Association of atypical enteropathogenic Escherichia coli (EPEC) with prolonged diarrhoea

Enteropathogenic Escherichia coli (EPEC) is one of the leading causes of diarrhoeal morbidity and mortality among children in developing countries (Clarke et al., 2002). Decades ago EPEC was also frequently seen among children in industrialised countries (Levine & Edelman, 1984). The EPEC diagnosis was at that time based on the identification of recognized EPEC serogroups. Then, for unknown reasons, a decline in the incidence of EPEC was observed in the industrialised world, and most routine laboratories stopped testing for EPEC. Only after the central mechanism of EPEC pathogenesis was discovered and new diagnostic methods were developed, did interest in this pathogen again gradually increase in the developed world.

The central mechanism of EPEC pathogenesis is a lesion called attaching and effacing (A/E), which is characterized by intimate adherence of the bacteria to the intestinal epithelium (Nougayrède et al., 2003). The eae gene located in the pathogenicity island locus of enterocyte effacement (LEE) and the bfpA gene located on a plasmid, called the EPEC adherence factor (EAF), have been used to classify this group of bacteria into typical and atypical strains (Kaper, 1996). E. coli strains of the A/E genotype (eae⁺) harbouring the EAF plasmid (bfpA⁺), are classified as typical EPEC'. Most such strains belong to certain O : H serotypes (Trabulsi et al., 2002). Strains of the A/E genotype, which do not posses the EAF plasmid (bfpA), are classified as atypical EPEC'. eae-positive E. coli strains harbouring Shiga toxin genes (stx1 and/or stx2) are classified as enterohaemorrhagic E. coli.

Typical EPEC is well recognized as a cause of gastroenteritis in infants (Levine & Edelman, 1984; Nataro & Kaper, 1998). The role of atypical EPEC in childhood diarrhoea is still controversial. Most case control studies have not been able to demonstrate any significant association between atypical EPEC and diarrhoea (Echeverria et al., 1991; Gomes et al., 1991; Morelli et al., 1994; Forestier et al., 1996; Knutton et al., 2001; Scaletsky et al., 2002a, b; Nunes et al., 2003; Regua-Mangia et al., 2004). However, the majority of these studies demonstrated a tendency for the group of subjects with diarrhoea to have a higher prevalence of atypical EPEC than healthy controls, and it has been argued that atypical EPEC are probably pathogenic, since such strains are capable of colonizing the intestinal mucosa and produce A/E lesions (Knutton et al., 2001). In volunteer studies some of the subjects who were given EAF plasmid-cured (Levine et al., 1985) or bfpA-mutated (Bieber et al., 1998) EPEC developed diarrhoea, although to a lesser extent than those who were given the original EPEC strains. There have also been reports where atypical EPEC was significantly associated with endemic diarrhoea (Scaletsky et al., 1999; Vieira et al., 2001; Dulguer et al., 2003) or was the cause of outbreaks (Viljanen et al., 1990; Hedberg et al., 1997; Yatsuyanagi et al., 2002; Jenkins et al., 2003).

Recently, we described a high prevalence of EPEC among Norwegian children with diarrhoea (Afset et al., 2003). The majority of EPEC isolates in that study were classified as atypical, and most of them (35/43 isolates) did not belong to EPEC serogroups (as defined by WHO, 1987). To investigate the prevalence of EPEC and its possible role in causing diarrhoea among Norwegian children, we conducted a case control study among children < 5 years old.

Patients and controls.
The study was conducted during the period March 2002 to January 2003 in the county of Sør-Trøndelag, Norway, and was approved by the Regional Committee for Medical Research Ethics, the Norwegian Data Inspectorate, and Norwegian Social Science Data Services. Cases were defined as subjects < 5 years of age with suspected gastroenteritis from whom the laboratory received a stool specimen. Included were outpatients and patients admitted to the hospital paediatric observation ward. Patients were not included if there was no growth in the stool culture, making EPEC analysis impossible. Consent for participation in the study was requested from the parents of the patients.

Healthy controls were recruited through Maternal and Child Health Centres in Trondheim, after informed consent from parents. They were matched to patients with respect to sex, time of the year of specimen collection (quarter) and age (12 months, 1324 months and 2560 months). Subjects were not included as controls if they had had diarrhoea within the last 4 weeks, and were excluded if there was no growth in the stool culture.

