Phenotypic and Molecular Epidemiology of ESBL-, AmpC-, and Carbapenemase-Producing Escherichia coli in Northern and Eastern Europe

Extended-spectrum beta-lactamases (ESBL) and AmpC producing-Escherichia coli have spread worldwide, but data about ESBL-producing-E. coli in the Northern and Eastern regions of Europe is scant. The aim of this study has been to describe the phenotypical and molecular epidemiology of different ESBL/AmpC/Carbapenemases genes in E. coli strains isolated from the Baltic States (Estonia, Latvia, and Lithuania), Norway and St. Petersburg (Russia), and to determine the predominant multilocus sequence type and single nucleotide polymorphisms diversity of E. coli isolates deduced by whole genome sequencing (WGS). A total of 10,780 clinical E. coli strains were screened for reduced sensitivity to third-generation cephalosporins. They were collected from 21 hospitals located in Estonia, Latvia, Lithuania, Norway and St. Petersburg during a 5 month period in 2012. The overall prevalence of ESBL/AmpC strains was 4.7% by phenotypical test and 3.9% by sequencing. We found more strains with the ESBL/AmpC phenotype and genotype in St. Petersburg and Latvia than other countries. Of phenotypic E. coli strains, 85% contained confirmed ESBL genes (including blaCTX–M, blaTEM–29, blaTEM–71), AmpC genes (blaCMY–59, blaACT–12/–15/–20, blaESC–6, blaFEC–1, blaDHA–1), or carbapenemase genes (blaNDM–1). blaCTX–M–1, blaCTX–M–14 and blaCTX–M–15 were found in all countries, but blaCTX–M–15 prevalence was higher in Latvia than in St. Petersburg (Russia), Estonia, Norway and Lithuania. The dominating AmpC genes were blaCMY–59 in the Baltic States and Norway, and blaDHA–1 in St. Petersburg. E. coli strains belonged to 83 different sequence types, of which the most prevalent was ST131 (40%). In conclusion, we generally found low ESBL/AmpC/Carbapenemase prevalence in E. coli strains isolated in Northern/Eastern Europe. However, several inter-country differences in distribution of particular genes and multilocus sequence types were found.


INTRODUCTION
Antimicrobial resistance is an emerging problem worldwide. Each year, 33,000 people die from an infection due to bacterial resistance to antibiotics in Europe. The burden of infections with bacterial resistance to antibiotics on the European population is comparable to that of influenza, tuberculosis and HIV/AIDS combined (Cassini et al., 2019). It has been estimated that by 2050, 10 million lives a year and a cumulative 100 trillion USD economic output are at risk worldwide due to the rise of drug resistant infections if we do not find proactive solutions to slow down drug resistance (O'Neill, 2016).
Resistance of Gram-positive bacteria is generally stable or even decreasing in Europe, whereas resistance to Gram-negative bacteria (such as Enterobacterales) has an increasing trend in several European countries (EARS-Net, 2018).
Prevalence of these beta-lactamases has been increasing all over the world, including European countries (Bevan et al., 2017). Data from the European Antimicrobial Resistance Surveillance Network shows that E. coli resistance to thirdgeneration cephalosporins is lower in Northern and higher in the Southern and Eastern Europe (EARS-Net, 2018). The proportion of invasive E. coli isolates resistant to thirdgeneration cephalosporins by EARS-Net 2017 report was 5.9% in Norway, 8.8% in Estonia, 16.8% in Lithuania, and 22% in Latvia (EARS-Net, 2018). Comparable data for Russia is absent. Data from WHO CAESAR 2016 report includes a limited number of strains from Western part of Russia and shows high proportion of invasive E. coli isolates resistant to thirdgeneration cephalosporins (66%; World Health Organisation, 2016). However, genes responsible for resistance to the thirdgeneration cephalosporins are not well described in this region (Edelstein et al., 2003;Naseer et al., 2009;Dumpis et al., 2010;Seputiene et al., 2010;Bevan et al., 2017).
The aim of this study has been to describe the prevalence and molecular mechanisms of resistance to third-generation cephalosporins in E. coli strains isolated from Estonia, Latvia, Lithuania, Norway, and St. Petersburg (Russia), and to determine the predominant multilocus sequence type and single nucleotide polymorphisms diversity of E. coli isolates deduced by whole genome sequencing (WGS).

