Prevalence of Virulence Genes and Antimicrobial Resistances in E. coli Associated with Neonatal Diarrhea, Postweaning Diarrhea, and Edema Disease in Pigs from Austria.

Increasing numbers of multi-resistant Escherichia (E.) coli from clinical specimens emphasize the importance of monitoring of their resistance profiles for proper treatment. Furthermore, knowledge on the presence of virulence associated genes in E. coli isolates from European swine stocks is scarce. Consequently, a total of 694 E. coli isolated between 2016 and 2018 from diarrheic piglets of Austrian swine herds were investigated. The isolates were tested for their susceptibility to twelve antibiotics using agar disk diffusion test and for the presence of 22 virulence associated genes via PCR. Overall, 71.9, 67.7, and 49.5% of all isolates were resistant to ampicillin, tetracycline, and trimethoprim-sulfamethoxazole, while resistance levels to gentamicin and fosfomycin were 7.7 and 2.0%, respectively. Resistance frequency to ciprofloxacin was higher than in previous studies. Isolates were more likely to be resistant to ampicillin if they were also resistant to ciprofloxacin. No isolate was resistant to meropenem or amikacin. Virulence genes were detected more frequently in isolates expressing hemolytic activity on blood agar plates. The detection rate of faeG was increased in fimH negative isolates. We assume, that hemolytic activity and absence of fimH could be considered as potential indicators for the virulence of E. coli in piglets.


Introduction
The emergence of multi-resistant bacteria is commonly considered to be a consequence of the misuse and overuse of antibiotics [1,2]. In particular, the emergence of extended-spectrum β-lactamases (ESBL) producing Escherichia (E.) coli [3] has led to major concerns over the past few years. A study demonstrated that distinct numbers of pathogenic porcine E. coli were resistant to antibiotics that are also commonly applied in swine stocks, such as tetracyclines and trimethoprim/sulfonamide [4]. An increasing number of E. coli isolates resistant to cephalosporins and fluoroquinolones has also been reported repeatedly [5,6]. Enterobacteriales (including E. coli, Proteus spp., Klebsiella pneumoniae) being

Antimicrobial Resistance Profiles
Resistances to ampicillin were predominant (71.9%), followed by resistances to tetracycline (67.7%) and trimethoprim-sulfamethoxazole (49.5%), whereas none of the isolates was resistant to meropenem or amikacin (Table 1). Approximately 2% of all isolates were resistant to fosfomycin and aztreonam. Among tested antibiotics that are regularly applied to treat colibacillosis in swine stocks, resistances against gentamicin (7.7%) were the least common (Table 1). In total, 16.4% of all isolates were resistant to ciprofloxacin. Isolates that were resistant to ciprofloxacin were more likely to be resistant to ampicillin (OR = 5.381 (95% CI 2.855-10.140)) and trimethoprim-sulfamethoxazole (OR = 3.482 (95% CI 2.554-4.747)). Compared to non-hemolytic E. coli, isolates with hemolytic activity showed higher resistance rates to trimethoprim-sulfamethoxazole, chloramphenicol, and gentamicin. Isolates that were determined as ESBL-producing E. coli were more frequently resistant to eight out of those ten tested antibiotics, for which resistances were observed, than non-ESBL-producing E. coli ( Table 1). Out of all ceftazidime resistant isolates, 88.2% could be classified as ESBL-producing E. coli. Generally, 27.1% of all isolates from suckling piglets and 16.2% of all isolates from weaned piglets were classified as ESBL-producing E. coli. No ESBL-producing isolate showed hemolytic activity. Weaned piglets had significantly lower resistance rates to tested antibiotics in all three calculated generalized linear models (Model 1: p = 0.0107, Model 2: p = 0.0397, Model 3: p = 0.0019). Examined isolates, originating from swine stocks supervised by herd veterinarian 1 exhibiting the lowest resistance rates to tetracycline, had the second highest resistance rates to ciprofloxacin ( Figure 1). However, considering age group and clinical symptoms, the data show no significant influence of the herd veterinarian on the resistance profiles in any of the generalized linear models.

