Sequence-based genotyping of extra-intestinal pathogenic Escherichia coli isolates from patients with suspected community-onset sepsis, Sweden

Extra-intestinal pathogenic Escherichia coli (ExPEC) strains are responsible for a large number of human infections globally. The management of infections caused by ExPEC has been complicated by the emergence of antimicrobial resistance, most importantly the increasing recognition of isolates producing extended-spectrum β -lactamases (ESBL). Herein, we used whole-genome sequencing (WGS) on ExPEC isolates for a comprehensive genotypic characterization. Twenty-one ExPEC isolates, nine with and 12 without ESBL-production, from 16 patients with suspected sepsis were sequenced on an Illumina MiSeq platform. Analysis of WGS data was performed with widely used bioinformatics software and tools for genotypic characterization of the isolates. A higher number of plasmids, virulence and resistance genes were observed in the ESBL-producing isolates than the non-ESBL-producing, although not statistically significant due to the low sample size. All nine ESBL-producing ExPEC isolates presented with at least one bla gene, as did three of the 12 without ESBL-production. Multi-locus sequence typing analysis revealed a diversity of sequence types whereas phylogroup A prevailed among isolates both with and without ESBL-production. In conclusion, this limited study shows that analysis of WGS data can be used for genotypic characterization of ExPEC isolates to obtain in-depth information of clinical relevance.


Introduction
The bacterium Escherichia coli (E.coli) can colonize the human gut in an asymptomatic manner [1,2], but also emerge from this reservoir and cause extraintestinal infections, including severe invasive diseases such as sepsis and meningitis [3][4][5].Approximately 20% of all clinically relevant isolates isolated from the bloodstream are E. coli, making it one of the most frequently isolated bacteria in the bloodstream [3].Isolates that have the ability to reach and effectively colonize niches outside the gut are called extraintestinal pathogenic E. coli (ExPEC) [6].ExPEC strains are responsible for a significant number of infections in humans worldwide, both community-acquired and healthcare-related infections [6,7].Moreover, ExPEC strains are known for their ability to acquire novel and troubling resistance genes [8].Increasing numbers of ExPEC infections in combination with increasing antimicrobial resistance will make the future management of E. coli infections both more challenging and costlier [9].Of particular concern is the increased prevalence of extended-spectrum β-lactamase (ESBL)-producing E. coli, which has resulted in an increased use of the "last-resort" antimicrobial drugs, typically carbapenems.ESBL genes, commonly bla CTX-M , bla TEM , and bla SHV , are harbored by self-transmissible conjugative plasmids that are horizontally shared within and between bacterial species [10].Although most reports on ESBL genes during the 1990s concerned bla TEM , and bla SHV , the recent global increase has been caused primarily by bla CTX-M .This genotype appears to be the most prevalent in all continents [11][12][13][14] and is often associated with a widely distributed variant of E. coli sequence type (ST) 131.
DNA profiling of ExPEC can be performed by different molecular technologies such as multi-locus sequence typing (MLST), amplified fragment length polymorphism (AFLP) and multiplex PCR.Indeed, these molecular techniques have advanced our understanding of ExPEC lineages by categorizing isolates into STs and phylogenetic groups.Today, ST131 is the predominant E. coli sequence type among ExPEC isolates worldwide [15] and it is associated with urinary tract infections (UTI), but also bladder and kidney infections as well as urosepsis [16].ST131 isolates are commonly reported to produce ESBL [17] with CTX-M-15 as the predominant CTX-M type [18][19][20].International studies have also reported the presence of other sequence types; ST69, ST73, and ST95 among large collections of ExPEC strains derived from human infections [21][22][23][24].ExPEC can also be sorted into four major phylogenetic groups A, B1, B2, and D [25] by using a multiplex PCR method [26,27].Multiplex PCR analysis of ExPEC isolates has shown that the virulent isolates mainly belong to group B2 and to a lesser extent, group D, whereas most commensals and less virulent strains belong to group A or B1 [27][28][29][30][31][32][33][34].
An increase in ESBL-producing Enterobacteriaceae has been observed in Sweden until 2020 [35][36][37], when the pandemic of covid-19 started.The many measures aimed at reducing the spread of covid-19 taken in Sweden also affected the spread of other transmissible diseases [37].In addition, travel and migration decreased.Brolund et al. [35,36] found that the increase in ESBL-producing Enterobacteriaceae noted in Sweden during 2007-2011 was largely accounted for by E. coli isolates.Triplex PCR analysis of 913 ESBL-producing E. coli isolated from patients with urinary tract infections found the phylogenetic group B2 in 41-47% of the isolates, group D in 28-29%, group A in 15-21% and B1 in 9-10%.A majority harbored the bla CTX-M-15 resistance gene, 34-38% of the isolates were ST131 of which 90-98% belonged to the B2 phylogenetic group.The typing of plasmids, containing the β-lactamase genes, should be carried out when performing epidemiological studies of ESBL-producing E. coli isolates [36].The plasmids often contain mobile genetic elements that may cause resistance genes, and other pathogenic markers, to move between plasmids and/or chromosomes.The most common plasmid in ESBL-producing E. coli is the IncF, belonging to a common group of classified plasmids.A high number of the ST131 are known to harbor the plasmid IncFII, which may be one reason why this clone is so widespread.
Whole-genome sequencing (WGS) of bacteria by high-throughput sequencing (HTS) technology is now widely used in microbiological research as a methodology for rapid molecular typing, providing large scale data of the virulence and genetic characteristics of bacterial strains.In a previous study, comparative genomics were used to study resistomes and virulomes in E. coli isolates from skin, soft tissue and extraintestinal infections [38].HTS of E. coli has also been proved as a powerful technique in several studies for outbreak investigations and surveillance.The technique was successfully used to investigate and genotype a multidrug-resistant E. coli outbreak in a neonatal unit at a hospital in Australia [39], in an outbreak of ESBL-producing E. coli in a nursing home in the Netherlands [40] and for routine typing, surveillance and outbreak detection of verotoxigenic E. coli isolates in Denmark [41].The aim of this limited study was to investigate and characterize ExPEC isolates by WGS-based analysis using bioinformatics tools.A total of 21 clinical ExPEC isolates, whereof nine with ESBL-production, isolated from different types of culture specimens collected during a prospective, consecutive sepsis study in Sweden [42], were included.

