Whole genome sequencing analysis of antimicrobial resistant Escherichia coli : from food to human

The main work of this study is to analyze antimicrobial resistant E. coli based on whole genome sequencing data. Three stages were included for isolates from different sources (ready-to-eat food, retail raw meats and human patients). At the first stage, a retrospective study for antimicrobial resistant E. coli in ready-to-eat (RTE) food sold in retail food premises in Singapore was performed in collaboration with Environmental Health Institute under NEA. A total of 99 E. coli isolates from poultry-based dishes (n=77) and fish-based dishes (n=22) were obtained between 2009 and 2014 during the surveillance project. All the isolates were included for disk diffusion testing for antimicrobial susceptibility testing. Of the 99 isolates, 24 (24.2%) were resistant to at least one antimicrobial agent. These isolates were then subjected to broth microdilution testing against 33 antimicrobial agents, including β-lactams, aminoglycosides, tetracycline, fluoroquinolones and polymyxin E (colistin), to determine the minimum inhibitory concentration (MIC) of isolates. Whole genome sequence (WGS) was carried out on the strains in order to correlate resistant phenotypes to putative antimicrobial resistance-related genes. Of the 24 isolates, 15 (62.5%) were found to be resistant to three or more classes of antimicrobials and thus were defined as multi-drug resistant strains. Two isolates (8.3%) were confirmed as Extended-Spectrum β-lactamase (ESBL)- producing E. coli by double disk synergy test. Based on WGS data, online analysis tool ResFinder detected 7 classes of antimicrobial resistance genes and resistance-related chromosomal point mutations in 19 of the 24 E. coli isolates. By analyzing the WGS contigs using BLASTn and KmerFinder, ESBL genes and transferable colistin resistance gene mcr-1 (2/24) and mcr-5 (1/24) were determined to be located on plasmids, which could pose a greater risk of AMR transfer among bacteria. Mutations were detected in four isolates within genes previously shown to confer resistance to quinolones (gyrA and parE) and tetracycline (rrsB). Prediction of AMR using WGS data was evaluated for six antimicrobials including ampicillin, chloramphenicol, colistin, fluoroquinolones, tetracycline and trimethoprim. The evaluation indicates WGS–based genotype and phenotype showed high consistency, however, for some antimicrobial resistance whose mechanism is not totally clear yet (such as colistin), there is challenge for resistance gene detection. To have a better understanding of genetic environment and mobility of colistin resistance gene mcr-5.1, the isolates carrying mcr-5.1 were sequenced using long-read sequencing technology. The plasmid sequences were assembled into a closed circle using both long-read sequencing data and short-read sequencing data. The blasting result showed the closest plasmid sequence to pSGMCR103 in NCBI is plasmid pYD786-3 (accession number KU254580.1) with 77% query coverage and 99% identity, which was carried by one E. coli isolate from human urine in USA. They share antimicrobial resistance gene aph(3’)-la, aadA1 (aminoglycoside resistance), blaTEM-176 (beta-lactam resistance) and sul3 (sulphonamide resistance). Gene mcr-5.1 was harbored on a Tn3 transposon-like element, which is similar with pSE13-SA01718 (accession number KY807921.1) carried by a Salmonella isolate reported before. Also, other insertion elements such as IS5, IS6, IS91, IS256 family were found on the plasmid, which may indicate the recombination activity of the plasmid. Moreover, the mobility of this plasmid was confirmed by the conjugation experiment. The frequency of conjugation after 24 hours is 10-6. After knowing the AMR profile in ready-to-eat food in Singapore, at the second stage, an important resistance type was studied: Extended-Spectrum Beta-Lactamase (ESBL)-caused resistance to most of beta-lactams. We collected 634 meat samples including chicken, pork and beef from 97 supermarkets and 65 wet markets in Singapore during June 2017-October 2018. The samples were enriched before bacteria isolation. Presumptive ESBLs were screened by Brilliance TM ESBL Agar and confirmed by Double Disk Synergy Test (DDST). E. coli isolates were identified by EMB agar and indole test. The genomic DNA of ESBL-producing E. coli were extracted and sent for WGS. Besides the analysis for AMR genes, MLST, annotation and genetic environment, these sequence collection was also compared with sequence data of ESBL E. coli isolated from community in Singapore for phylogenetic study based on SNPs. A