Food products are a key source for transmission of pathogenic bacteria, such as Salmonella and Escherichia coli. (Gizaw, 2019) E. coli colonizes the human gastrointestinal tract as a commensal but some pathotypes of E. coli can cause serious disease (Denamur et al., 2021). Pathogenic strains of E. coli can be grouped into intestinal (InPEC) and extraintestinal (ExPEC) pathotypes depending on whether they cause primarily gastrointestinal symptoms or whether they affect other organs, including urinary tract, bloodstream, wounds, kidney, brain, and other internal organs (Sora et al., 2021). ExPEC strains are responsible for a large fraction of urinary tract and bloodstream infections in humans (Sarowska et al., 2019) and can be classified into 4 pathotypes: uropathogenic E. coli (UPEC, primarily associated with urinary tract infections-UTIs), avian pathogenic E. coli (APEC, which causes colibacillosis in poultry but that can also infect humans), septicemia-associated E. coli (SEPEC) and neonatal meningitis-causing E. coli (NMEC) (Meena et al., 2023).

Community and hospital ExPEC infections are generally treated with 2^nd and 3^rd generation cephalosporins, fluoroquinolones, or with trimethoprim-sulfamethoxazole (Pitout, 2012). Resistance to cephalosporins can generally be attributed to the acquisition of genes with extended-spectrum β-lactamase (ESBL) activity; in ExPEC, the most frequent ones are bla _CTX-M variants. Resistance to fluoroquinolones is largely due to mutations that reduce the affinity of these drugs for their targets (gyrase -gyrAB- and topoisomerase IV -parCE-), although resistance can also be provided or enhanced by a fluoroquinolone-degrading enzyme (aac(6’)-lb-cr), by DNA-protecting enzymes (qnr genes), or by efflux regulation mutations (marR, acrR and soxR) (Huseby et al., 2017; Poirel et al., 2018). Resistance to dihydrofolate metabolism inhibitors is generally provided by dihydrofolate synthetase genes that are insensitive to sulfonamides (sul) or by dihydrofolate reductase mutants (dfr) that are insensitive to trimethoprim (Van Duijkeren et al., 2000). Topoisomerase genes represent core functions and are most frequently found in the chromosome. The other resistance genes are accessory and tend to be found in plasmids, which in turn facilitate their spread, mostly by conjugation. Over time, plasmid-borne genes tend to get incorporated into chromosomes with the assistance of mobile genetic elements (MGEs) such as insertion sequences and transposons (Wang Y. et al., 2022).

Sequence type 131 (ST131) is a globally-dominant multidrug resistance (MDR) clone and a major driver of the current ExPEC pandemic. Compared to other globally dispersed clones, ST131 stands out for its highly dynamic accessory genome (Decano et al., 2021). Like most UPEC strains, ST131 resides in the human gut as a commensal but can cause mild to severe infections of urinary tract and asymptomatic urinary bacteriuria (Biggel et al., 2020; Tchesnokova et al., 2020). The ST131’s genome exhibits a broad range of known or suspected virulence factors that likely contributed to its dominance in the clinic. These include siderophores, adhesins, toxins, protectins and other elements contributing to the successful establishment and persistence of an infection. Some of these factors are specific to ST131 sublineages, notably hly (hemolysin), iro (siderophore), AAF (aggregative adherence fimbriae), and papGII (P fimbrial tip adhesin variant).

