Emergence of Enteroaggregative Escherichia coli within the ST131 Lineage as a Cause of Extraintestinal Infections

E. coli ST131 is an important extraintestinal pathogenic lineage. A signature characteristic of ST131 is its ability to asymptomatically colonize the gastrointestinal tract and then opportunistically cause extraintestinal infections, such as cystitis, pyelonephritis, and urosepsis. In this study, we identified an ST131 H27 sublineage that has acquired the enteroaggregative diarrheagenic phenotype, spread across multiple continents, and caused multiple outbreaks of community-acquired ESBL-associated bloodstream infections in Denmark. The strain’s ability to both cause diarrhea and innocuously colonize the human gastrointestinal tract may facilitate its dissemination and establishment in the community.

KEYWORDS E. coli, ESBL, genomic, H27, resistance, ST131, enteroaggregative, evolution, outbreak, plasmids E scherichia coli sequence type 131 (ST131), which is the dominant multidrugresistant (MDR) extraintestinal pathogenic E. coli (ExPEC) lineage worldwide, causes a wide range of infections, including bloodstream and urinary tract infections (BSIs/ UTIs) (1)(2)(3). Its rise to global dominance and pathogenicity presumably was primed by the sequential acquisition of virulence-associated genes and then antibiotic resistance (4,5). ST131 exhibits predominantly serotype O25:H4 and is closely associated with fluoroquinolone resistance and the CTX-M-15 extended-spectrum ␤-lactamase (ESBL) (2,6). The global expansion of E. coli ST131 has been driven mainly by a single clade, designated H30 because of its tight association with allele 30 of the type 1 fimbrial adhesin gene, fimH. H30, also described as clade C, has a prominent MDR clade, H30R, which accounts for most fluoroquinolone resistance within ST131. H30R, in turn, has two main subclades: H30R, of which most isolates are ESBL negative, although some produce CTX-M-14 or CTX-M-27, and H30Rx, which accounts for almost all ST131associated CTX-M-15 production (6). In addition to H30, other distinct clades of ST131 are also circulating worldwide, most commonly associated with fimH alleles 41 and 22, also previously designated clades A and B, respectively (7). While these non-H30 clades are usually not associated with carriage of CTX-M genes, ST131-H22 and -H41 strains carrying bla CTX-M-14 , bla CTX-M-15 , or bla CTX-M- 27 have been reported (4,8), as have been food-associated ST131-H22 strains that carry bla TEM-1 (9). ST131 isolates typically carry multiple ExPEC-associated virulence genes encoding adhesins, toxins, and siderophores but reportedly rarely carry virulence genes typical of diarrheagenic E. coli (DEC) (10,11). However, among the 128 archival ST131 O25 isolates within an international World Health Organization (WHO) collection that we surveyed for temporal trends in antibiotic resistance and virulence traits, we identified 12 isolates (9%) that fulfilled molecular criteria for the enteroaggregative E. coli (EAEC) pathotype (12). Pulsed-field gel electrophoresis (PFGE) analysis revealed a cluster (ϳ80% similarity coefficient) comprising seven of these EAEC isolates. Of these seven isolates, six, including two urine isolates from patients with UTI, three fecal isolates from patients with diarrhea, and one lower respiratory tract isolate (associated symptoms unknown), were from Danish patients (1998 to 2000), supporting the occurrence in Denmark during these 3 years of an unrecognized EAEC ST131-associated outbreak of UTI and possibly also diarrhea (12).
EAEC strains are a well-established cause of endemic diarrheal illness in developing countries and of foodborne outbreaks in developed countries and more recently have been associated with extraintestinal infections (13). This pathotype gained particular attention following a major foodborne outbreak in Germany in 2011 that was caused by a Shiga toxin (Stx)-producing O104:H4 EAEC strain and resulted in 3,842 confirmed cases and 54 deaths (14). Additionally, it is concerning that a high prevalence of multidrug resistance among EAEC strains has been reported, and reports of CTX-Mtype, ESBL-producing EAEC have emerged around the world (15)(16)(17)(18).
