Characterizing carbapenemase-producing Escherichia coli isolates from Spain: high genetic heterogeneity and wide geographical spread

Introduction Carbapenemase-Producing Escherichia coli (CP-Eco) isolates, though less prevalent than other CP-Enterobacterales, have the capacity to rapidly disseminate antibiotic resistance genes (ARGs) and cause serious difficult-to-treat infections. The aim of this study is phenotypically and genotypically characterizing CP-Eco isolates collected from Spain to better understand their resistance mechanisms and population structure. Methods Ninety representative isolates received from 2015 to 2020 from 25 provinces and 59 hospitals Spanish hospitals were included. Antibiotic susceptibility was determined according to EUCAST guidelines and whole-genome sequencing was performed. Antibiotic resistance and virulence-associated genes, phylogeny and population structure, and carbapenemase genes-carrying plasmids were analyzed. Results and discussion The 90 CP-Eco isolates were highly polyclonal, where the most prevalent was ST131, detected in 14 (15.6%) of the isolates. The carbapenemase genes detected were bla OXA-48 (45.6%), bla VIM-1 (23.3%), bla NDM-1 (7.8%), bla KPC-3 (6.7%), and bla NDM-5 (6.7%). Forty (44.4%) were resistant to 6 or more antibiotic groups and the most active antibiotics were colistin (98.9%), plazomicin (92.2%) and cefiderocol (92.2%). Four of the seven cefiderocol-resistant isolates belonged to ST167 and six harbored bla NDM. Five of the plazomicin-resistant isolates harbored rmt. IncL plasmids were the most frequent (45.7%) and eight of these harbored bla VIM-1. bla OXA-48 was found in IncF plasmids in eight isolates. Metallo-β-lactamases were more frequent in isolates with resistance to six or more antibiotic groups, with their genes often present on the same plasmid/integron. ST131 isolates were associated with sat and pap virulence genes. This study highlights the genetic versatility of CP-Eco and its potential to disseminate ARGs and cause community and nosocomial infections.


Introduction
Antibiotic-resistant Escherichia coli is one of the principal pathogens that can cause community-acquired and nosocomial infections with increased morbidity and mortality rates (Hu et al., 2022).They are considered as one of the principal causes of Urinary Tract Infections (UTIs), bacteremia, and Intra-Abdominal Infections (IAI) (Balasubramanian et al., 2023).E. coli have a heightened ability to acquire Antibiotic-Resistance Genes (ARGs), such as the bla CTX-M-15 Extended Spectrum b-Lactamase (ESBL), and rapidly disseminate them throughout the community (Gonzaĺez et al., 2020).Carbapenemase-producing Escherichia coli (CP-Eco), though not as frequently isolated in the clinical setting as compared to other CP-producing Enterobacterales (such as Klebsiella pneumoniae and Enterobacter cloacae complex), is especially worrying.This is because they are increasing in prevalence (Cañada-Garcıá et al., 2022), creating a concern that they would be able to spread carbapenemase genes in the community in a similar manner as it was observed for ESBLs (Gonzaĺez et al., 2020).Moreover, these isolates are typically resistant to several other antibiotics, making their related infections difficult to treat (Boutzoukas et al., 2023).
All main carbapenemase families have been detected in CP-Eco (Grundmann et al., 2017), in addition to a wide range of virulence determinants that negatively affect clinical outcome (C ̌urováet al., 2020).All this led to the declaration by the World Health Organization that CP-Eco are a critical priority issue (Tacconelli et al., 2018).Globally, antibiotic-resistant E. coli has the highest incidence of hospital-acquired infections in high-to-mid income countries, causing three to twenty-five million infections per year (Balasubramanian et al., 2023).In Europe, the median number of infections caused by CP-Eco was 2,619 in 2015, with a median of 141 attributable deaths (Cassini et al., 2019).In Spain, the incidence of CP-Eco has evolved from isolated cases in 2013 (Oteo et al., 2015), to being detected in 10 different Spanish provinces in 2019 (Cañada-Garcıá et al., 2022).
Given the importance and the widespread dissemination of CP-Eco, the aim of this study is the phenotypic and genotypic characterization of a set of such isolates obtained from different Spanish regions over a period of five years.

