initiated by hydrogen peroxide, can induce Stx-encoding prophages [45]. By generating stx mutants, a strain can survive without automatically lysing and carrying the burden of toxin production. The loss of stx-encoding phage can thus offer a selective advantage. Molecular typing methods demonstrated the relatedness of the strains within single patients, which suggests that the change in genotype is caused by the loss of relatively little genome. Even though the exact mechanisms by which stx genes are lost are still unclear, these data might explain the genome rearrangement observed by means of PFGE. A spontaneous loss of stx 1 or stx 2 genes by EHEC isolates during laboratory subcultures that resulted in nontoxigenic derivates, can occur in particular non-o157 serotypes [46], and recently, spontaneous loss of both stx 1 and stx 2 genes in vitro has also been described in an E. coli O157:H7 clinical isolate [47]. We referred to this phenomenon as clonal turnover, which defines changes in clonal composition characterized by an appearance of new clonal genotypes and loss of old ones [46]. Clonal turnover can be the result of selection for a mutant and subsequent clonal replacement within individual patients or of massive genomic rearrangements, producing the appearance of a new clone. The dynamics of these processes and their function in the evolution of these pathogens warrant further investigation. However, compared with the slow clonal turnover in H. pylori [48 50], real-time, rapidly emergent mutations occur in enterohemorrhagic E. coli during human infection. Acknowledgments We thank Philip I. Tarr for fruitful and extensive discussions of the manuscript. Financial support. The Bundesministerium für Bildung und Forschung (BMBF) Project Network of Competence Pathogenomics Alliance (BD no. 119523 to M.B.), BMBF Verbundprojekt (no. 01KI 9903 to M.B.), and the Interdisciplinary Center of Clinical Research Münster (IZKF, Project no. Ka2/061/04 to A.M.). Potential conflicts of interest. All authors: no conflicts. References 1. Tarr PI, Gordon CA, Chandler WL. Shiga toxin producing Escherichia coli and the haemolytic uraemic syndrome. Lancet 2005; 365:1073 86. 2. Banatvala N, Griffin PM, Greene KD, et al. The United States National Prospective Hemolytic Uremic Syndrome Study: microbiologic, serologic, clinical, and epidemiologic findings. J Infect Dis 2001; 183: 1063 70. 3. Eklund M, Scheutz F, Siitonen A. 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 2001; 39:2829 34. 4. Elliott EJ, Robins-Browne RM, O Loughlin EV, et al. Nationwide study 790 CID 2005:41 (15 September) Mellmann and Bielaszewska et al.

of haemolytic uraemic syndrome: clinical, microbiological, and epidemiological features. Arch Dis Child 2001; 85:125 31. 5. Friedrich AW, Bielaszewska M, Zhang W, et al. Escherichia coli harboring Shiga toxin 2 gene variants: frequency and association with clinical symptoms. J Infect Dis 2002; 185:74 84. 6. Gerber A, Karch H, Allerberger F, Verweyen HM, Zimmerhackl LB. Clinical course and the role of Shiga toxin producing Escherichia coli infection in the hemolytic-uremic syndrome in pediatric patients, 1997 2000, in Germany and Austria: a prospective study. J Infect Dis 2002; 186:493 500. 7. Tozzi AE, Caprioli A, Minelli F, et al. Shiga toxin producing Escherichia coli infections associated with hemolytic uremic syndrome, Italy, 1988 2000. Emerg Infect Dis 2003; 9:106 8. 8. Werber D, Fruth A, Liesegang A, et al. A multistate outbreak of Shiga toxin producing Escherichia coli O26:H11 infections in Germany, detected by molecular subtyping surveillance. J Infect Dis 2002; 186: 419 22. 9. Sonntag A, Prager R, Bielaszewska M, et al. Phenotypic and genotypic analyses of enterohemorrhagic Escherichia coli O145 strains from patients in Germany. J Clin Microbiol 2004; 42:954 62. 10. Zhang WL, Bielaszewska M, Liesegang A, et al. Molecular characteristics and epidemiological significance of Shiga toxin producing Escherichia coli O26 strains. J Clin Microbiol 2000; 38:2134 40. 