Genetic traits of IncK2 plasmids and the Escherichia coli host underlying the association to the chicken gut

TL;DR

This study explores how IncK2 plasmids and Escherichia coli bacteria adapt to the chicken gut environment.

Contribution

The study identifies genetic traits of IncK2 plasmids and E. coli hosts that are specifically adapted to the chicken gut.

Findings

IncK2 plasmids show genetic conservation across different sources and countries.

Genome-wide association study identified genes linked to chicken gut adaptation.

Predicted gene functions suggest selective advantages in the chicken gut environment.

Abstract

This manuscript presents the phylogeny and conservation of IncK2 plasmids in Europe. It also provides insights into genetic traits responsible for IncK2 plasmids and its Escherichia coli host adaptation to the chicken gut. Fifty-eight E. coli isolates from nine European countries were sequenced using Illumina and Nanopore technology. Genetic analyses were performed to determine the relatedness of IncK2 plasmids and their E. coli hosts from poultry (80% of the total) and other sources. To analyse genetic traits associated with E. coli and IncK2 plasmid from chicken origin, a genome-wide association study (GWAS) was performed. The phylogenetic analysis of IncK2 plasmids revealed conservation across sources and countries of isolation. GWAS revealed multiple genes associated with IncK2 plasmids or its E. coli host from chicken origin. The predicted functions of these genes can indicate a selective advantage in the chicken gut, and that IncK2 plasmids are adapted to the physiological environment of the chicken gut.

Figures

An overview of the number of isolates included in the study.

An MST of the cgMLST analysis of 57 E. coli isolates. Nodes are colour-coded according to sources. The cluster definition was set at ten allelic differences.

A core genome SNP phylogenetic tree of 56 E. coli isolates. The tree was built based on 58,553 SNPs. Branches are colour-coded according to isolation country and samples are colour-coded based on the source. Samples 77-2014-01-7010 and VGOL0538 were omitted from the analysis by the parsnp due to issues with the sequence.

Phylogenetic trees of IncK2 plasmids. Strain names are highlighted according to sources, while coloured lines represent the country of isolation. Poultry isolates contain chicken and turkey isolates. (a) Gene presence–absence tree. (b) Core plasmid genome SNP tree. The tree was built based on 2,995 SNPs.

A Jaccard index comparison between plasmids originating from human (human–human), chicken (chicken–chicken) and between these two sources (human–chicken).

Alignment of the IncK2 plasmid sequences (representatives from the three clusters in Fig. 4b) demonstrating its plasticity.

Alignment of the hok/sok reference sequence (top sequence) to the corresponding region of a pig IncK2 plasmid PE132-134 (second sequence), a chicken IncK2 plasmid ST4CTX (third sequence) and a human IncK2 plasmid RIVM_C037143. Three plasmids from this study are representatives from the three main clusters in Fig. 4(b). The red arrow represents an ORF significant in the GWAS.

Tables

Table 1.. Results of the GWAS for E. coli isolatesThe table depicts genes that were significantly associated with chicken.

Gene name	P-value	Function
gnu	0.0154	Epimerase
hsdR	0.0197	Type I site-specific deoxyribonuclease
rafY	0.0199	Sucrose porin precursor
radC	0.0199	DNA repair protein
trkG	0.0202	Cation transport
Hypothetical gene	0.0279	Phytase family protein
Hypothetical gene	0.0279	WYL domain-containing protein
Hypothetical gene	0.0279	Chemotaxis protein
Phage protein	0.0379	Unknown
intS	0.0436	Prophage integrase
Hypothetical gene	0.0436	Cobalamin adenosyltransferase
Hypothetical gene	0.0489	c-type lysozyme inhibitor
tnpA	0.0489	Transposase
papH	0.0489	Pilus assembly
papC/fimD	0.0489	Outer membrane usher protein
papD	0.0489	Pilus chaperone
Hypothetical gene	0.0489	Autotransporter outer membrane beta-barrel domain-containing protein
Hypothetical gene	0.0489	Porin family protein

Table 2.. Results of the GWAS for IncK2 plasmidsThe table depicts two genes that were significantly associated with poultry and one gene positively associated with human.

