Quinolone resistance genes qnr , aac(6 ′ )-Ib-cr , oqxAB , and qepA in 1 environmental Escherichia coli : insights into their genetic 2 contexts from comparative genomics

16 Previous studies have reported the occurrence of transferable quinolone resistance 17 determinants in environmental Escherichia coli . However, little is known about their vectors 18 and genetic contexts. To gain insights into these genetic characteristics, we analyzed the 19 complete genomes of 53 environmental E. coli isolates, including 20 sequenced in this study 20 and 33 sourced from RefSeq. The following transferable quinolone resistance determinants 21 were detected: qnrS1 (n = 33), aac(6 ′ )-Ib-cr (n = 12), qnrS2 (n = 5), oqxAB (n = 4), qnrB4 (n 22 = 3), qnrD1 (n = 3), qnrB7 (n = 2), qnrB19 (n = 2), qepA1 (n = 1), and qnrA1 (n = 1). These 23 resistance genes were detected on plasmids of diverse replicon types; however, aac(6 ′ )-Ib-cr , 24 qnrS1 , and qnrS2 were also detected on the chromosome. The genetic contexts surrounding 25 these genes included not only those previously reported in clinical isolates but also novel 26 contexts, such as qnrD1 embedded within a composite transposon-like structure bounded by 27 Tn 3 -derived inverted-repeat miniature elements (TIMEs). This study provides deep insights 28 into mobile genetic elements associated with transferable quinolone resistance determinants, 29 highlighting the importance of genomic surveillance of antimicrobial resistant bacteria in the 30 environment.


INTRODUCTION
Escherichia coli is a commensal member of the gut microbiota of humans and animals but can also cause intestinal and extraintestinal infections. 1 Another concern besides pathogenicity is the rise of antimicrobial resistance (AMR) in E. coli. 2 This is due to the accumulation of AMR determinants, including acquired AMR genes and chromosomal point mutations. 3nolones are antibiotics that are widely used for treatment of infections caused by a variety of bacteria, including E. coli. 4 Quinolone resistance in E. coli is mainly attributed to mutations in the quinolone resistance-determining regions (QRDRs) of the topoisomerase genes gyrA and parC. 5,6Although these chromosomal mutations are generally not horizontally transferred, there are quinolone resistance determinants that can be transmitted between bacteria, called transferable mechanisms of quinolone resistance (TMQR). 7These resistance determinants include target protection (qnr), antibiotic efflux (mainly qepA and oqxAB), and antibiotic modification (aac(6′)-Ib-cr and crpP).Although these genes are usually called plasmidmediated quinolone resistance (PMQR) genes, the term PMQR is not always correct because these transferable genes can sometimes be integrated into the chromosome (hence we use the term TMQR, not PMQR, throughout this paper). 8It is known that the levels of resistance conferred by TMQR are relatively low; however, TMQR can play an important role in the development of quinolone resistance because (i) the effect of TMQR on quinolone minimum inhibitory concentrations (MICs) is additive to that conferred by other transferable or chromosome-mediated quinolone resistance determinants, and (ii) TMQR can facilitate the selection of mutants with higher-level quinolone resistance. 7,9ormation on the vectors and genetic contexts of AMR genes is important for multiple reasons: for example, (i) it allows identification of frequently associated/linked resistance genes and thus can provide information on co-resistance; (ii) it facilitates better understanding of their evolution and how resistance genes spread. 10,11Although some previous studies performed in-depth analysis of mobile elements carrying TMQR determinants in clinical E. coli, limited information exists with respect to the vectors and genetic contexts of TMQR determinants for environmental E. coli. 7,8Here, we analyzed the complete genomes of environmental E. coli isolates, sequenced in this study or obtained from the NCBI reference sequence (RefSeq) database, with the aim to acquire genomic insights into TMQR, including their vectors and genetic contexts.