Study design and bacterial collection
As part of the ongoing Antibiotic Resistance Surveillance Program, the Spanish National Center for Microbiology (CNM) receives clinical carbapenem-resistant E. coli isolates for molecular characterization from Spanish hospitals.Participation in this program is voluntary and the criteria to send the isolates to the CNM is having a Minimum Inhibitory Concentration (MIC) to meropenem of over 0.12 mg/L (Giske et al., 2017).
Of the total CP-Eco isolates received in this Surveillance Program between January 2015 and December 2020, a subset was selected for further analysis according to the following criteria: i) one isolate for each type of carbapenemase per year was included from each of the collaborating hospitals, ii) in the event that there was more than one isolate for each carbapenemase type in the same year, the selection was prioritized based on the type of infection in which they were involved according to the following order: 1) invasive infections 2) other infections 3) colonizations; iii) in all cases, the first isolate received per year was selected.These selection criteria allowed for a temporalspatial representative sample of the different types of carbapenemases circulating in E. coli in Spain.Not all hospitals had isolates from all the five years, and from some hospitals more than one isolate per year was included since they produced different carbapenemase types.The information sent along with the isolate from the participating hospitals included collection date, age and sex of the patients, sample type, whether the isolate was a colonizer or causing infections (as determined by the attending microbiologist), and isolate origin [hospital, healthcare center, or community setting; i.e.where the infection is contracted without interaction with healthcare settings during the last 48 hours (Plachouras et al., 2018)].

DNA extraction and identification of carbapenemase genes
All the carbapenem-resistant E. coli isolates received at the CNM from 2015 to 2020 were cultured on MacConkey agar (Becton Dickinson Microbiology Systems, Cockeysville, MD, USA) and incubated overnight at 37°C.DNA was extracted using the QIAamp ® DNA Mini Kit (QIAGEN, Hilden, Germany) according to the manufacturer's instructions.Real-time PCR assays targeting the carbapenemase gene families bla OXA-48 , bla KPC , bla VIM , bla IMP and bla NDM were performed (Ortega et al., 2016).All the isolates were stored at -80°C until used.

Antibiotic susceptibility testing
Antibiotic Susceptibility Testing (AST) was determined for the selected isolates with broth microdilution using the YDKMGN Sensititre ™ Gram Negative panels (Thermo Fisher, Waltham, MA, USA) (International Organization for Standardization (ISO), 2006).Etests using meropenem/vaborbactam and imipenem/relebactam and disc diffusion assays for cefiderocol (Liofilchem, Roseto degli Abruzzi, Italy) were performed and interpreted according to the EUCAST guidelines (European Society of Clinical Microbiology and Infectious Diseases (EUCAST), 2022).Additionally, E-tests using plazomicin (Liofilchem, Roseto degli Abruzzi, Italy) and fosfomycin (bioMeŕieux, Marcy l'Etoile, France) strips were performed and interpreted according to an FDA-approved breakpoint of susceptibility of ≤2mg/ L for plazomycin (Cañada-Garcıá et al., 2022), and according to the EUCAST cutoff values for urinary tract infections (UTI) for fosfomycin due to lack of clinical breakpoints for other types of samples (European Society of Clinical Microbiology and Infectious Diseases (EUCAST), 2022), respectively.

Whole-genome sequencing and read assembly
DNA was extracted from the selected isolates as described in section 2.2.Paired-end (2x150) libraries were then prepared using the Nextera DNA Flex Library Preparation Kit and sequenced using Illumina NextSeq 550 (Illumina Inc., San Diego, CA, United States) according to the manufacturer's instructions.Quality of the reads was assessed using FASTQC (version 0.11.9),followed by de novo assembly using Unicycler (version 0.4.8)(Wick et al., 2017).The quality of the assembly was assessed using QUAST (version 5.2.0) and annotation was performed using Prokka (version 1.14.6)(Seemann, 2014).The obtained assemblies were deposited in the European Nucleotide Archive under the accession number PRJEB70795.