Virulence Genes
Out of 694 tested isolates, 39 did not carry any of 22 examined virulence genes, while fimH was detected most frequently compared to other virulence genes (552/694). Due to the uncertainty about the impact of several virulence factors in the pathogenesis of neonatal diarrhea, PWD and ED only those which are known to be associated with ETEC or EDEC are visualized (Tables 2 and 3) and further discussed in more detail. Only a small number of isolates contained genes coding for F5 (8/694), F6 (3/694), and F41 (2/694). In total, 55 of 453 (12.1%) isolates originating from suckling piglets with diarrhea could be classified as Enterotoxigenic E. coli (ETEC), with the combination of faeG (F4) and elt being the most common ETEC (38/55). elt could be detected in 70.2% of all faeG-positive ETEC isolates by PCR (Table 2). Based on the combination of certain virulence genes, 43 out of 241 (17.8%) isolates from weaned piglets could be classified as ETEC. In this group, the combination of faeG (F4) and elt was also predominant (35/43), while fasA (F6), fim41A (F41), and stb were not detected in a single isolate. AidA was present in the majority of all fedA (F18) and stx2-positive isolates (12/14) ( Table 3). Number of E. coli isolates originating from weaned piglets with diarrhea or suspected edema disease that were positive for an adhesion gene, as well as a toxin gene.
In general, fedA (F18) was numerically the most frequent virulence gene in all aidA positive ETEC isolates (13/14). escV was present in 10 isolates originating from suckling piglets and in 18 isolates from weaned piglets, respectively.
In addition, cnf1 was exclusively found in isolates that were positive for papC (17/17), but only 27.0% of all papC-positive isolates were also positive for cnf1. The detection rate of faeG (F4) was significantly increased in fimH negative isolates (Standardized residual >1.96; p <0.05) but not vice versa (Standardized residual <1.96). The detection rate of fedA (F18) was decreased if the isolate was fimH negative (Standardized residual <1.96). Comparing the detection rate of faeG in different sampling materials, faeG was more frequently detected in isolates from fecal samples than in isolates from the intestines (p = 0.0116). The odds ratio of detecting faeG in E. coli isolates from fecal samples compared to isolates from intestinal samples was 2.732 (95% CI 1.304-5.721).

