Characterization of qnrB-carrying plasmids from ESBL- and non-ESBL-producing Escherichia coli

Escherichia coli carrying clinically important antimicrobial resistances [i.e., against extended-spectrum-beta-lactamases (ESBL)] are of high concern for human health and are increasingly detected worldwide. Worryingly, they are often identified as multidrug-resistant (MDR) isolates, frequently including resistances against quinolones/fluoroquinolones. Here, the occurrence and genetic basis of the fluoroquinolone resistance enhancing determinant qnrB in ESBL-/non-ESBL-producing E. coli was investigated. Overall, 33 qnrB-carrying isolates out of the annual German antimicrobial resistance (AMR) monitoring on commensal E. coli (incl. ESBL-/AmpC-producing E. coli) recovered from food and livestock between 2013 and 2018 were analysed in detail. Whole-genome sequencing, bioinformatics analyses and transferability evaluation was conducted to characterise the prevailing qnrB-associated plasmids. Furthermore, predominant qnrB-carrying plasmid-types were subjected to in silico genome reconstruction analysis. In general, the qnrB-carrying E. coli were found to be highly heterogenic in their multilocus sequence types (STs) and their phenotypic resistance profiles. Most of them appeared to be MDR and exhibited resistances against up to ten antimicrobials of different classes. With respect to qnrB-carrying plasmids, we found qnrB19 located on small Col440I plasmids to be most widespread among ESBL-producing E. coli from German livestock and food. This Col440I plasmid-type was found to be highly conserved by exhibiting qnrB19, a pspF operon and different genes of unassigned function. Furthermore, we detected plasmids of the incompatibility groups IncN and IncH as carriers of qnrB. All qnrB-carrying plasmids also exhibited virulence factors and various insertion sequences (IS). The majority of the qnrB-carrying plasmids were determined to be self-transmissible, indicating their possible contribution to the spread of resistances against (fluoro)quinolones and other antimicrobials. In this study, a diversity of different plasmid types carrying qnrB alone or in combination with other resistance determinants (i.e., beta-lactamase genes) were found. The spread of these plasmids, especially those carrying antimicrobial resistance genes against highest priority critically important antimicrobial agents, is highly unfavourable and can pose a threat for public health. Therefore, the dissemination pathways and evolution of these plasmids need to be further monitored.

Background The spread of antimicrobial-resistant bacteria is a global concern. Commensal E. coli can acquire and spread various antimicrobial resistance genes sometimes leading to an emergence of multidrug-resistant (MDR) isolates in the livestock, food, and human sector, worldwide. Among antimicrobial-resistant bacteria, extended-spectrum beta-lactamases (ESBL)-producing E. coli (ESBL-EC) are of particular concern. While ESBL-EC had become common in hospitalised and healthy people [1], they are also increasingly detected in livestock, food and the community [2]. Their occurrence in livestock and food poses a threat to public health due to a possible transmission to humans via direct contact to colonised animals or the consumption of contaminated food products [1]. Frequently, ESBL-EC carry additional AMR genes, which result in non-wildtype phenotypes against substances of other antimicrobial classes, which can affect an efficient treatment of infections during hospitalization. Prevailing reports highlighted that ESBL-EC isolated from food samples are often associated with increased minimum inhibitory concentrations (MICs) and resistances against avilamycin, colistin, quinolones and fluoroquinolones [3,4]. Especially, the co-occurrence of genes conferring resistances against quinolones or fluoroquinolones [further designated as (fluoro)quinolones] are of great concern. Fluoroquinolones are classified as highest priority critically important antimicrobials for the treatment of human infections [5], but are also used in the veterinary sector. (Fluoro)quinolone resistance is mainly caused by alterations in the quinolone resistance-determining regions (QRDR) within the E. coli chromosome. However, plasmid-mediated quinolone resistance (PMQR) genes are also responsible for a decreased susceptibility or resistance of isolates [6]. As PMQR genes are usually present on mobile genetic elements (MGE) their spread to different bacteria and environments is feasible. Colocalization of PMQR genes and other resistance determinants on MGEs can contribute to MDR development and persistence by co-selection [7]. Interestingly, characterised ESBL-EC were increasingly reported as carriers of qnr genes (especially qnrB) [8][9][10][11][12][13][14][15] in Europe, the United States, Asia and Africa [16], with different genes and variants of both determinants (bla and qnr, predominantly qnrB1 and bla CTX-M-9 , bla CTX-M-3 or bla SHV-12 ) co-localised on the same plasmids. As many different plasmidtypes were shown to be responsible for the spread of qnr genes, a deeper understanding of their occurrence, diversity and transmission pathways is necessary.
In this study, we investigated the composition of plasmids carrying qnrB variants from ESBL-EC and non-ESBL-EC originating from the annual resistance monitoring from livestock and food in Germany. Indepth characterization of the plasmids was conducted by phenotypic analyses in combination with whole-genome sequencing (WGS) and bioinformatics analysis of the isolates. The isolates were studied to gain insight into the properties of qnrB-carrying MGEs as well as into their transmission potential. Thus, the risk for the spread of MDR E. coli outgoing by these qnrB-carrying ESBL-EC is discussed.