Epidemiological setting and isolate collection
From September 2011 to June 2012, a prospective observational study of community-onset severe sepsis and septic shock in adults was conducted at Skaraborg Hospital, a secondary hospital with 640 beds, in the western region of Sweden [42].All patients ≥18 years consecutively admitted to the emergency department for suspicion of community-onset sepsis were asked to participate in the study.The study was approved by the Regional Ethical Review Board of .As the present study only focused on bacterial isolates recovered from cultures included in the routine patient care, no individual written consent was needed.During the prospective sepsis study, 1,827 pathogenic bacterial isolates were recovered from various types of culture specimens at extra-intestinal locations.These isolates were cryopreserved at the time of recovery by transferring colonial material to MicrobankTM vials (Pro-Lab Diagnostics, Ontario, Canada) stored at − 80 • C. E. coli was the most common bacterial finding with isolates, of which nine were phenotypically classified as ESBL-producing E. coli by conventional laboratory methods.For the present study, E. coli isolates, nine ESBL-producing E. coli and 12 E. coli, from 16 patients with suspected sepsis, were selected.Out of these patients, 15 had a final clinical diagnosis of pyelonephritis, and in one the diagnosis was decubitus with cellulitis.Patients with non-ESBL-producing E. coli isolates were as closely matched as possible to the patients with ESBL-producing E. coli infections according to sex, age, sampling time, and clinical diagnosis.Four patients had two isolates from different types of culture specimens and one patient had two isolates from the same type of culture specimen collected at different sampling time points, i.e., separate episodes.Eleven patients had one isolate.The isolates were recovered in specimens collected from urine (n = 15), blood (n = 4), and wound (n = 2).

Species identification by MALDI-TOF MS
Conventional typing of bacterial cultures and definite species identification with MALDI-TOF MS was performed on a Microflex LT mass spectrometer (Bruker Daltonics, United States) with BioTyper software v2.0 using default parameter settings as part of the routine clinical practice as described elsewhere [43].Spectral scores above 2.0 were used as a cut-off for correct species identification.At the time of the study, the Bruker microorganism database MBT Compass Library DB-4110 (Bruker Daltonics, Germany) released in April 2011 was used.

Phenotypic antibiotic susceptibility testing
Antibiotic susceptibilities were determined by accredited laboratory methods using the disc diffusion method on Mueller-Hinton media according to European Committee on Antimicrobial Susceptibility Testing (EUCAST) guidelines (https://eucast.org/).

