key: cord-0428006-100tpyif
authors: Nadimpalli, Maya L.; Stegger, Marc; Viau, Roberto; Yith, Vuthy; de Lauzanne, Agathe; Sem, Nita; Borand, Laurence; Huynh, Bich-tram; Brisse, Sylvain; Passet, Virginie; Overballe-Petersen, Søren; Aziz, Maliha; Gouali, Malika; Jacobs, Jan; Phe, Thong; Hungate, Bruce A.; Leshyk, Victor O.; Pickering, Amy J.; Gravey, François; Liu, Cindy M.; Johnson, Timothy J.; Hello, Simon Le; Price, Lance B.
title: Leakiness at the human-animal interface in Southeast Asia and implications for the spread of antibiotic resistance
date: 2021-03-16
journal: bioRxiv
DOI: 10.1101/2021.03.15.435134
sha: 3070eb70407464c25b0ee32e2f1a32b98009036a
doc_id: 428006
cord_uid: 100tpyif

International efforts to curb antimicrobial resistance have focused on drug development and limiting unnecessary use. However, in areas where water, sanitation, and hygiene infrastructure is lacking, and where biosecurity in food-animal production is poor, pathogen-flow between humans and animals could exacerbate the emergence and spread of resistant pathogens. Here, we compared mobile resistance elements among Escherichia coli recovered from humans and meat in Cambodia, a country with substantial connectivity between humans and animals, unregulated antibiotic use, and poor environmental controls. We identified multiple resistance-encoding plasmids and a novel, blaCTX-M and qnrS1-encoding transposon that were widely dispersed in both humans and animals, a phenomenon rarely observed in high-income settings. Our findings indicate that plugging leaks at human-animal interfaces should be a critical part of addressing antimicrobial resistance in low and middle-income countries.

4 Introduction:

Low and middle-income countries are projected to experience the greatest mortality and 56 economic fallout from the looming antimicrobial resistance crisis. Global public health 57 organizations are collaborating with countries to develop "One Health"-oriented national action 58 plans aimed at improving antimicrobial stewardship in both humans and animal production.

However, the importance of environmental controls at the human-animal interface has not 60 received the same level of attention as antimicrobial use. This lack of prioritization may be due (Gunilla Ström et al., 2018) suggesting that untreated drinking water may be a source of animal were ineffective (G. Ström et al., 2018) and antibiotics that are critically important for human 148 medicine (e.g. colistin) were used extensively in poultry production (Carrique-Mas et al., 2015) .

Such practices mean that the same antibiotic resistance genes could confer a selective 150 advantage in multiple hosts. Previously, we found that extended-spectrum β -lactamase (ESBL)-

producing Escherichia coli colonizing healthy community members in Phnom Penh, Cambodia

were highly similar to strains recovered from meat and fish sold at markets (Nadimpalli et al., Here, we analyzed ESBL-encoding elements among E. coli of human and animal origin to 175 determine if strains from these two sources shared a common pool of mobile resistance 176 elements. We first used long-read sequencing to assemble high-quality draft genomes of five 177 ESBL-producing E. coli isolates from the feces of healthy humans (n=2) and from pork meat 178 (n=2) and chicken (n=1) sold at Phnom Penh markets (Supplementary Text, Tables S1-S2,

Fig. S1). We identified the ESBL-encoding plasmids they harbored, annotated them, and then 180 screened Cambodian collections of ESBL-producing E. coli from healthy, gut-colonized humans 181 (n=88), clinical specimens (n=15), and meat and fish from markets (n=93) for their presence 182 (Data S1).

We identified four distinct bla CTX-M-55 -encoding plasmids that were shared with slight variations 186 among E. coli isolates of both human and animal origin (Fig. 3) . One IncHI1-type plasmid TnCTX-M/qnrS (Fig. 4) ; plasmid C did not. E. coli and S. enterica strains harboring reference 206 plasmids A through D were identified using the NASP pipeline; strains that mapped to at least 207 70% of the reference plasmid genome with at least 10x sequencing depth were considered 208 matches. Regions of similarity between these matches and each reference plasmid were 209 visualized using the GView server using default parameters (80% identity threshold for coding 210 sequences). Each concentric circle represents a unique match and is colored by origin. The 211 arrows indicate coding sequences identified on the reference plasmid.

In addition to being distributed among multiple vertebrate hosts, these four ESBL-encoding 214 plasmids were also detected among E. coli from diverse genetic backgrounds (Supplementary 215 Text, Data S1). This indicated that plasmid sharing across hosts was not exclusively driven by 216 the zoonotic transmission of specific bacterial clones; rather, frequent mixing of host-adapted 217 strains likely allowed for the uptake of each plasmid into diverse genetic contexts, a 1 1 phenomenon others have observed in hospitals in high-income settings (Sheppard et al., 2016) .

