Increasing Prevalence of ESBL-Producing Multidrug Resistance Escherichia coli From Diseased Pets in Beijing, China From 2012 to 2017

We investigated antimicrobial resistance trends and characteristics of ESBL-producing Escherichia coli isolates from pets and whether this correlates with antibiotic usage in the clinic. Clinical samples containing E. coli from diseased cats and dogs were screened for antibiotic sensitivity and associated genotypic features. We identified 127 E. coli isolates from 1886 samples from dogs (n = 1565) and cats (n = 321) with the majority from urinary tract infections (n = 108, 85%). High rates of resistance were observed for β-lactams and fluoroquinolones and resistance to > 3 antibiotic classes (MDR) increased from 67% in 2012 to 75% in 2017 (P < 0.0001). This was especially true for strains resistant to 6–9 antibiotics that increased from 26.67 to 60.71%. Increased rates in β-lactam use for clinical treatment accompanied these increasing resistance rates. Accordingly, the most frequently encountered subtypes were blaCTX–M (n = 44, 34.65%), blaCTX–M–65 (n = 19) and blaCTX–M–15 (n = 18) and qnrB (n = 119, 93.70%). The blaCTX–M-isolates possessed 36 unique pulsed field electrophoretic types (PFGEs) and 28 different sequence types (STs) in ST405 (7, 15.9%), ST131 (3, 6.8%), ST73, ST101, ST372, and ST827 (2, 4.5% each) were the most prevalent. This data demonstrated a high level of diversity for the blaCTX–M-positive E. coli isolates. Additionally, blaNDM–5 was detected in three isolates (n = 3, 2.36%), comprised of two ST101 and one ST405 isolates, and mcr-1 was also observed in three colistin-resistant E. coli with three different STs (ST6316, ST405, and ST46). Our study demonstrates an increasing trend in MDR and ESBL-producing E. coli and this correlated with β-lactam antibiotic usage for treatment of these animals. This data indicates that there is significant risk for the spread of resistant bacteria from pets to humans and antibiotic use for pets should be more strictly regulated.


INTRODUCTION
Antimicrobial resistance has become one of the most challenging problems for public health and results in 700,000 deaths annually (O'Neill, 2016). Antibiotic misuse has led to the spread of antibiotic resistance genes (ARGs) in humans, food animals, pets, songbirds, water, and soil and even agricultural plants, and this represents a significant threat to public health security (Carter et al., 2018;Hartantyo et al., 2018;Anderson et al., 2019;Chen et al., 2019;Gros et al., 2019;Sanchez et al., 2019;Vikesland et al., 2019). However, novel ARGs have emerged that encode resistance to carbapenems (bla NDM , bla IMP , bla VIM , and bla KPC ) (Perez and Bonomo, 2019) colistin (mcr-1 to 9) and tigecycline (tetX3 and tetX4) (Carroll et al., 2019;He et al., 2019;Wang et al., 2019). The current limited development of novel drugs and substitutes makes the use of ARGs monitoring even more important to develop comprehensive and integrative measures for antimicrobial resistance.
Over the past two decades, there has been a significant number of infections caused by bacteria expressing extendedspectrum-β-lactamase (ESBL) and carbapenemases (Logan and Weinstein, 2017). In particular, ESBL isolates have been found in humans (Pitout and Laupland, 2008), animals, the environment (water and soil) (Runcharoen et al., 2017), meat and even vegetables (Yang et al., 2019). ESBL are becoming more common because this phenotype is being selected for by the use and exposure to β-lactams, especially the cephalosporins. This has generated a vicious cycle of drug resistance and decreased therapeutic effects. The increasing use of cephalosporin has been linked to Escherichia coli infections in pigs (Hammerum et al., 2014) and a high frequency of ESBL-producing E. coli was directly linked with a high consumption of third-or fourthcephalosporins (Andersen et al., 2015).
One area that has not been thoroughly investigated is ARG presence in companion animals. This group comes in intimate contact with humans and pet contact can lead to bacterial spread to humans (Lloyd, 2007). In particular, E. coli is a common pathogenic agent isolated from pets (Mathers et al., 2015) and is often present in dogs and cats with urinary tract infections (UTIs), pyometra and respiratory tract infections (Karkaba et al., 2019;Moyaert et al., 2019). Multidrug resistance (MDR) E. coli isolates have emerged in companion animals in the United States and Europe (Morrissey et al., 2016), but data for China is lacking.
Here, we explored the effects of antibiotic dose on resistance phenotypes in pet bacterial isolates over a 6-year period. The aim of this study was to investigate the dissemination of ESBLproducing multidrug-resistant pathogens in diseased pets and the correlation between resistance rates and consumption of β-lactam antibiotics.

