Characterization of Shiga Toxin-Producing Escherichia coli Strains Isolated from Human
Patients in Germany over a 3-Year Period
Today, more than 200 different E. coli O:H serotypes are known to be associated with the production of Shiga toxins
J Clin Microbiol. 2004 March; 42(3): 1099-1108.
doi: 10.1128/JCM.42.3.1099-1108.2004.Copyright © 2004, American Society for Microbiology
Characterization of Shiga Toxin-Producing Escherichia coli Strains Isolated from Human
Patients in Germany over a 3-Year Period
Lothar Beutin,* Gladys Krause, Sonja Zimmermann, Stefan Kaulfuss, and Kerstin GleierDivision of Microbial Toxins,
Department of Biological Safety, Robert Koch Institute, D-13353 Berlin, Germany
*Corresponding author. Mailing address: Division of Microbial Toxins, Department of Biological Safety, Robert Koch
Institut, Nordufer 20, D-13353 Berlin, Germany. Phone: 49 30 45 47 2484. Fax: 49 30 45 47 2673. E-mail: [email protected]
Received September 22, 2003; Revised November 12, 2003; Accepted November 15, 2003.
This article has been cited by other articles in PMC.
We have investigated 677 Shiga toxin-producing Escherichia coli (STEC) strains from humans to determine their
serotypes, virulence genes, and clinical signs in patients. Six different Shiga toxin types (1, 1c, 2, 2c, 2d, and 2e) were
distributed in the STEC strains. Intimin (eae) genes were present in 62.6% of the strains and subtyped into intimins a1,
ß1, ?1, , ?, and ?. Shiga toxin types 1c and 2d were present only in eae-negative STEC strains, and type 2 was
significantly (P < 0.001) more frequent in eae-positive STEC strains. Enterohemorrhagic E. coli hemolysin was
associated with 96.2% of the eae-positive strains and with 65.2% of the eae-negative strains. Clinical signs in the
patients were abdominal pain (8.7%), nonbloody diarrhea (59.2%), bloody diarrhea (14.3%), and hemolytic-uremic
syndrome (HUS) (3.5%), and 14.3% of the patients had no signs of gastrointestinal disease or HUS. Infections with eae-
positive STEC were significantly (P < 0.001) more frequent in children under 6 years of age than in other age groups,
whereas eae-negative STEC infections dominated in adults. The STEC strains were grouped into 74 O:H types by
serotyping and by PCR typing of the flagellar (fliC) genes in 221 nonmotile STEC strains. Eleven serotypes (O157:[H7],
O26:[H11], O103:H2, O91:[H14], O111:[H8], O145:[H28], O128:H2, O113:[H4], O146:H21, O118:H16, and O76:[H19])
accounted for 69% of all STEC strains. We identified 41 STEC strains belonging to 31 serotypes which had not
previously been described as human STEC. Twenty-six of these were positive for intimins a1 (one serotype), ß1 (eight
serotypes), (two serotypes), and ? (three serotypes). Our study indicates that different types of STEC strains
predominate in infant and adult patients and that new types of STEC strains are present among human isolates.
The association of Shiga (Vero) toxin production in Escherichia coli with human pathogenicity was first described in 1979
(82, 85). However, it was the investigation of an outbreak caused by Shiga toxin-producing E. coli (STEC) O157 which
provided the major impetus to study these pathogens (65). In the following years, STEC strains were increasingly
isolated from humans with diarrhea and hemolytic-uremic syndrome (HUS) and from farm animals, which serve as a
natural reservoir for STEC (52, 86). Today, more than 200 different E. coli O:H serotypes are known to be associated
with the production of Shiga toxins (86; K. A. Bettelheim's VTEC table, May 2003 update, www.sciencenet.com.
