Molecular ecology of highest priority critically important antibiotic resistant Escherichia coli from mammals housed at an urban zoo

Abstract Objectives Zoos are environments where species of highly valued animals are kept largely separated from others and the wider world. We report the molecular ecology of critically important antibiotic resistant (ABR) Escherichia coli carried by 28 mammalian species housed in a zoo located in an urban residential district. Methods Over 3 months we collected 167 faecal samples from captive mammals and processed for E. coli resistant to third-generation cephalosporins (3GC-R) and fluoroquinolones (FQ-R). Isolates were sequenced using Illumina. Results We identified high rates of faecal sample-level positivity, with 50%, 57% and 36% of mammalian species excreting 3GC-R, FQ-R or dual 3GC-R/FQ-R E. coli, respectively. Isolates represented multiple ST and ABR mechanisms; CTX-M-15 and CMY-2 dominated for 3GC-R, and target-site mutation caused 75% of FQ-R. We identified multiple examples of ABR E. coli transmission between mammalian species in separate enclosures, and a variant of the epidemic plasmid pCT within the zoo. There was no evidence for ABR E. coli leaving the zoo, based on comparative analysis with E. coli from humans, cattle and dogs isolated from the 50 × 50 km region in which the zoo is located. Amoxicillin/clavulanate was the most widely used antibiotic in the zoo, and we identified four widely disseminated amoxicillin/clavulanate resistance mechanisms, including a previously unreported inhibitor-resistant TEM, and the carbapenemase OXA-181. Conclusions We conclude that the zoo studied here is a ‘melting pot’ for the selection and circulation of 3GC-R and FQ-R E. coli, but these circulating E. coli appear captive within the zoo.


Introduction
Studies on antibiotic resistance (ABR) in zoo animals are rare and predominantly phenotypic, 1 with one including MLST 2 and two identifying ABR genes through PCR. 3,4 Characterization of ABR bacteria is useful when considering selection and transmission of ABR, including zoonotic transmission. 5 Our previous work has investigated One-Health transmission of ABR Escherichia coli within a 50 × 50 km study area. Focusing on third-generation cephalosporin resistance (3GC-R) and fluoroquinolone resistance (FQ-R), we have sequenced E. coli isolates from dairy cattle, dogs and humans. [6][7][8][9][10][11][12] Within our study area, in an urban residential district, is a zoo. It is the world's oldest provincial zoo, having been open since 1836. The aim of the study reported here was to investigate the molecular ecology of 3GC-R and FQ-R E. coli excreted by mammals within this zoo, to compare them with those from humans, cattle and dogs within our study area, and to investigate ABR selection drivers within the zoo.

Study population
Random sampling of pooled faeces from animal enclosures at Bristol Zoological Gardens was performed by keepers during routine enclosure cleaning at three monthly timepoints between October and December 2020. Three replicates were taken to represent each enclosure at each timepoint, and samples were stored at 4°C and processed within 48 h. Each faecal sample was identified by keepers as having come from 1 of 28 mammalian species (Table S1, available as Supplementary data at JAC Online). Faeces from animals isolated due to poor health were not collected, so some enclosures were sampled on fewer than three occasions. Animals were not observed and there were no changes to routine husbandry practices or to the personnel entering each enclosure.

Laboratory and genomics analysis
Replicate samples from the same enclosure and timepoint were pooled and weighed. Sample processing and selection of ABR E. coli was as described previously, 12 using cefotaxime (2 mg/L) or ciprofloxacin (0.5 mg/L). Up to three isolates per selective agent per sample were analysed using three multiplex PCR assays: one to identify mobile β-lactamase genes, 6 one to detect various bla CTX-M groups, 6 and one to detect plasmid-mediated quinolone resistance (PMQR) genes. 10 One isolate per PCR profile, per selective agent, per sample collection timepoint, per mammalian species was selected as representative for WGS, performed by MicrobesNG and analysed for ABR and phylogeny as previously described. 10,12 Quality control data are provided in Table S2. E. coli Fis2-e4 (NCBI accession CP041992.1) was used as the reference. Disc susceptibility testing and MIC assays were performed and interpreted according to CLSI guidelines. [13][14][15] Results and discussion

