Prevalence and Characterization of Non-O157 Shiga Toxin-Producing Escherichia coli on Carcasses in Commercial Beef Cattle Processing Plants

Abstract

Beef carcass sponge samples collected from July to August 1999 at four large processing plants in the United States were surveyed for the presence of non-O157 Shiga toxin-producing Escherichia coli (STEC). Twenty-eight (93%) of 30 single-source lots surveyed included at least one sample containing non-O157 STEC. Of 334 carcasses sampled prior to evisceration, 180 (54%) were found to harbor non-O157 STEC. Non-O157 STEC isolates were also recovered from 27 (8%) of 326 carcasses sampled after the application of antimicrobial interventions. Altogether, 361 non-O157 STEC isolates, comprising 41 different O serogroups, were recovered. O serogroups that previously have been associated with human disease accounted for 178 (49%) of 361 isolates. Although 40 isolates (11%) carried a combination of virulence factor genes (enterohemorrhagic E. coli hlyA, eae, and at least one stx gene) frequently associated with STEC strains causing severe human disease, only 12 of these isolates also belonged to an O serogroup previously associated with human disease. Combining previously reported data on O157-positive samples (R. O. Elder, J. E. Keen, G. R. Siragusa, G. A. Barkocy-Gallagher, M. Koohmaraie, and W. W. Laegreid, Proc. Natl. Acad. Sci. USA 97:2999-3003, 2000) with these data regarding non-O157-positive samples indicated total STEC prevalences of 72 and 10% in preevisceration and postprocessing beef carcass samples, respectively, showing that the interventions used by the beef-processing industry effected a sevenfold reduction in carcass contamination by STEC.

Shiga toxin-producing Escherichia coli (STEC) has been implicated as the causative agent in several human diseases (22, 27). These diseases range from mild diarrhea to very severe and life-threatening conditions, such as hemolytic-uremic syndrome (HUS). Enterohemorrhagic E. coli (EHEC) refers to the subset of STEC strains found to cause human disease. The STEC strain most frequently associated with clinical disease in the United States is serotype O157:H7 (20, 22). However, several other serotypes (O26, O103, O111, O113, and O121) are commonly associated with severe disease outbreaks and, in some countries, are isolated from clinical cases more often than O157 (1, 22). It has been estimated that E. coli O157:H7 strains cause two-thirds of the human EHEC infections in the United States, with the remaining cases attributed to the non-O157 STEC population (20). This estimation may be biased toward O157 because several clinical laboratories only screen for E. coli O157:H7, leaving non-O157 EHEC-associated diseases underreported (20).

Cattle are considered the primary reservoir of both O157 and non-O157 STEC bacteria (4). Cattle frequently carry STEC without suffering from any pathological symptoms (6, 42). Several human infections have been linked to STEC present in undercooked ground beef (22, 36). During processing, fecal contamination of the carcass or transfer of bacteria found on the animal hide to the carcass can facilitate transmission of STEC into the food supply (2, 3, 13).

The complete list of bacterial virulence determinants necessary for STEC to cause EHEC-related disease is not known. However, the Shiga toxin is a key factor in pathogenesis (1, 27). Two classes of Shiga toxin have been identified. Shiga toxin 1 (Stx₁) is 98% homologous to the Stx produced by Shigella dysenteriae type 1 (22, 27). Stx₂ is approximately 60% homologous to Stx₁ and is antigenically distinct (22, 27). Recent studies have indicated that strains possessing stx₂ are potentially more virulent than strains carrying stx₁ or even strains carrying both stx₁ and stx₂ (11, 22, 23, 32, 39). Other virulence factors, such as intimin (eae) and EHEC hemolysin (hlyA), are thought to enhance pathogenicity but are not required for strains to cause severe disease, including HUS (1, 8, 22).

