The microbial ecology of Escherichia coli in the vertebrate gut

Abstract

Escherichia coli has a rich history as biology's ‘rock star’, driving advances across many fields. In the wild, E. coli resides innocuously in the gut of humans and animals but is also a versatile pathogen commonly associated with intestinal and extraintestinal infections and antimicrobial resistance—including large foodborne outbreaks such as the one that swept across Europe in 2011, killing 54 individuals and causing approximately 4000 infections and 900 cases of haemolytic uraemic syndrome. Given that most E. coli are harmless gut colonizers, an important ecological question plaguing microbiologists is what makes E. coli an occasionally devastating pathogen? To address this question requires an enhanced understanding of the ecology of the organism as a commensal. Here, we review how our knowledge of the ecology and within-host diversity of this organism in the vertebrate gut has progressed in the 137 years since E. coli was first described. We also review current approaches to the study of within-host bacterial diversity. In closing, we discuss some of the outstanding questions yet to be addressed and prospects for future research.

This review presents an overview of E. coli diversity studies encompassing human and other nonhuman vertebrate hosts in the 137 years since this organism was first described and outstanding gaps in our knowledge.

Introduction

Escherichia coli: a versatile organism

‘Tout ce qui est vrai pour le Colibacille est vrai pour l'éléphant" (All that is true of E. coli is also true of the elephant; Friedmann 2004). This widely quoted epigram reflects the exalted status of E. coli in microbiology as the favourite model organism. Many advances in biology have been driven by studies with E. coli, particularly, using the strain designated K-12 and its derivatives (Tenaillon et al. 2010, Friedmann 2014; Table 1).

Table 1.

The contribution of studies using E. coli across various fields.

Field(s)	Contribution of studies with E. coli	Reference(s)
Molecular biology, physiology, and genetics	The explanation of the genetic code	Crick et al. 1961
	DNA replication	Lehman et al. 1958
	Transcription	Stevens 1960
	The life cycle of bacterial viruses	Ellis and Delbrück 1939, Lwoff 1953
	The elucidation of the molecular basis of antibiotic tolerance	Hu and Coates 2005
	The discovery of restriction enzymes	Linn and Arber 1968, Meselson and Yuan 1968
	Swarming motility	Harshey and Matsuyama 1994, Inoue et al. 2007
	Gene regulation	Jacob et al.1960
	Elucidation of the structure and function of ATP synthase	Capaldi et al. 2000
Pharmaceuticals	The synthesis of recombinant proteins in vivo, such as insulin, which is used to treat millions of people with diabetes worldwide	Swartz 2001
	The synthesis of several other biopharmaceuticals, such as human interferon-β, interleukin-2, human growth hormone, and human blood clotting factors	Huang et al. 2012, Baeshen et al. 2015
Evolution	Escherichia coli is the model organism of choice in experimental evolution studies; e.g. in Lenski's long-term evolution experiment, on-going since February 1988 and spanning over 60 000 generations	Good et al. 2017
	The demonstration of the stochastic nature of mutations	Luria and Delbrück 1943, Lederberg and Lederberg 1952
	Mapping the trajectory of long-term fitness	Wiser et al. 2013
	The elucidation of how sexual recombination influences adaptation	Cooper 2007
	Insights into predator–prey interactions	Lenski 1988
	The evolution of a novel trait, aerobic citrate utilization	Blount et al. 2012
Genetic engineering and biotechnology	The development of recombinant DNA techniques and molecular cloning, the production of biofuels and industrial chemicals such as phenol	Kim et al. 2014
	Mannitol production	Kaup et al.
	Ethanol production	Hildebrand et al. 2013, Liu and Khosla 2010

Theodor Escherich, a Bavarian doctor, first described E. coli on 14 July 1885 in a lecture to the Society for Morphology and Physiology in Munich (Escherich 1988). Over 15 months, Escherich observed and isolated 19 different bacteria in Otto von Bollinger's bacteriology laboratory (Escherich 1988), including the Bacterium coli commune (‘the common colon bacillus’; Escherich and Pfaundler 1903, Shulman et al. 2007)—which we now know as E. coli—and Bacterium lactis aërogenes (now Klebsiella pneumoniae). Aided by Christian Gram's new staining technique (Escherich 1888, Bettelheim 1986, Friedmann 2014), Escherich described the organism as a Gram-negative bacillus of approximately 1.1–1.5 μm × 2.0–6.0 μm. The name E. coli was proposed by Castellani and Chalmers in 1919 to honor Escherich, and was officially adopted in 1958 by the Judicial Commission of the International Committee on Systematic Bacteriology. It was subsequently included in the Approved Lists of Bacterial Names in 1980 (Castellani and Chalmers 1919, Judicial Commission of the International Committee on Bacterial Nomenclature 1958, Skerman et al. 1980, Bettelheim 1986).

In the wild, E. coli exists as a common resident of the vertebrate gut and nonhost associated habitats such as water, soil, manure, and food (Macfarlane and Macfarlane 1997, Alm et al. 2011, van Elsas et al. 2011, Blount 2015). Due to its remarkable metabolic and regulatory abilities, E. coli can also survive under prolonged periods of nongrowth (Leimbach et al. 2013). Thus, E. coli represents a highly versatile species, capable of adapting to many different ecological habitats (Leimbach et al. 2013). Escherichia coli is also a versatile pathogen, eliciting a broad spectrum of diseases and responsible for at least 2 million human deaths per year (Kosek et al. 2003). The organism's role in intestinal and extraintestinal disease was recognized not long after its discovery (Friedmann 2014). Ørskov and Ørskov note, ‘Any E. coli strain can probably cause invasive disease given the right opportunities, and E. coli has therefore aptly been called an opportunistic pathogen’ (Orskov and Orskov 1992).

Escherichia coli is a facultative anaerobe, meaning that it can grow in aerobic and anaerobic conditions. It is the most common aerobe in the lower intestine of mammals (Tenaillon et al. 2010); however, it typically constitutes only 0.1%–5% of a gut microbial community comprised of over 500 other bacterial species (Tenaillon et al. 2010). This community is typically dominated by obligate anaerobes, including members of Firmicutes and Bacteroidetes—which make up at least 90% of the gut microbial population (Bäckhed et al. 2005, Claesson et al. 2009, Tenaillon et al. 2010). Nevertheless, E. coli can hold its ground in this highly competitive, ever-changing niche, existing in a life-long relationship with its host.

Escherichia coli pathotypes

Pathogenic E. coli are classified into ‘pathotypes’ or ‘pathovars’ (Kaper et al. 2004, Croxen and Finlay 2010; Table 2) based on several criteria, including the site of infection (e.g. uropathogenic E. coli (UPEC), named for their impact on the urinary tract, extraintestinal pathogenic E. coli (ExPEC), which cause infections in organs outside the gut), host (e.g. avian pathogenic E. coli (APEC), named after infections in avian species), site and host (e.g. neonatal meningitis E. coli (NMEC), which infect the cerebrospinal fluid in new-borns), and pathogenesis (e.g. Shiga-toxigenic E. coli (STEC)).

Table 2.

Classification and examples of the pathotypes of E. coli.

Pathotype/pathovar	Virulence mechanism(s)	Host range	Reference
InPEC	Locus of enterocyte effacement; pathogenicity island 1	Humans and all mammals	Robins-Browne et al. (2016)
InPEC	Small fimbrial adhesins; toxins; transcriptional activator gene; aggregative adhesion	HumansRobins-Browne et al. (2016)	Robins-Browne et al. (2016)
InPEC	Shiga toxin or verotoxins; afimbrial and fimbrial adhesins	Humans and piglets	Robins-Browne et al. (2016)
InPEC	Heat labile and heat-stable enterotoxins	Humans, ruminants, pigs, and dogs	Robins-Browne et al. (2016)
InPEC	Invasion and multiplication in enterocytes	Humans and primates	Robins-Browne et al. (2016)
InPEC	Adhesins	Humans and animals	Robins-Browne et al. (2016)
InPEC	Shiga toxin	Humans	[323]
InPEC	Adherent invasive phenotype	Humans and animals	Robins-Browne et al. (2016)
ExPEC	Adhesins, secretion and iron uptake systems, increased serum survival and cytotoxic proteins	Birds	Sheldon et al. 2010
ExPEC	Fimbrial adhesins; siderophores, resistance to complement	Humans and animals (especially dogs and cats)	Mainil 2013
ExPEC	Iron acquisition systems, degradation of interferon-gamma, and cleavage of the human defensin LL-37	Humans	Nicholson et al. (2016)
ExPEC	Fimbrial adhesins; siderophores; resistance to complement	All mammals and birds (especially poultry)	Mainil 2013
ExPEC	Unknown	Animals	Blum et al. 2015
ExPEC	Type II, IV, and VI secretion systems, long polar fimbriae (lpfA) and iron acquisition	Animals	Dogan et al. 2012
ExPEC	Cytotoxic Necrotizing Factors 1 or 2 and α haemolysin; fimbrial and/or afimbrial adhesins; siderophores; resistance to complement	Humans, animals, and ruminants	Mainil 2013

InPEC, Intestinal pathogenic E. coli; ExPEC, Extraintestinal pathogenic E. coli; EPEC, Enteropathogenic E. coli; EAEC, Enteroaggregative E. coli; EHEC, Enterohaemorrhagic E. coli; ETEC, Enterotoxigenic E. coli; EIEC, Enteroinvasive E. coli; DAEC, Diffusely adherent E. coli; STEAEC, Shiga-toxigenic Enteraggregative E. coli; AIEC, Adherent invasive E. coli; APEC, Avian pathogenic E. coli; UPEC, Uropathogenic E. coli; NMEC, Neonatal meningitis E. coli; SePEC, Human sepsis-associated E. coli; MPEC, Mammary pathogenic E. coli; EnPEC, Endometrial pathogenic E. coli; and NTEC, Necrotoxigenic E. coli.