Subjects in whom EPEC was identified were asked for follow-up specimens. The first of these specimens was taken after 1 month, and specimens were collected at monthly intervals until negative or until the subject had been followed for 6 months.

EPEC identification.
For EPEC isolation, all stool specimens from patients and controls were cultured on MacConkey agar (Difco). PCR (eae, stx) was primarily performed on 23 cm streaks from bacterial growth on the agar plate. If the streak was PCR-positive, ten colonies were subcultured from the same primary culture plate (stored at 4 °C) and retested. Isolates with the eae genotype identified as E. coli were then stored (one isolate per subject) at 70 °C until they were analysed for the presence of the bfpA gene. Biochemical identification of E. coli was done by API 10S/20E (bioMérieux).

For PCR testing, streaks from bacterial growth were suspended in 4 ml physiological saline, and further diluted 1 : 100. From subculture plates one colony was suspended in 1 ml physiological saline. Lysis of bacterial cells was done by heating at 94 °C for 15 min. Amplification of the eae and bfpA genes was performed in a total volume of 50 µl, containing 50 µM (each) of dATP, dCTP, dGTP and dTTP, 0.5 µM of each primer, 10x PCR buffer (Roche), 1.5 mM MgCl₂, 1 U AmpliTaq Gold polymerase (Roche), 0.025 % BSA and 2 µl bacterial lysate as template. Detection of stx1 and stx2 genes was carried out by multiplex PCR with conditions as above, except the use of 100 µM (each) dNTP and no BSA. The reaction mixture was heated to 94 °C prior to amplification and held at 72 °C for 7 min before cooling to 4 °C. Primer sequences and cycling conditions are listed in Table 1. Amplification products were analysed by 2 % agarose gel electrophoresis and visualized by staining with ethidium bromide. Clinical isolates of E. coli O157 : H7 (eae, stx1, stx2) and E. coli B171 (bfpA) were used as positive controls.

Table 1. PCR for the detection of EPEC, enterohaemorrhagic E. coli (EHEC), enteroinvasive E. coli (EIEC), enterotoxigenic E. coli (ETEC) and enteroaggregative E. coli (EAEC) eae, Attaching and effacing-associated gene; stx, Shiga toxin gene; bfpA, bundle-forming pilus gene; ial, invasion associated locus of EIEC.

Bacterial growth on a primary MacConkey plate was also tested for EPEC serogroups with polyspecific O-antisera Anti-Coli I (O26, O44, O114, O125, O142, O158) and Anti-Coli II (O55, O86, O111, O119, O126, O127, O128) according to the manufacturer's instructions (Sifin). If the primary culture was positive, ten colonies were subcultured and retested. Positive strains that were identified as E. coli were examined using monospecific O : K antisera at the Norwegian Institute of Public Health, Oslo (Sifin and in-house antisera).

Follow-up specimens were analysed by PCR for the eae gene using a streak of bacterial growth from the culture plate.

Other pathogens.
Whereas stool specimens from control subjects were tested for EPEC only, patients specimens were routinely examined for a range of pathogens in addition to EPEC. Specimens from patients were cultured for Salmonella, Shigella, Campylobacter, Yersinia, Aeromonas and Plesiomonas using lysine sucrose-urea agar (Jühlin & Ericson, 1961), SSI Enteric Medium (Blom et al., 1999) and selenite broth (Difco). For the isolation of Campylobacter species, specimens were cultured on charcoal cefoperazone desoxycholate agar (Mast Diagnostics). The specimens were examined for adenovirus by enzyme immunoassay (DakoCytomation), PCR (Krokstad et al., 2003) and viral cell culture, and for rotavirus with enzyme immunoassay (DakoCytomation).

Stool specimens that contained eae-positive isolates were additionally tested for the presence of other diarrhoeagenic (enteroinvasive, enterotoxigenic, enteroaggregative) E. coli by PCR (Table 1). These specimens were also examined with enzyme immunoassay for astrovirus and norovirus (DakoCytomation), Giardia lamblia (Remel) and Cryptosporidium (Cellabs).

Questionnaire.
Information regarding demographic data, possible risk factors for gastrointestinal infection, medical history and duration of disease was collected in a questionnaire as well as from the physician's referral form (patients). Information about hospital admission and discharge diagnosis was collected from hospital records.