The objective of this study was to determine the prevalence and virulence profiles of non-O157 STEC on beef carcasses at commercial processing plants in the United States during the summer months. Such data will contribute to the establishment of baseline prevalence levels of non-O157 STEC. The samples used for this study were originally collected by Elder et al. for a survey of O157:H7 prevalence (13). Extrapolation of the data from both studies provides an indication of the total STEC prevalence in U.S. beef-processing plants.

MATERIALS AND METHODS

Carcass samples.

Sponge samples were collected by Elder et al. from beef cattle carcasses during processing (13). Briefly, individual carcasses were followed through processing and sampled at various points. Speci-Sponges (Nasco, Fort Atkinson, Wis.) moistened with Butterfield's phosphate diluent and 0.1% Tween 20 were used to sample carcasses just prior to evisceration (preevisceration) and after all antimicrobial interventions when the carcasses were in the cooler (postprocessing). Lots of 35 to 85 animals were sampled. A lot is defined as cattle from a single source and maintained as a group throughout processing. Twenty percent of the carcasses in each lot were sampled. Antimicrobial interventions varied between plants and included steam vacuuming, hot water washing, organic acid washing, and steam pasteurization.

Bacterial culture.

Beef carcass sponge samples were enriched by incubation with 90 ml of Brilliant Green Bile 2% (40 g/liter) for 10 h at 37°C (13). Following enrichment, 1 ml of culture was added to 250 μl of 50% glycerol and stored at −70°C. The enrichments were stored frozen for approximately 2 years prior to initiation of this study.

PCR screening of samples.

Samples that were not previously determined to be positive for E. coli O157:H7 (13) were screened for the presence of stx₁ and stx₂ by PCR. The procedure used is a modification of that described by Paton and Paton (24). Briefly, a 50-μl sample was removed from the frozen enrichment and centrifuged for 1 min at 12,000 × g in a microcentrifuge and the supernatant was discarded. The pellet was resuspended in 200 μl of sterile water and centrifuged again, and the supernatant was discarded. The pellet was resuspended in 50 μl of sterile water and boiled for 15 min. The lysate was spun briefly, and 2 μl was used as template DNA in a 30-μl PCR.

Colony hybridization.

All samples previously determined to be E. coli O157:H7 positive (13), as well as samples determined by PCR to contain either stx₁ or stx₂, were assayed by colony hybridization for STEC detection. Samples from the frozen enrichments were serially diluted in buffered peptone water and plated onto tryptic soy agar in 182-mm-diameter petri dishes to achieve a colony count of ∼2,000 CFU per plate. The plates were incubated for 11 h at 37°C and then transferred to 4°C for 30 min. The colonies were blotted to Hybond N+ membranes (Amersham Pharmacia Biotech, Piscataway, N.J.), and hybridization (60°C) and detection were carried out by using the ECL Random Prime Labeling and Detection kit (Amersham) and following the manufacturer's instructions. The membranes were hybridized with 90 ng of both stx₁ and stx₂ probes in 50 ml of hybridization buffer. The DNA probes for random prime labeling were derived via PCR from E. coli O157 strain ATCC 43895 by using the following primers: stx₁-F (bp 77 to 102 of Sakai-VT1; 5′-CCTTAGACTTCTCGACTGCAAAGACG-3′), stx₁-R (bp 894 to 913 of Sakai-VT1; 5′-CCAGAGTGGATGAATCCCAC-3′), stx₂-LP43 (bp 57 to 78 of EDL933; 5′-ATCCTATTCCCGGGAGTTTACG-3′), and stx₂-LP44 (bp 622 to 643 of EDL933; 5′-GCGTCATCGTATACACAGGAGC-3′) (9).

Isolation and characterization of STEC.

