Isolation and detection of Shiga toxin-producing Escherichia coli in clinical stool samples using conventional and molecular methods

Shiga toxin-producing Escherichia coli (STEC) serotype O157 : H7 has emerged as one of the most important causes of foodborne infections worldwide. Clinical presentation of STEC infection varies from an asymptomatic state to bloody diarrhoea and life-threatening complications such as haemolytic uraemic syndrome (Karmali et al., 1985; Nataro & Kaper, 1998). Elderly adults and young children are most susceptible to STEC infections as compared to healthy adults (Nataro & Kaper, 1998) and production of Shiga toxin (Stx) is involved in haemorrhagic and systemic symptoms (Karch et al., 2005). Stx production is not restricted to serotype O157 : H7 strains, as over 100 STEC serotypes have been isolated from humans with diarrhoeal illness, including haemolytic uraemic syndrome (Johnson et al., 2006).

The significance of the non-O157 STEC as enteric pathogens is probably underestimated because there are no simple laboratory methods to detect and isolate all of these organisms. Most O157 STEC strains are readily identified based upon a unique phenotypic trait, but the majority of STEC serogroups cannot be routinely distinguished from benign E. coli during basic microbial culture. O157 strains can be detected by differential growth on sorbitol MacConkey agar (SMAC) or similar commercially available media (Church et al., 2007) because most O157 strains do not ferment sorbitol; however, there is no robust phenotypic screening test for differentiating non-O157 STEC from other E. coli strains that are prevalent in stools. High incidence rates of non-O157 STEC serogroups have been reported under conditions of enhanced surveillance (Fey et al., 2000; Jelacic et al., 2003; Lockary et al., 2007; Manning et al., 2007; Thompson et al., 2005). The non-O157 STEC serogroups have also resulted in outbreaks of human illness, including O111 : NM (Boudailliez et al., 1997; Caprioli et al., 1994), O26 : H11 (Miyajima et al., 2007), O121 : H19 (McCarthy et al., 2001) and O103 : H2 (Lindstedt et al., 2007), but generally not with the same frequency as O157 strains.

Current methods for the detection of non-O157 STEC depend primarily on the detection of Stx or toxin-encoding determinants (stx1 and stx2), followed by strain isolation and then serotyping. Conventional testing for the presence of Stx is by inoculating stool filtrate onto Vero cells to observe a cytopathic effect on the cells, and confirmation is performed by neutralization of the toxin by Stx monoclonal antibodies. This method is not suited for routine diagnostic application because it is labour-intensive, time-consuming (48–72 h) and requires cell culture facilities and materials. Alternatively, commercially available immunoassays that detect Shiga toxin directly in stools and PCR amplification platforms to detect stx genes in stool cultures allow for screening of STEC (Beutin et al., 2002; Park et al., 2003; Teel et al., 2007). A multitude of both real-time and conventional PCR assays have been described and evaluated for the detection of stx1 and stx2 genes in E. coli (Bastian et al., 1998; Belanger et al., 2002; Bischoff et al., 2005; Gannon et al., 1992; Nielsen & Andersen, 2003; Paton & Paton, 2002; Wang et al., 2002; Ziebell et al., 2002). Assays for Stx and stx, however, do not differentiate between O157 and non-O157 STEC, and commonly if a sorbitol-negative O157 isolate is recovered from a stool culture encoding stx determinants, then no further laboratory testing is completed for non-O157 strains that may also be present.