Statistical analyses.
Multiple logistic regression was used to study the potential association between EPEC and diarrhoea. The analyses were adjusted for matching factors (sex, age group and time of specimen collection), and were controlled for potential confounding from other risk factors. The data were also analysed using conditional logistic regression, with results similar to that of ordinary logistic regression analyses. The MannWhitney U-test was employed for testing of difference in number of eae-positive subcultures between patients and controls, whereas Fisher's exact test was used in the analysis of duration of EPEC carriage. P values < 0.05 were considered significant. Statistical analyses were performed using SPSS version 12.0 (SPSS) except in conditional logistic regression analyses where Stata version 8.0 (StataCorp 2003) was used.

During the period March 2002 to January 2003, 251 patients and 210 healthy controls < 5 years of age were included in the study. Demographic variables and time of specimen collection are presented in Table 2. The majority in the patient group (192/251 subjects, 76.5 %) were treated as outpatients, while 59 (23.5 %) were admitted to hospital. Of those admitted, 44 patients (17.5 % of the total) were discharged with a diagnosis of infectious gastroenteritis, whereas the remaining 15 patients had conditions other than infectious diarrhoea. Among these, respiratory tract infection (n = 8) and non-infectious gastrointestinal disease (n = 3) were most common, while four patients had unspecified symptoms or disease in other organ systems. It was possible to classify the duration of diarrhoea in 178 (70.9 %) of the 251 patients, as either short ( < 14 days) or prolonged (14 days) disease.

Table 2. Demographic variables and time of specimen collection in patients with diarrhoea (n = 251) and healthy controls (n = 210)

Completed questionnaires were returned from 192 (76.5 %) patients and 204 (97.1 %) controls. Patients and controls were compared with respect to potential risk factors for diarrhoea other than micro-organisms. A significant association was found for the factors: drinking water from private well [Odds ratio (OR) = 4.8, P = 0.003]; antibiotic treatment within the last 3 months (OR = 1.9, P = 0.03) and diarrhoeal illness among family members within the last 2 weeks (OR = 9.0, P < 0.001). An association was also seen in children < 1 year of age between diarrhoea and travel abroad within the last 6 months (OR = 3.3, P = 0.02). A table with risk factors for diarrhoea other than micro-organisms is available as supplementary data in JMM Online.

Potential enteropathogens in patients
A potential or established microbial pathogen was detected in 99 (39.4 %) of the 251 children with diarrhoea. Among these, a single pathogen was identified in 88 patients whereas more than one organism was found in 11 patients. Among all potential enteropathogens in children with diarrhoea, rotavirus and EPEC were the most commonly identified agents (Table 3).

Table 3. EPEC and other intestinal pathogens isolated from 251 Norwegian children with diarrhoea

EPEC
PCR from growth on the primary culture plate was positive for the eae gene in 45 (17.9 %) patients and in 27 (12.9 %) healthy controls. In 13 (7 patients and 6 controls) of these subjects, an eae-positive organism was not recovered after subculture of ten distinct colonies. Only subjects in whom eae-positive E. coli isolates were identified were recorded as EPEC-positive in the subsequent analyses. Thus, EPEC was identified in 38 (15.1 %) patients compared to 21 (10.0 %) of the healthy controls. Among the 38 patients with EPEC, one or two additional pathogens were identified in eight patients (Table 3).

Atypical EPEC (58/59 isolates, 98.3 %) constituted the majority of eae-positive isolates in this study. Only three of these isolates (all from patients) belonged to EPEC serogroups (Table 4). This finding is in accordance with that recently reported from the UK (Knutton et al., 2001) and Brazil (Dulguer et al., 2003), but not with the notion that such strains are rarely isolated from children with and without diarrhoea (Trabulsi et al., 2002). eae-positive E. coli strains not belonging to EPEC serogroups have been reported to constitute a heterogeneous group (Vieira et al., 2001; Trabulsi et al., 2002). However, most such strains have virulence profiles similar to atypical strains of EPEC serogroups (Vieira et al., 2001; Dulguer et al., 2003), and at least some of these strains are able to induce A/E lesions in human intestinal biopsies (Knutton et al., 2001; Vieira et al., 2001). We therefore chose to include all eae⁺, stx and bfpA E. coli strains in the group of atypical EPEC in this study, irrespective of the result of serogroup determination. The finding of only one isolate of typical EPEC in this study (Table 4) confirms our previous observation that typical EPEC is a rare cause of diarrhoea in Norwegian patients (Afset et al., 2003).