Gene name	P-value	Function	Positive association with
yffA	0.00518	Unknown	Human
ydeA	0.0305	l-Arabinose export	Chicken
Hypothetical gene	0.0451	Unknown	Chicken

Full text

Data Summary

The authors confirm that all supporting data, codes and protocols have been provided within the article or through supplementary data files. Sequencing data were submitted to the European Nucleotide Archive (ENA) with the project accession no. PRJEB73354.

Introduction

Antimicrobial resistance (AMR) is a major public health concern globally. The European Centre for Disease Prevention and Control (ECDC) reported that more than 670,000 infections each year are due to bacteria resistant to antimicrobials and ~33,000 people die as a direct consequence of these infections [1]. Acquired resistance among Enterobacterales is of special concern, particularly extended-spectrum cephalosporin resistance, since the related AMR genes are primarily carried on plasmids with the ability to spread horizontally. Currently, plasmid classification is foremost done by PCR-based replicon typing or sequence-based methods: replicon typing with, for example, PlasmidFinder [23]. Another typing method is based on conjugative relaxases [4]. Recently, Plasmer, a tool based on machine learning of shared k-mers and genomic features, was also introduced [5]. However, all these tools have insufficient discriminatory power for understanding the complex epidemiology and spread of plasmid-mediated AMR. Due to the high genetic plasticity of resistance plasmids, complete sequence analysis and an accurate plasmid taxonomy are needed to decipher plasmid dissemination routes across different ecological niches.

IncK2 plasmids belong to the I-complex, which includes IncB/O, IncI1, IncI2, IncI1ɣ, IncK1 and IncZ plasmids, widely present all over Europe [6]. IncK2 plasmids are associated with poultry, and IncK1 plasmids are found in a variety of sources [7]. The plasmidic AmpC (pAmpC) β-lactamase gene blaCMY-2 has frequently been reported in chickens and is predominantly associated with IncK2 plasmids [811]. Highly conserved IncK2 plasmids associated with Escherichia coli sequence type (ST) 38 were also identified in a longitudinal study of the Norwegian chicken production chain from 2011 to 2016 [12]. Moreover, IncK2 plasmids have been described to persist for up to 6 months on chicken farms [12]. The frequent occurrence in the European chicken population of conserved IncK2 plasmids carrying blaCMY-2 is at least partially explained by vertical transmission from the shared breeding stock [1314]. However, plasmids can also adapt to their bacterial hosts as well as their ecological niches. Previously, it has been postulated that IncK2 plasmids are often found in E. coli originating from chicken because of their elevated body temperature (41–42 °C) compared to mammals, leading to a higher (compared to IncK1) plasmid conjugation rate as well as levels of sigma-32 [1516]. The higher conjugation rate has been linked to the assembly of type IV secretion systems of IncFII plasmid, which leads to extra-cytoplasmic stress that is sensed by the two-component regulation system CpxAR [17]. A cytoplasmic stress leads to the activation of heat-shock genes, and this increases levels of alternative sigma-32, a transcriptional regulator that affects protein expression. Interestingly, CpxAR was also shown to benefit bacterial colonization, which further increases adaptation [18].

Despite the putative association between IncK2 plasmids and chickens due to a higher body temperature, little is known about the genetic determinants of this phenomenon. In addition, there are knowledge gaps regarding the involvement of the E. coli host in this adaptation. The research described in the current article highlights certain genes that drive the strong relationship between the IncK2 plasmid and its E. coli host present in the chicken reservoir. Moreover, this is the first study comprehensively describing the variation of IncK2 plasmids across European countries, Belgium, Denmark, France, Germany, Italy, the Netherlands, Norway, Poland, Portugal, Sweden and the UK, as well as in a non-European country, namely, Lebanon.

