O-Serogroups and Pathovirotypes of Escherichia coli Isolated from Post-Weaning Piglets Showing Diarrhoea and/or Oedema in South Korea

This study aimed to determine the prevalence of several pathovirotypes and evaluate the association of haemolysis with the virotypes of pathogenic E. coli isolated from post-weaning piglets in South Korea from 2015 to 2019. We isolated 890 E. coli and tested for O-serogroups, virulence genes, haemolysis, and multilocus sequence typing. The predominant virotypes were STb:EAST1:AIDA-I, F18b:Stx2e:AIDA-I, F18:STa:STb:Stx2e, and eae:Paa in enterotoxigenic E. coli (ETEC), Shiga toxin-producing E. coli (STEC), ETEC/STEC, and enteropathogenic E. coli (EPEC), respectively. Regarding serogroups, O139, O149, O141, and O121 were mostly detected in F18:Stx2e:AIDA-I, F4:LT:STb:EAST1, F18:STa:STb, and F18:Stx2e:EAST1, respectively. There was a significant change in the frequency of the O141:F18ac:STa:STb (an increase from 1.6% to 10.1%) and O139:F18ab:Stx2e:AIDA-I (a decrease from 13.0% to 5.3%) virotypes in ETEC and STEC, respectively, from 2015 to 2019. The O141:F18ac:STa:STb virotype was mostly detected in the central area and was spreading to the southern area. The odds ratios between haemolysis and virotypes were 11.0, 6.25, and 8.57 in F18:STa:STb, F18:Stx2e:AIDA-I, and F4:LT:STb:EAST1, respectively. Our findings provide insights regarding the recent prevalence of pathogenic E. coli in South Korea and could be used for the development of vaccines for E. coli responsible for PWD and ED in post-weaning piglets.

F4 (K88) and F18 fimbriae are strongly associated with PWD and ED. F4 fimbriae have three antigenic variants (ab, ac, and ad), with F4ac being the most common in ETEC [5]. On the other hand, there are two variants of F18 fimbria: F18ab and F18ac. F18ab is mostly detected in Stx2e-encoding STEC strains while F18ac is associated with the ETEC and ETEC/STEC strains. A new F18 variant (F18new) has recently been identified in Germany and Korea [6,7].
There have been efforts to diagnose and characterise pathogenic E. coli. Generally, routine colibacillosis diagnosis in piglets is often performed through isolation, haemolysis on blood agar, O and H serotyping, and determination of the virulence factors using polymerase chain reaction (PCR or multiplex PCR), with its virulence pattern being used for virotyping [5,8]. Recently, multilocus sequence typing (MLST) has been used to determine the epidemiologic relevance of pathogenic E. coli in humans and animals [9,10]. It has more discriminatory power than existing methods, including pulse-field gel electrophoresis (PFGE) and PCR-based genotyping [11,12].
Previous studies have described the association between haemolysis and certain virulence factors [1,5,13,14]. However, the diagnostic utility of haemolytic E. coli strains isolated from diarrheic samples remains unclear.
Although there have been numerous efforts to detect and characterise the pathogenic E. coli involved in PWD and ED, there remains a need for systematic national data to facilitate prevention strategies and vaccine development for colibacillosis.
This study aimed to conduct national surveillance for pathogenic E. coli isolated from piglets with suspected PWD and/or ED from 2015 to 2019, as well as to characterise their phenotypes and genotypes. Moreover, we aimed to determine the diagnostic and predictive utility of haemolysis for pathogenic E. coli involved in PWD and ED. These results could provide novel insights regarding the recent prevalence of pathogenic E. coli in South Korea and the vaccine development for E. coli involved in PWD and ED in post-weaning piglets.