Table 4. Classification of EPEC isolates identified among 251 children with diarrhoea and 210 healthy controls < 5 years old

Atypical EPEC was highly prevalent both in patients with diarrhoea (14.7 %) and in controls (10.0 %). Whereas a high prevalence in symptomatic children was as expected (Afset et al., 2003), the carrier rate in healthy controls was considerably higher than predicted and almost twice the rates of atypical EPEC in children without diarrhoea recently reported by other authors (Knutton et al., 2001; Gomes et al., 2002; Dulguer et al., 2003).

Although atypical EPEC was more commonly diagnosed in patients than controls, the association with diarrhoea was not statistically significant when all subjects with and without diarrhoea were compared (OR = 1.4, P = 0.3, Table 5). Controlling for other risk factors did not change this result. Patients with combinations of atypical EPEC and other enteropathogens were included in the statistical analyses. This was done as EPEC is often found together with other diarrhoeal agents (Afset et al., 2003), and might have a propensity to cause symptoms through synergism with other enteropathogenic agents (Fagundes-Neto & Scaletsky, 2000).

Table 5. Rate of atypical EPEC in selected groups of Norwegian children with diarrhoea compared with healthy controls

Atypical EPEC was uncommon among children less than 1 year of age, and was identified in three each of patients and controls (Table 5). This finding is in accordance with that found in the same age group in the UK (Knutton et al., 2001), but very different from that reported from Guinea-Bissau where half the infants in a cohort followed from birth were infected with atypical EPEC before 4 months of age (Valentiner-Branth et al., 2003). It is possible that Norwegian infants to some extent are protected from EPEC infection by maternal immunity, breastfeeding and the fact that most children start attending kindergarten after 1 year of age. In this study only two of the children less than 1 year of age attended kindergarten. However, these factors are unlikely to completely explain the great difference between infants and older children, where the prevalence of atypical EPEC was more than 10 % (Table 5). A considerable proportion of the infants were not breastfed (32/121 infants), and many were exposed to potentially infected older siblings (see supplementary table in JMM Online). The reason for this difference in EPEC rate between infants and older children in Norway and other industrialised countries therefore remains unclear, as does the difference compared with developing countries.

The prevalence of atypical EPEC was not markedly different between the groups of patients and controls less than 2 years of age. In those aged 2560 months, however, there was a tendency that atypical EPEC was more frequently diagnosed among patients than among controls (OR = 2.2, P = 0.09, Table 5). Atypical EPEC was more common among girls (15.9 %) than among boys (10.5 %). This was due to a higher prevalence among girls with diarrhoea, whereas the prevalence among healthy girls was comparable with that of boys in both groups (Table 5). Although diagnosed throughout all seasons of the year, atypical EPEC was found most frequently in the late summer/early autumn period. Almost half (27 isolates, 46.6 %) of the 58 isolates were found during the 3-month period August to October. Atypical EPEC was uncommon among hospitalized patients, and was detected in only three (5.1 %) of the 59 patients who were admitted. This result, which is in accordance with our previous study of patients < 2 years old (Afset et al., 2003), supports the notion that atypical EPEC rarely causes gastroenteritis severe enough to necessitate hospitalization.

Disease duration
When patients with diarrhoea were compared with respect to disease duration, the prevalence of atypical EPEC was found to be higher in those with diarrhoea lasting 14 days or more (22.5 %) than in patients with disease of shorter duration (10.2 %, Table 5). Atypical EPEC was significantly associated with prolonged diarrhoea (OR = 2.1, P = 0.04) when the subgroup of patients with protracted disease was compared with healthy controls (Table 5). This difference was caused by a much higher prevalence in female patients (40.6 %) than controls (9.2 %, OR = 5.6, P = 0.001), whereas no difference was observed for males.

