Phenotypic and genotypic characterization of sorbitol-negative or slow-fermenting (suspected O157) Escherichia coli isolated from milk samples in Lombardy region


Claudia Picozzi, Dipartimento di Scienze e Tecnologie Alimentari e Microbiologiche, Università degli Studi di Milano, via Celoria 2, 20133 Milan, Italy (e-mail:


Aims:  To investigate phenotypic and genotypic aspects of sorbitol-negative or slow-fermenting Escherichia coli, suspected to belong to O157 serogroup, isolated in Italy.

Methods and Results:  Milk samples originating from goats and cows were screened for the presence of E. coli O157 with cultural methods. Sorbitol-negative or slow-fermenting strains were subjected to phenotypic characterization, antibiotic resistance profiles, PCR reactions for detection of toxins (stx1 and stx2) and intimin (eaeGEN and eaeO157) genes and clustering by pulsed field gel electrophoresis (PFGE). Only one strain revealed to be O157. Susceptibility to 11 antibiotics highlighted the high resistance to tetracycline (50%), sulfonamide and streptomycin (33%). The stx2 gene was detected in two strains; only the strain identified as O157 exhibited an amplicon for both eae genes. PFGE identified seven distinct XbaI macrorestriction patterns at a similarity level of 41%.

Conclusions:  The use of sorbitol fermentation as cultural method is not sufficient for STEC discrimination while PCR assay proved to be a valuable method.

Significance and Impact of the Study:  The study reports presence of Shiga toxin-producing E. coli in raw milk, signalling a potential risk for humans.


Shiga toxin (ST)-producing Escherichia coli (STEC) have emerged as important food-borne pathogens associated with various human diseases, including watery diarrhoea, haemorrhagic colitis (HC), and the diarrhoea-associated (D+) form of the haemolytic-uremic syndrome (HUS) (Nataro and Kaper 1998).

More than 200 ST-producing E. coli serotypes have been isolated, but only a few are related to severe human diseases: most infections are caused by O157:H7 serotypes (Boyce et al. 1995).

The pathogenicity of STEC strains is mainly associated with the production of either one, or both, of the two phage encoded toxins (stx1 and stx2) thought to cause the vascular endothelial damage observed in haemorrhagic colitis and HUS patients. Another virulence factor contributing to STEC pathogenicity is the eae gene, coding for intimin, that is implicated in the formation of attaching-and-effacing (A/E) lesions in the intestine of the host (Law 2000).

The main reservoir of STEC appears to be cattle, although STEC isolates are frequently found in the faeces of, for example, sheep, goats, pigs, dogs and pigeons. The transmission of STEC occurs through the consumption of contaminated food and water, and through person-to-person and animal-to-person transmission (Ackers et al. 1988; Heuvelink et al. 1995; Bielaszewska et al. 1997).

In Italy, the mean annual incidence of HUS in people of ≤15 years of age from 1988 to 2000 was 0·28 per 100 000 population, and the most common serogroup was O157, although non-O157 was responsible for a considerable number of cases (Tozzi et al. 2003).

The management of patients with STEC infections continues to be non-specific and largely empirical, and the usefulness of anti-microbial therapy is still not clear.

The increased reportage of HUS cases mainly caused by E. coli O157:H7 strains has led to greater attention being given to the development of methods for detecting STEC strains; such methods include culture isolation, serological tests, DNA probes and PCR assays (Heuvelink et al. 1997). Conventional O157 serogroup detection techniques involve enrichment and isolation steps using selective and/or indicator media like E. coli (EC) broth, lauryl sulfate tryptose 4-methyl-umbelliferyl-β-d-glucuronic acid broth and eosin methylene blue agar. However none of these media give specific strain identification. Rapid methods able to identify E. coli O157:H7 in foods or faecal specimens are directed towards immunological or genetic targets. Although immunoassays are more rapid and less labour intensive than cultural techniques (Acheson et al. 1996), they are inadequate as independent procedures. Furthermore, false positives are not uncommon when interpreting immunoassay test results, and further confirmation is required. Also PCR assays targeting stx1 and stx2 genes and the ehx gene (Bastian et al. 1998) have been used to identify STEC. The identity of an isolate as a STEC cannot be confirmed by individual PCRs, therefore the amplification of the different genes, either by means of a multiplex PCR or separated PCRs, is required (Gannon et al. 1992; Fratamico et al. 1995; Fagan et al. 1999).