Whole-genome sequencing
Isolates were cultured according to conventional methods.Colonies were picked and diluted in 1 mL of phosphate buffered saline.DNA was extracted on MagnaPure Compact (Roche Diagnostics, Switzerland) with the MagnaPure Compact Nucleic Acid Isolation kit I -large volume and the protocol Total_NA_plasma_1000_v3_2. A sample volume of 1 mL and an elution volume of 100 μL was used.A BioSpec-nano (Shimadzu, Japan) was used to test the purity of the extracted nucleic acids.DNA extracts were kept at − 20 • C until transport on dry ice to Clinical Genomics, SciLifeLab, Solna, Sweden (https://clinical.scilifelab.se/).Sample preparation was performed following the guidelines in the Nextera XT DNA sample preparation guide (Illumina, United States).Double-stranded DNA measurements were performed with broad-and low range assay kits on a Qubit (Thermo Fisher Scientific, United States).Extracted E. coli DNA available at SciLifeLab was used as a positive control in parallel with the study samples throughout the sequencing process.Fragment analyses were performed with a Bioanalyzer (Agilent, United States) on the 21 PCR libraries to obtain abundances and the average length of fragments for each sample before pooling.Library preparation, which consisted of enzymatic tagmentation, PCR clean-up and index-ligation, was performed according to the Nextera XT guidelines (Illumina, United States).Samples were sequenced on the Illumina MiSeq platform at 2 × 300 base pair read length and a 1% PhiX spike-in was used for error estimation.

Quality trimming and assembly
Output Bcl-files were demultiplexed into fastq-files.The tool FastUniq [44] was used to remove exact duplicate reads and SeqPrep [45] was used thereafter to remove adapters and for quality trimming with default parameters.The sequencing quality was assessed using FastQC [46].Two assemblers, SPAdes genome assembler v3.5.0 [47] and RAY v2.3.1 [48] were separately used to assemble the read fragments into contigs.One dataset of SPAdes-assembled contigs was produced for each isolate for the total genome, and three datasets of RAY-assembled contigs were produced for each isolate; total genome, chromosomal and plasmids.Downstream analyses using bioinformatic tools hosted by Center for Genomic Epidemiology (CGE, https://cge.cbs.dtu.dk) and JSpeciesWS [49] were performed on the assembled contigs from both assemblers, while other analyses were performed only on the SPAdes-assembled contigs.

Sequence data analysis
Species identification was performed using the tool SpeciesFinder v.1.0hosted by the Center for Genomic Epidemiology (CGE, https://cge.cbs.dtu.dk)website, as well as by calculating the pairwise average nucleotide identity (ANI) based on BLAST+ (ANIb) in JSpeciesWS [49] against the reference genome E. coli IAI39 (GenBank accession number NC_011750.1).An ANI threshold of 96% or greater was considered to delineate species boundaries as a threshold of 96% correlates well to DNA-DNA hybridization [50,51].
The presence of plasmid replicons, acquired virulence genes, and acquired antibiotic resistance genes, including both chromosomal and plasmid-borne, were detected using the tools PlasmidFinder v.1.2,Vir-ulenceFinder v.1.2,and ResFinder v.2.1 available at the CGE website.For PlasmidFinder, the database "Plasmid enterobacteriaceae" was selected and 95% threshold for identity.Plasmid DNA sequences were then classified by PlasmidFinder based on the Inc-group-based classification framework.For VirulenceFinder, species was selected as E. coli and the threshold for identity was 98%.For ResFinder, the threshold for identity was 98%, and the selected minimum length 60%.In collaboration with 1928Diagnostics (Gothenburg, Sweden), additional investigation of the presence of resistance genes was made against the resqu resistance database v1.1.For this part of the analysis, adapter removal and quality trimming were made using Trimgalore! v0.4.0 [52].The read quality after processing was evaluated using FastQC [46].SPAdes v3.6.0 [47] was used to assemble the processed reads into contigs.Genes were identified by querying the assembled contigs using Blastx + v2.30 [53] against resqu database v1.1 (http://www.1928diagnostics.com/resdb) with a threshold e-value of 1e − 50 and filtering of low complexity regions disabled.Hits with 100% alignment coverage and >98% sequence identity were considered to be a positive evidence of a present resistance gene.
The CGE tools snpTree and CSIPhylogeny v.1.0were used for determination of single-nucleotide polymorphisms and construction of a phylogenetic tree.For snpTree, a reference genome, urinary tract infection isolate E. coli IAI39 (GenBank accession number NC_011750.1)with a complete genome size of 5.13 Mb, was used for alignment.CSI-Phylogeny was run with default parameters.
Phenotypes of the isolates were also predicted by the use of Traitar open access software package [55] for analysis of microbial traits.The analysis includes 67 different phenotypic traits divided into microbiological or biochemical categories such as enzyme (production), growth (conditions, amino acid, carboxylic acid, glucose, sugar), morphology (Gram-negative, Gram-positive, coccus, motility, spores, yellow pigment), oxygen (aerobe/anaerobe, catalase/oxidase enzyme), product (hydrogen-sulphide) and proteolysis (casein or gelatine hydrolysis) as examples.For phenotyping of each isolate, two different algorithms (phypat and phypat + PGL), were used for prediction by the software.
Phylogenetic group typing of the E. coli isolates was manually performed on the sequence data in assembled contigs by using available primer sequences for the genes gadA, chuA, yjaA and TSPE4.C2 [27,34].The E. coli reference genome UMN026 with a genome size of 5.2 Mb (GenBank accession number NC_011751.1)isolated from a urinary tract infection was used for comparison.