Discussion
Antimicrobial resistances were determined by agar disk diffusion test and not by determining distinct resistance encoding genes or resistance-mediating mutations. Furthermore, due to limitations of available clinical breakpoints towards most tested antimicrobials for porcine E. coli, clinical breakpoints for humans (Table 6) were applied for the interpretation of all twelve antimicrobial substances on disk diffusion tests instead. Therefore, antimicrobial resistance profiles should be interpreted with caution.
In the present study, resistance levels to ampicillin showed approximately the same level as in other European countries [19]. High resistance levels could be linked to the frequent use of β-lactam antibiotics to treat suckling-piglets against diseases like clostridiosis. This could also explain why resistance rates of isolates from suckling piglets to ampicillin were significantly higher than those from weaned piglets.
Tetracyclines are also more frequently applied to treat livestock in Austria than other antimicrobial substances [20]. Therefore, observed high resistance levels in our investigations might be the result of a widespread use of tetracyclines, as already assumed before [21]. In recent Spanish investigations using agar disk diffusion testing, resistances of clinical porcine E. coli isolates were also more frequently detected against amoxicillin and tetracycline than against ceftiofur, gentamicin, and enrofloxacin [22]. It can be assumed that resistances of porcine E. coli towards tetracyclines emerged, when antimicrobials were also used as growth promoters. Tetracyclines are not licensed for the treatment of diarrhea or edema disease in Austria but are mainly applied to treat respiratory diseases in fattening pigs.
Our standard microbiological diagnostic procedures do not test resistances to trimethoprim and sulfamethoxazole separately since both are only licensed in combination for the application in pigs. Resistance rates towards trimethoprim-sulfamethoxazole of porcine E. coli were relatively high in our investigations, which goes along with the results of other European surveys [19,23]. Similar findings were also reported in Germany, although the number of isolates with resistances to trimethoprim-sulfamethoxazole has slightly decreased over the past 14 years [24].
Chloramphenicol is prohibited as a therapy in livestock due to its toxic attributes [25]. Florfenicol though is commonly used to treat respiratory diseases in swine. In a Danish study, only 12% of all chloramphenicol resistant isolates were also resistant to florfenicol [26]. But even if the use of florfenicol in swine stocks is the reason of resistances to chloramphenicol in about 18.5% of all isolates, we have to keep in mind that in Austria florfenicol is not licensed to treat E. coli associated diseases. In our investigations, hemolytic isolates which were more frequently classified as ETEC had higher resistance rates to chloramphenicol than non-hemolytic isolates (Table 1).
An increase of gentamicin resistant porcine E. coli isolates may be attributed to the use of apramycin and other aminoglycosides in swine stocks since cross-resistances are reported [27]. Observed higher resistance rates of isolates with hemolytic activity to gentamicin and trimethoprim-sulfamethoxazole in comparison to non-hemolytic E. coli might be caused by the frequent use of both antibiotics to treat colibacillosis in the field. Noteworthy, virulence genes associated with pathogenicity were more frequently detected in isolates with hemolytic activity.
In our investigations, the probability of an isolate being resistant to ampicillin and trimethoprim-sulfamethoxazole was increased, if the isolate was resistant to ciprofloxacin. In trials, the number of ampicillin and trimethoprim-sulfamethoxazole resistant E. coli was increased after an intramuscular administration of ciprofloxacin [28]. In total, 16.4% of all isolates were resistant to ciprofloxacin, which is relatively high compared to investigations in other European countries [19]. In general, it can be assumed that the observed higher resistance levels to fluoroquinolones compared to cephalosporins in our investigations might be linked to their more frequent use in the treatment of E. coli-associated diseases in swine.
Observed resistances of E. coli to cephalosporins may reflect the wide spread of ESBL-producing E. coli, but could be also a potential result of their general use in Austrian swine stocks [20]. Out of all ceftazidime resistant isolates, 88.2% could be classified as ESBL-producing E. coli. However, only 30.9% of all isolates which were classified as ESBL-producing E. coli showed an in vitro resistance to ceftazidime. This could be due to the use of human clinical breakpoints, as well as using phenotypic methods to determine ESBL instead of testing genes.
Similar to ceftazidime, resistances to aztreonam were mainly observed in ESBL-producing isolates. However, compared to cephalosporins, aztreonam is not licensed to treat pigs in Austria. Despite ESBLs being commonly defined to confer bacterial resistance to penicillins and cephalosporins of the first, second, and third generation, as well as aztreonam, only 7.7% of our isolates that were classified as ESBL-producing were also resistant to aztreonam.
Although neither fosfomycin nor any related substance is licensed to treat pigs in the European Union, fosfomycin resistant isolates were observed in our investigations. Thus, swine might act as a reservoir for fosfomycin resistant E. coli in Austria. Isolates with plasmids carrying fosA3 that encodes resistance to fosfomycin, as well as ESBL genes, have already been detected in livestock [29]. Nevertheless, in our study no ESBL-producing isolate was resistant to fosfomycin.
