The highly diverse plasmid population found in Escherichia coli colonizing travellers to Laos and its role in antimicrobial resistance gene carriage

Increased colonization by antimicrobial-resistant organisms is closely associated with international travel. This study investigated the diversity of mobile genetic elements involved with antimicrobial resistance (AMR) gene carriage in extended-spectrum beta-lactamase (ESBL)-producing Escherichia coli that colonized travellers to Laos. Long-read sequencing was used to reconstruct complete plasmid sequences from 48 isolates obtained from the daily stool samples of 23 travellers over a 3 week period. This method revealed a collection of 105 distinct plasmids, 38.1 % (n=40) of which carried AMR genes. The plasmids in this population were diverse, mostly unreported and included 38 replicon types, with F-type plasmids (n=23) the most prevalent amongst those carrying AMR genes. Fine-scale analysis of all plasmids identified numerous AMR gene contexts and emphasized the importance of IS elements, specifically members of the IS6/IS26 family, in the evolution of complex multidrug resistance regions. We found a concerning convergence of ESBL and colistin resistance determinants, with three plasmids from two different F-type lineages carrying bla CTX-M and mcr genes. The extensive diversity seen here highlights the worrying probability that stable new vehicles for AMR will evolve in E. coli populations that can disseminate internationally through travel networks.


INTRODUCTION
Infections caused by antimicrobial-resistant organisms are harder to treat, lengthen hospital stays, increase mortality rates and place a significant financial burden on healthcare institutes [1]. It is increasingly important to characterize the mechanisms that allow antimicrobial resistance (AMR) to spread worldwide, compromising treatment options for many bacterial pathogens. The rapid spread of AMR has been closely associated with international travel [2][3][4][5]. Clinically relevant AMR determinants are commonly found in Gram-negative bacteria colonizing travellers returning from regions with the highest AMR prevalence, including Southeast Asia [3,6,7].
Escherichia coli, a Gram-negative human gut commensal and opportunistic extraintestinal pathogen, is an important vector for AMR [8,9]. As exemplified by pandemic multidrug-resistant (MDR) lineages such as ST131 [10,11], E. coli is capable of acquiring and maintaining multiple AMR determinants and exhibiting resistance to multiple classes of antibiotics. The presence of combinations of AMR genes can significantly impact therapeutic options. The limited treatment options for ESBL-resistant organisms make these a cause for concern. Colistin is a last-resort antibiotic included in the Reserve category of the World Health Organization (WHO) Essential Medicines List [12].
In E. coli and other Gram-negative bacteria, carriage of both colistin resistance genes and production of extended spectrum beta-lactamases (ESBLs) is concerning, as it suggests a potentially stable environment for the accumulation of further resistance, for example, carbapenemases, severely limiting treatment options [9,13,14]. ESBL resistance genes, including bla CTX-M variants, and mcr genes that confer colistin resistance, are commonly found in E. coli carried by returning travellers [3,6,7,15]. AMR genes can move intra-or inter-cellularly and accumulate at single sites in association with mobile genetic elements (MGEs) [16][17][18][19]. Plasmids are extrachromosomal genetic elements that can transfer horizontally between bacteria of the same or different species and are strongly associated with the spread of AMR [20,21]. In E. coli and other members of the Enterobacterales, AMR genes have been found in many different plasmid types [20]. Reports of stable, epidemic and internationally distributed plasmids [22][23][24] highlight the threat successful plasmid lineages pose and the importance of understanding the mechanisms by which they acquire and accumulate AMR genes.
We recently characterized the dynamics of the acquisition of MDR Gram-negative organisms in real time during travel to Vientiane, Laos. These MDR Gram-negative organisms had a surprisingly high co-prevalence of ESBLs and colistin resistance genes [7]. Here we explore the pattern of AMR carriage and context in E. coli from that cohort using long-read sequencing to understand the contexts of AMR genes and the role of MGEs, particularly plasmids, in the acquisition of drug-resistant E. coli by travellers to a region of high AMR prevalence. The 48 representative isolates for long-read sequencing in this study were selected for their AMR and plasmid replicon profiles and were collected on a daily basis in an area of high AMR prevalence, enabling continuous monitoring of the drug-resistant E. coli that colonized study participants. Continuous sampling facilitated the examination of common and circulating plasmids in the E. coli population, across a variety of sequence types (STs) and at different study time points [7]. Our data show the diversity and widespread distribution of numerous distinct AMR plasmids acquired by these travellers in Vientiane, Laos and the multiple different potential routes of AMR spread by plasmids and highlighted the complex nature of plasmids carrying both ESBL resistance and mcr genes.