In silico gene prediction to infer antibiotic resistance
The applicability of in silico gene prediction was assessed by comparing the antibiotic resistance profile inferred by the results from ResFinder and resqu database to the phenotypically determined antibiotic resistance pattern.The resistance spectrum for a given gene was estimated through literature studies.Only publications that fulfilled the following criteria were included: the antibiotic resistance in E. coli shall be reported; the sequence shall be reported; the resistance spectrum shall be determined through cloning of the gene into a susceptible bacterium.In cases where no good publications for a given antibiotic resistance gene was found, another publication fulfilling the same criteria for the parent gene was accepted.This applied mainly to β-lactamases (bla).

Statistical analysis
Statistical analyses were performed using R v.4.0.5 [56] (R Foundation for Statistical Computing, Austria).All tests were two-sided, and p < .05 was considered statistically significant.Poisson and quasi-Poisson regression analysis were used for comparisons of count data, and adjustment of p-values for multiple comparisons was made using the Holm method.Dispersion test was performed using the R package AER v.1.2-9.Boxplots were constructed using the R package ggplot v.3.3.5 [57].

Results
The HTS yielded approximately 25 million paired-end reads which, divided among the 21 isolates, equals a theoretical average coverage of 178X for each sample.All 21 bacterial isolates were approximately 5 Mb/bacterial isolate.The average number of contigs for each bacterial genome, including both chromosome and plasmids, was 626 contigs per isolate with the longest contig of 94,000 base pairs.SPAdes and RAY total genome contig assemblies gave identical results in all subsequent analyses performed at CGE and JSpeciesWS.
Prediction of bacterial species based on the 16S ribosomal DNA sequence was performed using the tool SpeciesFinder.In seven isolates, SpeciesFinder gave a clear result for E. coli.In the other isolates, E. coli could not be distinguished from Shigella.The species determination based on the calculation of the pairwise ANI against a reference genome identified all 21 isolates as E. coli consistent with the results from the species identification with MALDI-TOF MS.
An overview of the characteristics of the 21 isolates used in this study is shown in Table 1.Pairs of isolates from the same patient, although different types of culture specimens, gave almost identical results for identified plasmid replicons, resistance and virulence genes, and MLST sequence type.Three urine isolates (no.4, 18, and 19) were classified as ST131, and two urine isolates (no.8 and 11) were classified as ST69.No significant differences were observed in the number of plasmids and resistance genes between the two groups of isolates, non-ESBLproducing E. coli and ESBL-producing E. coli (Fig. 1A).In addition, no significant differences between the number of plasmids, virulence or resistance genes were found in isolates from different types of culture specimens (Fig. 1B).
All isolates gave sequence types by MLST #1.Five of the isolates showed an unknown sequence type by MLST #2 (data not shown).The sequence types obtained from MLST#1 of the paired isolates from the same five patients gave identical results.The two urine isolates collected from the same patient, but at different sampling times, also gave identical results.Typing by MLST#1 and phylogenetic grouping of all isolates are presented in Table 1.Phylogenetic group A was most common among the isolates in this study, 6/9 ESBL-producing E. coli and 8/12 non-ESBL-producing E. coli.The phylogenetic types of the pairs of isolates from the same patients gave identical results, except for the two ESBL-producing E. coli isolates from blood and urine (no. 5 and 6).These two isolates also differed in the number of plasmids and resistance genes, as the urine isolate showed antibiotic multi-resistance.