This study investigated E. coli sorbitol-negative strains isolated in Italy, making an analysis of their phenotype, antimicrobial profile, virulence genes and pulsed-field gel electrophoresis (PFGE) of digested DNA.

Materials and methods

Bacterial strains

We isolated, from a total of 144 cow and goat milk samples, 12 E. coli sorbitol-negative or slow-fermenting strains, suspected of being O157 (Table 1). EC selective broth (225 ml) with novobiocin (20 mg l−1) (Okrend et al. 1990) were added to 25 ml sample, homogenized in a stomacher and incubated at 37°C for 24 h. A loop of the culture was streaked on MacConkey sorbitol agar (March and Ratnam 1986) containing cefixime (0·05 mg l−1) and potassium tellurite (2·5 mg l−1) (CT-SMaC) and incubated at 37°C for 24 h. Sorbitol-negative or slow-fermenting colonies (white coloured), presumably E. coli O157, were then streaked on tryptic soya agar. All the bacteria were grown in tryptic soya broth (Oxoid, Basingstoke, UK) at 37°C and stored at −30°C with 20% glycerol.

Table 1.  List of sorbitol-negative or slow-fermenting Escherichia coli strains
StrainSerotypingSorbitol fermentationGrowth at different temperatureOrigin
  1. OR, rough strains; NS, not serotyped.

  2. *Strains with a ‘slow’ sorbitol fermentation.

Ec43a1NS++Goat's milk
Ec43a2NS++Goat's milk
Ec43bNS++Goat's milk
Ec44NS+Goat's milk
Ec45NS+Goat's milk
Ec46NS++Goat's milk
Ec50NS++Goat's milk
Ec51NS−*++Goat's milk
Ec52NS−*++Goat's milk
Ec34O157:H-+Goat's milk
Ec55OR+Goat's milk
Ev14OR−*+Cow's milk

Phenotypic characterization

Gram stain, growth in tryptic soya broth at different temperatures ranging from +4 to +48°C and the ID32E biochemical test (bioMérieux, Marcy-l'Etoile, France) were used to classify the isolates. The serotyping of E. coli was performed by slide agglutination with antisera (Staten Serum Institute, Copenhagen, Denmark) against different O and H serogroups. Strains giving clumping with distilled sterile water were defined as rough (OR).

Susceptibility testing was performed by the disk diffusion method (NCCLS 2003) with disks containing an antibiotic (Mast Group Ltd, Merseyside, UK). The results were analysed according to the recommendations of the Antibiogram Committee of the French Society for Microbiology (1999). The strains were tested against the following antibiotics suggested by the European Co-ordinating Centre of Enteric Pathogens Surveillance (ENTER-NET Program). The symbols and concentration (μg disc−1) of the respective antibiotics are given in parenthesis: nalidixic acid (NAL, 30), ampicillin (AMP, 10), cefotaxime (CEF, 30), ciprofloxacin (CIP, 5), chloramphenicol (CHL, 30), gentamicin (GEN, 10), kanamycin (KAN, 30), streptomycin (STR, 10), sulfonamide (SUL, 300), tetracycline (TET, 30) and trimethoprim (TMP, 5).

PCR assay for stx1, stx2 and eae genes

Escherichia coli colonies were transferred from tryptic soya agar to 1 ml of sterile distilled water; the bacterial suspensions were heated at 100°C for 15 min, chilled in ice and then stored at −30°C until used in PCR assays. To evaluate the presence of stx genes the primer pairs SLTI and SLTII (Gannon et al. 1992) were used. The eae gene was detected using the primer pairs EAE and EAE157 (Gannon et al. 1997). The authors’ amplification conditions and procedures (reaction mixtures, temperatures of amplification cycles) for each primer set were adhered to. The STEC strain NCTC12079 (serotype O157:H7) served as the positive control and negative, no template, PCR control was included for each amplification reaction set. The reactions were carried out with a T-Gradient PCR thermal cycler (Biometra, Göttingen, Germany). After each PCR assay, 10 μl of the amplification product were analysed on 1·5% agarose gels containing 0·4 μg ml−1 ethidium bromide, visualized with UV illumination, and photographed. Each agarose gel electrophoresis run included DNA molecular size standards (100 bp Ladder; MBI Fermentas, St Leon-Rot, Germany).