EAEC pathogenesis involves adherence to human intestinal mucosa by virtue of aggregative adherence fimbriae (AAF) and subsequent biofilm formation. The AAF are located on a large plasmid, designated pAA, that also encodes a suite of other EAEC virulence factors. These include AggR, a global regulator of EAEC virulence; dispersin, required for proper dispersal of AAFs on the bacterial surface; the AatPABCD transporter system, which mediates dispersin secretion; and Aar, a recently described negative regulator of AggR (19,20).
To understand the emergence and underlying genetic acquisitions leading to ESBL-producing EAEC ST131, we investigated the relatedness of a large collection of temporally and spatially diverse ST131 isolates. This investigation revealed the emer-and ORF4, that are assumed to play a role in isoprenoid biosynthesis (23). The two exceptional isolates, C1883-99 (H27) and C167-00 (H22), lacked most of these EAECspecific genes.
Invasive bla CTX-M-101 -containing EAEC ST131 isolates from Danish patients. To determine whether EAEC ST131 strains were present among contemporary Danish patients, we searched for ST131-and EAEC-specific virulence genes within a published collection of 552 whole-genome-sequenced ESBL-producing E. coli isolates from Danish patients with bloodstream infections (BSIs) between 2014 and 2015 (24). ST131 accounted for 50% of the isolates (24). Among the 258 ST131 isolates, 25 (9.7%) qualified molecularly as EAEC, of which 24 (9%) harbored the genes encoding AAF/V, AggR, Aar, Aap, AatPABCD, and ORF3/4. The remaining isolate contained the genes encoding AatPABCD (Table S2). All 25 isolates carried fimH27 and bla CTX-M-101 . In total, 27 isolates in the collection were bla CTX-M-101 positive, of which 93% qualified molecularly as EAEC. The EAEC pathotype did not occur in conjunction with any other ESBL-encoding gene. As shown by Roer et al., the 27 ST131 strains with bla CTX-M-101 formed a distinct cluster with nine or fewer SNP differences (24), supporting a recent emergence or clonal spread.
To assess whether a corresponding CTX-M-101-producing EAEC ST131 strain could be detected in the urine of the source patients for the blood isolates, genome sequencing was performed on all available urine isolates from 8 of the 25 patients with a CTX-M-101-producing EAEC ST131 BSI. Notably, as all eight patients had presented with UTI on one or more occasions, multiple ST131-positive urine samples were available for four of the patients. The urine samples were taken in a time frame ranging from 3 weeks prior to 8 months after the BSI episode. SNP analysis demonstrated that pairs of blood and urine isolates from the same patient were nearly identical (i.e., differed by Յ6 SNPs), whereas different pairs were separated by 0 to 11 SNPs. The respective blood and urine isolates from five of the eight cases were nearest neighbors in the phylogeny, strongly implying their distinct relatedness (Fig. S1). For the remaining three cases, the clustering was inconclusive. Combined, these findings suggest strongly that the ESBL-producing EAEC ST131 strains were capable of long-term persistence in these hosts and/or their immediate environment and of causing repeated episodes of UTI.
pAA plasmid characterization in ST131 isolates. To characterize in detail the pAA plasmids present in the H27 ESBL-producing EAEC ST131 strains, we applied MinION sequencing (Oxford Nanopore Technologies) to obtain the complete plasmid sequences for five bla CTX-M-101 -containing EAEC isolates and five EAEC isolates from the WHO collection. The plasmid from a single representative bla CTX-M-101 -containing isolate (ESBL20150001) was designated pAA-ST131 and used for in-depth in silico analyses.
The complete sequence of pAA-ST131 was 142,646 bp in length, with an average GϩC content of 48.7% (GenBank accession no. KY706108) (Fig. 2). A total of 197 ORFs were predicted and annotated, and 142 were assigned functionally. Two replicons were identified, RepFII and RepFIB, with the multireplicon F plasmid FAB pMLST formula of F1:A-:B33. The pAA-ST131 plasmid contained several genes associated with plasmid stability, including parA, parM, and stbB, plus three toxin-antitoxin (TA)-based addiction systems: ccdAB, vagCD, and relBE (Fig. 2). Furthermore, pAA-ST131 harbored a 33-kb complete tra region encoding transfer components (24 tra genes, 8 trb genes, and finO), implying that the plasmid is transmissible. It also contained genes encoding putative transmembrane proteins and proteins involved in catabolism and metabolism, plus several integrated mobile elements. It contained, as expected, all of the EAEC virulence factor genes already identified by Illumina sequencing but no additional known virulence factor genes or antibiotic resistance genes, including bla CTX-M-101 .