Samples and Identification of Bacterial Isolates
Animal samples were collected at the Veterinary Teaching Hospital of China Agricultural University (VTH-CAU), between January 2012 and June 2017. This study was approved by the China Agricultural University Animal Ethics Committee and the approval document (No. AW08104102-2) (see Supplementary  Information). We collected 1886 samples from dogs (1565) and cats (321) and where some samples were gathered from different infections in the same pet. All samples consisted of urine (UTI) samples (1398, 74.1%) and samples from pyoderma (125, 6.7%), ear swabs (6.1%), effusions (115, 6.1%) and other specimens (132, 7.0%), which included pus from the uterus and soft tissue infections and several trachea lavage fluids.
Bacterial strains were recovered using blood agar and MacConkey agar plates that were incubated at 37 • C for 24 h and single pink colonies were collected from each isolation plate. Subsequently, the DNA of individual clones was extracted by Fast Pure Bacteria DNA Isolation Mini Kit (Vazyme Biotech, Nanjing, China) and used as templates for PCR. PCR amplification of the 16S rDNA gene were performed for all isolates as previously described (Brianna et al., 2013), and amplicons were sequenced to confirm bacterial genus using the BLAST algorithm 1 .

Transconjugation Assays and Whole Genome Sequencing
Conjugation assays were performed between clinical isolates and E. coli J53 to evaluate whether bla NDM and mcr-1 were mobilizable. Transconjugants were selected on MacConkey agar containing 100 mg/L sodium azide and 1 mg/L meropenem or colistin. Presumptive transconjugants were identified using PCR screening for bla NDM and mcr-1.
Transconjugant DNA was extracted and used for whole genome sequencing (WGS). A library of 250-bp paired-end was constructed by using a NEXT Ultra DNA Library Prep kit (New England Biolabs, Beverley, MA, United States) and sequenced using an Illumina HiSeq 2500 system at Bionova Biotech (Beijing, China). Raw data was de novo assembled using the SPAdes algorithm v.3.10.0. ARGs and plasmid incompatibility groups were analyzed using ResFinder v.3.2 2 , RASTtk v.2.0 3 and PlasmidFinder 2.1 4 , respectively.

PFGE and MLST Typing of E. coli Strains
The clonal relatedness of bla CTX-M positive was determined using pulsed-field gel electrophoresis (PFGE) typing conducted as previously described (Tenover et al., 1995). PFGE patterns were visually inspected and gel images were analyzed using InfoQuest FP software (Biorad, Hercules, CA, United States). Group analysis of PFGE profiles was performed using the Dice coefficient and the unweighted pair group method with arithmetic means. Simultaneously, multilocus sequence typing (MLST) analysis was conducted using the following E. coli gene set: recA, adk, fumC, icd, mdh, purA, and gyrB. The results were interpreted using the MLST database 5 . Sequence types (STs) of 131 clades in E. coli isolates were also screened using multiplex conventional PCR assays as previously described (Matsumura et al., 2017).