The pathogenic strains of E. coli have acquired specific virulence factors that enable them to adapt to new niches and cause a wide variety of diseases (Kaper et al. 2004, Denamur et al. 2020). These include adhesion/colonization factors, toxins, and effectors enabling pathogenic strains to colonize sites such as the urethra and small intestine and affect various fundamental eukaryotic processes (Kaper et al. 2004; Tables 3 and 4). For example, pathogenic E. coli can adhere to host epithelial cells using adhesion molecules known as intimin (Dutta et al. 2015). Pathogenic E. coli are also known to induce inflammation in host cells, resulting in the production of inducible nitric oxide synthase by the host innate immune cells. This results in the accumulation of nitrate, which the pathogenic strains can utilize to produce energy (Kamada et al. 2013). UPEC strains are equipped with type I fimbriae, AfA/Dr adhesins and pyelonephritis-associated pili (PAP) that enable them to colonize and infect the urinary tract. NMEC and sepsis-associated E. coli are armed with the K1 polysaccharide capsule that facilitates their evasion from host complement-mediated killing (Pitout 2012, Leimbach et al. 2013). Differential utilization of nutrients is another way in which pathogenic strains of E. coli can overcome the resistance mounted by the commensal strains. For example, pathogenic EHEC can use the sugars galactose, mannose, ribose, and hexuronates as carbon sources, while commensal E. coli cannot (Fabich et al. 2008, Le Bouguénec and Schouler 2011). Besides, EHEC possesses the eut operon, which enables it to assimilate nutrients such as ethanolamine, whereas commensal E. coli lack this operon, and are therefore, unable to utilize ethanolamine (Perna et al. 2001, Fabich et al. 2008, Bertin et al. 2011).

Table 3.

Colonization and fitness factors (adapted from Kaper et al. (2004)).

Virulence factor	Pathotype/pathovar	Effect(s)
icsA/virG	EIEC	Nucleates actin filaments
Intimin	EPEC/EHEC	Adhesin, inducing TH1 response
Dr adhesins	DAEC/UPEC	Adhesin, binds to decay-accelerating factor and activates phosphatidylinositol 3-kinase, induces MHC class I chain-related gene A
P (Pap) fimbriae	UPEC	Adhesin, also induces cytokine expression
Colonization factor antigens	ETEC	Adhesin
S fimbriae	UPEC/NMEC	Adhesin
Bundle-forming pili (BFP)	EPEC	Type IV pili
Aggregative adherence fimbriae	EAEC	Adhesin
paa	EPEC/EHEC	Adhesin
toxB	EHEC	Adhesin
Efa-1/LifA	EHEC	Adhesin
Long polar filaments	EHEC/EPEC	Adhesin
saa	EHEC	Adhesin
ompA	NMEC/EHEC	Adhesin
Curli	Various	Adhesin, binds to fibronectin
ibeA/B/C	NMEC	Stimulates invasion
aslA	NMEC	Stimulates invasion
Dispersin	EAEC	Stimulates colonization; facilitates mucous penetration
K antigen capsules	MNEC	Antiphagocytic activity
Aerobactin	EIEC	Siderophore, iron acquisition
Yersiniabactin	Various	Siderophore, iron acquisition
ireA	UPEC	Siderophore, iron acquisition
iroN	UPEC	Siderophore, iron acquisition
chu (shu)	EIEC/UPEC/NMEC	Siderophore, iron acquisition
Flagellin	All	Motility, inducing cytokine expression through Toll-like receptors
Liposaccharides	All	Inducing cytokine expression through Toll-like receptors

Please see Table 2 for the full names of the pathovars/pathotypes.

Table 4.

Escherichia coli toxins and effectors (adapted from Kaper et al. (2004)).

Virulence factor	Pathotype/pathovar	Toxin class	Effect(s)
Heat-labile enterotoxin	ETEC	AB subunit/type II effector	ADP ribosylates and activation of adenylate cyclase, leading to ion secretion
Shiga toxin	EHEC	AB subunit	Depurination of rRNA, inhibiting protein synthesis and inducing apoptosis
Cytolethal distending toxin	Various	ABC subunit	DNase activity, blocks mitosin in G2/M phase
Shigella enterotoxin 1	EAEC/EIEC	AB subunit	Ion secretion
Urease	EHEC	ABC subunit	Cleaves urea to NH3 and CO3
EspC	EPEC	Autotransporter	Serine protease, cleavage of coagulation
EspP	EHEC	Autotransporter	Serine protease, cleavage of coagulation factor V
Haemoglobin-binding protease	ExPEC and APEC	Autotransporter	Degradation of haemoglobin to release haem/iron
pet	EAEC	Autotransporter	Serine protease; ion secretion and cytotoxicity
pic	UPEC, EAEC, and EIEC	Autotransporter	Protease/mucinase
sat	UPEC	Autotransporter	Vacuolation
sepA	EIEC	Autotransporter	Serine protease
sigA	EIEC	Autotransporter	Ion secretion
Cycle-inhibiting factor	EPEC and EHEC	Type III effector	Blocks mitosis in G2/M phase, resulting in the inactivation of cdk1
espF	EPEC and EHEC	Type III effector	Opens tight junctions and reduces apoptosis
espH	EPEC and EHEC	Type III effector	Modulates filopodia and pedestal formation
map	EPEC and EHEC	Type III effector	Disrupts mitochondrial membrane potential
tir	EPEC and EHEC	Type III effector	Nucleates cytoskeletal proteins, loss of microvilli and GAP-like activity
ipaA	EIEC	Type III effector	Actin depolymerization
ipaB	EIEC	Type III effector	Apoptosis, Interleukin-1 release and membrane insertion
ipaC	EIEC	Type III effector	Actin polymerization
ipaH	EIEC	Type III effector	Modulation of inflammation
ipgD	EIEC	Type III effector	Inositol 4-phosphatase and membrane blebbing
VirA	EIEC	Type III effector	Microtubule destabilization and membrane ruffling
stcE	EHEC	Type II effector	Cleavage of C1-esterase inhibitor and disruption of the complement cascade
hlyA	UPEC	Repeats-in-toxin (RTX) toxin	Cell lysis
ehx	EHEC	RTX toxin	Cell lysis
Cytotoxic necrotizing factors (1 and 2)	NMEC, UPEC, and NTEC	-	Altered cytoskeleton and necrosis
LifA/Efa	EPEC and EHEC	-	Inhibits lymphocyte activation and adhesion
Shigella enterotoxin 2	EIEC and ETEC		Ion secretion
Heat-stable enterotoxin a	ETEC	Heat-stable enterotoxins	Activating guanylate cyclase, leading to ion secretion
Heat-stable enterotoxin b	ETEC	Heat-stable enterotoxins	Ion secretion via increasing intracellular calcium
Enteroaggregative E. coli heat-stable enterotoxin	Various	Heat-stable enterotoxin	Activating guanylate cyclase, leading to ion secretion

Please see Table 2 for the full names of the pathovars/pathotypes.

Despite its designation as a separate genus, Shigella is classified as an intestinal pathogenic E. coli (InPEC) pathotype (InPEC encompasses Adherent Invasive E. coli (AIEC) and diarrheagenic E. coli). It possesses virulence traits and pathogenicity that closely resemble enteroinvasive E. coli (EIEC) and is, thus regarded as an EIEC pathotype (Pupo et al. 2000, Escobar-Páramo et al. 2003). Phylogenomic data support the classification of Shigella as an E. coli pathotype (Lan et al. 2004, The et al. 2016), even though its nomenclature has been retained solely for historical reasons and to avoid confusion in the clinical setting. Chaudhuri and Henderson (2012) advocate that the clinical and academic community working on E. coli/Shigella adopts a similar approach as was used in the classification of species such as Salmonella typhimurium, Salmonella enteritidis, and Salmonella typhi as Salmonella enterica subgroups so that the Shigella species are re-designated as subspecies within E. coli to avoid their neglect in E. coli studies.

Despite the importance of pathotyping to the epidemiology and pathogenesis of strains, the recent discovery of complex hybrid pathotypes has challenged some of the classical pathotype definitions (Denamur et al. 2020), as illustrated in the following examples.

The strain that caused an outbreak of foodborne illness in Germany in 2011 (Frank et al. 2011), which swept across most of Europe, killing 54 individuals and causing approximately 4000 infections including 900 cases of haemolytic uraemic syndrome (HUS), was, in fact, an EHEC-EAEC hybrid strain (Karch et al. 2012, Denamur et al. 2020) belonging to serotype O104:H4 and sequence type ST678. This strain combined virulence characteristics of EHEC (Shiga toxin production) and adherence typical of EAEC strains, despite lacking the Type III secretion and tir/intimin system. It also possessed virulence factors commonly found in ExPEC strains, such as yersiniabactin and aerobactin (iron acquisition factors) and demonstrated expanded-spectrum beta-lactamase (ESBL) resistance (Karch et al. 2012).

A hybrid clone belonging to serotype O80:H2 and sequence type ST301 (clonal complex 165) has emerged in France, Belgium, and Switzerland, capable of causing HUS and bacteraemia (Mariani-Kurkdjian et al. 2014, Soysal et al. 2016, Thiry et al. 2017, Nüesch-Inderbinen et al. 2018, De Rauw et al. 2019). This strain possesses all the virulence factors typical of EHEC strains, such as intimin, Shiga toxin production, and enterohaemolysin, yet it belongs to the phylogenetic group A, unlike other EHEC strains (Mariani-Kurkdjian et al. 2014). It possesses a large plasmid (> 100 kb), which encodes a resistance cassette, providing resistance characteristics to a wide range of antimicrobials, including cotrimoxazole, tetracyclines, streptomycin, and penicillin (Soysal et al. 2016, Cointe et al. 2020). The carriage of large plasmids is commonly associated with ExPEC strains (Peigne et al. 2009, Nicholson et al. 2016).