Our findings suggest that in regions with high leakiness, focusing exclusively on the spillover of 220 specific bacterial clones may fail to capture the spread of mobile resistance elements across a 221 diversity of host-adapted strains. Frequent plasmid sharing could also create opportunities for 222 the evolution of novel, antimicrobial-resistant human pathogens.

Next, we compared the four ESBL-encoding plasmid types to determine if the ESBL genes 225 themselves shared a common origin. We identified a putative, bla CTX-M -encoding transposon that 226 was integrated across three of the four plasmids (Fig. 4) . This ~6 kb transposon, hereafter Cambodia (Fig. 2b, Fig. S2, Data S1 

The pervasive leakiness between humans and animals in countries such as Cambodia may 247 make it impossible to establish the host origins of mobile resistance elements. All three 248 plasmids identified in this study that encoded TnCTX-M/qnrS (Fig. 4) effective approach for addressing antimicrobial resistance in such settings. Instead, cost-258 effective strategies for implementing environmental controls at the human-animal interface, i.e., 259 "plugging the leaks," should actively be explored.

This study was limited by our lack of access to samples collected directly from food animals, 262 resulting in our use of meat and fish products as proxies. Theoretically, the bacteria that we 263 isolated from these products could have originated from human contamination. However, 80% 264 of the strains from meat and fish were resistant to phenicols (compared to <30% of the strains 265 from humans) (Nadimpalli et al., 2019b) , suggesting that food animals, which are regularly given 

Whole genome sequencing and isolate selection

We previously isolated and characterized ESBL-producing Escherichia coli (ESBL-Ec) 306 from gut-colonized, healthy humans (n=88) and market-origin chicken, pork, and fish (n=93) at 307 the Institut Pasteur du Cambodge from 2015-2016 (Nadimpalli et al., 2019b) . Briefly, food-origin 308 third-generation cephalosporin-resistant (3GCR) E. coli were isolated from 60 specimens each 1 5 of fish and pork and 30 specimens from chicken; human-origin 3GCR-E. coli were isolated from 310 fecal swabs of recently pregnant healthy women living in the same neighborhood where 311 markets were sampled. For one isolate per sample, ESBL production was confirmed via double-312 disk synergy testing and species identification by bioMérieux's API 20E system. DNA extraction, 313 library preparation, and whole genome sequencing were carried out on the Illumina NextSeq 314 500 platform using a 2x150 paired-end protocol (Nadimpalli et al., 2019b) . The raw reads were 315 pre-processed and assembled using SPAdes (Bankevich et al., 2012) , assigned a multilocus 316 sequence type (MLST) using the Achtman scheme (Larsen et al., 2012) , and screened for 317 acquired resistance genes using ResFinder v3.1.0 (selected threshold equal to 90% nucleotide 318 identity) (Zankari et al., 2012) .

As previously reported, we identified the bla CTX-M-55 gene to be the most common ESBL 

Thus, for the present study, we selected five E. coli CC10 isolates from food (n=3) and 326 humans (n=2) for long-read sequencing using the MinION platform (Oxford Nanopore 327 Technologies) to resolve the genetic context of their ESBL genes (Fig. S1 ).

Long-read sequencing, hybrid genome assembly and genome characterization Table S1 ). We used Unicycler v0.4.4 (Wick et al., 338 2017b) to construct high-quality hybrid assemblies using the Oxford Nanopore Technology long 339 reads and pre-existing Illumina short-read sequences from the same isolates, trimmed with 340 Trimmomatic v0.36 (Bolger et al., 2014) to remove sequencing adaptors and low quality (Q≤20) 341 ends. We used the CLC Genomics Workbench v10.1.1 (Qiagen) to manually correct the 342 assemblies, which were then annotated with the Rapid Annotation Using System Technology 343 Server (RAST) v2.0 (Aziz et al., 2008) . We screened the assemblies for acquired antibiotic 344 resistance genes using ResFinder v2. 1 (Camacho et al., 2009 ) (selected threshold equal to 345 90% nucleotide identity). Plasmid replicons were detected using PlasmidFinder v2.0 (Carattoli et 346 al., 2014) (selected threshold equal to 90% identity) and incompatibility group F (IncF) subtyping 347 was performed using PubMLST (Fig. S1) .

Draft genomes have been uploaded to NCBI's GenBank under Bioproject number 349 PRJNA566431. Individual accession numbers are available in Table S2 . that was smaller than a plasmid between strains (e.g. transposon) (Fig. S1) . We further 355 annotated the identified region using BLASTP and characterized insertion elements using 356 ISFinder (Siguier et al., 2006) . We screened multiple ESBL-producing Enterobacteriaceae collections from Cambodia 361 for the bla CTX-M -encoding elements identified in this study. These collections included the 362 aforementioned collection of ESBL-producing E. coli from food and healthy humans (n=181) 363 (Nadimpalli et al., 2019b) , ESBL-producing E. coli from clinical specimens, including urine and 364 blood (n=15) (Nadimpalli et al., 2019b) , ESBL-producing Salmonella enterica from food (n=26) 365 (Nadimpalli et al., 2019a) , and ESBL-producing Klebsiella pneumoniae from food (n=8) (Data 366 S1). All methods for isolating, characterizing, and sequencing the E. coli and Salmonella 367 collections have been described above and/or previously been published in detail (Nadimpalli et 368 al., 2019b (Nadimpalli et 368 al., , 2019a , K. pneumoniae were isolated from the same meat and fish samples as E.