Statistical Analysis
Tests of statistical significance was determined using Fisher's exact test with Yates continuity correction in GraphPad Prism 6 (San Diego, CA, United States) and the level of significance
We performed a correlation analysis between antibiotic usage and the rate of antibiotic resistance between 2014 and 2017. Amoxicillin-clavulanic acid (1.3 × 10 7 mg/yr), doxycycline (5.7 × 10 6 mg/yr), ampicillin (1.9 × 10 6 mg/yr), and enrofloxacin (8.1 × 10 5 mg/yr) were the most widely used antibiotics at the animal facility. With the widespread use of FQs in VTH-CAU, >50% of the clinical E. coli isolates showed resistance to FQs since 2012. Interestingly, even though fluoroquinolone usage had decreased dramatically, resistance rates to this antibiotic class remained high. In addition, the use of β-lactams, including amoxicillin-clavulanic acid, ampicillin, and ceftriaxone, had increased over the study period and was positively correlated with an increase in β-lactam resistance that was also related to dosage. Meropenem and imipenem were not used at the facility and we found no resistance from 2012∼2015 although three resistant E. coli were detected in 2016 (Supplementary Figure S3).
We further examined our mcr-1 and bla NDM isolates and tested whether these genes were present on mobile elements. In our conjugation tests, only bla NDM in three E. coli isolates were successfully transferred at frequencies of 4.86 × 10 −8 -8.02 × 10 −7 . Two complete 46,161 bp bla NDM-5 -harboring plasmids pP16NDM-502 (MN701974) and pP16NDM-503 (MN701975) were obtained from strains 16DU02 and 16DF03 transconjugants, respectively. The backbone sequences were assembled and contigs and gaps were identified by additional PCR and sequence analyses. The other NDM-1-carrying plasmid was unsuccessfully assembled because of fragmentary and short contigs (Supplementary Figure S6). The two completely assembled plasmids were all in the IncX3 replication group. In addition, bla NDM-5 was contained within an insertion sequence (IS) cassette ( ISAba125-IS5-bla NDM -ble-trpF-dsbC-IS26). bla NDM-5 and ble MBL were the only ARGs present in the two plasmids. Homology analysis revealed that pP16NDM-502 and pP16NDM-503 were ≥ 99% identical to the following IncX3 bla NDM plasmids: (i) pNDM_MGR194 (KF220657) from a Klebisella pneumoniae human isolate in India, (ii) p1079-NDM (MG825384) from a chicken E. coli isolate in China. (iii) pL65-9 (CP034744) from E. coli in goose in China, (iv) pZHDC40 (KY041843) from E. coli human isolate in China, (v) pQDE2-NDM (MH917280) from K. pneumoniae human isolate in China, (vi) pCRCB-101_1 (CP024820) from a Citrobacter freundii human isolate in Korea (vii) p128379-NDM (MF344560) from Enterobacter hormaechei from a human sample in China and (viii) pAD-19R (KX833071) from a chicken E. coli isolate in China (Supplementary Figure S6). We were unable to transfer the mcr-1 gene by conjugation in three separate tests and we could not assemble a complete plasmid sequence. However, no complete plasmid sequence of mcr-1 was successfully assembled by using three clinical isolates genomes. Although the analysis of incompatibility group in three clinical isolates genomes by analyzed PlasmidFinder-2.0 Serve revealed they possessed IncFIB, IncFIC, IncFII, and IncHI2 typical fragments, all mcr-1 genes weren't located on those fragments (data not showed).

PFGE and MLST Typing
In our group of 127 E. coli isolates, most (n = 109, 85.8%) were successfully characterized by PFGE typing and included 38/44 bla CTX-M -positive isolates. These 38 isolates were obtained from cats (6) and dogs (32) and could be subdivided into 36 unique PFGE patterns. Interestingly, two isolates obtained from urine (15cu184) and perirenal effusion (15cu186) samples showed identical PFGE patterns and carried the same bla CTX-M-127 subtype but were recovered from different animals (Figure 2).

