Exchange of Carbapenem-Resistant Escherichia coli Sequence Type 38 Intercontinentally and among Wild Bird, Human, and Environmental Niches

ABSTRACT

Carbapenem-resistant Enterobacteriaceae (CRE) are a global threat to human health and are increasingly being isolated from nonclinical settings. OXA-48-producing Escherichia coli sequence type 38 (ST38) is the most frequently reported CRE type in wild birds and has been detected in gulls or storks in North America, Europe, Asia, and Africa. The epidemiology and evolution of CRE in wildlife and human niches, however, remains unclear. We compared wild bird origin E. coli ST38 genome sequences generated by our research group and publicly available genomic data derived from other hosts and environments to (i) understand the frequency of intercontinental dispersal of E. coli ST38 clones isolated from wild birds, (ii) more thoroughly measure the genomic relatedness of carbapenem-resistant isolates from gulls sampled in Turkey and Alaska, USA, using long-read whole-genome sequencing and assess the spatial dissemination of this clone among different hosts, and (iii) determine whether ST38 isolates from humans, environmental water, and wild birds have different core or accessory genomes (e.g., antimicrobial resistance genes, virulence genes, plasmids) which might elucidate bacterial or gene exchange among niches. Our results suggest that E. coli ST38 strains, including those resistant to carbapenems, are exchanged between humans and wild birds, rather than separately maintained populations within each niche. Furthermore, despite close genetic similarity among OXA-48-producing E. coli ST38 clones from gulls in Alaska and Turkey, intercontinental dispersal of ST38 clones among wild birds is uncommon. Interventions to mitigate the dissemination of antimicrobial resistance throughout the environment (e.g., as exemplified by the acquisition of carbapenem resistance by birds) may be warranted.

IMPORTANCE Carbapenem-resistant bacteria are a threat to public health globally and have been found in the environment as well as the clinic. Some bacterial clones are associated with carbapenem resistance genes, such as Escherichia coli sequence type 38 (ST38) and the carbapenemase gene bla_OXA-48. This is the most frequently reported carbapenem-resistant clone in wild birds, though it was unclear if it circulated within wild bird populations or was exchanged among other niches. The results from this study suggest that E. coli ST38 strains, including those resistant to carbapenems, are frequently exchanged among wild birds, humans, and the environment. Carbapenem-resistant E. coli ST38 clones in wild birds are likely acquired from the local environment and do not constitute an independent dissemination pathway within wild bird populations. Management actions aimed at preventing the environmental dissemination and acquisition of antimicrobial resistance by wild birds may be warranted.

INTRODUCTION

It is important to assess the possible exchange of antimicrobial resistance (AMR) within and between human and environmental niches and how it might contribute to the evolution and spatial dispersal of clinically important bacteria such that interventions may be implemented to limit the further spread (1, 2). Carbapenem-resistant Enterobacteriaceae (CRE) are an urgent threat globally and have been isolated from clinical settings, as well as the commensal gut microbiota of healthy individuals or wildlife, on six continents (3–9). It remains unclear, however, whether the evolution and dissemination of CRE in wildlife and human niches is parallel (i.e., distinct clones circulating in each niche) or is characterized by the frequent exchange of a limited number of globally distributed epidemic clones (10, 11).

Carbapenem-resistant Escherichia coli sequence type 38 (ST38) is considered an emerging epidemic clone that has caused healthcare-associated outbreaks in multiple countries, yet it has also been reported from wild birds on at least four continents (4, 12). This sequence type is associated with the carbapenemase gene bla_OXA-48 (13–16), which was originally described in an IncL/M epidemic plasmid found in clinical Klebsiella pneumoniae isolates in Turkey (17). Chromosomal integration of bla_OXA-48 has since been identified or suspected in many cases (15, 18). No other carbapenem-resistant E. coli sequence type is as frequently reported and geographically widespread in wild birds as OXA-48-producing ST38 (12, 19).

We previously identified highly genetically similar OXA-48-producing E. coli ST38 isolates from gulls sampled in Alaska, USA, and Turkey using short-read whole-genome sequencing (12), though no obvious routes of dissemination between the gulls inhabiting these two distant locations were identified. Given the widespread distribution of OXA-48-producing E. coli ST38 in wild birds, including genomic evidence of closely related isolates sampled from different continents, we aimed to investigate whether strains of OXA-48-producing E. coli ST38 found in wild birds are distinct from strains isolated from humans or the environment or whether they reflect carbapenem-resistant clones circulating in spatial proximity in multiple niches. We compared E. coli ST38 genome sequences generated by our research group, as well as publicly available genomic data, to (i) understand the frequency of intercontinental dispersal of E. coli ST38 clones isolated from wild birds, (ii) more thoroughly measure the genomic relatedness of carbapenem-resistant isolates from gulls sampled in Turkey and Alaska, USA, using long-read whole-genome sequencing and assess the spatial dissemination of this clone among different niches, and (iii) determine whether ST38 isolates from different niches (i.e., humans, environmental water, wild birds) have different core or accessory genomes (e.g., AMR genes, virulence genes, plasmids) which might elucidate bacterial or gene exchange among niches.