Methods

Strain selection

Previously, whole-genome sequenced Enterobacterales isolates included in this study were obtained from various sampling contexts performed by the different institutes. The main criterion for inclusion was the presence of an IncB/O/K/Z plasmid. However, the collections of sequenced isolates from the partner institutes were biassed towards ESBL/pAmpC encoding isolates, with a large proportion of isolates harbouring blaCMY-2 or other genes encoding extended cephalosporinase.

The collection contained 398 isolates, of which 137 harboured IncK2 plasmid (34.4%). Most of the strains containing IncK2 originated from poultry (80.3%). All isolates were identified as E. coli, except for one Klebsiella pneumoniae. The available draft genomes from Illumina sequences from selected isolates were mapped against a reference IncK2 plasmid, which had been isolated from chicken cloaca and is 79,297 bp in size (NZ_KR905386.1) using blastn to detect the presence of the IncK2 reference plasmid genome. A phylogenetic tree based on gene presence–absence of all IncK2 plasmids was constructed with Roary v3.13.0 and visualized with iTOL v6.8.1 [1920] (data not shown). Based on this analysis, a set of isolates was selected for long-read sequencing with the aim of including two isolates per country carrying IncK2 plasmids originating from chicken. These were representatives from each phylogenetic IncK2 cluster, as well as from all other sources, including cattle, humans, pigs, turkeys and the environment.

Sequence data and whole-genome bioinformatic analyses

Long-read sequencing was performed using Oxford Nanopore Technology (ONT), with Flowcell versions R9 and the basecalling performed using the high- or super-accurate algorithm (Table S1, available in the online Supplementary Material).

The generated long- and short-read sequence data were assembled using the FullForce Plasmid Assembler (FFPA) based on the hybrid assembly method in Unicycler v0.4.7–0 [21]. If the E. coli genome was not closed, genomic contigs were predicted with RFPlasmid [22]. Isolates with the genomic contigs predicted with RFPlasmid have ‘_rfplasmid’ added to their names (Table S1). If the plasmid sequence was not closed after assembly with FFPA, another assembly was performed with Flye v2.9.1, followed by polishing of the sequence with short reads using BWA-MEM and Pilon [2325]. Isolates assembled with Flye and subsequent polishing have ‘_pilon’ added to their names. After both assembly rounds, only isolates with closed plasmids were further analysed (Table S1).

The annotation of the whole genome of the isolate was done with Prokka v1.14.6 [26]. A gene presence–absence phylogenetic tree was created with Roary v3.13.0 and visualized with iTOL v6.8.1 [1920]. The species confirmation, as well as the E. coli ST, was determined using MLST (Achtman) v2.11 [27]. The E. coli phylotype was deduced using EzClermont v0.6.3 [28]. Core genome MLST (cgMLST) analysis and an MST were obtained using SeqSphere+ (Ridom) v9.0.8 (2023/06). In the cgMLST scheme, a total of 2,519 targets were included and 2,513 were used for calculations. The E. coli serotype was also determined with this software using the available E. coli CGE SerotypeFinder task template [29]. Potential avian pathogenic E. coli (APEC) phenotypes were determined using APECtyper [30].

Plasmid bioinformatic analyses

A core genome alignment of IncK2 plasmids was performed with blastn and visualized with Circos, the core genome being defined as a set of genes present in all of the studied plasmids [31]. Mobility and host range of IncK2 plasmids were predicted with MOB-suite [32]. AMR genes were annotated with Abricate v0.93 using the ResFinder database v4.0 [3334]. An SNP-based phylogenetic tree and SNP analysis were performed with parsnp v1.2 (with minimal cluster size set to 200 and vcf option for SNP analysis with the S4CTX plasmid as a reference) [35]. A figure combining both phylogenetic trees was made with R package cophylo. The Jaccard index for IncK2 plasmids was calculated with Mash v2.3 [36]. A genome-wide association study (GWAS) was performed with pyseer v1.3.10, using clusters of orthologous gene presence–absence [37]. All closed plasmids typed as IncK2 were downloaded from PLASDB (accessed 07 March 2023) and added to the analysis [38]. A full list of these additional sequences is included in Table S2. Samples of chicken origin were compared to those of other sources. Additionally, a GWAS of human IncK2 plasmids compared to the remaining plasmids was performed.

