Non-O157 Shiga Toxin Producing Escherichia coli in Foods

Transcription

1 1721 Journal of Food Protection, Vol. 73, No. 9, 2010, Pages Copyright G, International Association for Food Protection Review Non-O157 Shiga Toxin Producing Escherichia coli in Foods EMILY C. MATHUSA,* YUHUAN CHEN, ELENA ENACHE, AND LLOYD HONTZ Grocery Manufacturers Association, 1350 I Street N.W., Suite 300, Washington, D.C , USA MS : Received 21 January 2010/Accepted 8 May 2010 ABSTRACT Non-O157 Shiga toxin producing Escherichia coli (STEC) strains have been linked to outbreaks and sporadic cases of illness worldwide. Illnesses linked to STEC serotypes other than O157:H7 appear to be on the rise in the United States and worldwide, indicating that some of these organisms may be emerging pathogens. As more laboratories are testing for these organisms in clinical samples, more cases are uncovered. Some cases of non-o157 STEC illness appear to be as severe as cases associated with O157, although in general cases attributed to non-o157 are less severe. There is much variation in virulence potential within STEC serotypes, and many may not be pathogenic. Of more than 400 serotypes isolated, fewer than 10 serotypes cause the majority of STEC-related human illnesses. Various virulence factors are involved in non-o157 STEC pathogenicity; the combined presence of both eae and stx genes has been associated with enhanced virulence. A scientific definition of a pathogenic STEC has not yet been accepted. Several laboratories have attempted to develop detection and identification methods, and although substantial progress has been made, a practical method of STEC detection has yet to be validated. Worldwide, foods associated with non-o157 STEC illness include sausage, ice cream, milk, and lettuce, among others. Results from several studies suggest that control measures for O157 may be effective for non-o157 STEC. More research is needed to uncover unique characteristics and resistances of non- O157 STEC strains if they exist. The public health significance of non-o157 STEC and the implications for industry practices and regulatory actions are discussed. Escherichia coli is one of the most studied microorganisms (11). Part of the Enterobacteriaceae family, E. coli is a facultatively anaerobic, rod-shaped, gram-negative bacterium. It is a non spore-former, and some strains may be motile with peritrichous flagella. Certain strains of E. coli have been recognized as human pathogens since the 1940s; subsequently, additional strains have been linked to human illness. Several serotypes of STEC have been linked to foodborne illness, but not all STEC strains are capable of causing human disease (38). The different enteropathogenic groups of E. coli are broken down into six pathotypes (46). The pathotype of interest in this review is verocytotoxigenic E. coli (VTEC), also known as Shiga toxin producing E. coli (STEC) (11). The terms Shiga toxin (Stx) and verotoxin are interchangeable; they refer to the production of cellular cytotoxins (53, 64, 79, 102). Enterohemorrhagic E. coli (EHEC) strains are a subset of STEC that cause bloody diarrhea (79). All EHEC strains are considered to be pathogenic, because they are correlated to specific clinical connotations, whereas not all STEC strains are pathogens. EHEC strains are defined as possessing a ca. 60-MDa plasmid, expressing Stx, and having the ability to cause attaching and effacing (AE) lesions on epithelial cells (78). * Author for correspondence. Tel: ; Fax: ; emathusa@gmaonline.org. In 2003, Karmali et al. (67) proposed that STEC be classified into five seropathotypes, A to E. Seropathotype A consists of serotypes considered to be most virulent; this group consists of O157:H7 and O157:NM (nonmotile). Seropathotype B consists of serotypes that are able to cause severe illness and outbreaks but occur less frequently. Examples of serotypes in this group according to the data reported include O26:H11, O103:H2, O111:NM, O121:H19, and O145:NM. Organisms belonging to seropathotype B are the focus in this review. Seropathotype C is composed of serotypes that are infrequently associated with sporadic hemolytic uremic syndrome (HUS) but not often with outbreaks, including O91:H21 and O113:H21. Seropathotype D consists of serotypes able to cause diarrhea, and seropathotype E includes all the other STEC serotypes that have not been implicated in human disease (53, 67). Scheutz (92) argued that there are problems in using this classification system for STEC, because it associates serotype with degree of illness instead of the organism s virulence profile. Scheutz suggests a classification of STEC into three groups. The first group would include HUSinducing and/or epidemic outbreak potential STEC, the second group would consist of human diarrhea inducing STEC, and the third group would be animal-associated STEC (92). STEC strains may be placed into each category based on information on pathogenicity of the specific strain, and the group classification may change as new data become available.