Characteristics of ESBL-/non-ESBL-EC carrying qnrB
Based on available E. coli sequences from annual resistance monitoring programs conducted by the National Reference Laboratory for Antimicrobial Resistances (NRL-AR) hosted at the German Federal Institute for Risk Assessment, 33 ESBL (n = 29)/non-ESBL-EC (n = 4) carrying qnrB were identified and subjected to further characterization. The isolates originate from samples of livestock and food taken between 2013 and 2020 and were provided from different German federal state laboratories. In Table 1, essential information on the main characteristics of these isolates is given.
Antimicrobial susceptibility testing (AST) confirmed non-wildtype phenotypes for the majority of the investigated isolates against the beta-lactam antimicrobials ampicillin, cefotaxime and ceftazidime as well as against the (fluoro)quinolones nalidixic acid and ciprofloxacin for the ESBL-EC. Thus, these isolates exhibited at least resistances against critically important antimicrobials of Keywords: E. coli, qnrB, Fluoroquinolone, Plasmids, Inc-group Table 1 Main characteristics of ESBL-EC (A, n = 29) and non-ESBL-EC (B, n = 4) associated with qnrB from the WGS collection of the NRL-AR. Besides basic metadata (year of sampling and source), information on phenotypic resistance profiles, in silico-based prediction of acquired resistance determinants and chromosomal sequence alterations associated with (fluoro) quinolone resistance development as well as the MLST type is given If no gene variant is given, the bioinformatics analysis was unable to specify the gene variant two different classes. However, a few isolates also exhibited non-wildtype phenotypes against other important antimicrobials like colistin (n = 8). The phenotypic data was in good agreement with the AMR genes detected by bioinformatics analysis. In silico-typing revealed the presence of multiple AMR genes per isolate (Supplemental Table 1 and Table 1). Most frequently, the beta-lactamase gene bla TEM-1 and bla TEM-135 were detected, while qnrB19 represented the dominant (fluoro)quinolone resistance determinant. Analysing the point mutations within the chromosomal sequence of gyrA, gyrB, parC, parE, pmrA, pmrB, folP, 23S/16S rRNA as well as the ampC and rpoB regions, of which some of the genes are known to be associated with (fluoro)quinolone resistance, we found yet uncharacterised alterations in the above-mentioned sequences within every isolate. For ten isolates, previously characterised point mutations were detected, known to be involved in a decrease of the susceptibility against (fluoro)quinolones. The most frequently appearing point mutations within the E. coli are alterations of gyrA leading to a decreased susceptibility against nalidixic acid and ciprofloxacin (Table 1). Besides, we detected mutations leading to changes of the amino acid sequence of ParC and ParE also affecting the (fluoro) quinolone susceptibility of the isolates.
To assess a potential transferability of the AMR genes and the association of qnrB genes and mobile elements, WGS data of the isolates were used. Potential plasmid sequences of all isolates could be assigned to 24 distinct incompatibility groups (Supplemental Table 2).

Characterization of qnrB-carrying plasmids
Among the 33 qnrB-carrying E. coli isolates, four harboured qnrB1, one qnrB2 and 28 qnrB19. The in silico analysis further revealed an association of qnrB to contigs of plasmidic origin. Twenty-four isolates harboured qnrB on a Col440I replicon carrying plasmid contig. Furthermore, individual plasmid types represented by replicon sequences of the incompatibility groups IncN and ColE10 were detected (n = 2). For the remaining isolates (n = 7), the qnrB-carrying contig could not be associated with a specific replicon type backed in the PlasmidFinder database. Overall, we were able to close the qnrB-carrying plasmid genomes of 12 isolates, solely through assembling with unicycler. Interestingly, the majority of those closed plasmids belonged to the Col440I type, while one was assigned to ColE10 (Table 2). We were able to close the remaining plasmids carrying the qnrB gene on a Col440I plasmid through primer walking.
For the determination of the qnrB-plasmid diversity, the genetic basis of the most prominent types was analysed, including Col440I-, IncN-and IncH2/ IncH2A-plasmids.