A BLASTP comparison was made of pAA-ST131 with five pAA plasmids from EAEC-type strains (each containing aggR, aap, and genes encoding an AAF variant), i.e., GenBank accession numbers (strain name) NC_018666 (strain 2011C-3493), NC_011752 (strain 55989), FN554767 (strain 042), and NC_008460 (strain DIJ1) and Sequence Read Archive (SRA) number SRA055981 (strain 226). This analysis showed that all these plasmids shared common features (Fig. 2), including EAEC virulence genes, but also varied substantially in genetic content, as described previously for E. coli virulence plasmids (25). Interestingly, regions containing putative metabolic, catabolic, and transmembrane proteins were unique to pAA-ST131. Although the sequences of the associated chromosomes were available for typing for only some of these EAEC isolates, analysis of available data showed that the isolates represented at least three unrelated STs (ST40, ST414, and ST678). Additionally, although, like pAA-ST131, four of the five reference pAA plasmids also contained the RepFIB and/or RepFIIA replicon, only three contained both, and only two contained the tra region.
Standard agarose gel-based plasmid profiling demonstrated that all 10 of the ST131 MinION-sequenced WHO-or bla CTX-M-101 -containing EAEC strains, for which complete plasmid sequences were obtained, contained only a single plasmid of various sizes (Fig. S2). Isolates ESBL20150196 and ESBL20150300 (both obtained from Danish patients in 2015) each harbored a single 31-kb plasmid, which, based on in silico mapping, was the result of a single major deletion that resulted in the loss of all EAEC-specific virulence genes (Fig. S3). The two remaining ESBL-producing isolates, plus three of the WHO collection isolates, harbored plasmids nearly identical to pAA-ST131 (ESBL20150001). Isolates C546-00 and C1883-99, from the WHO collection, had slightly truncated plasmids in which a single or double deletion event, respectively, had resulted in the loss of some EAEC-specific genes (Fig. S3). None of the ESBL-producing isolates contained additional plasmids (Fig. S2), implying a chromosomal localization for bla CTX-M-101 .
We next assessed whether the 12 EAEC ST131 isolates from the WHO collection were clonally related to other available EAEC ST131 isolates. For this, a SNP-based phylogeny was constructed, including (i) all 121 ST131 isolates from the WHO collection (12); (ii) the 93 German/U.S. ST131 isolates (6); (iii) the 27 Danish bla CTX-M-101 -containing ST131 isolates from 2014 to 2015 (24); (iv) all 42 international fimH27-carrying ST131 isolates from EnteroBase; and (v) all seven non-H27, aggR-positive ST131 isolates from Entero-Base. Focusing on isolates within the H22 clade (Fig. 3A), WHO collection EAEC isolate Colors represent different putative functions: gray, hypothetical proteins; red, EAEC-specific virulence factors; blue, plasmid replication and maintenance; maroon, catabolism and metabolism; orange, membrane and transporter proteins; green, conjugational transfer proteins (tra and trb genes); light blue, regulatory genes; purple, miscellaneous; and black, mobile elements. Within the circles representing pAA plasmids from other EAEC strains, the darkest color indicates Ͼ90% nucleotide identity, the lightest color Ͼ80% identity.
In the unrooted phylogeny, most EAEC isolates nested within the H27 clade (Fig. 3A). To improve resolution, this clade's 79 isolates (76 with fimH27, 2 with fimH5, 1 with fimH54) were analyzed separately. The 27 bla CTX-M-101 -containing Danish EAEC isolates formed a distinct cluster (Fig. 3B). In contrast, none of the fimH27-carrying international isolates clustered either with the Danish isolates or with one another, suggesting that the outbreak was confined to Denmark and that no other focal outbreaks were captured. Instead, the WHO collection H27 EAEC isolates (11 Danish and 1 Vietnamese) were intermingled with isolates from the United Kingdom, Thailand, and Canada, suggesting the global spread of a common-ancestry strain.