2 1722 MATHUSA ET AL. J. Food Prot., Vol. 73, No. 9 ILLNESSES AND OUTBREAKS ATTRIBUTED TO NON-O157 STEC In the United States, active surveillance of infections attributed to non-o157 STEC began in 2001 (29). The number of non-o157 STEC infections reported in the United States between 2000 and 2005 increased from 171 to 501 cases, suggesting a higher burden of illness than previously thought (1). An increase in testing for Stx in diarrheal cases and for STEC serotypes other than O157 is likely the reason for the increase in incidence of cases and outbreaks attributed to non-o157 STEC (60). Human infections with STEC can occur with ingestion of contaminated food or water or by direct contact with animals. Transmission can also occur through person-toperson contact (53). Illnesses reported due to STEC serotypes other than O157 are on the rise worldwide (78). It is estimated that 20 to 50% of all STEC infections can be attributed to non- O157 strains, but the percentages differ greatly from country to country and among regions within a country (66, 78).Itis estimated that less than 10% of HUS cases in North America are caused by non-o157 STEC strains (52). In Germany, Italy, and the United Kingdom, it is estimated that non-o157 STEC strains have caused 10 to 30% of sporadic cases of HUS (23). Estimating the true percentage of infections caused by non-o157 STEC strains is difficult, because these strains are not routinely subject to testing (78). There is currently no convenient method that can reliably screen for non-o157 STEC strains, which complicates the determination of incidence of disease due to these organisms (78). There have been at least 22 outbreaks attributed to non- O157 STEC strains in the United States since 1990 (19, 38, 51). The sources for the non-o157 STEC strains were not determined in some of these outbreaks, and the vehicles included foods and nonfoods (Table 1). Food vehicles implicated in the outbreaks included milk (26), salad bar (27), punch (21), apple cider (19), and iceberg lettuce (9, 51). Mead et al. (77) estimate that, in the United States, E. coli O157:H7 causes 73,000 illnesses annually and non- O157 STEC strains cause at least 37,000 illnesses annually. Illnesses attributed to non-o157 STEC strains are most frequently reported in the summer months (21). In 1994, there was an outbreak involving postpasteurization contaminated milk, with 11 confirmed and seven suspected cases of illness. Sixteen of the patients developed bloody stools, diarrhea, and abdominal cramps. Isolates from three patients were identified as E. coli O104:H21 and were positive for Stx2. A confirmed case was defined as acute infection with E. coli O104:H21 isolated from patient stool with serological confirmation. Based on a case-control study, one brand of milk was significantly associated with the illnesses. The incident strain could not be isolated from samples taken from the dairy at which the milk was produced (26). In 1999, an outbreak of E. coli O111:H8 associated with ice and salad from a salad bar occurred at a camp in Texas. Of 521 campers interviewed, 58 had symptoms that met the definition of illness: either bloody diarrhea or nonbloody diarrhea accompanied by abdominal cramps and occurring within 14 days of the start of camp. Two patients developed HUS. The meal served on the first night of camp was significantly associated with development of illness. Several possible food items were identified, but two were significantly and independently associated with illness: ice and salad from the salad bar. E. coli O111:H8 was isolated from 2 of the 11 stool specimens submitted by ill patients. PCR was used to determine that both stx 1 and stx 2 were present. No food samples from the indicated meal were available for testing (27). According to a newspaper report that was cited in the review by Eblen (38), E. coli O121:H19 was linked to an outbreak of illness associated with iceberg lettuce from a fast food restaurant in Of 73 people who became ill, three developed kidney failure. Iceberg lettuce was pinpointed by local health officials, because it was the only common food among sickened patients. The iceberg lettuce was not tested. Before the outbreak occurred, the implicated restaurant passed a health inspection, and no issues were discovered when another health inspection was performed during the outbreak (9, 38). To date there are no conclusive epidemiological data that link meat products to non-o157 STEC illness in the United States. In 2006, there were two individual cases of illness in the United States involving ground beef that may have been due to non-o157 STEC, but efforts to pinpoint the source and the pathogen were inconclusive. The first involved a patient who consumed undercooked ground beef. An indistinguishable strain of E. coli O103 was detected using pulsed-field gel electrophoresis (PFGE) from both the patient and leftover ground beef. The original source of E. coli O103 in this case was left undetermined because of possible cross-contamination of a meat grinder (38). In the second case, a patient ill with E. coli O157 provided ground beef samples in which Stx was present but from which no O157 could be recovered. When the ground beef sample was sent to the Centers for Disease Control and Prevention (CDC) for characterization, E. coli O6:H34 was found, but the ground beef could not be confirmed as the source of illness (38). Recently, in 2008, there was an outbreak attributed to E. coli O111:NM in Locust Grove, Oklahoma, that involved 341 cases of illness, 70 hospitalizations, and 1 death. The outbreak was linked to a local restaurant, but the contamination route to the restaurant and/or food source remain undetermined. Cross-contamination of food through handling, surface contact, and storage areas was likely. It was reported that several employees of the restaurant worked on days they experienced diarrhea, but the strain of E. coli was not isolated from employee stool specimens. Clinical specimens were tested for Stx using a Shiga toxin enzyme immunoassay, screened for stx 1 and stx 2 genes using real-time PCR, and typed by PFGE. Six Xbal PFGE patterns were determined among E. coli O111 isolates. These patterns were uploaded to the national PulseNet database. The state investigation of this outbreak found many asymptomatic infections of E. coli O111, in which