qnrB19 genes on Col440I plasmids
The qnrB-associated Col440I plasmids of this study exhibited genome sizes ranging between 2,700 and 3,500 bp with GC-contents of 47.7% to 50.9%. Therewith, we found five reference plasmids representing the different Col440I plasmid-types (Table 2). According to a phylogenetic comparison as shown in Fig. 1, the qnrB19 Col440I plasmids clustered into three distinct clades. While clade (A) was only represented by the reference plasmid pEC14-9 (NC_013782), the remaining clades exhibited a slight diversity in their assigned reference genomes. However, Clade (B) was best assigned to the reference plasmid p14-7355.2 (NZ_CP039609), while the most distant clade (C) matched the best to the reference plasmid p3_12888 (NZ_CP045448).
All Col440I plasmids were equipped with a pspF operon transcriptional activator, upstream of the coding sequence (CDS) of the pentapeptide-encoding gene qnrB19. Close to the replication initiation protein, the gene for the transcription factor Sp1 was identified, as well as a putative CDS of a yet uncharacterised function (Fig. 2). The main differences of the phylogenetic clusters are caused by the occurrence of these sequences encoding proteins of unknown function. On the Col440I genomes, no further AMR, virulence, or biocide resistance genes were detected. Interestingly, these plasmids also lack insertion sequences potentially associated with a further dissemination of the qnrB gene. In Fig. 2, the genetic diversity of the Col440I-plasmid is shown. Furthermore, the organization and function of the CDS predicted on the Col440I-genomes are given. Based on the prevailing data a core genome of this plasmid type can be assigned to the presence of qnrB19 and pspF gene.
By analysing the Col440I genomes with the in silico mob-suite tool, the plasmids were assigned to be mobilizable. In general, the origin of transfer (oriT) of the plasmids was subjected to the conjugative transfer system of the MOB P type, but no type IV secretion system (T4SS) was detected. Thus, the plasmid is transmissible by a helper plasmid but not self-transmissible. However, in vitro studies using the E. coli isolates for filter mating with the sodium acid-resistant E. coli strain J53, yield negative results under double selective conditions (NAL/ SAC), suggesting that the plasmid could not be transmitted due to a lack of a suitable helper plasmid or based on missing determinants or altered sequences involved in mobilization.

Characterization of IncN plasmids carrying qnrB2
Out of the prevailing qnrB collection, only one ESBLproducing E. coli, from cattle, represented a qnrB2 determinant. This one was located on an IncN plasmid. qnrB2 was identified on a 39,563 bp contig, which could not be circularised to a complete plasmid genome using unicycler alone. However, reference-based mapping with plasmidID against a complete plasmid database revealed a close relationship (92.33% reference coverage) to the E. coli plasmid pHHA45 (NC_019098). This plasmid represents an IncN plasmid of 39,510 bp with a GC content of 50.5%. A sequence comparison of both plasmid sequences revealed the presence of repetitive sequences on pHHA45, which might be responsible for failing circularisation by using the unicycler program.
However, the derived contigs of the isolate comprised all essential information for the evaluation of the impact of this IncN plasmid type. In proximity to qnrB2, the plasmid carried a bla CTX-M-1 gene, which is also present on the reference plasmid pHHA45 (NC_019098) (Fig. 3). The bla CTX-M-1 gene is flanked by  The schematic illustration provides no information on the actual plasmid size but the size and position of the annotated genes. The ruler given below the gene map represents an artificial indicator for the size of the region IS26 elements. Upstream to qnrB2, the dihydropteroate synthase (encoded by folP) and bla CTX-M-1 flanked by IS26 elements are located. Downstream of qnrB2, sapA encoding a peptide transport periplasmic protein and a further dihydropteroate synthase was identified. Moreover, the multidrug transporter gene emrE, associated to the transposase ISSsu9, was located on this plasmid. The mob-suite analysis predicted that the IncN plasmid is self-transmissible (conjugative). In silico analysis revealed the presence of MOB F relaxase and the MpfT mating pair formation (mpf) system. The predicted self-transmissibility of the plasmid could be confirmed by in vitro filter mating studies. We observed a transmission rate of 10 5 to 10 6 per donor cell.

