The strains we studied were collected from patients of low socioeconomic status in three cities: São Paulo and Ribeirão Preto, in São Paulo State, and Rio de Janeiro, in Rio de Janeiro State, Brazil. The São Paulo strains were collected from 505 diarrheic and 505 nondiarrheic children (1–4 years of age) who visited Hospital Infantil Menino Jesus (April 1989–March 1990) (7). These strains had been previously characterized for various traits (6); in the present study, we tested them for new gene sequences. The Rio de Janeiro strains were collected from 372 diarrheic and 74 nondiarrheic children <5 years of age at the Instituto de Puericultura e Pediatria Martagão Gesteira, a public hospital at the Federal University of Rio de Janeiro (January 1998–December 1999, and May–December 2001). Strains from Ribeirão Preto were derived from 294 diarrheic children (<9 years of age) and adults (18–52 years), including 42 adults with AIDS. Fecal samples from these patients were sent to the Regional Laboratory of Instituto Adolfo Lutz by Hospital Santa Lydia and different clinics in the vicinity (August 2000–June 2002). This study has been approved by the Universidade Federal de São Paulo, Escola Paulista de Medicina Ethical Committee for human experimentation.

In all studies, five lactose-fermenting isolates and one nonlactose-fermenting isolate of each morphologic type, present in each fecal sample, were biochemically characterized as E. coli. Other well-established bacterial enteropathogens (Salmonella spp., Shigella spp., Aeromonas spp, Campylobacter spp., and Yersinia enterocolitica) and rotavirus were also searched for by standard methods (8).

All E. coli isolates were tested by colony hybridization with cloned or amplified genetic sequences for enterotoxigenic E. coli, enteroinvasive E. coli, EPEC (eae and EAF probes), STEC (stx probes), and enteroaggregative E. coli, as previously described (6). The E. coli strains that were eae+ EAF– stx– were serotyped at the Instituto Adolfo Lutz (National Reference Center for E. coli Serotyping) by using antisera O1 to O173 and H1 to H56.

In São Paulo and Rio de Janeiro, the eae+ EAF– stx– E. coli strains of non-EPEC serogroups occurred in similar frequencies in diarrheic and nondiarrheic children: 32 (6.3%) compared with 27 (5.3%), and 19 (5.1%) compared with 4 (5.4%), respectively. In Ribeirão Preto, such strains were found in 17 (5.8%) patients: 13 from children (1 month–9 years of age) and 4 from adults with AIDS (27–52 years of age). A total of 99 strains (one from each patient) were selected for further analysis. These strains had diverse serotypes (Table 1); 25 (25.2%) strains were nonmotile, 3 were rough, and 47 (47.5%) did not react with the O antisera tested. Among the 49 O-typable strains, 29 serogroups and 35 serotypes were found. The most frequent serotype was O51:H40 (10.1%), which occurred in all three areas studied. Most of the other serotypes occurred in one or two strains.

All strains were tested for adherence to HeLa cells (3- and 6-hour assays) (9). Four of them promoted sporadic adherence, four were nonadherent, and one was cytodetaching. For 88 of the 90 adherent strains, the adherence patterns could only be determined in 6 hours. Seventy-two (80.0%) of the 90 strains had variations of the localized adherence (LA) pattern of typical EPEC, which is characterized by compact bacterial clusters (10). These variant patterns included the following: LA-like pattern, which showed loose bacterial clusters (11); a pattern that showed loose and compact clusters; and a pattern identical to LA, despite its detection in only 6 hours (LA6). Other less frequent patterns included the following: the diffuse adherence typical of diffusely adhering E. coli, the aggregative adherence typical of enteroaggregative E. coli (2), and an association of diffuse adherence and LA or of aggregative adherence and LA. These mixed patterns were retained when individual colonies were tested. The aggregative adherence/LA pattern (two strains) was only recognized in the 3-hour assays. The prevalence of the different patterns varied by area of study, but the variations of LA were the most prevalent in all (72.7%) (Figure 1).

The ability to promote attaching and effacing lesions was tested by the fluorescent actin staining test (FAS) (7) in 94 strains; the 5 nonadherent or cytodetaching strains were not tested. Seventy (74.4%) of the strains tested were positive: 43 (72.9%), 15 (65.2%), and 12 (70.2%) of the strains from São Paulo, Rio de Janeiro, and Ribeirão Preto, respectively. Moreover, four distinct segments of the LEE region were found in all strains studied, as detected by hybridization with specific LEE sequences (LEE A, B, C, and D) (12), which suggests that all bear a complete LEE region.

