key: cord-0017377-xfws5ndv
authors: Foster-Nyarko, Ebenezer; Alikhan, Nabil-Fareed; Ravi, Anuradha; Thomson, Nicholas M.; Jarju, Sheikh; Kwambana-Adams, Brenda A.; Secka, Arss; O’Grady, Justin; Antonio, Martin; Pallen, Mark John
title: Genomic diversity of Escherichia coli isolates from backyard chickens and guinea fowl in the Gambia
date: 2020-11-30
journal: Microb Genom
DOI: 10.1099/mgen.0.000484
sha: cf6a13263c6971139bcc930c82a9b2c27ae4b2a4
doc_id: 17377
cord_uid: xfws5ndv

Chickens and guinea fowl are commonly reared in Gambian homes as affordable sources of protein. Using standard microbiological techniques, we obtained 68 caecal isolates of Escherichia coli from 10 chickens and 9 guinea fowl in rural Gambia. After Illumina whole-genome sequencing, 28 sequence types were detected in the isolates (4 of them novel), of which ST155 was the most common (22/68, 32 %). These strains span four of the eight main phylogroups of E. coli, with phylogroups B1 and A being most prevalent. Nearly a third of the isolates harboured at least one antimicrobial resistance gene, while most of the ST155 isolates (14/22, 64 %) encoded resistance to ≥3 classes of clinically relevant antibiotics, as well as putative virulence factors, suggesting pathogenic potential in humans. Furthermore, hierarchical clustering revealed that several Gambian poultry strains were closely related to isolates from humans. Although the ST155 lineage is common in poultry from Africa and South America, the Gambian ST155 isolates belong to a unique cgMLST cluster comprising closely related (38–39 alleles differences) isolates from poultry and livestock from sub-Saharan Africa – suggesting that strains can be exchanged between poultry and livestock in this setting. Continued surveillance of E. coli and other potential pathogens in rural backyard poultry from sub-Saharan Africa is warranted.

The domestic chicken (Gallus gallus domesticus) is the most numerous bird on the planet, with an estimated population of over 22.7 billion -10 times more than any other bird [1] . Since their domestication from the red jungle fowl in Asia between 6000 and 8000 years ago [2, 3] , chickens have been found almost everywhere humans live. Other poultry, such as turkeys, guinea fowl, pheasants, duck and geese, are derived from subsequent domestication events across Africa, Europe and the Americas [4] . For example, the helmeted guinea fowl (Numida meleagris) originated in West Africa, although domesticated forms of this bird are now found in many parts of the tropics.

Poultry are reared for meat, eggs and feathers [5] . Poultry production is classified into four sectors, based on the marketing of poultry products and the level of biosecurity [6] . Intensive poultry farming falls under sectors 1 to 3, characterized by moderate to high levels of biosecurity, while sector 4 pertains to the 'backyard' , 'village' or 'family' poultry system, with few or no biosecurity measures.

Backyard poultry fulfil important social, economic and cultural roles in many societies. Seventy per cent of poultry production in low-income countries comes from backyard poultry [7] . The sale of birds and eggs generates income, while occasional consumption of poultry meat provides a source of protein in the diet. It is estimated that meat and eggs from backyard poultry contribute about 30 % of the total animal protein supply of households in low-income countries [8] .

In rural Gambia, backyard poultry can be offered as gifts for newlyweds or sold to solve family needs such as paying school fees, buying new clothes or other household needs [9] . The proximity between backyard poultry and humans may facilitate transmission of pathogens such as Escherichia coli between the two host species.

E. coli is a generalist bacterium that commonly colonizes the gastrointestinal tract of mammals and avian species [10] . Based on their pathogenic potential, E. coli can be divided into three categories: commensals, diarrhoeagenic E. coli and extraintestinal pathogenic E. coli (ExPEC). ExPEC frequently colonize the gut asymptomatically; however, they possess a wide range of unique virulence factors that enable them to colonize extraintestinal tissues in humans, pets and poultry [11, 12] . A sub-pathotype of ExPEC strains, known as avian pathogenic E. coli (APEC), causes colibacillosis -an extraintestinal disease in birds, with manifestations such as septicaemia, air sacculitis and cellulitis [13] . As a result of the high mortality and condemnation of birds associated with avian colibacillosis [14] , antimicrobials are often used in intensive farming systems to prevent bacterial infections and treat sick birds -a practice that has been linked to the development of antimicrobial resistance (AMR) in poultry.