All short- and long-read whole-genome sequences were submitted to the public repository ENA (PRJEB73354). The accession numbers can be found in Table S1.

Statistical methods

The Jaccard index estimation was calculated, based on two k-mer sets per plasmid pair, with Mash using MinHash algorithm. Analysis of the significance of plasmid similarity, represented as the Jaccard index, was performed with the Kruskal–Wallis statistical test followed by Dunn’s test (GraphPad Prism v10.2.2) to determine exactly which groups are different (P<0.05 as the significance level).

Statistical significance of GWAS results was calculated with fixed-effect analysis. Logistic regression was performed on each k-mer for binary phenotypes using pyseer v1.3.10 with P<0.05 as the significance level.

Results

Genome analysis of E. coli isolates

Out of 62 selected E. coli isolates, 58 met the set criteria and were included for further genetic analyses (Fig. 1).

An overview of the number of isolates included in the study.

The most prevalent E. coli phylotype was B2 (n=14, 24.1%), followed by A (n=12, 20.7%), D (n=10, 17.2%), B1 (n=8, 13.8%), F (n=5, 8.6%), C (n=3, 5.2%) and E and G (n=2, 3.4%) each. Two isolates were typed as U/cryptic (3.4%) (Table S1). Overall, the E. coli isolates belonged to 33 different STs with ST69 (n=7, 12.1%), ST429 (n=5, 8.6%), ST359 (n=4, 6.9%) and ST373 (n=3, 5.2%) being the most common. Determining the E. coli serotype enabled further subtyping of the isolates to help identify serotypes associated with pathogenicity. For example, among the seven ST69 E. coli isolates, five serotypes were detected, i.e. O17/O44/O77:H18 (n=2), O23:H6 (n=2), O15:H1, O15:H18 and O17/O77:H18. Similarly, the four ST359 isolates could be subtyped in several serotypes, i.e. O7:H4, O100:H25, O115:H21 and O131:H12, while in contrast, all five ST429 belonged to O2:H1. In total, 43 different serotypes were detected among the 58 E. coli investigated with O2:H1 being the most prevalent with 6 isolates (Table S1). Next, a cgMLST analysis was performed to further assess the genomic similarity of E. coli isolates. One E. coli isolate had to be excluded from this analysis and the overall study because it did not meet the criterium of more than 90% allelic targets being present. A minimum spanning tree (MST) of cgMLST types revealed wide diversity overall (Fig. 2). The cluster definition was set at ten allelic differences. Cluster 1 contained three isolates originating from chickens in Sweden (CK67ctx-18_pilon), Norway (1-2016-40-13912_pilon) and Italy (18029473), and one human isolate from Denmark (ESBL20170103_rfplasmid). Moreover, all isolates in cluster 1 contained an almost identical IncK2 plasmid (100% identity over 99% of the sequence). Cluster 2 consisted of two human isolates (63HR1-5 and 65GR1-2) that shared 100% identity over 99% of the sequence and contained almost identical plasmids (100% identity over 99% of the sequence). Cluster 3 contained two identical chicken E. coli isolates 18028027 and 18053823 (99.99% identity over 99% of the sequence) with identical IncK2 plasmids. A core genome SNP phylogenetic tree confirmed that the isolates within clusters are nearly identical (Fig. 3).

An MST of the cgMLST analysis of 57 E. coli isolates. Nodes are colour-coded according to sources. The cluster definition was set at ten allelic differences.

A core genome SNP phylogenetic tree of 56 E. coli isolates. The tree was built based on 58,553 SNPs. Branches are colour-coded according to isolation country and samples are colour-coded based on the source. Samples 77-2014-01-7010 and VGOL0538 were omitted from the analysis by the parsnp due to issues with the sequence.