Escherichia coli are classified into eight major phylogenetic groups (phylotypes or phylogroups) linked with the ecological niche, lifestyle, and pathogenic potential of strains (Walk et al. 2009, Alm et al. 2011). Phylogroups B2, D, F, and H appear to be the most basal taxa, with phylogroup E emerging before phylogroups C, B1, and A, which are thought to be the most recently diverged lineages (Lu et al. 2016, Beghain et al. 2018). The most anciently diverged encompass mostly extraintestinal pathogenic strains. In contrast, the most recently diverged lineages span strains that are associated with life-threatening intestinal diseases such as dysentery and HUS (Picard et al. 1999, Escobar-Páramo et al. 2004a). Hybrid clones exhibiting characteristics of the B2 phylogenetic backbone typical of ExPEC strains and EPEC and STEC attributes have been recently reported in the literature, further blurring the lines of pathotype boundaries that have been used to define E. coli pathogens for so long. Isolates belonging to serotype O153:H10 and harboring eae (depicting an atypical EPEC-ExPEC hybrid pathotype) have been detected in meat, poultry farms, human diarrhoeagenic samples, and wildlife from northwest Spain (Díaz-Jiménez et al. 2020). Similarly, isolates belonging to serotype O137:H6 and ST2678 and positive for eae, bfpA, and stx2f genes—demonstrating typical EPEC-STEC hybrid pathotype—have been isolated in exotic psittacine birds (Gioia-Di Chiacchio et al. 2018). In addition, strains belonging to serotype O2:H4 and ST141 and demonstrating uropathogenic traits have been found with some characteristics of the EHEC pathotype (Bielaszewska et al. 2014, Gati et al. 2019). Also, an ST12 strain belonging to serotype O4:H1 was found to harbor the locus of enterocyte effacement pathogenicity island and the bundle forming pili protein (encoded by the bfp gene) typical of the EPEC pathotype but simultaneously causing diarrhoea and bacteraemia in the same patient (Kessler et al. 2015). Furthermore, some EPEC isolates that harbor the heat-stable enterotoxin produced by ETEC strains have been described (Dutta et al. 2015).

These examples amply demonstrate that pathotyping is limited in its capacity to adapt to new strains that fail to respect the currently utilized pathotype boundaries. However, it can be argued that the emergence of pathovars that exhibit traits uncharacteristic of their genetic background is not unexpected, given that pathotypes are usually defined based on horizontally acquired elements like plasmids and genomic islands. Thus, while the majority of a pathovar's membership may show similarity in pathogenic phenotype, horizontal genetic exchange would result in an elastic continuum within a pathovar's definition.

A further limitation lies in the use of negative criteria to identify pathogens. For instance, a strain is described as a typical EPEC if it presents the locus of enterocyte effacement but does not produce Shiga toxin. On the other hand, if a strain both lacks Shiga toxin and bfp, it is classified as an atypical EPEC (Kaper et al. 2004). As Robins-Browne et al. (2016) point out, ‘characterizing pathogens based on their lack of one or more virulence determinants may group several types of distantly related or unrelated bacteria together and cause some distinct pathogenic categories with uncharacterized virulence determinants to be overlooked’.

The taxonomy of Escherichia

In the premolecular era, several non-coli species were designated under the Escherichia genus based on DNA relatedness/hybridization and overall phenotypic similarity: Escherichiablattae (1973) (Burgess et al. 1973), Escherichiavulneris (1982) (Brenner et al. 1982a), Escherichiahermannii (1982) (Brenner et al. 1982b), and Escherichia fergusonii (1985) (Farmer et al. 1985).

Lawrence et al. (1991) utilized DNA sequences of the slowly evolving genes gap and ompA (encoding glyceraldehyde-3-phosphate dehydrogenase (GAPDH) and outer membrane protein 3A, respectively), representing conserved genes across 11 species of enteric bacteria, to map the phylogenetic relationships among the above mentioned five species. Their analysis revealed these species to be distantly related, rather than a monophyletic group. Apart from E. fergusonii, the other species were more divergent from E. coli than Salmonella. As Salmonella is traditionally considered most closely related to E. coli, the results justified the redesignation of E. blattae, E. hermannii, and E. vulneris in alternative genera (Lawrence et al. 1991).

Subsequently, in 1991, Albert et al. (1991) isolated a diarrhoeagenic isolate with EPEC-like phenotypic and genetic features from a 9-month-old girl with watery diarrhoea in Bangladesh, initially designated Hafnia alvei. Subsequent identification and characterization of five ‘H. alvei-like’ isolates by DNA–DNA hybridization, phenotypic characterization, and 16S rDNA sequencing led to their redesignation as Escherichia albertii (Huys et al. 2003). Members of this species were later found to be closely related to Shigella boydii serotype 13—a divergent lineage in Escherichia; the E. albertii/S. boydii serotype 13 lineage estimated to have diverged from a most-recent common ancestor circa 28 million years ago (Hyma et al. 2005).

Consequently, the taxonomy of the genus Escherichia has been modified with the reclassification of E. hermannii, E. blattae, and E. vulneris to other genera and the description of five cryptic clades within Escherichia (Walk et al. 2009, Priest and Barker 2010, Hata et al. 2016, Jain et al. 2018). Until recently, the genus was comprised of three named species: E. albertii, E. coli, and E. fergusonii and five cryptic clades, designated Escherichia clades I–V (Denamur et al. 2020). Of the three named species within the genus, E. albertii is the most divergent species, while E. fergusonii is closely related to Escherichia coli sensu stricto (Clermont et al. 2011).

The name Escherichia marmotae has recently been validly published for Clade V (Liu et al. 2015, 2019), although the species is not restricted to marmots (Ocejo et al. 2020, Gilroy et al. 2021). This builds on an earlier suggestion to classify clade V as a novel species and clades III and IV combined as a second novel species, based on digital DNA–DNA association and Average Nucleotide Identity (ANI) data (Walk et al. 2009; an ANI of 95% between two genomes is generally taken as the standard for the demarcation of prokaryotic species; Richter and Rosselló-Móra 2009, Jain et al. 2018). Furthermore, cryptic clade I strains possess the trademark virulence traits of E. coli, besides their potential to cause human infections (Steinsland et al. 2010, Clermont et al. 2011, Ingle et al. 2011). Thus, E. coli and Escherichia clade I are now designated as Escherichia coli sensu lato and the classic E. coli (designated phylogroups A–G) as E. coli sensu stricto (Clermont et al. 2013). Recently, Gilroy et al. (2021) assigned the species name Escherichia whittamii to clade II, in honor of the American bacteriologist, Thomas S. Whittam, for his contributions to the study of E. coli. van der Putten et al. have also recently proposed that the cryptic clades III and IV be named Escherichia ruysiae sp. nov (van der Putten et al. 2021).

Walk et al. (2009) estimated the lineages' divergence times that gave rise to each of the cryptic clades using a minimum evolution tree. Based on the assumption that E. coli split from S. enterica between 100 and 160 million years ago, they estimated that the Escherichia lineages shared a common ancestor between 48 and 75 million years ago. They suggested that E. albertii, E. fergusonii, Escherichia clade II, and clade V split between 38 and 75 million years ago, whereas E. coli, Escherichia clade I, clade III, and clade IV split between 19 and 31 million years ago (Walk et al. 2009).

As evidenced by earlier efforts at classifying E. blattae, E. hermanii, and E. vulneris, metabolic characteristics such as the utilization of specific carbon sources or the production/catabolism of certain biochemical compounds are insufficient to delineate strains into species and species groups accurately. First, multiple genes are often required for the expression of a particular phenotype (Lawrence et al. 1991). Second, convergent evolution in distantly related species confounds the delineation of species based on these phenetic characteristics. Similarly, DNA hybridization fails to consider the relatedness among congeneric species reliably (Lawrence et al. 1991).

The Genome Taxonomy Database Toolkit (GTDB-Tk), a tool for the automatic classification of draft bacterial and archaeal genomes, was published recently (Chaumeil et al. 2019). GTDB-Tk inputs genome assemblies in FASTA format and predicts the placements within domains based on identifying a set of 120 bacterial and archaeal marker genes and domain-specific reference trees. Then, species designations are computed using a GTDB reference tree, the relative evolutionary divergence parameter and ANI values. However, GTDB automatically assigns a random alphanumeric designation to novel species, which do not scale well to the increasing number of newly identified novel species and are often confusing and user-unfriendly (Gilroy et al. 2021). To address this gap, Pallen et al. (2020) recently developed an automated combinatorial approach to creating more than 1 million Linnaean binomials for Bacteria and Archaea.

The emergence of antimicrobial resistance in E. coli

Escherichia coli has long been used as an indicator bacterium for monitoring the faecal contamination of food and water and antimicrobial resistance (AMR) in enteric bacteria of animals and humans (Teuber 1999). For example, E. coli has been employed as the model organism in determining the level of AMR in bacteria from people in close contact with food animals, such as those who work at abattoirs and veterinarians—with the observation that such people harbor significantly higher levels of resistant E. coli compared to the overall community (Bongers et al. 1995). The ability to acquire and transfer AMR traits was recognized in E. coli as far back as the 1960s: 26%–61% and 50%–76% of human and pig multidrug resistant E. coli strains, respectively were found to transfer resistance to lab strains of E. coli K-12 in conjugation experiments (Datta 1969, Guinée et al. 1970, Linton et al. 1972, Gyles et al. 1977, Davies and Stewart 1978, Bourque et al. 1980, Nijsten et al. 1996).

According to the World Health Organization (WHO), the ever-increasing levels of AMR is one of the most significant threats to human health (www.who.int/entity/drugresistance/en), with substantial economic implications (Brown et al. 2017). Resistance to antimicrobials represents a daunting challenge to treating many infections, not least those caused by E. coli.

As a pathogen, E. coli is a prominent cause of urinary tract infections (UTIs), gastroenteritis, and bloodstream infections, among others; and as such, antibiotics are frequently applied to treat E. coli infections (Foxman 2010, de Kraker et al. 2013, Riley 2020). The use (and abuse) of antibiotics in treatment is linked with antibiotic resistance development (Webber and Piddock 2005, Ventola 2015). Microorganisms exhibit a natural ability to resist antimicrobials’ action—a phenomenon referred to as intrinsic resistance (Cheng et al. 2016). In addition, AMR can arise due to a gene mutation or the acquisition of resistance determinants via horizontal gene transfer (de la Cruz and Davies 2000). Horizontal gene transfer can take the form of free DNA uptake (transformation), plasmid-mediated transfer of resistance gene traits (conjugation), or via phage-mediated transfer (transduction; Burmeister 2015, Lee 2019).