Impact Statement
The global spread of antimicrobial resistance (AMR) is closely associated with international travel. AMR is a severe global concern and has compromised treatment options for many bacterial pathogens, among them pathogens carrying extendedspectrum beta-lactamase (ESBL) and colistin resistance genes. Colonizing multidrug-resistant (MDR) organisms have the potential to cause serious consequences. Infections caused by antimicrobial-resistant bacteria, including MDR bacteria, are associated with longer hospitalization, poorer patient outcomes, greater mortality and higher costs compared to infections with susceptible bacteria. This study elucidates the numerous different types of plasmids carrying AMR genes in colonizing ESBLproducing Escherichia coli isolates found in faecal samples from travellers to Vientiane, Laos. Here we add to known databases of AMR plasmids by adding these MDR plasmids found in Southeast Asia, an area of high AMR prevalence. We characterized novel AMR plasmids, including complex ESBL (bla CTX-M ) and colistin (mcr) resistance co-carriage plasmids, emphasizing the potential exposure of travellers to Laos to a wide variety of mobile genetic elements that may facilitate global AMR spread. This in-depth study has revealed further detail concerning the numerous factors that may influence AMR transfer, and therefore potential routes of AMR spread internationally, and is a step towards finding methods to combat AMR spread.

Study design and sample source
The E. coli isolates used here were collected as part of a study [7] looking at the dynamics of gut colonization of 21 volunteers attending a medical course in Vientiane, Laos. Faecal samples were taken daily during the 22 day period. Samples were processed, shipped, stored and handled as previously described by Kantele et al. [7]. Here ESBL-positive isolates were cultured from faecal samples after initial screening on CHROMagar ESBL agar plates (CHROMagar, Paris, France) at the Microbiology Laboratory of Mahosot Hospital, Vientiane, Laos and after transportation further screened with chromID ESBL chromogenic medium (bioMérieux) by the University of Helsinki, Helsinki, Finland [7]. These ESBL isolates [7] were used in this study. In order to explore the pattern of AMR in this traveller dataset and the role of MGEs we prepared hybrid assemblies and annotated plasmid sequences identified to locate resistance genes and potential routes for spread.

DNA extraction and sequencing
E. coli were cultured overnight on UTI chromogenic agar (Sigma) at 37 °C. After purity checks single colonies were subcultured overnight in lysogeny broth (LB) (Miller) (shaking, 37 °C). For the majority of isolates DNA was extracted using Monarch Genomic DNA Extraction kit (NEB), but in some instances a lower quantity and quality yield was obtained. In these instances, we observed atypical precipitates and opted for extraction using phenol/chloroform with cetyltrimethylammonium bromide (CTAB). The extracted DNA was sequenced over four runs of MinION (Oxford Nanopore Technology) using R9.4.1 flow cells. Three runs were prepared using Ligation Protocol (LSK-SQK109) and one run was prepared with the Rapid Barcoding Sequencing kit (SQK-RBK004), with both protocols modified to a one-pot implementation.