3 J. Food Prot., Vol. 73, No. 9 LITERATURE REVIEW ON NON-O157 SHIGA TOXIN PRODUCING E. COLI IN FOODS 1723 TABLE 1. Selected outbreaks of non-o157 STEC in the United States and worldwide Year Serotype(s) present Country (state) No. of persons ill a HUS Source ID method b Reference(s) 1986 O111:H8 Germany 4 1 Undetermined 1990 O111 USA (OH) 55 (5 conf.) Yes Undetermined O104:H21 USA (MT) 18 (conf.) Yes Milk O111:NM Australia 158 (26 conf.) 23 Sausage PCR O121 USA (MT) 40 Unknown Undetermined O121 USA (CT) 11 (conf.) Yes Lake water O111:H8 USA (TX) 56 (conf.) Yes Salad bar O145:H28 Germany 2 No Undetermined O26:H11 Germany 3 3 Undetermined O103 USA (WA) 18 (conf.) Yes Punch O26:H11 Germany 11 No Day care, beef PCR, PFGE O145:H28 Germany 6 1 Undetermined O111 USA (SD) 3 No Day care O26 USA (MN) 4 No Lake water O145:H28 Germany 2 No Undetermined O111 c USA (NY) 212 (conf.) No Apple cider 19, O45:NM, O45:H2 USA (NY) 52 Food handler PCR, PFGE O45 USA Day care O121 USA Day care O121:H19 USA (UT) 73 No Iceberg lettuce O103:H25 Norway 17 (conf.) 11 Lamb sausage MLVA O145, O26 Belgium 12 5 Ice cream PFGE O26 Denmark 20 Beef sausage PCR, PFGE O111 USA (OK) 341 Yes Restaurant PCR, PFGE 83 a conf., confirmed. b PFGE, pulsed-field gel electrophoresis; MLVA, multiple locus variable-number tandem repeat analysis. c Cryptosporidium was also isolated. serum immunoglobulin (Ig) M antibodies to O111 were found in patients that experienced no clinical illness. Of the 135 persons from whom serum or plasma specimens were tested for E. coli O111 IgM antibodies, 66 (49%) were asymptomatic, 8 (6%) had mild illness, 12 (9%) were suspected cases, 22 (16%) were probable cases, and 26 (19%) were confirmed cases (83). In many outbreaks involving non-o157 STEC, the source of infection remains unidentified (78), especially for outbreaks that occurred prior to 2000 (Table 1). In Australia in 1995, an outbreak of HUS was linked to semidry uncooked, fermented sausage contaminated with STEC O111:NM. Sixteen (70%) of the patients required dialysis, and one patient died. Stool samples were screened for genes encoding Stx using PCR; 87% were positive for both stx 1 and stx 2,4% were positive for stx 2 only, and 9% were negative. E. coli O111:NM was isolated from 16 of the stool samples, and other E. coli strains were recovered from three of the patients. Another 62 patients with bloody and nonbloody diarrhea who had consumed the implicated sausage were reported, but E. coli O111:NM was isolated from only 3 of the patients. Of 10 sausage samples taken from patients homes, 8 were positive for stx, and E. coli O111:NM was isolated from 4 (25). In 2007, there was an outbreak of E. coli O145 and O26 in Belgium associated with ice cream produced on a dairy farm. Twelve people became severely ill, with five children developing HUS. E. coli O145 was isolated from stool samples from three HUS patients; from one stool sample, E. coli O26 was also isolated. Stool samples were cultured on sorbitol-containing MacConkey (SMAC) agar, and colonies were identified as E. coli O145 or O26 through biochemical tests, PCR, and an agglutination assay. For the E. coli O145 strains, PCR analysis revealed the presence of stx 1, stx 2, eae, and ehxa (enterohemolysin). PFGE was used to compare the genetic profiles of all STEC strains isolated. Undistinguishable strains of E. coli O145 and O26 were found in stool samples, ice cream samples, and environmental samples collected on the dairy farm (33). In Denmark in 2007, an outbreak of E. coli O26:H11 was associated with organic, fermented beef sausage. Twenty people were involved in the outbreak, with the majority of cases in children. Two unopened samples and two opened samples of sausage tested positive for the infection strain of E. coli O26:H11. Leftover beef used to make the sausage also tested positive for the strain, which was stx 1 positive and eae positive. The reported symptoms of illness were mild, but there was one case involving bloody diarrhea. Several samples of stool also tested positive for other diarrheal pathogens (two Campylobacter species, two Yersinia enterocolitica, one norovirus, and two eae-positive but stx-negative E. coli strains) (43, 44). In the United States, Canada, United Kingdom, and Japan, E. coli O157:H7 is currently the STEC serotype most frequently linked to illness, but in other countries, other STEC serotypes have been associated with disease and outbreaks (21, 38, 42, 99). In Europe, Argentina, Australia, Chile, and South Africa, non-o157 STEC infections are just

4 1724 MATHUSA ET AL. J. Food Prot., Vol. 73, No. 9 as prevalent, if not more prevalent, than E. coli O157:H7 infections, according to some assessments (21, 38, 64, 78, 99, 101). In 1999 in Germany, two-thirds of the STEC infections reported were due to non-o157 STEC. In Germany, the STEC serotype O26 was the second most frequently reported, after O157, and accounted for 20% of all reported STEC infections (101). An unpublished study by Acheson in 2001 reported a similar incidence of O157 (54%) and non-o157 STEC (46%) from clinical stool samples (38). Acheson concluded that certain strains of non-o157 STEC, including O26, O45, O103, O111, and O145, are just as prevalent and clinically significant as E. coli O157 in the United States (38). Worldwide, disease caused by non-o157 STEC is considered an emerging problem (102). A 2009 study done by Hedican et al. (57) and the Minnesota Department of Health compared the characteristics of infections attributed to O157 versus non-o157 STEC. All stool cultures were collected between 2000 and 2006 and were received from two sites in Minnesota, a metropolitan health maintenance organization laboratory and a hospital laboratory that served a small city and a rural area. They found that O157 STEC infections were more likely than non-o157 STEC infections to result in bloody diarrhea (78 versus 54%), hospitalization (34 versus 8%), and HUS (7 versus 0%). They also noted that when only isolates that harbored stx 2 genes were considered, O157 STEC cases were still more likely to result in bloody diarrhea and hospitalizations than the non-o157 STEC cases. Of the non-o157 STEC cases, 74% were represented by just five serotypes, including O26 (27%), O103 (21%), O111 (19%), O145 (5%), and O45 (4%) (57). The incidence of illnesses associated with these serotypes correlated to Minnesota cases differ slightly from percentages seen for the United States and worldwide, indicating that their prevalence may be unique country to country and region to region. Similar information on comparisons of O157 and non-o157 STEC infections, but on a national level, based on FoodNet data, were reported by Gould of the CDC at a public meeting in Washington, DC, in late 2009 (51). It was reported that factors such as age, gender, and seasonality of O157 and non-o157 STEC infections are similar. Gould noted that non-o157 STEC infections are more sporadic than infections of O157 and are correlated with fewer outbreaks. E. coli O157 has a much higher incidence of HUS (6.3% O157 versus 1.7% non-o157), hospitalizations (42% O157 versus 12% non-o157), and deaths (0.6% O157 versus 0.1% non-o157). Another interesting difference was seen between infections of O157 and non-o157 STEC: the incidence of international travel was five times greater for patients with non-o157 