To determine the mosaicism of the EAEC-specific virulence genes among fimH27carrying ST131 isolates, sequence reads from all 79 H27 clade isolates were mapped against pAA-ST131 from ESBL20150001. Intriguingly, genes of pAA-ST131 were highly conserved across the H27 EAEC isolates (Fig. 3B) when assessed based on continuous reference mappings of Ͼ1,000 bp in length using the Northern Arizona SNP Pipeline (NASP) against the pAA-ST131 reference plasmid. We next performed a SNP-based analysis of pAA-ST131 across all 57 H27 isolates with more than 70% coverage of the plasmid, with three H22 clade EAEC isolates from the WHO collection used as an outgroup. This identified 197 SNPs within the ϳ25% of pAA-ST131 that were conserved Emergence of ESBL-Producing EAEC ST131 ® across all isolates, equivalent to ϳ1 SNP per 4,000 bp conserved plasmids among the H27 isolates compared to 1 per ϳ5,000 in the core chromosome of the same isolate collection. Phylogenetic analysis based on these SNPs showed that pAA-ST131 is highly conserved among the H27 isolates but differs considerably between the H27 and non-H27 ST131 isolates (Fig. S4), strongly suggesting a single acquisition of the plasmid by an H27 clade ancestor, with subsequent partial deletions within certain descendants.
Phenotypic confirmation of pAA plasmid-mediated adhesive properties. To determine whether the AAF encoded by the pAA plasmids in H27 isolates is functional and potentially facilitates adherence to the intestinal mucosa, we compared the ability of two of the ESBL EAEC ST131 isolates, ESBL20150001 and ESBL20150196, carrying a full-length and truncated pAA plasmid lacking the AAF-encoding genes, respectively, to adhere to T84 colonic epithelial cells. ESBL20150001 adhered significantly better than did ESBL20150196, whereas both isolates adhered significantly better than did a laboratory E. coli K-12 strain (Fig. S5).

DISCUSSION
E. coli ST131 is an extensively reported MDR ExPEC lineage associated with both UTI and BSI (1). Many typical ExPEC-associated virulence factors, including P fimbriae, hemolysins, and factors conferring increased serum survival and iron uptake, have been identified in ST131 isolates (11,12), but overall little is known about which genes (whether they promote virulence or other phenotypes) have made this clonal lineage so successful. In contrast, virulence traits associated with diarrheagenic E. coli (DEC) pathotypes have thus far largely been absent from ST131 isolates (12). Their absence might be considered surprising because ST131 sublineages vary extensively according to acquired virulence genes (2, 28) and exhibit high levels of genomic plasticity. This plasticity includes frequent recombination and plasmid flux, particularly involving IncF-type plasmids, which facilitates the spread of antibiotic resistance and virulence genes (4,29). Importantly, in ST131 the fitness costs that result from these genomic events are offset by compensatory mutations (7). Here, we document the ST131 H27 sublineage to have acquired the enteroaggregative diarrheagenic phenotype followed by spread across multiple continents and to have caused outbreaks of communityacquired ESBL-associated bloodstream infections in Denmark.
In developed countries, EAEC is known mostly as a cause of mild to moderately severe, self-limiting diarrhea. Indeed, long-term carriage of EAEC has been suggested to lead to colonization rather than infection (13). In contrast, in developing countries EAEC is a leading cause of childhood diarrhea (13,30). The pathogenic potential of EAEC is underscored by its ability to cause major foodborne outbreaks of diarrheal disease (21,31,32). Intriguingly, recent studies have associated EAEC with UTI, suggesting that what classically has been regarded as a DEC pathotype also qualifies as ExPEC and causes both diarrhea and UTI (33)(34)(35).
Like other DEC pathotypes, EAEC has been shown to encompass diverse genetic lineages, reflecting a high level of phylogenetic heterogeneity (18,36). Although until recently no EAEC ST131 associated with extraintestinal infections had been documented, recent reports have described the occurrence in certain ST131 sublineages with EAEC-specific traits in geographically distinct areas (12,15,18). ESBL-producing ST38 strains, and other sequence types of various serotypes, have been implicated as emerging hybrids of UPEC and EAEC that cause mainly UTI in Germany, the Netherlands, and the United Kingdom (37). In a study of fecal and urine isolates from Danish patients in 1998 to 2000, we made the novel observation of EAEC ST131 isolates of serotype O25 (12). Subsequently, another study documented a bla CTX-M-14 -containing EAEC ST131 O25:H4 strain in stool samples of diarrheic patients in Japan from 2003 onwards (15).