The development of molecular methods for identification of STEC serotypes other than O157, and for differentiation of STEC serotypes from benign E. coli, has facilitated the isolation of a wider spectrum of STEC subtypes. The high degree of genetic diversity between E. coli O-antigen gene clusters allows for the development of O-specific PCR reagents (Louie et al., 1998; Perelle et al., 2007); however, PCR schemes currently detect a subset of the relevant non-O157 serogroups. Molecular serogrouping of the O-antigen can be achieved by PCR and DNA sequencing of the gnd locus that is co-inherited with the O-antigen gene cluster (Gilmour et al., 2007a). This method has been developed for use with DNA template prepared from pure culture because gnd is encoded in all E. coli strains and stool cultures contain a complex mixture of E. coli strains. To detect pathogenic E. coli in stool cultures in the presence of other E. coli strains, the espZ locus encoded in the locus of enterocyte effacement pathogenicity island contains polymorphisms that correlate to different STEC serotypes that can be detected using either an allele-specific liquid microsphere suspension assay or real-time PCR assay (Gilmour et al., 2006). The lpfA_OI-154 fimbrial locus specific to O157 and genetically related O55 strains (Shen et al., 2005; Toma et al., 2004) was targeted in a LUX real-time PCR assay and all of these platforms were previously used to detect a co-infection of O177 : NM and O55 : H7 STEC in a 4-year-old hospitalized patient with haemolytic uraemic syndrome (Gilmour et al., 2007b).

The Alberta Provincial Laboratory for Public Health collected stool samples from 876 paediatric patients with gastroenteritis. Stools containing STEC were identified using conventional cytotoxicity and PCR assays and a commercial immunoassay, and both O157 and non-O157 strains were subsequently isolated. Molecular methods were used to detect STEC serogroup-specific genetic determinants in stool enrichment cultures and were compared with the results of the conventional screening and culture techniques.

Stool culture and bacterial strains. Stool samples from children (ages 0–16) with gastroenteritis were collected by the Calgary Laboratory Services, DynaLIFE_dx Medical Laboratories and the University of Alberta Hospital Microbiology Laboratory from 1 May to 31 August 2002. During this study period, one stool specimen per patient was submitted for routine culture of O157 STEC using SMAC media. STEC was also detected in stools using a verocytotoxicity assay, PCR and Premier EHEC enzyme-immunoassay (Meridian). The commercial assay was completed following the manufacturer's directions and the cytotoxicity assay was performed on stool filtrates as previously described by Karmali et al. (1985). Briefly, dilutions of each filtrate were prepared using PBS (undiluted, 1/5, 1/25 and 1/125), and tested for Vero cell death. After 48 h, the last dilution that showed any cytotoxicity was used in the neutralization assay using Stx1 or Stx2 antisera. PCR was performed as previously described (Gannon et al., 1992), using template prepared as described below.

Enrichment cultures of the stool samples were prepared to generate template for molecular assays and to isolate STEC. For each sample, 200 µl watery stool or pea size material was inoculated into 5 ml trypticase soy broth and incubated at 37 °C overnight. This enrichment culture was then refrigerated while PCR for stx determinants was performed. For this process, 200 µl was transferred from the middle of the enrichment culture tube into a 1.5 ml screw-cap centrifuge tube, centrifuged for 3 min at 14 000 g, and the supernatant was removed and the pellet was washed with 1 ml 12 mM Tris buffer (pH 7.4). The wash buffer was removed after centrifuging for 3 min at 14 000 g, and the pellet was resuspended in rapid lysis buffer (100 mM NaCl; 10 mM Tris/HCl, pH 8.3; 1 mM EDTA, pH 9.0; 1 % Triton X-100) and boiled for 15 min (Holland et al., 2000). The supernatant of this preparation was removed, used as template in PCR assays and then subsequently stored at –70 °C. To isolate STEC from stool that was positive by either the PCR or enzyme-immunoassay screening methods, 100 µl of the refrigerated enrichment culture was plated onto SMAC and MAC plates. Ten random lactose-fermenting colonies were picked from the MAC plates and screened for stx determinants by PCR, and positive colonies were serotyped by conventional agglutination using antisera prepared at the National Microbiology Laboratory.

