Molecular Evidence for Intra- and Inter-Farm Spread of Porcine mcr-1-Carrying Escherichia coli in Taiwan

From January 2013 to December 2018, 90 Escherichia coli isolates were collected from 90 sick pigs on 58 farms in seven cities in Taiwan. The minimum inhibitory concentrations (MICs) of the isolates to 14 antimicrobial agents were determined on the VITEK 2 system (bioMérieux, Marcy-l’Etoile, France), and the resistance to colistin was assessed via the reference broth microdilution (BMD) method. The mobilized colistin resistance genes (mcr) were determined for the colistin-resistant isolates, which displayed BMD MICs ≥ 4 mg/L. Genotypes of the mcr-positive E. coli isolates were determined by multilocus sequence typing, arbitrarily primed polymerase chain reaction (PCR), and pulsed-field gel electrophoresis. All of the isolates were tested for susceptibility to carbapenems. Fifty isolates (55.6%) were resistant to colistin, 39 of which (78%) were positive for the mcr-1 gene. E. coli isolates harboring mcr-1 were most frequent in 2017 (15/18, 83.3%), followed by 2018 (13/23, 56.5%), 2015 (7/21, 33.3%), and 2016 (3/24, 12.5%). A total of 18 sequence types (STs) were identified among the 39 porcine mcr-1-carrying E. coli isolates; 13 were ST2521 (33.3%) isolated in 2017 and 2018. Five genotypes (clones) were identified, and the same genotypes were in sick pigs on the same farm and different farms. This suggests intra- and inter-farm spread of porcine mcr-1-carrying E. coli. The results presented here indicate high colistin resistance and wide mcr-1 E. coli prevalence among the sick pigs sampled in 2015–2018 in different regions of Taiwan.

The use of colistin in animal husbandry in Taiwan has been restricted since July 2007, but mcr-1-harboring E. coli isolates are still a concern. Liu et al. (2018) documented high rates of mcr-1-harboring E. coli (mainly enterohemolytic E. coli) in sick pigs from southern Taiwan. A high rate (46.2%) of colistin resistance among zoonotic Salmonella spp. has also been reported (Chiou et al., 2017). The mcr-1 gene has been identified in humans and retail meats (Kuo et al., 2016;Lai et al., 2018). Kuo et al. (2016) described 18 colistin-resistant E. coli isolates positive for mcr-1 in ground beef, chicken, and pork. The prevalence of mcr-1 in meat-associated E. coli isolates has increased in Taiwan, with rates of 1. 1, 6.6, and 8.7% in 20121, 6.6, and 8.7% in , 20131, 6.6, and 8.7% in , and 20151, 6.6, and 8.7% in , respectively (Kuo et al., 2016.
To our knowledge, no longitudinal studies have examined the prevalence of colistin-resistant E. coli in sick pigs. This study investigated the annual prevalence of colistin resistance and mcrharboring E. coli in sick pigs from different farms in Taiwan.

Bacterial Isolates From Sick Pigs
A total of 90 E. coli isolates were obtained from 90 sick pigs from 58 farms in different geographical regions of Taiwan between January 2013 and December 2018. A total of 88 isolates were collected from seven counties in southern (n = 4), northern (n = 1), middle (n = 1), and eastern Taiwan (two isolates were from unknown sources; Table 1). Most of the isolates were collected between 2015 and 2018 (n = 84, 93.3%), from southern Taiwan (n = 82, 91.1%). The age of the sick pigs was known for 79 isolates, 78 (98.7%) were <10 weeks old. Isolates from rectal swabs (n = 32) and mesenteric lymph nodes (n = 32) comprised 71.2% of the samples. The isolates were sent to the Animal Disease Diagnostic Center, College of Veterinary Medicine, National Pingtung University of Science and Technology, Pingtung, Taiwan, for isolation and identification of the enterohemolytic E. coli.

Identification of the Enterohemolytic E. coli
Escherichia coli isolates were identified as enterohemolytic strains by the presence of β-hemolysis on Trypticase soy agar supplemented with 5% sheep blood and by the presence of at least one virulent gene (STa, STb, LT, F18, or aidA) (Chapman et al., 2006;Lai et al., 2017Lai et al., , 2018.