au/vtectable.htm). Certain STEC strains belonging to serogroups O26, O103, O111, O145, and O157 were more
frequently isolated from humans with severe diseases such as hemorrhagic colitis and HUS. Accordingly, these highly
virulent STEC strains were also designated as enterohemorrhagic E. coli (EHEC) (42, 52). The search for additional
virulence markers in these pathogens revealed that most EHEC strains carry a plasmid which encodes a hemolysin
(EHEC hemolysin) and the chromosomally located locus of enterocyte effacement (LEE) pathogenicity island (16, 43,
70, 84). The genes carried by the LEE enable the bacteria to produce attaching and effacing lesions in the host
intestinal mucosa cells, which increases the virulence of the bacteria for humans (35, 44, 60). Intimate attachment of
bacteria to the host cell is mediated by the binding of intimin, the product of the eae gene, to the translocated intimin
receptor (80). Nucleotide sequencing of the LEEs from STEC O157 and enteropathogenic E. coli (EPEC) strains
revealed differences in the genes coding for intimate attachment of bacteria to the enterocytes (48, 59, 89), and more
than 10 genetic variants of the eae gene have been identified in STEC and EPEC strains (32, 34, 55, 77, 88). Some
intimin types, such as intimin a, were found to be associated with EPEC, whereas others, such as intimins ?, , and ?,
were found in STEC strains (1, 55, 77, 88). The association between infections with intimin-positive STEC and severe
disease in humans was demonstrated previously (13), but it was also shown that intimin is not essential for the virulence
of certain STEC strains for humans. Other colonization mechanisms, such as adhesins and pili, were identified in eae-
negative strains (58, 75), and certain LEE-negative strains and serotypes of STEC were associated with bloody diarrhea
(BD) and HUS (13, 14, 36). These STEC strains might possess other virulence factors which have not yet been
characterized.The virulence of STEC for humans may also be related to the Stx type which is produced by the bacteria.
The presence of the stx2 gene in the infecting strain was previously reported to correlate with severe disease in humans
(13), and the administration of purified Stx2, but not of Stx1, was shown to cause HUS in experimentally treated primates
(74). A variety of genetic variants of stx1 and stx2 were detected by nucleotide sequence analysis of stx genes (27, 30,
56, 61, 73). Some of the stx1 and stx2 variants were found to be associated with STEC from sheep (stx1ox3/stx1c and
stx2d-ount) (15, 39, 64, 79), pigs (stx2e) (83), or pigeons (stx2f) (72). Some genetic variants of stx1 (stx1ox3/stx1c) and
stx2 (stx2e and stx2d-ount) are not present in classical EHEC strains but are frequently found in eae-negative STEC
strains from patients with uncomplicated diarrhea or asymptomatic infections (23, 24, 25, 39, 61). Other variants, such
as stx2e, were rarely or not (stx2f) associated with STEC from humans (24).E. coli strains belonging to serotypes O157:
[H7], O145:[H28], O111:[H8], O103:H2, and O26:[H11] are recognized classical EHEC types which occur in different
countries worldwide (86; www.sciencenet.com.au/vtectable.htm). Diagnostic tools such as indicator media, O-antigen-
agglutinating antisera, and magnetic beads coated with O-antigen-specific capture antibodies were developed for the
enrichment and isolation of EHEC strains from fecal, environmental, and food samples. These tools proved to be useful
for the detection of some but not all human-pathogenic STEC types (68). New emerging EHEC clones formed by O118:
H16 and O121:H19 strains were recently described and were associated with BD and HUS (4, 31, 46, 76). These
findings show that the list of human pathogenic STEC types is far from being completed, and further work has to be
done to characterize human STEC strains for their serotypes, their virulence markers, and their associations with
disease. This was the aim of our study, where we have investigated 677 STEC isolates which were collected between
1997 and 1999 from human patients in Germany.
Studies from different countries have shown that humans can be infected with a large spectrum of serologically different
STEC types (86; www.sciencenet.com.au/vtectable.htm). In an earlier study, we analyzed 89 human non-O157 STEC
strains which were isolated in 1996 (10). As a result, eae-positive O118:H16 strains were identified as emerging EHEC
strains in Germany, and some eae-negative STEC strains belonging to serogroups O91, O128, and O146 were
frequently found among human clinical isolates. These findings encouraged us to examine larger numbers of human
STEC strains in order to characterize the STEC strains associated with human infections in more detail and also to
detect possible new STEC types.