11. Richardson SE, Karmali MA, Becker LE, Smith CR. The histopathology of the hemolytic uremic syndrome associated with verocytotoxin-producing Escherichia coli infections. Hum Pathol 1988; 19:1102 8. 12. Karch H. The role of virulence factors in enterohemorrhagic Escherichia coli (EHEC) associated hemolytic-uremic syndrome. Semin Thromb Hemost 2001; 27:207 13. 13. Schmidt H, Beutin L, Karch H. Molecular analysis of the plasmidencoded hemolysin of Escherichia coli O157:H7 strain EDL 933. Infect Immun 1995; 63:1055 61. 14. Yu J, Kaper JB. Cloning and characterization of the eae gene of enterohaemorrhagic Escherichia coli O157:H7. Mol Microbiol 1992;6: 411 7. 15. Donnenberg MS, Tzipori S, McKee ML, O Brien AD, Alroy J, Kaper JB. The role of the eae gene of enterohemorrhagic Escherichia coli in intimate attachment in vitro and in a porcine model. J Clin Invest 1993; 92:1418 24. 16. Jerse AE, Yu J, Tall BD, Kaper JB. A genetic locus of enteropathogenic Escherichia coli necessary for the production of attaching and effacing lesions on tissue culture cells. Proc Natl Acad Sci U S A 1990; 87: 7839 43. 17. Nataro JP, Kaper JB. Diarrheagenic Escherichia coli. Clin Microbiol Rev 1998; 11:142 201. 18. Dobrindt U, Hochhut B, Hentschel U, Hacker J. Genomic islands in pathogenic and environmental microorganisms. Nat Rev Microbiol 2004; 2:414 24. 19. Herold S, Karch H, Schmidt H. Shiga toxin encoding bacteriophages: genomes in motion. Int J Med Microbiol 2004; 294:115 21. 20. O Brien AD, Newland JW, Miller SF, Holmes RK, Smith HW, Formal SB. Shiga-like toxin-converting phages from Escherichia coli strains that cause hemorrhagic colitis or infantile diarrhea. Science 1984; 226: 694 6. 21. Matsushiro A, Sato K, Miyamoto H, Yamamura T, Honda T. Induction of prophages of enterohemorrhagic Escherichia coli O157:H7 with norfloxacin. J Bacteriol 1999; 181:2257 60. 22. Shaikh N, Tarr PI. Escherichia coli O157:H7 Shiga toxin encoding bacteriophages: integrations, excisions, truncations, and evolutionary implications. J Bacteriol 2003; 185:3596 605. 23. Misselwitz J, Karch H, Bielazewska M, et al. Cluster of hemolyticuremic syndrome caused by Shiga toxin producing Escherichia coli O26:H11. Pediatr Infect Dis J 2003; 22:349 54. 24. Wong CS, Jelacic S, Habeeb RL, Watkins SL, Tarr PI. The risk of the hemolytic-uremic syndrome after antibiotic treatment of Escherichia coli O157:H7 infections. N Engl J Med 2000; 342:1930 6. 25. Schmidt H, Geitz C, Tarr PI, Frosch M, Karch H. Non-O157:H7 pathogenic Shiga toxin producing Escherichia coli: phenotypic and genetic profiling of virulence traits and evidence for clonality. J Infect Dis 1999; 179:115 23. 26. Schmidt H, Rüssmann H, Schwarzkopf A, Aleksic S, Heesemann J, Karch H. Prevalence of attaching and effacing Escherichia coli in stool samples from patients and controls. Zentralbl Bakteriol 1994; 281: 201 13. 27. Cebula TA, Payne WL, Feng P. Simultaneous identification of strains of Escherichia coli serotype O157:H7 and their Shiga-like toxin type by mismatch amplification mutation assay-multiplex PCR. J Clin Microbiol 1995; 33:248 50. 28. Bitzan M, Moebius E, Ludwig K, Müller-Wiefel DE, Heesemann J, Karch H. High incidence of serum antibodies to Escherichia coli O157 lipopolysaccharide in children with hemolytic-uremic syndrome. J Pediatr 1991; 119:380 5. 29. Prager R, Strutz U, Fruth A, Tschäpe H. Subtyping of pathogenic Escherichia coli strains using flagellar (H)-antigens: serotyping versus flic polymorphisms. Int J Med Microbiol 2003; 292:477 86. 30. Karmali MA, Petric M, Bielaszewska M. Evaluation of a microplate latex agglutination method (Verotox-F assay) for detecting and characterizing verotoxins (Shiga toxins) in Escherichia coli. J Clin Microbiol 1999; 37:396 9. 31. Beutin L, Montenegro MA, Orskov I, et al. Close association of verotoxin (Shiga-like toxin) production with enterohemolysin production in strains of Escherichia coli. J Clin Microbiol 1989; 27:2559 64. 