E. coli isolates
Twenty-four E. coli isolates obtained from surface water and wastewater in Japan were characterized in this study.All these isolates were confirmed to carry TMQR determinants by molecular methods such as PCR.These 24 isolates included 19 isolates from our previous collections: (i) three isolates from river water samples collected in 2010, 12 (ii) 10 isolates from river water samples collected between 2011-2013, 13 and (iii) six extended-spectrum βlactamase (ESBL)-producing E. coli isolates obtained from municipal wastewater and hospital wastewater in 2015. 14Five isolates were obtained in the present study: three isolates were screened from E. coli isolated from municipal wastewater using CHROMagar ECC (Kanto Chemical Co., Tokyo, Japan) supplemented with cefotaxime (1 mg/L) between 2019-2020, and two isolates were screened from E. coli isolated from river water using CHROMagar ECC supplemented with ciprofloxacin (0.03 mg/L) between 2021-2022.Antimicrobial susceptibility testing of the 24 isolates was performed by microdilution using Dry Plate Eiken (Eiken, Tokyo, Japan) according to CLSI specifications (see Table S1 for antimicrobials used).
The MIC for ciprofloxacin was also determined using the Etest (bioMérieux, Marcy-l'Étoile, France).The results were evaluated according to CLSI criteria 15 and EUCAST epidemiological cutoff (ECOFF) values (https://mic.eucast.org/search/).All 24 isolates were subjected to short-and long-read sequencing in the present study as described below (note that short reads were obtained for some of the isolates in our previous studies but we obtained short reads again in this work because the short reads in our previous studies had low/uneven sequencing depth).

Genome sequencing and assembly
DNA was extracted from each isolate using a DNeasy PowerSoil Pro Kit (Qiagen, Hilden, Germany).Short-read sequencing libraries were prepared using an Illumina DNA PCR-Free Prep kit (Illumina, San Diego, CA).The libraries were sequenced on a NovaSeq 6000 platform (Illumina), generating paired-end reads of 151 bp each.Long-read sequencing libraries were prepared using the SQK-LSK109 kit (Oxford Nanopore Technologies, Oxford, UK) and sequenced on the MinION with FLO-MIN106 flow cells.Short reads were trimmed using fastp (v0.23.2), 16 and long reads were filtered using Filtlong (v0.2.0, https://github.com/rrwick/Filtlong).Hybrid assembly was performed with a long-readfirst approach, i.e., long-read assembly using Flye (v2.9.1-b1780) 17 followed by long-read polishing with Medaka (v1.8.0, https://github.com/nanoporetech/medaka)and short-read polishing with Polypolish (v0.5.0). 18Hybrid assembly was also performed with a short-readfirst approach using Unicycler (v0.5.0). 19If the long-read-first approach generated a complete genome, the polished Flye assembly was chosen as a final assembly.We note that long-readfirst assemblies sometimes lack small plasmids, so we recovered small plasmids that were not present in the Flye assembly but were present in the Unicycler assembly. 20If the long-readfirst approach did not generate a complete genome but the short-read-first approach did, the Unicycler assembly was chosen as a final assembly.
If the above approaches failed to generate a complete genome, we performed long-read sequencing again using DNA prepared from the same bacterial pellet but with a different extraction method, which employs enzymatic digestion followed by proteinase K digestion and AMPure XP (Beckman Coulter, Inc. Brea, CA, USA) bead purification, aiming to reduce DNA shearing and to obtain longer reads.The obtained long reads with less fragmentation were subjected to hybrid assembly as described above.

Retrieving complete E. coli genomes with TMQR determinants from the NCBI database
RefSeq E. coli genomes with an assembly level of "complete genome" (n = 2815) were downloaded using the ncbi-genome-download tool (v0.

Data availability
The complete genomes and sequence reads obtained in the present study have been deposited in GenBank and the NCBI SRA under BioProject PRJNA1078256 (also see Table 1 for the nucleotide accession numbers of replicons with TMQR determinants).Table S1 also summarizes the nucleotide accession numbers of closed replicons for completed genomes and SRA accession numbers for the remaining genomes.