Phylogenetic analyses and diversity
Ridom SeqSphere+ (version 8.3.1;Ridom, Münsten, Germany) was used to perform a core-genome Multi-Locus Sequence-Typing analysis (cgMLST) using a built-in scheme for E. coli containing 2,515 core genes, and to construct a minimum spanning tree based on allelic differences.ARIBA (version 2.6.2) (Hunt et al., 2017) was used to determine STs in accordance with the University of Warwick scheme.Diversity of the samples was calculated using a Simple Diversity Index [SDI (Gastmeier et al., 2006)].Analysis of the fimH gene was done to subclassify ST131isolates using the FimTyper tool [CGE server; https://cge.cbs.dtu.dk;(accessed 23/ 02/2024)].

Antibiotic resistance genes, virulenceassociated genes, and plasmids
ARGs were analyzed by ARIBA (version 2.14.6) (Hunt et al., 2017) using the CARD database (https://card.mcmaster.ca;(accessed 23/02/2024)) and ResFinder (CGE server) with ID thresholds of 100% for b-lactamase variants and 98% for other resistance genes.In isolates resistant to cefiderocol, the sequences of the genes encoding PBP3 and the siderophore CirA, previously related to resistance to this antibiotic (Wang et al., 2022), were analyzed with ARIBA.

Statistical analyses
The differences in prevalence of the ARGs, virulence-related genes, STs, and sample characteristics were evaluated using the Fisher's exact test using GraphPad Prism (version 3.02; GraphPad Software, Inc., San Diego, CA, USA).P values of less than 0.05 were considered statistically significant.

Data availability
All the data generated for this study is available in the manuscript and its related Supplementary Material.The sequenced raw reads have been deposited in the open-access European Nucleotide Archive database under the accession number PRJEB70795.

Characteristics of all the CP-Eco isolates received from 2015 to 2020
From January 2015 until December 2020, a total of 593 CP-E. coli isolates were confirmed as carbapenems producers by the Spanish CNM Reference Lab after screening for MIC to meropenem of over 0.12 mg/L by the collaborating hospitals.Supplementary Table S1 shows the characteristics and origin of the isolates, as well as the information received from the participating hospitals (total and per year).The isolates were received from 25 Spanish provinces and 59 hospitals.Twohundred-ninety-nine (50.4%) of the isolates were from females and the average patient age was 77.8 years.Three-hundred-one (50.8%) isolates were colonizers, and the rest were causing infections (24 were unspecified).The most common infection was UTI (175; 29.5%), followed by wound/ulcer infections (32; 5.4%), bacteremia (25; 4.2%), and IAI (17; 2.9%).Four-hundred-fortythree (74.7%) were healthcare-associated cases, 391 from hospitals and 52 from other healthcare centers, and 126 (21.2%) were from community settings; 24 were unspecified.
Using real-time PCR carbapenemase genes were detected in all 593 isolates where bla OXA-48-like was detected in 485 (81.8%), bla VIM-like in 71 (12%), bla NDM-like in 17 (2.9%), and bla KPC-like in 17 (2.9%)isolates.Three isolates harbored two carbapenemases, one of each of the following combinations bla OXA-48 and bla KPC , bla OXA- 48 and bla NDM , and bla VIM and bla KPC .In the community setting, the most frequent carbapenemase was bla OXA-48-family , detected in 109 (86.5%) isolates.The relative frequency of these genes per year was maintained throughout the study period and there was no statistically relevant correlation between the types of carbapenemase gene, sample type, location of isolation, infection, colonization, and type of carbapenemase harbored by the isolates (P > 0.05).