PCR and liquid microsphere suspension assays. Oligonucleotide primers and fluorescent probes are described in Table 1. Allelic discrimination of espZ variants was performed as previously described (Gilmour et al., 2006) using primer set GIL245 and GIL246-L (Table 1) and a liquid microsphere suspension array containing four fluorescently coded microspheres each covalently coupled with a probe specific for one of the four espZ alleles γ1, β1, γ2 and ε. Molecular serogrouping of the O-antigen was performed by PCR and DNA sequencing a region of the gnd locus using templates prepared directly from the stool enrichment culture (Gilmour et al., 2007a). Characterization of gnd DNA sequence data was completed with the E. coli O-Typer webtool () (M. W. Gilmour and G. Van Domselaar, unpublished).

Table 1. Oligonucleotides used in this study

Detection and isolation from stool
A collection of 876 stool samples from paediatric patients under the age of 17 with presentation of gastroenteritis was screened for STEC using a cytotoxicity assay, commercial immunoassay and a conventional PCR targeting Shiga toxin determinants. A total of 45 stool samples presumptively containing STEC were identified from this collection (Table 2) and routine culture for O157 STEC using SMAC media isolated O157 : H7 strains from 20 stools. To isolate non-O157 STEC, enhanced culture isolation was attempted using MAC media and screening individual lactose-fermenting isolates using stx PCR. Of the 45 tested stool cultures, 40 stools yielded STEC isolates and serotyping identified 20 O157 : H7 and 22 non-O157 STEC isolates, including serotypes O26 : H11 (11 isolates), O121 : H19 (three isolates) and one isolate of each of the following serotypes: O26 : NM, O103 : H2, O111 : NM, O115 : H18, O121 : NM, O145 : NM, O177 : NM and O5 : NM (Table 2). Notably, multiple STEC serotypes were detected and isolated from two clinical stool samples. Stool sample 03-3900 yielded O157 : H7 and O26 : H11 strains, and sample 03-3425 yielded O157 : H7 and O103 : H2 strains. In addition, four non-toxin-producing E. coli isolates (O25 : H1, O8 : H19, O2 : H4, O1 : H7) were carried through the screening process and serotyped (Table 2). The relative ratio of O157 versus non-O157 STEC isolates recovered was similar to that observed during other enhanced clinical stool STEC isolation studies (Fey et al., 2000; Jelacic et al., 2003; Lockary et al., 2007; Manning et al., 2007; Thompson et al., 2005).

Table 2. E. coli isolates from clinical stool samples and molecular characterization of stool culture enrichments Stool culture screening assays included an immunoassay (Premier), cytotoxicity (VCA) and a PCR assay for stx determinants. Routine culture for O157 was on SMAC media and enhanced isolation was on MAC media (see Methods). Subsequent assays using DNA template prepared from the stool culture included gnd-based molecular serogrouping, LUX real-time PCR detection of the O157-specific locus lpfAOI-154 and liquid microsphere suspension typing of the espZ allele.

For 5 of the 45 stool samples that were expected to contain a STEC strain, none were isolated. Screening of individual lactose-fermenting colonies for STEC after culture on MAC plates may have been insufficient, or these STEC strains were in a viable but non-culturable state. Of the initial screening methods (cytotoxicity assay, stx PCR and Stx immunoassay), at least one method had a negative result for each of these five stool samples (Table 2). For the 40 stool samples from which STEC was isolated, the cytotoxicity assay was positive for 38 samples, stx1 or stx2 PCR was positive for 40 samples, and the Stx immunoassay was positive for 33 samples. These screening assays, including stx PCR, offered little insight into the serotypes ultimately isolated from the stool samples because they are limited to detecting only the two major stx alleles, stx1 and stx2. For example, for the two stool cultures that contained an O157 and a non-O157 strain, the recovered O157 : H7 strains in each instance encoded both stx1 and stx2 determinants, whereas the O26 : H11 and O103 : H2 isolates both encoded stx1. If only routine O157 culture and stx PCR were performed, the stx1 signal contributed by (and the presence of) the non-O157 strains would be concealed in the presence of O157 : H7 isolates.