The complex population structure of ST131 has been elucidated (Price et al., 2013; Petty et al., 2014). Two ancestral clades have been identified, namely clade A (which carries the fimH41 allele) and clade B (which carries the fimH22 and fimH35 allelic variants). Strains belonging to these two clades, which are now infrequent, are generally sensitive to fluoroquinolones and rarely carry ESBL plasmids, although they occasionally have bla _CTX-M genes integrated in the chromosome. Clade C is an emerging epidemic lineage characterized by multidrug resistance and higher virulence (Denamur et al., 2021). This clade appears to have evolved from clade B by the acquisition of the fimH30 allele and of several prophages and it expresses serotype O25b:H4. Within clade C, three subclades can be distinguished by divergent antimicrobial resistance profiles: C0, which is sensitive to fluoroquinolones, and C1 and C2, which have acquired fluoroquinolone resistance mutations in gyrA and parC. C1 and C2 in turn, differ in the bla _CTX-M gene associated with them: subset C1 (with fimH30R1), is typically associated with bla _CTX-M-14 or with bla _CTX-M-27, whereas subset C2 (with fimH30RX), is generally associated with bla _CTX-M-15. The bla _CTX-M-15 gene was initially carried by IncF plasmids but more recently the mobilization of this gene to the chromosome has been reported through transposition events frequently involving insertion sequences belonging to the IS26 family (Dhanji et al., 2011; Stoesser et al., 2016). The C2 subclade has been found to be enriched in highly virulent and multidrug resistance isolates, with the virulence associated with the presence of the gene papGII (Pajand et al., 2021; Biggel et al., 2022).

ST131’s main reservoir is thought to be the human intestine and the usual route of transmission to be person-to-person contact (Lopez et al., 2014; Pitout and Finn, 2020); however, an increasing (although still small) number of studies are reporting its isolation from non-human sources, such as food, food animals, pets, and environmental sources, suggesting that these may represent vectors for the transmission of this strain (Meena et al., 2023). ST131 strains recovered from food have been associated primarily with meat products, particularly pork and chicken (Meena et al., 2021); because of ST131’s ability for long-time residence in the gastrointestinal tract, animal carcasses contaminated with intestinal content are considered the likely source in meat products. It also needs to be noted that, in addition to contaminating food, E. coli food strains can also function as reservoirs of virulence and antibiotic resistance genes, which can be frequently exchanged with clinical strains via horizontal gene transfer (HGT) (Balbuena-Alonso et al., 2022).

Here we report the isolation, phenotypic characterization, and genomic sequence of an E. coli isolate from a spinach sample that we named A23EC. We describe the presence of a large number of virulence factors, mainly located in three pathogenicity islands, that suggest a uropathogenic pathotype. Also, we show that this strain presents a multidrug resistance profile that is consistent with its antibiotic resistance gene (ARG) content, and we describe a conjugative plasmid that is capable of self-transmission even at room temperature. Finally, we compare A23EC’s genomic sequence with that of other fully-assembled ST131 genomes in GenBank and find that A23EC belongs to a virulent, emerging sublineage of the ST131 subclade C2. We identify genetic markers unique to this sublineage and report on their alignment with earlier studies of virulent sublineages of the C2 subclade that were based on partially assembled sequences. These results highlight the pressing need for rigorous epidemiological surveillance of ExPEC in vegetables with a One Health perspective.

Here we describe the isolation of a strain (A23EC) from a spinach sample at point-of-sale in Mexico and present a detailed genomic characterization of this isolate. Based on its sequence, this strain belongs to the C2 subclade of ST131, which is recognized as a pandemic clone that is highly virulent, multidrug resistant, and widely distributed around the world (Petty et al., 2014).

A phylogenetic analysis of all complete genomes corresponding to ST131 E. coli deposited in the Genbank database as of June 2022, places A23EC in a monophyletic sublineage of subclade C2, which we named C2b, clustered with seventeen other strains. These 18 genomes showed four unique commonalities, illustrated in Figure 1 : 1) They carry a TnMB1860 transposon structure flanked by IS26 elements. This transposon is chromosomally integrated in all cases and its chromosomal integration site in these genomes is consistent with a single capture event. 2) They carry a PAI II536-like pathogenicity island with an additional cnf1 gene. 3) They consistently belong to virotype E. 4) Thirteen of the eighteen isolates (including A23EC) also carry a F(31,36):A(4,20):B1 plasmid. We did not detect this pMLST type in any other ST131 sample included in our analysis, suggesting that it is specific for sublineage C2b samples.

By contrast, strains belonging to the C2a sublineage included in the analysis lack the TnMB1860 transposon and frequently carry bla _CTX-M-15 in plasmids, belong mostly to virotypes A or C (only two out of thirteen belong to virotype E); and consistently have F2:A-:B- plasmids when plasmids are present.