Characterization of the plasmids carrying the carbapenem resistance genes
The median size of the detected plasmids was 63,598 bp with a median coverage of 98% and a median similarity of 100% with the respective entries in Genbank.IncL plasmids were the most frequently detected (n=42; 46.7%).Thirty-two of these harboured bla OXA-48 , eight harboured bla VIM-1 , and two harboured bla OXA-162 .The second most frequent plasmid type was IncFII detected in 12 (13.3%)isolates, eight of which harbored bla OXA-48 and four harbored bla NDM .Five of the IncF-harboring isolates also belonged to clade C of ST131.IncN was detected in Minimum spanning tree showing the population structure of the 90 isolates analyzed through core-genome multi-locus sequence typing (scheme of 2,515 genes).The distance between the genes are not up to scale, black lines represent differences higher than 50% of the alleles considered in the scheme (>1257 alleles), blue lines differences between 10-50% (252 to 1257) and red lines differences lower than 10% (<252).Sequence types with four or more isolates are circled and amplified.The blue shading highlights the only group (containing two ST127 isolates) that have a genetic distance of ≤ 5 alleles (this group has also been amplified).
five (5.3%) isolates where four harbored bla VIM-1 and one harbored bla KPC-2 .IncP6 was detected in four isolates harboring bla K PC-2 , and IncFIB was detected in four of the five isolates harboring bla KPC-3 .Other plasmid types were detected with a frequency of two or less and it was not possible to fully reconstruct the carbapenemase-harboring plasmids in 20 isolates (Supplementary Table S2).

Antibiotic resistance profiles
Table 2 shows the AST results that were analyzed according to the categories described in section 2.3.All isolates had MICs>0.12mg/L to meropenem, harbored at least one gene of carbapenem resistance, and were resistant to at least one carbapenem (Table 2).The two isolates producing two carbapenemases presented relatively low MICs of 4 mg/L to imipenem (both) and 4 and 1 mg/L to meropenem (each).Antibiotics with in-vitro activity against more that 90% of the isolates were colistin, plazomicin, cefiderocol, and meropenem/ vaborbactam.The mcr genes and known mutations in the PhoPQ and PmrAB encoding genes related to colistin resistance were not detected in the colistin-resistant isolate.
Six of the seven isolates resistant to cefiderocol were NDMproducers (4 bla NDM-5 and 2 bla NDM-1 ) and one had bla VIM-1 .Four NDM-producers belonged to ST167 and had an insertion in the pbp3 gene (Y334insYRIN), in addition to two other mutations in the same gene (E349K and I532L).Three of these isolates additionally had a frameshift mutation in the siderophore gene cirA (V89fs).Another NDM-producer (ST131) had a truncation in cirA (W243trunc), and the final NDM-producer belonged to ST405, had the Y334insYRIN insertion in the pbp3 gene, and a frameshift mutation in cirA (T508fs).
Finally, 82 isolates (91.1%) had fosfomycin MICs of ≤ 8 mg/L and were considered susceptible (using EUCAST criteria for UTIs).Of those, 21 isolates (25.6%) had chromosomal mutations associated with resistance to fosfomycin.Of the 8 isolates resistant to this antibiotic, two harbored fosA genes, three had the chromosomal mutations ptsI_V25I/uhpT_E350Q, one had the chromosomal mutation uhpT_E350Q, and in the last two no known genetic mutation or gene was identified.There was a significant correlation between harboring mutations or genes associated with fosfomycin resistance and phenotypic resistance to this antibiotic (P = 0.003).

Characterization of the virulenceassociated genes
Table 4 shows the detected virulence-associated genes (13 virulence genes considered), and the number of infection-causing isolates harboring them.The isolates had a median of two virulenceassociated genes (range=0-9) where adhesin-coding genes were the most prevalent (68.9%).Thirty-eight isolates (42.2%) had four or more virulence-associated genes, 19 of which were causing infections.The most common infection in this group was UTI (9), followed by bacteremia (5), IAIs (2), wound infections (2), and abscess (1) (Supplementary Table S2).Isolates with 4 or more virulence-associated genes included three isolates belonging to ST131 and two belonging to ST88; and had similar averages of ARGs and susceptibility rates to the different antibiotic categories.Nineteen of these 34 isolates had bla OXA-48 , 8 had bla VIM-1 , 3 had bla NDM-5 , 2 had bla NDM-1 , 1 had bla KPC-3 , and 1 had both bla OXA-48 and bla KPC-3 .Of the isolates pertaining to ST131, all had fimH and over half of them harbored pap and ompA.Over half of those pertaining to ST88 harbored fimH, iroN, and ompA.
The adhesin-coding pap and the toxin-coding sat genes were found to be more frequent among isolates pertaining to ST131 (P = 0.02 and 0.0004, respectively).Particularly, five of the 9 isolates positive for pap and five out of the six positive for sat belonged to clade C. Additionally, the adhesion-coding fimH gene had almost double the frequency among isolates harboring bla OXA-48 as compared to isolates harboring other carbapenemase genes (P = 0.04).There were no statistically significant differences in the virulence-associated genes between bloodstream isolates and isolates from other sources.

