Different prevalence of STEC in pigs were reported previously [24, 25, 27–29]. Kaufmann et al.  compared the STEC shedding rate in pigs at slaughter, which varied widely and ranged from 2.1% to 70% depending on the health conditions of the pigs and the detection method used. As shown in this study the anatomic sites sampled also affected the rate of isolation and consequently affected the prevalence in the population reported. Fecal samples were commonly used [24–26]. In our study we sampled the small intestinal content, the colon content and the feces. The prevalence of STEC in the colon (47.24%) was almost 2.5 times higher than in feces (19.33%) (P < 0.05) and 4.4 times higher than in the small intestine (10.83%) (P < 0.05). STEC strains are thought to mostly colonize the colons of humans  and it is likely to be the same for pigs.
In this study, 93 isolates were recovered from 62 of the 255 stx-positive samples, giving a culture positve rate of 24.31%, this result is similar to that of Botteldoorn et al., in which STEC isolates were obtained from 31% of the stx PCR-positive pig samples. Failure to isolate STEC from the stx-positive samples may due to the perturbation of high levels of background microflora, the loss of Stx prophages during subculture, the presence of other bacteria carrying stx or low levels of STEC in the samples.
In the present study, 12 serogroups and 19 serotypes were identified. The majority of these serotypes have been isolated from swine, sheep, cattle, food, and water in other countries [24, 31–36]. The most prevalent serotype is O20:H30/[H30], which was also reported in cattle and sheep in different countries [31, 32]. Six serotypes (O100:H20/[H20], O143:H38/[H38], O87:H10, O172:H30/[H30], O159:H16, O9:H30/[H30]) were rarely found in STEC isolates isolated from swine and other ruminants, implying that these serotypes may be restricted to the swine populations in these regions and their environments. Serotypes O86:H11, O20:NM, O100:NM, O9:NM, O172:NM and O114:NM have previously been described among STEC isolated from human patients [37–42]. Serotype O157:H7, which is common serotype causing human disease in some countries, was not detected.
A possible reason for no isolation of O157:H7 might be the method used. Isolation of O157 STEC often requires more targeted methods, such as the use of O157 immunomagnetic beads to capture the bacteria from enrichment broth and then culture on selective media . We previously used immunomagnatic separation to successfully isolate O157 STEC from pigs, although that was in an outbreak setting and was in a different geographic region . In this study we used CHROMagar™ ECC only and didn’t specifically target O157 STEC. CHROMagar™ ECC has been used by others for isolation of STEC from pigs . However, that study did not isolate O157 STEC either. Therefore, the CHROMagar™ ECC may not be an ideal media for O157 STEC isolation.
We used sorbitol-MacConkey agar as a quick method to pick potential O157 colonies since sorbitol fermentation is a traditional feature for differentiating O157:H7 which is sorbitol-negative although there are sorbitol-positive O157 STEC . In this study, a fair proportion (43%) of non-O157 STEC is actually sorbitol-negative. Therefore sorbitol fermentation is not a good indicator for O157:H7.
We analyzed multiple colonies from 21 samples to determine diversity within a sample (Figure 2). Two samples contained isolates with identical properties, suggesting they are the same strain, while the majority of the samples contain isolates belonging to the same sequence type but differing by one or more of the phenotypic or genetic properties tested, indicating that they are variants of the same clone. The most common variations are non-expression of the H antigen, variation of antibiotic resistance and/or variation in PFGE patterns. However 4 samples contained 2 different STs. Samples S15, S41, S49 and S50 all contain the prevalent ST993 and an additional ST, being ST10, ST88, ST710 and ST540 respectively, suggesting 2 different clones infecting the same pig.
Many studies have underlined the potential key role of the Stx2 subtypes in the severity of disease. Although Stx2e is not a potent subtype , strains harboring Stx2e have been isolated from patients with diarrhea . Intimin contributes to the development of A/E lesions and is a key virulence for some STEC serotypes , while ehxA can be found in many STEC serotypes, such as O157:H7 and O26:H11 that are associated with diarrheal disease and HUS [7, 50]. However, Sonntag et al. reported that the stx
2e-positive E. coli isolated from healthy pigs rarely contains genes for intimin and enterohemolysin . The prevalence of ehxA is very low in our samples at 2.15%, consistent with the findings of Sonntag et al. .