Overview of 3GC-R and FQ-R E. coli
One hundred and sixty-seven faecal samples were collected from the enclosures of 28 mammalian species across three monthly timepoints. Not all animals could be sampled at each timepoint. 3GC-R or FQ-R E. coli were found in at least one sample from at least one timepoint in 14 (50%) and 16 (57%) animal species, respectively. Ten (36%) species excreted both 3GC-R and FQ-R E. coli, and eight (29%) did not excrete E. coli resistant to either (Table S1).
One hundred and thirty-one resistant E. coli isolates were analysed by PCR to identify mobile 3GC-R and PMQR genes, and one isolate per PCR profile per mammalian species per sample visit was selected for WGS, making 51 isolates in total. Eleven E. coli STs were identified through WGS of 26 3GC-R E. coli, with ST1722 the most common (Table S3). 3GC-R was due to bla CTX-M-15 (58% of isolates), bla CMY-2 (35%) or bla CTX-M-14 (8%). In total, there were 14 ST/ABR gene combinations (Table S3). 3GC-R was predominantly plasmid mediated (Table S4). An identical (based on read mapping) ∼90 kb IncI1 bla CTX-M-15 plasmid (assembled as a single contig) was found in 42% of sequenced 3GC-R isolates. This plasmid matched at 98% coverage with 99.7% identity with pJIE139 (accession: EU418926.1) carried by an Australian human urinary E. coli. 16 A second IncI plasmid, being ∼94 kb and carrying bla CMY-2 , was found in six E. coli STs, comprising 35% of isolates and 25% of sampled mammalian species (Table S4). Read mapping against its closest match, p92 (accession number: CP023376) from dogs in the UK, 17 identified 70 SNP differences plus five insertions/deletions. There were three variants of this plasmid in the zoo, each differing by one SNP or insertion.
Among 25 sequenced FQ-R isolates, 14 ST/ABR gene combinations were identified (Table S5). Most FQ-R was caused by quinolone resistance-determining region (QRDR) mutation, with PMQR genes present in two STs carrying qnrS1 and one carrying qnrS2 plus aac(6′)-lb-cr (Table S5). In all cases, aac(6′)-Ib-cr was immediately upstream of bla OXA-1 , with qnrS2 also being adjacent, encoded on a novel IncFIB plasmid most similar to p14EC033f (accession number: CP024153.1). Figure 1 shows a mid-rooted phylogenetic tree based on core genome alignment of all 51 sequenced zoo isolates. In most cases SNP distances across groups of isolates from the same ST were <15, even across multiple mammalian species, indicative of ongoing circulation of isolates between animals within the zoo. However, none of the ST/ABR gene combinations found in the zoo (Tables  S3 and S5) were found among 609 3GC-R and/or FQ-R E. coli from humans, cattle or dogs in the same study area. [6][7][8]10,12 One example of potential ABR plasmid dissemination into the zoo was a bla CTX-M-14 plasmid (Table S4), with only five SNP differences from the IncK plasmid pCT. 18 pCT is classed as an epidemic plasmid for bla CTX-M-14 , being responsible for 30% of bla CTX-M-14 horizontal transfer in E. coli from humans, cattle and poultry in the UK, 19 and pCT has been found in our study region in E. coli from humans and cattle but not dogs. 6,8,12

β-Lactam/β-lactamase inhibitor resistance
Isolates representative of 3GC-R and FQ-R ST/ABR gene combinations were subjected to disc susceptibility testing (Table S6). Amoxicillin/clavulanate non-susceptibility was seen in all isolates producing CMY-2 or OXA-1, as expected. 20 All CTX-M producers, including those carrying bla TEM-1 , were amoxicillin/clavulanate susceptible, except for one ST1722 combination (Table S6). This was found to carry a bla TEM-33-like gene, which is not recorded on any public sequence repository. Like TEM-33, this novel TEM has a leucine at position 69, which has been associated with reduced inhibitor binding. 21 The only difference from TEM-33 is a Trp133Cys substitution. Additionally, a Pa −10 promoter mutation was found upstream of this bla TEM-33-like gene, where all bla TEM-1 genes found at the zoo had WT promoters ( Figure S1). Pa −10 promoter mutations confer amoxicillin/clavulanate resistance because they cause TEM overproduction. 20 We confirmed that the bla TEM-33-like gene is located on a p0111-type plasmid sharing 92% identity with pA16EC0618-5 (accession: CP088846.1). All sequenced bla TEM-33-like -containing plasmids were identical, confirming widespread circulation in the zoo.

Antibiotic usage
Analysis of electronic medical records allowed cumulative dispensing of antibiotics for treatment of mammals at the zoo to be reported ( Table 1). The highest usage (by weight) was of ciprofloxacin, amoxicillin, and amoxicillin/clavulanate. In terms of number of doses administered, however, and the breadth of species receiving these doses, amoxicillin/clavulanate was by far the most widely used, comprising 51.8% of all antibiotic doses (Table 1). It may well be, therefore, that high amoxicillin/clavulanate usage is a driver of the extensive array of amoxicillin/clavulanate resistance mechanisms seen among 3GC-R and/or FQ-R E. coli in the zoo. Importantly, this can co-select for resistance to piperacillin/tazobactam (CMY, OXA-1/-181), 3GCs (CMY), FQs [AAC(6′)-Ib-cr, QnrS, encoded alongside amoxicillin/clavulanate resistance genes] and ertapenem (OXA-181).  (13) Not all the mammalian species present at the zoo and receiving treatment over the 5 year period of this medicines usage survey were available for faecal sampling in this study. Numbers in brackets relate to the 28 mammalian species sampled in this study, if the total includes additional species not sampled.
Despite being in the heart of a residential part of the city, we found no evidence of 'escape' of the novel TEM, OXA-181 or OXA-1 plasmids identified above, or any other ST/ABR gene combination found in the zoo into E. coli colonizing dogs or cattle, or infecting humans in the same city, or surrounding region. ABR was found to be readily circulating within the zoo, however, even between animals in enclosures separated by some distance. This could be due to shared food or food preparation space, or because keepers or their equipment move between enclosures. However, wider sampling of people, equipment and the environment would be required to investigate further.