Although previous studies have focused on the detection of AMR and documented the emergence of multiple-drug resistance (MDR) in this niche [15] [16] [17] [18] , little is known about the population structure of E. coli in rural backyard poultry. The Gambia does not have genomic data on E. coli from poultry prior to this study, and data on the circulating MLST types among poultry E. coli strains from sub-Saharan Africa is limited. However, reports from Ghana, Senegal and Nigeria have indicated the prevalence of ST624, ST69, ST540, ST7473, ST155, ST297, ST226, ST10, ST3625 and ST58 among E. coli isolates from commercial poultry [19] [20] [21] [22] . Given the increased exposure to humans, the natural environment and other animals, the population of E. coli in birds raised under the backyard system may differ considerably from those reared in intensive systems. It is also possible that the lineages of E. coli within local genotypes of rural poultry might differ between geographical regions. Previous studies have suggested that several E. coli clones are shared between poultry and humans, including isolates recovered from clinical cases. These include ST10, ST69, ST95, ST117, ST131, ST155, ST371, ST100, ST88 and ST23, ST38, ST3541, ST3018, ST58, ST6359, ST1011, ST746 and ST2676 [21, [23] [24] [25] [26] [27] [28] [29] . The absence of biosecurity measures in backyard poultry farming increases the potential for zoonotic transmission of pathogenic and/or antimicrobial-resistant strains to humans.

In a recent study of commercial broiler chickens, multiple colony sampling revealed that a single broiler chicken could harbour up to nine sequence types of E. coli [30] . However, within-host diversity of E. coli in backyard poultry, particularly in guinea fowl, has not been well studied and so we do not know how many lineages of E. coli can co-colonize a single backyard bird. To address these gaps in our knowledge, we exploited whole-genome sequencing to investigate the genomic diversity and burden of AMR among E. coli isolates from backyard chickens and guinea fowl in rural Gambia, West Africa.

The study population comprised 10 local-breed chickens and 9 guinea fowl from a village in Sibanor in the Western Region of the Gambia (Table 1) . Sibanor covers an area of approximately 90 km 2 and is representative of rural areas in the Gambia [31] . It has a population of about 10, 000. Most of the villagers are subsistence farmers growing peanuts, maize and millet. Households within this community comprise extended family units of up to 15 people, which make up the 'compound' . All guinea fowl were of the pearl variety, characterized by purplish-grey feathers dotted with white.

Domestic birds play a crucial role in human society, in particular contributing to food security in low-income countries. Many households in sub-Saharan Africa rear free-range chickens and guinea fowl, which are often left to scavenge for feed in and around the family compound, where they are frequently exposed to humans, other animals and the environment. Such proximity between backyard poultry and humans is likely to facilitate transmission of pathogens such as Escherichia coli or antimicrobial resistance between the two host species. Little is known about the population structure of E. coli in rural chickens and guinea fowl, although this information is needed to contextualize the potential risks of transmission of bacterial strains between humans and rural backyard poultry. Thus, we sought to investigate the genomic diversity of E. coli in backyard poultry from rural Gambia.

The sampling was done in November 2016. Poultry birds were first observed in motion for the presence of any abnormalities. Healthy-looking birds were procured from eight contiguous households within 0.3-0.4 km of each other and transported to the Abuko Veterinary Station, the Gambia in an airconditioned vehicle. A qualified veterinarian then euthanized the birds and removed their caeca under aseptic conditions. These were placed into sterile Falcon tubes and flash-frozen on dry ice in a cooler box. The samples were transported to the Medical Research Council Unit The Gambia at the London School of Hygiene and Tropical Medicine labs in Fajara, where the caecal contents were aseptically emptied into new Falcon tubes for storage at −80 °C within 3 h. A peanut-sized aliquot was taken from each sample into a 1.8 ml Nunc tube containing 1 ml of skim-milk-tryptone-glucose-glycerol (STGG) transport and storage medium (Oxoid, Basingstoke, UK), vortexed at 4200 r.p.m. for 2 min and frozen at −80 °C. Fig. 1 summarizes the sample processing flow.

The caecal-STGG suspension was removed from −80 °C storage and allowed to thaw briefly on wet ice. A 100 µl aliquot was then taken into 900 µl of physiological saline (0.85 %) and taken through four 10-fold serial dilutions. A 100 µl aliquot each was then taken from the dilutions and uniformly streaked onto tryptone-bile-X-glucoronide agar plates using the spread plate technique. The inoculated plates were incubated at 37 °C for 18-24 h under aerobic conditions. Following overnight incubation, colony counts were determined for raised, translucent and entire colonies that exhibited bluish-green pigmentation typical of E. coli. Up to five candidate colonies were selected per sample and subcultured on MacConkey agar. These were incubated at 37 °C in air for 18-24 h and stored in 20 % glycerol broth at −80 °C. The isolates from chickens were designated C1-C10, while those from guinea fowl were prefixed by GF1-GF9, followed by the respective colony number (1 up to 5).

