Whole genome sequencing, molecular typing and in vivo virulence of OXA-48-producing Escherichia coli isolates including ST131 H30-Rx, H22 and H41 subclones

Carbapenem-resistant Enterobacteriaceae, including the increasingly reported OXA-48 Escherichia coli producers, are an emerging public health threat worldwide. Due to their alarming detection in our healthcare setting and their possible presence in the community, seven OXA-48-producing, extraintestinal pathogenic E. coli were analysed by whole genome sequencing as well as conventional tools, and tested for in vivo virulence. As a result, five E. coli OXA-48-producing subclones were detected (O25:H4-ST131/PST43-fimH30-virotype E; O25:H4-ST131/PST9-fimH22-virotype D5, O16:H5-ST131/PST506-fimH41; O25:H5-ST83/PST207 and O9:H25-ST58/PST24). Four ST131 and one ST83 isolates satisfied the ExPEC status, and all except the O16:H5 ST131 isolate were UPEC. All isolates exhibited local inflammatory response with extensive subcutaneous necrosis but low lethality when tested in a mouse sepsis model. The bla OXA-48 gene was located in MOBP131/IncL plasmids (four isolates) or within the chromosome (three ST131 H30-Rx isolates), carried by Tn1999-like elements. All, except the ST83 isolate, were multidrug-resistant, with additional plasmids acting as vehicles for the spread of various resistance genes. This is the first study to analyse the whole genome sequences of bla OXA-48-positive ST131, ST58 and ST83 E. coli isolates in conjunction with experimental data, and to evaluate the in vivo virulence of bla OXA-48 isolates, which pose an important challenge to patient management.

and bla OXA-48 were also experimentally detected by PCR amplification. Plasmid genes are underlined (see Table 2 for details). f ST, sequence type according to Achtman; PST, sequence type according to the Pasteur Institute. g The virotype was determined by PCR based on the presence or absence of 13 virulence genes 33  other carbapenems was low. In fact, all isolates except Ec-HUCA-4 were susceptible to meropenem, so this drug could have been included in the therapy regimen for the treatment of the affected patients (Table 1). In addition, all isolates were positive in the modified Hodge and the Carba NP tests, and contained the bla OXA-48 gene (Table 1). In Ec-HUCA 1 to 4 the bla OXA-48 gene was carried by conjugative plasmids of about 60 Kb, experimentally assigned to the IncL group. The bla OXA-48 gene was chromosomally located in Ec-HUCA 5 to 7. All isolates except Ec-HUCA 3 were also resistant to other antimicrobial agents, including fluoroquinolones, aminoglycosides, trimethoprim-sulfamethoxazole and broad spectrum cephalosporins ( Table 1). The latter resistance was detected in four isolates (Ec-HUCA 2, 5, 6 and 7), which tested positive for bla CTX-M-15 by PCR amplification and sequencing. Five out of the seven isolates were identified as ST131 by the Achtman scheme 16 and further differentiated into PST43 (Ec-HUCA 5 to 7), PST9 (Ec-HUCA 4) and PST506 (Ec-HUCA 1) according to the Pasteur Institute scheme 17 (Table 1). All ST131 isolates belonged to serotype O25:H4, except Ec-HUCA 1 which was O16:H5. The three O25:H4-ST131/PST43 isolates carried fimH30, were resistant to fluoroquinolones and positive for bla CTX-M-15 , thus belonged to subclone H30-Rx. The remaining two isolates (Ec-HUCA 2 and 3) showed different serotypes (O9:H25 and O25:H5), STs (ST58 and ST83), PSTs (PST24 and PST207) and fimH alleles (H27 and H21) ( Table 1). ST58 has been identified in isolates of different origins (human, animal and environment), and was frequently associated with CTX-M production, particularly CTX-M-1 18, 19 . However, besides Ec-HUCA-2, only one other ST58 OXA-48-producer has been reported 20 . In contrast to ST131 and ST58, little information exists on ST83, which was previously found in uropathogenic E. coli from cats 21 . In addition, a query of the Enterobase database (version 6 th May 2017) identified 22 ST83 entries from different sources (human, companion animals, wildlife and environment) and geographical regions. The presence of bla OXA-48 in the otherwise susceptible Ec-HUCA 3, demands further surveillance of the ST83 lineage. By PCR, the ST131 and ST83 isolates were assigned to phylogroup B2, while the ST58 isolate belonged to phylogroup B1.