Strain Collection
During a 5 month period in 2012, E. coli clinical isolates from 21 hospitals located in Estonia (n = 5), Latvia (n = 4), Lithuania (n = 3), Norway (n = 1), and St. Petersburg (Russia) (n = 8) were screened for reduced susceptibility to the third-generation of cephalosporins. Briefly, all clinically relevant materials (such as blood, pus, urine, and respiratory tract samples) taken in case of infection from any kind of patients (all ages, outpatients or hospitalized in any department) and sent to microbiology laboratories for culture were included in the study. Surveillance, environmental and clinically irrelevant samples were excluded. All non-duplicate E. coli isolates interpreted as a probable cause of infection were included to the study (excluding clinically irrelevant cases, such as probable colonization or contamination from indigenous microbiota), and tested for third-generation cephalosporins (at least for ceftazidime and ceftriaxone and/or cefotaxime). Duplicates were defined as the same species isolated from the same patients during the study period and showing the same resistance pattern. Thus, the first isolate from a patient was always included. In case of similar isolates that were found from different materials taken at the same time, invasive isolate was preferred (for example, an isolate from blood was collected instead of sputum). Written instructions for sampling and laboratory procedures, and laboratory materials (ESBL/AmpC, confirmations kits, quality control strains, and if needed antibiotic disc) were distributed to all participants. Beforehand our project's country managers and technical coordinators participated in a training course to ensure similar handling of samples, performance of laboratory techniques and quality control.

ESBL/AmpC Screening and Confirmation
Susceptibility testing used disk diffusion according to the guidelines of valid versions at the time of testing of European Committee of Antimicrobial Susceptibility Testing (EUCAST, as in the Baltic countries and Norway) or the Clinical and Laboratory Standards Institute (CLSI, used in St. Petersburg, Russia). Initial antimicrobial susceptibility testing was performed in each laboratory for local standard panel that includes mandatorily ceftazidime and ceftriaxone and/or cefotaxime. In E. coli isolates with reduced susceptibility to third-generation cephalosporins, ESBL and AmpC cephalosporinases production was confirmed in a local laboratory with a ESBL + AmpC confirmation kit (Rosco Diagnostica, Taastrup, Denmark) provided by the project coordinator. E. coli isolates with the ESBL/AmpC phenotype were stored and sent to Estonian reference center (Human Microbiota Biobank, University of Tartu, Tartu, Estonia) 1 for deposition and future characterization. Identification of all strains was confirmed by MALDI-TOF MS (MALDI Biotyper, Bruker Daltonics GmbH, Germany).

Bacterial DNA Extraction and Whole Genome Sequencing (WGS)
All E. coli isolates with ESBL/AmpC phenotype were sequenced. Briefly, DNA templates for sequencing were generated by growing cultures of E. coli isolates overnight on the Mueller-Hinton agar (Oxoid Limited, United Kingdom). The total bacteria DNA from the strains were extracted using QIAamp DNA Mini Kit (Qiagen, Germany).
Bacterial genomic DNA was quantified using the Qubit R 2.0 Fluorometer (Invitrogen, Grand Island, NE, United States). 1 ng of sample DNA was processed for the sequencing libraries, using Illumina Nextera XT sample preparation kit (Illumina, San Diego, CA, United States). The DNA normalization step was skipped; instead, the final dsDNA libraries were quantified with the Qubit R 2.0 Fluorometer and pooled in equimolar concentrations. The library pool was validated with the 2200 TapeStation (Agilent Technologies, Santa Clara, CA, United States) measurements) and qPCR used the Kapa Library Quantification Kit (Kapa Biosystems, Woburn, MA, United States) to optimize cluster generation. A total of 96 bacterial genomic libraries were sequenced with 2 × 101 bp paired-end (PE) reads on the HiSeq 2500 rapid-run flow cell (Illumina, San Diego, CA, United States). Demultiplexing was done with CASAVA 1.8.2. (Illumina, San Diego, CA, United States), allowing 1 mismatch in index reads.
All sequenced genomes were assembled de novo with assembler Velvet version 1.2 (Zerbino and Birney, 2008). Before assembly, all reads with low quality were removed after quality control with fastq_quality_trimmer (with parameter values -l 40, -t 30) and fastq_quality_filter (-q 25 -p 90) from FASTX-Toolkit 2 . Velvet was run with different parameter values (-max_gap_count -max_divergence -cov_cutoff -ins_length -min_pair_count) until the best match of E. coli MLST genes was retrieved.