STEC infection (51). In 2009, McPherson et al. (76) collected information on serogroup-specific risk factors of STEC infections in Australia from 2003 through Questionnaires were used to collect data on clinical illness, foods consumed, and exposure to environmental sources from individuals from six different jurisdictions in Australia. Interviewees included 43 case patients infected with O157 STEC, 71 case patients infected with non-o157 STEC, and 304 control subjects. Of the non-o157 STEC infected patients, 14 cases could be attributed to O111, 7 cases to O26, and 1 case each to O103, O113, and O172. Infections due to O157 STEC were positively associated with eating at a restaurant or catered event, eating hamburgers, prior use of antibiotics, and family occupational exposure to red meat. There was a negative association between eating homegrown vegetables, fruits, and herbs and O157 STEC infection. Infections due to non-o157 STEC were positively associated with eating at a catered event, eating chicken, meat, or corned beef from a delicatessen, camping, family occupational exposure to animals, and living on or visitation to a farm. For non-o157 STEC infections, there was a negative association to eating pork, eggs, raw and homegrown vegetables, fruits, and herbs (76). PATHOGENESIS OF NON-O157 STEC There is extensive variation within serotypes of STEC in the severity of illness caused, and more than 120 different serotypes have been associated with illness (78, 92). In the United States between 1983 and 2002, the six most commonly occurring serotypes of non-o157 STEC associated with disease were, in descending order, O26, O111, O103, O121, O45, and O145 (3, 21, 53). According to preliminary data presented by Gould (51) in 2009, these six serotypes made up 82% (n ~ 803) of FoodNet human isolates of non-o157 STEC between 2000 and STEC infection in humans may result in no illness or mild to severe symptoms and, in some cases, may lead to more severe disease such as hemorrhagic colitis, HUS, and thrombotic thrombocytopenic purpura (102). Twardon et al. (99) speculate that fewer than 10 bacterial cells of E. coli O26 are able to infect humans; however, no data were provided by the authors for this postulation. Gyles (53) suggested it to be fewer than 50 cells to a few hundred organisms based on information on E. coli O157. It is estimated that the infectious dose for non-o157 STEC may be higher than that for E. coli O157, which has been shown to be 10 to 100 cells (45). Anarticle by Paton et al. (85) on an outbreak of HUS in dry fermented sausage that was contaminated with non-o157 STEC found low levels (,100 CFU/g) of E. coli present in sausages eaten by ill patients. In this outbreak, E. coli O111:NM was indicated as the causative agent for illness. STEC O111:NM was isolated from both patients and reserved sausage samples. PCR was used to determine that only 0.4 to 1.4% of E. coli isolated from the sausage were STEC. Of the STEC strains isolated from the sausage on MacConkey agar, generally less than 10% were identified as STEC O111:NM by colony immunoblotting. The authors suggest that there may have been as little as one cell of STEC O111:NM per 10 g of the sausage, which would indicate a low infectious dose for this organism in certain foods (85). Characteristics of disease related to non-o157 STEC. The incubation period of STEC is usually 3 to 4 days, but can be as long as 5 to 8 days or as short as 1 to 2 days. Initial symptoms include crampy abdominal pain, a short-lived fever, and nonbloody diarrhea. Vomiting can

5 J. Food Prot., Vol. 73, No. 9 LITERATURE REVIEW ON NON-O157 SHIGA TOXIN PRODUCING E. COLI IN FOODS 1725 occur during the diarrhea stage of illness, but is observed in only about half of the patients. In 1 to 2 days, diarrhea may become bloody with increased abdominal pain, and this may last for up to 10 days. Most cases of infection with STEC will resolve without sequelae, but 10% of patients, most commonly young children (younger than 10 years old) and the elderly, may experience the development of HUS (44, 53, 78). Hemorrhagic colitis is characterized by severe abdominal cramps and watery, then grossly bloody, diarrhea with little to no fever. HUS was initially described in 1955 and linked to Shiga toxin producing Shigella dysenteriae. HUS is characterized by acute renal failure, thrombocytopenia, and microangiopathic hemolytic anemia. Stx is responsible for damage to both intestinal and renal tissue (78). Patients suffering from thrombotic thrombocytopenic purpura experience the same clinical symptoms as HUS, accompanied by fever and formation of thrombi that may lead to severe neurological disorders (102). Bloody diarrhea is more common with E. coli O157:H7 than with non-o157 STEC. It is estimated that O157 causes at least 80% of HUS cases associated with STEC infections, while less than 10% of HUS cases can be attributed to non- O157 STEC (5, 52, 71). Some Shiga toxigenic non-o157 E. coli, including serotypes O26 and O111, have been associated with hemorrhagic colitis and HUS (78, 79). Some cases of illness from infection with non-o157 STEC have resulted in symptoms similar to those for E. coli O157:H7 (53, 78). Although in some reported cases the degree of illness due to non-o157 STEC has been just as severe as illness due to E. coli O157:H7 (78), in most of the reported cases it appears that the overall illness associated with non-o157 STEC is less severe than illness due to E. coli O157:H7, and fewer hospitalizations are reported (60, 79). The disease process for STEC first requires the organism to overcome host defense mechanisms and establish itself in the intestine. Acid resistance of STEC is important for its survival in the harsh acidic environment of the gastrointestinal tract. STEC strains that possess the eae (E. coli attaching and effacing or intimin) gene can produce products involved in cell attachment. During attachment, eae-positive STEC strains form an AE lesion on intestinal epithelial cells. The AE lesion results in structural changes in the epithelial cells such as loss of microvilli, pedestal formation, and accumulation of cytoskeletal proteins, allowing adherence of the bacteria to the host cell surface. After attachment, Stx is absorbed into the host cell through a transcellular pathway (78). STEC infection appears to be localized without septicemia, but the toxin produced is absorbed from the intestine and causes the systemic effects of the disease (53). Translocation of the toxin into the bloodstream is believed to be aided by damage of the intestinal epithelium by lipopolysaccharide or the toxin itself (78). Virulence factors. Over 200 serotypes of E. coli can produce Stx, but only about 50 of these serotypes have been associated with bloody diarrhea or HUS in humans (78). Shiga and Shiga-like toxins can be produced by