The isolates were also analysed by XbaI-PFGE electrophoresis. The dendrogram generated from comparison of the macrorestriction profiles was consistent with the genetic diversity of the isolates. Thus, only the three ST131/PST43 H30-Rx isolates of virotype E clustered with similarity ≥85%. The remaining two ST131 isolates (Ec-HUCA 1 and 4) presented 76.7% and 65% similarity with that cluster (Supplementary Figure S1). Although the high diversity of the ST131 clonal group has been widely reported 3,22 , it was unexpected for this specific group of five ST131 OXA-48-producing isolates, since they were obtained in the same health area over a relatively short period of time.
Sequencing of OXA-48-producing isolates and plasmid reconstruction. The draft genomes of the seven E. coli isolates yielded 22 to 135 contigs larger than 1 Kb, with assembly sizes ranging from 4.9 Mb (Ec-HUCA 3) to 5.4 Mb (Ec-HUCA 4); (average of 5.2 ± 0.137 Mb). This is the first time that the genomes of OXA-48-producing ST131 E. coli isolates of subclones H30-Rx, H22 and H41, ST58 and ST83 were sequenced. In silico determinations of serotypes, STs/PSTs and fimH alleles fully agreed with experimental data.
In order to establish the plasmid content of the seven E. coli genomes, the PLACNET protocol 23 was used for plasmid reconstruction (Supplementary Figure S2). As shown in Table 2, plasmids were found in all isolates in numbers ranging from one (Ec-HUCA 5) to seven (Ec-HUCA 4). In the genomes of Ec-HUCA 2 and 4, PLACNET identified two plasmids that could not be separated and accounted for a total of 348 Kb  Figure S3).
Overall, the seven E. coli genomes included 23 plasmids, which were classified as MOB P51 /ColE1-like, MOB P131 /IncL, MOB F12 /IncF, MOB F11 /IncN, MOB P3 /IncX and MOB P12 /IncI (Table 2). For one MOB Q12 plasmid the incompatibility group was not determined and two other plasmids could not be affiliated with any of the MOB categories. Finally, a MOB H2 relaxase was identified in the chromosome of Ec-HUCA 1, consistent with the presence of an integrative and conjugative element.
In four out of the seven E. coli isolates (Ec-HUCA 1 to 4; Table 2), the MOB P131 /IncL plasmids were experimentally and in silico linked with the bla OXA-48 gene. Their genetic relationship was assessed by building a phylogenetic tree, which also included other IncL and IncM plasmids present in the GenBank-NCBI database (Fig. 1). Assignment to one or the other Inc group was corroborated in silico, or newly determined using the primers reported by Carattoli et al. 8 . The tree separated IncL and IncM plasmids in two clusters, the latter with two subclusters, IncM1 and IncM2, as previously observed 8 . Interestingly, bla OXA-48 was not only carried by IncL plasmids but also by IncM1 plasmids, formerly classified as IncL/M 24,25 . Besides, it was found also on IncA/C plasmids 26 . The backbones of the IncL plasmids from the HUCA isolates were identical to each other and to nine other IncL plasmids, originating in Klebsiella pneumoniae, Citrobacter freundi, Raoultella planticola and E. coli. These results underline the prevalence of an IncL plasmid lineage which plays a major role in the horizontal spread of bla OXA-48 between members of the Enterobacteriaceae 5 . BRIG comparison of the IncL plasmids is shown in Fig. 2. The pEc-HUCA and pE71T 27 plasmids are highly similar to pOXA-48 11 . pCF29 (Accession Number LN864820) showed a large gap at the mobilization-transfer region (24-39 Kb); whereas pCTX-M3 28 and pNDM-HK 29 lack the bla OXA-48 region (2.5-8 Kb), as expected.