Bacterial Isolates
Annually, approximately 1000 faecal and/or intestinal samples obtained from postweaned piglets with diarrhoea and/or oedema are submitted to four Animal Disease Diagnostic Labs (national lab: Gimcheon, Korea, private labs: Iksan, Korea, Cheongju, Korea, and Osong, Korea). From the samples, 890 samples containing E. coli were identified from 850 farms from 2015 to 2019. Farm information regarding the submission date, farm location, and affected age were obtained using a request form. The farm regions were divided into central (454 farms: Gyeonggi, Gangwon, Chungbuk, and Chungnam provinces) and southern (396 farms: Gyeongbuk, Gyeongnam, Jeonbuk, and Jeonnam provinces) areas.
Samples were streaked onto MacConkey (BD, Franklin Lakes, NJ, USA) and blood agar (Asan, Korea), followed by aerobic incubation for 18 h at 37 • C. Distinctive colonies that showed nearly pure culture on MacConkey (BD, Franklin Lakes, NJ, USA) and blood agar were transferred onto blood agar. After incubation, haemolysis was determined in the blood agar and isolates were identified as E. coli using the VITEK II system (bioMérieux, Craponne, France). Finally, the isolates were frozen in 10% glycerol at −70 • C until further use.

O-Serogroups
O-serogroups were determined through PCR, as previously described [15], as well as through slide agglutination using rabbit polyclonal antisera to 181 O antigens [16]. Briefly, each overnight LB broth (BD, Franklin Lake, NJ, USA) culture was pelleted through centrifugation at 8000× g. Subsequently, the pellet was homogenised and heated at 100 • C for 5 min. The centrifuged supernatant was used as a DNA template. Upon amplification of the Og44, Og153, Og117, Og13, Og89, Og186, and Og46 primers, the isolates were incubated in tryptic soy broth (BD, Franklin Lakes, NJ, USA) overnight at 37 • C and heated for 2.5 h at 100 • C. Moreover, autoagglutination was determined using a mixture of 0.85% sodium chloride (NaCl) before testing the O-serogroup.

Multilocus Sequence Typing (MLST)
All 50 F18:STa:STb isolates underwent MLST genotyping, as previously described [12]. The PCR amplification and sequencing of seven housekeeping genes (adk, fumC, gyrB, icd, mdh, purA, and recA) were performed based on the protocols specified on the E. coli MLST website (http://mlst.warwick.ac.uk/mlst/dbs/Ecoli, accessed on 17 January 2021). All the primer sequences of these seven housekeeping genes are available at http://mlst. warwick.ac.uk/mlst/dbs/Ecoli/documents/primersColi (accessed on 17 January 2021). Allele numbers for seven gene fragments of each isolate were obtained through comparison with the corresponding alleles available in the MLST E. coli database (http://mlst.warwick. ac.uk/mlst/dbs/Ecoli, accessed on 17 January 2021); additionally, the sequence type (ST) of each isolate was determined by combining the seven allelic profiles. MLST clusters were calculated using BioNumerics version 8.0 (Applied Maths, bioMérieux, St. Maartens-Latem, Belgium). The farm locations with the most prevalent STs were depicted on a map based on information regarding the latitude and longitude obtained from the Korean Animal Health Integrated System.

Statistical Analysis
Proportion differences between the haemolysis and virulence genes (virotypes) were tested using the χ 2 test; however, Fisher's exact test was used when the number of isolates in the contingency table was below five. Statistical significance was defined as a p-value < 0.05. Odds ratios (OR) and log ORs were determined as comparative measures. All statistical analyses were conducted using the Jamovi statistical package (version 1.6.15, The jamovi project).

O-Serogroup and Virotype
There were likely associations between certain O-serogroups and virotypes (Tables 1 and  S2

MLST Genotypes of the F18:STa:STb
A total of 50 F18:STa:STb isolates were tested to identify the ST distributions ( Figure  4). Among them, 45 isolates were genotyped and assigned to 10 STs. The remaining five isolates could not be assigned to any ST since they could not be identified in the ST database in PulsNet (https://pubmlst.org/, accessed on 17 January 2021). The isolates were mostly isolated from farms located in the densely populated central area ( Figure 5). The most frequent sequence types were ST7323 (28), ST760 (7)