Basic characteristics of E. coli isolates sequenced in this study
Twenty-four E. coli isolates with TMQR determinants were sequenced in the present study.
Antimicrobial susceptibility testing revealed that the isolates were resistant to a wide variety of antibiotics: nonsusceptibility rates ranged from 0% (piperacillin-tazobactam, ceftolozanetazobactam, cefmetazole, imipenem, meropenem, and amikacin) to 70.8% (ampicillin and piperacillin) (Table S1).Nonsusceptibility rates for quinolones were 29.2% for nalidixic acid, 37.5% for ciprofloxacin (evaluated by Etest), and 37.5% for levofloxacin.All isolates were non-susceptible to nalidixic acid, ciprofloxacin, and levofloxacin when they carried a QRDR mutation(s).Isolates with TMQR determinants but without QRDR mutations were susceptible to nalidixic acid but either susceptible or non-susceptible to ciprofloxacin and levofloxacin.This is consistent with previous studies reporting that isolates with TMQR determinants but without QRDR mutations sometimes show an unusual phenotype of nalidixic acid susceptibility and ciprofloxacin resistance. 7We also note that isolates with TMQR determinants but without QRDR mutations were all classified as non-wild type to ciprofloxacin and levofloxacin according to the ECOFF criteria, indicating that interpreting results with ECOFF values can detect TMQR determinants when isolates do not carry QRDR mutations.
Of the 24 isolates, the genomes of 20 isolates could be completed, and these 20 completed genomes were further analyzed to elucidate the genetic contexts of TMQR determinants.
This structure was previously reported in the class 1 integrons In36 and In37, 27 though a structure (intI1-bla VIM-1 -aac(6')-II-dfrA1-ΔaadA-smr-ISPa21) seen in another class 1 integron, In4873, was detected in the upstream region in p009_A. 28The association of In4873, which carries a bla VIM metallo-β-lactamase gene, and the qnrA-containing region in In36/In37 is worrisome from a drug-resistance perspective but seems to be rare: blastn analysis found only one closely related structure in plasmid pME-1a (CP041734.1)from a clinical Enterobacter hormaechei isolate, which carried bla VIM-4 instead of bla VIM-1 .
Two isolates sequenced to closure in the present study carried qnrB7.qnrB7 was previously reported to be associated with the ESBL gene bla SHV-12 . 26However, reports of complete plasmid (or chromosome) sequences carrying qnrB7 are scarce; blastn analysis identified only four complete plasmid sequences carrying qnrB7.One of them, the IncX3 plasmid pLHST2018_IncX3 harbored by a clinical Salmonella enterica isolate from India, carried both qnrB7 and bla SHV-12 . 29In our isolates, qnrB7 was also carried by IncX3 plasmids and associated with bla SHV-12 (Figure 2b).Various insertion sequences, including IS26 and IS3000, and remnants of transposons were found around qnrB7 and bla SHV-12 , which seem to have contributed to the mobilization of these resistance genes.Genetic contexts of qnrB7 and bla SHV- 12 were almost identical in pKTa005_4 and pKFu015_3 except that a 1.9 kbp fragment of IS3000 was inserted next to another IS3000 in pKTa005_4.The same genetic context was identified in pLHST2018_IncX3 but was interrupted by IS26 in deoR.
Two genomes, one sequenced by us and the other from RefSeq, carried qnrB19.qnrB19 is usually mobilized by transposition units (TPU) mediated by ISEcp1, such as Tn2012 and Tn5387, or small ColE1-like plasmids containing fragments of these transposons. 30The most common qnrB19-carrying ColE1-like plasmid is a 2699 bp plasmid named pPAB19-1/pSGI15/pECY6-7/pMK100, with others being minor variants of this plasmid. 30The two qnrB19-positive genomes carried this gene on small plasmids, p20A24_1 (3180 bp) and p32-4_E (4417 bp), both with a ColE-like replicon Col(pHAD28). 31These plasmids contained putative ColE1-like ori (RNAI, RNAII, and oriV), oriT, and Xer sites (sites used for converting plasmid multimers into monomers) (Figure 2c).p20A24_1 was found to be closely related to plasmid pPAB19-3 (JN985534) from a clinical E. coli isolate, with some minor differences (e.g., the Tn5387 fragment was ~240bp longer in p20A24_1).p32-4_E was identified as a cointegrate plasmid comprising the common qnrB19-carrying plasmid pPAB19-1 (GQ412195) and a 1718 bp plasmid almost identical (single nucleotide difference) to pMB9272_7 (CP103528), with putative oriT regions being the cointegration points.pMB9272_7 carries only two genes encoding hypothetical proteins, and no plasmid replicon was identified by PlasmidFinder.Cointegrates comprising ColE1-like plasmids and large plasmids were previously shown to play an important role in the evolution of AMR. 31 The example of p32-4_E indicates cointegrates of ColE1-like plasmids and other small plasmids may also contribute to the spread of AMR.
Three genomes sequenced to closure in the present study carried qnrD1.qnrD1 was first described on a 4270 bp plasmid, p2007057, in a clinical Salmonella enterica strain isolated in China. 32qnrD1 has since been detected on multiple plasmids; however, blastn analysis identified that qnrD1 is rare in Enterobacteriaceae and mostly detected on ~2.7 kbp plasmids in Morganellaceae.Two of the three genomes carried qnrD1 on a plasmid identical to p2007057 (Figure 3).This plasmid contains four ORFs, one of which is mob, in addition to qnrD1.The remaining genome carried qnrD1 on a 6657 bp plasmid, named pKFu015_4, with no close blastn hits (all hits were below 50% query coverage).The qnrD1 region, including qnrD1 and ORF2, in pKFu015_4 was identical to ~2.7 kbp Morganellaceae plasmids such as pRS12-11 (KF364953) and was ~95% identical to the corresponding region in p2007057 (and thus pKMi029_6 and pKTa005_7).This plasmid also carried regions corresponding to the ori, bom (basis of mobility), rom (RNA one inhibition modulator), mob, and cer (ColE1 resolution sequence) regions in plasmid ColE1. 33Analysis of sequences surrounding the qnrD1 region revealed that there were 209 bp directly oriented repeats upstream and downstream of the region.This repeat carried imperfect 33 bp inverted repeats (IRs) closely related to some Tn3family transposons (e.g., TnEc4), though it did not encode a transposase.The blastn analysis revealed that this repeat is identical to a Tn3-derived inverted-repeat miniature element (TIME), named TIME IS101 . 34Pairs of TIMEs are known to be involved in mobilization of intervening sequences when the corresponding Tn3 family transposases are provided in trans, and these composite transposon-like structures were previously named TIME-COMP. 34,35Thus, the TIME-COMP structure, TIME IS101 -qnrD1-ORF2-TIME IS101 , seems to have been mobilized as a single unit into a ColE1-like plasmid, generating pKFu015_4.The presence of 5 bp direct repeats (AGCTA) surrounding this structure supports this idea.Moreover, we also found a ColE1-like plasmid without insertion of this TIME-COMP structure among the genomes sequenced in this study (pKMi029_5, CP147114), further supporting this.qnrD1 was previously suggested to be mobilized as a mobile insertion cassette (mic) element (explained below), 36 but this observation indicates another mode of qnrD1 mobilization, namely TIME-COMP.
qnrS1 was the most prevalent TMQR gene both in our isolates (n = 12) and in RefSeq genomes (n = 21), indicating the high prevalence of this gene in environmental E. coli.qnrS1 was detected on plasmids of diverse replicon types, and on the chromosome in two genomes (Table 1 and Table 2).Analysis of the flanking sequences revealed that the genetic contexts of qnrS1 can largely be classified into 13 types (type A to type M in Figure 4).Types B, E, and I are the contexts found only in the genomes analyzed in this study and no other hits were found by blastn analysis.Type A is the context detected in the original qnrS1-harboring plasmid pAH0376 from a clinical Shigella flexneri isolate and was also detected in five isolates analyzed in this study (note that the available pAH0376 sequence ends at tnpR and the presence of ISKpn19 in the plasmid is unknown). 37Insertion of a TPU (ISEcp1-bla CTX-M-15 -Δorf477) into tnpA of Tn2 in type A, which is a context prevalent in genomes deposited in GenBank (e.g.,