Identification of small plasmids from Illumina dataset
NCBI blast (v2.5.0+) was used to search draft genomes for plasmid-specific 100 bp signature sequences (Table S3). Signature sequences, as described previously [22], comprise 50 bp each of a mobile genetic element and its immediately adjacent sequences such that they span the specific junction generated by a given insertion or deletion event. This facilitated the detection of specific plasmid lineages, whether they were represented by single or multiple contigs in the Illumina dataset [7].

Isolates from Laos contain a broad diversity of plasmids and resistance genes
A total of 163 complete plasmids were obtained from the hybrid-assembled genomes of 48 E. coli isolates that colonized 21 participants and their contacts (Table S4). Isolates harboured one to eight plasmids each, except for a single ST1722 isolate that did not contain any plasmids. The 163 plasmids were typed by size, replicon type and AMR gene carriage, and identical or almost-identical plasmids (Table S3) were deduplicated, resulting in a total of 105 distinct plasmids. These 105 plasmids were named pLAO1-pLAO105 (Tables S1 and S4) and ranged in size from 1531 to 259 739 bp, with around half (53.3 %, n=56) smaller than 25 kb. Forty of the 105 plasmids (38.1 %) contained one or more AMR genes. Plasmids that did not carry AMR genes were generally smaller (mean size: 23 554 bp), than those that did (mean size: 97 260 bp). However, pLAO10 (GenBank accession OP242224), pLAO78 (OP242289) and pLAO84 (OP242239) are notable ColE1-like plasmids that carry one-five AMR genes each and range in size from just 5540 bp to 22 368 bp (Table S4).
We found evidence suggesting the circulation of plasmids within Lao E. coli in the traveller population. One example, the ColE1-like plasmid pLAO84 (GenBank accession: OP242239), which carries the tetracycline resistance determinant tet(A), was found in the complete genomes of E. coli of two different STs, ST195 and ST34, that were acquired by two different participants (Pt33 and Pt40). Mapping contigs assembled from Illumina data against the complete pLAO84 sequence confirmed plasmid circulation, with its complete or fragmented sequence detected in seven additional E. coli isolates (Fig. S1). These isolates were obtained from six different participants (including Pt33 and Pt40) over a 10 day period. It is not possible to determine where or when pLAO84 transferred between its hosts, or how the hosts were acquired by these participants. Another example of a potential circulating plasmid, the Q1 plasmid pLAO60 (GenBank accession OP242237) that carries aminoglycoside and beta-lactam resistance genes (Table S4), was present in the complete genome of the ST542 isolate (LA124) and in one genome (LA230) in the wider Illumina dataset (Table S4 and Fig. S2). The Q1 plasmid in isolate LA230 was missing a 101 bp segment that appears to have been lost in a homologous recombination event.
Another notable finding was the presence of phage and virulence plasmids in this collection. A Y-type phage-plasmid, pLAO59 (GenBank accession OP242238), appears to have lost key genes for phage body synthesis, potentially in deletion events mediated by ISKpn26 and IS1294, and instead carries genes that confer resistance to six antibiotic classes and genes that confer resistance to mercury, copper and silver (Fig. S3). The FII-18:FIB-1 pLAO32 is related to the colicin V (ColV) virulence-resistance plasmid pCERC3 [40] and contains virulence genes, including those for the aerobactin and Sit siderophore systems, in addition to multiple drug resistance genes (Fig. S3), but lacks the genes for ColV.