IncK2 plasmid analysis

Out of the 50 closed plasmids (Table 1), 29 (58%) originated from chickens, 16 (32%) from humans (which included the plasmid from the single K. pneumoniae isolate in the study), two (4%) from pigs and one (2%) from cattle, the environment and turkey, respectively. IncK2 plasmids were on average 94.1 kb in size (ranged from 73.7 to 150 kb). All of the plasmids are expected to be conjugative and their predicted host range spans the family of Enterobacteriaceae. The pAmpC gene blaCMY-2 was encoded on 40 out of 50 (80%) plasmids and was surrounded by widely spread on different plasmid types ISEcp1, blc (lipocalin) and sugE (quaternary ammonium compound resistance). In nine IncK2 plasmids, of which 18% contained blaCMY-2, the resistance genes sul1, aac(3)-VIa and ant(3″)-Ia were associated with a class 1 integron (Table S1). The ESBL gene blaSHV-12 was identified on three plasmids (6%) and was flanked by deoR (DNA-binding transcriptional repressor) and two copies of IS26. The β-lactamase determinant blaTEM-1B was found on three plasmids (6%) associated with Tn1331 (Tn2-like element), while one plasmid carried blaTEM-1C (2%) as part of Tn3 (Table S1). Of the 50 plasmids, four did not contain any known resistance genes.

Core genome and phylogenetic tree of IncK2 plasmids

The core plasmid genome of IncK2 is 4.4 kb and includes transfer and pilus formation genes, DNA primase (dnaG) and recombinase genes (rci), excA, yggA, ygcA and yhgA, as well as several hypothetical genes (Table S3). A dendrogram of IncK2 plasmids based on both gene presence/absence (Fig. 4a) and a phylogenetic tree based on core genome SNPs (Fig. 4b) was constructed. The first of the phylogenetic trees illustrates the plasticity and gene gain/loss, whereas the second dendrogram depicts the conservation of the plasmid core genome. For the gene presence/absence and core genome SNP tree, the relationship between the isolates remained the same – human and poultry isolates clustered closer within the source than between sources. A Jaccard index analysis showed that IncK2 plasmids from human E. coli were statistically more conserved compared to IncK2 plasmids from chickens to each other (P<0.0001) (Fig. 5). Moreover, IncK2 plasmids from humans showed a higher degree of similarity to each other (P=0.0015) than those from chickens.

Phylogenetic trees of IncK2 plasmids. Strain names are highlighted according to sources, while coloured lines represent the country of isolation. Poultry isolates contain chicken and turkey isolates. (a) Gene presence–absence tree. (b) Core plasmid genome SNP tree. The tree was built based on 2,995 SNPs.

A Jaccard index comparison between plasmids originating from human (human–human), chicken (chicken–chicken) and between these two sources (human–chicken).

Fig. 3 also shows the presence of highly similar IncK2 plasmids among chicken isolates (n=3, CK67ctx-18_pilon, 1-2016-40-13912_pilon and 18029473) and one human isolate (ESBL20170103_rfplasmid) belonging to cgMLST cluster 1 (Fig. 2). The number of SNPs between these plasmids ranged between zero and three. These isolates (together with seven additional isolates) carried an alternative partitioning system stbABC, RelE/ParE toxin–antitoxin system, tetracycline resistance and mercury tolerance genes. The number of average core plasmidome SNPs among human and chicken isolates was very similar (156 vs 163). No clustering of IncK2 plasmids was detectable based on the country of origin. In contrast, the IncK2 plasmids carried by E. coli isolated from chicken differed more from each other, for example, by the presence of different resistance genes (Table S1).

The core plasmid genome of IncK2 consisted of 42 genes, which included transfer and pilus formation genes, DNA primase (dnaG) and recombinase genes (rci), excA, yggA, ygcA and yhgA, as well as several hypothetical genes.

The alignment of the whole plasmid sequences (Fig. 6) showed a different order of genes between different isolates. That means that either core genes were arranged differently on IncK2 plasmids or different accessory genes were interspersed between core genes. These results suggest a certain plasticity of IncK2 plasmids.

Alignment of the IncK2 plasmid sequences (representatives from the three clusters in Fig. 4b) demonstrating its plasticity.

