Prevalence of childhood diarrhoea-associated Escherichia coli in Thailand

Diarrhoea caused by Escherichia coli infection is one of the major health problems for children in many developing countries and travellers to those countries (Adachi et al., 2001; Ogata et al., 2002; Robins-Browne & Hartland, 2002). Diarrhoeagenic E. coli are recognized as five major pathotypes: enterotoxigenic (ETEC), enteropathogenic (EPEC), enteroinvasive (EIEC), Shiga-like toxin-producing (STEC) or enterohaemorrhagic (EHEC) and enteroaggregative (EAEC) E. coli (Nataro & Kaper, 1998). Diffuse-adherent E. coli (DAEC), also known as diarrhoea-associated haemolytic E. coli (DHEC), and cytolethal distending toxin-producing E. coli (CDT-EC) have also been described recently as diarrhoeagenic (Clarke, 2001; Nataro & Kaper, 1998). Their epidemiology and diarrhoeagenic potential are, however, not yet clear (Scaletsky et al., 2002).

Standard methods currently used in the identification of major E. coli pathotypes are based on distinct sets of virulence markers, such as toxins [heat-labile and heat-stable enterotoxins LTh and STh or STp (ETEC) (Kuhnert et al., 2000) and Shiga-like toxins SLT1 and SLT2 (STEC) (Nataro & Kaper, 1998)], adhesins [intimin (Cravioto et al., 1996) and EPEC adherence factor (EAF) (Donnenberg et al., 1997) (EPEC)] and invasins (EIEC) (Robins-Browne, 1987), and their cell-adherence characteristics (Clarke, 2001; Nataro & Kaper, 1998). None of these methods is available for use in routine clinical laboratories in developing countries such as Thailand. Although surveillance of diarrhoeagenic E. coli using standard identification methods had been carried out sporadically in Thailand in the past until 1986 (Chatkaeomorakot et al., 1987; Sunthadvanich et al., 1990), these surveillance studies were not designed to detect all major pathotypes.

In this study, we have categorized E. coli isolates collected over a 5-year period (19962000) from Thai children with acute diarrhoea by examining their virulence markers and cell-adherence patterns. The isolates were further examined for their O antigens. The prevalence of diarrhoeagenic E. coli pathotypes, with their age distribution and the isolation rate per year, is presented.

Clinical samples and bacterial strains.
Rectal swabs (one per patient) were collected from 2100 Thai children less than 12 years of age with acute diarrhoea who attended 15 different hospitals across Thailand between 1 January 1996 and 31 December 2000.

Rectal swab samples collected in Cary-Blair transport medium were inoculated directly onto MacConkey agar, sorbitol MacConkey agar, thiosulfate/citrate/bile salt/sucrose agar, SalmonellaShigella agar, xylose lysine desoxycholate agar, selenite broth and alkaline peptone water for culture overnight at 37 °C. Sorbitol non-fermenting colonies were again tested with O157 : H7 antisera. One to three colonies of each sorbitol non-fermenting, lactose-fermenting and lactose non-fermenting isolate with typical E. coli morphology were initially selected and examined further biochemically following Edwards and Ewing's identification methods (Ewing, 1986). E. coli isolates that had been biochemically confirmed at the hospitals concerned were submitted to our institute (National Institute of Health, Thailand), where they were stored on Dorset egg yolk agar (Nissui Pharmaceutical Inc.) until used. Samples derived from mixed infections with salmonellae, shigellae and vibrios were excluded from our study. A total of 2629 isolates thus obtained as single pathogens were included in this study.

E. coli strains 1298 (invE⁺), EDL931 (stx1/2⁺), 682 (eltIA⁺), 825 (stIA⁺, STp), 1296 (stIA⁺, STh) and 1228 (eaeA⁺, bfpA⁺, EAF⁺), which were kindly provided by the National Institute of Public Health, Tokyo, Japan, were used as positive control strains and JM109 was used as a negative control strain for virulence markers. E. coli strains 28-3, 1228 and JM109 were respectively used as controls for aggregative adherence (AA), localized adherence (LA) and non-adherence on HeLa cells.