Genomic DNA was extracted from overnight broth cultures prepared from each single colony sub-culture using an in-house 96-well plate lysate method as described previously [32] . The DNA was eluted in Tris/Cl (pH, 8.0) and quantified using the Qubit high-sensitivity DNA assay kit (Invitrogen, MA, USA). DNA samples were kept at −20 °C until the Illumina sequencing library preparation. Broth cultures were spun at 3500 r.p.m. for 2 min and lysed using lysozyme, proteinase K, 10 % SDS and RNase A in Tris EDTA buffer (pH 8.0).

Whole-genome shotgun sequencing of the DNA extracts was performed for all the study isolates on the Illumina NextSeq 500 instrument (Illumina, San Diego, CA, USA) using a modified Illumina Nextera library preparation protocol as described previously [32] . We ran the final pooled library at a concentration of 1.8 pM on a mid-output flow cell (NSQ 500 Mid Output KT v2 300 cycles; Illumina catalogue no. FC-404-2003) according to the manufacturer's instructions. Following sequencing, FASTQ files were downloaded from BaseSpace to a local server hosted at the Quadram Institute Bioscience.

The raw sequences were initially analysed on the Cloud Infrastructure for Microbial Bioinformatics [33] . This included concatenating paired-end short reads, quality checks with FastQC v0.11.7 [34] , trimming of low-quality reads (median quality below a Phred score of~30 and read lengths below 36 bp) and Illumina adapters with Trimmomatic v0.39 [35] and assembly by Spades v3.13.2 [36] . The quality of the assemblies was checked using QUAST v5.0.0, de6973bb [37] and annotation of the draft genomes was carried out using Prokka v1.13.3 [38] . We used the mlst software (https:// github. com/ tseemann/ mlst) to call multilocus sequence types (MLSTs) using the Achtman scheme [39] , based on the seven house-keeping genes, adk, fumC, gyrB, icd, mdh, purA and recA. We used Snippy v4.3.2 (https:// github. com/ tseemann/ snippy) for variant calling and to generate a core-genome alignment, from which a maximum-likelihood phylogenetic tree was reconstructed using RAxML v8.2.4 [40] , based on a general time-reversible nucleotide substitution model with 1, 000 bootstrap replicates. We included representative reference genome sequences for the major phylogroups of E. coli and Escherichia fergusonii as an outgroup (File S1). Given that recombination is widespread in E. coli and tends to blur phylogenetic signals [39] , we used Gubbins (Genealogies Unbiased By recomBinations In Nucleotide Sequences) [41] to detect and mask recombinant regions of the core-genome alignment prior to the phylogenetic reconstruction. We used the GrapeTree [42] to visualize and annotate phylogenetic trees. We calculated pair-wise single-nucleotide polymorphism (SNP) distances between genomes from the core-genome alignment using snp-dists v0.6 (https:// github. com/ tseemann/ snp-dists).

Subsequently, the short-read sequences were uploaded to EnteroBase [43] , an online genome database and integrated software environment that currently hosts more than 138 164 E. coli genomes, sourced from all publicly available sequence databases and user uploads. EnteroBase routinely retrieves short-read E. coli sequences from the public domain, performs quality control and de novo assemblies of Illumina short-read sequences, annotates these and assigns seven-allele MLST (ST) and phylogroups from genome assemblies using standardized pipelines. In addition, EnteroBase assigns unique core-genome MLST (cgMLST) numbers to each genome, based on the typing of 2, 512 genes in E. coli.

We utilized the hierarchical clustering (HierCC) algorithm in EnteroBase to assign our poultry genomes to eleven stable clusters designated as HC0 up to HC1100, based on pair-wise differences between genomes at cgMLST alleles. In Salmonella, the HC100 or HC200 clusters seem to correspond to long-term strain endemicity, while in E. coli, HC1100 corresponds to the seven-allele MLST clonal complexes [43] . The HierCC algorithm therefore lends itself as a very useful tool for the analysis of bacterial population structures at multiple levels of resolution. In a recent study of the population structure of Clostridioides difficile, Frentrup et al. [44] showed that HierCC allows closely related neighbours to be detected at 89 % consistency between cgMLST pair-wise allelic differences and SNPs. We determined the closest relatives to our study E. coli isolates using the HC1100 cluster and reconstructed neighbour-joining trees using NINJA [45] . In order to compare the strain distribution that we observed among our study isolates with what pertains in poultry E. coli isolates from elsewhere, we further retrieved genomic assemblies from all publicly available poultry E. coli isolates, stratified by their source continent and reconstructed NINJA neighbour-joining trees depicting the prevalence of STs per continent.