Neither meropenem-nor amikacin resistant E. coli were detected, both not licensed for use in animals in the EU. The absence of isolates being resistant to amikacin in swine stocks was also observed before [30], while Carbapenem resistant E. coli from European swine stocks have occasionally been detected [31], concretely in Germany [32,33] and Italy [34].
Data on the amount of used antibiotics by the different herd veterinarians were not available for this analysis. Those data could have helped to explain potential associations between the enhanced and blanket usage of certain antibiotics by each veterinarian. Interestingly, isolates from the herd veterinarian with the lowest resistance rates to tetracycline had the second highest resistance rates against ciprofloxacin, which could reveal the preference of certain veterinarians to use certain antibiotics.
The impact of certain virulence genes, like astA, aidA, and cnf1, might be higher than expected, whereas genes previously exclusively associated with pathogenicity like sta and stb were detected in only a few isolates from clinically diseased piglets. Furthermore, many of these virulence factors are not contained in commercial vaccines.
The absence of any investigated virulence gene in only a small number of all isolates (39/694) has also been observed in several other studies [35,36]. However, in those studies a PCR for fimH, which was the most frequently detected virulence gene in our investigation, was not applied. In an Australian investigation fimH was detected in all E. coli isolates from piglets with neonatal diarrhea [16], although an association between neonatal diarrhea and fimH has not been made so far. Nevertheless, the intestines are described to be the main reservoir of UPEC [37], which could explain the detection of fimH in so many isolates derived from feces or intestines.
The observed significantly increased detection rate of faeG in isolates negative for fimH could define fimH as an indicator of non-pathogenic E. coli colonizing the intestines. A correlation between the occurrence of fimH and faeG in porcine E. coli has not been described before. While fimH is a gene often associated with UPEC, F4 (faeG) are associated with ETEC. Thus, detection of both genes in the same isolate emphasizes the insufficiency of the classification of E. coli due to distinct virulence gene patterns. The occurrence of fimH and faeG in the same isolate has been observed in our investigations in several isolates.
E. coli isolates being positive for faeG were more likely to be detected in isolates from fecal samples compared to isolates from intestines. Therefore, fecal samples could be more adequate to detect faeG than intestinal samples.
Compared to previous studies, the number of isolates which were classified as ETEC out of all isolates from weaned piglets (43/241) was relatively low [35,36]. In accordance with results of other studies we detected astA more often in isolates which were positive for elt [9,38] and faeG (F4) was detected more frequently than fanC (F5), fasA (F6), and fim41A (F41) [35,36,38]. The low detection rates for fanC, fasA, and fim41A were explained to be the result of collecting samples only from piglets which were older than 14 days. However, recently no European study on the occurrence of virulence genes in E. coli isolates from younger suckling piglets has been published. However, in our study a high number of samples was also taken from suckling piglets with less than 14 days of age. Thus, we assume that a shift away from ETEC containing F5, F6, and F41 might have happened over the last decades.
Although the combination of faeG and elt was also noted frequently in our investigations, only one isolate was also positive for stb, whereas the combination of faeG, elt, and stb was observed most frequently in ETEC in other mentioned studies [36,38].
The impact of AIDA-I on the pathogenesis of PWD and ED is discussed controversially. However, it has already been proven that strains containing only aidA and stb were able to colonize the intestine, form biofilms, and induce diarrhea [12]. A strong linkage between fedA (F18) and aidA has already been described [39]. That result correlates well with our findings, in which the combination of faeG and aidA only occurred three times (twice together with fedA), whereas fedA and aidA were detected together in 13 isolates. In fact, aidA has already been discussed to play a significant role in the pathogenesis of ED [39].
Despite the fact that hemolytic activity is not a necessary attribute for pathogenic E. coli, it has been demonstrated several times that the majority of ETEC isolates from swine exhibits hemolytic activity on blood agar plates [38,40]. This correlates well with our investigations, in which several virulence genes associated with pathogenicity (aidA, faeG, fasA, fedA, fim41A, elt, sta, astA, stx2) were detected more frequently in hemolytic E. coli than in E. coli without hemolytic activity ( Table 4,5). In addition, the occurrence of these genes was much lower in ESBL-producing isolates (Table 4,5), from which none showed hemolytic activity.
Data about the impact of cnf1 in the pathogenesis of PWD is rare, while it has been described to be a typical gene of Necrotoxic E. coli (NTEC), along with genes encoding P and S fimbriae (papC, sfa) [18]. The correlation between the occurrence of cnf1 and papC has also been observed before [18]. Therefore, the role of several described genes in the pathogenesis of colibacillosis should not be underestimated.