Genetic environment of the bla OXA-48 gene. As already indicated, bla OXA-48 was carried by MOB P131 / IncL plasmids in four E. coli isolates, but chromosomally located in another three ( To identify the insertion sites of bla OXA-48 within the chromosomes of Ec-HUCA 5 to 7, assembled contigs were analysed. Three contigs containing sequences homologous to pOXA-48a (NC_019154 11 ) and pRA35 (LN864821 9 ) were identified, and two of them were flanked by E. coli chromosomal genes. The three isolates shared an identical 21.9 Kb fragment in which the inverted and deleted version of Tn1999.2 was found, followed by genes previously detected in pOXA-48a and pRA35. This fragment, which is flanked by IS1R, constitutes an IS1R composite transposon, termed Tn6237 9 . With the single exception of an isolate from Lebanon 9 , this is the first report of a Tn6237 insertion into the chromosome of ST131. The precise insertion site of the transposon was established by PCR mapping (Supplementary Figure S4). In Ec-HUCA 6 and 7, Tn6237 was placed between the tdcA (DNA-binding transcriptional activator) and tdcB (threonine dehydratase) genes, which are contiguous in other E. coli genomes. In the case of Ec-HUCA 5, Tn6237 was followed by tdcB but morA (putative dehydrogenase) was placed upstream. This organization may have resulted from inversion of the chromosomal segment spanning tdcA to morA. Consistent with its expected lack of specificity, IS1-flanked Tn6237, was also found in three other insertion sites within the E. coli chromosome, in isolates from the United Kingdom, Czech Republic and Lebanon 9,30,31 .
Resistance genes for other β-lactam and non-β-lactam antimicrobials. According to their phenotypes (Table 1), additional resistance genes were found in all OXA-48 isolates, contained in MOB F12 /IncF, MOB F12 /IncF + IncI1 and MOB F11 /IncN plasmids or in the chromosome (Table 2). Thus, diverse plasmids act as vehicles for the spread of resistance genes between E. coli clones/subclones circulating in the same health area. Remarkably, bla CTX-M-15 appeared in different locations within the four ST131 isolates carrying it. Specifically, bla CTX-M-15 was chromosomally located in two out of the three isolates belonging to the H30-Rx subclone (Ec-HUCA 5 and 6), whereas in Ec-HUCA 7 it was carried by a MOB F11 /IncN (ST7) plasmid of 48.7 Kb, together with qnrS1 and dfrA14, for plasmid-mediated quinolone resistance (PMQR) and trimethoprim resistance, respectively. Interestingly, a slightly smaller MOB F11 /IncN (ST7) plasmid of 45.8 Kb contained qnrS1 and dfrA14, but not bla CTX-M-15 , in Ec-HUCA 6. The bla CTX-M-15 gene of the non-ST131 isolate (Ec-HUCA 2), was carried by a conjugative MOB F12 /IncF plasmid of ca. 150 Kb 15 .
In the three ciprofloxacin resistant ST131 H30-Rx-OXA-48 isolates (Ec-HUCA 5 to 7; MIC > 32 mg/L), in silico analysis of the genomes revealed the distinct gyrA/parC allele combination (gyrA-S83L/D87N; parC-S80I/ E84V) previously reported for the H30 subclone 32 . The S83L mutation was also found in the gyrA gene of Ec-HUCA 1 and 4, which were resistant to nalidixic acid but susceptible to ciprofloxacin. Apart from qnrS1 found in Ec-HUCA 6 and 7, other PMQR genes, including additional qnr genes and qep or oqx genes, tested negative by bioinformatic methods. The genetic bases of most other resistances could be also established by in silico analysis (Tables 1 and 2), being of note the presence of a W30R mutation in the folA gene (dihydrofolate reductase) of Ec-HUCA 5, which was resistant to trimethoprim-sulfamethoxazole but lacked sul and dfr genes.  Table S1). The virulence profile of the O16:H5-ST131 isolate did not match any of the 12 virotypes included in the scheme 33   PCR-screening results were complemented by in silico analysis of the genome sequences, which allowed the identification of additional VFs, mainly encoded by genes located in the chromosome but also on plasmids ( Table 2). Like before, the highest VF score was obtained for Ec-HUCA 4, followed by the ST131 H30-Rx-OXA-48 isolates (Ec-HUCA 5 to 7), which shared the same profile consisting of four operons and 15 individual genes. In contrast, a score of 16 was obtained for Ec-HUCA 1, the ST131 isolate with undefined virotype.