It has been argued that the association between EPEC and prolonged diarrhoea does not necessarily represent a causal link (Levine & Edelman, 1984; Vallance et al., 2003). Rather than being cause of the disease, EPEC colonization could be promoted by the intestinal epithelial changes seen in chronic diarrhoea. Studies in patients with inflammatory bowel disease do shed some light on this question (Schultsz et al., 1997; Darfeuille-Michaud et al., 1998). It was shown that adult patients with ulcerative colitis and Crohn's disease did not have an increased rate of eae⁺ E. coli compared to healthy controls. These studies therefore do not support the theory of EPEC colonization as a result of chronic intestinal disease.

It is well known that typical EPEC can cause protracted diarrhoea (Levine & Edelman, 1984; Donnenberg, 1995; Fagundes-Neto & Scaletsky, 2000), but it has been less clear whether atypical EPEC has a similar potential. To our knowledge, an association between atypical EPEC and prolonged diarrhoea has previously only been reported by Scaletsky and colleagues in a study of infants where half of the patients with atypical EPEC had persistent diarrhoea (Scaletsky et al., 1999). The present study does indicate a role of atypical EPEC in prolonged diarrhoea, although lack of data on disease duration in some patients, as well as other factors mentioned above preclude any firm conclusion. The finding of a possible sex-related difference in the association of atypical EPEC and prolonged diarrhoea is interesting, but difficult to explain biologically and needs confirmation in further studies.

Relative quantity of atypical EPEC in stool specimens
Patients and healthy controls with atypical EPEC were compared with respect to the relative quantity of eae-positive organisms in their stools. It was found that a pure culture (10/10 subcultures) of atypical EPEC was more common in patients (17/37, 45.9 %) than in controls (5/21, 23.8 %). However, atypical EPEC was present in a large proportion of subcultures in both groups. Therefore, the overall difference in relative quantity between patients and controls was not statistically significant (P = 0.3, data not shown). This finding is not in accordance with the observation for typical EPEC which is usually found to be present in a higher number in patients with diarrhoea than in healthy carriers (Levine & Edelman, 1984), and is not in favour of the theory of a causative association between atypical EPEC and diarrhoea. However, it is difficult to pick colonies from a culture plate in a random fashion, and the subjective element involved might have influenced the results.

Duration of EPEC carriage
A total of 50 of the 57 subjects with atypical EPEC provided at least one follow-up specimen. Among these, altogether 14 specimens were positive for the eae gene at the first follow-up, four of 20 (20 %) healthy controls and 10 of 30 (33.3 %) patients, respectively. However, the difference in interval from primary to follow-up specimen collection between patients (median 38 days) and controls (median 59 days) could explain this difference. Three subjects (5.8 %), one of whom was a healthy control, still had EPEC in their stool after 3 months.

In patients where it was possible to classify diarrhoea with respect to duration, colonization with atypical EPEC at 1 month follow-up was compared with disease duration. Although not significant, the proportion of patients still colonized after 1 month tended to increase with increasing length of diarrhoea. Among eight patients with diarrhoea < 14 days, one subject (12.5 %) was still colonized at 1 month follow-up, compared with five of 16 (31.3 %) and four of six patients (66.7 %) with diarrhoea lasting 1428 days and >28 days, respectively (P = 0.12).

Based on the high rate among healthy controls, it is obvious that a large proportion of atypical EPEC infections must represent asymptomatic carriage rather than symptomatic infection. This does not exclude the possibility that some atypical EPEC strains may have the potential to cause symptomatic disease in some instances. The association between atypical EPEC and prolonged diarrhoea in this study supports this view. Individual atypical EPEC strains may have additional virulence factors (Scaletsky et al., 2002c; Dulguer et al., 2003), and host factors might be involved, either through immunity or other mechanisms. An example of such a host factor could be polymorphism of the IL8 promoter gene, which has recently been associated with variable susceptibility to enteroaggregative E. coli diarrhoea (Jiang et al., 2003). Further studies are necessary to clarify the role of such factors.

Conclusions
Atypical EPEC was found to be slightly more prevalent in patients than controls, without any overall significant association with diarrhoea. However, a significant association was observed with diarrhoea duration of 14 days or more, a finding that may indicate a role of atypical EPEC in prolonged disease.

We thank the staff at Maternal and Child Health Centres in Trondheim for kind co-operation in the recruitment of healthy controls, Jørgen Lassen, Norwegian Institute of Public Health, for O : K serogrouping of E. coli isolates, and the technical staff at the laboratory for skilful technical assistance.