Sample Collection
Between 2016 and 2018, 694 E. coli strains were isolated from feces and intestines derived from clinically diseased suckling and early weaned piglets from Austrian swine stocks in the course of routine diagnostics performed at the Institute for Microbiology, University of Veterinary Medicine Vienna (Table S1). E. coli isolates were identified using standard microbiological procedures. Animals were selected by the herd veterinarian based on the presence of typical clinical symptoms of neonatal diarrhea, postweaning diarrhea, or edema disease, since the aim of the vet was to obtain results of antimicrobial susceptibility testing of the causative pathogens for the choice of an appropriate treatment. Intestines (n = 312) and feces (n = 372) were collected from 215 different farms. Depending on the veterinarian, the number of farms by herd veterinarian from which specimens were taken varied between one and 21.

Antimicrobial Susceptibility Testing
Susceptibility testing was performed by agar disk diffusion according to the recommendations given by the CLSI document M100 (28th ed.) (Clinical and Laboratory Standards Institute (CLSI), 2018). Human clinical breakpoints (M100-28) were applied for the interpretation of zone diameters of all tested antimicrobial substances (Table 6). Isolates were classified as ESBL-producing if they showed an increase of over 5 mm in a zone diameter for either ceftazidime or cefotaxime tested in combination with clavulanate compared to the zone diameter of ceftazidime or cefotaxime alone in agar disk diffusion (Becton Dickinson, Heidelberg, Germany).

Evaluation
All results were summarized retrospectively using TIS®(Tierspitalinformationssystem Orbis VetWare, Agfa HealthCare, Bonn, Germany) and Microsoft Excel. In total, 453 isolates were recovered from suckling piglets and 241 from weaned piglets. Overall, 403 isolates were characterized as non-hemolytic E. coli, 142 as hemolytic E. coli and 166 isolates as ESBL-producing E coli without hemolytic activity. Due to the relevance of antimicrobials in public health and/or their importance for therapeutic use in swine and humans, resistance patterns to the following twelve antimicrobial substances or combinations were analyzed: ampicillin, tetracycline, trimethoprim/sulfamethoxazole, chloramphenicol, gentamicin, tobramycin, ciprofloxacin, ceftazidime, aztreonam, fosfomycin, amikacin, and meropenem (Becton Dickinson, Heidelberg, Germany).

Statistical Analyses
The prevalence of virulence genes and antimicrobial resistance in dependence of the age (suckling piglets and weaned piglets), as well as the phenotype of E. coli (non-hemolytic E. coli, hemolytic E. coli, ESBL E. coli), was assessed via Microsoft Excel. Furthermore, Pearson Chi Squared and Fisher' Exact Tests were calculated using the R statistical computing environment (R Core Team R: A language and environment for statistical computing. R foundation for statistical computing, Vienna, Austria, version 3.5.3) to verify potential dependencies between 1.: ampicillin, ciprofloxacin, and trimethoprim-sulfamethoxazole; 2.: fimH and faeG; 3.: papC and cnf1, as well as 4.: the detection rate of faeG depending on the sampling material; 5.: the detection rates of virulence genes (aidA, faeG, fedA, fim41A, elt, sta, astA, stx2) depending on the presence of hemolytic activity; and 6.: considering the age group and clinical symptoms, the influence of the herd veterinarian on antimicrobial resistance rates was investigated using different generalized linear models.
We aimed to reduce the models to include only the most relevant factors and applied the Akaike information criterion (AIC) to determine the best model. The AIC uses the parsimony principle to reduce the number of factors considered in the model and incorporates both the complexity of the estimated model and their associated goodness-of-fit to the data [44]. In detail, first model analyzed the influence of age group, clinical symptoms and herd veterinarian on a binary outcome (defined as no detected resistance within the tested antibiotics and at least one detected resistance), after application of AIC only the age group was kept in the model. The second model is similar, with a slightly different binary outcome (defined as less than three detected resistances within the tested antibiotics and at least three detected resistances), after application of AIC only age group and herd veterinarian were kept in the model. The third model used the relative amount of resistances within the tested antibiotics as metric outcome; all three variables were kept in the model after using stepwise AIC. A P-values of <0.05 were considered as significant. The used R packages were "gmodels", "oddsratio", "MASS" and "ggplot2" [45][46][47].

Conclusions
Resistance levels to ciprofloxacin are higher than previously reported. Resistances to substances that are not applied to swine stocks like fosfomycin emphasize the potential of transmission between different species. Hemolytic activity and absence of fimH are potential indicators for the pathogenicity of E. coli but not necessary attributes. Furthermore aidA, astA, and cnf1 might play a key role in the pathogenesis of colibacillosis, but further investigations are needed.