The developed films from the colony blots were aligned with the original plates, and up to three suspect colonies per sample were patched in a grid on tryptic soy agar. The patch plates were again colony blotted to identify those colonies that were correctly picked. The positive patches were then streaked for isolation on sorbitol MacConkey agar (SMAC). From each SMAC streak plate, two colonies were picked and screened for the EHEC genes stx₁, stx₂, eae, and EHEC hlyA by multiplex PCR (24). Any colony possessing stx₁ or stx₂ was assayed for expression of the O157 antigen by latex agglutination with the Dryspot E. coli O157 Latex Test (Oxoid, Basingstoke, England). Colonies giving a positive agglutination reaction were discarded. All non-O157, stx-containing isolates were confirmed to be E. coli by biochemical identification with the Sensititre Gram-Negative Autoidentification (AP80) system (Accumed International, Westlake, Ohio) or API 20E strips (bioMérieux, Hazelwood, Mo.).

ETEC and EPEC PCR.

Selected strains were screened by PCR for virulence attributes commonly associated with enterotoxigenic E. coli (ETEC; heat-stable and heat-labile toxins) and enteropathogenic E. coli (EPEC; bundle-forming pilus and EAF plasmid) by using methods described previously (15, 35, 43).

Serotyping.

All STEC isolates were serotyped in accordance with their O and H antigens by the E. coli Reference Center (Pennsylvania State University, University Park).

Statistics.

Confidence intervals were calculated by using the method presented by Wardrup (40).

RESULTS

Prevalence of non-O157 STEC.

Any sample that produced at least one non-O157 E. coli isolate harboring a Shiga toxin gene was considered positive for non-O157 STEC. One hundred eighty (53.9%) of 334 preevisceration samples and 27 (8.3%) of 326 postprocessing samples were positive for non-O157 STEC (Table 1). Multiple isolates were recovered from 90 of the 180 positive preevisceration samples and 17 of the 27 positive postprocessing samples. A total of 361 non-O157 STEC isolates were recovered.

TABLE 1.

Summary of STEC prevalence^a

Sample category	Total no. of samples	Total no. (%) non-O157 STEC positive	95% CIb	No. (%) of samples:		Total no. (%) STEC positived
				O157:H7 positive,c non-O157 STEC positive	O157:H7 negative,c non-O157 STEC positive
Preevisceration	334	180 (53.9)	48.5-59.2	144, 84 (58.3)	190, 96 (50.5)	240 (71.9)
Postprocessing	326	27 (8.3)	5.3-11.3	6, 0	320, 27 (8.4)	33 (10.1)

^aValues represent numbers of carcass samples carrying at least one non-O157 STEC isolate.

^b95% CI, 95% confidence interval.

^cSee Elder et al. (13).

^dSTEC-positive samples (O157 and non-O157) from this work and that of Elder et al. (13).

At preevisceration, 58.3% of samples that were previously determined to carry E. coli O157:H7 (13) also were positive for non-O157 STEC (Table 1). Of those samples that were negative for O157 (13), 50.5% produced a non-O157 STEC isolate. However, PCR results indicated that 76.8% of these samples contained either one or both of the Shiga toxin genes (data not shown). Twenty-seven (8.4%) of 320 O157:H7-negative postprocessing samples were positive for non-O157 STEC (Table 1). However, PCR results indicated that 43 (13.4%) of the samples contained either one or both of the Shiga toxin genes (data not shown). There were no E. coli O157:H7-positive postprocessing samples (13) that also carried a non-O157 STEC isolate (Table 1). Seven of the non-O157 STEC-positive postprocessing samples were from carcasses that were not found to be positive preevisceration.

Twenty-eight (93.3%) of 30 lots produced at least one preevisceration sample that was positive for non-O157 STEC (Table 2). The prevalence of positive preevisceration samples within a lot ranged from 0 to 100%. Eleven lots that had at least one positive preevisceration sample did not have a positive postprocessing sample. Seventeen (56.7%) of 30 lots produced at least one postprocessing sample that was positive for non-O157 STEC (Table 2). All 17 positive postprocessing lots were also positive preevisceration. The prevalence of positive samples within postprocessing lots ranged from 0 to 33.3%.