Virulence gene content and studies of virulence "in
With regard to non-ST131 isolates, 27 VFs were detected in Ec-HUCA 2 and only 13 in Ec-HUCA 3. Interestingly, the plasmid virulence genes of Ec-HUCA 2 and 4 were nearly identical, and some virulence genes have chromosomal and plasmid copies (iss in Ec-HUCA 2, and mchF and iss in Ec-HUCA 4), providing regions of homology for interaction between the two replicons. Finally, the intrinsic extraintestinal virulence of the OXA-48-producing E. coli was assessed in a mouse sepsis model. Within a seven day experiment, all mice challenged with E. coli CFT703 (positive control) died, compared with none of the mice challenged with E. coli MG1655 (negative control), which remained healthy. All OXA-48 isolates showed lethality as low as ≤10% and only three (Ec-HUCA 2, 6 and 7) caused the death of one out of the ten inoculated mice. In contrast, all isolates caused local inflammatory response, with extensive subcutaneous necrosis, in the surviving mice (Supplementary Table S2). Previously, we observed different virulence patterns in the final lethality, the rapidity in causing death and the inflammation-causing ability of ST131 isolates in correlation with the virotype, with the highest lethality (≥80% of mice challenged killed) shown by virotypes A, B, C and D1. By contrast, isolates within virotypes D2, D3 and D4 led to different outcomes, and isolates of virotype E showed the lowest final lethality, varying from 10 to 40% of the challenged mice. We also observed that certain ST131 isolates of virotypes C, D, and E induced an acute inflammatory response in the inoculation region 22 , like those in this study. There are few comparable in vivo studies and this is the first one assaying OXA-48-producing isolates. Results derived from the sepsis model might be consistent with the clinical nature of the isolates, not involved in severe disease but recovered from surgical wounds and UTIs. Future studies would be necessary to investigate the mechanisms responsible for the differences in lethality within ST131 virotypes.
Phylogenomics of the OXA-48-producing isolates. The phylogenetic context of the OXA-48-producing E. coli was assessed by comparison with 28 E. coli genomes with different STs, PSTs, virotypes and fimH alleles (Fig. 4,  Supplementary Table S3). All ST131 isolates grouped in a single cluster which was further divided into subclusters according to clade, virotype and serotype. In agreement with their assignment to phylogroup B2, the ST131 cluster was genetically closer to ST83 (also B2) than to ST58 (phylogroup B1). It is of note that only two SNPs differences Escherichia coli isolates and representative IncL, IncM1 and IncM2 plasmids. Each ring corresponds to a plasmid (indicated at the right of the figure together with the color code). Plasmid pOXA-48 (inner black ring) was used as a reference. Their genes are shown in the outer ring, represented by black arrows that also indicate the direction of transcription.
were found between the COG-based core genomes of Ec-HUCA 6 and 7, and that both differed from Ec-HUCA 5 by 14 SNPs. Comparing these three highly similar isolates using SNP analysis and conventional PFGE, we found that both approaches discriminated between Ec-HUCA 5 vs 6 and 7, the latter displaying 100% identity between them and 88.4% with Ec-HUCA 5 by PFGE (Supplementary Figure S1). As shown in Table 1, these clonal isolates were detected in a primary-care centre, in the emergency unit of the HUCA and in the geriatric unit of a long-term facility. Thus, the H30-Rx-OXA-48 subclone, with the chromosomally located carbapenemase gene, is circulating both in the community and health-care institutions, which indicates transmission between the two settings. The number of SNPs between these and other isolates progressively increased according to the ST/PST and phylogroup affiliation, with averages of 3,755, 8334, 19,787 and 79,640, with regard to Ec-HUCA 4, 1, 3 and 2, respectively.
According to our results, diverse OXA-48-producing E. coli clones are circulating in Oviedo, a situation justified in part by the conjugative potential of the IncL plasmid carrying the bla OXA-48 gene. Additional plasmids also play a role as vehicles of resistance and/or virulence genes. Treatment of the affected patients represents a serious challenge since all except one isolate were MDR. So, antimicrobial stewardship policies, new antimicrobial therapy approaches and control measures are necessary to combat the infections caused by these bacteria, and to control further dispersal.

Material and Methods
Epidemiological background of the OXA-48-producing isolates. The isolates were recovered from wound infections or UTIs between 2012 and 2015 (Table 1). Like Ec-HUCA 1 to 3 15 , Ec-HUCA 4 caused a hospital-acquired infection and affected a critical patient exposed to long-term hospitalization and prolonged antimicrobial treatment in the HUCA. The remaining isolates were recovered from patients attended at a primary care centre (Ec-HUCA 5), the emergency unit of the HUCA (Ec-HUCA 6), and the geriatric unit of a long-term care facility associated with the hospital (Ec-HUCA 7).  MICs for erythromycin were determined by broth microdilution following CLSI guidelines 36 . MICs for carbapenems (ertapenem, imipenem and meropenem) were obtained with Etest strips (bioMérieux, Marcy-l'Étoile, France). Results were interpreted according to CLSI breakpoints 36 . Carbapenemase production was confirmed by the modified Hodge and Carba NP tests 37 . Identification of genes encoding resistance to carbapenems and broad-spectrum cephalosporins, and plasmid analysis were performed as reported 8,15,38 . The genetic context of the bla OXA-48 gene was determined by PCR mapping (see Supplementary Table S4 and Supplementary Figure S4).