several other bacilli, including Enterobacter cloacae, Citrobacter freundii, and Aeromonas hydrophila (79, 99). The ability of an E. coli strain to produce Stx alone does not automatically confer pathogenicity without other virulence factors (78, 100). There are two types of Stx, Stx1 and Stx2. Stx1 is identical to the toxin produced by Shigella dysenteriae type 1 (53). Variants of stx genes have been reported, such as stx 1a, stx 1b, stx 1c, stx 1d, stx 2a, stx 2b, stx 2c, stx 2d, stx 2e, stx 2f, and stx 2g (13, 64). Certain variants, including stx 2a and stx 2c are more likely to be associated with hemorrhagic colitis and HUS (13). Several other variants of Stx show no clinical significance (53, 78). A single STEC strain may express Stx1, Stx2, or both toxins (78, 79). Expression of Stx2 has been associated with a higher risk for developing HUS, especially when the organism is also eae positive (21, 52, 66). It has been suggested that E. coli producing Stx2 is involved in most HUS cases because E. coli O157:H7 strains that are isolated from patients with HUS usually produce only Stx2 or both Stx1 and Stx2. E. coli producing only Stx1 has not been isolated from patients with HUS (90). Stx2 has also been shown to be 1,000 times more toxic for human renal microvascular endothelial cells than Stx1, which may be due to major differences in crystal structure between the two toxins (53). Boerlin et al. (18) found a strong statistical association between non-o157 STEC serotypes O26, O103, O111, and O145 expressing stx 2 and the severity of human disease. They determined that possession of the stx 2 gene makes the organism significantly more likely to cause serious disease, including bloody diarrhea and development of HUS (18, 42). Friedrich et al. (47) used PCR to screen 626 STEC isolates from stool samples collected in Germany from 1996 to 2000 to determine serotype and detect the presence of stx 1, stx 2, and stx 2 variants, and the eae gene. Serotypes of non-o157 STEC were isolated from patients with HUS, including O26, O103, O111, and O145. The most frequently isolated non-o157 STEC serotype from patients with HUS was O26. Identical strains of non-o157 STEC were isolated from both asymptomatic patients and those with diarrhea. STEC strains O26:H11/NM, O145:NM, O103:H2/H18/NM, and O111:NM were isolated from patients with HUS, patients with diarrhea but no HUS, and asymptomatic patients. The stx 2 variants detected included stx 2c, stx 2d, and stx 2e, with stx 2c as the most frequent variant, found in 148 (23.6%) of the 626 isolates. Variants stx 2d and stx 2e were eae negative and not detected in any of the non-o157 STEC serotypes of interest. Of the 626 isolates, there were 87 non-o157 STEC isolates harboring stx 2 and nine carrying stx 2c. Friedrich et al. (47) found that of 87 isolates of non-o157 STEC that did harbor stx 2, which included O26, O103, O121, and O145, 83 (95.4%) carried the eae gene. Of the non-o157 STEC isolates harboring the stx 2c variant 33.3% were eae positive. Of the 28 O157 isolates (from the pool of 626 isolates) with the stx 2c variant, 100% were eae positive. The authors concluded that STEC stains harboring the stx 2c variant are able to cause HUS, but isolates with either the stx 2d or stx 2e variant result in milder illness unlikely to produce sequelae

6 1726 MATHUSA ET AL. J. Food Prot., Vol. 73, No. 9 (47). Another study by Beutin et al. (15) found that high production of Stx2e by human-associated STEC strains did not result in diarrheal disease. Strains harboring stx 2e genes were negative for eae and ehxa genes. The authors concluded that Stx2e-producing strains are not good colonizers of the human intestine, probably due to the lack of receptors on human enterocytes, and that strains producing only Stx2e are not able to cause severe disease (15). Stx is encoded by phages inserted into the E. coli chromosome (53, 78, 79). Stx is made up of the basic A-B subunit structure. The B pentamer of the toxin binds to a specific receptor, globotriaosylceramide, on the intestinal cell surface, permitting internalization. The Stx2e variant, which is associated with disease in swine, uses globotetraosylceramide as its receptor. The toxin molecule is taken up into the cell through receptor-mediated endocytosis. The membrane vesicle containing toxin may fuse with lysosomal vesicles, resulting in destruction of the toxin, or may be transported to the Golgi apparatus and endoplasmic reticulum. The A subunit of the toxin protein possesses enzymatic activity that cleaves a specific adenine base from the 28 S rrna, inhibiting protein synthesis (78). This can result in apoptosis, programmed cell death, due to ribocytotoxic stress response (53). Important virulence factors include expression of the eae gene and the hly (hemolysin) gene (53). Another hemolysin gene present in some STEC strains, ehxa, is correlated with virulence of EHEC (64). The eae gene expresses intimin, also called the eae protein, which is important in the production of AE lesions in the intestine. A pathogenicity island called the locus of enterocyte effacement (LEE) encodes proteins necessary for the formation of the AE lesion. LEE encodes for a type III secretion apparatus, a protein translocation system, and an adherence system that consists of the eae protein, which is the outer membrane protein, and its receptor, translocated intimin receptor. The translocated intimin receptor protein becomes inserted into the host cell outer membrane where it acts as the receptor for the eae protein on the bacterial cell surface (53). These genes are more common in STEC strains that are correlated to illness, but strains lacking these genes reportedly have caused clinical illness (79, 80). E. coli O113:H21 does not possess the LEE pathogenicity island but has been the cause of sporadic illness and outbreaks. The illness cases attributed to E. coli O113:H21 were reported to be just as severe as those caused by E. coli O157:H7 (80). Fluid secretion associated with diarrhea occurs with death of absorptive villus tip intestinal epithelial cells by Stx. It is believed that a STEC strain s ability to produce AE lesions is sufficient to cause nonbloody diarrhea, but Stx production is essential for the development of bloody diarrhea and hemorrhagic colitis. Expression of hemolysin is widely distributed among non-o157 STEC strains and causes lysis of red blood cells in vitro. Approximately 90% of all STEC strains possess genes encoding hemolysin (78). Other toxins produced by STEC may play a role in the etiology of human disease. Cytolethal distending toxin is produced by a few eae-negative STEC strains that have been associated with disease (17, 53). Subtilase cytotoxin is also produced by an eae-negative STEC strain, O113:H21, and the gene is detected in many other STEC strains (53, 80). Newton et