TABLE 2.

Prevalence of non-O157 STEC by lot

Plant	Sampling	Lot	na	No. of samples positive/totalb (%)
				Preevisceration	Postprocessing
A	1	1	84	6/18 (33.3)	0/18 (0)
		2	35	5/8 (62.5)	1/7 (14.3)
		3	68	1/15 (6.7)	0/16 (0)
	2	1	80	14/18 (77.8)	2/18 (11.1)
		2	41	8/10 (80.0)	2/9 (22.2)
		3	62	11/14 (78.6)	2/13 (15.4)
		4	44	8/13 (61.5)	1/13 (7.6)
B	1	1	46	0/11 (0)	0/9 (0)
		2	37	0/9 (0)	0/9 (0)
		3	39	6/9 (66.7)	1/9 (11.1)
		4	48	3/12 (25.0)	2/12 (16.7)
	2	1	36	2/9 (22.2)	0/9 (0)
		2	36	6/8 (75.0)	0/9 (0)
		3	36	8/9 (88.9)	0/7 (0)
		4	40	10/10 (100)	1/10 (10.0)
C	1	1	41	6/10 (60.0)	0/10 (0)
		2	76	10/17 (58.8)	1/17 (5.9)
		3	42	7/10 (70.0)	1/10 (10.0)
		4	38	10/10 (100)	2/9 (22.2)
	2	1	38	10/10 (100)	0/10 (0)
		2	44	5/9 (55.6)	1/10 (10.0)
		3	40	9/11 (81.8)	2/11 (18.2)
		4	46	9/12 (75.0)	4/12 (33.3)
D	1	1	37	4/7 (57.1)	1/8 (12.5)
		2	39	4/7 (57.1)	1/5 (20.0)
		3	38	2/8 (25.0)	0/8 (0)
		4	42	1/8 (12.5)	2/8 (25.0)
	2	1	65	4/14 (28.6)	0/13 (0)
		2	42	2/11 (18.2)	0/11 (0)
		3	58	9/17 (52.9)	0/16 (0)
Cumulative totals				180/334 (53.9)	27/326 (8.3)

^aNumber of animals in lot.

^bValues represent the number of positive samples over the total number of carcasses sampled per lot.

Total STEC prevalence.

Elder et al. originally collected the carcass samples used in this study for a survey of E. coli O157:H7 prevalence (13). The E. coli O157:H7 prevalence in these samples was determined to be 43.4 and 1.9% for preevisceration and postprocessing, respectively (13). By merging the data from the two studies, the total STEC prevalence on carcasses was determined to be 71.9% preevisceration and 10.1% postprocessing (Table 1).

Characterization of isolates.

All non-O157 isolates were confirmed as E. coli by biochemical assay, and all isolates were confirmed by PCR to carry the stx₁ and/or stx₂ genes. Eight percent of the isolates (29 of 361) did not ferment sorbitol on the basis of growth on SMAC. The distribution of virulence factors among the non-O157 STEC isolates is summarized in Table 3. With regard to Shiga toxin genes, PCR results indicated that 170 (47.1%) of the isolates carried stx₁, 132 (36.6%) had stx₂, and 59 (16.3%) possessed both stx₁ and stx₂. The eae gene was present in 43 isolates (11.9%), and EHEC hlyA was found in 98 (27.1%) of the non-O157 STEC isolates. Strains carrying only stx₁, without any other virulence factors for which we tested, accounted for the dominant share of isolates, at 42.1% (152 of 361). Forty isolates (11.1%) possessed the combination of virulence factors (eae, EHEC hlyA, and at least one Shiga toxin gene) that has been associated with severe disease in humans (1, 12, 14). Only four of these isolates were recovered from postprocessing samples. Of the 90 preevisceration samples that yielded multiple isolates, 25 had isolates of different virulence genotypes. Only one postprocessing sample yielded isolates of different virulence genotypes.