Finding Beta-Lactamase Genes From Assembled Genomes
Beta-lactamase genes were retrieved from the Comprehensive Antibiotic Resistance Database [CARD database;(McArthur et al., 2013)]. Thereafter, the sequences were searched with BLAST (identity cut-off 90% and alignment length 90% of shortest sequence) from assembled genomes. The assembled contigs were considered to originate from either the plasmid genome or the chromosomal genome based on a BLAST search. We used complete plasmid genomes and complete chromosomal genomes of E. coli from NCBI genomes database for the BLAST search (the best match based on BLAST Score and E-value are used for deciding the origin of a contig).

Multi-Locus Sequence Typing (MLST)
For accurate multi-locus sequence typing of assembled E. coli genomes, a dedicated MLST tool was used, created by Torsten Seemann 3 , that calculates the MLST profile based on a BLAST (Altschul et al., 1997) alignment of the input sequence file and the specified allele set. Public E. coli database (Achtman scheme) for molecular typing was downloaded (with given date: 11.06.2019) from PubMLST 4 . Raw reads from isolates with undetermined MLST types were submitted to Enterobase, which assigned five new sequence types (9656, 9692, 9693, 9694, and 9696). In order to visualize evolutionary relationships between bacterial strains, we used PHYLOViZ 2.0a (Nascimento et al., 2017) that generates complete minimum spanning trees with goeBURST Full MST algorithm.

Core Genome Analysis
Parsnp program from Harvest suite (Treangen et al., 2014) was run to create core genome alignment. The alignment was used to calculate the maximum likelihood phylogenetic tree, with RaxML under GTR-GAMMA model and with 100 bootstrap replicates (Stamatakis, 2014).

SNP Analysis of ST131 Isolates
ST131 isolates were aligned with Parsnp using Escherichia coli EC958 as a reference ST131 strain (Forde et al., 2014). Core genome SNPs from the alignment were extracted with harvest-tools and pairwise SNP distances were used for UPGMA tree calculation conducted in MEGA7 (Kumar et al., 2016). Phylogenetic trees were visualized using iTOL (Letunic and Bork, 2019).

Statistical Analysis
Statistical analysis used Past 3.22 5 . The prevalence of strains, genes, ST and clones were compared by Chisquared test or Fisher's exact test; p < 0.05 was considered statistically significant.

Phenotypic and Genotypic Epidemiology of ESBL/AmpC/Carbapenemases Producing E. coli Strains
A total of 10,780 consecutive E. coli isolates from Estonia, Latvia, Lithuania, Norway and St. Petersburg (Russia) were screened for reduced sensitivity to third-generation cephalosporins. Of these, 5,486 (51%) were recovered from stationary patients and 5,294 (49%) from outpatients. A total of 508 (4.7%) E. coli strains showed ESBL/AmpC phenotype. Significant inter-country differences were found regarding the prevalence of E. coli showing ESBL/AmpC phenotype (Table 1).
ESBL associated TEM genes were found only in 2 cases: bla TEM−29 in Estonian and bla TEM−71 in Lithuanian.