Plasmids were present in 8/9 ESBL-producing E. coli isolates and in 6/12 non-ESBL-producing E. coli (Table 2).In 6/9 ESBL-producing E. coli, the FII plasmid was found, but only in 1/12 non-ESBLproducing E. coli.In the ST131 ESBL-producing E. coli isolate (no.4), this plasmid was present, but it was not found in the two non-ESBLproducing E. coli ST131 isolates (no.18 and 19).All the detected virulence genes are presented, per isolate and phylogenetic group, in Fig. 2.
The four most common virulence genes, iss, sat, gad and prfB, corresponded well with what was expected to be present in uropathogenic E. coli.
A summary of the β-lactamase (bla) genes detected by ResFinder and resqu is presented in Fig. 3A.The nine phenotypically classified ESBLproducing E. coli isolates all presented with one or two bla genes.Among the 12 non-ESBL-producing E. coli isolates, three were shown to harbor one bla gene from the TEM family.These three strains also presented with a higher number of other resistance genes.In seven out of the 12 non-ESBL-producing E. coli isolates, no resistance genes were predicted.Of these seven isolates, six showed no plasmid replicons by PlasmidFinder.Overall, ResFinder and resqu identified the same genes with one exception, for isolate no. 8 resqu reported blaCMY-59 where ResFinder reported blaCMY-2.The resistance spectrum inferred by the antibiotic genes identified by resqu and ResFinder were compared to the phenotypic antibiotic resistance (Fig. 3B).Overall, the inferred resistance correlated well to the phenotype with no false positives.Both databases had difficulty predicting the gene causing ciprofloxacin resistance, where the resistance in only one of eight isolates could be explained.
For five patients, two samples each were included in the study (Table 1).The paired samples originating from the same patient were For all of these sample pairs, except for the sample pair derived from patient 5 (i.e., isolates no. 5 and 6), identical resistance genes were detected using ResFinder and resqu.Isolate no.6 contained one additional plasmid and three additional resistance genes (bla TEM-1 , aph(3')-I and aph( 6)-I) when compared to isolate no. 5.For snpTree analysis, 66.5% of the reference genome (E. coli IAI39) was covered by all study isolates, resulting in 3.4 Mb common positions.For each isolate, 33-35,000 single nucleotide polymorphisms (SNPs) were observed compared to the reference genome and these were used for the construction of a phylogenetic snpTree (Fig. 4).The paired isolates clustered together in the tree, indicating similar detected SNPs in the chromosomal data between the pairs.No clustering was observed between the nine ESBL-producing E. coli isolates.
Results for the 67 predicted phenotypic microbial traits by the software Traitar are presented in Fig. 5.In all 21 E. coli isolates, 33% (22/67) of the traits were predicted to be present by both algorithms as indicated in the heatmap.These were typical E. coli traits such as bacillus or coccobacillus morphology, Gram-negative, catalase and facultative growth on ordinary blood agar.In total, 43% (29/67) of the traits were not predicted in any of the isolates.Among these traits were coccus morphology, spore formation, yellow pigment, oxidase, aerobe and anaerobe.For 24% (16/67) of the traits, the prediction differed between the isolates; traits were present in some isolates by one or both algorithms, or absent.Some of these traits were motility, hydrogen sulphide and acetate utilization.Among the paired isolates from five patients, only one pair gave exactly the same prediction of traits (patient no. 5, isolates no. 5 and 6).