We used ARIBA v2.12.1 [46] to detect virulence factors, antimicrobial resistance genes and plasmid replicons among our study isolates. Briefly, this tool scans the shortread sequences against the core Virulence Factors Database (VFDB) [47] (virulence factors), ResFinder (AMR) [48] and PlasmidFinder (plasmid-associated genes) [ 

Two novel strains recovered from guinea fowl were longread sequenced on the Oxford Nanopore platform as follows. Prior to sequencing, DNA fragments were assessed using the Agilent 2200 TapeStation (Agilent catalogue no. 5067-5579) to determine the fragment lengths. Long-read sequencing was carried out using the rapid barcoding kit (Oxford Nanopore catalogue no. SQK-RBK004). Libraries were prepared following the manufacturer's instructions. An input DNA concentration of 400 ng was used for the library preparation and a final concentration of 75 µl of the prepared library was loaded onto an R9.4 MinION flow cell. The final concentration of the library pool was assessed using the Qubit high-sensitivity DNA assay (Invitrogen, MA, USA).

The long reads were base-called with Guppy, the Oxford Nanopore Technologies' post-sequencing processing software (https:// nanoporetech. com/). The base-called FASTQ files were then concatenated into a single file each and demultiplexed based on their respective barcodes, using the qcat Python command-line tool v1.1.0 (https:// github. com/ nanoporetech/ qcat). We performed hybrid assemblies of the Illumina and Nanopore reads with Unicycler v0.4.8.0 [51] . The quality of the hybrid assemblies was assessed with QUAST v5.0.0, de6973bb [37] . The hybrid assemblies were then analysed for the presence of plasmids and prophages using ABRicate PlasmidFinder and PHASTER [52] respectively. Annotations of the assemblies were carried out using Prokka v1.13.3 [38] .

We analysed 19 caecal samples obtained from 10 chickens and 9 guinea fowl. Fifteen out of the 19 (79 %) samples yielded growth of E. coli on culture, from which 68 colonies were recovered (5 colonies from each of 13 birds, 2 from a single bird, and 1 colony from another bird).

We recovered 28 7-allele sequence types (STs), of which ST155 was the most common (22/68, 32 %). Four of the STs were novel -two from chickens and two from guinea fowl. The allelic profiles of the novel strains are provided in File S2. Seventeen of the 28 STs have previously been isolated from humans or other vertebrates, 6 (ST942, ST2165, ST2461, ST4392, ST5826 and 6186) have not been seen in humans before and 1 (ST6025) only occurred in 1 other isolate in EnteroBase, beside the study strain. However, the source of isolation of this other isolate was not available ( Table 2 ). The isolates were spread over phylogroups B1, A, B2 and D, but most belonged to phylogroups B1 and A, which are home to strains associated with human intestinal infections and avian colibacillosis [53, 54] (Fig. 2) . Hierarchical clustering resolved the study strains into 22 cgMLST complexes, indicating a high level of genomic diversity (File S2).

We generated complete, circular genome assemblies of the two novel sequence types isolated from guinea fowl: ST10654 (GF3-3) and ST9286 (GF4-3). Although neither strain encoded AMR genes or plasmids, GF3-3 contained three prophages (two intact, one incomplete), while GF4-3 harboured four prophages (three intact, one incomplete) (File S3).

Several birds (12/19, 63 %) were colonized by two or more STs; in most cases, the STs spanned more than two phylotypes (Table 1 ). In two chickens, all five colony picks belonged to distinct STs. We observed some genetic diversity among multiple colonies of the same ST recovered from the same host (Table 3a) . Most of these involved variants that differed by 0-4 SNPs, i.e. variation likely to have arisen due to within-host evolution. However, in one instance, pair-wise SNP differences (ranging from 4 to 255) suggested independent acquisition of distinct clones. Pair-wise SNP analysis also suggested transmission of strains (including MDR isolates) between chickens and between chickens and guinea fowl (Table 3b , c) from the same household (File S4).