DISCUSSION
In this study, we investigated the antibiotic resistance profiles and trends in MDR E. coli isolates collected from diseased dogs and cats. We identified 127 E. coli isolates and most were FIGURE 2 | XbaI pulsed-field gel electrophoresis (PFGE) analysis of E. coli containing bla CTX variants. associated with UTI (84.2%) accounting for 20.4% (108/529) of the confirmed bacterial UTI cases. In contrast, in the United States nearly 30% of UTI isolates from pets were E. coli (Ling et al., 2001;Hall et al., 2013). Previous studies in European countries based on 22,256 isolates from dogs and cats with UTI between 2008 and 2013 showed that E. coli was the most common pathogen in both dogs (59.50%) and cats (59.31%) (Marques et al., 2016). Isolation of E. coli from the respiratory tract is much less frequent and occurred in 10-15% of dog respiratory tract infections (Rheinwald et al., 2015;Morrissey et al., 2016).
Our 127 clinical E. coli isolates showed high prevalence rates of resistance to β-lactams (58.3-77.9%) and quinolones (52.8-57.4%). This pattern was quite different from the data of dog and cat E. coli isolates from United States that identified resistance to ampicillin at 40%, cephalexin 98% and doxycycline at 100% (Thungrat et al., 2015). An Australian study revealed that canine clinical E. coli isolates had low rates of resistance to quinolones (9.1-9.3%) and among 392 canine UTI isolates, 9.9-10.2% were resistant to third-generation cephalosporins (Saputra et al., 2017).
In the current study, the overall MDR frequency of 73.2% was higher than that observed in studies from the United States (52%) and Poland (66.8%) (Rzewuska et al., 2015;Thungrat et al., 2015). In contrast, a European multi-centre study on AMR of various bacteria isolated from companion animals with UTIs showed a much lower frequency of MDR among E. coli isolates (1.4-29.7%). The high MDR rates observed in the current study indicated that currently available antimicrobial treatment options for E. coli infections in companion animals are limited. We also identified an increase in E. coli MDR prevalence between 2012 and 2017 and increased rates of thirdgeneration cephalosporins and amoxicillin-clavulanic acid usage was accompanied by increased AMR and MDR rates. Based on these findings, we speculate that the rising incidence of multidrug-resistant E. coli might be associated with the heavy use of these antibiotics in companion animals at the hospital.
Among the 44 CTX-M-producing E. coli isolates in this study, bla CTX-M-65 (43.2%) and bla CTX-M-15 (40.9%) were the most bla CTX-M variants. Currently, more than 220 CTX-M-lactamases have been reported and clustered into five subgroups, containing CTX-M-1, -2, -8, -9, and -25 depending on amino acid sequence homology (Peirano and Pitout, 2019). CTX-M-15 (CTX-M-1) is the most frequent CTX-M variant worldwide (Karim et al., 2001) especially in South-East Asia, China, South Korea, and Japan (Peirano and Pitout, 2019). Since 2000, bla CTX-M-15 has emerged worldwide and is the most prevalent ESBL globally (Peirano and Pitout, 2019). The spread of bla CTX-M-15 contributed to the dissemination of MDR bacterial isolates in animals but also in humans; a significant public health concern (Mugnaioli et al., 2006;Canton et al., 2012;Liu et al., 2016b). While the bla CTX-M-15 variant was also common in canine E. coli isolates from both Shanxi, China (n = 40) and the United States (n = 50), bla CTX-M−123 predominated in those studies (Liu et al., 2016a,b). The bla CTX-M-65 was the most prevalent subtype we found and is frequently detected in Salmonella isolates. In a previous study, the bla CTX-M-65 gene (n = 131) was identified among 153 ESBL-positive Salmonella isolates from poultry slaughterhouses (n = 121) and humans (n = 10) (Bai et al., 2016). Similarly, a study on chickens and pigs in China surveyed seven bla CTX-M-65 among Salmonella isolates . All these studies indicated more frequent occurrences in China and is supported by data from 2005 to 2017 in another study (Bevan et al., 2017). Additionally, bla CTX-M-55 prevalence has increased in recent years in China among animal and human isolates, although not in our present study. Conversely, bla CTX-M-14 has a remarkable reduction based on data for China (Bevan et al., 2017). We found only one bla CTX-M-14 isolate while this gene is globally present (Bevan et al., 2017). Although other bla CTX-M variants, such as bla CTX-M-64 , -174 , -116 , and -127 rarely occur, their presence in diseased pets suggests that diversity and evolution of bla CTX-M had occurred in these companion animals.