al. (80) suggest that subtilase cytotoxin emerged as a virulence factor in the absence of LEE, and this toxin likely plays a role in the progression of severe disease. Although E. coli O113:H21 is eae negative, it has been associated with HUS, which further complicates the definition of pathogenicity for these organisms as a whole (11). Several other gene products have been suggested to have possible virulence roles for STEC, including adhesins, such as the VTEC auto-agglutinating adhesin (saa), proteases, iron acquisition systems, lipopolysaccharide, and flagellin (53, 64). The virulence of the subtilase cytotoxin of LEE-negative STEC is partially dependent on flagellin, showing that some of these products may work with other virulence factors to impart pathogenicity (80). Given that there is no satisfactory animal model that mimics the disease in humans, it is difficult to determine how significantly these factors contribute to virulence, if at all (53, 102). Much of the research on non-o157 STEC has focused on the serotype O26. A study by Zhang et al. (103) examined the molecular characteristics of 55 STEC O26 strains collected in Germany and the Czech Republic between 1965 and Virulence genes that were found in O26, such as hlya, catalase peroxidase (katp), and a serine protease (espp) that cleaves human coagulation factor V, are also found in STEC O157. They found that all the STEC O26 strains possessed a high-pathogenicity island that O157 does not that contains genes encoding pesticin receptor ( fyua) and a siderophore, called yersiniabactin. An interesting discovery was made regarding the type of stx gene contained by STEC O26 strains over time. Through PCR analysis, they found that 16 of 18 strains collected from 1965 to 1996 expressed stx 1 alone, with only two additional strains expressing stx 1 after The 37 strains that expressed stx 2 alone or in combination with stx 1 were isolated between 1995 and These results indicate that there was a shift from stx 1 to stx 2 expression among STEC O26. Of the 55 STEC O26 isolates, 16 clonal subgroups were determined by PFGE, showing the diversity of this serogroup. Using PFGE, Zhang et al. (103) discovered the emergence of a new clonal subgroup, A, with a set of unique virulence genes, including stx 2, hlya, and the etp (EHEC type II secretion pathway) cluster. Originally found only in STEC O157, the etp gene cluster, which encodes a type II secretion system which allows for extracellular excretion of proteins, was seen in several O26 strains with identical plasmid profiles, and only after 1995 (94, 103). Four clusters of outbreaks were linked to this subgroup A of STEC O26. The STEC O26 of subgroup A were shown to have a high pathogenic potential for humans, so any disease outbreaks correlated to these organisms should be closely monitored by public health authorities (103).

7 J. Food Prot., Vol. 73, No. 9 LITERATURE REVIEW ON NON-O157 SHIGA TOXIN PRODUCING E. COLI IN FOODS 1727 A shift in the expression of virulence factors and emergence of virulence strains among STEC strains is also suggested by evidence for O157. E. coli O157:H7 was first reported as a cause of foodborne illness in 1983 by Riley et al. (89), after investigating outbreaks in 1982 involving undercooked ground beef. Before these incidents, this serotype was almost never isolated (10, 78, 89). After the link between E. coli O157:H7 and foodborne illness was made, laboratories around the world reviewed all E. coli strains collected between 1973 and Only one E. coli O157:H7 was isolated by the CDC laboratories, out of 3,000 serotyped isolates, and the Public Health Laboratory in the United Kingdom also found just one O157:H7 isolate out of 15,000 serotyped isolates. Only six O157:H7 isolates were found out of 2,000 isolates from patients with diarrhea by Canada s Laboratory Centre for Disease Control. Although illness from O157:H7 STEC could have been hidden in the overall burden of illness from EHEC, the limited isolation of O157:H7 prior to 1982 suggests that the presence of this serotype may have increased since that time instead of having previously been missed (78). SOURCES FOR STEC AND DISTRIBUTION Ruminants, especially cattle, are an important reservoir for STEC strains (10, 42, 53, 61). STEC strains have been recovered from cattle, sheep, goats, pigs, cats, deer, horses, dogs, birds, and flies (53, 78, 81). In North America, cattle are the significant reservoir for STEC strains, but in other countries such as Australia, sheep are the most important carrier (53). In the United States, beef carcass processing is the main area targeted for interventions to reduce contamination (53). Generally, non-o157 STEC strains are found in cattle at a much higher prevalence than E. coli O157 (10). Ina study by Beutin et al. (12), STEC strains were isolated in 63.2% of feces samples from cattle in one herd (n ~ 19) over a period of 6 months. Of the 33 serotypes of STEC isolated, none were O157. Stx was detected by the Vero cell test, and the presence of stx 1 and stx 2 was determined by colony blot hybridization with digoxigenin-11-dutp labeled gene probes. Almost all of the STEC serotypes produced Stx2; only one strain produced Stx1. All the strains but one were negative for the eae gene (12). Most cattle colonized by STEC are asymptomatic due to the absence of the globotriaosylceramide receptor in their intestinal cells that is specific for Stx proteins (99). Rates of colonization of STEC in cattle have been found to be as high as 60%, but are more typically in the range of 10 to 25% (12, 78). In 2007, Hussein estimated that the prevalence of non-o157 STEC in dairy cattle may be as high as 74% (61, 63). Non-O157 STEC strains isolated from dairy cattle belonged to 152 different serotypes, with an estimated 49% of these being pathogenic when defined as a STEC that produces one or more of the following virulence factors: Stx1, Stx2, hlya, EHEC-hlyA, and/or intimin (61). Another study by Hussein on non-o157 STEC in cattle at slaughter found prevalence rates of 2.1 to 70.1% (62). The rate is variable and thought to depend on environmental factors and management practices (62). A 2003 study by Barkocy- Gallagher et al. (6) found the prevalence of non-o157 STEC in beef cattle at the time of slaughter to be between 13.9 and 27.1% depending on the season. Studies have shown that there is a higher frequency of fecal shedding of STEC by cattle in warmer months than colder months, with a correlating higher incidence of human illness in summer months (53, 78). Age may also play a role in fecal shedding of STEC in cattle, with the lowest shedding rates in calves before weaning, the highest rates in the postweaning period, and intermediate rates in adult cattle (53). Studies have shown that many bovine isolates of non-o157 STEC are less likely to carry important virulence