TABLE 3.

Virulence factors of non-O157 STEC isolates

STEC virulence factor(s)	No. of isolatesa			O serogroupsb
	Total	Preevis	Post
stx1	152	135	17	2, 3, 8, 45, 68, 74, 81, 119, 121, 128, 132, 142, 172, NT, rough
stx2	93	78	15	2, 6, 8, 10, 15, 20, 55, 88, 113, 132, 168, 171, x3, x6, NT, rough
stx1, stx2	15	15	0	8, 15, 113, 117, 171, NT
stx1, eae	2	2	0	59, 119
stx1, hlyA	8	3	5	104, NT
stx2, hlyA	19	17	2	2, 8, 88, 113, 139, x25, NT
stx1, stx2, hlyA	31	23	8	2, 55, 87, 88, 106, 113, 139, 153, NT, rough
stx1, stx2, eae	1	1	0	NT
stx1, eae, hlyA	8	6	2	103, 109, 136, 145, NT
stx2, eae, hlyA	20	20	0	6, 10, 117, 165, NT
stx1, stx2, eae, hlyA	12	10	2	21, 88, 121, 131, 153, 162, 165, NT
Total	361	310	51

^aPreevis, preevisceration; Post, postprocessing.

^bO serogroups of isolates from each virulence genotype are listed. NT, O antigen nontypeable; rough, autoagglutinable strains. Bold indicates a serogroup previously associated with human disease.

Serotype distribution.

Seventy-five percent of the isolates could be serogrouped in accordance with their O antigen, but only 21% could be typed by their H antigen (data not shown). Forty-one O serogroups were identified (Table 4). No isolates of serogroup O26 or O111 were recovered. Eight serogroups (O2, O8, O88, O113, O121, O132, O142, and O171) accounted for 49% of the STEC isolates. O142, a serogroup rarely associated with STEC (34), was the most prevalent serogroup among both preevisceration and postprocessing isolates. Two isolates were recovered that were of a serogroup (OX25) not previously associated with STEC. O serogroups that have been associated with human disease accounted for 178 (49%) of the 361 isolates recovered.

TABLE 4.

Serogroup distribution of non-O157 STEC isolates^a

Serogroup	No. of isolatesb
	Total	Preevis	Post
O142	54	46	8
O121	31	31	0
O2	22	19	3
O171	18	18	0
O113	15	12	3
O132	14	13	1
O8	13	11	2
O88	10	10	0
O6	9	8	1
O139	9	5	4
O172	9	7	2
OX3	9	3	6
O104	5	1	4
O117	5	5	0
O15	4	4	0
O165	4	4	0
O3	3	3	0
O55	3	3	0
O153	3	3	0
O168	3	0	3
O10	2	2	0
O45	2	2	0
O103	2	2	0
O109	2	0	2
O119	2	2	0
O145	2	2	0
OX25	2	2	0

^aBold indicates a serogroup previously associated with human disease. Serogroups with only one isolate: O20, O21, O59, O68, O74, O81, O87, O106, O128, O131, O136, O162, and OX6. Not listed are 87 Ont and 4 Orough isolates.

^bPreevis, preevisceration; Post, postprocessing.

Sixty-two preevisceration samples yielded multiple non-O157 isolates with typeable O antigens. Within 16 of these samples, isolates of different serogroups were identified. In two such samples, all three isolates were of different serogroups. Fifteen postprocessing samples produced multiple isolates of detectable O serogroups. Within each sample, all of the isolates belonged to a single serogroup.

One isolate was of a serotype (O6:H6) that has been previously associated with ETEC. PCR testing indicated that this isolate did not harbor genes for either the heat-labile or the heat-stable toxin that is indicative of ETEC (data not shown). Several isolates of serogroups commonly associated with EPEC were screened by PCR for the presence of the EAF plasmid or the gene coding for bundle-forming pili (bfp), both of which are usually found in EPEC strains. None of the STEC strains tested were positive for either marker (data not shown).