RESULTS

E. coli ST38 from wild birds.

The EnteroBase database included whole-genome sequencing data from 49 E. coli ST38 isolates originating from wild birds sampled on five continents. Sixty-seven percent (33/49) of the isolates originated from gulls, the remainder being from ducks (n = 6), cranes (n = 3), storks (n = 2), and single isolates from kite, grackle, owl, pigeon, vulture, and unknown hosts. The isolates were grouped into two main clades that were not differentiated by continent (Fig. 1). Ten isolates, all from gulls, harbored bla_OXA-48, including five from Turkey, four from Alaska, USA, and one from Spain (Fig. 1). No other carbapenemase genes were identified among wild bird isolates. The isolate from Spain (2462b1) harbored an IncL plasmid that likely carried bla_OXA-48 (see Table S1 in the supplemental material) and was phylogenetically divergent from those from Alaska and Turkey. A total of 36 unique HC10 clusters were identified. Four isolates originating from gulls in Alaska and two isolates originating from gulls in Turkey belonged to the same HC10 cluster (20636) and the same HC5 cluster (121628). The three other isolates from Turkey belonged to different HC10 clusters. No other avian origin isolates sampled from different continents belonged to the same HC10 cluster.

FIG 1.

Maximum likelihood phylogenetic tree of all publicly available ST38 E. coli genomes originating from wild avian sources. Tips are colored according to the continent from which the isolates originated, and the matrix indicates the presence of antimicrobial resistance genes. Carbapenemase genes are shown in green. The dotted box includes isolates that belonged to the same HC10 cluster.

HC10 cluster 20636.

Within the EnteroBase database, 22 isolates belonged to the same HC10 cluster (20636) as the gull isolates from Alaska and Turkey. These isolates originated from North America, Europe, and Asia and were sampled from humans, gulls, and unknown hosts. All harbored bla_OXA-48, and none harbored IncL/M plasmids (Fig. 2). A total of 237 virulence genes were found, of which 189 were found in all 22 isolates (Table S2).

FIG 2.

Maximum likelihood phylogenetic tree of HC10 20636 isolates from EnteroBase. Tips are colored according to the continent, and isolates belonging to the HC5 121628 cluster are indicated with a gray icon depicting the host (human or gull). The matrix indicates the presence of antimicrobial resistance genes, with carbapenemase genes shown in green.

Three isolates from humans sampled in the United States and Canada belonged to the same HC5 cluster as the six gull isolates from Alaska and Turkey. The human samples were collected in the United States in 2017 and 2018 and Canada in 2019, whereas the gull fecal samples were collected in Turkey in 2015 and Alaska in 2016. Phylogenetic analysis based on a total of 28 single nucleotide polymorphisms (SNPs) suggested that the human and gull isolates from North America shared a more recent common ancestor than those from Turkey (Fig. 2). The gull isolates from Alaska differed from the human isolates by 9 to 14 SNPs and the gull isolates from Turkey by 13 to 15 SNPs. The complete genome sequence of isolate 15.TR.026_OXA from Turkey confirmed the chromosomal location of bla_OXA-48 as well as its synteny with A1_136_Kasilof from Alaska within the 19,867-bp bla_OXA-48 insertion region (Fig. S1).

Comparison of E. coli ST38 from different niches.

The presence of AMR genes, virulence factors, and plasmids was assessed in silico for 49 isolates each from wild birds, humans, and environmental water. A total of 57 AMR genes, 385 virulence factor genes, and 44 plasmid replicons were identified (Tables S1, S3, and S4). Discriminant analysis of principal components (DAPC) results suggested that niches could be differentiated based on the accessory genome content, with virulence factors being the most discriminatory (Fig. 3A to C). The proportion of isolates assigned to the correct group was 0.71 for AMR genes, 0.87 for virulence factor genes, and 0.65 for plasmid replicons. Phylogenetic analysis of all 147 E. coli ST38 isolates showed extensive genetic diversity with isolates from different niches distributed throughout the phylogenetic tree (Fig. 3D).

FIG 3.

Plots conveying the relationships among E. coli ST38 isolates from three niches: avian, human, and water. (A to C) Scatterplots of discriminant analysis of principle components based on accessory genome content of (A) AMR genes, (B) virulence factors, and (C) plasmid replicons. (D) Maximum likelihood phylogenetic tree of E. coli ST38 isolates.