Characterization of IncH2/IncH2A plasmids carrying qnrB1
Four of the investigated E. coli revealed plasmid sequences associated with qnrB1 and IncHI2-IncHI2A replicon sequences. Here, we were able to assign contigs, bioinformatically clustering together, from our isolates to a reference plasmids, carrying the qnrB1 as well as the IncH replicon sequence on the same plasmid. Therewith, these contigs of our isolate fully covered the qnrB1 and IncH replicon sequence region. Two of the isolates were recovered from cattle, one from pig and one from poultry. De novo assemblies yielded qnrB1-carrying contigs of 6 and 12 kb for three and one isolate, respectively. Using the plasmidID tool we were able to identify p23_A-OXA140 (NZ_ CP048350; 99% coverage) and pCRENT-193_1 (NZ_ CP024813; 99% coverage) as the closest relatives. The respective reference plasmids ranged between 279 and 298 kb in size and exhibited GC contents of 48%. Both reference plasmids are 92% identical at nucleotide level to each other. Here, qnrB1 was located downstream to the pspF and a Tn3 family transposase gene (Fig. 4).
Here, the IS6 family transposons (IS26) flanked an area, containing the AMR genes qnrB1, aac(3)-lle, aac(6')-Ib-cr5 and bla OXA-1 . All AMR genes detected on these plasmids are located in a region of 40 kb. in silico analysis revealed a self-transmissibility of the plasmid due to the detection of the Mpf T system and the MOB H relaxase type. The plasmid pCRENT-193_1(NZ_CP024813) carries multiple AMR genes like aac(3)-IIe, aac(6')-Ib-cr5, aadA1, aph(3'')-Ib, aph(6)-Id, bla CTX-M-15 , bla OXA-1 , bla TEM-1 , catA1, catB3, dfrA14, qnrB1, sul2 and tet(A). The same genes were detected in the WGS data of 18-AB00078 analysed in this study. A similar resistance gene profile, only lacking aadA1 and catA1, was detected on the p23_A-OXA140 genome (NZ_CP048350). However, aadA1 and catA1 were present in the respective matching plasmids from our investigated isolates and resembled to the reconstructed qnrB1 plasmid. Furthermore, we detected the hipA gene, coding for a serine/threonine-protein kinase toxin, which is the toxic component of a type II toxin-antitoxin (TA) system. The binding partner HipB encoding gene was not detected in proximity of hipA and could not be identified in the complete WGS dataset representing the complete E. coli isolate. We also detected the tellurium ion resistance gene terC, often associated with pathogenic bacteria [17], in the investigated plasmids. Overall, all qnrB-carrying plasmids were shown to be self-transmissible under natural condition based on the observed transfer rates of 10 2 to 10 4 per donor cell.

Characteristics of ESBL and non-ESBL E. coli carrying qnrB-plasmids
In this study, 33 ESBL-/non-ESBL-EC carrying qnrB on an extrachromosomal element were characterised in detail. This co-occurrence of ESBL and (fluoro)quinolone resistance genes in E. coli poses a threat to public health, as these antimicrobials are highest priority critically important substances in human medicine.
Among the investigated isolates, a broad variety of sources of qnrB-carrying E. coli was found. While poultry seems to be the predominant source for qnr genes [18], plasmids with PMQR were also found in other sources. The presence of qnrB in ESBL-EC from poultry is frequently reported. The latest summary report of the European Food Safety Authority (EFSA) on AMR in zoonotic and indicator bacteria from humans, animals and food [18] also addresses this trend. Although the incidence of ESBL-EC was generally low, it was most often detected in broiler isolates (Member State group level of up to 30%). However, EFSA further reported a high level of ciprofloxacin-and nalidixic acid-resistant E. coli especially from broilers (median 73.5% for ciprofloxacin and 64.1% for nalidixic acid) and turkeys. Thus, poultry seems to be a common reservoir for ESBL-EC and (fluoro)quinoloneresistant E. coli. General, the occurrence of MDR E. coli, as characterised in this study, along the food chain poses a risk for a transmission of these bacteria to human via food products.
Interestingly, qnrB-positive ESBL-EC were also detected among isolates of the international high-risk clone ST131 [19] and O89 serotype. ST131 isolates are known to represent a predominant sequence type among extraintestinal pathogenic E. coli, which comprise ESBLpositive as well as (fluoro)quinolone-resistant isolates [20]. E. coli of the serotype O89 are often associated with MDR [21]. Based on the results of this study, a similar association was observed for these E. coli types. The occurrence of qnrB-carrying plasmids in various STs of ESBL-EC further demonstrated that these plasmids exhibit a broad adaptability to E. coli of different ST.
It had been discussed before, that qnrB represents the dominant PMQR group in humans, while qnrS seemed to be more frequent in the environment [22], the veterinary and food sector [23][24][25][26]. This emphasises the need for a better understanding of the composition and impact of qnrB-carrying plasmids to estimate the transmission possibilities from animal to humans.
Here, the ESBL-EC were mostly phenotypically resistant against penicillins, cephalosporins and (fluoro) Fig. 4 Organization of the qnrB1 coding region on the IncH plasmids. The region represents the contig as detected for the four investigated qnrB1 carrying E. coli of our study. This contig, bioinformatically mapped to the reference plasmids containing the IncH replicon sequence, did always consist of the same structure as presented here quinolones. Further, every isolate carried a bla gene coding for different TEM-or less frequently CTX-enzyme variants. For qnrB, the variant qnrB19 seemed to be predominant. Interestingly, some isolates showed genes encoding penicillinases and ceftazidime resistance. This observation needs to be further verified and characterised in detail to determine, if also other mechanisms like mutations in PBP3 and efflux pumps can cause these effects. However, all isolates of this study further exhibited phenotypical resistance to antimicrobials of other classes and were shown to carry various AMR genes cooccurring in the same isolate.
The detected virulence factors (ompA, csg, fim, chu, iroN, kpsM, irp, iuc, pap, pic and vat) may contribute to an increase in the pathogenic potential of these E. coli. The outer membrane protein A (OmpA) contributes to pathogenesis. The capsular antigen (KpsM) represents a protection factor against phagocytosis. The siderophore aerobactin gene (iuc) as well as irp, iroN and chu are associated with iron uptake often present in uropathogenic E. coli (UPEC). The genes pap (coding for P fimbriae), fim (type 1 fimbriae) and csg (curli fibers) contribute to the adhesion properties of the E. coli. The serine protease autotransporter encoding gene pic and the vacuolating autotransporter encoding gene vat do represent toxins. All detected factors individually contribute to an increase in the pathogenic potential of E. coli [27]. The presence of these virulence factors, in qnr-carrying ESBL-EC demonstrates an aggregated risk. As virulence factors are also frequently present on plasmids [28], their potential spread can increase the clinical impact of the bacterium dramatically. Further subtyping results, i.e., the phylotype and detected fim variants are given in Supplemental Table 3.
While PMQR genes are the main contributors for horizontal (fluoro)quinolone resistance transmission, alterations in the sequences of the DNA gyrase and topoisomerase IV genes are the main reason for resistance against (fluoro)quinolones in E. coli [29]. We detected previously determined single nucleotide polymorphisms (SNPs) leading to high-level resistances within ten of 33 isolates. The SNPs where mainly identified in the genes gyrA and parC. Especially the mutations in the S83L (in GyrA) and D87N (in ParC), as found here, are common [30,31]. These mutations, in combination with the carriage of a qnr gene, are responsible for the (fluoro) quinolone resistance phenotype [32]. However, we further detected yet uncharacterised SNPs in the QRDR for every analysed isolate (Supplement Table 4). As the study only focused on (fluoro)quinolone-resistant E. coli, it might be possible that these mutations also contribute to the observed (fluoro)quinolone resistance. Another appropriate interpretation might be a higher contribution of the qnrB genes to (fluoro)quinolone resistance, than commonly expected [33].