The characterization of clonal types in E. coli populations by multilocus enzyme electrophoresis and by multilocus
nucleotide sequencing has shown that the O:H serotype is a good indicator for the identification of strains belonging to
distinct clonal groups (17, 22). However, many EHEC and STEC isolates are phenotypically NM (www.sciencenet.com.
au/vtectable.htm) and therefore cannot be grouped into distinct O:H serotypes. In order to detect the H types of NM
STEC strains, we have characterized the fliC genes in these strains by PCR-RFLP typing. This method has previously
been shown to be suitable for the grouping of STEC O91:NM and O128:NM strains into serotypes O91:[H14] and O128:
[H2], respectively, and the close relationship between motile and NM strains belonging to the same serotype was
confirmed by pulsed-field gel electrophoresis typing (79). The contribution of molecular H-antigen typing for the
identification of STEC serotypes is emphasized by the facts that 221 (32.6%) of the strains from our study were
phenotypically NM and that only 6 of these were negative in the fliC PCR and were classified as H antigen negative. High
numbers of NM strains were also detected in other studies of EHEC O26, O111, O145, and O157 (13, 24) and STEC
O91, O113, and O174 (old designation, OX3) strains (28, 62, 63). In our study, NM and motile strains belonging to the
EHEC O-antigen groups O26, O103, O111, O118, O121, O145, and O157 could be assigned to single O:H types by fliC
genotyping (Table 1). NM STEC strains belonging to O-antigen groups which were associated with more than one H-
antigen type (O91, O113, O115, O125, and O177) could be typed accordingly (Table 2). Twenty-four of 53 known E.
coli H types (54) were found in the STEC strains, and 15 H types were associated with strains belonging to more than
one O-antigen group (Table 3). Only 10 H types (H2, H7, H8, H11, H16, H19, H25, H28, H30, and H33) were linked with
the 424 eae-positive STEC strains from our study, and these were distributed over 21 O groups and Ont and O rough
strains. These findings indicate that the determination of the fliC type may be a useful diagnostic approach for the
detection and characterization of STEC strains.
In order to search for new STEC types which are not yet known to occur in humans, we used a reference list which
summarizes published data on the serotypes and origins of non-O157 STEC strains (www.sciencenet.com.au/vtectable.
htm). By this list, we could identify 41 STEC strains belonging to 31 different serotypes which were not previously
described as human STEC (Tables 1 and 2). Some of these “new” O:H types were already reported to be STEC strains
from animals, and it was shown that certain STEC serotypes are closely associated with some animal host species (12,
26, 29, 41, 53, 78, 79). Based on these reports, we made an estimate about the possible animal source of the STEC
strains from our study. Most of the eae-positive human STEC strains listed in Table 1 belong to serotypes which are
closely associated with cattle. The eae-negative STEC strains were more diverse in their relations to animal hosts.
Serotypes O91:[H14], O128:H2, and O146:H21, which represent about 30% of the human eae-negative STEC strains
(Table 2), were reported to be associated with sheep (5, 12, 20, 41, 78, 79), whereas others, such as O22:H8, O113:
[H4], and O113:[H21], are common in cattle (29, 63). Animal and human STEC strains which belong to the same
serotype were found to be similar in their virulence markers, and the transmission of STEC from animals to humans has
been reported (86). According to these findings, cattle and sheep represent an important source of STEC types which
were frequently isolated from humans in our study.
Previous studies have shown that the virulence of STEC for humans may be related to the type of Shiga toxin which is
produced by the bacteria. Of the different Shiga toxins, Stx2 (stx2) was found to be related with high virulence and was
significantly associated with STEC strains from BD and HUS patients (13, 24). Similar findings were made in our study,
as toxin type 2 was present in 20 of 21 STEC strains from HUS patients (Table 5). Moreover, toxin type 2 (stx2) was
found more frequently in eae-positive STEC strains than in eae-negative STEC strains (Tables 1, 2, and 4). The
combination of the stx2 and eae genes was found to be significant in another study, which was performed on STEC
strains of different origins and serotypes (13).