coli and Salmonella, following the same culturing methods used to isolate 3GCR-E. coli 370 (Nadimpalli et al., 2019b) . K. pneumoniae were confirmed by MALDI-TOF. Library preparation 371 was performed using the Nextera XT V2 300-cycle Kit (Illumina) and whole genome sequencing 

(BWA) and identified SNPs using GATK UnifiedGenotyper. Strains that mapped to at least 70% 382 of the reference plasmid with at least 10x sequencing depth were considered to harbor the 383 plasmid. Regions of similarity between these matches and each reference plasmid were 384 visualized using the GView server using default parameters (80% identity threshold for coding 385 sequences, https://server.gview.ca/). To screen for b) TnCTX-M/qnrS we took a 2-step 1 8 approach. First, we extracted the protein-coding segments of this region and built a BLAST 387 database. We then generated de novo assemblies of the Cambodian ESBL-producing E. coli 388 genomes (n=196) using SPAdes (Bankevich et al., 2012) and queried them for the protein-389 coding genes in the transposon using BLASTN. We analyzed BLAST results using a custom R 390 script and arbitrarily included as a hit any contig where at least 50% of the proteins were present 391 with 99% coverage and 95% identity and where at least genes encoding a transposase protein 392 and a CTX-M variant were present. Second, we screened the Cambodian ESBL-producing E.

coli collections (n=196) for TnCTX-M/qnrS using NASP, as described above. Strains that 394 mapped to at least 99% of the reference region with at least 10x sequencing depth were 395 considered to harbor the region of interest. Findings from the two approaches were comparable.

Thus, we proceeded to use only the NASP pipeline to screen the remaining Cambodian 397 Salmonella (n=26) and K. pneumoniae (n=8) collections for the TnCTX-M/qnrS element. We 398 only report matches identified through the NASP pipeline here.

We constructed core genome phylogenies of our E. coli and S. enterica datasets to 400 visualize the presence of a) each reference plasmid and b) TnCTX-M/qnrS by isolate source 401 and sequence type or serotype. For each tree, we selected a published reference genome that 402 belonged to the same sequence type or serotype as other isolates in each dataset; i.e.,

CP011113 for E. coli (ST10) and CP010282 for S. enterica (serotype Newport). SNP alignments 404 were obtained by running NASP as described above and removing recombined regions using 405 Gubbins v2.3.4 (Croucher et al., 2015) . We annotated the resulting phylogenetic trees using 406 iTOL (Letunic and Bork, 2016 TnCTX-M/qnrS element (Fig. S1) Table S1 . Assembly statistics for draft Escherichia coli genomes constructed using Oxford 550 Nanopore Technology (ONT) long reads and Illumina short reads. 

Interactive tree of life (iTOL) v3: an online tool for the display and 472 annotation of phylogenetic and other trees

Ministry of Agriculture, Forestry, and Fisheries (MAFF). 2017. Annual Report for Agriculture

Ministry of Agriculture

Forestry, and Fisheries (MAFF)

Attributable sources of community-acquired carriage of 480 Escherichia coli containing β -lactam antibiotic resistance genes: a population-based 481 modelling study

CTX-M-55-type ESBL-489 producing Salmonella enterica are emerging among retail meats in Phnom Penh

Meat and Fish as Sources of Extended-494 Spectrum β -Lactamase-Producing Escherichia coli

Ministry of Planning

If it's a broad spectrum, it can 499 shoot better": inappropriate antibiotic prescribing in Cambodia

NASP: an accurate, rapid method for the identification of SNPs in WGS datasets 504 that supports flexible input and output formats

Nested Russian Doll-Like Genetic Mobility Drives Rapid 509 Dissemination of the Carbapenem Resistance Gene bla KPC

ISfinder: the reference centre 512 for bacterial insertion sequences

Manure management and public health: Sanitary and 515 socio-economic aspects among urban livestock-keepers in Cambodia

Antimicrobials in small-scale urban pig farming in a lower middle-519 income country -arbitrary use and high resistance levels

Water, sanitation and hygiene: UNICEF Country Programme 2019-522 2023

Global trends in antimicrobial use in food animals

Completing bacterial genome assemblies with 527 multiplex MinION sequencing

Unicycler: Resolving bacterial genome 530 assemblies from short and long sequencing reads

Pulling Together to Beat Superbugs: Knowledge and Implementation 533 Gaps in Addressing Antimicrobial Resistance

Identification of acquired antimicrobial resistance genes