The core plasmid genome SNP analysis showed that all IncK2 differed between 0 and 1,275 SNPs (average of 170 SNPs). Plasmids with no SNP difference also had highly similar structures (Fig. S2).

Genome-wide association study

E. coli isolates

In order to assess the association between E. coli genomes and the chicken reservoir, a GWAS was performed. The E. coli genomes of chicken origin were compared with those obtained from other sources. A list of chromosomal genes that were significantly associated with the chicken source is presented in Table 1.

One of the genes significantly associated with chicken was rafY, which is a glycophorin. This determinant is part of a raffinose (trisaccharide composed of galactose, glucose and fructose) utilization operon (sugar metabolism) consisting of raffinose permease, an invertase and a galactosidase encoded by rafB, rafD and rafA, respectively [39]. Interestingly, out of all isolates carrying rafY (45), only 23 contained a complete rafABD operon. A full operon was carried by 16 E. coli isolates (69.6%) from chickens, 5 (21.7%) from humans and 2 (8.7%) from cattle.

The hsdR gene, which encodes for a type I site-specific deoxyribonuclease, is part of a gene cluster consisting of hsdRMS [40]. Only 17 out of 25 isolates carrying hsdR had a full hsdRMS cluster. Multiple variants of hsdR, hsdM and hsdS were present among the analysed E. coli chromosomes.

The papC, papD and papH genes are part of the pap operon encoding elements of P pili and have been described to be associated with pyelonephritis and cystitis in humans [41]. There was a large variation in the presence of the pap operon components within the analysed E. coli set. The pap gene variants that were significantly associated with the chicken source were present in 12 out of a total of 58 isolates. Four out of 12 of these isolates are APEC E. coli.

IncK2 plasmids

Similar to the analysis performed with E. coli genomes, GWAS was also performed with the IncK2 plasmids. Plasmids of chicken origin were compared to the ones obtained from other sources. Additionally, a GWAS of human IncK2 plasmids compared to the remaining (non-chicken) ones was performed. To increase the statistical power, publicly available sequences of IncK2 plasmids were also included in this analysis (Table S2).

Two proteins were significantly associated with chicken IncK2 plasmids in the dataset used (Table 2).

The first gene with a positive association was ydeA (most similar to UniProt protein A0A0H3ETB0), which encodes a l-arabinose export protein. The protein structure of YdeA closely resembles MFS proteins, such as the chloramphenicol exclusion protein CmlR [42]. The second significant hit was a hypothetical gene adjacent to the hok/sok toxin–antitoxin system. The protein translated from this gene is identical to the NCBI Reference Sequence WP_001546474 initially annotated as ‘antitoxin’, later changed to ‘hypothetical protein’. Fig. 7 depicts the alignment of the hok/sok system (middle sequence) with chicken (plasmid 3818) and cattle (plasmid 3460) IncK2 plasmids. The red arrow depicts the hypothetical gene, which shares 60 nucleotide identity with the hok/sok system.

Alignment of the hok/sok reference sequence (top sequence) to the corresponding region of a pig IncK2 plasmid PE132-134 (second sequence), a chicken IncK2 plasmid ST4CTX (third sequence) and a human IncK2 plasmid RIVM_C037143. Three plasmids from this study are representatives from the three main clusters in Fig. 4(b). The red arrow represents an ORF significant in the GWAS.

In our dataset, the only gene positively associated with human IncK2 plasmids was yffA which has an unknown function.