Molecular characterization of stool enrichment cultures
Preliminary serotype identification using molecular methods can provide the first indication of relatedness between samples prior to culture isolation and serotyping, and can be valuable for laboratories that lack the capacity to perform serotyping. To determine the molecular profile of the stool enrichment culture in relation to the STEC strains independently isolated from those same stool samples, molecular assays targeting the lpfA_OI-154 fimbrial loci and the espZ virulence loci were performed on DNA extracts from enriched stool cultures. The lpfA_OI-154 fimbrial locus is specific for O157 and genetically related O55 strains. The espZ gene encodes a type III secreted effector protein and could be typed into four alleles: γ1 (encoded by serotypes O157 : H7 and O145 : NM), β1 (O26 : H11, O26 : NM, O5 : NM and O177 : NM), γ2 (O111 : NM) and ε (O121 : H19 and O103 : H2) (Gilmour et al., 2006).

For all stool samples from which an O157 : H7 strain was isolated (n=20), all but one enriched stool DNA extract encoded lpfA_OI-154 (extract 03-1226; Table 2). Similarly, for stool cultures containing a single STEC strain of serotype O157 : H7, O26 : H11, O26 : NM, O121 : H19, O121 : NM, O111 : NM, O145 : NM, O177 : NM or O5 : NM, the expected espZ allele was detected in almost all cases (Table 2). The exceptions were sample 03-1226, containing an O157 : H7 isolate (but no γ1-espZ allele was detected), sample 03-3004, containing a O115 : H18 strain that did not encode espZ, and sample 03-3681, which yielded a stx1+ O111 : NM strain but both γ1-and γ2-espZ alleles were detected, indicating that a second STEC strain may have been present. Detection of the γ1-espZ allele suggests that an O157 strain could have been present, but this stool did not encode lpfA_OI-154, potentially excluding an O157 or O55 strain (Gilmour et al., 2007b). For the stool cultures containing two STEC serotypes, both the γ1-espZ and ε-espZ alleles were detected in sample 03-3425 (corresponding to O157 : H7 and O103 : H2 isolates), but only the β1-espZ allele was detected in sample 03-3900 (corresponding to the O26 : H11 isolate). However, this DNA extract encoded lpfA_OI-154, so the combination of results from both the lpfA-LUX and espZ-microsphere assays accounted for both STEC strains (O157 : H7 and O26 : H11).

For each of the five samples that had a positive result in at least one of the screening assays (cytotoxicity assay, stx PCR and Stx immunoassay), but no STEC was isolated, there was also a positive result in the espZ allelic discrimination assay (Table 2). Two of these stool cultures encoded stx1 and either γ1-espZ or ε-espZ alleles, and for two stx2+ cultures the β1-espZ and ε-espZ alleles were observed. For the fifth culture (03-3936) that did not yield an STEC isolate, the only positive screening assay was the commercial enzyme-immunoassay; however, two espZ alleles were detected (γ1 and β1). None of these five cultures encoded lpfA_OI-154, potentially excluding O157 or O55 strains. These data cannot predict individual serotype identities for the presumptive STEC strains that were present; however, the detection and typing of a virulence determinant provided further support that STEC was present in these samples and provided more depth to the genetic characterization of stool samples than contributed by stx detection alone.