Pulsed-field gel electrophoresis

Pulsed field gel electrophoresis (PFGE), modified from the Pulse Net methods of the Center for Disease Control and Prevention (CDC 1998), was used to subtype the isolates. Colonies were plated on TSA, incubated overnight at 37°C, and then suspended in cell suspension buffer (100 mmol l−1 Tris; 100 mmol l−1 EDTA, pH 8) to an optical density of 1·3 ± 0·1 at 610 nm. A 400-μl volume of this suspension was mixed with 20 μl of proteinase K (20 mg ml−1; Sigma Aldrich, Milan, Italy) and an aliquot of 300 μl was mixed with 400 μl of 1% Molecular Biology Certified Agar (Bio-Rad, Hercules, CA, USA) in TE buffer (10 mmol l−1 Tris; 1 mmol l−1 EDTA, pH 8) containing 1% sodium dodecyl sulfate. This mixture was then pipetted into reusable plug moulds (Bio-Rad). After the plugs solidified, they were transferred to 5 ml of lysis buffer (50 mmol l−1 Tris; 50 mmol l−1 EDTA, pH 8; 1% sarcosine) with 25 μl of proteinase K. Plugs were lysed at 54°C with shaking at 80 rev min−1 for 2 h. The lysis buffer was then removed and the plugs were washed twice with distilled water and three more times with TE buffer at 54°C with shaking. Prior to digestion of the agarose-embedded DNA, the plugs were cut into appropriately sized slices that were preincubated for 10 min at 37°C in complete restriction enzyme buffer without restriction enzyme. XbaI (MBI Fermentas GmbH, St Leon-Rot, Germany) was used for genomic DNA digestion. The fragments obtained with this restriction enzyme were resolved by contour-clamped homogenous electric field (CHEF)-PFGE with a Gene Navigator apparatus (Pharmacia, Uppsala, Sweden). An isolate of E. coli, CDC strain G5244, was included every five samples as reference standard. Gels were made of 1·2% pulsed field certified agarose (Bio-Rad) in 0·5x Tris-borate-EDTA buffer (44·5 mmol l−1 Tris; 44·5 mmol l−1 boric acid; 1 mmol l−1 EDTA, pH 8) at 12°C with a voltage gradient of 6 V cm−1. The pulse time was increased from 2·4 to 52·4 s over a 23-h period. The gels were then stained for 30 min in 250 ml of distilled water containing 20 μl of ethidium bromide (10 mg ml−1). Macrorestriction patterns were visualized by UV illumination, photographed and analysed, and compared using BioNumerics software (Applied Maths, Kortrjik, Belgium). We used the band-based Dice similarity coefficient and the unweighted pairs geometric matched analysis (UPGMA) dendrogram type with a position tolerance setting of 0·5% for optimisation and position tolerance of 1·5% for band comparison. Restriction profiles of all the isolates were normalized to the known molecular size bands of the E. coli G5244 standard strain. Strains were considered indistinguishable only if their XbaI patterns were not able to be distinguished by the BioNumerics software.


Phenotypic analysis

All the strains were Gram negative. Of the 12 isolates, seven strains (58%) grew at +4°C after 7 d of incubation, while all grew at +10°C and at +45°C (Table 1). Three isolates (25%) were identified as slow fermenting, as sorbitol fermentation took place after 24 h. Biochemical tests conducted with an ID32E kit confirmed the identification of the species as E. coli. After the agglutination assay one strain was serotyped as O157:H-. Among the other strains, two (Ec55 and Ev14) were found to be rough strains (OR), giving clumping with distilled sterile water and therefore not typeable; the remaining strains did not give agglutination with any of the sera at our disposal.

The Multiple Antibiotic Resistance (MAR) Index is defined as a/b where a represents the number of antibiotics to which the particular isolate is resistant, and b the number of antibiotics to which the isolate is exposed (Table 2). The partial sensitivity (±) to an antibiotic shown by certain strains was considered as negative (sensitive) in determining the MAR Index; three E. coli strains had a MAR Index more than 0·2 (high-risk sources) while four were found to be sensitive to all the antimicrobial agents tested. Sharp peaks of resistance to antimicrobial agents such as tetracycline (50%), streptomycin and sulfonamide (33%), were observed. One isolate (EV14) was resistant to ampicillin (8%). The O157:H- strain was sensitive to most antibiotics employed.