In contrast, toxin type 1c (genotype stx1ox3/stx1c) STEC strains were shown to be closely associated with sheep (15,
39, 78, 79). In our study, toxin type 1c was present in 41% of the eae-negative STEC strains and spread over 24
different serotypes. Toxin type 1c was often present in combination with toxin type 2c (stx2d-ount or stx2d-OX3a) but not
with toxin type 2 (Table 2). STEC strains with toxin type 1c and/or 2c from our study were frequently isolated from
patients with milder disease (abdominal pain or nonbloody diarrhea), corresponding to previously published results (24,
25, 61). The lower virulence of these STEC strains for humans may be explained by the lack of the attaching and
effacing property and of Stx2; both factors were reported to be major virulence attributes of STEC strains causing
severe disease in humans (13, 24).
A mucus elastase-activatable variant of Stx2 called stx2d (stx2vha and stx2vhb) was previously described for an STEC
strain of serotype O91:H21 (40, 49). Toxin type 2d-positive STEC strains of serotypes O91:H10, O91:H21, O174:NM,
and O174:H21 were isolated from patients with BD and HUS (30, 49, 62), and a linkage was found between severe
gastrointestinal disease in patients and the presence of stx2d (31). In our study, we could identify two known (O91:[H21]
and O174:H21) and four new (O113:[H21], O148:H8, O163:[H19], and Ont:[H21]) serotypes of toxin 2d strains (Table 2).
Two of these serotypes (O148:H8 and O174:H21) were from patients with BD.
It was reported that toxin type 2e strains are rarely isolated from humans and cause milder disease (24). Similar findings
were made in our study (Table 4). The production of Stx2e is characteristic for porcine STEC strains, which cause
edema disease in pigs (26, 83). However, none of the strains from our study belonged to typical porcine serotypes,
indicating that these are not as important as human pathogens.
Intimins ? and a were previously not described as being associated with human STEC (55, 77). In this study, we have
detected intimin ? in eight STEC strains with flagellar type H25 (O109:H25, O156:H25, Ont:H25, and Orough:H25) which
were positive for EHEC hlyA and, except in Ont:H25 (toxin type 2) strains, for toxin type 1 (Table 1). Intimin ? was
recently described to occur in bovine STEC strains which belonged to serotypes other than those of our human STEC
isolates (33, 38). Intimin a was previously detected in EPEC but not in STEC strains (7, 32, 55) and was detected in our
study in two strains representing a new STEC serotype, O177:[H7] (toxin type 1, EHEC hlyA negative).
Intimin ß is reported to be the most frequent type of intimin in EPEC and STEC strains from humans and animals, and its
presence is associated with multiple E. coli serotypes (1, 7, 19, 55). In this study, intimin ß1 was detected in 123 STEC
strains and 12 serotypes. Intimin ß1 was found as a characteristic trait of all STEC strains with flagellar type H11 (Tables
1 and 3). Fourteen intimin ß1-positive STEC strains belonging to eight serotypes (O5:H11, O43:[H30], O68:H11, O68:
H25, O77:H11, O123:H11, O177:H11, and O177:[H25]) were identified as new groups of human STEC strains in this
Intimin was first described for EHEC O103:H2 strains (55) and more recently for emerging EHEC O121:H19 strains (31,
76). Apart from EHEC O103 and O121 strains, we have detected intimin in two new groups of STEC O123:H2 and Ont:
H2 strains (Table 1). Intimin was also found in many STEC strains which were O rough; 14 of these were positive for
flagellar type H2, toxin type 1, and EHEC hlyA and resembled strains of the EHEC O103:H2 clone. Characteristic
combinations between H types, toxin types, and EHEC hlyA and intimin types were found in other STEC O rough strains,
which may indicate that these strains originated from classical EHEC O157:H7 and O111:H8 strains (Table 1).