Note that all strains belonging to sublineage C2a were collected before 2014 (mean isolation date, 2009), whereas 16 of the 18 strains belonging to sublineage C2b were collected after 2014 (mean collection date, 2017), suggesting that C2b represents an emerging sublineage of the C2 subclade; this could be confirmed by further analysis with a larger number of strains.

Recent studies of ST131 clade C clinical strain collections already point to the emergence of new lineages of subclade C2 exhibiting higher virulence and antibiotic resistance. These virulent lineages are consistent with C2b, although in these studies, the identification of genomic markers was less comprehensive because most of the sequences were not completely assembled. Specifically, one study looked into the association of the gene papGII with the expansion of ST131 (Biggel et al., 2022). This report noted a large expansion of papGII+ isolates within the C2 subclade and distinguished three papGII+ sublineages: L1, L2 and L3. We noticed that the three fully sequenced strains from that study correspond to sublineage C2b in our analysis (marked with a triangle next to the accession number in Figure 1 ) and carry all the genetic markers that we defined as diagnostic for C2b. In that study, these strains were ascribed to L1 sublineage of papGII+ C2 strains (along with 233 additional samples). The L1 sublineage was defined by the dominant presence of F(31,36):A(4,20):B1 alleles in IncF plasmids, and by the dominant presence of bla _CTX-M-15 as ESBL-encoding gene; further, the sub-branch they were ascribed to, L1a, consisting of 184 sequences is (like C2b) characterized by the insertion of TnMB1860 in the vicinity of the metG gene and by the presence of the two virulence genes hly/cnf1 in the papGII+ PAI. Thus, our proposed C2b sublineage appears to align perfectly with the L1a sub-branch of the Biggle et al., 2022 study (Biggel et al., 2022). They found two additional distinct papGII+ sublineages within the C2 subclade. Given that these additional sublineages (L2 and L3) were less frequent and more geographically restricted (Ludden et al., 2020; Biggel et al., 2022), their absence in our study can likely be attributable to our much smaller sample size.

A second study identified a monophyletic cluster of 22 C2 strains enriched for virulence and for ARG markers that was named by the authors “C2 subset” (Pajand et al., 2021). Similar to our proposed C2b sublineage, this C2 subset was characterized by the presence of hlyABCD virulence genes found in a PAI II₅₃₆-like genomic island containing cnf1, by largely belonging to virotype E (therefore being papGII+), and by frequently carrying plasmids with F31 or F36:A4:B1 replicons. This study also describes the presence of aac(3)-IIa and of a IS15DIV-bounded transposon with aac(6’)-lb-cr, bla _OXA-1, ΔcatB3, although they report the frequent presence of aac(3)-IId instead of that of aac(3)-IIe (an arrangement similar to that of MB1860TU_A, given that IS15DIV is very closely related to IS26), and bla _CTX-M-15 with an ISEcp1 upstream of and in the same orientation as bla _CTX-M-15, the reversed wbuC and a Tn2 transposon (possibly MB1860TU_B). The concordance between the markers associated with this “C2 subset” and our proposed C2b sublineage, is striking.

We noted the presence of a plasmid in strain A23EC. This plasmid (named pA23EC) was classified as PTU-FE by COPLA and has four replicons, three of which belong to incompatibility group F: F(31,36):A(4,20):B1, which can be annotated as F-:A-:B1. As it happens, PTU-FE F-:A-:B1 is one of four PTU-replicon combinations previously proposed to mediate most of the flow of resistance and virulence genes between food and clinical strains, so the present work supports the idea that the plasmid flow between food and clinical strains is preferentially mediated by a specific subset of plasmids (Balbuena-Alonso et al., 2022), although the presence of A23EC in food may be incidental in this case. The fourth replicon belongs to the Col156 type, which has been previously described in isolates of the C1 subclade, carrying CTX-M-14 or -27 (Kondratyeva et al., 2020) (seen also in Figure 1 ). The presence of Col156 replicons, which are infrequent, in separate subclades of ST131 raises the possibility of plasmid exchanges across ST131 subclades.