MLST Genotypes of the F18:STa:STb
A total of 50 F18:STa:STb isolates were tested to identify the ST distributions ( Figure 4). Among them, 45 isolates were genotyped and assigned to 10 STs. The remaining five isolates could not be assigned to any ST since they could not be identified in the ST database in PulsNet (https://pubmlst.org/, accessed on 17 January 2021). The isolates were mostly isolated from farms located in the densely populated central area ( Figure 5). The most frequent sequence types were ST7323 (28), ST760 (7), ST155 (3), and ST3054 (2). These four ST isolates accounted for 80% of the 50 E. coli F18:STa:STb isolates and were mostly detected in the central area. The ST7323, ST3054, and ST760 were generally detected from 2017 to 2019; however, ST155 was only detected in the central area in 2019 (Figures 4 and 5). Table 2 presents the association between virotypes (virulence genes) and haemolysis (Table S3). Generally, isolates encoding the virulence genes of F4, F18, LT, STa, and Stx2e were associated with haemolysis (p < 0.01). However, the virulence genes of STb, EAST1, eae, Paa, and AIDA-I was negatively associated with haemolysis (p < 0.01). Further, the virulence genes of F18 (OR, 33.3), Stx2e (OR, 6.18), and LT (OR, 4.28) were strongly associated with haemolysis.   Table 2 presents the association between virotypes (virulence genes) and haemolysis (Table S3). Generally, isolates encoding the virulence genes of F4, F18, LT, STa, and Stx2e were associated with haemolysis (p < 0.01). However, the virulence genes of STb, EAST1, eae, Paa, and AIDA-I was negatively associated with haemolysis (p < 0.01). Further, the virulence genes of F18 (OR, 33.3), Stx2e (OR, 6.18), and LT (OR, 4.28) were strongly associated with haemolysis.