32. Friedrich AW, Borell J, Bielaszewska M, Fruth A, Tschäpe H, Karch H. Shiga toxin 1c-producing Escherichia coli strains: phenotypic and genetic characterization and association with human disease. J Clin Microbiol 2003; 41:2448 53. 33. Zhang W, Bielaszewska M, Kuczius T, Karch H. Identification, characterization, and distribution of a Shiga toxin 1 gene variant (stx (1c)) in Escherichia coli strains isolated from humans. J Clin Microbiol 2002; 40:1441 6. 34. Oswald E, Schmidt H, Morabito S, Karch H, Marches O, Caprioli A. Typing of intimin genes in human and animal enterohemorrhagic and enteropathogenic Escherichia coli: characterization of a new intimin variant. Infect Immun 2000; 68:64 71. 35. Zhang WL, Kohler B, Oswald E, et al. Genetic diversity of intimin genes of attaching and effacing Escherichia coli strains. J Clin Microbiol 2002; 40:4486 92. 36. Brunder W, Khan AS, Hacker J, Karch H. Novel type of fimbriae encoded by the large plasmid of sorbitol-fermenting enterohemorrhagic Escherichia coli O157:H( ). Infect Immun 2001; 69:4447 57. 37. Friedrich AW, Nierhoff KV, Bielaszewska M, Mellmann A, Karch H. Phylogeny, clinical associations, and diagnostic utility of the pilin subunit gene (sfpa) of sorbitol-fermenting, enterohemorrhagic Escherichia coli O157:H. J Clin Microbiol 2004; 42:4697 701. 38. Janka A, Bielaszewska M, Dobrindt U, Greune L, Schmidt MA, Karch H. Cytolethal distending toxin gene cluster in enterohemorrhagic Escherichia coli O157:H- and O157:H7: characterization and evolutionary considerations. Infect Immun 2003; 71:3634 8. 39. O Brien AO, Lively TA, Chen ME, Rothman SW, Formal SB. Escherichia coli O157:H7 strains associated with haemorrhagic colitis in the United States produce a Shigella dysenteriae 1 (Shiga) like cytotoxin. Lancet 1983; 1:702. 40. Sonntag AK, Prager R, Bielaszewska M, et al. Phenotypic and genotypic analyses of enterohemorrhagic Escherichia coli O145 strains from patients in Germany. J Clin Microbiol 2004; 42:954 62. 41. Prager R, Liesegang A, Voigt W, Rabsch W, Fruth A, Tschäpe H. Clonal diversity of Shiga toxin producing Escherichia coli O103:H2/H(-) in Germany. Infect Genet Evol 2002; 1:265 75. 42. Tenover FC, Arbeit RD, Goering RV, et al. Interpreting chromosomal DNA restriction patterns produced by pulsed-field gel electrophoresis: criteria for bacterial strain typing. J Clin Microbiol 1995; 33:2233 9. 43. Beutin L, Marches O, Bettelheim KA, et al. HEp-2 cell adherence, actin aggregation, and intimin types of attaching and effacing Escherichia EHEC Evolution In Vivo CID 2005:41 (15 September) 791

coli strains isolated from healthy infants in Germany and Australia. Infect Immun 2003; 71:3995 4002. 44. Tarr PI, Neill MA, Clausen CR, Watkins SL, Christie DL, Hickman RO. Escherichia coli O157:H7 and the hemolytic uremic syndrome: importance of early cultures in establishing the etiology. J Infect Dis 1990; 162:553 6. 45. Wagner PL, Acheson DW, Waldor MK. Human neutrophils and their products induce Shiga toxin production by enterohemorrhagic Escherichia coli. Infect Immun 2001; 69:1934 7. 46. Karch H, Meyer T, Rüssmann H, Heesemann J. Frequent loss of Shigalike toxin genes in clinical isolates of Escherichia coli upon subcultivation. Infect Immun 1992; 60:3464 7. 47. Feng P, Dey M, Abe A, Takeda T. Isogenic strain of Escherichia coli O157:H7 that has lost both Shiga toxin 1 and 2 genes. Clin Diagn Lab Immunol 2001; 8:711 7. 48. Falush D, Kraft C, Taylor NS, et al. Recombination and mutation during long-term gastric colonization by Helicobacter pylori: estimates of clock rates, recombination size, and minimal age. Proc Natl Acad Sci U S A 2001; 98:15056 61. 49. Robins-Browne RM, Bordun A, Tauschek M, et al. Escherichia coli and community-acquired gastroenteritis, Melbourne, Australia. Emerg Infect Dis 2004; 10:1797 805. 50. Gartner JF, Schmidt MA. Comparative analysis of locus of enterocyte effacement pathogenicity islands of atypical enteropathogenic Escherichia coli. Infect Immun 2004; 72:6722 8. 792 CID 2005:41 (15 September) Mellmann and Bielaszewska et al.