Twenty isolates (20/68, 29 %) harboured at least one AMR gene and 16 (16/68, 24 %) were MDR, i.e. positive for genes predicted to convey resistance to three or more classes of antibiotics ( Fig. 3; File S5) . Fourteen of the 16 MDR isolates belonged to ST155 -representing 64 % (14/22) of the ST155 isolates recovered in this study. Notable among the resistance genes detected was the class A broad-spectrum 'C' denotes chickens and 'GF' denotes guinea fowl. Table 3b . Single-nucleotide polymorphism differences between isolates recovered from chicken 3, chicken 5, chicken 6 and guinea fowl 7. All the isolates in this transmission network encoded resistance to ≥3 classes of antimicrobials Interestingly, the MDR isolates also harboured more genes encoding putative virulence factors than did less-resistant isolates (Fig. 2) . Overall, 125 unique virulence-associated genes were detected from the study isolates (File S6) . Notably, the virulence and AMR profiles of co-colonizing STs tended to differ from each other. Scrutiny of annotated assemblies revealed that the resistance genes were often co-located on the same contig as one of the IncF plasmids. In three birds (guinea fowl 2, guinea fowl 5 and guinea fowl 7), co-colonizing strains (belonging to different STs) shared the same plasmid profile. The results of ARIBA ResFinder, PlasmidFinder and VFDB were 100 % concordant with those produced by ABRicate for our study isolates.

Hierarchical clustering analyses provided evidence of genomic relationships between strains from poultry and those from humans (Table 4) ; however, this warrants further investigation using samples collected from poultry and humans living in close proximity from the same setting. Significant among these were ST2772 and ST4392, which were separated from human isolates belonging to these STs by just 41 and 68 alleles in the core-genome MLST scheme, respectively (Figs 4 and 5). Similarly, ST86, ST6186 and ST602 were closest to isolates from livestock (Figs S9-S11), suggesting possible exchange of strains between livestock species.

By contrast, three of the novel STs from this study (ST10654, ST9285, ST9286) were genetically distinct from anything else in the public domain. These belonged to unique HC1100 clusters in the cgMLST scheme and did not have any relatives in the seven-allele MLST scheme, even after allowing for two mismatches. Two of these (ST10654 from Guinea fowl 3 and ST9286 from Guinea fowl 4) now have complete genomic assemblies.

Phylogenomic analyses of 4, 846 poultry E. coli isolates from all over the world revealed that ST155 is common among poultry isolates from Africa and South America (Figs S1 and S2). In contrast, ST117 is prevalent among poultry isolates from Europe and North America (Figs S3 and S4) , with ST156 and ST254 being the most common E. coli STs found in poultry from Asia and Oceania, respectively (Figs S5 and S6 ).

Our phylogenetic analyses revealed that ST155 strains from Africa were dispersed among other ST155 isolates from the rest of the world; however, the majority of ST155 strains from this study belonged to a tight genomic cluster, comprising isolates from poultry and livestock from sub-Saharan Africa (separated by 38-39 alleles), except for a single isolate sourced from poultry in the USA. In the cgMLST scheme, all the study ST155 isolates fell into four HC100 sub-clusters (100 alleles difference) (Fig. S7) . The largest sub-cluster (sub-cluster 1, HC100_43137) comprised ST155 isolates from this study and isolates from Uganda and Kenya; while sub-clusters 2 (HC100_73903), 3 (HC100_73905) and 4 (HC100_93719) occurred in the Gambia only, although distantly related to isolates from humans and a companion animal (Fig. S8) . Antimicrobial resistance was high across the continents, with the highest prevalence of MDR in South America (100/131, 77 %), followed by Asia (175/249, 70 %) and then Africa (392/591, 66 %) ( Table 5 ; File S8). Of note, the highest percentages of resistance globally were those for broad-spectrum beta-lactamases, while the lowest percentages of resistance were to colistin (File S8). Interestingly, the prevalence of colistin resistance was highest in Europe but did not occur in Oceania and North America.

Here, we have described the genomic diversity of E. coli from backyard chickens and guinea fowl reared in households in rural Gambia, West Africa. Backyard poultry from this rural setting harbour a remarkably diverse population of E. coli strains that encode antimicrobial resistance genes and virulence factors that are important for infections in humans. Furthermore, we provide evidence of sharing of strains (including MDR strains) from poultry to poultry and between poultry, livestock and humans, with potential implications for public health.