Discussion
In this study, we have characterized a limited set of clinical E. coli isolates using the HTS analysis pipeline.Furthermore, we compared the genomic makeup of ESBL-producing E. coli and non-ESBL-producing E. coli collected from patients with suspected community-onset sepsis in the western region of Sweden.Overall results indicated that the number of plasmids, virulence and resistance genes were higher in ESBLproducing E. coli than non-ESBL-producing E. coli (Fig. 1, Table 1), although not statistically significant due to the low sample size.This result is consistent with a previous study reporting a higher rate of virulence and resistance genes in ESBL-producing E. coli compared with non-ESBL-producing E. coli [58].
Regarding E. coli infections, the ESBL-producing E. coli is becoming more common among community-acquired isolates [59].The extent of entry of multidrug-resistant E. coli from the community into the hospital, and subsequent clonal spread amongst patients is still unclear.
Recent studies suggest that multidrug resistant E. coli are mostly acquired in the community and are not replaced by hospital-associated clones in hospitalized patients [60].In this limited study, no conclusions regarding the transmission of resistance can be made, but the community-acquired E. coli in this region of Sweden are probably still susceptible to most antibiotics.The only β-lactamase (bla) gene identified by ResFinder among non-ESBL-producing E. coli isolates was bla-TEM-1B (Fig. 3B).
The phenotypic microbial traits as predicted by Traitar did not differ  between the nine ESBL-producing ExPEC isolates and the 12 non-ESBLproducing isolates (Fig. 5).Several phenotypic characteristics commonly seen in E. coli, such as Gram-negative, facultative, growth on MacConkey agar, nitrate-to-nitrite conversion, and indole production [61], were predicted for all isolates regardless of ESBL-production or not.Phylogenetic group A was most common among the isolates in this study when using PCR-primers to define the group directly on the sequence data in assembled contigs.This result is contradictory to several other studies conducted in Europe and the USA using PCR on bloodstream infections caused by E. coli, where >70% of the isolates belonged to phylogenetic groups B2 and D [62][63][64][65][66][67].Similar patterns have been reported in another Swedish study, looking into the major trends of epidemiology concerning Swedish ESBL-producing E. coli clinical isolates during 2007-2012 [35].Nevertheless, our study focused on E. coli isolated from patients with urinary tract infections, pyelonephritis, that has led to severe symptoms and suspicion of sepsis.In Brolund et al. (2014), there was no information regarding the severity of the patients experiencing urinary tract infections [35].However, WGS of 312 blood-or urine-derived isolates of ExPEC in the USA showed phylogenetic group distribution A 8.3%, B1 8.7%, B2 65.7% and D 16.3% [68].Yet, another European study identified A as the second most common phylogenetic group among E. coli isolated from adult patients with sepsis [69].Actually, group A has been found with increased frequency in nosocomial blood stream infections and sepsis in compromised hosts [63,65,70].E. coli from groups A and B1 with low virulence capacity were still able to cause extraintestinal infections, sepsis and pyelonephritis.However, our contradictory results may be due to the low number of ESBL-producing E. coli isolates in this study, and may not necessarily reflect the phylogenetic group distribution in Sweden.1).The phylogenetic tree is drawn to scale with branch lengths measured in the number of substitutions per site.The scale axis is provided below the tree.Globally, ST131 is the predominant E. coli lineage among ExPEC isolates [17].In the present study, three ST131 isolates were identified, one belonging to group A (isolate no.4) and two belonging to group B1 (isolates no.18 and 19).This is partly in contrast to previous studies in Sweden which reported that group B2 is the most prevalent among ST131 isolates while the other phylogenetic groups A, B1 and D occur to a lesser extent [35,71].
Four isolates in this study, two ESBL-producing E. coli and two non-ESBL-producing E. coli, were classified as phylogenetic group D. The phylogenetic groups of isolates from the same patients gave identical results, except for the two ESBL-producing E. coli isolates from blood and urine (isolates no. 5 and 6).These two isolates also differed in the number of plasmids and resistance genes, as the urine isolate showed multi-resistance (Table 1).
A study from United Kingdom identified E. coli sequence type ST69, ST73, ST95 and ST131 as the most common among isolates from patients with urinary tract infections and sepsis [72].The strains differed markedly as either susceptible to all tested antibiotics (45%), or resistant to amoxicillin only (43%).Overall, multi-resistance was most prevalent in E. coli ST69, with 79% of the isolates predicted to be resistant to at least three classes of antibiotics.The present study identified two multi-resistant ST69 isolates, one ESBL-producing E. coli and one non-ESBL-producing E. coli, presenting with 11 and 7 resistance genes, respectively (Table 1).
The population structure of drug-susceptible, -resistant and ESBLproducing E. coli from community-acquired urinary tract infections has also been studied in Denmark.Overall, it was found that ST131, ST73 and ST69 were dominating types among all isolates, in accordance with previous studies [73].In Sweden, ST131 is also the major sequence type of ExPEC isolates [35].Other sequence types were ST38, ST69, ST405, ST617 and ST648, each representing 2-6% of the isolates.In this study, three ST131 urine isolates were found, but more than ten other sequence types were also observed which indicate that different sequence types may cause clinical disease and pyelonephritis.
The main weaknesses of this study are the small number of isolates and the limited geographical area.Although our results therefore should be interpreted with caution, they are still interesting as there are not many studies comparing the WGS of ESBL-producing and non-producing ExPEC isolates from humans.