DISCUSSION
This study describes the phenotypic and molecular epidemiology of E. coli strains with reduced susceptibility to third-generation cephalosporins in Northern and Eastern Europe by screening of more than 10,000 E. coli strains. The overall prevalence of ESBL/AmpC strains was 4.7% by phenotypical test and 3.9% by sequencing. We found more strains with the ESBL/AmpC phenotype and genotype in St. Petersburg and Latvia than in other countries. According to our knowledge this is the first study analyzing beta-lactamases epidemiology of E. coli using WGS and describing in detail resistance genes, distribution of MLST and SNP clones in this region.
Although several reports have been previously published, scope, methodology and data quality in these studies vary. Edelstein's group investigated E. coli strains from Russia, and found that the prevalence of phenotypic ESBL-positive strains was close to 16%; however, prevalence figures in different institutions varied from 10 to 90% (Edelstein et al., 2003).
WHO CAESAR study reports high prevalence (66%) of thirdgeneration cephalosporins resistance in invasive E. coli strains in Russia, however the number of strains was small (World Health Organisation, 2016). This high variance might be dependent upon different antibiotic use policies in different Russian regions and hospitals. Our results also showed a relatively high prevalence of ESBL/AmpC phenotype in St. Petersburg area compared to other countries. However, it is impossible to draw final conclusions about the overall prevalence of ESBL/AmpC/Carbapenemases in Russia at large, since the strains were collected only from St. Petersburg region, and thus reflects the situation in only one city.
When comparing ESBL/AmpC prevalence in different studies several aspects should be taken into account. Different methods and criteria have been used in different studies such as decreased sensitivity to third-generation cephalosporins as an indicator of ESBL; phenotypic confirmation test for ESBL alone or ESBL combined with AmpC. In studies where molecular methods were applied, different approaches have been used: searching only for CTX-M types or including also TEM, SHV and AmpC type genes.
Besides differences in detection methodologies several other factors might influence the results and potential cause over as well as underestimation of ESBL/AmpC prevalence and resistance percentages. One such factor is the use of different sampling practices in different institutions noted also by EARS-Net as a factor that should be taken into account in interpreting intercountry differences (EARS-Net, 2018). A similar limitation is present in all studies using clinical strains from routine cultures, including our study.
We found more strains with ESBL/AmpC phenotype than strains with known ESBL/AmpC gene. Several reasons can cause this: other mechanisms such as possible hyperproducers of intrinsic (chromosomal) cephalosporinase combined or not with alteration in porin channels can lead to resistance to thirdgeneration cephalosporins; bla genes databases are not complete -we probably don't know all ESBL/AmpC genes or not all are submitted to databases, furthermore these genes are changing and new variants may not be recognized; we found in our strains several genes (such as SHV and TEM variants) without information about their belonging to particular Bush-Jacoby functional group (ESBL/AmpC or not). Only well described ESBL/AmpC genes were included in this study.
The most common ESBL genes in our study were bla CTX−M−15 and bla CTX−M−14 . These enzymes have been reported throughout Asia, Africa, Europe, America and Australia (Livermore et al., 2007;Sidjabat et al., 2010;Iroha et al., 2012;Chen et al., 2014;Pietsch et al., 2017). So far, the CTX-M-15 genotype appears to be the most prevalent in all continents, and our findings are in accordance with previous reports (Sidjabat et al., 2010;Canton et al., 2012;Iroha et al., 2012;Voets et al., 2012;Brolund et al., 2014;Chen et al., 2014;Bevan et al., 2017;Jorgensen et al., 2017;Pietsch et al., 2017). CTX-M-15 dominates in Germany and the Netherlands, but more recent studies show an increased proportion of CTX-M-1 compared to CTX-M-14 (Voets et al., 2012;Pietsch et al., 2017). CTX-M-14 has been found to be less prevalent in most countries with some exceptions [China, South-East Asia, South Korea, Japan, and Spain; (Onnberg et al., 2011;Copur Cicek et al., 2013;Helldal et al., 2013;Bevan et al., 2017)]. Although we found CTX-M-14 in all investigated countries it was less prevalent than CTX-M-15. We found in few cases (0.7%) the combination of CTX-M-14 and CTX-15. In some regions this combination was frequently reported (Park et al., 2012). Increasing prevalence of CTX-M-27 has been reported worldwide. This genotype is a single nucleotide variant of CTX-M-14 showing higher MIC to ceftazidime and therefore use of ceftazidime would theoretically select it (Bevan et al., 2017). We found only a few CTX-M-27 strains from Estonia, Latvia and Norway. As in previous studies we found the majority of bla CTX−M in plasmids and a minority (20%) in chromosome. However, frequency of chromosomal location of bla CTX−M (mainly bla CTX−M−14 and bla CTX−M−15 ) varies in different regions and studies from<5% in some European countries to 27% in recent Japanese study (Rodríguez et al., 2014;Hamamoto and Hirai, 2019).
In previous studies AmpC prevalence in E. coli was usually low, however in some regions prevalence up to 9% has been reported (Pascual et al., 2016;Zhou et al., 2017;Kazemian et al., 2019;Ribeiro et al., 2019). In our study AmpC prevalence was<1% except for in St. Petersburg (2.1%). In the previous studies, CMY-2 was usually the most common AmpC, however DHA-1 has been reported as dominant in some studies (Brolund et al., 2014;Soraas et al., 2014;Pascual et al., 2016;Kazemian et al., 2019;Ribeiro et al., 2019). In our study bla CMY−59 was dominating in the Baltic States and Norway but bla DHA−1 in St. Petersburg. There are only a few reports about finding bla CMY−59 in clinical strains (Roy et al., 2011;Ranjbar et al., 2013). In some AmpC epidemiology studies, common predominance of "CMY-2 like" genes has been reported without exact gene determination that makes it difficult to compare our data with others (den Drijver et al., 2018;Pietsch et al., 2018).
Only one NDM-1-producing E. coli was found during our study. Carbapenem resistance is still rare among E. coli strains in Europe (0-1.6%) and bla OXA−48 is the most commonly observed carbapenemase. At the same time carbapenem resistant K. pneumoniae is more common in Europe (0-64.7%) with bla KPC and bla OXA−48 predominance (Grundmann et al., 2017;EARS-Net, 2018). However, outbreak of NDM-1producing K. pneumoniae has been reported in St. Petersburg (Pavelkovich et al., 2014). No co-production of NDM-1 and CMY-39 has been reported previously. Prevalence of carbapenemases among other Enterobacterales is probably rare. In our study in Northern and Eastern Europe (2015, including nine countries) only one bla OXA−48 was found in 88 Enterobacterales strains (other than K. pneumoniae) with reduced susceptibility to carbapenems; in the same settings ca 50% of K. pneumoniae strains with reduced susceptibility to carbapenems (n = 171) harbored carbapenemase gene (our unpublished data).
ST131 was also the most common genotype in our study. More than 50% of the Latvian and over one-third of Estonian, Lithuania, and St. Petersburg's E. coli strains belonged to this group. When applying SNP analysis to ST131 strains several clones with cross-border spreading were found.
In general, prevalence of ESBL, AmpC and Carbapenemases genes was low in investigated E. coli strains. However, several inter-country differences notably in distribution of particular genes, MLST groups and SNP clones, were described.