Prevalent qnrB-carrying plasmids in ESBL-/non-ESBL-EC
We identified a 39.5 kb IncN plasmid carrying qnrB2 in combination with bla CTX-M-1 surrounded by IS26 elements as previously described [34,35]. The bla CTX-M-1 in proximity to IS26 elements was also detected on other plasmid types, suggesting that transmission of this specific region took place via IS26-mediated transfer [36]. The folP gene, identified upstream of the qnrB2 gene, is another characteristic of this plasmid. The folP gene encodes the dihydropteroate synthase (DHPS) enzyme, which is usually encoded on the chromosome and represents the target of sulfonamides [37,38]. The presence of folP on the plasmid may represent a genetic advantage for E. coli, as it ensures the folate biosynthesis pathway. Furthermore, the MDR transporter, encoded by ermD was detected. It represents a small MDR transporter known to confer resistance to a broad spectrum of disinfectants and quaternary cation compounds [39]. The gene ermD was detected close to the ISSsu9, also suggesting ISmediated transfer. In this study, we isolated the plasmid from bovine E. coli. However, similar IncN plasmids were identified in isolates of various sources (food, livestock and humans) from the Czech Republic, Poland, Denmark and Italy [40]. Dolejkska et al. [41] described comparable plasmids also carrying a bla CTX-M-1 , in addition to qnrS1 or qnrB19, but to the best of our knowledge not together on the same IncN plasmid. As this plasmid was determined to be conjugative as well as to be a broad-host range plasmid [40], the risk resulting from this special IncN plasmid and the evolvement of AMR gene accumulation should be further monitored. Another detected plasmid type in this study were plasmids of the IncHI2-IncHI2A incompatibility group. We detected this plasmid type in association with a qnrB1 gene in isolates of different animal sources, which suggests a possible broad dissemination. We identified multiple AMR genes on our qnrB1-carrying plasmids. All IncH plasmids from this study exhibited the following AMR genes: aac(3)-IIe, aac(6')-Ib-cr5, aadA1, aph(3'')-Ib, aph(6)-Id, bla CTX-M-15 , bla OXA-1 , bla TEM-1 , catA1, catB3, dfrA14, qnrB1, sul2 and tet(A), leading to resistance against aminoglycosides, beta-lactams, chloramphenicol, trimethoprim, (fluoro) quinolones, sulphonamides, and tetracyclines. IncH-like plasmids were previously reported as accumulators, carriers and spreaders of various resistances [42]. The carriage of multiple AMR genes, as present for this plasmid type, probably presents a risk when transmitted. The cooccurrence of genes conferring resistance against two broad antimicrobial classes is alarming. With the presence of bla CTX-M-15 , bla OXA-1 and bla TEM-1 , three different beta-lactamase genes were located on the same qnrB1carrying plasmid. The IS26 element in proximity to qnrB1 and bla OXA-1 is known to be responsible for spreading multiple AMR genes [43]. Varani et al. described the importance of IS26 in clinical settings. They mentioned an increased frequency of plasmids carrying IS26 involved in aggregation of antimicrobial resistance genes [43]. Harmer and colleagues described IS26 as key element for the dissemination of AMR genes in Gram-negative bacteria [44]. We further detected hipA on the IncH plasmid. We were not able to detect the antitoxin component hipB, neither in other plasmids nor in the chromosomal DNA of the respective isolates. Thus, it remains unclear how the E. coli copes with the burden of the toxin produced from hipA. We can assume that the presence of this component is a benefit for the plasmid stability within the isolate. The detected terC virulence factor on the plasmid is quite common for IncH plasmid types. It was described as responsible for the control of resistance to infections by some bacteriophages [45]. Thus, it may confer another advantage for the host to retain the respective plasmid. Further, IncH-type plasmids were often detected in animal and human isolates and seemed to be disseminated among different sources contributing to the spread of AMR genes from animals to humans or vice versa [45]. As we determined the self-transmissibility of the IncH-like plasmids characterised within this study, a possible spread of this plasmid carrying multiple AMR genes is indeed given.
The most frequently detected qnrB plasmid type in the ESBL-/non-ESBL-EC in this study belonged to the Col440I-like group. All Col440I plasmids carried a qnrB19 gene and a pspF operon, as well as the gene for the transcription factor sp1. The protein sequence of the hypothetical protein as well as the non-coding regions altered within the different clusters. With the detection of the mob P relaxase gene, this plasmid was categorised as mobilizable, but not self-transmissible. The same plasmid (100% identity) has been described before with exactly the same genome structure but was assigned to a different plasmid type. Karczmarczyk and colleague identified this "ColE-like" plasmids from food samples in Colombia to carry qnrB19 [46]. Pallecchi et al. characterised the same structure as ColE-like plasmid. They found this small qnrB19-carrying plasmid in E. coli from humans around Latin America with a high frequency and suggested a major role of this small plasmid in qnrB dissemination. As the plasmid is small and contains only a few genes, the authors hypothesised that it could have undergone a subsequent excision [47]. This was supported by other studies, describing qnrB19 within a comparable genetic environment in larger plasmids, associated with ISEcp1C-based transposons [48]. Moreno-Switt et al. [49] described that this small qnrB19-carrying plasmid was reported in Europe, the U.S.A. and South America in Salmonella obtained from food, animals and humans. They demonstrated how this qnrB19-carrying plasmid type was transmitted between different Salmonella serotypes through a P22-mediated transduction, probably explaining the frequent detection of this small plasmid. Although we only detected qnrB19 on the small Col440I plasmids, there are some studies available, presenting the qnrB19 on different plasmids also containing bla TEM-1 or bla SHV-12 [50]. Blasting the Col440I plasmid against the NCBI database, we detected an 11.3 kb plasmid (FDAAR-GOS_1249) containing the backbone of the qnrB19-carrying plasmid but also additional genes, like the plasmid mobilization gene encoding MOB C .
Overall, we were able to thoroughly determine and characterize the structures of the plasmids carrying the qnrB gene. However, as we investigated the isolates with short-read sequencing the limitations for closing these plasmids has to be mentioned. Although, in silico estimation of the whole plasmid from short reads is getting more reliable, an optimised approach would include long-read sequencing of the plasmids of interest. As previously shown, especially for the determination of large plasmid genomes long read sequencing is necessary [51]. Due to the occurrence of mobile genetic elements or repetitive sequences, short read sequencing techniques represent limitations for addressing this issue.