DISCUSSION

The results of this investigation provide important molecular epidemiological insights into E. coli ST38 as it pertains to wild birds and other niches. For example, our results suggest that E. coli ST38 isolates from wild birds are diverse and geographically widespread. In addition, evidence for the intercontinental exchange of closely related (i.e., same HC10 cluster) E. coli ST38 clones from wild bird hosts is uncommon. Finally, carbapenem-resistant E. coli ST38 isolates from wild birds are not genetically distinct from those from humans or environmental water, though there is some evidence of differences in the occurrence of virulence genes, which might be explained by differential selection pressures or sampling biases inherent in our analyses.

Our initial query of EnteroBase yielded whole-genome sequencing data from 49 E. coli ST38 isolates originating from wild birds sampled from Africa, Asia, Europe, North America, and Oceania, though the only isolates originating from different continents that belonged to the same HC10 cluster were those from Alaska and Turkey. Gulls sampled in Turkey in 2015 harbored carbapenem-resistant E. coli ST38 isolates that were assigned to multiple HC10 clusters. Carbapenem resistance was presumably conferred by chromosomally encoded bla_OXA-48, as inferred from long-read sequencing of isolate 15.TR.026_OXA. These findings are consistent with the endemicity of chromosomally encoded bla_OXA-48 E. coli ST38 in that region since at least 2004 (20). The isolates from Alaska originated from gulls sampled in two locations several weeks apart, though there were no reported human clinical cases in the state (3). Isolates from gulls in Alaska and Turkey differed by as few as 13 SNPs, and the two complete genomes revealed identical OXA-48 insertion regions.

Very closely related bla_OXA-48-positive E. coli isolates were recovered from gulls sampled in Alaska and Turkey, though closely related isolates were also identified in samples collected from humans and unknown sources in North America, Europe, and Asia. Isolates obtained from gulls in Turkey in 2015 and humans in Canada in 2019 differed by only 13 to 14 SNPs, comparable to the number of genetic differences between clones from gulls in Alaska and Turkey. This finding is also consistent with mutation rates previously reported for E. coli (21). Collectively, these results indicate there is bacterial exchange between humans and wild birds, rather than parallel evolution/maintenance of bacterial clones within each niche. It is likely that the intercontinental dissemination of these closely related clones was facilitated by humans, given the prior widespread reports of carbapenem-resistant E. coli ST38 among various global populations (13), the close genetic similarity of clones from gulls sampled in Alaska and Turkey to human isolates, and the lack of migratory connectivity exhibited by gulls from these two regions (22). However, given the limited number of published E. coli ST38 genomes originating from wild birds and HC10 cluster 20636 at the time of analysis, additional transmission pathways may not have been captured.

Based upon further querying of EnteroBase data and subsequent analyses, we inferred that E. coli ST38 isolates from wild birds, humans, and environmental water are not segregated into distinct phylogenetic clades but rather are found throughout the phylogeny. This provides further evidence that E. coli ST38 strains, including carbapenem-susceptible strains, are exchanged among niches, as has been observed previously in other E. coli sequence types (23, 24). Accessory genome content indicates some clustering by niche but also considerable overlap among wild birds, humans, and environmental water. The DAPC provides some evidence for discrimination of isolates by niche, as inferred from the occurrence of virulence factors, and less so of AMR genes and plasmid replicons, raising the prospect that selection pressures exerted within niches could lead to differential evolutionary trajectories. The number of accessory genome targets included in the DAPC analysis likely influenced the discriminatory power, because the power increased as the number of genetic targets increased. It is important to consider that the data used in this analysis relied on publicly available data that are inherently biased. For example, the isolates derived from human sources were generally from clinical samples and may therefore have had a higher proportion of AMR and virulence genes. Additionally, since all publicly available wild bird isolates were analyzed, as opposed to a subset of randomly selected human and environmental water isolates, isolates derived from wild birds sampled at the same location and time point were included and more likely to be clonally related. There were also spatial biases, as surveillance, reporting, and sequencing of bacteria from different niches in different regions of the globe are not uniform.

Identifying specific transmission pathways of bacteria and AMR genes is challenging, compounded by multiple sources of contamination (e.g., production animals and wastewater), inter- and intraspecific horizontal gene transfer (e.g., multidrug resistance plasmids), and limitations in identifying directionality (e.g., zoonotic or zooanthroponotic transmission). Despite belonging to the same sequence type, extensive genetic diversity was observed among 147 E. coli ST38 isolates from three niches. Although we can infer that this sequence type is widespread in multiple environments and hosts, genotypes may exist that are adapted to, or circulating within, a specific niche. However, this was not observed in the one genotype further investigated in this study. Isolates belonging to HC10 cluster 20636 had mutations in fewer than 10 core genes yet were detected on three continents and within at least two different niches. These results highlight the complexity in AMR epidemiology and advocate for conservative inferences regarding source attribution and transmission pathways in the absence of extensive sampling.