Discussion

Sequence analysis of E. coli carrying an IncB/O/K/Z plasmid in this study revealed a high variation of cgMLST types present. However, one cluster contained ST429 and ST9298 isolates from humans (n=1) and chickens (n=3) isolated in multiple countries. In addition, these isolates carried highly similar IncK2 plasmids. These results indicate the existence of a successful E. coli clone carrying an IncK2 plasmid that has spread over Europe and between humans and chickens. Moreover, another cgMLST cluster of E. coli carrying two almost identical IncK2 plasmids from two different persons sampled almost 1 month apart from the same poultry farm was found in this study. Additionally, one more cluster of highly genetically similar E. coli isolates carrying identical IncK2 plasmids was isolated in Italy. Although detailed information on the farm of origin related to these meat samples collected in the frame of the EU harmonized AMR monitoring programme was not available, the carcasses came from the same slaughterhouse owned by a company that raises and slaughters most of the chickens in Italy. The existence of successful bacterial clones and highly related plasmids from different sources is always concerning. However, it should be highlighted that the usage of cephalosporin in poultry is banned within the EU, and there is an overall decreasing trend of extended-spectrum cephalosporin resistance in poultry [4344].

The phylogenetic analysis of IncK2 plasmids in this study also revealed conservation across sources and countries of isolation. Overall, based on the data included in this study, IncK2 plasmids from humans were more similar to each other than the ones from chickens. Multiple factors affect plasmid evolution in the gut environment. One plausible cause of higher variation in chickens is their high body temperature, which has been shown to induce more mutations and recombination compared to the same strain cultured at 37 °C [45]. Moreover, environmental temperature will likely affect the conjugation rate of the plasmid [15], which can lead to changed fitness cost and adaptive changes in the plasmid sequence. In addition, IncK2 plasmids, included in this study, showed a degree of core plasticity. Plasticity of plasmid pINV, resulting in loss of the PAI, was shown to increase the growth rate of Shigella flexneri [46]. More research needs to be conducted in order to understand the role of plasmid plasticity in IncK2 plasmid’s evolution and host adaptation. The resident gut microbial community was found to limit the evolution and spread of AMR plasmids both in the human and chicken gut [4748]. Antibiotic treatment history will also affect the evolution of plasmids in the digestive tract [49]. Moreover, the genetic composition of the animal/human source, bacterial host or even other mobile elements present in the cell will affect the comportment of the AMR plasmid [5052]. It should also be noted that the number of IncK2 plasmids from humans present in the dataset was much lower compared to the ones from chickens, which will affect the assessment of overall plasmid variability. However, considering all possibilities that influence plasmid’s variability, it is very difficult to pinpoint the driving factor to explain why chicken IncK2 plasmids are more diverse compared to the ones that originate from humans.

GWASs are powerful tools to study links between genotype and phenotype and were previously successfully applied to investigate AMR determinants and virulence genes in E. coli and K. pneumoniae [5354]. GWAS has also been used to study bacterial-source adaptation and showed the importance of vitamin B5 biosynthesis (panBCD) in adaptation to the cattle gut [55]. Another study examined host factors affecting conjugation in E. coli and concluded that four genes fliF, fliK, kefB and ucpA were important for conjugation [39]. One of the limitations of GWAS is relying on assemblies that correctly assign genes to the chromosome and the appropriate plasmid type. Although using a hybrid assembly greatly reduces the risk of genes being misassigned, such a possibility cannot be completely excluded. The obtained phylogenetic results are influenced by the relatively small sample size and the geographic region being limited almost exclusively to Europe. Adding more samples isolated in more distant locations will improve the understanding of the epidemiology and evolution of IncK2 plasmids.

GWAS was performed on E. coli isolates to study the genes associated with the chicken source. The genes associated with chicken were three pap genes: papC, papD and papH. Pap proteins are responsible for the assembly of P pili, and the papC gene found in this dataset is linked with the virulence of APEC [56]. Another protein linked to the virulence of APEC that was significant in the GWAS was a c-type lysozyme inhibitor, which is responsible for the serum resistance of E. coli in the chicken environment [57].