Antimicrobial Susceptibility Testing
The minimum inhibitory concentrations (MICs) of 14 antimicrobial agents, including colistin, were determined on the VITEK 2 system (bioMérieux, Marcy-l'Etoile, France) ( Table 2) and were interpreted according to guidelines of the Clinical and  To examine the inter-test agreement between the methods for determining colistin susceptibility, the essential and categorical agreements and very major error (VME) were evaluated. The essential agreement between the BMD method and VITEK 2 was measured as the difference between MICs ≤ ± 1 log 2 dilution, using the BMD method as the reference standard. Categorical agreement between the two methods was measured as the percentage of isolates with concordant test results (i.e., intermediate vs. resistant). A VME was defined as inconsistent results, e.g., a colistin-resistant isolate from the BMD method that was considered colistin-intermediate by VITEK 2 (Chen et al., 2014).

Detection and Sequencing of the mcr Genes
Polymerase chain reaction (PCR) amplification of the wholecell DNA from colistin-resistant isolates (via the BMD method) was performed using previously described primers for mcr-1, mcr-2, mcr-3, mcr-4, and mcr-5 (Rebelo et al., 2018). The PCR products were sequenced.
The DNA extraction and purification for PFGE followed previous descriptions (Tenover et al., 1995;Hsueh et al., 2002). DNA was digested using the SmaI restriction enzyme, the restriction fragments were separated using a CHEF-DR III unit (Bio-Rad Laboratories, Hercules, CA, United States), a Essential agreement between the BMD broth microdilution method and the VITEK 2 susceptibility testing method was defined as the difference between MICs ≤ ± 1 log 2 dilution, using the BMD method as the reference standard. b Categorical agreement between the two susceptibility testing methods was measured as the percentage of isolates with concordant test results. c Seven (8.2%) isolates identified as colistin-resistant with the reference BMD method but were colistin intermediate with the VITEK 2 susceptibility method.  and the pulsotypes were analyzed on the Bio-Rad CHEF-Mapper apparatus (Bio-Rad Laboratories). Cluster analysis was performed in BioNumerics version 5.0 (Applied Maths, Sint-Martens-Latem, Belgium) using the unweighted pair-group method with arithmetic averages. The Dice correlation coefficient was used to analyze similarities between the banding patterns (tolerance = 1%). Isolates with identical PFGE patterns were considered the same strain (same pulsotype) and isolates with PFGE patterns >80% similar were considered closely related. Isolates exhibiting identical ST, RAPD patterns, and pulsotypes, or closely related strains by PFGE, were considered identical strains (clones).

Antimicrobial Susceptibilities
The MICs of 13 antimicrobial agents against the 90 porcine E. coli isolates are shown in Table 2. The susceptibility to ceftazidime and cefepime was 56.7 and 94.4%, respectively, and the susceptibility to ciprofloxacin and levofloxacin were both 54.4%. Only 11.1% of the isolates were susceptible to trimethoprimsulfamethoxazole. All of the isolates were susceptible to ertapenem, imipenem, and meropenem, and all were inhibited by tigecycline at 2 mg/L. The colistin-resistant isolates and  those harboring mcr-1 displayed similar susceptibilities to ciprofloxacin, levofloxacin, and amikacin ( Table 3). The colistinintermediate isolates had higher susceptibility to gentamicin than the colistin-resistant or mcr-1 positive isolates (both p < 0.0001) ( Table 3). The strains did not differ in their resistance to the other antibiotics. The distribution of colistin MICs determined by the VITEK 2 susceptibility system and the reference BMD method are depicted in Figure 1. The respective rates of colistin-intermediate isolates were 54.1% (n = 48) and 44.4% (n = 40). The essential agreement of MICs was 97.8%, and the rate of agreement was 91.1% (Table 4). A VME occurred in 16.7% (8/48) of the VITEK 2 susceptibility determinations.