The gnd locus is encoded at the terminus of the O-antigen gene cluster and DNA sequencing of a 643 bp subregion is indicative of individual O-serogroup STEC lineages (Gilmour et al., 2007a). The gnd locus is encoded in all E. coli strains so we hypothesized that amplification from stool culture enrichments (likely to contain a multitude of non-pathogenic E. coli strains comprising the patient's normal flora) might result in a mixture of PCR products of near-identical lengths from each sample and would therefore not be amenable to DNA sequencing. To determine the utility of the gnd molecular serogrouping method on this type of template, PCR and DNA sequencing was attempted for 45 stool enrichments, including the samples for which no STEC was isolated but STEC was suspected (Table 2). Surprisingly, 14 of the sequenced gnd amplicons were 100 % identical to the isolate from the respective stool culture, including 11 O157 gnd sequences corresponding to O157 : H7 isolates, and one corresponding match for each of O111, O121 and O25 (which is not a toxin-producing isolate). In addition, for samples 03-1510 and 03-3425 the gnd sequence was 99 % identical to O157 reference sequences and this corresponded to the O157 : H7 strains recovered from the stool cultures. An O103 : H2 isolate was also isolated from sample 03-3425; however, this second gnd allele in the culture population was seemingly not co-amplified. The gnd sequence of sample 03-3904 was 99 % identical to the reference O117 gnd sequence but only a O121 : NM strain was isolated (matching the observed espZ-ε allele).

The database of gnd alleles was not comprehensive for all E. coli serogroups, rather it was targeted to all STEC serogroups reported by Canadian public health laboratories (Gilmour et al., 2007a). For 10 stool cultures the gnd amplicon sequence was less than 99 % identical to any allele present in the database, and accordingly these did not correlate to the known serogroup of the isolated strain (Table 2). No amplicons were produced for the remaining 12 preparations, indicating that templates prepared from stool containing a complex mixture of E. coli strains are not ideal for amplification of gnd. For those templates from which gnd could be amplified, the subsequent DNA sequencing and allele assignment was capable of determining O-serogroups of STEC when a threshold of ≥99 % was applied for the global pairwise DNA sequence identity between stool sequence and reference gnd sequence. This approach will not be appropriate for routine diagnostics of clinical specimens, but may be of use during outbreak situations and to detect viable but not culturable STEC strains when the threshold sequence identity is applied and in combination with the espZ typing method.

Conclusions
Using non-differential culture techniques to isolate STEC from clinical stool samples, without introducing a bias towards serogroup O157 strains, we observed an approximately even isolation rate of O157 versus non-O157 strains. This was comparable to the isolation rates observed in other studies where differential media were not used. Furthermore, the ability to detect STEC strains in clinical stool enrichment cultures using serogroup-specific molecular assays was evaluated against these isolation techniques. The sensitivity of each of the lpfA, espZ and gnd assays was reduced compared to stx PCR, as at least one false-negative result occurred with each. However, in all other cases the lpfA assay accurately predicted the presence of O157 strains, and the espZ assay could indicate the presence of STEC strains and also provide a broad level of typing between samples. These data included two incidences of co-infection with both an O157 and a non-O157 STEC strain that were detected by a combination of these methods. For clinical samples that STEC strains were not successfully cultured from, these methods also provided cumulative genetic profiles indicating the presence and subtype of possible STEC strains. Therefore, if used in clinical microbiology settings these platforms would accurately indicate when more comprehensive culture-based efforts are required to isolate STEC strains from stool samples. A PCR and DNA sequencing method originally developed for molecular serogrouping pure STEC isolates was also evaluated on stool enrichment cultures. The majority of results were inconclusive, and this partly reflects the likely amplification of non-STEC E. coli contributed by the stool and the lack of representation for these benign E. coli in the reference sequence database. However, one-third of stool cultures were accurately characterized when a threshold of ≥99 % pairwise DNA sequence identity was applied. These assays cumulatively provide a molecular toolbox to detect STEC and serogroup lineages directly from clinical stool enrichment cultures, and will be useful to direct culture-based isolations of STEC strains.

The authors thank J. McCrea, A. Deroschers, K. Trout and A. Andrysiak for assistance with real-time PCR and microsphere array experiments, S. Steven for culture and picking of the colonies, J. Walsh for performing serotyping of stool isolates and C. Gebhart for assistance with conventional PCR.