Another gene significantly associated with the chicken E. coli isolates was rafY. The RafY protein is a porin responsible for the uptake of maltose, raffinose and sucrose, which allows a significantly higher growth rate of E. coli grown in a media containing these oligosaccharides [58]. Soybean is the predominant protein source for chickens worldwide, and the main components of its oligosaccharides are raffinose and stachyose [5960]. The fact that the chicken diet is rich in raffinose and chicken E. coli carry a gene that allows the uptake of this oligosaccharide suggests adaptation of these bacteria to its source or a selection advantage caused by feed. Another similar example of plasmid adaptation to feed has been described for IncHI1 plasmids isolated from horses [61]. These IncHI1 plasmids were found to carry a fos operon that allows degradation and utilization of fructooligosaccharides (scFOS), which are commonly used as prebiotics in horse diets. Another adaptation to the feed might be the association of E. coli isolates with the protein phytase that catalyses the hydrolysis of phytic acid. Chicken feed contains phytic acid (phytate), which is a source of phosphorus for the animal, but has also been shown to have an antibacterial effect, and it is possible that phytase neutralizes the antibacterial effect of phytic acid in the chicken gut [6263]. The GWAS on the IncK2 plasmids also identified genes associated with the chicken source and possibly linked to the use of specific feeds. One of the proteins that was significantly associated with the chicken IncK2 plasmids was the l-arabinose exporter YdeA. The chicken diet contains high doses of non-starch polysaccharides (NSPs), which adversely affect growth performance [64]. Supplementation of the chicken diet with enzyme cocktails has been proposed to improve the digestibility of NSPs, which has been shown to significantly increase levels of l-arabinose and other sugars [6566]. Recently, the European Food Safety Authority (EFSA) approved chicken feed additive products containing these enzymes that degrade NSP for use in Europe [6769]. The presence of a gene encoding for a l-arabinose export protein might therefore be an adaptation to the arabinose-rich environment of the chicken gut.

Another gene significant for the IncK2 plasmids isolated from chickens was a hypothetical protein adjacent to the hok/sok system. Interestingly, hok/sok was shown to shorten the lag phase, increase the exponential growth rate and increase overall cell densities of E. coli grown at 41–42 °C [70]. More research is needed to examine the function of the hypothetical gene and its relationship to the hok/sok system. The only gene that was significantly associated with IncK2 plasmids from humans was yffA, a gene of unknown function. This is particularly interesting considering that human IncK2 plasmids are more conserved compared to chicken IncK2 plasmids.

Previous reports demonstrate the vertical spread of clones and plasmids through the chicken production pyramid, as well as highly similar isolates from human and poultry [147172]. IncK2 plasmid transfer was also shown from E. coli to Salmonella Heidelberg in chicken ceca, which could potentially represent a public health risk [73]. More extensive surveillance and a comprehensive database are necessary to determine the prevalence of this plasmid variant. The possible risk for human health posed by bacteria and plasmid highly adapted to the chicken gut (which can easily contaminate the meat during processing) should also be assessed.

Presented data implicates genetic factors that might contribute to IncK2 plasmid’s adaptation to the chicken gut environment. Conducting in vitro assays is necessary to confirm that results obtained based on genetic data modelling are in agreement with actual IncK2 plasmid adaptation determinants. Performing a study that would assess the influence of maltose (present, for example, in soy), phytate and l-arabinose on the chicken gut microbiome and plasmidome would be a first step towards understanding the adaptation of bacteria carrying AMR plasmid to its host. In the future, these results could lead to interventions regarding poultry diet in order to reduce virulence- or antimicrobial-associated plasmids residing in gut bacteria.

In conclusion, the IncK2 plasmids in this dataset did not show high core genome diversity, and human IncK2 plasmids were more conserved compared to chicken IncK2 plasmids. The genes ydeA (responsible for l-arabinose transport) and one hypothetical gene were more likely to be associated with chicken IncK2 plasmids. In addition, multiple genes were associated with E. coli carrying IncK2 plasmids isolated from chickens. The presence of these genes both on the IncK2 plasmid and on the E. coli chromosome in isolates from chickens suggests a convergent adaptation of both to the chicken gut environment. More research is needed to understand the evolution of IncK2 plasmids in different niches and to confirm their significance for E. coli and IncK2 plasmids, including adaptation of these to chickens and their diet.

Supplementary material

10.1099/mgen.0.001535Uncited Supplementary Material 1.

10.1099/mgen.0.001535Uncited Supplementary Material 2.