total of 225 ESBL-producing E. coli were isolated from 184 samples. The prevalence of ESBL in chicken, pork and beef was 51.2% (109/213), 26.9% (58/216), 7.3% (15/205), respectively. The most common AMR genes in all 225 ESBL isolates were beta-lactam-resistance genes (100%), aminoglycoside resistance gens (92.4%), sulphonamide resistance genes (86.2%). In terms of beta-lactam resistance genes, 172 of isolates (76.4%) carry blaCTX-M genes, 102 (45.3%) of isolates carry blaTEM genes and 52 of isolates (23.1%) carry blaSHV genes. Besides these most common three beta-lactamase genes, blaCMY-2, blaOXA and blaDHA were also found. Gene blaCTX-M-55 (57/225, 25.3%) and blaCTX-M-65 (40/225, 17.8%) were the most frequent ESBL genes. Among all these classes of antimicrobials, beta-lactam-resistance genes and aminoglycoside resistance genes exhibit great variety. The last-resort antimicrobial colistin resistance mcr genes exist in 15.6% of all isolates (33 isolates carry mcr-1, one carries mcr-3.1 and one carries mcr-5). Phylogeny tree based on SNPs of our isolates and previous ESBL isolates from human community in Singapore shows obvious separate human clusters and food clusters, however, two E. coli isolates from human fell into food clusters and showed high similarity with our isolates from meats, which indicates the possible transmission of resistant E. coli from meats to human may exist. Occurrence of AMR genes especially for last resort drug resistance genes was observed, raising concerns on food safety and public health. At the last stage, we applied WGS to the analysis of clinical isolates. We collaborated with the university in Thailand to get 28 ESBL-producing E. coli isolated from diarrhea patients hospitalized at the Phayao Ram Hospital in Thailand. Result shows all E. coli carried CTX-Ms beta-lactamase (including CTX-M-14, CTX-M-15, CTX-M-27, CTX-M-55), and half of these isolates (14/28) belong to the important pathogenic cluster ST131. CTX-M-55s were detected only in non-ST131s. Two serotypes O16:H5 (6/14) and O25:H4 (8/14) were observed in ST131 isolates. Generally, ST131 isolates showed different virulence factor patterns with non-ST131 isolates. BLAST results indicate that for half of ST131 isolates, blaCTX-M genes are located on chromosome adjacent to insert sequences (IS). For the other half ST131s, blaCTX-M genes are located on plasmids. Besides CTX-Ms, other beta-lactamases such as TEM-1B, OXA-1 and CMY-2 were also observed in our study. Phylogenetic analysis for a global collection and our clinical isolates in Thailand based on SNPs showed the closest isolates with our isolates are from Thailand, Singapore, Australia, Laos and New Zealand. A special strain cluster O16:H5-ST131 was found from 2015-2017 as well as other previous Thailand studies. These isolates showed high similarity in term of serotype, MLST, virulence factor / AMR patterns, and phylogeny, which indicates the persistence and spread of this cluster. This study provides an insight on characteristics of clinical ESBL-producing E. coli with special focus on ST131 in Thailand. It is noteworthy that four isolates in our study showed same serotype, ST (O16:H5-ST131), same virulence factor pattern (cnf1, iha, sat, senB) and even resistance gene pattern. In terms of SNP analysis, these isolates were located the same cluster 3 and the SNP difference range 17-20, which indicate the high similarity of these strains. But actually they are isolated in different years from 2015-2017, respectively. In cluster 3, other four isolates from Thailand (ERR1218557, ERR1218609, ERR1218624, ERR1218628) in a previous study also showed less than 30 SNPs difference with our four isolates. This also strengthens the hypothesis that these isolates diverged from one ancestor and this cluster of isolates are persistent in Thailand in recent years. This strain should raise our attention of further spread. The phylogenetic study indicates that the closest isolates with Thailand ST131 isolates are from Oceania and Southeast Asia. Our data based on whole genome add more evidence on ubiquity of ESBL-ST131 E. coli in Thailand. The co-existence of multi-resistance and multi-virulence factors adds challenges to clinical treatment. Although resistances to carbapenem and/or colistin are rare in this study, more epidemiology studies are needed to verify their actual prevalence. Overall speaking, WGS is a useful tool for prevalence and epidemiological analysis and source tracking. With the development of sequencing technology and the decrease of cost, whole-genome-based analysis is becoming more and more necessary for AMR study.Doctor of Philosoph