Antimicrobial susceptibility testing and plasmid analysis.
Typing, subtyping and phylogenetic grouping of the isolates. E. coli isolates were characterized with regard to O:H serotype and fimH alleles (for type 1 fimbrial adhesion) [39][40][41] . The STs were established following the MLST schemes of Achtman (http://mlst.warwick.ac.uk/mlst/dbs/Ecoli) and the Pasteur Institute (http://bigsdb.pasteur.fr/ecoli/ecoli.html). The XbaI-PFGE profiles were determined according to PulseNet protocol (http://www.pulsenetinternational.org/), imported into BioNumerics (Applied Maths, St-Martens-Latern Belgium) and clustered by Dice/UPGMA. Fifty genes/alleles encoding virulence factors (VF) were screened by PCR (Supplementary Table S5). Isolates were presumptively designated as ExPEC if positive for two or more of five markers and as uropathogenic (UPEC) if positive for three or more of four markers (Supplementary   42,43 . The virotype of the ST131 isolates was established according to the scheme described by Dahbi et al. 33 . Assignment to the main phylogroups (A, B1, B2 and D) was based on the protocol of Clermont et al. 17 .
Genome sequencing, assembly and analysis. Total DNA from E. coli isolates was extracted with the QIAmp DNA Mini Kit (Qiagen). Libraries were prepared using the TruSeq PCR-free DNA Sample Preparation Kit (Illumina) at the sequencing facility of the University of Cantabria. Paired-end 100 bp reads (550 bp insert size) were sequenced in a HiSeq. 2500 (Health in Code Facility). Reads were assembled with the VelvetOptimiser. pl script of Velvet software 44 . Serotype, MLST, fimH alleles and virulence gene profiles were in silico determined with SerotypeFinder v1.1 45,46 , and homemade MLST (Achtman and Pasteur schemes), fimH and virotype databases 47 . Antimicrobial resistance genes were detected using the ARG-ANNOT 48 and ResFinder 49 databases; virulence gene content was established with VirulenceFinder and homemade databases 50 .
Plasmid reconstruction from WGS data and analysis. Plasmid reconstructions were based on the PLACNET method 23 . Contig analysis was performed against complete bacterial genomes and plasmids from GenBank-NCBI. Relaxase proteins (REL) and Replication Initiation Proteins (RIP) were identified using in-house databases 23,51 . Incompatibility groups and pMLST subtypes were experimentally 8,52 and/or in silico determined (http://pubmlst.org/plasmid/). Reconstructed bla OXA-48 IncL plasmids and references belonging to the IncL and IncM groups were compared using BRIG 53 , and a phylogenetic tree was built from variable positions (SNPs) in genes encoding core proteins.
Mouse lethality assay. A mouse sepsis model was used to assess extraintestinal virulence of the isolates 22 .
For each one, 10 outbred female RjOrl:Swiss mice (3-4 weeks old; Janvier Labs, France) received a subcutaneous injection into the nape of the neck of approximately 2 × 10 8 CFU of log-phase bacteria. After inoculation, mice were clinically inspected along one week. Time of death and local presence of lesions (acute inflammation in the region of inoculation) were recorded for each mouse. Surviving mice were euthanatized on day seven by cervical dislocation. In each assay, two control isolates were included: E. coli K-12 MG1655 which does not kill mice by seven days post-challenge, and E. coli CFT073 which shows a lethality of ≥80% by seven days post-challenge. Results of lethality were indicated as the number of mice killed within 24 h and within seven days post-injection.
Ethics statement. All experimental protocols dealing with bacteria from human samples were approved by the ethics committee of the HUCA.
All animal experimentation was conducted following European (Directive 2010/63/EU on the protection of animals used for scientific purposes) and National (RD 53/2013) regulations for transport, housing and care of laboratory animals. The protocol used was approved by the Animal Welfare Committee of the Veterinary Faculty in Lugo, University of Santiago de Compostela (AE-LU-002/14-1).