Plasmid pA23EC had a complete set of mobilization genes, suggesting that this plasmid is capable of conjugation. Indeed, we detected conjugation at both 20 and 37°C. Conjugation frequencies were similar at both temperatures, which is surprising, as lower temperature slows growth down; however, it appears that the frequency of conjugation of A23EC to the C600^Rif+ recipient strain is not influenced by incubation temperature. Thus, our report is the first to show efficient conjugation for a subclade C2 strain of ST131 at a temperature under 37°C. These observations suggest that pA23EC has the potential to spread via conjugation, not only in human hosts but also in environmental reservoirs. Also note that pA23EC is classified as PTU-FE, which exhibits a host range of III on a six-grade scale, with a level of promiscuity to the level of family, and therefore has the potential to contribute to genetic exchange across multiple species in the Enterobacteriaceae.

A23EC is a multidrug-resistant strain, with resistance to aminoglycosides, penicillin, cephalosporins, tetracyclines, quinolones and monobactam. The prevalence of strains that are resistant to all first-line drugs is rising at an alarming rate (Manges, 2016), and multidrug-resistant ExPEC has been categorized by the WHO as a high-risk pathogen of critical priority (Tacconelli et al., 2018). Genotypically, we found nine ARGs in A23EC. Six of these mapped to TnMB1860 in the chromosome, in two separate TUs. The first TU (MB1860TU_A) carried aac(6’)-lb-cr, bla _OXA-1, ΔcatB3, aac(3)-IIe and tmrB, whereas the second TU (MB1860TU_B) carried bla _CTX-M-15. Two additional ARGs were found elsewhere in the chromosome >1 Mb away from TnMB1860, namely aac(3)-IIa and mdfA; tetA/tetR were the only ARGs found in the pA23EC plasmid.

The ARGs that we identified are likely responsible for the observed resistance to penicillins and first-generation cephalosporins (bla _OXA-1 and bla _CTX-M-15), to synthetic cephalosporins (bla _CTX-M-15), monobactams (bla _CTX-M-15), (Zhu et al., 2022), fluoroquinolones (aac(6’)-lb-cr, and mutations at position S83L, D87N of gyrase and position S80L of ParC (Redgrave et al., 2014; Huseby et al., 2017), to gentamicin (aac(3)-IIe), to tobramycin (aac(6’)-lb-cr and aac(3)-IIe), to amikacin (aac(6’)-lb-cr, intermediate resistance) (Ojdana et al., 2018; Stogios et al., 2022) and to tetracycline (tetA/tetR) (Møller et al., 2016).

The genes mdfA and tmrB could enhance resistance to a variety of drugs rather than being primarily responsible for resistance to a specific drug. MdfA (also known as cmlA or cmr) is a proton-dependent pump with a very wide range of substrates that include chloramphenicol, erythromycin, roxithromycin and certain aminoglycosides and fluoroquinolones (Edgar and Bibi, 1997). TmrB is an ATP-binding membrane protein that protects against tunicamycin exposure, binding tunicamycin and functioning either as an efflux pump or as a permeability barrier for this drug (Noda et al., 1992; Noda et al., 1995). Given that tunicamycin is an experimental drug not used as an antibiotic in the clinic or as growth promoter for animals, the frequent presence of the tmrB gene in resistance-determining regions (Grevskott et al., 2020; Pungpian et al., 2022) is intriguing and points to a possible role as modulator of resistance to other antibiotics.

Chloramphenicol acetyl transferases acetylate the antibiotic chloramphenicol at the 30-hydroxyl position using acetyl coenzyme as an acetyl donor (White et al., 1999). CatB3 is a B-type acetyltransferase, which tends to have low activity against chloramphenicol, and forms homotrimers. Trimer formation requires the C-terminal α-helical domain and is important for catalysis because the acetyl acceptor site of each protein is located in a pocket formed between monomers of the trimer. Therefore, the loss of the C-terminal α-helical domain in A23EC’s ΔcatB3 is expected to destabilize the trimer (Alcala et al., 2020). Belonging to a CAT family with low activity against chloramphenicol to begin with and having a truncation that likely suppresses its catalytic activity could explain the sensitivity of A23EC to chloramphenicol despite carrying ΔcatB3. Indeed, in a previous report, a strain with this exact truncation was reported as sensitive to chloramphenicol (Hubbard et al., 2020). However, the fact that the two deletions observed in ΔcatB3 of A23EC are in-frame, and that this allele is fully conserved in our proposed C2b sublineage and even in other STs (strain CP048934.1 is classified as ST315) suggests that ΔcatB3 might retain some residual function that is being maintained through selection.