All bla CTX -carrying isolates in the current study had MDR profiles and showed high resistance rates to β-lactams, quinolones, doxycycline, gentamycin, and chloramphenicol. This suggested that other resistance genes may be co-transferred with bla CTX , making it even more difficult to eliminate the spread of MDR. Most bla CTX-M positive strains in other studies were also resistant to quinolones due to topoisomerase modifications of qnr genes (Lahlaoui et al., 2014) as we found in the present study. Additionally, mutations in gyrA and parC also can be responsible for quinolone resistance. Our results showed S83L and D87 alterations to N, Y, G generated in gyrA were commonly associated with resistance to fluoroquinolones and mostly generate high level resistance (Basu and Mukherjee, 2019).
Only three E. coli isolates containing bla NDM-5 were identified in our study animals with UTI in late 2016. Five (3.94%) were colistin resistance and three carried mcr-1. Previous reports implied that the transmissible bla NDM and mcr-1-carrying plasmids play a major role in the dissemination of these genes . For example, mcr-1, initially named as mobile colistin resistance gene, was generally considered as mediating the rapid spread of bacterial colistin resistance worldwide due to its mobile plasmid association. In contrast, we found that mcr-1 could not be mobilized by conjugation. In a previous study, 14/23 mcr-1-positive isolates were successfully transferred and six plasmids were non-transferable (Zhou et al., 2017). The mcr-1 gene was located in the chromosome of E. coli from a goose isolate . Furthermore, ISApl1 transposon can mediate mcr-1 transfer from chromosome to plasmids and this is the reason for its current global distribution (Wang R. et al., 2018). So, mcr-1-carried in a non-transferable plasmid or chromosome may be spread by ISApl1 or others transposons. In our study, we could not assemble a complete plasmid sequence for the mcr-1 genomic data; and mcr-1 genes were not present on those plasmid fragments and may be chromosomal. There are nine variants of mcr and we identified only mcr-1 (Carroll et al., 2019). This gene is the most prevalent variant globally and has been detected in almost 40 countries/regions across five continents in both involving developed and non-developed countries. Moreover, the mcr-1 gene was found in more than 11 bacterial species and in diverse locations such as rivers, public beaches, well water, wastewater, hospital sewage, foods (vegetables and meats), animals (wild birds, housefly/blowfly, cattle, pigs, poultry, and companion animals) (Feng, 2018). Our study provided new evidence for the above claim and we identified three mcr-1 positive E. coli in diseased pets. Recently, carbapenem-resistant Enterobacteriaceae (CRE) have posed a threat to humans and animals because they exhibited resistance to most β-lactams including carbapenems, further compromising treatment of MDR infections (Gupta et al., 2011;Potter et al., 2016).
Carbapenem-resistant Enterobacteriaceae is mediated largely by the production of carbapenemase especially for NDM isolates. NDM-5, an NDM-1 variant, exhibited increased enzyme activity to carbapenems (Rogers et al., 2013) and its gene bla  has been reported worldwide (Khan et al., 2017) and is the most prevalent variant in China . In support of this, 84 (52%) NDM-5-producing E. coli were collected from 161 bla NDM carrying CRE in chickens . Our study revealed that pets have become a reservoir of NDM-5-producing E. coli. The bla NDM-5 gene is also associated with different plasmids such as IncFIA/B, IncFII, IncN, and IncX3 (Sun et al., 2015;Tyson et al., 2019;Zhang et al., 2019) and IncX3 was the dominant type Zhang et al., 2019). Similarly, the bla NDM-5 -harboring plasmid was assigned to the IncX3 type, and their transferability was confirmed in our study indicating a risk of bla NDM-5 plasmid transfer between bacteria. However, further studies should be conducted to determine the origins of bla NDM and mcr-1 in the E. coli isolates from companion animals at the hospital. This will be helpful in designing measures to control the spread of MDR isolates.
All CTX-M-producing E. coli in our study displayed a diversity of PFGE patterns and STs demonstrating that bla CTX-M encoding ESBLs are present in diverse E. coli. Additionally, our diseased pet samples were from animals having no prior contact suggesting that clonal spread had a low frequency and it was not likely that clonal spread of bla CTX-M -positive E. coli occurred between the pets. But ESBL or other resistance genes can transfer between different E. coli by transferable genetic elements, as previously reported (Kim et al., 2019). However, two E. coli isolates (15cu184 and 15cu186) from same cat but different sample sources shared the same PFGE, ST, and bla CTX-M−127 types, indicating that clonal spread may be occurring. It was reported that dissemination of Enterobacteriaceae can occur between pets and their owners by both horizontal transfer and clonal expansion (Yao et al., 2016).