Escherichia coli frequently displays resistance to multiple classes of antimicrobials (MDR)—with observed rates of MDR identified among strains that cause UTI and bacteraemia exceeding 50% (Croxall et al. 2011, Alhashash et al. 2013, de Kraker et al. 2013). The occurrence of MDR in E. coli has been focused on a small number of widely dispersed clones, mediated by ESBL and carbapenemase-encoding plasmids (Dunn et al. 2019). Worryingly, the rates of E. coli extraintestinal infections such as bacteraemia and UTIs have been rapidly increasing in recent years due to an increase in the number of antibiotic-resistant infections caused by ‘superbugs’ such as the ST131 clone (Croxall et al. 2011, Banerjee and Johnson 2014, McNally et al. 2019).

Although several strains have been linked with the global emergence and dissemination of MDR in E. coli (e.g. strains belonging to ST88, ST410, ST648, ST405, and ST73), none equal ST131 in their extent (Dunn et al. 2019). The MDR ST131 clone belongs to a sublineage designated clade C which arose from two predominantly drug-susceptible clades by acquiring large MDR plasmids which confer ESBL and metallo-beta-lactamases encoding resistance to carbapenems and cephalosporins up to third generation; as well as point mutations leading to fluoroquinolone resistance (Ben Zakour et al. 2016, Stoesser et al. 2016).

McNally et al. (2019) recently analysed an extensive collection of more than 1000 ST131 strains. They reported that the acquisition of colonization and fitness factors and the accumulation of genes encoding dehydrogenase enzymes involved in anaerobic metabolism account, at least in part, for the successful expansion of this clone (McNally et al. 2019). These results indicate hyper-resistant clones’ ability to outcompete resident nonpathogenic strains of the same species, and thus facilitate their long-term colonization in the gut (Dunn et al. 2019, Benz et al. 2021). The fact that MDR strains such as ST131 are well-suited to colonize the gut competitively suggests that AMR traits can be readily transferred from resistant strains to co-colonizing susceptible strains, with the potential to hamper the treatment of future infections caused by such strains.

Accumulating data shows that ExPEC are frequently isolated from diseased companion animals and livestock—highlighting the potential for zoonotic as well as anthroponotic transmission (Achtman et al. 1986, Johnson et al. 2008a,b, 2009, Ewers et al. 2010, Zogg et al. 2018). High rates of AMR among E. coli isolates from livestock and poultry have been documented, linked with the agricultural use of antimicrobials (Bennani et al. 2020). The use of antimicrobials as growth promoters in poultry feed was banned in Europe in 2006 (Castanon 2007), with a resultant sharp drop in AMR rates among isolates from livestock and poultry (Bennani et al. 2020). However, this may be less well controlled in other parts of the world, particularly in low-to-middle income countries. Worryingly, the usage of antimicrobials in developing countries is likely to increase as intensive farming practices are adopted (Van Boeckel et al. 2015).

Several studies have investigated the link between antimicrobial usage and AMR in animals and AMR in humans. In 2015, a systematic review (Lazarus et al. 2015) of the published literature to quantify the zoonotic transfer of ESBL-encoding extraintestinal E. coli infections found that six studies established the zoonotic transfer of AMR by whole bacterial transmission using molecular methods, in particular, via poultry in the Netherlands. A further 13 molecular studies suggested the mobile-genetic element-mediated transfer of AMR from resistant organisms in animals to susceptible strains in humans, while four observational epidemiological studies inferred the zoonotic transfer of AMR to humans.

Although the authors cautioned that their conclusions might not be geographically generalizable, it appears that a proportion of human extraintestinal infections caused by ESBL-encoding strains originate from food-producing animals. Further, E. coli strains from commercial broilers have recently been reported to share resistance profiles with strains recovered from human extraintestinal infections (Borges et al. 2019). This is worrying and highlights the need for further investigation, especially with whole-genome sequence-based studies (Webber and Piddock 2005).

Escherichia coligenomics

Frederick Blattner first conceived the E. coliK-12 genome sequence project in 1983 (Blattner 1983); however, due to funding and technological challenges, it was not until 15 years later that the project (which was 6-year-long) was finally completed (Blattner et al. 1997). ‘E. coli was the obvious choice for a sequencing effort’, noted Frederick Neidhardt in Pennisi (1997), since more was known about the organism than any other (Hayashi et al. 2006), and so much information had been gleaned from studies with E. coli, that, as Blattner stated, ‘Figuring out the microbe's genetic code would help integrate all those years of study’ (Pennisi 1997). Blattner and colleagues submitted the final 2.0 Mb of the 4.6 Mb E. coli genome to GenBank on 16 January 1997 (Pennisi 1997). This was closely followed by the deposition in GenBank of an incomplete genome sequence of the closely related strain, W3110, 7 days later (Pennisi 1997) (the complete genome sequence of which was published in 2006 (Hayashi et al. 2006)). Based on atypical codon usage and base composition, Lawrence and Ochman (1998) inferred 18% of the K-12 genome to have arisen from horizontal transfer (earlier studies had shown that horizontally acquired genes exhibited atypical codon usage, base composition, and dinucleotide frequencies; Muto and Osawa 1987, Lawrence and Ochman 1997). This estimate was later revised to 24.5%, using improved methods (Lawrence and Ochman 2002).

Next, the genome sequences of two EHEC O157:H7 isolates were swiftly published, expanding the number of E. coli genomes for comparative studies. First came the EDL933 strain isolated from Michigan ground beef connected to the 1982 multistate outbreak by Perna et al. (2001), then RIMD 0508992—the strain that caused a large outbreak in 1996 in Sakai city in Osaka prefecture, Japan, involving at least 6000 school children (Watanabe et al. 1996)—the latter sequenced by Hayashi et al. (2001). Comparisons with K-12 revealed a shared sequence of 4.1 Mb, representing a common chromosomal ‘backbone’ of E. coli, with the remaining sequence comprising strain-specific clusters varying in size, encoding putative virulence factors, prophages, and prophage-like elements (Hayashi et al. 2001, Perna et al. 2001). These analyses provided evidence of extensive horizontal gene transfer, with the description of many ‘K’ and ‘O’ islands—depicting introgressed DNA present only in K-12 but not in O157:H7 or only in O157:H7, respectively. The existence of a shared E. coli backbone was confirmed by a three-way comparison of the K-12 and O157:H7 strains with the genomic sequence of the third E. coli strain to be completed: the ExPEC strain CFT073 (Welch et al. 2002). A surprising finding was that all three strains shared only 39.2% of the combined nonredundant set of proteins. The role of horizontal gene transfer in the pathogenic strains' evolution was also highlighted by the presence of several pathogenicity islands exhibiting atypical codon usage interrupting the common backbone (Welch et al. 2002).

With the availability of more E. coli genome sequences, the core and accessory genome concept was defined to represent a conserved set of roughly 2200 genes common to all E. coli strains and strain-specific sequences respectively (Rasko et al. 2008, Touchon et al. 2009, Chaudhuri et al. 2010). The core genomic sequence lends itself as a useful tool for the phylogenomic comparison of isolates, provided the effect of homologous recombination (estimated to affect about a 10th of the E. coli core genome; Mau et al. 2006) is accounted for (Chaudhuri et al. 2010).

The pan-genome concept

The concept of a ‘pan-genome’ was proposed by Tettelin et al. (2005), during an attempt to utilize genome sequence information from Group B streptococcus to predict proteins that might be exposed on the organism's surface and could be exploited as vaccine candidates (Tettelin et al. 2005, Tettelin and Medini 2020). By this concept, the pangenome of each bacterial species is defined by three distinct components: namely, its core genome, representing the genes found in each isolate of the species; the accessory genome, depicting the genes present in several but not all isolates of the species, and strain-specific genes detected in one isolate only. Through genomic comparisons of nineteen GBS isolates, Tettelin and colleagues uncovered the first evidence that closely related isolates differed significantly in their gene content. A single isolate of a particular species was insufficient to capture the species' genome (Tettelin et al. 2005, Tettelin and Medini 2020).

In a study that compared 61 sequenced genomes of E. coli, Lukjancenko et al. (2010) predicted a pangenome comprised of 15 741 gene families, with only 993 (6%) of the gene families present in every genome (the core genome)—indicating an accessory genome of more than 90% in E. coli. This equates to an accessory or variable gene content of approximately 4/5th of any given E. coli genome (Lukjancenko et al. 2010). These results corroborate those reported by Rasko et al. (2008) and others before them (Ochman and Jones 2000, Fukiya et al. 2004). Rasko et al. identified a pangenome comprised of more than 13 000 genes via comparison of seventeen E. coli reference genome sequences encompassing human commensal and distinct clinical groups of E. coli and a core gene set of about 2200 genes conserved in all isolates. It has become apparent that the more E. coli genomes are sequenced and compared, the more the pangenome continues to increase—what has been referred to as an ‘open pangenome’ (Rasko et al. 2008)—and the core genome shrinks (Robins-Browne et al. 2016). Thus, the accessory genome content contributes crucial insights which, coupled with single nucleotide polymorphisms (SNPs) in the core genome, can be employed to track the evolutionary history of natural isolates, as has been recently demonstrated by McNally et al. (2016) and Decano and Downing (2019). The accessory genome content has arisen from repeated gene acquisition and the contemporaneous loss of sequences is thought to account for the distinctions between divergent lineages within the same species (Ochman and Jones 2000). They include genes encoding virulence determinants, bacteriophages, virulence factors, and acquired AMR determinants (Robins-Browne et al. 2016).

Whole genome sequencing (WGS) offers many advantages for the diagnosis and understanding of the pathobiology of E. coli strains, in that it is possible to predict most of the pathotypes of E. coli based on the presence of well-recognized virulence factors, as well as elucidating the full array of virulence factors possessed by individual strains within a particular pathotype. Furthermore, typing schemes that combine several genes within the core genome (e.g. core genome MLST available on platforms such as EnteroBase; Zhou et al. 2020) and others that would incorporate the accessory genome are expected to become the mainstay of E. coli analysis, particularly as the accessibility of whole-genome data continues to increase (Robins-Browne et al. 2016).