Our previous PFGE analysis of Danish ST131 O25 E. coli isolates from the WHO collection demonstrated that 7 of the 12 identified EAEC strains were highly similar and appeared to have been part of an unrecognized UTI and diarrhea outbreak in Denmark from 1998 to 2000 (12). Although the genomic approach we used here to analyze the same isolates confirmed the suspected outbreak, it also yielded certain substantially different conclusions than those from the PFGE analysis, including the identification of a second distinct cluster of EAEC urine isolates. These discrepancies between WGS SNP-based analysis and PFGE analysis among ST131 isolates correspond with previous findings (6).
Here, we also document the current presence of an H27 ESBL-producing EAEC ST131 sublineage as a cause of bacteremia in patients admitted to hospitals in all five regions of Denmark. The isolates, collected across Denmark over a 16-month period in 2014 to 2015, all carried bla CTX-M-101 . Notably, heretofore the only other reports of CTX-M-101positive E. coli were from China (38)(39)(40). Unfortunately, travel information was not available for all the Danish patients with CTX-M-101-producing E. coli. Roer et al. used a genomic approach to establish that the Danish ESBL EAEC isolates were highly clonal, strongly suggesting a recent common source (24). Here, by analyzing sequential urine isolates from eight of the Danish patients with ESBL-producing EAEC ST131 H27 bacteremia, we found that five patients had recurrent UTI episodes caused by their BSI strain that were separated in time by up to 8 months from one another and/or the BSI episode. The most plausible explanation for this phenomenon is that the Danish ESBL-producing EAEC ST131 outbreak strain is capable of persistently colonizing patients, resulting in occasional clinical manifestations, although conceivably the patients may have been exposed repeatedly to an external source.
Using the EnteroBase collection of available ST131 genome sequences, we identified the global presence of fimH27-carrying EAEC ST131 isolates. These isolates originated from Africa, Asia, Europe, and North America, and, together with the Danish EAEC ST131 outbreak isolates, spanned more than 2 decades. Strikingly, all the EAEC ST131 H27 strains studied here proved to be clonal, i.e., to share an ancestor. Moreover, they share a virulence plasmid, which we designated pAA-ST131, which appears to have been acquired once in the H27 clade and transmitted vertically, although it exhibits considerable variation in some isolates. MinION-based sequencing of the pAA-ST131 plasmid in representative ESBL-producing EAEC ST131 strains identified an array of classical plasmid-encoded EAEC-defining virulence genes, including those encoding AAFs and the global regulator of virulence AggR. The AAFs play a central role in EAEC pathogenesis in part by facilitating adherence to the intestinal mucosa (20). Supporting a similar potential role in the EAEC ST131-H27 clade, a representative Danish ESBLproducing EAEC isolate harboring a full-length pAA-ST131 plasmid adhered significantly better to colonic epithelial cells than another closely related Danish outbreak strain harboring a truncated plasmid without the AAF-encoding genes.
The EAEC ST131 sublineage described here is reminiscent of an E. coli O78:H10 strain that caused a UTI outbreak in Copenhagen in 1991, in that both fulfilled the molecular criteria for EAEC and also contained multiple ExPEC virulence genes (34). That the O78:H10 outbreak strains' EAEC-associated virulence factors increased uropathogenicity (35) suggests that this is true also for EAEC ST131 strains. The O78:H10 EAEC outbreak strain, however, was never found outside the Capital Region of Denmark and was never identified as having caused BSI. In contrast, ST131 has proved highly successful globally and is a leading cause of BSI, which makes the potential of future outbreaks of UTI and bacteremia caused by this H27 sublineage, or other cases of EAEC ST131, a cause for serious concern.