Discussion
In this study, 90 representative CP-Eco isolates obtained from 25 Spanish provinces from 2015 to 2020 were characterized.The inclusion criteria were designed to achieve the greatest possible representation of geographic areas, clones, and carbapenemases, as well as to avoid overrepresentation due to outbreaks and/or endemics.They do not reflect the real prevalence of the carbapenemase genes; an approach to this more general view is provided in section 3.1, which mirrored the epidemiological situation of CP-Eco in Spain (Cañada-Garcıá et al., 2022).Notably, CP-Eco isolates were found in 25 Spanish provinces, and major STs, such as ST131, showed inter-provincial spread.The studied CP-Eco isolates were highly polyclonal, unlike other Enterobacterales, such as Klebsiella pneumoniae that tend to show less genetic diversity (Oteo et al., 2015;Laźaro-Perona et al., 2022).
The high-risk ST131 was the most prevalent in this collection; this ST also represented 36% of CP-Eco isolates from 16 countries between 2008and 2013(Patiño-Navarrete et al., 2020).In Spain, ST131 represented 16.5% of CP-Eco isolates collected between 2012-2014 (Ortega et al., 2016).Furthermore, isolates belonging to this ST harbored all the carbapenemase genes detected in this study, on different plasmids and in different combinations with other acquired ARGs.This demonstrates its potential to act as a vector for disseminating resistance genes in different scenarios.Half of the ST131 isolates caused infections, and all of them were community-acquired.Moreover, a significant association was found between ST131 and the virulence genes sat (toxin-coding) and pap (adhesin-coding), highlighting its potential to cause lifethreatening community-acquired infections.
The ST167/NDM-5 clone, observed in this study, had been previously described (Poirel et al., 2022b) and had shown to spread rapidly across European borders (Linkevicius et al., 2023).These isolates were also resistant to cefiderocol and detected in three different provinces.Of note, 6/7 of the cefiderocol-resistant isolates were also NDM-producers, an association that has been previously drawn, especially among ST167 isolates having the same mutations in pbp3 as shown in our study and a premature stop codon in the siderophore gene cirA (Wang et al., 2022).Additionally, NDMproducers were found to have tighter inhibition zones towards cefiderocol compared to non-NDM-producers, in line with recent findings where NDM-producers had elevated MICs towards this antibiotic (Poirel et al., 2022a).This hints at an interplay between NDM and cefiderocol resistance that warrants further investigation.
Another finding in our study is the detection of a relatively high rate of MBL-producers in isolates recovered from the community setting.The high average age of these patients could suggest an origin in nursing homes that are known reservoirs of CP-producing bacteria (Palacios-Baena et al., 2016) and whose flow of admissionsdischarges from and to the community is difficult to trace.Moreover, bla VIM-1 was detected in small IncL plasmids similar to the highly transmissible ones that carried bla OXA-48 (Boyd et al., 2022).This combination was detected in isolates belonging to different STs, as previously described in K. pneumoniae (Cañada-Garcıá et al., 2023), and was predominantly found in healthcareassociated settings.This raises concern regarding the fast dissemination of bla VIM-1 , especially in Spain where the prevalence of VIM-producers is higher than other countries (Johnston et al., 2021;Cañada-Garcıá et al., 2022).Of note, bla OXA-48 was in IncFII plasmids in eight isolates, an observation that was reported in another study though bla OXA-48 was part of a different transposon (Moussa et al., 2020).
The CP-Eco isolates harboring bla OXA-48 were significantly more susceptible to imipenem and meropenem compared to CP-Eco isolates harboring other carbapenemases, in accordance with the known weaker hydrolytic activity of this enzyme towards carbapenems (Loṕez-Camacho et al., 2019).Moreover, all OXA-48-like-producers and all but one KPC-producer did not harbor any other ARG on their plasmids, unlike other Enterobacterales such as K. pneumoniae where it is typical to find other ARGs with bla KPC (Cañada-Garcıá et al., 2022;Cañada-Garcıá et al., 2023).In contrast, 30-40% of MBL-producers were resistant to more than six antibiotic categories as defined in this study.All the "new" antibiotics and antibiotic/inhibitor combinations, except