LEE insertion sites were detected by a combination of polymerase chain reaction (PCR) assays with primers for the selC junctions and for conserved sequences of selC and pheU (12,13). LEE was inserted in selC in 46 strains: 24 (40.7%), 13 (56.6%), and 9 (53.0%) strains from São Paulo, Rio de Janeiro, and Ribeirão Preto, respectively. In addition, LEE was probably inserted in pheU in 29 (49.1%) and 3 (13.0%) of the São Paulo and Rio de Janeiro strains, respectively. In 13 strains, LEE is probably inserted in another site, since both loci were intact. The LEE insertion site was undetermined in eight strains because both selC and pheU were disrupted, and the primers for the LEE junctions in selC yield no amplification. Strains with an undetermined LEE insertion site occurred in all three areas studied.

Strains were also tested for 24 DNA sequences of established or putative virulence properties of pathogenic E. coli by colony hybridization (6). DNA probes were obtained from cloned genes (bfpA, perA, E-hly, EAEC, daaC, cdt, cnf, hly) (6) or by PCR amplification, which used as templates the genomic DNA of EAEC prototype strains 042 (aafC, aggR, aspU, shf, irp2, pet, and pic) and 17-2 (aggC and astA); extraintestinal pathogenic strains (ExPEC) J96 (pap) and KS52 (afa), and E. coli HB101 (pANN 801-13) (carrying the sfa probe). PCR primers and assay conditions used were described previously (6,14).

Hybridization with 17 of the 24 sequences tested was detected among the strains; hly and irp2 (31.3% each) and astA (29.3%) were the most frequent. Thirty-four different combinations of these 17 sequences were found (Table 2). Their prevalence varied by location, but 25 (73.5%) occurred in two or fewer strains. Among the less frequent combinations found, some were of genes of ExPEC and EAEC, and others of genes of EPEC (bfpA) and EHEC (E-hly). Moreover, 30.3% of the strains lacked all 24 DNA sequences tested, comprising the most frequent subgroup of strains in all three areas (Table 2). Although these strains carried only the eae gene and the four LEE probe sequences (LEE+ only strains), they may have carried virulence sequences other than those tested. Thus, one should not emphasize the virulence potential of these LEE+ strains solely on the basis of findings of significant differences in their frequencies between cases and controls.

DNA sequences similar to bfpA were detected in 14 (14.1%) of the 99 strains studied, however, only 2 expressed Bfp in Western blot experiments (not shown); these two strains also carried perA and presented AA/LA in 3 hours. The HeLa pattern of the remaining bfpA+ strains varied, but none of them had compact clusters in 3 hours, which is typical of LA. Thus, Bfp expression was found only in strains presenting aggregative adherence/LA in 3 hours, as in typical LA of EPEC (1).

PCR assays with specific primers for the variable region of intimin were used to identify five intimin types (α, β, γ, δ, and ε) (15,16). Most strains had a nontypable intimin (64.6%), but the distribution of these strains varied (approximately 70% in São Paulo and 29%–35% in Rio de Janeiro and Ribeirão Preto). Recently, new schemes were proposed to identify intimin subtypes, which were not tested (17,18). The prevalence of typable intimins varied among the three areas analyzed. Intimin subtypes β (11.1%) and γ (12.1%) prevailed, and intimin ε was not found (Figure 2). The intimin types of two strains were not determined because amplification products of the expected size were obtained with four intimin pairs of primers.

In this study, we sought to verify the frequency with which eae+ EAF– stx– E. coli strains of non-EPEC serogroups occur in persons of poor socioeconomic status in three Brazilian cities; we also compared these strains' genotypic and phenotypic characteristics. Although these strains occurred in 5% to 6% of the populations studied, including nondiarrheic children (in São Paulo and Ribeirão Preto), 73%–88 % of them were dissociated from other well-established enteropathogens (not shown).

Although O51:H40 was the most frequent serotype found and occurred in all three areas studied, the non-EPEC eae+ EAF– stx– strains comprised a large variety of serotypes, and many were O nontypable. Moreover, the strains had diverse adherence patterns and various combinations of pathogenic E. coli DNA virulence sequences; the prevalence of these properties varied among the areas studied. Altogether, these data show that eae+ EAF– stx– E. coli strains outside the EPEC serogroups are even more diverse than already observed (6). As we have emphasized previously, such diversity challenges the diagnosis of these putative pathogens (6).

All strains carried an apparently complete LEE region, and approximately 75.0% of them had the potential to promote attaching and effacing lesions in HeLa cells, as detected by FAS. Thus at least these FAS+ strains are potentially enteropathogenic, since they are capable of inducing attaching and effacing lesions in vitro and may occur in diarrheic patients of various ages and in patients with AIDS. In the EPEC meeting held in 1995, a consensus definition of atypical EPEC was established, namely, that they are EAF–, eae+ strains that promote attaching and effacing lesions (19). Therefore, the FAS+ strains of our study could be classified as atypical EPEC. Whether these strains have additional virulence properties not present in typical EPEC remains to be elucidated. Studies on the virulence potential of selected strains at the cellular and molecular levels will certainly contribute to further understanding of this group of strains and aid in discriminating enteropathogenic strains within the group.