Identification of virulence markers by multiplex PCR.
E. coli isolates were subjected to two PCR assays using two multiplex primer sets (Table 1). The first set was used to identify ETEC, EIEC and STEC (Itoh et al., 1992), and the second set was used to detect EPEC (Franke et al., 1994; Sueyoshi et al., 1996). Boiled lysates from overnight-grown bacterial colonies on LB plates were used as PCR templates. PCR assays were carried out in 25 µl reaction mixtures consisting of 1 µl template DNA, 20 mM Tris/HCl, pH 8.4, 50 mM KCl, 1.5 mM MgCl₂, 0.25 mM dNTPs (New England Biolabs), 0.1 µM of each primer (0.2 µM for stIA primers) and 1 U Taq DNA polymerase (Gibco). The reaction mixtures were run in a thermal cycler (model 9700; Perkin-Elmer) with the following cycling profile: 94 °C for 5 min, 25 cycles of denaturation at 94 °C for 1 min, annealing at 48 °C for 1.5 min and primer extension at 72 °C for 2 min and a final extension at 72 °C for 5 min. The annealing temperature for EPEC PCR was 50 °C. Amplified products were resolved by 2 % agarose gel electrophoresis and visualized under UV transillumination after ethidium bromide staining. DNA templates from positive and negative control strains for virulence markers and a minus-template sample were included in each PCR.

Table 1. Multiplex PCR primer sets used to identify recognized virulence markers of ETEC, EIEC, STEC and EPEC LTh, Heat-labile enterotoxin; STh/STp, heat-stable enterotoxin (human/pig alleles); SLT1/2, Shiga-like toxins 1 and 2; BFP, bundle-forming pili. Primers for stIA can detect both STh and STp.

HeLa cell-adherence assay.
Cell-adherence patterns of E. coli isolates were examined using monolayers of HeLa cells (ATCC CCL-2) as described by Nataro et al. (1987). HeLa cells were grown to 7080 % confluence on circular coverslips (13 mm diameter) in 24-well tissue culture plates (Nalge Nunc) with DMEM supplemented with 10 % fetal bovine serum in a 37 °C incubator with 5 % CO₂. After washing three times with PBS, fresh DMEM containing 1 % methyl α-D-mannoside was added to the wells to inhibit type 1 fimbriae-mediated cell adherence. Cells were infected with bacteria (approx. 2 x 10⁶ cells) that had been grown statically in LB broth at 37 °C for 18 h, to give an m.o.i. of 1 : 20. After 3 h incubation, cells were washed three times with PBS, fixed with 100 % methanol for 10 min and stained with 10 % Giemsa stain for 30 min. Cell-adherence patterns were determined using a light microscope (Nikon) following the criteria described by Nataro et al. (1987). Each assay was done in duplicate using positive and negative control strains.

Determination of diarrhoea-associated E. coli pathotypes.
E. coli isolates were categorized using the following criteria: isolates positive for eltIA, stIA or both as ETEC; isolates positive for invE as EIEC; isolates positive for stx1/2, eaeA or both as STEC; isolates negative for stx1/2 and positive for eaeA as EPEC; EPEC with LA pattern on HeLa cells as typical EPEC; EPEC with non-LA pattern on HeLa cells as atypical EPEC; isolates negative for all tested virulence markers but positive for AA pattern on HeLa cells as EAEC; and isolates negative for both tested virulence markers and AA pattern on HeLa cells as non-diarrhoeagenic E. coli. In this study, we used the AA phenotype as the marker for EAEC, because the known virulence-associated genes of EAEC, which have been widely used as genotypic markers in identifying EAEC, are also present in non-EAEC strains (Elias et al., 2002).

Serogrouping.
O antigen determination was done by slide agglutination test using a heated suspension (100 °C for 1 h) of bacterial cells and eight polyvalent and 43 monovalent O antisera (Denka Seiken Co.) that are targeted against common O-serogroups of EPEC, ETEC, EIEC and STEC. The following 43 serogroups were determined: O1, O6, O8, O15, O18, O20, O25, O26, O27, O28ac, O29, O44, O55, O63, O78, O86a, O111, O112ac, O114, O115, O119, O124, O125, O126, O127a, O128, O136, O142, O143, O144, O146, O148, O151, O152, O153, O157, O158, O159, O164, O166, O167, O168 and O169. Results were confirmed by the test-tube agglutination test (Ewing, 1986).

Statistical analysis.
The S² and Fisher's exact tests were used to report the significance of differences between variables.

E. coli isolates (n = 2629) were initially examined for their virulence markers by two multiplex PCR assays. Strains negative for all tested virulence markers (n = 2453) were further examined for their HeLa cell-adherence patterns. Additionally, 85 EPEC strains were included in the HeLa cell-adherence assay to differentiate typical EPEC from atypical EPEC. We did not investigate DAEC or CDT-EC in this study as their role in diarrhoea is still not clear (Scaletsky et al., 2002).

Prevalence of diarrhoea-associated E. coli
Tables 2 and 3 show the characteristics and isolation frequency of diarrhoea-associated E. coli during the 5-year study period. The mean isolation rates per year were 16.9 % (range 13.322.6 %) for total diarrhoeagenic E. coli, 10.2 % (812.5 %) for EAEC, 3.2 % (08 %) for EPEC, 3.0 % (25.4 %) for ETEC, 0.5 % (01 %) for EIEC and 0.04 % (00.1 %) for STEC. EAEC, EPEC, ETEC, EIEC and STEC respectively accounted for 60.4 % (269/445), 19.1 % (85/445), 17.5 % (78/445), 2.7 % (12/445) and 0.2 % (1/445) of diarrhoeagenic E. coli.