In our study, ST405 E. coli (n = 7) was the most prevalent isolates carrying bla CTX-M . They presented different bla CTX-M subtype genes including bla CTX-M-15 (n = 3), bla CTX-M-65 (n = 3), and bla CTX-M−64 (n = 1). In a clinical study from King Abdulaziz Medical City (KAMC) in Riyadh (Alghoribi et al., 2015), six ST405 E. coli harboring bla CTX-M were detected in UTI and ST405 (P ≤ 0.02) were significantly associated with ESBL production. ST405 extended-spectrum β-lactamase-producing E. coli (ESBL-EC) was also detected in animals, in which one ST405 ESBL-EC was screened in barbary macaques (Macaca sylvanus) in Algeria, notably, it held the mcr-1, bla TEM−1 , and qnrB19 genes and bla CTX-M-15 (Bachiri et al., 2017). We observed a similar pattern and a UTI25 (ST405 ESBL-EC) isolate harbored mcr-1, bla TEM , qnrB, and bla CTX-M−64 . Additionally, ST405 ESBL-EC (16XXI8) carried bla NDM-5 , bla OXA , bla TEM , and qnrB. These data indicated ST405 ESBL-EC has become the vector of multiple resistance genes, containing carbapenemand colistin-resistance genes. In a Sweden project, ST405 ESBL-EC was also found in UTI from a diseased cat (Bogaerts et al., 2015). We found seven ST405 ESBL-EC that were all recovered from dogs but in different years. Thus, companion animals already contain the hosts of ST405 ESBL-EC. In addition, except for ST405, we also identified two ST101 E. coli carrying bla NDM-5 (16DU02 and 16DF03). They were from the same animal source and possessed the same ST, PFGE, plasmid type and bla NDM-5 , suggesting a clonal origin. ST101 was reported to be strongly associated with the NDM genotype although most were NDM-1 rather than NDM-5. Recently, five bla NDM-5positive E. coli were identified associated with ST101 and ST1196 (Ranjan et al., 2016;Aung et al., 2018). In contrast, only two ST101 NDM-5-producing E. coli were observed in another study and most were ST167 and ST410 but were clonally spread in a hospital (Sun et al., 2019). This was similar to our study and suggested that ST101 may be becoming the most important clone for dissemination of bla NDM-5 , similar to ST167 and ST410.
The clone ST131 was observed in three E. coli isolates. This is the predominant E. coli lineage in extraintestinal pathogenic E. coli (ExPEC) isolates worldwide and are associated with global community and nosocomial dissemination (Marie-Hélène et al., 2014). Currently, ST131 is classified as three clades: A, B, and C (Peirano and Pitout, 2019). In our study, all ST131 belonged to clade C and ST131 is in clade C 80% of the time (Peirano and Pitout, 2019). Clade C evolved from clade B and further evolved into C1 and C2. Recently, a new C1 subclade C1-M27 was identified in animal and human (Matsumura et al., 2016), including companion animals (Melo et al., 2019), and we found one C1-M27 subclades that are rare in China, but prevalent in Europe (Ghosh et al., 2017;Merino et al., 2018). The other ST131 belonged to subclade C1-nM27. C1 is commonly associated with quinolone resistance, in agreement with our results. The ESBL-EC isolates were commonly associated with ST131, where 40-80% of ESBL ExPEC belong to ST131 E. coli (Marie-Hélène et al., 2014), and CTX-15 was the most prevalent ESBL enzyme in ST131 ESBL-EC (Alghoribi et al., 2015). Our study demonstrated that bla CTX-M-65 was found in two ST131 ESBL-EC and one bla CTX-M-15 isolate. Almost all ST131 isolates were resistant to fluoroquinolones, especially for ESBL CTX-M-15 isolates (Marie-Hélène et al., 2014). We found similar results that almost all CTX-M ESBL isolates (40/44) were resistant to quinolones. ST131 ESBL E. coli possessed more frequent resistance to amikacin than non-ST131 ESBL isolates, but showed more frequent susceptibility to gentamicin or trimoxazole. On the other hand, the non-ESBL isolates such as E. coli ST131 were more frequently resistant to fluoroquinolones than non-ST131 isolates. Therefore, the resistance to quinolones may be associated with ST131 rather than ESBL presence (Marie-Hélène et al., 2014). These findings indicated that quinolones resistance may have been the predecessor of ESBL enzymes. We found that ST131 and ST405 were correlated to ESBL production as previously found (Alghoribi et al., 2015). Other STs in our study occurred infrequently. Recently, a water sample study revealed that ESBL-producing E. coli isolates were present 15 different STs (ST10, ST46, ST48, ST58, ST69, ST101, ST117, ST131, ST141, ST288, ST359, ST399, ST405, ST617, and ST4530) (Said et al., 2016), We detected ST46, ST101, ST131, ST359, and ST405 in our study. Multiple ST types (ST46, ST1286, ST10, ST29, ST101, and ST354) have also been found in chickens and they carried mcr-1 and produced ESBLs (Wu et al., 2018). In our work, 28 diverse STs in ESBL-EC were found.