Conclusions
Our results indicate the utility of WGS-based characterization of ExPEC isolates to obtain in-depth information such as resistance patterns, virulence genes, phylogenetic groups, sequence types, etc.Despite the limited sample size, we could observe a higher number of plasmids, virulence and resistance genes in ESBL-producing ExPEC isolates than in non-ESBL-producing when comparing the genomic makeup.

Fig. 2 .
Fig. 2. Virulence genes identified by VirulenceFinder in clinical ExPEC isolates (n = 21).The absence or presence of the gene is indicated as follows: blank cell -the gene is not detected; • the gene is detected ≥98 sequence similarity; • the gene is detected with 100% sequence similarity.Isolates no.1-9 were phenotypically classified as ESBL-producing ExPEC, indicated by dark grey columns in the figure, while isolates no.10-21 were classified as non-ESBLproducing ExPEC, indicated by pale grey columns.

Fig. 3 .
Fig. 3. Resistome and phenotypic antibiograms.A. Resistome showing resistance genes for each isolate detected on WGS-data by ResFinder and resqu.Pale greyresistance gene not detected by neither ResFinder nor resqu, orangeresistance gene detected by ResFinder only, greyresistance gene detected by resqu only, dark redresistance gene detected by both ResFinder and resqu.At the top, black dots indicate the antibiotics to which each gene confers resistance.B. Antibiograms, or antibiotic susceptibility pattern, along with ST for each of the 21 isolates sequenced.Blank cells indicate that the specific antibiotic has not been phenotypically tested for that isolate.R resistant, S susceptible, I indeterminate resistant, ST sequence type, AMP ampicillin, CFR cefadroxil, CTX cefotaxime, CAZ ceftazidime, CTB ceftibuten, CIP ciprofloxacin, CLI clindamycin, ERY erythromycin, FOF fosfomycin, FLX flucloxacillin, CHL chloramphenicol, MEC mecillinam, MEM meropenem, NIT nitrofurantoin, PCV penicillin V, TZP piperacilllin-tazobactam, TOB tobramycin, SXT trimethoprim-sulphamethoxazole, TMP trimethoprim.

Fig. 4 .
Fig.4.Phylogenetic tree of the 21 E. coli isolates and the reference strain (E. coli IAI39) as estimated by snpTree on chromosome datasets.The reference strain E. coli IAI39 is drawn in red, non-ESBL-producing E. coli in black, ESBL-producing E. coli in green.The samples are indicated by isolate number (see also Table1).The phylogenetic tree is drawn to scale with branch lengths measured in the number of substitutions per site.The scale axis is provided below the tree.

Table 1
Summary of the main characteristics of the 21 E. coli isolates included in the study.
a Isolates no.1-9 are phenotypically classified as ESBL-producing E. coli, whereas isolates no.10-21 as non-ESBL-producing E. coli.b Threshold for sequence identity.c Isolates collected from the same patient at the same time point, but from different body locations.d Urine isolates collected from the same patient at two different time points, i.e., two separate episodes.Fig. 1. A. Number of plasmids, virulence genes, and resistance genes identified in ESBL-producing E. coli (n = 9) and non-ESBL-producing E. coli (n = 12).No significant differences between E. coli ESBL and non-ESBL-producing E. coli were detected using either Poisson or quasi-Poisson regression followed by adjustment of the resulting p-values by the Holm method (all values of p > .05).B. Number of plasmids, virulence genes, and resistance genes identified by culture specimen type, i.e., urine (n = 15), wound (n = 2), and blood (n = 4).No significant differences between culture specimen types were detected using either pairwise Poisson or quasi-Poisson regression followed by adjustment of the resulting p-values by the Holm method (all values of p > .05).collected either from different body locations or at different episodes.

Table 2
Plasmids identified by PlasmidFinder by isolate no. and sequence type.