DATA AVAILABILITY STATEMENT
The datasets generated for this study can be found in the NCBI GenBank https://www.ncbi.nlm.nih.gov/bioproject/ PRJNA528606.

ETHICS STATEMENT
Approval was not required as per the local legislation. Institutions used only samples sent for routine diagnostics, no additional sampling being necessary. No patient data was used, and all strains were coded and processed anonymously (it is impossible to identify any patient by strain number).

AUTHOR CONTRIBUTIONS
ES: principal investigator, preparation of the manuscript. RA: bioinformatics. ArB: BEEP/BARN coordinator in Latvia, data and strain collection, critical reading of the manuscript and SNP analyses. AnB: principal investigator, preparation of manuscript, BEEP/BARN Estonian coordinator. AgB: SNP analyses. SE: BEEP/BARN coordinator in Russia, data and strain collection, critical reading of the manuscript. KH: molecular studies, critical reading of the manuscript. MI: BEEP/BARN international technical coordinator, data preparation of the manuscript. LK: BEEP/BARN coordinator in Russia, data and strain collection, critical reading of the manuscript. SK: ARMMD coordinator, preparation of the manuscript. TK: WGS data analyses. MM: BEEP/BARN coordinator in Russia, data and strain collection, critical reading of article. JM BEEP/BARN coordinator in Lithuania, data and strain collection, critical reading of the manuscript. KP: design of molecular studies and strains preparation. MR: WGS data analyses and preparation of the manuscript. TR: strains characterization and responsible for culture collection. PN: scientific coordinator and preparation of manuscript.

ACKNOWLEDGMENTS
We thank the Swedish Institute for Communicable Disease Control for laboratory assistance. We also thank all the clinical microbiology laboratories for their contribution of isolates and data, and Irja Roots for their technical assistance. The final version of our report was prepared for us by BioMedES United Kingdom (www.biomedes.biz).