CP052151
) but not detected in this study, seems to have been followed by Two genomes sequenced to closure in this study and two RefSeq genomes carried qnrS2.qnrS2 was previously reported to be part of a mic element bracketed by 22 bp imperfect IRs in the Aeromonas punctata plasmid p37. 40A mic element is a nonautonomous element bracketed by two (imperfect) IRs.A mic does not carry a transposase gene but can be mobilized in trans by a transposase that can recognize the IR.In two RefSeq genomes, qnrS2 was embedded within the same genetic context on IncX1 plasmids, but the mic element was truncated by IS26 (Figure 5).The IS26-qnrS2-ΔmpR-IS26 structure, which was also detected in E. coli plasmids such as pRW7-1_235k_tetX (MT219825), was followed by the fosA3 genetic context type K, 41 linking qnrS2 to fosA3, Δbla TEM , and bla CTX-M-55 .Two ST2179 genomes sequenced to closure in this study carried qnrS2 on the chromosome.In both genomes, qnrS2 was within the truncated mic element and was surrounded by directly-oriented IS26 sequences.This IS26-qnrS2-ΔmpR-IS26 structure (note that this structure is different from that in the IncX1 plasmids, in that the IS26 sequences are directly-oriented and a ~200 bp fragment is present between the truncated mic and the left IS26) was also detected in some Enterobacteriaceae plasmids such as pSH16G4525 (MH522424).Interestingly, this qnrS2 region was duplicated in 19M19.In both ST2179 genomes, the qnrS2 region was followed by E. coli chromosomal genes, but the gene contents were different, likely due to the rearrangement caused by intramolecular replicative transposition of IS26. 35
aac(6′)-Ib-cr has been identified in a gene cassette as part of class 1 integrons. 42Although aac(6′)-Ib-cr is sometimes linked to the 5′-conserved segment (5′-CS) and the 3′-conserved segment (3′-CS) in class 1 integrons, integrons containing aac(6′)-Ib-cr are often interrupted by IS26. 8,43Our analysis supports this observation, having identified deletion/truncation of either or both of 5′-CS and 3′-CS by IS26 in 11 of 12 cases (Figure 6).Apparently, the truncated structures seen in the 11 genomes were derived from the class 1 integron intI1aac(6′)-Ib-cr-bla OXA-1 -catB3-arr-3-qacEΔ1-sul1.Among them, the IS26-aac(6′)-Ib-cr-bla OXA- 1 -ΔcatB3-IS26 structure was prevalent and detected in seven genomes (58%), while the upstream and downstream sequences were divergent.The blastn analysis revealed that this structure is prevalent among the public genomes (more than 1000 hits with >99% coverage and with >99% identity).It should be noted, however, that the IS26 sequences are inversely oriented and thus this structure is neither a transposon nor a pseudo compound transposon. 44 prevalence of this structure among various plasmids/chromosomes can partially be explained by the presence of other mobile genetic elements surrounding it, which might have aided the movement of multiresistance regions containing this structure.