Risk posed by qnrB-carrying plasmids in ESBL-/ non-ESBL-EC
We detected different qnrB genes on plasmids within the E. coli isolates of this study. All investigated isolates were resistant to (fluoro)quinolones. Usually, plasmidic factors were accounted only with a decrease of the susceptibility of the isolate not necessarily resulting in a non-wildtype phenotype of the isolates. However, not all isolates carried a known mutation within the respective QRDRs. Thus, it might be possible that the presence of a qnrB gene without any other yet characterised chromosomal alteration in the PMQR or the presence of other plasmidic factors can lead to a resistance phenotype for (fluoro)quinolones. Different studies had already explained, how the qnr genes are able to alter the resistance against (fluoro)quinolones due to mutations in the chromosomal QRDR regions and how the presence does allow other antimicrobial resistance genes to enter and persist. Thus, Li and colleagues reported on how QnrB promotes DNA replication stress that leads to an increased bacterial mutation risk. In their investigations, they measured a two-fold increase in the mutation rate, when QnrB is expressed. Further, they found how QnrB is responsible for the accumulation of mutations, also including quinolone resistance mutations. Overall, they suggested that QnrB could be in charge for promoting the persistence of plasmids leading to resistance against respective antimicrobial agents [52]. These results let us assume, that the presence of qnrB genes in our isolates are also responsible for an environment allowing some mutations to occur and therewith promote the presence of different resistance profiles.
When the presence of qnrB was investigated for the first time in-depth, the observation of its association with ESBL-producing bacteria was mentioned [53]. Jacoby et al. explained how the qnrB-carrying isolates, primarily detected in the U.S.A. and India, were always present with bla SHV-12 or bla CTX-M-15 genes on the same plasmid [53]. The combined presence of extended-spectrum betalactamases encoding bla and qnr genes together within the same isolate, and/or on the same plasmid, is highly unfavourable for medical treatment. Both genes are able to confer resistance against two important classes of antimicrobial agents. Their spread to different sources, as different bacterial species, different environments or to humans is a high risk. Kawamura et al. described how ESBL-EC have become common among healthy people worldwide. They explained how most ESBL-EC usually acquired co-resistance to (fluoro)quinolones and other clinically important antimicrobial agents [1]. As we detected the ESBL and non-ESBL-EC, carrying resistance determinants against (fluoro)quinolone within isolates recovered from livestock and food, one could assume, that the respective plasmids had spread over different areas, thus, demonstrating the necessity of a One-Health approach when estimating the risk especially arising from qnr-carrying ESBL-EC and from non-ESBL-EC.