CONCLUSION
We found a high prevalence of MDR E. coli isolates from diseased cats and dogs in Beijing and this rate has markedly increased over the last 6 years. The widespread use of thirdgeneration cephalosporins and amoxicillin-clavulanic acid at the veterinary teaching hospital has likely contributed to the increasing frequency of β-lactam resistance in these isolates. These strains carried bla CTX , bla NDM-5 , and mcr-1 and most possessed an MDR profile. Diversity analysis of the PFGE patterns and STs of these clinical E. coli isolates from different origins suggested that the dissemination of bla CTX have broad reservoirs of E. coli. Future studies should be undertaken to identify the MDR transmission mechanisms and establish national standards for the rational use of antibiotics in companion animals.

DATA AVAILABILITY STATEMENT
All datasets generated for this study are included in the article/Supplementary Material.

ETHICS STATEMENT
This study was approved by the Agricultural University Animal Ethics Committee, China. The animals were given the best practice veterinary care and informed consent was granted by the owners.

AUTHOR CONTRIBUTIONS
ZX was responsible for the study design. YC, ZL, YZ, and ZZ assisted in the data collection. ZL, YC, and LL interpreted the data. ZL, YC, and ZX completed the report writing. All authors revised, reviewed, and approved the final report.

FUNDING
This work was supported by the Beijing Science and Technology Plan Project (Z171100001517008) and the National Natural Science Foundation of China (31422055).