Genetic contexts of oqxAB
Mobile oqxAB genes are commonly found within the IS26-oqxA-oqxB-oqxR-IS26 structure, named Tn6010. 45All four oqxAB genes identified in the present study were situated within Tn6010 (or a truncated Tn6010) and carried by plasmids of different Inc types (Figure 7).
While the downstream sequences flanking Tn6010 were divergent in the four plasmids, three RefSeq plasmids carried common sequences upstream of Tn6010, containing tnpA of Tn2, IS26, bleO, nimC/nimA, and ΔISEnca1.The blastn analysis using this upstream structure + Tn6010 as a query sequence identified more than 100 genomes carrying this structure.On the other hand, the upstream structure of Tn6010 was replaced by a truncated Tn1721 and a truncated Tn2 in pJKHS004_1, linking tetA and a truncated bla TEM to oqxAB.No genomes were found by blastn analysis to carry the multiresistance region detected in pJKHS004_1.

Genetic context of qepA1
qepA genes are typically associated with ISCR3 and embedded within complex integrons. 46 isolate sequenced in this study carried qepA1, and the gene was located between a truncated intI1 and ISCR3 on an incF plasmid (C4:F18:A-:B1), pKMi012_1 (Figure 8).This region was followed by a ΔintI1/groEL-dfrB4-qacEΔ1-sul1-Δorf5 region, the chromate resistance (chrA) region, and the macrolide resistance (mph(A)) region.The blastn analysis of the entire resistance region bounded by IS26 revealed almost identical sequences (100% query coverage and >99% identity) in three E. coli plasmids, namely pJJ1887-5 (CP014320.1)from a clinical isolate in the USA, pM216_mF (LC492469.1)from a clinical isolate in Myanmar, and pMyNCGM604 (LC744503.1)from an isolate of unknown origin in Myanmar, indicating a wide distribution of this multiresistance region.

Study limitations
This study has some limitations.The genomes of four isolates (KMi014, KTa007, KTa009, JKHS019) could not be completed despite repeated long-read sequencing.This is probably due to the presence of long repeats in these genomes.We also included RefSeq genomes in our analysis to extend our dataset, but most RefSeq genomes were from Japan and Switzerland.
This seems to reflect the situation where long-read sequencing is still not widely used to study antibiotic resistance in environmental bacteria.