DISCUSSION

The data presented here show that 58.3% of the beef carcasses tested in large processing plants in the United States carried at least one type of non-O157 STEC prior to evisceration, but the prevalence was reduced, by using various antimicrobial intervention strategies, to 8.3% of the carcasses carrying non-O157 STEC at postprocessing. The samples surveyed were collected during the summer months, which have been associated with a peak in the prevalence of bovine STEC carriage (37). Therefore, STEC carcass contamination may be much less frequent at other times of the year.

Rogerie et al. reported a lower postprocessing non-O157 STEC prevalence (1.9%) on carcasses sampled during the summer at processing plants in France (29). Similarly, the non-O157 STEC prevalence on carcasses processed in Hong Kong was reported to be 1.7% (18). These apparently lower prevalence levels could come from a variety of factors but are most likely due to the use of dissimilar methods of STEC isolation. The methods used in both of these reports screened 100 or fewer colonies per sample, whereas we screened approximately 2,000 colonies per sample. In the previous studies, 10.7 and 11.4% of the postprocessing carcass samples were identified by PCR as carrying Shiga toxin genes, indicating a prevalence level similar to that described here (8.3% by colony hybridization, 13.4% by PCR) (18, 29). Neither of the previous studies reported STEC prevalence on carcasses prior to the application of antimicrobial interventions.

The hides and feces of animals presented for slaughter have been shown to be major sources of pathogens in processing plants (2, 13). To our knowledge, there have been no previous studies reporting non-O157 STEC prevalence on cattle hides to date. The prevalence of non-O157 STEC in the feces of beef cattle in the United States has scarcely been studied. One report using a small sample of heifers (n = 23) found fecal prevalence to range from 0 to 10% (33). Reported levels of non-O157 STEC in the feces of dairy cattle in the United States have ranged from 5.8 to 19.0% (10, 38, 41). There have been several studies of fecal prevalence of non-O157 STEC in beef cattle in other countries. Two studies testing healthy cattle in France recovered non-O157 STEC from 7.9 and 34% of the samples tested (28, 29). PCR results reported by these groups indicate that even higher levels of STEC were actually present, as stx genes were identified in 18 and 70% of the samples tested (28, 29). Kobayashi et al. detected stx genes, by using a nested PCR approach, in 100% of the cattle fecal samples obtained at an abattoir in Japan (17).

It is not clear what proportion of non-O157 STEC bacteria detected in cattle feces or on beef carcasses is able to cause disease in humans. Gyles et al. put forward the idea that STEC virulence is likely to be on a continuum, implying that virtually all STEC bacteria could be pathogenic under the proper circumstances (16). This would mean that less virulent strains could cause the same types of disease as highly virulent strains, given a sufficiently large dose and a sufficiently low immune status of the infected individual. Although there are most certainly nonpathogenic STEC bacteria in the continuum, there are not enough data to allow a determination of which STEC bacteria in the food supply represent a potential threat to human health.

Even though the set of virulence factors necessary to cause EHEC-related disease has not been defined strictly, associations between the carriage of certain genes and the ability to cause severe disease in humans have been made. Toxin-profiling studies of O157:H7 clinical isolates by Ostroff et al. showed that patients infected with isolates carrying only stx₂ were 6.8 times more likely to develop severe disease than those infected with strains carrying stx₁ or both stx₁ and stx₂ (23). Another study determined that, within a particular serotype, the carriage of stx₂ increases an isolate's likelihood of association with severe disease fivefold (7). It has also been demonstrated that Stx₂ has a lower 50% lethal dose than does Stx₁ when administered to mice (32). Therefore, isolates carrying stx₂ potentially represent an increased threat to human health. Of the non-O157 STEC isolates recovered in this study, 36.6% carried stx₂ either alone or in combination with the accessory virulence factors intimin and/or EHEC hemolysin.