Table 2. Determination of pathotypes among 2629 childhood diarrhoea-associated E. coli isolates on the basis of virulence marker profiles and HeLa cell-adherence patterns Virulence markers were examined by multiplex PCR.

Table 3. Prevalence of childhood diarrhoea-associated E. coli in Thailand, 19962000

The relatively high prevalence of EAEC in this study (10.2 %) and in previous studies from Calcutta (9 %; Dutta et al., 1999), northern India (12.3 % in acute diarrhoea and 34.5 % in persistent diarrhoea; Bhan et al., 1989) and southern Israel (25.9 %; Porat et al., 1998) is consistent with reports that EAEC is the pathotype responsible for persistent diarrhoea among international travellers who visit developing countries (Adachi et al., 2001; Jiang et al., 2002; Vargas et al., 1998).

By the definition adopted at the Second International Symposium on EPEC in Sao Paulo in 1995, typical EPEC possess the bundle-forming pili (BFP)-producing EAF plasmid with the LA pattern on cultured cells, whereas atypical EPEC lack this plasmid and the LA pattern (Nataro & Kaper, 1998; Trabulsi et al., 2002). We used the LA pattern on HeLa cells as the only marker for classifying typical EPEC, because the methods commonly used to detect the BFP-producing EAF plasmid, such as hybridization with EAF and bfpA probes (Nataro & Kaper, 1998) and PCR amplification of the EAF region and bfpA gene (Franke et al., 1994), can be misleading, as some LA-expressing EPEC strains, which are positive for either EAF or bfpA, may not carry the true BFP-producing EAF plasmid (Nataro & Kaper, 1998; Trabulsi et al., 2002). Accordingly, 71.8 % (61/85) of our EPEC strains were non-adherent to HeLa cells and therefore classified as atypical EPEC, whereas the remaining 24 EPEC strains exhibited the classical LA pattern on HeLa cells (Table 4). In addition, we also observed that 50 % (12/24) of LA-positive EPEC strains were positive for eaeA and bfpA but negative for EAF and that 29 % (7/24) of LA-positive EPEC strains were positive for eaeA and EAF but negative for bfpA. Only 20.8 % (5/24) were positive for eaeA, bfpA and EAF. Some atypical EPEC strains have been shown to express the intimin-mediated LA-like pattern (Pelayo et al., 1999; Trabulsi et al., 2002), but all our atypical EPEC strains were non-adherent. Although geographical specificities may exist, the high presentation of atypical EPEC in this study and in previous reports from Brazil (Gomes et al., 1989) and other industrialized countries (Nataro & Kaper, 1998) underscores the emergence of atypical EPEC strains worldwide.

Table 4. Distribution of O serogroups in E. coli isolates of this study Values in parentheses are percentages.

Among ETEC isolates, 53.8 % (42/78) and 38.5 % (30/78) were respectively positive for stIA and eltIA and 7.7 % (6/78) were positive for both eltIA and stIA. Taken together, 61.5 % (48/78) of ETEC were stIA positive. ETEC are responsible for 2040 % of diarrhoeal cases amongst travellers and ST-producing ETEC, in particular, are known to be associated with the majority of endemic cases (Nataro & Kaper, 1998). Although the occurrence of ETEC in Thai children had been rather steady during the last 5 years, most strains (61.5 %) in this study were stIA positive, indicating that there is a persistent risk of ETEC-associated outbreak in Thailand. On the other hand, similar to reports from south-western Nigeria (Okeke et al., 2000), southern Israel (Porat et al., 1998) and Bangladesh (Albert et al., 1995), EIEC and STEC seem to be only minor diarrhoeagenic pathogens for children in Thailand, since their occurrence was negligible among diarrhoea-associated E. coli in three studies, including this one, from Thailand (Chatkaeomorakot et al., 1987; Sunthadvanich et al., 1990).

In two previous studies from Thailand in 1985 and 1986 (Chatkaeomorakot et al., 1987; Sunthadvanich et al., 1990), ETEC (6 and 7 %, respectively), EIEC (< 1 and 2 %), STEC (0 and 0 %) and EAF-positive EPEC (4 and 6 %), as determined by probe hybridization, colony hybridization and HeLa adherence-assay, were recovered from 393 children in 16 district hospitals (Sunthadvanich et al., 1990) and 278 children in the Children's Hospital in Bangkok (Chatkaeomorakot et al., 1987). Compared with these data, the prevalence of ETEC (3 %) and EAF-positive EPEC (0.5 %) in this study was significantly lower (P < 0.002 and P < 0.0001, respectively). Nonetheless, the overall prevalence of major diarrhoeagenic E. coli pathotypes in Thailand did not change much during the 5-year study period.