Despite the genetic evidence that all 25 bla CTX-M-101 -containing Danish EAEC isolates were derived from a common ancestor, implying a common source, we have been unable to establish a patient link or explain the spread of the sublineage across regions using the limited available epidemiological data. From all but one of the 25 patients, an ESBL-producing E. coli was isolated from urine the same day that CTX-M-101-producing E. coli was detected from blood (data not shown). UTI is often caused by enteric E. coli strains that enter the urinary tract via the fecal-perineal-urethral route, and, in some instances, may have as their proximate external source food products or animals (41). The gender distribution among the present 25 cases (68% females) differs significantly (P ϭ 0.01 by Pearson's 2 test) from that across the entire collection of 258 ESBLproducing ST131 bloodstream isolates from Danish hospitals in 2014 to 2015 (42% females) (24). A similar overrepresentation of women was observed during the 2011 multinational European outbreak caused by a novel multipathotype, Shiga-toxinproducing EAEC O104:H4 strain (42), for which bean sprouts were the most likely vehicle of infection. Interestingly, the high proportion of female cases was thought to reflect the tendency of women to be more health-conscious, which, by analogy, suggests a food-related source for the Danish ESBL-producing EAEC ST31 outbreak. Further screening of, e.g., fecal and urine samples from patients who present with diarrhea or UTI, is warranted to determine the true clinical impact of ESBL-producing EAEC ST131-H27 strains and to provide the demographic and epidemiological data needed to identify potential sources.
In conclusion, we hypothesize that the acquisition of the pAA plasmid made the ESBL-producing EAEC ST131-H27 sublineage highly successful at persistently colonizing patients, thereby allowing it to occasionally cause UTIs and diarrhea. The presence of multiple ExPEC virulence factors, including P fimbriae, ␣-hemolysin, and Sat, in turn likely facilitates dissemination from the urinary tract to the bloodstream. Furthermore, we have demonstrated the ability of a non-H30 ST131 sublineage to acquire EAECspecific pAA virulence plasmids and to disseminate across multiple continents over the past 2 decades. Apart from virulence/colonization-associated genes, this EAEC ST131-H27 strain has also acquired bla CTX-M-101 and has caused BSI outbreaks in several geographic regions in Denmark seemingly associated with recurrent infections. These findings emphasize the potential for pathogens to evolve, potentially generating important new pathotypes that require continuous vigilance.

MATERIALS AND METHODS
Bacterial isolates, sequencing, and genomic data. WGS was performed on a previously described international collection of 128 ST131 E. coli human isolates of serotype O25:H4 or O25:H-(1968 to 2011) from the WHO Collaborating Centre for Reference and Research on Escherichia and Klebsiella (www.ssi.dk) (12). DNA samples were prepared for multiplexed, paired-end sequencing using a combination of Illumina MiSeq and HiSeq (6,12). We also included sequence data from a published E. coli ST131 data set comprising 93 U.S. and German isolates of human origin (1967 to 2011) (6) and sequences from 27 ESBL-producing ST131 bloodstream isolates (carrying bla CTX-M-101 ) collected from Danish patients within the national surveillance program for antimicrobial resistance (DANMAP) for ESBL-producing E. coli (2014 to 2015) (24). Furthermore, data from all Ͼ3,500 E. coli ST131 genomes available at EnteroBase (http://enterobase.warwick.ac.uk, accessed 15 November 2017) were analyzed for EAEC characteristics (presence of the aggR gene) by BLASTN analysis, and positive isolates were included. Finally, we sequenced and included 13 E. coli isolates obtained from urine from the source patients for eight of the ESBL-producing blood isolates. The sequences were analyzed using the Bacterial Analysis Platform (BAP) from the Center for Genomic Epidemiology (43).
Plasmid sequencing and analysis. Plasmid DNA was sequenced on both a MiSeq instrument (Illumina) and a MinION flow cell (Oxford Nanopore Technologies). The MiSeq library was made using the Nextera XT kit (Illumina), and sequencing was performed as a paired-end 250-bp run, yielding 372,720 reads with an average length of 237 bp. The MinION library was prepared using the Genomic Sequencing SQK-MAP006 kit and was sequenced on a FLO-MAP003 early access flow cell according to the manufacturer's instructions. Fast5 read files were subjected to base calling via a two-direction workflow using Metrichor software (ONT) to yield two-dimensional read files. Mixed assembly was performed by combining MiSeq and MinION reads using the SPAdes assembler (v3.9.0). Finally, CLCbio Genomics Workbench (v9.5.2) was used for end trimming of the assembled plasmid and for final error correction by mapping trimmed MiSeq reads against the plasmid contig obtained after the mixed SPAdes assembly.pAA-ST131 was annotated using RAST (44), with putative hypothetical genes curated manually using NCBI BLASTN and BLASTP searches. A BLASTN atlas of pAA-ST131 and other pAA virulence plasmids was constructed using BLAST Ring Image Generator v0.95 (BRIG) (45).