DISCUSSION
The prevalence of the mcr-1 gene in porcine E. coli isolates from different geographical regions of Taiwan may be unrelated to a specific clonal population. The strains may carry different plasmids encoding the mcr-1 gene. The MLST of 32 mcr-1positive E. coli isolates from 18 retail meat samples and 14 human samples in Taiwan (2010-2015) revealed 18 distinct STs, including ST38 (n = 8), ST117 (n = 5), and two each of ST701, ST744, and ST428. The STs of the remaining isolates were distinct (Kuo et al., 2016). ST744 was also documented in this study, in addition to many other STs. For example, ST2521 was the most common ST, documented since 2017, but it had never been observed in pigs or humans (Kuo et al., 2016;Hadjadj et al., 2017;Kong et al., 2017;Lai et al., 2018;Wang et al., 2020). Our results suggest that further investigation is required to assess the clinical significance of ST2521 among mcr-1 positive E. coli isolates. Our findings also indicate that mcr-1-positive E. coli isolates from different geographical regions of Taiwan may have variable STs.
The RAPD patterns showed no evidence of clonal dissemination of the mcr-1-positive isolates between humans and pigs. In contrast, we identified five clone isolates from the MLST results, RAPD patterns, and pulsotypes. One clone was from sick pigs on different farms and also sick pigs on the same farm. This may indicate intra-and inter-farm spread of porcine E. coli harboring mcr-1. Thus, active and regular screening of mcr-1-containing E. coli isolates from humans and animals is imperative.
Animal-to-human transmission remains a serious concern. A study conducted in Laos reported two cases of colistin-resistant E. coli; one in a boy with no recent history of antibiotic usage and another in a pig that belonged to the boy's family. Both isolates belonged to the same novel ST and displayed the same virulence and PFGE patterns (Olaitan et al., 2015). The boy normally fed the pig without protective equipment (e.g., boots). The presence of the same ST4015 mcr-1-carrying E. coli strain in the boy and pig indicates possible horizontal transmission of colistinresistant E. coli. Another study described ST648 in two travelers and ST3997 in two villagers, both of which may represent interhuman transmission (Hadjadj et al., 2017). However, the clonality of the 39 E. coli isolates harboring the mcr-1 gene was diverse, and 18 different STs were detected (though some STs appeared more than once). Previously, we described six mcr-1-carrying E. coli isolates from patients with bacteremia; two were ST69, but the rest only occurred once (ST1196, ST361, ST1463, and ST1011) (Lai et al., 2018). The human mcr-1-positive E. coli isolates had different pulsotypes, and the STs were different from the porcine mcr-1-carrying E. coli isolates. Although we failed to detect mutual transmission between humans and pigs in Taiwan, we cannot exclude its occurrence. Further surveillance is necessary to identify potential transmissions of the mcr-1-carrying E. coli between animals and humans.
We observed a high VME (16.7%) in the VITEK 2 susceptibility testing of colistin compared to the BMD method. Our findings are contradictory to a previous study in human bacteremic E. coli isolates (no VME) and indicate that the VITEK 2 method is a low-sensitivity tool for identifying colistin resistance in porcine E. coli isolates (Lai et al., 2019). Gentamicin is frequently used to treat colibacillosis in pigs, especially in neonatal piglets (via intramuscular or oral administration). Of the 90 E. coli isolates, the resistance to gentamicin was 55.6% (50/90). High levels of gentamicin resistance were reported in E. coli isolates from sick pigs in several countries -32.7% in the United States (Jiang et al., 2019), 46% in Belgium, 45% in Poland, 20% in Spain, and 77% in Korea (Lee et al., 2009;Luppi, 2017). Plasmid-mediated antimicrobial resistance genes are transmissible and cross-resistance between gentamicin and other aminoglycosides such as apramycin (a veterinary drug) has been described (Jensen et al., 2006). However, the prevalence of antimicrobial resistance genes in aminoglycosides requires further investigation.
This study has several limitations. First, a small number of porcine E. coli isolates were evaluated and all were from sick pigs. This might limit the power of intra-and inter-farm spreading analysis of porcine E. coli isolates harboring mcr-1. The inclusion of more E. coli isolates from healthy and sick pigs raised on different farms may better describe the prevalence of mcr-1-harboring E. coli in swine and porcine farms. Second, the mechanisms mediating colistin resistance among the mcr-1-negative colistin-intermediate E. coli isolates were not investigated. Third, we investigated mcr-1 to mcr-5, but at least nine mcr genes have been reported. The roles of mcr-6 to mcr-9 should be investigated.
In conclusion, the occurrence of mcr-1-positive E. coli isolates in sick pigs has continuously increased in Taiwan. Regular screening for the mcr-1 gene in E. coli in sick pigs and their environment must be performed to prevent the spread of these resistant organisms.

DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

ETHICS STATEMENT
This study was approved by the Research Ethics Committee of National Taiwan University Hospital (201609066RINB).

AUTHOR CONTRIBUTIONS
S-CH, C-CL, and P-RH designed the experiments. S-CH, M-TC, and C-HL executed the lab experiments. S-CH, C-CL, Y-TH, C-HL, and C-NL analyzed the data. S-CH, C-CL, and P-RH prepared the manuscript. S-CH, C-CL, Y-TH, C-HL, M-TC, C-NL, and P-RH read and approved the final version of the manuscript.

FUNDING
This work was supported by grants from the Centers for Disease Control and Prevention, Minister of Health and Welfare, Executive Yuan, Taiwan (MOHW106-CDC-C-114-114701). The funder was not involved in the study design, data collection, analysis, or interpretation, writing the manuscript, or the decision to submit this work for publication.