Age distribution of diarrhoea-associated E. coli
The prevalence of diarrhoeagenic E. coli in four different age groups among 905 children whose ages were recorded was 25 % (48/192) in those aged 05 months, 24.8 % (64/258) in those aged 611 months, 12.1 % (28/232) in those aged 12 years and 16.1 % (36/223) in those aged 212 years (Fig. 1). The isolation rate of diarrhoeagenic E. coli was significantly higher in children less than 1 year of age, compared with children older than 1 year (P < 0.0001). The prevalence of EAEC was 19.3 % (37/192) in 05 months, 18.2 % (47/258) in 611 months, 9.1 % (21/232) in 12 years and 8.1 % (18/223) in 212 years. EAEC were significantly more common in children less than 1 year old, compared with the older children (P < 0.0001). EPEC were isolated from 3.1 % (6/192), 4.3 % (11/258), 1.7 % (4/232) and 2.2 % (5/223), respectively, of children in age groups 05 months, 611 months, 12 years and 212 years, whereas ETEC were isolated from 2.6 % (5/192), 2.3 % (6/258), 1.3 % (3/232) and 5 % (11/223) of children in the corresponding age groups. ETEC were more common in children older than 2 years, compared with the younger children (P < 0.05). We did not find any significant difference in the age distribution of EPEC. EIEC were isolated from only two children in age group 212 years. STEC were not identified in children of known age.

(24K): Fig. 1. Isolation frequencies of EAEC (filled bars), EPEC (hatched bars), ETEC (shaded bars) and EIEC (open bars) among diarrhoea-associated E. coli from 905 Thai children with diarrhoea. mo, Months; yr, years.

Serogroups
With the 43 monovalent antisera used in this study, 38 % (169/445) of diarrhoeagenic E. coli and 23.9 % (522/2184) of non-diarrhoeagenic E. coli were typable for their O antigens. In addition, 55.1 % (43/78), 66.7 % (8/12), 100 % (1/1), 45.9 % (39/85) and 29 % (78/269) of ETEC, EIEC, STEC, EPEC and EAEC, respectively, were O-antigen typable and they were distributed respectively into 14, 4, 1, 14 and 18 different serogroups. O antigen-typable non-diarrhoeagenic E. coli were distributed into 32 serogroups. Twenty-seven serogroups were observed in more than one category, accounting for 62.8 % of total serogroups tested: 13 serogroups (O6, O8, O15, O18, O25, 86a, O119, O126, O127a, O128, O146, O159 and O166) were identified in three or more different categories, including non-diarrhoeagenic E. coli. There were only four serogroups that were exclusively associated with a single pathotype: O20 (ETEC), O124, O152 and O164 (all EIEC).

That about 71 % of EAEC, 54 % of EPEC, 45 % of ETEC and 33 % of EIEC strains were non-typable while 24 % of non-diarrhoeagenic E. coli strains were typable indicates that a significant number of isolates would have been misidentified by serogroup-based diagnosis. Besides, although only about 9 % of serogroups were identified exclusively in single pathotypes, more than 60 % of serogroups tested were not restricted to any pathotype, signifying the unrestricted nature of serogroups among different pathotypes.

Taken together, EAEC were the most common pathotype in every age group examined and in every year during the 5-year study period. Inasmuch as this is the first study to report EAEC isolates from Thailand, further characterization of these strains is required in order to understand better their role in diarrhoeal diseases in Thailand. The findings in this study will be of great importance in implementing management guidelines for E. coli-associated diarrhoea in Thailand. Given that Thailand is one of the popular tourist destinations in Asia, hosting 78 million international travellers every year, the results presented here would also have a significant impact on the management of E. coli-associated acute and persistent diarrhoea among travellers.

In conclusion, this study highlights that O antigen examination alone is of little value in epidemiological studies related to diarrhoeagenic E. coli and that the comprehensive surveillance of diarrhoeagenic E. coli in developing countries remains an important part of preventative and control measures in reducing the overall incidence of diarrhoeal diseases around the world.

We are grateful to the technicians and physicians who collected the samples for this study. Funding for this project was provided in part by NIH Thailand and grant no. 00L01411 from the Japan Society for the Promotion of Science (JSPS) and the Ministry of Education, Culture, Sports, Science and Technology, Japan. O.-A. R. is a fellow of the RONPAKU (Dissertation PhD) program (JSPS).