Variation within individual plasmid types, with diverse and complex resistance regions
In addition to the variety of different plasmid types, long-read sequence data allowed us to observe diversity amongst plasmids of the same type (Tables S4 and S5, Figs S1-S4). There was considerable genetic diversity in plasmids carrying the FII-2 replicon (Fig. 2). FII-2 plasmids pLAO44 (GenBank accession OP242233) and pLAO37 (OP242229) in ST69 and ST101 strains only contained FII-2 replicons, while FII-2 plasmid pLAO82 (OP242230) in an ST34 isolate carried an additional θ-RNA replicon and plasmids pLAO100 (ST40, OP242240) and pLAO103 (ST457, OP242243) carried an additional FIB-10 replicon (Figs 2 and 3). Diversity also occurred in the resistance regions of these plasmids ( Fig. 2 and Fig. S4). FII-2 plasmids carried multiple resistance genes, including various combinations of bla CTX-M-27 , bla CTX-M-55 and mcr-3.4 (Figs 2 and 3). Plasmid types FII-2 (pLAO37), FII-2:θ-RNA (pLAO82) and FII-46:FIB-like (pLAO69, OP242232) incorporated resistance regions with an accumulation of multiple resistance genes and co-carriage of mcr and bla CTX-M genes (Fig. 3b). FII-2:θ-RNA, pLAO82, demonstrated even more complexity with further accumulation of resistance genes as it also carried an additional mcr-3.4 located between copies of IS26 and ISKpn40 (Fig. 3). There is widespread global distribution of the resistance regions found in these FII-2 plasmids, which have been identified in multiple countries, some of which are not close to Laos (Table S6). We found no link between the ST of the isolate and the type of plasmid AMR genes carried by that ST (Fig. 3a).

Co-carriage of bla CTX-M and mcr genes occurs in multiple resistance region configurations
Three plasmids in this collection carried both bla CTX-M and mcr genes (Fig. 3b). All three bla CTX-M /mcr co-carriage plasmids were F-types, but they differed in size and plasmid replicon type. The 98 237 bp plasmid, pLAO37, only carried an FII-2 replicon and was found in five ST101 E. coli isolates. The 110 949 bp pLAO82 was a multi-replicon co-integrate plasmid carrying both FII-2 and θ-RNA replicons and was found in one ST34 E. coli isolate. The 105 425 bp plasmid, pLAO69, carried FII-46 and FIB-20 replicons, and was found in two ST10 E. coli isolates. All three of these plasmids have typical F-type plasmid backbones that include   genes for replication, stable maintenance, conjugative transfer and establishment in new hosts [40,41]. Each plasmid contained a complete and uninterrupted transfer region, suggesting that all three have the capacity for self-mediated conjugation [42].
Detailed comparison of the FII-2 plasmids pLAO37 and pLAO82 showed that the backbones are almost identical apart from a recombination patch (approximately 7 kb). pLAO37 and pLAO82 have the same FII-2 repA1 gene. The AMR genes in both pLAO37 and pLAO82 are located in a complex region bounded by IS26 at the left end and Tn1721 at the right end (Fig. 3b). These regions comprise sequences from multiple mobile genetic elements with distinct origins and include genes that confer resistance to beta-lactams (bla CTX-M ), colistin (mcr-3.4), aminoglycosides (aacC2d), chloramphenicol (catA1) and quinolones (qnrS1). An additional θ-RNA replicon in pLAO82 is part of a small plasmid that has been captured and incorporated into the resistance region. pLAO37 and pLAO82 contain an extra partial Tn21 and partial IS26 region inserted between the tmrB and mcr genes (Fig. 3b). The pLAO82 resistance region also includes an additional copy of mcr-3.4. The mcr-3.4 in pLAO37 was truncated by ISKpn26, which removed the terminal 32 bp of the gene. This configuration of ISKpn26 and mcr-3.4 is not present in any other sequence deposited in the GenBank non-redundant nucleotide database. A novel transposon, Tn7514, was present in both pLAO37 and pLAO82 but was inserted in two different backbone locations (Fig. 2b and Tables S7 and S8) [43].
Although largely syntenic, the nucleotide identity of the pLAO69 backbone differs significantly from that of pLAO37 and pLAO82, consistent with its distinct replicon type. The FII-46 repA1 gene of pLAO69 is only 94 % identical to FII-2 repA1 of pLAO37 and pLAO82. The pLAO69 transfer region also differed significantly, matching only 89 % of pLAO37 and pLAO82 transfer region with an overall identity of 97.2 % in a blastn comparison. pLAO69 carried mcr-3.1, which differs from the mcr-3.4 found in pLAO37 and pLAO82 (Fig. 3b). However, inspection of the resistance region in pLAO69 that contains mcr-3.1 revealed that it is largely composed of the same confluence of mobile elements found in the resistance regions of pLAO37 and pLAO82 (Fig. 3b). In addition to their identities, the configuration of these elements was the same in pLAO69 and pLAO37/pLAO82.