The ecology of E. coliin the gut

Microbial ecology is the study of the diversity, distribution, and abundance of microorganisms, and how microorganisms interact with each other and their environment to generate and preserve such diversities (Xu 2006). There are two areas of focus that have encapsulated microbial ecological studies to date: namely, the examination of microbial diversity—'who is there,’ i.e. the identification and characterization and estimation of abundance across a variety of niches; and the study of microbial activity, i.e. what microorganisms are doing and how they are doing it—including their biotic and abiotic interactions and how they impact the observed diversity and the ecosystem.

Members of the gut microbiota have co-diversified with their hosts over millions of years and exist in a mutualistic relationship with their host, in which the gut microbiota carry out vital functions for their hosts and in return occupy a nutrient-rich environment (Round and Mazmanian 2009, Nicholson et al. 2012). The gut microbial communities' composition is thought to be influenced by factors such as diet, physiology of the gut, host phylogeny, and diet (Ley et al. 2008, Muegge et al. 2011, Wu et al. 2011, David et al. 2014). A healthy gut microbiota plays critical roles in developing the host immune system, and is required for homeostasis in adult life (Hooper and Macpherson 2010, Flint 2012). For example, the gut microbiota cells help maintain the balance between host metabolism and the immune system and in the large intestine, metabolize the indigestible components of the diet (Bäckhed et al. 2005, Turnbaugh et al. 2009). The gut microbiota also detoxify toxic products and serve as a barrier against the colonization of opportunistic pathogens (termed as colonization resistance; Round and Mazmanian 2009, Claus et al. 2016). The mechanisms by which the resident intestinal microbiota elicit resistance against the colonization and invasion of pathogens include: the direct competition for nutrients, the modification of metabolites such as bile salts and short-chain fatty acids that render them toxic to invading pathogens, the alteration of pH and oxygen tension, the induction of host antimicrobial peptides, the expression of a dense mucous, IgA and cellular immunity, and direct attacks through the production of bacteriocins or Type IV secretion (Brown and Valiere 2004, Kamada et al. 2013).

Host-directed pathways act to break down complex protein, carbohydrate, and fat in ingested food in eukaryote hosts into simple forms, thereby facilitating the absorption of nutrients. In addition, members of the microbiota contribute to the catabolic process in producing short-chain fatty acids utilized by the host and support the growth of the bacterial community (Sorbara and Pamer 2019). The members of the microbiota have evolved to maximize the available nutrients such that under homeostatic conditions, an exogenous species needs to overcome the intense competition for limited substrates to engraft themselves in the available niches. Resident E. coli use sugars such as fucose, mannose, galactose, ribose, arabinose, and N-acetyl glucosamine in the mucous layer—which are required by pathogenic strains such as E. coli O157—and in this way, mount colonization resistance to invading pathogens (Maltby et al. 2013). Besides, commensal E. coli employ the preferential utilization of nutrients to out-compete exogenous strains. For example, resistance to EHEC by commensal E. coli is due to the competition for nutrients such as amino acids and organic acids (Chang et al. 2004, Fabich et al. 2008, Leatham et al. 2009, Maltby et al. 2013).

Competition for trace metals also appears to be significant in colonization resistance, as demonstrated in murine studies, where the Nissle 1917 E. coli strain was found to protect against Salmonella infection by competing for iron (Deriu et al. 2013).

However, intestinal microorganisms constitute a persistent invasion threat, considering their vast numbers and the large intestinal surface area (Hooper and Macpherson 2010). Emerging evidence suggests that dysbiosis—defined as a shift in the composition of the intestinal microbiota, and thus an alteration of the relationship between the host and the gut microbiota—is linked to the development of various diseases, such as inflammatory bowel disease, obesity, allergy, autoimmune disease, and irritable bowel syndrome (Round and Mazmanian 2009, Codling et al. 2010, Kadooka et al. 2010, Kang et al. 2013, Miyoshi et al. 2014, Quince et al. 2015, Lynch and Boushey 2016, Tanaka and Nakayama 2017).

Estimates show that anaerobic bacteria outnumber E. coli anywhere from 100:1 up to 10 000:1 (Berg 1996). The prevalence of E. coli in the various hosts they colonize varies widely (0%–100%), being influenced by host characteristics such as body size, microbiota, diet, and digesta retention times (Tenaillon et al. 2010). Over 90% of humans carry E. coli, while about 25%–56% of wild mammals appear to be colonized by the organism (Mitsuoka et al. 1976, Gordon and Cowling 2003, Penders et al. 2006, Tenaillon et al. 2010, Lescat et al. 2013). The prevalence rate in human-associated animals (such as chickens and cats) is estimated to be above 60% (Lescat et al. 2013). We recently reported a prevalence of 56% from a range of nonhuman primate species with varying degrees of contact with humans (Foster-Nyarko et al. 2020b).

In the gut, E. coli reside in the mucous covering of the epithelial cells and is shed with the degraded mucus components into the intestinal lumen and subsequently excreted in faeces (Poulsen et al. 1994, Poulsen et al. 1995). Human faeces typically contain between 10²and 10⁹ colony-forming units (cfu) of E. coli per gram (Slanetz and Bartley 1957, Bettelheim et al. 1972, Mitsuoka et al. 1976, Penders et al. 2006), while an estimated 10⁴–10⁶ cfu can be detected in the faeces of domestic animals (Slanetz and Bartley 1957). Data on the quantity of E. coli in stools of wild animals is, however, lacking. As first observed by Escherich (1886), E. coli is one of the first bacterial colonizers of the infant gut (Mueller et al. 2015), achieving concentrations of up to 10⁹ cfu per gram of the stools of infants (Mitsuoka et al. 1976, Penders et al. 2006). Subsequently, anaerobic members of the microbiota expand and dominate the gut (Syed et al. 1970). Given that E. coli is a facultative anaerobe, its ability to utilize oxygen probably helps create an anaerobic environment favoring the blooming of strict anaerobes (Mueller et al. 2015). As a gut microbiota member, E. coli produces vitamin K and mounts resistance against colonization by pathogens (Suvarna et al. 1998, Richter et al. 2018). Thus, E. coli exists in a mutualistic relationship with the human host, although it is mostly described as a commensal (Blount 2015, Martinson and Walk 2020).

Given that most E. coli reside innocuously in the gut, an important ecological question that has been plaguing microbiologists is what makes E. coli an occasionally devastating pathogen (Lescat et al. 2013, Martinson and Walk 2020)? To address such questions requires an enhanced understanding of the ecology of the organism as a commensal. However, nonpathogenic E. coli have been traditionally underrepresented in ecological studies of this species (Martinson and Walk 2020). More studies exploring the populations of resident or nonpathogenic E. coli within and between hosts and how these populations vary over time are needed to shed light on this evolutionary puzzle.

Humans are exposed to E. coli through multiple routes (Reeves et al. 2011, Akoachere et al. 2014, Nwankwo et al. 2014, Pal et al. 2015, Ercumen et al. 2017, Aijuka et al. 2018, Derakhshandeh et al. 2018, Abia and Ubomba-Jaswa 2019, Toombs-Ruane et al. 2020, Yar 2020): such as the consumption of contaminated food and water; through fomites, e.g. on bank coins and notes (particularly in the cracks of creased notes) and cell phones; pets and domestic animals and the environment. This high level of exposure is reflected in reports of more than one strain in normal stools (Sears et al. 1950, Sears and Brownlee 1952, Wiedemann et al. 1971, Bettelheim et al. 1972, Shooter et al. 1977, Chen et al. 2013, Dixit et al. 2018, Richter et al. 2018).

A total of two plausible theories explain the fate of swallowed strains (Sears et al. 1950, Sears and Brownlee 1952, Bettelheim et al. 1972). The displacement theory suggests that newly ingested strains may fail to establish themselves, in which case they are voided out, or if they succeed in establishing themselves, will displace the ‘resident’ strains present. Alternatively, the dominant–minor strain theory posits that freshly ingested strains do not replace the established strains within the gut but co-exist as minority or transient strains, albeit in small numbers and may be detected from time to time in the stool along with the dominant strain.

Experimental studies with the Nissle 1917 strain have shown that not all strains are equal in their propensity to establish themselves following immigration (Tannock et al. 2011). Evidence suggests that repeated exposure may facilitate immigration and the establishment of strains (Johnson and Clabots 2006, Johnson et al. 2008a). Host-related factors appear to contribute to the pattern of phylogroup distribution in the gut. In human studies, phylogroups E and F have been found to encourage a more heterogeneous population of E. coli following their establishment in the gut. Conversely, the establishment of phylogroup B2 strains in the gut appears to limit the immigration and establishment of other phylogroups (Gordon et al. 2015). Also, locale-specific differences in the distribution of E. coli phylotypes have been reported, linked with differences in climate and diet (Duriez et al. 2001, Escobar-Páramo et al. 2004b). For instance, strains belonging to phylogroups A and B1 are often prevalent in stools sourced from people living in developing countries. On the other hand, stools from people in developed countries are frequently dominated by E. coli belonging to phylogroups D and B2 (Skurnik et al. 2008, Massot et al. 2016).

The co-existence of multiple strains in a single host raises an essential question about the factors that govern residency in the gut, i.e. how incoming strains overcome the colonization resistance posed by the existing E. coli population. The current body of evidence suggests the following.

Freter's successful competition hypothesis

Freter theorized that successful colonization occurs due to successful competition for nutrients (Freter et al. 1983a,b, Pereira and Berry 2017). Accordingly, the gut microbiota composition is determined by several limiting substrates, which different members of the microbiota can utilize with variable efficacy. Conway and colleagues demonstrated this principle in E. coli strains (Escherichia coli strains HS and Nissle 1917 vs. E. coli O157:H7) using carbohydrate metabolism in a mouse model (Maltby et al. 2013, Conway and Cohen 2015). This theory is in tandem with Gause's exclusion principle, which precludes two organisms’ co-existence if they share the same limiting resource (Hardin 1960). Iron competition appears to influence the colonization of resident strains, as demonstrated by studies that found strains lacking siderophore genes to have a reduced ability to establish themselves in mouse models, compared to wild-type strains. Conversely, resident strains in the human gut have been found to encode siderophores (Nowrouzian et al. 2001), signaling their potential contribution to successful colonization. The mechanisms by which two organisms exclude each other from a particular niche might be through either direct (e.g. through bacteriocin production or phage to damage or kill competitors) or indirect (e.g. through passive resource utilization) competition (Ghoul and Mitri 2016, Pereira and Berry 2017, Granato et al. 2019).