In conclusion, our results suggest the exchange of E. coli ST38 strains among wild birds, humans, and the environment. Carbapenem-resistant E. coli ST38 clones in wild birds are likely acquired from the local environment and are not independently disseminated within wild bird populations. The intercontinental movement of humans is likely driving the observed spatial dissemination of bla_OXA-48-positive E. coli ST38 clones among wild birds.

MATERIALS AND METHODS

Long-read sequencing.

Long-read sequencing was performed on E. coli ST38 isolate 15.TR.026_OXA (previously referred to as E3ec26oxaOxa), which was cultured from a yellow-legged gull (Larus michahellis) fecal sample collected in Turkey in January 2015. This isolate had previously been sequenced using short-read Illumina technology and was found to be closely related to isolates cultured from gulls in Alaska (12). DNA was extracted using the MagNA Pure compact nucleic acid isolation kit (Roche, Stockholm, Sweden), libraries were prepared using the SMRTbell template preparation kit according to the manufacturer’s instructions (Pacific Biosciences, Menlo Park, USA), and sequencing was performed using a PacBio RS II platform at the Uppsala Genomics Center. A hybrid assembly of long and short reads was performed by first polishing the long PacBio reads with the 30% shortest reads in the data set using the CLC Bio PacBio de novo assembly pipeline (Qiagen), followed by another round of polishing using the Illumina short reads. The genome was then de novo assembled using both the PacBio and Illumina reads with the CLC Genomics Workbench v11. The genome was annotated using the NCBI Prokaryotic Genome Annotation Pipeline v4.6 (25). This completed genome was compared to the completed genome of A1_136_Kasilof (GenBank accession number CP040390), which was previously sequenced using both long- and short-read sequencing (3), to assess synteny within the OXA-48 insertion region.

Evaluation of intercontinental dissemination of clones.

All E. coli ST38 isolates of wild avian origin were retrieved from EnteroBase (https://EnteroBase.warwick.ac.uk/species/ecoli/search_strains) on 12 September 2022 using the search terms “wild animal” under “Source Niche” and “avian” under “Source Type.” Within EnteroBase, all avian origin sequences were mapped to the completed A1_136_Kasilof genome from Alaska (GenBank accession number CP040390), and a maximum likelihood phylogenetic tree was created based on single nucleotide polymorphism (SNP) profiles (26). Furthermore, assemblies were downloaded from EnteroBase to assess the presence of AMR genes, virulence genes, and plasmid replicon types. These were identified using ABRicate v1.0.1 (Seemann) and the NCBI (27), E. coli virulence factor, and PlasmidFinder (28) databases, respectively.

Within EnteroBase, E. coli isolates are assigned to hierarchical clusters (HCs) based on their core genome multilocus sequence type (MLST) assignment (29). Isolates belonging to the same HC10 cluster are defined as having different alleles in no more than 10 genes out of a total of 2,513 core genes interrogated. Similarly, isolates belonging to the same HC5 cluster differ by no more than 5 genes. We used a threshold of 10 core gene differences to identify closely related isolates based on previous studies investigating E. coli and Salmonella outbreaks in humans (30, 31). EnteroBase was then queried for all isolates belonging to identified HC10 clusters to identify other niches or hosts harboring that HC10 clone. A maximum likelihood phylogenetic tree based on SNP differences among isolates within the HC10 clone was generated within EnteroBase, as described above.

Accessory genome comparison among niches.

An equivalent number of randomly selected E. coli ST38 sequences, compared to the wild bird isolates investigated, were retrieved from EnteroBase from isolates originating from human and environmental water samples using the terms “human” and “environment” for “Source Niche” and “human” and “water/river” for “Source Type,” respectively. A maximum likelihood phylogenetic tree based on the SNPs identified in all isolates was created within EnteroBase (26). AMR genes, virulence factors, and plasmid replicon types were identified in each genome using ABRicate as described above for wild avian origin isolates. We compared whether presence of AMR genes, virulence factors, or plasmid content overlapped among different niches using discriminant analysis of principal components (DAPC) (32), implementing the function dapc in the R software package adegenet (33) and retaining the principal components to maximize predictive success and minimize the root mean square error.

Data availability.

The complete genome sequence of 15.TR.026_OXA has been deposited at GenBank under accession number CP032145.