EHEC hemolysin, which causes an enterohemolytic phenotype on blood agar, was detected in many STEC strains of
different origins (12, 13, 28, 29, 69) and was found to be significantly associated with eae-positive STEC strains
belonging to classical and emerging EHEC types (13, 28, 46, 69, 76). Similar findings were made in our study, where
EHEC hemolysin was detected in 96.2% of the eae-positive STEC strains. In contrast, alpha-hemolysin, which is known
as a characteristic trait of porcine STEC strains (26), was detected in only two (0.3%) of the human STEC strains. The
presence of toxin type 2e in the two alpha-hemolytic human STEC strains indicates that these strains could have
originated from pigs.
It was previously reported that aggregative adherence and EaggEC-specific DNA sequences were found in STEC O111:
H2 strains from HUS patients in France (51). EaggEC-specific DNA sequences were not found in any of the 677 STEC
strains from our study, indicating that this pathotype is not common among STEC strains from Germany.
Ninety-four (87.0%) of 108 STEC strains which were isolated from patients with BD or HUS belonged to EHEC-related O
groups and/or carried virulence markers (intimin, EHEC hlyA, and Stx2) which were previously associated with severe
disease in humans (Table 5). Thirteen strains from patients with BD and one strain from a patient with HUS did not
belong to classical EHEC serotypes and were negative for intimin (Table 5). These 14 strains were distributed over 13
serotypes, and some of these (O22:H8, O105:H18, O174:H2, and O174:H21) were already described as isolates from
patients with BD or HUS (11, 23, 66). Our findings support previous studies indicating that certain serotypes of eae-
negative STEC strains may cause severe disease (14, 23, 37, 63). The pathomechanism by which these atypical EHEC
strains cause disease is not well known. Toxin types 2 and 2d may contribute to the virulence of atypical EHEC strains.
In our study, toxin type 2 was found in 7 of 14 eae-negative strains (50%) from patients with BD and HUS but only in 19
(9.4%) of 202 eae-negative strains from all other patients, and two patients with BD were infected with toxin type 2d
The search for new STEC types in a large group of human patients resulted in the detection of 41 strains and 31
serotypes which have not been described before as human STEC. Nineteen strains distributed over 11 serotypes
represented new types of eae-positive STEC, and 16 of these expressed EHEC hemolysin; both properties are virulence
attributes of classical EHEC strains (52). None of these strains were from patients with BD or HUS. Toxin type 2, which is
associated with increased virulence of STEC, was found in only five of these strains belonging to serotypes O77:H11
and O177:[H25] (Table 1). The small number of cases which involved infection with the new types of eae-positive STEC
strains does not permit an estimate of the virulence potential of these strains.
We had previously reported that infections with eae-positive STEC are associated with young age but that eae-negative
isolates are more frequently isolated from adult patients (5). These findings were confirmed (P < 0.001) in the present
study, which was performed on a larger number of isolates. Protective immunity to intimin may be acquired in early
childhood due to infections with eae-positive EPEC and STEC strains (7, 21, 47), and this may explain why these strains
are less frequently isolated from adults. On the other hand, adults are principally more exposed to STEC strains from
nonhuman sources due to occupational contact with animals, food, and the environment, and the majority of STEC
strains from these sources are negative for intimin (6, 20, 29, 50). This may explain the high frequency of infections with
eae-negative STEC strains in adults. Severe disease such as BD or HUS was more frequent in young patients, which
corresponds to the virulence attributes of their STEC isolates.
Our study shows that different types of STEC strains predominate in infant and adult patients and that new types of
STEC strains can be identified by subtyping of virulence genes and by serotyping of new O-antigen groups, including
O175 to O181. The fliC PCR allowed the determination of H-antigen types in 221 (32.6%) STEC strains which were
phenotypically NM. The finding of Ont strains in this study (Tables 1 and 2) suggests that further O types need to be