Efficient utilizers

Some colonizers thrive because they utilize available nutrients much more efficiently than others who use the same nutrients. In murine studies, mice fed with E. coli MG1655, nonmotile flhD mutants were found to persist in stools collected 3 days postfeeding and dominate the population (roughly 90%) by day 15. The mutants colonized better than the wild-type parent strain and grew in the caecal mucus faster than the wild-type counterparts (Leatham et al. 2005, Fabich et al. 2011). Further analyses using high-throughput genomic approaches revealed that the flhD mutants possessed an enhanced ability to oxidize several carbon sources because the loss of FlhD conferred an increased expression of genes involved in carbon and energy metabolism. (The flhDC operon encodes the FlhD4C2 regulatory complex, which has been shown to negatively regulate the genes involved in galactose transport and the citric acid cycle while positively regulating the genes involved in ribose transport; Gauger et al. 2007, Fabich et al. 2011).

Restaurant hypothesis

Leatham-Jensen et al. (2012) have described how polysaccharide-degrading anaerobes break down polysaccharides into sugars, which they serve to E. coli cells within a shared biofilm (Ng et al. 2013). This is an example of ‘syntrophy’—an association where one organism feeds on the nutritional products of another (Dolfing 2014)—as has been observed between Bacteroides ovatus and Bacteroidesthetaiotaomicron (Tuncil et al. 2017). The biofilms that feed these E. coli strains are referred to as ‘restaurants’. Many such restaurants are thought to exist, proposed to comprise a mix of different commensal strains, each serving different nutrients to the commensal E. coli strains residing therein (Conway and Cohen 2015).

Different nutritional requirements

This hypothesis suggests that pathogenic strains may utilize other nutrients than those of the commensal residents. For example, it has been shown that E. coli strain HS, E. coli Nissle 1917, E. coli MG1655, and E. coli EDL933 utilize unique metabolic niches in the mouse intestine (Fabich et al. 2008, Maltby et al. 2013, Conway and Cohen 2015). Escherichia coli HS uses six out of the 12 sugars available in the mucous layer, namely, arabinose, galactose, gluconate, lactose, ribose, and N-acetylglucosamine. Escherichia coli MG1655 uses five sugars, namely, fucose, arabinose, gluconate, N-acetylneuraminate, and N-acetylglucosamine. Escherichia coli Nissle 1917, on the other hand, utilizes the following seven sugars: fucose, galactose, arabinose, N-acetylglucosamine, gluconate, N-acetylneuraminate, and mannose. Lastly, E. coli EDL933 also uses a unique selection of carbon sources: galactose, arabinose, hexuronates, N-acetylglucosamine, mannose, sucrose, and ribose.

Different biogeographical niches

Pathogenic strains of E. coli may colonize different biogeographic regions of the gut from commensal strains. Scientists from the University of Oklahoma have shown that E. coli EDL933, a strain of enterohaemorrhagic E. coli O157:H7 colonizes a different niche in the mouse intestine to human commensal strains (Leatham et al. 2009, Maltby et al. 2013). In the bovine host (where E. coli O157:H7 commonly colonizes the gut innocuously), studies in a 12-month-old naturally colonized steer revealed a unique tropism of E. coli strain O157:H7 for the rectal mucosa adjoining the recto-anal junction (Naylor et al. 2003).

Hierarchical nutrient utilization

Here, closely related bacterial species may co-colonize the gut by hierarchical utilization of similar nutrients, as has been demonstrated between two generalists, B. ovatus and B. thetaiotaomicron (Tuncil et al. 2017). Thus, direct competition for substrates is excluded by each species utilizing the available glycans with differing priorities (i.e. order and speed of consumption).

Exploiting colonization resistance for therapy

Antibiotic treatment has significantly reduced mortality and morbidity associated with potentially life-threatening diseases, and thus saved millions of lives (Pamer 2016). However, broad-spectrum antibiotics rarely target pathogens alone and concomitantly result in deleterious effects on the commensal bacterial populations, resulting in increased susceptibility to infections due to alteration of the host microbiota (dysbiosis; Dethlefsen et al. 2008, Buffie et al. 2012, Buffie and Pamer 2013, Ravi et al. 2019, Ribeiro et al. 2020). This problem is compounded by the fact that many pathogens have become increasingly antimicrobial resistant. Consequently, the administration of live bacteria (probiotics) to restabilize the altered microbiota and restore the colonization resistance conferred by the resident microbiota has increasingly been the focus of intense research (Buffie et al. 2012, Buffie and Pamer 2013, Pamer 2016, Morrow and Wischmeyer 2017). This heightened interest has been fuelled by the success of faecal transplantation in patients suffering from diarrhoea associated with Clostridiodes difficile infection (Gough et al. 2011, Youngster et al. 2014, Manges et al. 2016).

Probiotics are defined as living microbes of human origin that can colonize host sites such as the gut and oropharynx when ingested in adequate quantities, and thus deliver benefits to the host (Hill et al. 2014, Morrow and Wischmeyer 2017). Probiotic strains are attractive as preventive measures against gut infections. Bacterial candidates that have been proposed to promote high-level colonization resistance to infection include members of the lactate-producing genera, e.g. Lactobacillus (Gorbach 2000) and Bifidobacteria (Sivan et al. 2015)—although evidence of their effectiveness in health in humans, reducing infections or fostering longevity is sparse (Pamer 2016).

Escherichia coli Nissle 1917 (also known as Mutaflor) is one of the most extensively researched probiotic strains globally (Jacobi and Malfertheiner 2011). Alfred Nissle observed that this strain ameliorates ulcerative colitis (Nissle 1951), following his isolation of the strain in 1917 from the stool of a soldier who did not suffer from diarrhoea like his comrades did during the First World War (Nissle 1959, Nissle 1961). The available data shows that Mutaflor possesses many traits that favor its suitability as a probiotic strain. For example, the strain forms a biofilm well and, in this way, outcompetes numerous pathogenic and nonpathogenic strains such as EPEC and ETEC (Hancock et al. 2010). The complete genome sequence information revealed that Mutaflor contains several fitness traits such as adhesion factors and an array of iron uptake systems that enable it to outcompete other bacteria and block the adherence and invasion of pathogenic strains (Lodinová-Zádniková and Sonnenborn 1997, Boudeau et al. 2003, Altenhoefer et al. 2004). The strain also boosts intestinal barrier function and protects against epithelial disruption by EPEC (Zyrek et al. 2007). Moreover, E. coli Nissle 1917 induces the release of beta-defensin-2 (an antimicrobial peptide) from human epithelial cells, thus eliciting a broad antimicrobial response against both Gram-positive and Gram-negative bacteria and fungi and viruses and, thus inhibits invasion and colonization of other bacteria (Wehkamp et al. 2004).

Accumulating evidence suggests that probiotics comprised of a combination of strains are effective, owing to the symbiosis among strains—including those from different genera (Timmerman et al. 2007). An example is the VSL#3 multistrain probiotic consortium—comprising Eubacterium faecium, Streptococcus thermophilus, Lactobacillus acidophilus, Bifidobacterium breve, Bifidobacterium infantis, Lactobacillus delbrueckii subspecies bulgaricus, Bifidobacterium longum, Lactobacillus plantarum, and Lactobacillus casei—which has been shown to treat ulcerative colitis effectively (Venturi et al. 1999, Timmerman et al. 2007). Researchers working with E. coli EDL 933 found a combination of two strains—E. coli Nissle 1917 and E. coli HS, both promising probiotic candidates—can overlap nearly all the nutrient requirements of E. coli EDL 933 (Meador et al. 2014). However, the same combination of commensals could not prevent colonization by uropathogenic strain E. coli CFT073 and the enteropathogenic strain E. coli E2348/69. Different pathogenic strains, thus occupy distinct nutrient niches within the gut microbiome.

As interest in probiotics heightens, there is the need to understand the mechanics of how potential probiotic strains compete or co-exist with the commensal bacteria in the gut. We do not yet fully understand how probiotic strains interact with the usual residents of the gut. Previous attempts at unravelling the interactions between pathogens and commensals in the gut have been mainly carried out using well-characterized reference strains within in vitro environments. It will be desirable to test these hypotheses with the ‘real world’ strains in the natural habitat (i.e. the gut).

Within host bacterial diversity of E. coli

As previously discussed, continuous exposure to E. colivia multiple routes contributes to the turnover of strains in the vertebrate gut, thus resulting in considerable diversity among the population of E. coli that exists within a single individual and between different hosts. Besides immigration events, the diversity of E. coli is enriched by within-host evolution events (Didelot et al. 2016).

Members of a bacterial community undergo changes as they interact with each other and their hosts or environment. The sources of novel variation that account for within-host evolution may arise from the following.

Point mutations

These involve the substitution, insertion or deletion of a single nucleotide. Point mutations represent the smallest evolution unit (Bryant et al. 2012). Reeves et al. reported a within-host point mutation of approximately 1 per year per genome (corresponding to 1.44 × 10^–7 accumulation rate per site per year) for E. coli (Reeves et al. 2011), while Lee et al. estimate a 2.54 × 10^–10 mutation rate per site per generation (Lee et al. 2012).

Insertions and deletions (indels) and genomic rearrangements [466, 467]

Phylogenetic analysis of the O104:H4 strain that caused the large outbreak of gastroenteritis and HUS in Germany in 2011 revealed that the acquisition of a Shiga toxin 2-encoding prophage and an ESBL CTX-M-15-encoding plasmid contributed to its emergence [468–471]. Similarly, the acquisition of virulence gene-encoding plasmids into many ancestral Shigella spp. was crucial to the evolution of Shigella as human pathogens (Gati et al. 2019).