CONCLUSIONS
Here, we performed in-depth analysis of the genetic contexts of TMQR determinants in environmental E. coli.The genetic contexts described in this study included those closely related to the contexts previously reported in clinical isolates.We also detected novel genetic contexts, such as qnrD1 embedded within TIME-COMP.Some of the TMQR genes were frequently associated with other AMR genes in the described contexts (e.g., qnrB7 and bla SHV- aac(6′)-Ib-cr) are located not only on plasmids but also on the chromosome, encouraging the use of the term TMQR rather than PMQR to avoid erroneous interpretations in future studies.
Overall, this study provides valuable insights into mobile genetic elements associated with TMQR determinants and highlights the importance of genomic surveillance of antimicrobial resistant bacteria in the environment.a Detailed information on each isolate, such as other AMR genes detected, is available in Table S1.
b The context type is shown for qnrS1.
c GyrA codons 83 and 87 and ParC codons 80 and 84 are shown in this order.SDSE is the wildtype.The number of amino acid substitutions is shown in parentheses.
d The genomes of these four isolates could not be completed.
a Detailed information on each isolate, such as other AMR genes detected, is available in Table S2.
b The context type is shown for qnrS1.nucleotide identity with ISKpn19 in the type G context, but this region is not shaded).Tn6292 and its derivatives (Tn6360 and Tn6361) are shown for the purpose of comparison.See Table 1 for information on genomes carrying each context type.
(i) truncation of tnpR by IS26 (type B), (ii) transposition of a longer TPU, including the original TPU (ISEcp1-bla CTX- M-15 -Δorf477) and the qnrS1 region, into the chromosome by ISEcp1 (type C), or (iii) partial deletion of Tn2 and ISEcp1 and IS26-mediated truncation of ISKpn19 (type D).Truncation of type A by two IS26 elements might have led to generation of type E. ISKpn19 elements are present both upstream and downstream of qnrS1 in type F. The blastn analysis identified the type F context in only two replicons, E. coli plasmid pA1.S1.126.c2(CP146587) from humanurine and E. coli plasmid pKT58A (JX065631) from a wild bird in Slovakia.38In type G, ISEcl2 is truncated by IS26, and class 1 integrons with different cassette arrays are present upstream of qnrS1.The type H context contained the ftsI-bla LAP-2 -ISEcl2 structure upstream of qnrS1, which was reported previously in clinical Klebsiella pneumoniae and was also detected in multiple Enterobacteriaceae plasmids by blastn analysis.39Truncation of type H by IS26 at bla LAP-2 seems to have given rise to type I. Type J and type K contexts are inserted in the same position in each IncX1 plasmid and flanked by 5-bp direct repeats of ATAAC (or the reverse complement sequence GTTAT).Insertion of Tn2 in the ancestral IncX1 plasmid followed by IS26-mediated rearrangements, including insertion of the region containing qnrS1, can explain these contexts.Type L and Type M seem to have been derived from the qnrS1-bearing transposon Tn6292.Tn6292 itself was not detected in the present study, though blastn analysis identified that intact Tn6292 is almost restricted to Enterobacteriaceae plasmids from China.

Figure 3 .
Figure 3. Structures of plasmids carrying qnrD1.Genes and elements are shown as in earlier

Figure 4 .
Figure 4. Genetic contexts of qnrS1.Genes and elements are shown as in earlier figures.The

Figure 5 .
Figure 5. Genetic contexts of qnrS2.Genes and elements are shown as in earlier figures.

Figure 7 .
Figure 7. Genetic contexts of oqxAB.Genes and elements are shown as in earlier figures.The

Figure 8 .
Figure 8.The genetic context of qepA1.Genes and elements are shown as in earlier figures.
3.1, https://github.com/kblin/ncbigenome-download) in July 2023.Metadata information, including strain name and isolation source, was extracted from the downloaded files, and genomes determined to be of

TABLES 573 Table 1 .
Characteristics of 24 E. coli isolates sequenced in this study a 574

Table 2 .
Characteristics of 33 RefSeq E. coli genomes with TMQR determinants a 583