Several epidemiological studies have associated the possession of eae by STEC with severe disease (1, 5, 8). In addition, EHEC hlyA has been found in a high proportion of non-O157 STEC strains causing human disease (12, 16, 30). Therefore, irrespective of serotype, carriage of the combination of stx, eae, and EHEC hlyA may be a good indicator of the pathogenic potential of STEC strains (21). We identified 36 isolates (11.6%) from 31 preevisceration samples and 4 isolates (7.8%) from 2 postprocessing samples that carried stx, eae, and EHEC hlyA.

HUS cases have been reported in which the STEC isolates did not contain eae (8, 26, 44). It has been assumed that these isolates used other means of attachment to the intestinal epithelium (25). Without knowing the ingested dose in these cases and the immunological status of the persons preinfection, the virulence of such isolates cannot be estimated. We isolated 58 strains, 12 from postprocessing samples, that possessed stx₂ and matched the serogroups (O2, O6, O55, O113, O168, and OX3) of eae-negative strains isolated from HUS patients (8, 26, 44).

Serotype has been reported to be the most consistent factor for predicting the disease-causing potential of an STEC strain (16). More than 150 serotypes have been associated with human disease (www.sciencenet.com.au/vtectable.htm), while more than 200 serotypes of STEC have been isolated from cattle (16, 22). Non-O157 serogroups commonly identified in clinical cases in the United States include O111, O26, O121, and O103 (1). Surprisingly, no isolates of serogroup O111 or O26 were recovered from beef carcasses in this study. The serogroup (O142) that was most common among the isolates we identified has rarely been associated with STEC (34).

O121 was the most common serogroup we recovered that has been associated with human disease previously (19, 31). Only 1 of the 31 O121 isolates carried stx₂, eae, and EHEC hlyA in addition to stx₁. Overall, of the 361 isolates described here, 178 (49%) belonged to an O serogroup previously associated with human disease. Twelve isolates belonged to serogroups previously associated with human disease (O6, O103, O121, O145, O153, and O165) and carried eae and EHEC hlyA. However, none of these 12 isolates were recovered from postprocessing samples.

A comparison of the number of samples that were positive for stx genes by PCR and the number of samples from which STEC were isolated indicates that the STEC prevalence levels presented here underestimate the actual STEC prevalence. PCR screening showed that 146 (76.8%) of 190 O157-negative samples possessed Shiga toxin genes; however, isolates were only recovered from 96 (50.5%) of these samples. Secondly, PCR results for several samples indicated the presence of both stx₁ and stx₂ while the isolates recovered from those samples contained stx₁ or stx₂ but not both. These differences are not surprising, since the sensitivity of PCR is greater than that of colony hybridization. The PCR method used here has been shown to detect one STEC cell in a background of 10⁴ cells, which is at least 10-fold more sensitive than the colony hybridization method used herein to isolate STEC (24). In addition, the samples used in this study were from enrichments in selective media that had been frozen for 2 years. Under such conditions, it is likely that some STEC bacteria would no longer be viable. The nonviable organisms would contribute to the number of PCR-positive samples but would not be isolated by colony hybridization. Therefore, a possibility exists that the prevalence of non-O157 STEC has been underestimated.

In addition, the variation of serotypes and virulence genotypes detected would likely have been greater if more isolates had been characterized. As was seen with some carcasses in this study, several different serogroups can be present on a single carcass. Since we only selected, at most, three presumptive STEC isolates from each sample, the serogroup and virulence genotype populations were certainly not exhausted for some carcasses.

In summary, we report here that a non-O157 STEC prevalence of greater than 50% exists on beef carcasses prior to evisceration during processing in U.S. plants. The non-O157 STEC load on carcasses was reduced to 8.3% by in-plant antimicrobial interventions. Subsequently, non-O157 STEC bacteria possessing accessory virulence factors indicative of pathogenic strains were identified on only two carcasses (0.6%) postprocessing.