Identification of SNPs.
SNPs in the core chromosome or plasmid, depending on analysis, were identified using the Northern Arizona SNP Pipeline (NASP) (46). Briefly, duplicate regions of the reference chromosome of strain JJ1886 or pAA-ST131 plasmid (GenBank accession no. CP006784 and KY706108, respectively) were identified by aligning the reference against itself with NUCmer (47), followed by mapping of Illumina raw reads against the reference using the Burrows-Wheeler Aligner (BWA) (48) with identification of SNPs using GATK (49). Purging of recombinant regions in the chromosome-based SNPs was performed using Gubbins v2.2 (50).
Phylogenetic analysis. Relatedness between isolates according to core genome SNPs was inferred using RAxML v8.2.10 using the GTRCAT model (51) with 100 bootstrap replicates. Relatedness of the plasmids was inferred using PHyML with Smart Model Selection (52) with tree searching using subtreepruning-regrafting and 100 bootstrap replicates. Plasmid sequences from all 10 MinION sequences were analyzed for deletion and insertion using the progressiveMauve algorithm in Mauve v1.4.0 as implemented in Geneious Prime v2020.0.4 (Biomatters, Ltd.).
Bacterial growth conditions. E. coli strains were routinely cultured at 37°C on Luria-Bertani (LB) agar and in LB broth. For cell adhesion experiments, E. coli overnight cultures grown in LB broth were subcultured into Dulbecco's modified Eagle medium (DMEM) with 0.45% glucose (Invitrogen).
Cell adhesion assay. The human colon carcinoma-derived cell line T84 was maintained in DMEM with nutrient mixture F-12 (DMEM-F-12) containing 2 mM L-glutamine (ATCC) supplemented with 10% fetal bovine serum (FBS). For the adhesion assay, cells were grown to 80% confluence in 24-well tissue culture plates (10 5 cells/well), washed in DMEM (Invitrogen), and incubated with individual E. coli strains at a multiplicity of infection (MOI) of approximately 10 in DMEM for 1 h at 37°C. Quantification of cell-associated bacteria was performed by removing nonadhering bacteria, followed by washing and cell lysis. Serial dilutions of lysed cells with bacteria were plated and the number of CFU determined. Percent adherence was calculated by dividing the final number of CFU/ml by the initial number of CFU/ml for each replicate. Strain differences were calculated using a one-way analysis of variance (randomized design). Post hoc comparisons were performed using Tukey's test. Both statistical tests were calculated with R v3.4.1.
Data availability. The accession numbers for the Illumina sequences generated from the 134 E. coli ST131 isolates presented in this study are available in the European Nucleotide Archive (ENA; https:// www.ebi.ac.uk/ena) under accession number PRJEB27194. Sequences can also be located in the ENA using the following study summary: "Emergence of enteroaggregative Escherichia coli within the ST131 lineage as a cause of extraintestinal infections." The sequence of the pAA-ST131 plasmid has been deposited in GenBank under accession number KY706108.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.

ACKNOWLEDGMENTS
We thank Michala T. Sørensen at Statens Serum Institut for performing the plasmid gel electrophoresis. We also thank Karin Sixhøj Pedersen at Statens Serum Institut for her excellent WGS work.
This work was supported financially by the Danish Council for Research, grant DFF-1331-00161, to C. Struve, and the Danish Ministry of Health as part of the Integrated Surveillance of ESBL/AmpC-Producing E. coli and Carbapenemase-Producing Bacteria. This work was also supported in part by the Office of Research and Development, U.S. Department of Veterans Affairs, grant number 2I01CX000920-04, to J. R. Johnson.