Conclusion
In this study, we described and analysed the presence of different ESBL-/non-ESBL-EC carrying a qnrB resistance gene. We found qnrB1 and qnrB2 genes to be present on larger plasmids, carrying multiple antimicrobial resistance genes, including different bla genes (e.g. bla CTX-M-15 , bla TEM1 , bla OXA-1 ). Further, we found qnrB19 in particular on small Col440I plasmids. The presence of these PMQR genes together with extended-spectrum beta-lactamases encoding different bla genes in the same isolate or even on the same plasmid harbors a risk for public health. Especially the small Col440I plasmids seem to play an important role in the dissemination of qnrB19 genes, as they frequently had been described in E. coli and other Enterobacteriaceae from various sources and areas. The spread of MDR conferring plasmids contributes to an impaired treatment possibility. Thus, their evolution should be studied further, especially with an One-Health approach.

Isolate selection
For this study, E. coli collected during the German AMR monitoring of commensal (ZoMo-monitoring) and ESBL-/AmpC-producing (ESBL-monitoring) from food and livestock were chosen, as directed by the Federal Office of Consumer Protection and Food Safety (BVL). In addition, isolates carrying qnrB genes from research projects (indicated in the designation of MO) were added. Characterization of the isolates was based on WGS data obtained from Illumina NextSeq sequencing. ESBL-/ non-ESBL-EC carrying qnrB genes were further investigated here. In total 33 epidemiologically unrelated E. coli, carrying qnrB were included. All isolates were initially cultivated on lysogeny broth [Luria Broth Base (Miller's LB Broth Base), Invitrogen-Thermo Fisher, Darmstadt, Germany] for 16-18 h at 37 °C.

Antimicrobial susceptibility testing
Antimicrobial susceptibility testing (AST), of E. coli was conducted by determining minimum inhibitory concentrations (MICs) (including sulfamethoxazole, trimethoprim, ciprofloxacin, tetracycline, meropenem, azithromycin, nalidixic acid, cefotaxime, chloramphenicol, tigecycline, colistin, ampicillin, gentamicin and ceftazidime). Broth microdilution was performed according to EUCAST recommendations on a standardised European antimicrobial test panel (EUVSEC/EUVSEC2; Sensititre ™ , TREK Diagnostic Systems, UK). Therewith, all antimicrobials were tested in ranges as given by the European Commission Implementing Decision No. 2013/652/EU [54]. The results were interpreted according to EUCAST epidemiological cut-off values (ECOFFs) [55]. For quality assessment during MIC evaluation, the E. coli isolate ATCC 25,922 was included in every measurement. An ESBL-phenotype was assigned when the following phenotypic patterns were detected: cefotaxime or ceftazidime > 1 mg/L and meropenem ≤ 0.12 mg/L and cefoxitin ≤ 8 mg/L and no combination of cefotaxime/clavulanic acid and/or ceftazidime/ clavulanic acid. Strains were further determined as phenotypically resistant against (fluoro)quinolones when expressing a MIC of 32 mg/L for nalidixic acid and/or 0.12 mg/L for ciprofloxacin. For better monitoring, we chose all isolates representing these criteria or one MIC step lower.