factors other than stx, such as eae and hlya, in comparison to human isolates, indicating that these organisms may be less virulent (2, 18, 69). Over 435 different serotypes of STEC have been recovered from cattle, and more than 470 STEC serotypes have been isolated from humans, with great overlap. Only a fraction of these STEC serotypes are capable of causing illness. Of human STEC isolates, fewer than 10 O groups are responsible for the majority of illnesses (53, 78). FOODS ASSOCIATED WITH NON-O157 STEC Foods from which non-o157 STEC strains have been isolated and/or associated with illness include sausage, ice cream, postpasteurization contaminated milk, punch, and iceberg lettuce (21, 38, 44, 101). Bettelheim (10) suggested that many of the foods from past outbreaks associated with illness due to E. coli O157 were likely to also contain non- O157 strains, but that only O157 was sought. Studies have screened grocery items, such as delicatessen salad, raw milk, raw beef, minced meat, pork, lamb, poultry, fish, shellfish, and cheese, and were able to detect non-o157 STEC at different frequencies (Table 2) (35, 38, 86, 88, 91). A study in the United States by Samadpour et al. (91) sampled raw meat, poultry, and seafood samples for stx genes using DNA probes and found them in samples of beef (23%), veal (63%), pork (18%), chicken (12%), turkey (7%), lamb (48%), fish (10%), and shellfish (5%). After determination of serotypes in the samples, they found that several different non-o157 strains, but no O157 strains, were present. Comparisons of electrophoretic typing patterns found that the isolates had a close relationship to isolates from human and animal disease cases (91). A 2002 study by Arthur et al. (2) looked at the prevalence of non-o157 STEC on beef carcasses in U.S. processing plants and found that 53.9% were positive for at least one strain prior to evisceration. This level was reduced to only 8.3% following processing interventions, including steam vacuum, hot water, organic acids, and steam pasteurization (2). Studies from around the world have reported differing postprocessing prevalence of non-o157 STEC on beef carcasses, but this may be due to different STEC isolation methodologies (69). In 2006 in France, Perelle et al. (86) screened samples of raw milk (n ~ 205) and minced meat (n ~ 300) using PCR-ELISA and found the prevalence of STEC-positive

8 1728 MATHUSA ET AL. J. Food Prot., Vol. 73, No. 9 TABLE 2. Occurrence of STEC in foods Product tested % positive all STEC a % positive non-o157 STEC a Test methods Reference Beef 23 DNA probes for stx genes 91 Veal 63 Pork 18 Chicken 12 Turkey 7 Lamb 48 Fish 10 Shellfish 5 Beef carcasses PCR targeting stx genes and colony 2 Treated beef carcasses hybridization for STEC, serotyping Raw milk b PCR-ELISA targeting stx genes, multiplex 86 Minced meat b real-time PCR Beef 4 Not reported PCR targeting stx genes, API testing for E. 88 Cheese 1 Lysate from FSIS archived ground beef samples coli, serotyping Not reported 1.3 c PCR targeting O-antigen, stx, and eae genes a Results from PCR screening tests in which an isolate was not obtained for confirmation testing are presumptive positive, not confirmed positive. b These values represent the fraction of samples that tested PCR positive for one or more of the serotypes O26, O103, O111, O145, and O157. c This value represents the fraction of samples that tested PCR positive for the stx and eae genes, as well as positive for one of the six serotypes (i.e., O26, O103, O121, O45, O111, or O145). 50 samples was 17.4%. Of the 205 raw milk samples, 43 (21%) were positive for STEC. Of the 300 minced meat samples, 45 (15%) were positive for STEC. Of the 88 positive STEC samples, 74 (84%) were confirmed positive for stx using a 59-nuclease PCR assay. When multiplex real-time PCR was used to screen for specific serotypes, including O26, O103, O111, O145, and O157, they were found in 2.6% of the raw milk samples and 4.8% of the minced meat samples. Of the 45 samples of STEC-positive minced meat, 7 included serotype O145 and 2 had serotype O103. Of the 43 samples of STEC-positive raw milk, 9 had serotype O145, 2 had serotype O103, and 1 had serotype O26. Many of the samples had more than one of the specific STEC serotypes sought. The incidence of E. coli O157 in minced meat and raw milk was 1%, which is in line with worldwide values of incidence; but the incidence of E. coli O145 was surprisingly higher, 3% of the samples (86). Survey data were converted to most-probable-number counts following the previously proposed Halvorson and Ziegler (55) calculation and showed that the contamination was only 1 to 2 most-probable-number STEC cells per kg of sample. Perelle et al. (86) determined that the contamination of the beef and raw milk samples was very low and that the potential risk of consumer infection by these strains from the samples is likely very minor. Another French study by Pradel et al. (88) looked at the prevalence of STEC in beef samples and cheese samples. At least one strain of STEC was found in 4% of beef samples and 1% of cheese samples. The investigators screened 220 STEC isolates including isolates of the beef and cheese samples as well as isolates from stool samples from cattle and hospitalized patients. Of the STEC isolates, only 5% carried the eae gene, 15% harbored the stx 1 gene, 53% harbored the stx 2 gene, and 32% had both genes. The authors concluded that the majority of the STEC isolates from beef samples and cheese samples were unlikely to be pathogenic in humans based on the lack of virulence characteristics associated with clinical isolates (88). In early 2010, results of PCR screening tests for the stx, eae, and the O26, O103, O121, O45, O111, and O145 genes in U.S. Food Safety and Inspection Service (FSIS) archived lysates of ground beef samples were reported (50). PCR testing of 224 E. coli O157:H7 sample enrichments yielded the following percent positives for each genetic target: O26 (3.1%), O103 (3.6%), O121 (1.8%), O45 (20.1%), O111 (0.4%), and O145 (0.0%) (50). These samples had previously tested negative for E. coli O157:H7. It was noted that E. coli O111 and O145 did not grow well in the E. coli O157:H7 enrichment broth. Among the 224 samples, it was found that only 1.3% of sample enrichments were positive for all three factors: one of the top six serotypes, stx, and eae (50). Furthermore, these PCR screening tests yielded presumptive-positive results. The archived lysates of ground beef samples contain lysed cells from sample enrichment and, thus, isolates are unavailable for confirmation testing. The information presented above suggests that using the results of serotype screening alone could be misleading if it is assumed that all positive results represent pathogenic non-o157 STEC. If appropriate virulence factors are not targeted as part of food sample screenings, it will be difficult to know whether or not identified STEC strains are pathogenic. DETECTION AND IDENTIFICATION METHODS Currently there exists no standard cultural method to identify non-o157 STEC, but many laboratories worldwide are attempting to develop a method (11). The non-o157 STEC serotypes of interest differ from country to country,