One of the C2b genomes (strain 11A CP049077.2) carries the TnMB1860 transposon integrated in the chromosome and also in plasmid p11A_2. This results in a gene dosage duplication for all the genes encoded in TnMB1860 that likely makes this transposon structure and all the ARGs that it contains redundant in strain 11A. The concurrent presence of a given ARG in a chromosome and in a plasmid within the same isolate is not uncommon and is interpreted as an intermediate stage in the incorporation of genetic content from a plasmid into the chromosome (Wang et al., 2022). This interpretation is also consistent with the incomplete penetrance of the F(31,36):A(4,20):B1 plasmid in sublineage C2b, which appears to have been lost independently four times ( Figure 1 ).

The A23EC strain bears the seven virulence markers that are characteristic of ExPEC, as expected for ST131 ( Figure S1 ) (Johnson et al., 2003; van Hoek et al., 2016); the presence of chuA, fyuA, and yfcV identifies A23EC as a potentially UPEC strain and its papGII+ status suggests it may be particularly virulent. Indeed, the PapGII fimbrial tip adhesin was shown to promote inflammation in renal tissue through transcriptional reprogramming of kidney cells (Biggel et al., 2020), and papGII+ strains have been reported to be enriched in blood isolates relative to urine/UTI infections (Ambite et al., 2019). Other virulence genes in A23EC including P fimbriae, hemolysins, siderophores, toxins and capsular synthesis, all located within PAIs, have previously been linked to the development of invasive infections (Sabaté et al., 2006; Tsoumtsa Meda et al., 2022). Outside PAIs, we identified fimH, which is associated with adherence on biotic and abiotic surfaces (Cookson et al., 2002), and iss, sitA and ompT, which are genes associated with serum resistance and favoring colonization and invasion mainly described in APEC strains (Olsen et al., 2012), as well as the presence of toxin gene usp associated with strains causing pyelonephritis, prostatitis and bacteremia (Nipič et al., 2013).

A high virulence of the C2b sublineage of the C2 subclade is also supported by the sources of isolation of C2b samples. Out of eighteen samples, ten were isolated from patients suffering from septicemia (58.8%) and none from urine. For comparison, out of thirteen C2a samples, five were isolated from urine (38.5%) and only one was isolated from the bloodstream (7.7%). Consistent with these observations, Pajand et al., 2021 report that the “C2 subset” (which as explained above aligns with our proposed C2b sublineage) accounted for the excess resistance and virulence of subclade C2 relative to C1 subclade strains (Pajand et al., 2021). However, the observed difference in sample origin between the C2a and C2b sublineages of the C2 subclade could also be attributed to unknown variables affecting the sampling and is based in both cases on a small number of samples.

Based on epidemiological surveillance studies that have been carried out in different parts of the world in the “One Health” context, non-animal agricultural products have been proposed to be important vectors for the circulation of multidrug-resistant ExPEC strains between humans, animals and the environments (Meena et al., 2023). Our report of an isolate belonging to the C2 subclade of ST131 contaminating green leafy vegetables adds support to this hypothesis. We searched for additional relevant reports in the literature. Table S4 lists examples of E. coli strains isolated from vegetables, along with their phylotypes (when known) and their corresponding reference. A few sequence types stand out, namely ST10, ST38, ST23, ST69 and ST155, which to our knowledge have been independently reported 8, 4, 3, 3, and 3 times, respectively out of a total of 36 annotated examples. Note that while most of the isolates reported likely represent commensals (based on their phylotype), about ⅓ of them (including ST69) belong to likely ExPEC phylotypes (B2, and D), supporting the idea that agricultural products can indeed serve as vectors for the transmission of ExPEC strains.