Discussion
There have been numerous efforts to characterise pathogenic E. coli based on the O and H antigen serogroups, phenotypes (pathotypes and virotypes), and genotypes (e.g., through PFGE and MLST) [1,10,17]. There have been numerous studies to detect and characterise the pathogenic E. coli isolated from piglets showing PWD and ED in diagnostic labs. However, there is a need for comprehensive national surveillance to facilitate prevention strategies and vaccine development for colibacillosis. Accordingly, this study conducted national surveillance for pathogenic E. coli isolated from piglets with suspected PWD and/or ED in four diagnostic labs from 2015 to 2019; moreover, it characterised the phenotypes and genotypes. Additionally, we evaluated the diagnostic importance of haemolysis as a predictive factor for pathogenic E. coli in PWD and ED.
The underlying reason for the shift of prevalent virotypes isolated from pathogenic E. coli since 2017 remains unclear. However, using vaccines and feed additive antibiotics for diarrhoea and oedema could partially have influenced the observed shifts even if environmental factors, including feeds, temperature, and hygiene, are predisposing factors for PWD and ED [7]. The decrease of the frequency of ETEC (F4:LT:STb) and STEC (F18:Stx2e:AIDA-I) would have been associated with the use of vaccines for pili and enterotoxin (F4 and LT) since the 1990s and approval of the vaccine (Stx2e subunit) for ED in 2019, respectively; however, there were no differences in the frequency of STEC (Stx2e) from 2015 to 2019 [7].
ETEC-expressing fimbriae (mainly F4 and F18) and toxins have generally been detected; however, fimbriae negative virotypes (STb:EAST1:AIDA-I, STa:STb, and eae) have been detected in PWD [1]. Among them, the F4:LT:STb virotype is among the most prevalent in the USA [4], Europe [19], and Vietnam [13]. The F18:STa:STb is commonly detected in PWD; further, it is associated with O138 and O141 [1,8]. The role of F4-and F18-negative virotypes in PWD remains unclear. Moreover, ETEC isolates of the STb or STb:EAST-1 virotypes from weaned pigs may produce a non-fimbrial apparatus (AIDA-I) [20,21], which was originally detected in E. coli isolates from humans with diarrhoea [1]. Ngeleka and others (2003) reported that 20.5% of enterotoxigenic E. coli were STb: EAST1:AIDA-I or STb:AIDA-I, which can induce diarrhoea in an experimental infection using newborn pigs. EPEC is considered a PWD-related pathotype and does not encode any virulence factor of classic ETEC strains [22].
The F18ab and F18ac subtypes are closely associated with STEC and ETEC strains, respectively, which cause ED and PWD [23]. F18ab was common in O139-STEC encoding Stx2e and AIDA-I; moreover, F18ac was associated with O141-ETEC encoding STa and STb. In this study, F18new was associated with O121-STEC encoding Stx2e and EAST1. F18new variant isolates were associated with the ETEC, rather than the STEC, pathotype based on clinical and antigenic sequence differences [6,7].
MLST has been used in epidemiologic studies on cases or outbreaks attributable to pathogenic E. coli from human and animals [10,24]. Although three different MLST approaches have been used in the public database, the current study mostly used Achtman methods with seven-locus sequencing. The most frequent sequence types of F18:STa:STb were ST7323 (28), ST760 (7), ST155 (3), and ST3054 (2). These four ST isolates accounted for 80% of the 50 F18:STa:STb isolates, with most being isolated from the central area. The ST7323, ST3054, and ST 760 were generally detected from 2017 to 2019. Contrastingly, ST155 was newly detected in the central area in 2019, with subsequent spreading to other areas.
The relationship between haemolysis and the pathotype has been evaluated with respect to being a diagnostic criterion. In this study, isolates encoding the virulence genes of F4, F18, LT, STa, and Stx2e were associated with haemolysis (p < 0.01). ETEC isolated from PWD cases are mostly haemolytic (ETEC F4:LT:STb:EAST1, 9% or F18:Stx2e:AIDA-I, 7.2%); however, non-haemolytic strains can be observed. Luppi and others (2016) reported that 97.6% of ETEC cases were haemolytic. The remaining 2.4% of non-haemolytic ETEC isolates, which underwent consistent tests for haemolytic activity, were recovered in France, Italy, and Germany; moreover, they shared similar virotypes (F4, STa, and STb). Some O-serogroups (O138, O139, and O141) in ED are strongly associated with haemolysis [8]. However, a previous study reported that haemolysin is not an essential virulence factor in a challenge experiment involving a mutant missing haemolysin [25]. In our study, F18 and F4 were strongly associated with haemolysis in PWD and ED isolates. However, the ORs of the strains encoding eae, Paa, EAST1, and AIDA-I were < 1, which indicates a negative relationship between virulence genes and haemolysis. In virulence gene combinations, the ORs of the F18:STa:STb and F18:Stx2e:AIDA-I isolates were slightly lower than those for other F18-positive virotypes. However, the OR value for F4:LT:STb:EAST1 was higher than those for the isolates singularly encoding each gene (F4, LT, STb, and EAST1). This is consistent with previous findings regarding F4-and F18-expressing strains [1,19]. However, isolates encoding enterotoxin (STb, EAST1) and non-fimbrial genes (eae, Paa, and AIDA-I) were non-haemolytic. Therefore, haemolytic colonies are likely to be associated with F4and F18-positive ETEC and STEC.

Conclusions
Our findings indicated that the STb:EAST1:AIDA-I, F18:Stx2e:AIDA-I, and eae:Paa virotypes were the most prevalent in ETEC, STEC, and EPEC isolated in piglets with PWD and/or ED. Notably, from 2015 to 2019, there was a decrease and increase in the frequency of the F18:Stx2e:AIDA-I and F18:STa:STb virotype, respectively. F18-and F4-positive ETEC and STEC were more likely to be associated with haemolysis in PWD and ED isolates. Our data provide insights regarding the recent prevalence of pathogenic E. coli in South Korea and could facilitate vaccine development in E. coli responsible for PWD and ED in post-weaning piglets.