imipenem/relebactam, showed over 75% in-vitro activity towards the CP-Eco isolates.Resistance to imipenem/relebactam, meropenem/vaborbactam or ceftzidime/avibactam was mainly due to MBL production (Galani et al., 2019).Three of the only four isolates resistant to at least one of these combinations had insertions in the OmpC porin gene, in comparison with sequences from OXA-48-producing isolates of this study only resistant to ertapenem but susceptible to the rest of the carbapenems and 3rd generation cephalosporins.The presence of mutations in the main porins of Enterobacterales has previously been related to resistance to the new beta-lactamase inhibitors combinations (Navarro-Carrera et al., 2022).VIM-producers appeared to be more resistant towards imipenem/relebactam than imipenem alone, which has been previously reported in another study (Hernańdez-Garcıá et al., 2022).However, this was mainly caused by the EUCAST cutoff values (European Society of Clinical Microbiology and Infectious Diseases (EUCAST), 2022) used to determine susceptibility since six of the eight VIM-1 producing isolates resistant to imipenem/relebactam but susceptible to imipenem had MICs of 4 mg/L for both antibiotics.There were also seven NDM-producers that were susceptible to meropenem/vaborbactam according to the EUCAST guidelines.However, these isolates had elevated MICs, and most would have been considered resistant in case the CLSI guidelines were followed (Jean et al., 2019).Six isolates were resistant to plazomicin, five of them harbored bla N D M -l i k e and were the only isolates in which rmt methyltransferases were detected.Furthermore, unlike other Enterobacterales in which rmt methyltransferase gene seems to disseminate via clonal expansion (Arca-Suaŕez et al., 2022), in our collection this gene was detected in isolates from different STs and were present in different plasmids.
Fosfomycin is one of the main antibiotics used to treat community-acquired uncomplicated cystitis caused by E. coli in Spain (Sojo-Dorado et al., 2022).In this study, most of the isolates that had mutations associated with resistance to this antibiotic harbored chromosomal mutations, and only two isolates had fos genes, similarly to what was previously described (Falagas et al., 2019).Moreover, the fosA7 genes detected in the chromosome of one of the isolates has also been described to be able to be present there (Pan et al., 2021).The correlation between resistance to fosfomycin and the presence of resistance mechanisms against this antibiotic must be taken with caution since cut-off points for UTI have been used (the only ones established by EUCAST), of which there were only 26 isolates in this collection.Moreover, although a good agreement was reported between using E-test strips and disk diffusion assays (Hirsch et al., 2015;Goer et al., 2022), the EUCAST does not explicitly recommend the use of E-tests for the determination of cutoff values for fosfomycin.Nevertheless, these isolates could be displaying a heterorresistant phenotype that appeared to be susceptible in our testing methods and criteria (Campos et al., 2020).
In conclusion, our data shows a wide range of genetically diverse CP-E. coli isolates present across several Spanish geographical areas, with a heterogeneous set of ARGs, plasmids, and virulence-associated genes.These findings highlight the potential of this bacteria to acquire and disseminate ARGs, and to cause life-threatening, difficult-to-treat infections in different settings, including the community.The spread of CP-Eco in the community, like what already happened with CTX-M-15producing E. coli, would pose a threat of great impact to the general health of the population.The integration of WGS in the surveillance and diagnosis of CP-Eco, in line with ECDC recommendations, can help control and minimize this impact.

TABLE 1
Genes and mutations associated with antibiotic resistance detected through whole-genome sequencing.

TABLE 2
Antibiotic susceptibility profiles of the 90 selected isolates determined through broth microdilution and interpreted according to the EUCAST guidelines.

TABLE 3
Number and percentages of the CP-Eco isolates that are resistant to the tested antibiotics grouped by the carbapenemase gene that they harboured.

TABLE 4
Virulence-associated genes detected in the all 90 carbapenemase-producing E. coli isolates, in addition to the two most frequent sequence types and the number of isolates harboring these genes that were causing infections.