9 J. Food Prot., Vol. 73, No. 9 LITERATURE REVIEW ON NON-O157 SHIGA TOXIN PRODUCING E. COLI IN FOODS 1729 and there is no widely accepted selective-differential media available to determine the individual serotypes. Cultural methods with selective and differential media. The current cultural method for isolation of E. coli O157 is based on the inability of this organism to ferment sorbitol, although a few strains are able to ferment sorbitol (53). Most E. coli strains are capable of fermenting sorbitol. Using SMAC to isolate suspected E. coli will result in clear colonies for E. coli O157. Bright pink to mauve colonies indicate sorbitol-fermenting organisms, which include most non-o157 and other common fecal microflora. Grampositive microorganisms will be inhibited on this medium by crystal violet and the bile salts mixture in the formulation. Differentiation of non-o157 STEC colonies on SMAC is not possible (74). Researchers have been working on developing media to detect non-o157 STEC. In 2008, Possé et al. (87) developed a set of novel differential media for the isolation and confirmation of non-o157 STEC strains (O26, O103, O111, and O145) from food and feces. The first medium is based on a mixture of carbohydrate sources, b-d-galactosidase activity, and selective reagents that result in color-based differentiation of the four specified non-o157 STEC strains. The composition of this differential medium starts with MacConkey agar base and is supplemented with sucrose, sorbose, bile salts, 5-bromo-4-chloro-3-indolyl-b-D-galactopyranoside (X-gal), isopropyl-b-d-thiogalactopyranoside, novobiocin, and potassium tellurite. The growth of the four different non-o157 STEC serotypes on this medium produces different colored colonies. STEC O26 colonies appear bright red to dark purple, O103 and O111 colonies are blue-purple, and O145 colonies are green. The second group of media is for confirmation of serotypes based on specific carbohydrate utilization. These agars contain phenol red broth base supplemented with dulcitol, L-rhamnose, D- raffinose, or D-arabinose (87). Unpublished studies in our laboratories (40) using the chromogenic agars described by Possé et al. (87) showed that while some of the serotypes may present the colony color as indicated in that publication, the color and the entire aspect of the colonies may change as a function of incubation time, how crowded or isolated the colonies are, or the medium or food matrix from which they are isolated. When the colonies are small and crowded they may look significantly different than when they are well isolated and larger. For example, O26 streaked on the chromogenic agar from a pure culture in tryptic soy broth grew either as small purple colonies with a darker center or large pink colonies with a darker center and circled by a blue-grayish edge; the colonies were also shiny and smooth. For the same serotype inoculated in irradiated ground beef and streaked on the same chromogenic agar, resulting colonies were small, blue, rough, granulated, flat, and dry, or were large, pink, shiny, and smooth as seen from colonies isolated from tryptic soy broth. When six STEC serotypes were streaked on the chromogenic agar it was difficult to differentiate between serotypes (40). Another unpublished study conducted in 2009 used Rainbow agar to detect non-o157 STEC serotypes (49). Different color reactions on the Rainbow agar indicate which serotype may be present. Serotypes O26, O103, and O121 may appear pink or magenta on this agar; O45, O111, and O157 may appear gray, light blue, or light purple; and O145 may appear dark blue (49). Hiramatsu et al. (58) have developed a selective medium specific for E. coli O26 using rhamnose, called rhamnose-macconkey (RMAC). The study showed that all O26 strains, 31 total, were able to ferment rhamnose, while 108 other STEC strains could not. All STEC O26 colonies were colorless on RMAC, while the vast majority of other STEC (89 of 93 strains, which included serotypes O157 and O111) produced red colonies. Most non-stec strains (50 of 59 strains) were unable to grow on RMAC. Other studies have also shown that the O26 serotype is unique in that it is able to ferment rhamnose, a characteristic that could be used in its differentiation from other STEC serotypes (24, 58). Another indicator for STEC is the production of enterohemolysin, and a medium which detects enterohemolysin-producing organisms has been developed (11). Catarame et al. (24) reviewed many commercially available media for their ability to recover STEC serotypes O26 and O111 from minced beef. Different combinations of enrichment procedures and incubation time and temperature were tried, as well as novel media formulated with a range of selective antibiotics and carbohydrates. Tryptic soy broth containing cefixime and vancomycin was used as the enrichment medium for both serotypes, with the addition of potassium tellurite to optimize the enrichment for serotype O26. A couple of O111 strains were sensitive to potassium tellurite. Catarame et al. found that the optimum recovery of STEC O26 was on MacConkey agar modified by replacing lactose with rhamnose and supplemented with cefixime and potassium tellurite. Suspect colonies of O26 appear brown or red on this medium. STEC O111 was best recovered on chromocult agar supplemented with cefixime, cefsulodin, and vancomycin; colonies indicative of O111 appear purple. Before plating on selective agars, O26 and O111 cells were concentrated using immunomagnetic separation (IMS). The authors concluded that the serotype-specific enrichment broth, IMS extraction, and selective agar with serological and biochemical confirmation testing are effective methods for the recovery of these STEC serotypes (24). Immunological methods. IMS and plating is a highly sensitive method currently used to detect E. coli O157:H7 and other organisms. In this method, microscopic, ironcored beads are coated with specific antibodies to E. coli O157:H7, allowing for the organism s capture when a sample is passed over the beads. The bead-cell complexes are then captured using a magnetic concentrator. The cells can be removed from the beads and plated on agar, such as SMAC (100). IMS has also been used to detect STEC in fecal samples from animals shedding low numbers of STEC. Currently IMS is being used in the detection of O26, O103, O111, and O145 STEC (11, 53). Different kits for detection of STEC have been developed, but not all have been validated (11).