Bacterial genomes can also undergo rearrangements during DNA recombination, replication, or error-prone DNA repair (Periwal and Scaria 2015, Rodgers and McVey 2016). Genomic rearrangements alter chromosomes or large chromosomal regions. They involve the processes of deletion, duplication, insertion, inversion, or translocation. Homologous recombination leads to the reassortment of genes between chromosomal pairs; however, the genome's arrangement remains unchanged. Other forms of recombination may result in rearrangements of genomic DNA. DNA rearrangements contribute to gene expression and function and may contribute to genetic diversity in E. coli (Iguchi et al. 2006, Periwal and Scaria 2015). In long-term evolution experiments involving 12 E. coli populations propagated for more than 25 years in glucose-restricted environments, large-scale genomic rearrangements were found to occur in the same loci in at least two populations and contributed to genomic plasticity (Raeside et al. 2014). Moreover, most rearrangements were observed to confer higher fitness, highlighting the significant influence of large-scale rearrangements in genomic evolution.

Other factors that shape within-host diversity include genetic drifts: random processes by which allelic frequencies change over time due to birth and death of individuals within the population (Charlesworth 2009, Kuo et al. 2009), as well as by both purifying and diversifying selection (Didelot et al. 2016).

Recombination and horizontal gene transfer (HGT)

HGT involves the uptake of extracellular DNA (transformation), cell-to-cell transfer of genetic material via surface appendages (conjugation), or viral import (transduction) from genetically distant relatives. The acquisition of genetic sequences from unrelated organisms via HGT results in faster diversification of the genome compared to point mutation alone (Ochman et al. 2000)—consequently termed ‘evolution in quantum leaps’ (Hacker and Carniel 2001). Fragments of the chromosomal genome can be replaced with homologous sequences from another cell through homologous recombination, which plays a significant role in the evolution of E. coli (Didelot et al. 2012). In particular, the presence of mixed infection facilitates homologous recombination by providing material for import in the chromosome—as has been exemplified in Helicobacter pylori, where evolution is noted to accelerate up to a 100-fold in the presence of mixed infections (Kennemann et al. 2011, Didelot et al. 2013, Cao et al. 2015). Thus, homologous recombination is a strong driver of within-host evolution. Alternatively, nonhomologous sequences can be gained and incorporated into the genome—a phenomenon, i.e. compensated by genome degradation (Andersson and Andersson 1999).

Hypermutators

Besides spontaneous mutations, genetic variation can arise due to the breakdown of DNA repair mechanisms such as the mismatch repair system, producing bacteria with decreased replication fidelity (Mather and Harris 2013). In E. coli, the loss of mismatch repair is known to increase mutation rates by up to 200-fold (Marinus 2010). Lee et al. (L2012) investigated the accumulation of spontaneous mutations over thousands of generations in wild-type E. coli compared to a mismatch repair-defective variant with the mutL gene encoding MutL, one of the key enzymes required for mismatch repair, deleted. In their experiment, the mutant (MutL^–) strain showed a 233-fold increase in mutation rate to Nal^R (resistance to nalidixic acid) compared to the wild type.

In particular, hypermutation is advantageous during infection when the ability to adapt quickly can facilitate evasion of host immune defences and antimicrobial therapy (Mather and Harris 2013). Hypermutator strains have been identified during an MRSA outbreak in a neonatal intensive care unit (Köser et al. 2012) and among Staphylococcus aureus isolates from cows suffering from benign forms of mastitis (Wang et al. 2013). Strikingly, in one study where cystic fibrosis patients with chronic respiratory problems were followed up over 38 years, nearly half of them harbored hypermutator Pseudomonas aeruginosa isolates (Marvig et al. 2013). Of note, the accumulation of SNPs via hypermutation can hamper the analysis of transmission and result in inaccurate conclusions (Mather and Harris 2013). In particular, the use of SNP thresholds to identify and delineate transmission lines could potentially exclude hypermutator strains or the inclusion of such strains may bias results (Mather and Harris 2013).

Investigating within-host bacterial diversity

To study within-host bacterial evolution in detail requires the application of whole-genome sequencing to multiple clinical samples taken from an individual host (these samples may be collected longitudinally or simultaneously from a single body site or several sites) or multiple isolates or genomes derived from a single clinical sample.

Sampling a single colony does not sufficiently capture the E. coli strain diversity in stools, as established by several studies (Bettelheim et al. 1972, Shooter et al. 1977, Dixit et al. 2018, Richter et al. 2018). Lidin-Janson et al. (1978) estimated that sampling five colonies enables detecting dominant genotypes at a > 99% probability. At the same time, Schlager et al. (2002) showed that sampling 28 colonies provide a > 90% chance of detecting minor genotypes in a stool sample.

In its simplest form, two genomes or more can be compared to each other by counting the number of positions where they differ. Within-host diversity is determined by comparing two or more genomes from the same host (Golubchik et al. 2013). Comparing two genomes from different hosts lends insight into chains of transmission. These approaches have been successfully applied to several organisms beside E. coli (Croxall et al. 2011, Alhashash et al. 2013, Didelot et al. 2013, Halachev et al. 2014, Worby et al. 2014, Kay et al. 2015, Ssengooba et al. 2016, Herranz et al. 2017, Trauner et al. 2017, Hall et al. 2019, Ley et al. 2019).

Genotyping multiple isolates per host can also facilitate the identification of multiply infected individuals, representing the carriage of bacterial subpopulations descended from a distinct founder strain. Multiple concurrent infections are clinically crucial for the evolution of pathogens, particularly in facilitating recombination between divergent strains as discussed above and hence, the emergence of novel genotypes (Wymant et al. 2018). Clinically, multiple infections are known to affect disease progression. For example, in HIV infection, the disease is reported to accelerate in the presence of dual infections (Cornelissen et al. 2012). A total of two plausible scenarios could explain the origin of multiple infections. In the first instance, two distinct strains could be transmitted simultaneously from an external source that harbors sufficient strain diversity. Alternatively, the two strains could have arisen from separate sources at different times.

Studies of the within-host diversity of E. coli

Studies of the genetic within-host diversity of E. coli in the human gut date back to well over a hundred years ago and have involved a wide variety of methodologies—encompassing both microbiologic and molecular techniques. The earliest studies involved the use of serotyping and subsequently multilocus enzyme electrophoresis. Later studies employed multilocus sequence typing and WGS to characterize strains and infer within-host diversity and evolution of strains. Below, we present a summary of studies investigating the within-host diversity of E. coli since 1890.

Pre-PCR era

The first evidence of temporal changes in the population of E. coli in the human gut was documented by Smith in 1899 (Smith 1899). In this pioneering study (reviewed in Wallick and Stuart 1943), Smith analysed 48 isolates obtained from four separate cultures of the stools of a single normal infant and reported a high degree of temporal antigenic stability. Subsequently, Totsuka (1902; reviewed in Wallick and Stuart (1943) and Martinson and Walk (2020)) self-collected and cultured his own stool once a week for 12 weeks and recovered up to 32 isolates per culture, yielding a total of 332 isolates. Upon agglutinating these against antisera, which he had prepared against isolates obtained from the first two cultures, he provided evidence of gradual shifts from one antigenic type to the other over the course of time in the intestine of one individual.

In a similar study, Wallick and Stuart (1943) isolated 650 E. coli strains from one individual over the course of 14 months and described four distinct serotypes that persisted for a month at a time—including co-occurrence of dominant and minority strains. They also inferred household transmission based on antigenic identity (serotype).

Sears, Brownlee, and Uchiyama (Sears et al. 1950, Sears and Brownlee 1952) investigated changes in E. coli serotypes over time in five adults sampled over 3 and 30 months. They were the first to coin the term ‘residents’ and ‘transients’ to describe strains of E. coli ‘which establish themselves firmly and continue to multiply over extended periods of time’ and those that persisted for only a few weeks at most respectively. For the first time, they also investigated the production of bacteriocins by resident clones to ward off the invasion of other clones.

Next, Robinet (Robinet 1962, Branche et al. 1963) confirmed the periodic fluctuations of antigenic types in six healthy individuals followed up for 6 months and concluded that host antibody production does not explain the turnover of resident strains of E. coli in the gut. Robinet was involved with further work investigating bacteriocin production by the resident clones by testing the activity of strains collected in 1 month against those recovered the previous month. These studies concluded that the turnover of strains was lower when the resident strains produced colicins.

Shooter et al. (1977) investigated the serotype dynamics of E. coli isolated from nine adults over 3 months. They combined H- and O-antigen testing for the first time and observed that several strains with an identical O antigen, in fact, displayed different H antigens. Following on from this, Selander and colleagues (Selander and Levin 1980, Caugant et al. 1981) sampled E. coli from a single individual for 11 months and, using multilocus enzyme electrophoresis, provided seminal insight on the genetic structure of the resident and transient clones in the healthy adult gut. Importantly, they documented the loss and acquisition of plasmids in identical clones over time and the gain and loss of novel clones over the study period as essential contributors to the generation of E. coli diversity in the gut.

Finally, Tzabar and Pennington (1991) explored the dynamics of E. coli carriage in a group of individuals residing at the British Antarctic Survey research base for 26 weeks. Using MLEE, they provided evidence of strain-sharing among the members of the research station—although several isolates were lost to viability, hampering the study conclusions on the clonal turnover over time.

PCR era

Several studies (Johnson and Clabots 2006, Johnson et al. 2008b, Damborg et al. 2009) have explored the diversity of E. coli among adults, children and companion animals living in the same household over time. Key findings include the detection of clones that appeared to persist for 4–45 weeks, as well as documenting evidence of strain sharing among members of a household and their pets. Shared strains between humans and companion animals were found to be of phylogroups D and B2, with an ST73 strain (phylogroup B2) causing UTI in a family dog.

Martinson et al. (2019) studied eight healthy adults via biweekly sampling for 6 months to 2 years and observed that the resident clones often belonged to phylogroups A, B2, and F, including multiple resident clones persisting for more than a year in one individual.

Some studies have pointed to the possible exchange of plasmid-borne AMR genes between co-colonizing strains within a single individual. First, Stoesser et al. (2015) applied multiple colony sampling (16 colonies from each of eight faecal samples) followed by whole-genome sequencing to investigate the transmission of E. coli strains harboring ESBL genes collected in Cambodia. They reported substantial core- and accessory genome diversity, with a median of four STs recovered per individual. Remarkably, different clones from a single individual tended to share the same bla_CTX-M variant, and identical clones were found with different bla_CTX-M variants. Significant accessory genome diversity was also observed within and between clones—highlighting the utility of multiple colony sampling and whole-genome sampling in the analysis of E. coli within-host diversity.