Insertion sequence type, abundance and diversity are associated with AMR gene carriage
The vast majority (n=36) of the 40 AMR plasmids contained one or more IS. Three exceptions were plasmids that contained AMR genes found in association with gene cassettes [bla VEB in pLAO60 and aadA2, ant(3'″)-Ia, cmlA1, dfrA12 in pLAO78], or miniature inverted-repeat tandem elements (MITEs) [tet(A) in pLAO84]. The remaining plasmid, pLAO41, contained mcr-1 but no detectable IS elements.
AMR plasmids carried almost three times more unique families of IS than their non-AMR counterparts (Fig. 4a), with an average of 6.75 unique IS families in AMR plasmids (range 1-11), vs 2.4 in non-AMR plasmids (range 1-8).
AMR plasmids contained an abundance of IS6/26 and IS1-family elements. Of the 36 AMR plasmids that did carry IS elements, all but one carried at least one IS6-family element (Fig. 4a), and the IS1 family was present in 83 % (n=30). IS3 and IS5 were also common, present in 75 % (n=27) and 78 % (n=28) of AMR plasmids, respectively. Differences in IS6 family element carriage were seen between AMR and non-AMR plasmids, with AMR plasmids containing a mean of five IS6 family elements per plasmid (range 0-9). In contrast, only 24 % (n=4) of non-AMR plasmids carried IS6 family elements, with only 1-2 IS6 elements per plasmid (Fig. 4a). IS6 family elements were present in all plasmids that co-carry bla CTX-M and mcr genes (Fig. 4b), where they were the most prevalent IS family. There was no clear association between IS families and co-carriage of the mcr and bla CTX-M genes (Fig. 4b). In plasmids where there was co-carriage of mcr and bla CTX-M genes, IS6 and IS3 were the most abundant IS families, with all co-carriage plasmids additionally carrying IS1380, IS1 and IS4. In each co-carriage plasmid, more than six different IS families were identified, and of the IS identified between 30-46 % were of the IS6 family (Fig. 4b), with IS26 the most prominent element.