Our results reflect the rich diversity that exists within the E. coli population from backyard poultry. Although our sample size was small (19 birds), we recovered as many as 28 STs of E. coli, 4 of which have not been seen before -even though more than quarter of a million E. coli strains had been sequence typed to date (March 2020). Three of our novel STs differed by >945 alleles from their nearest relative. Two of these now have complete assemblies. Also, some of the strains from this study were found in unique cgMLST HierCC clusters containing strains only from this study.

Our results confirm previous reports that phylogroups B1 and A are the dominant phylogroups among E. coli isolates from both intensive and backyard poultry [55] [56] [57] [58] . Hierarchical clustering analysis suggested that ST155 is common in African poultry. However, most of our ST155 strains belong to a unique cgMLST cluster containing closely related (38-39 alleles differences, and so presumably recently diverged) isolates from poultry and livestock from sub-Saharan Africa, suggesting that strains can be exchanged between livestock and poultry in this setting.

Rural backyard poultry can act as a source of transmission of infections to humans, due to the absence of biosecurity and daily contact with humans [59] . Indirect contact might occur through food or through contact with faeces; for example, by children who are often left to play on the ground [60] .

We observed a high prevalence of AMR genes among E. coli isolates sourced from African poultry. Similarly, high rates of genotypic MDR were detected among poultry E. coli isolates from the rest of the world, with ESBL (various types) being the most significant resistance gene detected. Poultry-associated ESBL genes have also been found among human clinical isolates [61] . Strikingly, most of our ST155 isolates encoded resistance to ≥3 classes of clinically relevant antibiotics, with the highest percentages seen for bla TEM-1 beta-lactamase and tetracycline (tetA). This is worrying, as beta-lactamasepositive isolates are often resistant to several other classes of antibiotics [62, 63] .

Our results are consistent with previous studies that reported ST155 isolates to be commonly associated with MDR [64, 65] , but differ from other studies that have reported a low prevalence of AMR in backyard poultry. For example, in a study that The absence of resistance in that study was attributed to a lack of exposure to antimicrobials. Similarly, E. coli from organic poultry in Finland were reported to be highly susceptible to most of the antimicrobials studied and no ESBL resistance was detected [67] .

Although tetracycline is commonly used in poultry farming for therapeutic purposes [68] , resistance to this antibiotic is known to be prevalent in poultry, even in the absence of the administration of this antibiotic [69] . Our results also suggest that IncF plasmids may play a role in the dissemination of AMR in our study population. Conjugation assays are needed to confirm the association of these plasmids with the observed resistance genes and the mobilisability of the plasmids and thus, the potential for exchange among co-colonizing strains in a single host; however, these could not be performed due to coronavirus disease 2019 (Covid-19) restrictions.

Many sub-Saharan countries lack clear guidelines on the administration of antibiotics in agriculture, although an increasing trend in the veterinary use of antimicrobials has been documented [70] . The use of antimicrobials in developing countries is likely to increase because of increasingly intensive farming practices [71] . Europe has banned the use of antimicrobials as growth promoters since 2006 [72] and the use of all essential antimicrobials for prophylaxis in animal production since 2011 [73] . However, AMR may be less well controlled in other parts of the world.

Although APEC strains span several phylogroups (A, B1, B2 and D) and serogroups [54] , the majority of APEC strains encode virulence genes associated with intestinal or extraintestinal disease in humans. These include adhesion factors, toxins, iron acquisition genes and genes associated with serum resistance, such as fyuA, iucD, iroN, iss, irp2, hlyF, vat, kpsM and ompT. Although APEC isolates present different combinations of virulence factors, each retains the capability to cause colibacillosis [13, 74] . We did not detect haemolysin or serum survival genes in our study isolates; however, we recovered some of the known markers of intestinal and extraintestinal virulence in some study isolates, such as the enteroaggregative E. coli heat-stable enterotoxin and the vacuolating autotransporter toxin (vat, astA), invasion and evasion factors (kpsM, kpsD, pla) and adherence factors (fim and pap genes) that are associated with intestinal and extraintestinal infections in humans. Thus, these strains could cause disease in humans, should they gain access to the appropriate tissues.

Several birds were colonized with two or more STs and at least two phylotypes of E. coli. This level of diversity is probably a consequence of the frequent exposure of backyard poultry to the environment, livestock and humans. Co-colonization of single hosts with multiple strains may facilitate the spread of AMR-and virulence-associated genes from resistant strains to other bacteria via both horizontal and vertical gene transfer [75] . A high co-colonization rate of E. coli has been described in humans [76, 77] and in non-human primates [32] , involving pathogenic strains of E. coli. Recently, Li et al. reported three to nine sequence types of colistin-resistant E. coli to co-exist within a single broiler chicken [30] . Here, we report co-colonization with different lineages of E. coli in backyard chickens and guinea fowl. Unsurprisingly, co-colonizing strains often had different AMR and virulence patterns.