Table 2.  Antibiograms of strains
  1. MAR, multiple antibiotic resistance; +, resistant; ±, partially sensitive; −, sensitive.


PCR assays

The data indicate that the PCR assays are specific, with the length of the products being as expected. Of the 12 E. coli strains, stx2 was the only toxin type, as it was carried from two isolates: Ec34 (O157:H-) and Ev14 (OR). Only E. coli Ec34 was positive for the eaeGEN gene, which is strongly correlated with disease in humans, and to eaeO157 gene.

PFGE typing

Several efforts to obtain intact DNA from two E. coli isolates (Ec51 and Ec52) were unsuccessful, and DNA degradation was always observed, therefore they could not be typed by PFGE. Among the 10 that could be typed, seven different PFGE profiles (excluding reference strain G5244) were identified at a similarity level of 41% (Fig. 1). Ec34 strain (O157 serogroup) is not related to the other isolates and clusters to a 40% level of similarity with the reference strain G5244.

Figure 1.

Clustering based on PFGE patterns after digestion with enzyme XbaI

On considering the E. coli strains that were negative to all the virulence genes tested in this work, there was a greater restriction profile similarity among those collected from the same farm than among those collected from different farms, as already reported (Cobbold and Desmarchelier 2001). Actually Ec50 and Ec55, that were isolated in different months from milk samples of the same cheese-maker, were found to be closely related (level of similarity >97%). Likewise Ec43a1, Ec44, Ec45 and Ec46 that had the same PFGE profile, clustered at a level of similarity of 87% with Ec43a2 and Ec43b, and all they had been collected from a single sampling at another dairy farm.


There are no common biochemical characteristics associated with the great majority of STEC serotypes. Only the recognition of the serogroup O157 was facilitated by its inability to ferment sorbitol after overnight incubation but there is always the need of serotype confirmation through other tests. Moreover, over the past decade it has emerged that STEC strains of serotype O157:H-, which do ferment sorbitol rapidly, are an important cause of human disease in continental Europe (Bielaszewska et al. 1998). Thus, as a result of phenotypic variability, there could be an increased likelihood of misdiagnosing O157 infections using only classical methods. Therefore, inspection services would enhance awareness to the possibility of mislaying O157 serogroup if sorbitol fermentation is regarded as a discriminating test for the pathogen identification.

Indeed the emergence of antibiotic-resistant strains is a major therapeutic problem. This phenomenon has been brought about, and extended, by the transmission of resistant isolates, through the mobility of both local and worldwide populations, and by the consumption of foodstuffs derived from animals treated with antimicrobial agents. The present study has revealed some strains to have multiple resistance to different antibiotics. Among the eight isolates that showed a multiple resistance phenomenon, 75% was resistant to tetracycline, an antibiotic widely employed in feedlots.

The PCR techniques used in the research proved valuable for the identification of STEC strains and for the specific identification of E. coli O157. This broad-based detection was achieved by targeting conserved regions of two virulence factors, stx and eae genes. Our results demonstrated low incidence of STEC (1·4%), and in particular O157 serogroup (0·7%), in raw milk samples; even though this kind of foodstuff always represents a potential source of STEC infections. Moreover, both the STEC strains we isolated, are positive for stx2 that seems to be more important in the development of HUS than stx1 (Nataro and Kaper 1998).

In recent years, DNA macro-restriction analysis by PFGE has been used increasingly for the molecular subtyping of a wide range of bacterial and fungal pathogens. Different authors (Heuvelink et al. 1998; Radu et al. 2001) demonstrated that PFGE is a technique of very high discriminating power for STEC isolates, and we found it a useful tool to differentiate our strains. Nevertheless we noticed that the antimicrobial resistance profile that was highlighted did not turn out to be correlated with the clustering provided by PFGE analysis. Our investigation has provided a new insight into the degree of diversity of the PFGE patterns in Escherichia coli isolates, reflecting the complexity of this organism.

Further studies are necessary to determine pathogenicity of the isolated STEC strains, employing virulence factors other than stx and eae.


We wish to thank Enne de Boer (Kvw, Zutphen, the Netherlands), Wilma Hazeleger (WUR, Wageningen, the Netherlands), Kim van der Zwaluw (RIVM, Bilthoven, the Netherlands) and Mirella Pontello (Centro Enteropatogeni Italia Settentrionale, Milano, Italy) for their very helpful counsel.