DISCUSSION
In this study we conducted an in-depth investigation of the role of plasmids in the alarmingly high levels of AMR found in E. coli that colonized the gastrointestinal tract of travellers to Laos [7]. We have revealed an enormous diversity in the plasmids of this E. coli population, particularly in their resistance regions, many of which contained multiple AMR genes. Concerningly, the majority of the 40 AMR plasmids (n=30, 75 %) contained ESBL genes, a colistin resistance gene or both. Our data showed the abundance and importance of F-type, X-type, Q-type and ColE1-like plasmids as vectors for AMR gene spread in E. coli in Laos. AMR plasmids accounted for 38.1 % (n=40) of the 105 distinct plasmids and 17 of these 40 AMR plasmids identified were F-type (42.5 %), highlighting their importance as carriers of AMR genes (including ESBLs, mcr). F-type plasmids are known to carry AMR genes [44] and are an important factor in the high incidence of AMR carriage in this study. Although four types dominated, the AMR plasmid population identified in travellers to this region of Laos was extremely diverse, with 29 different plasmid types, including phage and virulence plasmids, found to carry AMR genes. Both AMR and non-AMR plasmids were identified throughout the study, including several from baseline faecal samples [7], where plasmids may have been acquired from travel to Laos or potentially from other travel.
A variety of less anticipated vehicles for AMR genes were identified in these Lao E. coli. Small ColE1-like θ-RNA plasmids are common in E. coli [21,45] and their high prevalence would be expected. These θ-RNA (n=26) were the most prevalent plasmids in the collection, followed by F-types (n=23) (Fig. 1), but in contrast to F-type plasmids, only a small number of θ-RNA plasmids (n=3) carried AMR genes. This is consistent with previous findings that ColE1-like plasmids occasionally carry resistance genes, including aminoglycoside and beta-lactam resistance determinants [46]. While relatively uncommon, the importance of small, high-copy-number plasmids for AMR should not be underestimated, as these can serve as platforms for the evolution of new resistance phenotypes [46,47]. The tet(A)-carrying ColE1-like plasmid pLAO84 clearly demonstrates the capacity these small plasmids have for disseminating resistance genes. pLAO84 was found in multiple E. coli STs in this collection, suggesting that it was circulating in the local E. coli population before being acquired by multiple study participants (Fig. S1). Additionally, the presence of a plasmid that was almost identical to pLAO84 in GenBank (GenBank accession CP057097.1) indicates that this ColE1-like plasmid lineage has already spread internationally, as it was present in Escherichia fergusonii isolated from pig faeces in the UK. A Y-type phage-plasmid (pLAO59, OP242238, File S8) highlighted an additional opportunity for AMR spread, since there is a chance for co-selection of this MDR plasmid due to carriage of metal resistance genes, e.g. for silver and copper. pLAO59 appeared not only to have lost key phage genes required for the lytic lifestyle [48], apparently as a result of deletions by insertion sequences, but carried multiple AMR genes and heavy metal resistance genes. Co-resistance and cross-resistance can cause co-selection of bacteria carrying metal resistance and AMR genes, with the metal resistance gene causing maintenance of the AMR gene [49,50]. In Vientiane, Laos it has been reported that environmental samples sourced near municipal solid waste landfill showed heavy contamination with heavy metals, including copper at levels higher than WHO permissible standards [51], indicating the possibility that co-selection by metal resistance genes is a real environmental pressure in the Vientiane area from which these isolates were collected. Silver and copper can co-select for various AMR genes in E. coli, including tetracycline and sulphonamide resistance genes, which were also found in pLAO59 [52,53].
Multiple combinations of AMR resistance genes were found in multiple genetic contexts associated with various mobile genetic elements. mcr genes, for example, were found next to multiple TEs (Fig. 3, Table S4) and mainly on F-type plasmids, consistent with the literature indicating that mcr-3.4 is associated with the FII-type [54,55]. ISEcp1 is known to contribute to the spread of bla CTX-M [56], but interestingly, only one complete ISEcp1 was identified (pLAO32) as part of a bla CTX-M-55containing transposition unit flanked by the 5 bp target site duplication TAACA. Most bla CTX-M genes in this collection were associated with complete copies of IS26 and partial ISEcp1 (Figs 2-4, Figs S3 and S4). IS26 from the IS6/26 family is known to play a key role in AMR gene dissemination [57][58][59][60]. IS26 carriage predisposes plasmids to insertion of additional IS26 and any associated AMR genes, which facilitates the accumulation of AMR genes at single sites, leading to multidrug resistance phenotypes [57,61]. This appears to be the case with pLAO37, pLAO82 and pLAO69 featuring IS26 and, based on what is known of IS26 behaviour [57], IS26 is likely to have played an important role in the assembly of their complex co-carriage resistance regions (Fig. 3).
Analysis of plasmid contigs from 49 E. coli hybrid assemblies has confirmed the vast and diverse genetic context of AMR in the Kantele et al. dataset isolated in Laos, and highlighted that in addition to multiple unique colonizing strains [7], there are multiple distinct plasmids present in this dataset. Long-read sequencing has been critical in highlighting the key role insertion sequence elements may play in the formation of this complex set of MDR plasmids that risk spreading AMR. This previously unreported cohort of MDR plasmids offers the alarming prospect that one of these will create a stable configuration for the creation of a successful pandemic MDR plasmid.