Other virulence factors may contribute to the pathogenicity of STEC. Although the role of EAST1 toxin in virulence to pigs has not been clearly determined, several studies have shown that astA gene is widely present among STEC isolates from both diarrheal and healthy pigs [15, 24, 26]. astA gene was also the most prevalent virulent gene (53.76%) among the 20 virulence genes tested in our study.
HPI was originally identified in Yersinia and now found in a range of pathogens , including the HUS-associated E. coli HUSEC041  and the 2011 German HUS outbreak strain O104:H4 . HPI had previously been detected in Stx2e- producing STEC strains from humans only . In this study we found 4 stx
2e STEC isolates, all ONT:H19/[H19], harbored the 2 HPI genes fyuA and irp although the frequency is low at 4.3%.
Fimbrial adhesins play an important role in colonization of the pig intestine and STEC strains may express up to 5 antigenically distinct fimbrial adhesins, F4, F5, F6, F18 and F41 . Different types of fimbriae can be associated with STEC diarrhea in animals of different ages [15–18]. In this study, only 4 isolates contained a fimbrial adhesin (F18). None of the other fimbrial adhesins (F4, F5, F6, F17 and F41) was detected. Of the nonfimbrial adhesin-encoding genes, paa was found in 7 isolates (7.5%), but efa1, toxB, lpfA
O113 and saa were not detected in any of the 93 STEC isolates. Eighty-two STEC isolates did not carry any of the adherence-associated genes tested.
Coombes et al.  reported that non-LEE encoded T3SS effector (nle) genes of non-O157 STEC strains are correlated with outbreak and HUS potential in humans. It will be interesting to examine our STEC isolates for the presence of the nle genes in future studies.
Many non-O157 STEC isolated from humans and animals have shown resistance to multiple antimicrobials [26, 55, 56], including resistance to trimethoprim-sulfamethoxazole and β-lactams [56, 57]. STEC isolates from swine feces in the United States show high resistance rates (>38%) to tetracycline, sulfamethoxazole and kanamycin but susceptible to nalidixic acid (resistance rate 0.5%) . In our study, we found that only 1 of the 12 categories of antimicrobial resistance types (carbapenems) and 2 of the 23 antimicrobial agents (imipenem and meropenem) were active against all the STEC isolates. The high prevalence (>50%) of resistance to tetracycline, trimethoprim-sulfamethoxazole, nalidixic acid and kanamycin is similar to that of other studies in China [55, 58]. In a study  of STEC from diseased pigs in Guangdong province, China, the majority of the isolates (95%) were resistant to more than 3 antimicrobials and the resistance rates to chloramphenicol (89%) and streptomycin (83%) were far higher than that of our study (37.63% and 48.39%, respectively). We also found that isolates from Chongqing showed a higher rate than those from the other 2 cities in this study. It should be noted that all samples collected from Chongqing were fecal samples while those from Beijing and Guizhou were small intestinal contents and colon contents samples, which may affect resistance profiles if different E. coli strains have a preference for the anatomic sites. However, it is more likely that the difference reflected the presence of resistant E. coli strains in different regions. Chongqing was dominated by the multidrug resistant ST3628. The differences in drug resistance rates between cities may be related to the differences in the prevalence of drug resistant STs.
Comparison with STs observed in human infections gives an indication of the potential risk for human infection of the swine STEC. We constructed an MST containing our STs, the 32 STs of the HUSEC collection and 52 human STEC STs from the E. coli MLST database (Figure 3B). None of the 21 STs in this study was identical to any of the 32 STs of HUSEC collection . We only found one ST, ST993, which was observed in human infections. When comparison was made at clonal complex level, some of our STs fell into the same clonal complex as the human STs (Figure 3B). ST10 clonal complex contained 2 of our STs (ST10 and ST3628), 1 HUSEC ST (ST43) and 1 human STEC ST (ST719) from the MLST database. However, Hauser et al. found that 8 of the 35 STEC STs they isolated from foods shared the same STs with HUSEC strains and were similar in their virulence gene composition . Since the STECs from foods and HUSEC collection were from the same geographical region, it is likely some of the HUSEC STECs were from local sources and not globally distributed. Our STECs from pigs may cause local human infections but there is no surveillance of human STECs in the regions where we sampled the swine STECs.