An obvious limitation of our study is the small sample size. This study could have also been enhanced by sampling E. coli from humans within close proximity to our backyard birds, but we could not perform an analysis of E. coli from sympatric humans from our study setting due to logistic reasons and funding limitations of our study. Nonetheless, the inclusion of publicly available sequences strengthens our analysis and inference of the population of E. coli in this setting. We also could not perform phenotypic susceptibility testing on all isolates. We acknowledge that a minor percentage of genotypic resistance predictions fail to correspond with phenotypic resistance [78] .

Taken together, our results indicate a rich diversity of E. coli within backyard poultry from the Gambia, characterized by strains with a high prevalence of AMR and the potential to contribute to infections in humans. This, coupled with the potential for the exchange of strains between poultry and livestock within this setting, might have important implications for human health and warrants continued surveillance. 

The broiler chicken as a signal of a human reconfigured biosphere

Investigating the global dispersal of chickens in prehistory using ancient mitochondrial DNA signatures

Chicken domestication: an updated perspective based on mitochondrial genomes

Village poultry: still important to millions, eight thousand years after domestication

Family poultry: multiple roles, systems, challenges, and options for sustainable contributions to household nutrition security through a planetary health lens

Recommendations on the Prevention, Control and Eradication of Highly Pathogenic Avian Influenza in Asia

Smallholder poultry: contributions to food and nutrition security, in encyclopedia of food security and sustainability

Poultry as a Tool in Poverty Eradication and Promotion of Gender Equality

Rural household chicken management and challenges in the upper river region of the Gambia

Diarrheagenic Escherichia coli

The link between phylogeny and virulence in Escherichia coli extraintestinal infection

Identification of forces shaping the commensal Escherichia coli genetic structure by comparing animal and human isolates

Characterizing the APEC pathotype

Diseases of Poultry, 12th. Iowa: Iowa State University Press

Impacts of small-scale chicken farming activity on antimicrobial-resistant Escherichia coli carriage in backyard chickens and children in rural Ecuador

Identification and antimicrobial susceptibility profile of Escherichia coli isolated from backyard chicken in and around ambo, central Ethiopia

Antimicrobial resistance genes in Salmonella and Escherichia coli isolates from chicken droppings in

Characterization of avian pathogenic Escherichia coli isolated from free-range helmeted guineafowl

Molecular characterization of Escherichia coli isolated from chickens with colibacillosis in Senegal

Whole-genome sequencing of Escherichia coli isolated from contaminated meat samples collected from the Northern Region of Ghana reveals the presence of multiple antimicrobial resistance genes

Detection and characterization of ESBL-producing Escherichia coli from humans and poultry in Ghana

Detection and molecular characterisation of extended-spectrum β-lactamase-producing enteric bacteria from pigs and chickens in Nsukka

Chicken as reservoir for extraintestinal Pathogenic Escherichia coli in humans

Multilocus sequence typing and virulence gene profiles associated with Escherichia coli from human and animal sources

Molecular characteristics of mcr-1-carrying plasmids and new mcr-1 variant recovered from polyclonal clinical Escherichia coli from Argentina and Canada

mcr-1 is borne by highly diverse Escherichia coli isolates since 2004 in food-producing animals in Europe

Do human extraintestinal Escherichia coli infections resistant to expandedspectrum cephalosporins originate from food-producing animals? A systematic review

Extended-spectrum and AmpC β-lactamase-producing Escherichia coli in broilers and people living and/or working on broiler farms: prevalence, risk factors and molecular characteristics

Escherichia coli and urinary tract infections: the role of poultry-meat

Within-Host heterogeneity and flexibility of mcr-1 transmission in chicken gut

Effects of community-wide vaccination with PCV-7 on pneumococcal nasopharyngeal carriage in the Gambia: a cluster-randomized trial

Genomic diversity of Escherichia coli isolates from non-human primates in the Gambia

CLIMB (the cloud infrastructure for microbial bioinformatics): an online resource for the medical microbiology community

FastQ screen: a tool for multi-genome mapping and quality control

Trimmomatic: a flexible trimmer for illumina sequence data

SPAdes: a new genome assembly algorithm and its applications to single-cell sequencing