Whole-genome sequencing
For conducting whole-genome-sequencing (WGS), genomic DNA of E. coli was extracted using the Pure-Link Genomic DNA Mini Kit (Invitrogen-Thermo Fisher) according to the manufacturer's recommendation. The library for sequencing purposes was generated with the Nextera DNA Flex Library Preparation Kit (Illumina ® , San Diego, CA, USA) as previously described by Borowiak et al. [56]. Short-read sequencing was performed in paired-end WGS mode in 2 × 151 cycles with the Illumina ® NextSeq ™ 500/550 Mid Output Kit v2.5 (300 Cycles). If appropriate for plasmid closing, a primer-walking approach was applied. Therefore, the gDNA was used as template with primers (Supplement Table 5) derived from the assembled plasmid genomes. The amplification was performed in a Bio-Rad Thermal Cycler (Bio-Rad, Feldkirchen, Germany). The purification of the PCR amplification products was performed using the illustra GFX PCR DNA and Gel Band Purification Kit (Cytiva Europe, Freiburg, Germany). Sanger sequencing was performed by Eurofins Genomics (Eurofins Genomics, Ebersberg, Germany).

Bioinformatics analysis
Raw reads were trimmed with Aquamis [57] and de novo assembled using unicycler [58]. Quality assessment of the assemblies was conducted with quast [59]. Analysis of virulence factors, antimicrobial resistance genes, serotype and plasmid markers was conducted with bakcharak [60]. The prediction of the 7-locus MLST for each E. coli was screened with the tool v0.2 (https:// github. com/ tseem ann/ mlst) [61]. The annotations of the generated fasta files were achieved with prokka v1.14.5 [62]. To estimate the regions of local similarity between sequences BLAST was utilised (https:// blast. ncbi. nlm. nih. gov/ Blast. cgi). We used the PointFinder tool v2.2 [63] for detecting alterations in the chromosome responsible for (fluoro)quinolone resistance in E. coli. The refSNPer tool v1.0.0 was used as described before [64] to find best matching reference plasmids. To confirm the estimation, we screened our contigs, whether they were assignable to the estimated reference plasmid. We conducted this screening with the gplas tool v0.6.0. The gplas tool bins the respective contigs and therefore allows accurate plasmid-prediction [65]. To visualize and confirm these findings we used the mapping-based and assembly-assisted plasmid identification tool plas-midID v1.6.5 (https:// github. com/ BU-ISCIII/ plasm idID) with a plasmid database containing all available closed plasmids from NCBI. For detecting all present MGEs within one isolate, the MGEfinder v1.0.6 was used [66]. For estimating the conjugative properties of the isolates, through detection of the MOB and mpf determinants, mob-suite v3.0.1 was utilised [67]. To detect the presence of Toxin-Anti-Toxin systems, we used the sling tool v2.0 [68]. For estimation of the phylotype, the phylotyper superphy (https:// github. com/ super phy/ insil ico-subty ping) was used. Finally, BRIG was used for visualizing the circular comparison between genomes [69]. For generating and visualizing the phylogram, Clustal Omega was used to generate the newick file [70]. This format was then used in iTOL [71]. Further, the alignment was visualised in geneious [72]. An arrow-based alignment was created with the gggene v0.4.1 extension of ggplot2 in R (https:// wilkox. org/ gggen es/). If not mentioned otherwise, all tools were used with default parameters.

in vitro filter mating studies
The transferability of the qnrB-carrying plasmids was determined by filter mating studies on solid LB agar. Liquid cultures of donor and recipient bacteria were mixed in a ratio of 1:2 (500:1000 µl), centrifuged at 5,000 × g for 5 min. Thereafter, the supernatant was discharged, and the pellet was resuspended in 150 µl LB broth, applied on a 0.2 µm pore-size filter on an LB agar plate and subjected to incubation at 37 °C. After 24 h, bacteria were resuspended from pore-size filters in 4 ml LB broth and 100 µl of the bacterial suspension was applied on double selective agar plates [nalidixic acid (NAL) 8 mg/L and sodium acid (SAC) 100 mg/L]. After 24 h of incubation at 37 °C, the plates were interpreted for presumptive transconjugants. To confirm successful plasmid transmission, the transconjugants were subcultivated on double selective LB agar (NAL/SAC) and further analysed for the presence of the qnrB gene by PCR.