Humans represent the main reservoirs of E. coli ST131 by colonizing the intestine (Morales Barroso et al., 2018; Sarkar et al., 2018; Johnson et al., 2022). When isolated in foods, the source of ST131 E. coli are typically animal products such as meat and dairy products (Platell et al., 2011; Liu et al., 2018) and the C2 (O25:H4/H30) subclade of E. coli is much less frequent in foods other ST131 lineages such as H22 (Manges, 2016; Massella et al., 2020). The isolation of a bla _CXT-M-15-bearing ST131 strain from an agricultural product has only been previously reported once, from a bitter cucumber imported from the Dominican Republic (Müller et al., 2016), and we ignore further details about its molecular classification. Thus, to our knowledge, this is the first report of an ST131 clone belonging to the C2 subclade (O25:H4/H30) contaminating green leafy vegetable. The isolation of an emerging, potentially uropathogenic strain of ST131 from spinach is relevant to public health because the consumption of fresh produce has increased as a result of a healthier lifestyle (Castro-Rosas et al., 2012).

Three other strains of the C2b sublineage described here were isolated from wastewater ( Figure 1 ). Further, ST131 clade C strains have been observed to survive wastewater treatment and release to surface water (Tausova et al., 2012; Zurfluh et al., 2013; Müller et al., 2016). Thus, A23EC could have reached spinach via contaminated water. Admittedly, the present study cannot determine the point at which A23EC contaminated the spinach, and it is therefore possible that contamination happened after the spinach was harvested (during processing, transport or at the supermarket itself) but we can say that this strain is able to persist in spinach long enough for transmission. Whether the ability of A23EC and possibly other C2b strains to persist in fresh vegetables represents a new adaptation or it was simply previously missed is unclear. The presence of virulence factors facilitating adherence to human and animal cells (papABCD, papGII, iha, yfcV, kpsm-TII-k5, fimABCDEFGHI and csgABCDEFG) (Sarowska et al., 2019) may be relevant. This is particularly true of adhesins CsgA and CsgB, which have been described to be significantly involved in adhesion and colonization in spinach leaves (Saldaña et al., 2011; MacArisin et al., 2012), and allow their proliferation on the food surface through the formation of biofilms (Zhao et al., 2022). In addition, previous studies have demonstrated the ability of E. coli strains to reside within the internal cavity of stomata and internal tissues; this internalization protects the bacteria from disinfecting and bactericidal products, thereby increasing their survival. Internalization also contributes to inefficient washing and sanitizing treatments in vegetables (Gullian-Klanian and Sánchez-Solis, 2018; Querido et al., 2020). However, these previous studies have focused on STEC and EHEC strains and may have missed the presence of ExPEC in these foods.

Some reports interpret the occurrence of ST131 strain in different hosts such as companion and food animals or in different environmental niches such as sewage and other aquatic environments as overflow from its main niche (Melo et al., 2019; Finn et al., 2020); in contrast, other studies report specialization in ST131 strains depending on their origin, suggesting that they are adapting to different niches. This raises complex questions about the role of reservoirs in ST131 evolution and spread (Bonnet et al., 2021; Denamur et al., 2021). The isolation of ST131 strains in foods such as A23EC paves the ground for understanding the epidemiology and evolutionary dynamics of ST131 through rigorous and systematic monitoring using selected molecular markers.

In conclusion, our results add to previous knowledge about the global dissemination of ST131 by confirming the emergence of a distinct sublineage of subclade C2 (C2b), that has the potential to be highly pathogenic and that bears a transposon structure that has the potential to facilitate the evolution of carbapenem resistance among non-carbapenemase-producing enterobacterales (Shropshire et al., 2021). The genetic content of virulence and resistance genes and associated mobilization elements described here for the A23EC strain, together with the plasticity of E. coli ST131 genome, suggests that this ExPEC strain has the potential to evolve persistence in new environments and to infect humans and/or animals through new routes of transmission. These observations call for a more comprehensive surveillance and monitoring system for ExPEC strains in non-clinical settings.

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/ Supplementary material .

MB-A organized the database and performed the formal analysis. MB-A, GC-C, EC-L, and MC participated in the research. MC, PL-Z, and RR-G supervised the experimental and bioinformatics analysis. MC performed extensive review of the original draft. RR-G performed resources and project administration and funding acquisition. All authors contributed to the article and approved the submitted version.