QUAST: quality assessment tool for genome assemblies

Prokka: rapid prokaryotic genome annotation

Sex and virulence in Escherichia coli: an evolutionary perspective

RAxML version 8: a tool for phylogenetic analysis and post-analysis of large phylogenies

Rapid phylogenetic analysis of large samples of recombinant bacterial whole genome sequences using Gubbins

Grape-Tree: visualization of core genomic relationships among 100,000 bacterial pathogens

The EnteroBase user's guide, with case studies on Salmonella transmissions, Yersinia pestis phylogeny, and Escherichia core genomic diversity

Global genomic population structure of Clostridioides difficile

Large-scale neighbor-joining with NINJA

ARIBA: rapid antimicrobial resistance genotyping directly from sequencing reads

VFDB 2019: a comparative pathogenomic platform with an interactive web interface

Identification of acquired antimicrobial resistance genes

In silico detection and typing of plasmids using Plas-midFinder and plasmid multilocus sequence typing

Agar and broth dilution methods to determine the minimal inhibitory concentration (MIC) of antimicrobial substances

Unicycler: resolving bacterial genome assemblies from short and long sequencing reads

PHASTER: a better, faster version of the PHAST phage search tool

Characterization and rapid identification of phylogroup G in Escherichia coli, a lineage with high virulence and antibiotic resistance potential

Human and avian extraintestinal pathogenic Escherichia coli: infections, zoonotic risks, and antibiotic resistance trends

Phylogenetic group determination of Escherichia coli isolated from animals samples

ECOR phylotyping and determination of virulence genes in Escherichia coli isolates from pathological conditions of broiler chickens in poultry slaughter-houses of southeast of Iran

Virulence gene profiles, antimicrobial resistance and phylogenetic groups of fecal Escherichia coli strains isolated from broiler chickens in Algeria

Molecular characterization of Escherichia coli Isolated from chickens with colibacillosis in Senegal

An epidemiologic simulation model of the spread and control of highly pathogenic avian influenza (H5N1) among commercial and backyard poultry flocks in South Carolina, United States

Clonal differences between Non-Typhoidal Salmonella (NTS) recovered from children and animals living in close contact in the Gambia

High heterogeneity of Escherichia coli sequence types harbouring ESBL/AmpC genes on IncI1 plasmids in the Colombian poultry chain

The role of epidemic resistance plasmids and international high-risk clones in the spread of multidrug-resistant Enterobacteriaceae

Extended-Spectrum β-lactamaseand AmpC-Producing enterobacteria in healthy broiler chickens

Characterization of Escherichia coli carrying mcr-1-plasmids recovered from food animals arom Argentina. Frontiers in Cellular and Infection Microbiology

Longitudinal study of ESBL Escherichia coli carriage on an organic broiler farm

Comparative possession of Shiga toxin, intimin, enterohaemolysin and major extended spectrum beta lactamase (ESBL) genes in Escherichia coli isolated from backyard and farmed poultry

Zoonotic public health hazards in backyard chickens

Effects of orally administered tetracycline on the intestinal community structure of chickens and on tet determinant carriage by commensal bacteria and Campylobacter jejuni

Epidemiology of resistance to antibiotics links between animals and humans

Antibiotic resistance in Escherichia coli in husbandry animals: the African perspective

Global trends in antimicrobial use in food animals

History of the use of antibiotic as growth promoters in European poultry feeds

Restrictions on antimicrobial use in food animal production: an international regulatory and economic survey

Diagnostic strategy for identifying avian pathogenic Escherichia coli based on four patterns of virulence genes

Antibiotic resistance-the need for global solutions

The homogeneity of the faecal coliform flora of normal school-girls, characterized by serological and biochemical properties

Clonal diversity of Escherichia coli colonizing stools and urinary tracts of young girls

Discordant bioinformatic predictions of antimicrobial resistance from whole-genome sequencing data of bacterial isolates: an inter-laboratory study

We thank Dr Duto Fofana and Dr Ousman Ceesay of the Department of Livestock Services, Veterinary Services, Ministry of Agriculture, the Gambia for kindly providing access to their veterinary station at Abuko. We gratefully acknowledge the Research Support Office at the Medical Research Council Unit The Gambia at the London School of Hygiene and Tropical Medicine, in particular Ms Isatou Cham, for project management support. We also thank Mr Steven Rudder and Mr David Baker for their assistance with the Illumina and Oxford Nanopore sequencing. 

The authors declare that there are no conflicts of interest.

The study was approved by the Medical Research Council Unit The Gambia at London School of Hygiene and Tropical Medicine's Scientific Co-ordinating Committee (SCC 1615).