Second, Knudsen et al. (2017, 2018) collected E. coli (and other enterobacteria) from the stools of children with cystic fibrosis or cancer along with matched healthy controls over a 9-month period, picking up to five colonies per faecal sample. Children with cystic fibrosis or cancer received antibiotic treatment over the course of the study. Of the isolated E. coli (90% of children with cystic fibrosis, 93% of children with cancer, and 94% of healthy controls), the prevalence of antibiotic-resistant enterobacteria did not significantly differ between the children who received antibiotics and the healthy controls at the start and at the end of the study, suggesting that the level of antibiotic resistance they observed arose from the community. As a follow-up study, seven isolates collected from three consecutive samples collected from one child with cystic fibrosis was characterized by WGS. Here, three distinct strains were detected, and the different strains were found to harbor Inc1 plasmids encoding bla_CTX-M-1. However, the plasmids from the three different strains were found to differ by only a few SNPs and varied with limited regions, suggesting recombination events. This study documented the horizontal transfer of bla_CTX-M-1-harboring plasmids within a single individual.

Stegger et al. (2020) analysed 20 E. coli isolates from each of nine urine samples collected from nine women who presented with UTI at a general practice in Zealand, Denmark, from which a single clone was detected in eight out of the nine samples. The authors then selected a total of 40 isolates belonging to the same clone: 10 each from two urine samples and two rectal swabs isolates from two healthy individuals (collected in a previous study; Nielsen et al. 2014) and investigated the intraclonal diversity of E. coli among the commensal and uropathogenic strains. A low intraclonal diversity was observed for each clone in both the commensal and pathogenic strains (0–2 nonsynonymous SNPs). The authors reached an interesting conclusion, stating that ‘sampling of one colony would be enough for surveillance, outbreak investigations and clonal evolution’, although there is overwhelming evidence to support the opposite. Evidence from their own study showed that among other clones, a variation in gene content of 2–15 genes was detected for all clones—which would have been unnoticeable had they isolated only a single colony from each sample.

We recently surveyed E. coli in healthy children (less than 5 years) from a rural Gambian community and assessed the relative contribution of independent immigration and establishment of strains vs. within-host evolution to the generation of within-host diversity in the healthy gut (Foster-Nyarko et al. 2021). Our results showed a dominance of independent immigration and establishment of strains, with several strains harboring putative virulence factors commonly associated with Enteropathogenic and Enteroaggregative E. coli—thus, adding to the body of evidence indicating a blurring of the lines of demarcation between resident strains and their pathogenic variants, while also adding significantly to the number of whole-genome sequences from commensal E. coli isolates, particularly from sub-Saharan Africa.

Prospective studies of travellers to low-income countries have provided vital insight into the acquisition and loss of strains, particularly, MDR clones. In a recent report (Foster-Nyarko et al. 2021), Kantele et al. investigated the dynamics of intestinal carriage of MDR bacteria during travel to a low-income country viadaily stool sampling for the duration of their visit—unlike most previous studies, which only involved before and after samples. The study involved participants attending a medical course in Vientiane, Laos, together but staying at three different hotels, visiting several restaurants together or in small groups, and participating in daily rounds at the local hospitals. Remarkably, all the study participants acquired ESBL-producing bacteria at least at one point during the study—dominated by E. coli, with 95% of the participants acquiring between two and seven strains and multiple transient colonization events, with 20% of the participants sharing some common strains over the study period.

In contrast to the abundance of studies in humans, few studies have investigated the dynamics of E. coli residency in nonhuman mammals. In a temporal survey of E. coli residency in 54 mountain brushtail possums sampled on four occasions over the course of a year, Blyton et al. reported phylogroup B2 to dominate the resident E. coli population (Blyton et al. 2013). This study concluded that E. coli was rapidly gained and lost among these nonhuman mammals, with just 36% of resident strains recovered at one timepoint recovered at the subsequent timepoint. Recent data from our lab involving a range of diurnal monkey species in the Gambia also showed a dominance of E. coli belonging to phylogroup B2 (Foster-Nyarko et al. 2020a), including a prevalence of well-characterized lineages associated with human extraintestinal infections or AMR. However, the strains that occurred in these monkeys were genetically divergent from those sourced from humans, with little AMR among the simian isolates. Several monkeys were colonized by two or more strains, with the evidence of strain-sharing among monkeys sampled from the same park.

Data from poultry further support the observation that co-colonization with multiple strains is prevalent in nonhuman vertebrate hosts, reflecting the findings of previous studies in humans. Li et al. (2019) sampled > 100 colonies per caecum sample from nine mcr-1-positive broiler chickens from three provinces in China (a total of 962 E. coli isolates). A high rate of co-colonization (3–9 STs per chicken) was observed, with several birds harboring one to five Inc type plasmids encoding mrc-1—this high level of heterogeneity facilitating the transmission of mcr-1 among these chickens. The authors concluded that the ‘gut is a ‘melting pot’ for active horizontal transfer of the mcr-1 gene’. We recently investigated the within-host diversity of E. coli from backyard poultry (10 backyard chickens and nine guinea fowl) drawn from rural Gambia, and found similarly high colonization with multiple strains in noncommercial birds (Foster-Nyarko et al. 2020b). Worryingly, a dominance of ST155—a well-characterized multidrug resistant clone—was observed in poultry E. coli isolates from Africa and South America. However, the isolates from these backyard birds carried few ESBL genotypes compared to what was found among other isolates from the African continent (Foster-Nyarko et al. 2020b).

In summary, the body of evidence shows that colonization with multiple strains is common in the vertebrate gut, with the rapid gain and loss of strains contributing to the within-host diversity of E. coli. At the same time, some clones can establish themselves in the gut for extended periods. As evidenced by the existing data, this level of diversity positions the gut as a hot spot for the active horizontal transfer of AMR and virulence traits within and between strains. Other vertebrate hosts may carry lineages well-established in humans—ndicating bacteria that have probably existed in the specific host species for long periods and passed down through vertical transmission. The direct contribution of nonhuman vertebrate hosts to the spread of AMR and virulence traits in humans remains to be established, despite the prevalence of isolates with similar profiles as their close relatives circulating in humans. Significant gaps remain in our understanding of the ecology of this familiar ally and foe of the vertebrate host, key examples of which are outlined below.

Conclusions and perspectives

It has been said: ‘All cell biologists have at least two cells of interest: the one they are studying and E. coli’ (Neidhardt et al. 1996). Yet, in the 137 years since E. coli was first described, most studies on the ecology of E. coli have been biased to pathogenic strains. Only a handful of studies have focused on the diversity of E. coli populations in the healthy human gut, particularly in the post-PCR era. Even fewer studies have explored the within-host diversity of nonpathogenic E. coli in nonhuman vertebrates in the post-PCR period, leaving much to be learned in terms of the resident populations of E. coli in the healthy vertebrate gut. However, such studies are vital to our understanding of the evolution of E. coli in the health and disease of both humans and animals.

An exploration of the commensal population of E. coli in nonhuman vertebrate hosts such as poultry and nonhuman primates is timely and relevant for several reasons. First, emerging infections are often linked with pathogens (such as E. coli) that inhabit both humans and nonhuman vertebrates and can cross the species barrier, potentially causing zoonotic as well as anthroponotic infections. Understanding the diversity within the commensal population would provide vital insights into the evolution of pathogens and AMR within this ecological context. Second, understanding the within-host evolution of E. coli during health may also inform new therapeutic approaches that exploit our understanding of the gut microbiome, such as faecal microbiome transplantation, probiotic development, and the control of foodborne diseases.

In particular, E. coli diversity studies based on whole-genome sequence data from sub-Saharan Africa are scarce. Very few studies have examined the population of E. coli among healthy individuals in this setting, particularly in healthy children. However, given the increased exposure to the environment, unsanitary conditions, and proximity to animals especially in rural areas, such studies are likely to yield crucial insights into the dynamics of strain turnover and residency of E. coli in the gut, ultimately facilitating our understanding of infection by this organism.

Furthermore, the bulk of existing surveys of E. coli in poultry have been biased towards the surveillance of AMR in commercial birds; thus, we know very little of the dynamics of colonization and the diversity of E. coli in this niche. Also, backyard birds, which are kept in proximity to humans, have been largely neglected in ecological studies. However, given the frequent contact with humans, studies involving backyard poultry might yield critical insights into the potential for the dissemination of strains and AMR between the two host species.

The direct contribution of human-habituated animals to E. coli transmission remains unknown. To address this would require future studies incorporating samples from whole households and animals living in proximity. The inclusion of nonhuman primate species in future studies of resident E. coli diversity would elucidate the full scale of the impact of human encroachment on natural habitats. The recent advances in Illumina sequencing platforms twinned with the potential of long-read sequencing (such as Pacific Biosciences sequencing and Oxford Nanopore technology) provide a timely opportunity to apply these approaches to these lines of study.

Outstanding questions that need addressing in future studies include: what factors govern co-existence or competition between the different E. coli strains co-colonizing the vertebrate gut? Do they exploit different microgeographical or nutrient niches in the gut? What is the importance, if any, of flagellar motility? How far do within-species antibacterial factors (e.g. colicins and type VI secretions systems) play a role in competitive exclusion? Can differences in the distribution of E. coli strains in vivo be accounted for by behaviors in vitro (e.g. competitive growth under different nutrient conditions)?

Biolog assays (Biolog Inc., Hayward, CA) are one way to screen for the key metabolic differences between different strains that may account for their co-existence or competitive exclusion within the gut. Candidate strains could then be selected for onward in vitro competition assays via individual and competitive fitness experiments and indirect antagonism assays as have been used by Durso et al. to assess fitness differences and competition between commensal strains and O157:H7 strains (Durso et al. 2004). Applikon Biotechnology (https://www.applikon-biotechnology.com/en/products/cultivation-systems/micro-matrix/) have designed mini-bioreactors that can be used for multiple competition assays in parallel, using 24-well microtiter plates with built-in controls for pH, temperature, and dissolved oxygen. Such experiments may shed light on the physiological and biochemical properties underlying the co-colonization of different strains in the vertebrate gut.

Conflicts of interest

None declared.