Genomic Surveillance of Ceftriaxone-Resistant Escherichia coli in Western New York Suggests the Extended-Spectrum β-Lactamase blaCTX-M-27 Is Emerging on Distinct Plasmids in ST38

Extended-spectrum β-lactamase (ESBL)-producing Enterobacteriaceae pose significant treatment and infection prevention challenges. Escherichia coli sequence type (ST) 131 associated with the blaCTX-M-15 gene has been the dominant lineage of ESBL-producing E. coli in the US and worldwide. In this study, our objective was to determine the β-lactamase profile, means of dissemination, prevalence, and the clonal identity of ESBL-producing E. coli in our region of Western New York. Whole-genome SNP-based phylogenomics was used to assess 89 ceftriaxone-resistant (CTR) E. coli. Isolates were collected from both inpatients and outpatients and from urine and sterile-sites over a 2 month period in 2017 or throughout the year, respectively. ST131 was the predominant ST (46.0%), followed by ST38 (15.7%). The blaCTX-M-15 gene was commonly found in 53.7% of ST131 isolates, whereas the blaCTX-M-27 gene was found in 26.8% of ST131, though was significantly associated with ST38, and was found in 71.4% of those strains. When compared to ST131, ST38 E. coli exhibited increased frequency of resistance to nitrofurantoin and decreased frequency of resistance to ciprofloxacin and ampicillin-sulbactam. Using Nanopore long-read sequencing technology, an analysis of the ESBL genetic context showed that the blaCTX-M-15 gene was chromosomal in 68.2% of ST131, whereas the blaCTX-M-27 gene was plasmid-borne in all ST131 and 90% of ST38 isolates. Notably, the blaCTX-M-27 gene in ST38 resided on highly-related (99.0–100.0% identity and 65.0–98.0% query coverage) conjugative IncF plasmids of distinct plasmid multi-locus sequence types (pMLSTs) from those in ST131. Furthermore, ST131 and ST38 were found to harbor different antibiotic resistance gene and virulence factor profiles. These findings raise the possibility of an emerging ESBL-producing E. coli lineage in our region.


INTRODUCTION
Extended-spectrum β-lactamase (ESBL)-producing organisms are responsible for ~26,000 drug-resistant infections and ~1,700 deaths per year in the US, where they are categorized as a serious and increasing threat within the Centers for Disease Control and Prevention's (CDC) 2019 Antibiotic Resistance Threat Report (CDC, 2019). Among hospitalized patients, ESBL-producers may account for up to 11.6 and 16.1% of Escherichia coli causing urinary tract infection (UTI) and bloodstream infections (BSIs), respectively (Mendes et al., 2019). At present, CTX-M β-lactamases are the prevailing family of ESBLs and include more than 150 genes (Zhao and Hu, 2013). They may have originated as chromosomally-encoded enzymes in Kluyvera spp. before spreading to Escherichia, Klebsiella, and other enteric bacteria (Rossolini et al., 2008). Documented mechanisms of mobilization include capture by the insertion elements ISEcp1 and ISCR1, as well as bacteriophages (Poirel et al., 2008). While chromosomal integration is reported, CTX-M β-lactamase genes are more frequently associated with IncF plasmids (Doumith et al., 2012;Stoesser et al., 2016).
The genetic context of bla CTX-M-15 and other CTX-M family genes in ESBL-producing E. coli in the US remains relatively undefined. In this study, we used bacterial whole-genome sequencing (WGS) to investigate the genomic epidemiology of 89 ceftriaxone-resistant (CTR) E. coli with respect to clonality, susceptibility, multi-drug resistance (MDR), and β-lactamase profiles. One of our goals was to compare the prevailing clonal types and ARGs between urine and sterile-site isolates from both inpatients and outpatients. Plasmid CTX-M gene context was further examined for all isolates using long-read sequencing. Complete plasmid sequences and chromosomal integration sites were mapped. Thus, this study represents a detailed snapshot of the genomic landscape, including the apparatus of horizontal transmission, of ESBL-producing E. coli isolated in our region of Western New York (NY).

Clinical Laboratory Setting and Isolate Selection
This study was performed under University of Rochester Medical Center (URMC) IRB protocol RSRB00068143. Eighty-nine CTR E. coli isolates collected as part of routine clinical care at the URMC Clinical Microbiology laboratory in Rochester, New York were selected for this project. The laboratory provides diagnostic services to a population of ~0.5 million people in Western NY and services several area hospitals, urgent cares, nursing homes, and outpatient practices. To identify potential ESBL-producing organisms, we selected unique patient isolates for WGS based on CTR. Urine CTR E. coli were collected at convenience during the months of October and November in 2017 (53 isolates), and CTR E. coli derived from sterile-site infections were collected throughout 2017 (36 isolates: 28 BSIs, 4 bone, 2 peritoneal fluid, 1 joint fluid, and 1 drain). In their respective timeframes, this captured 53/67 (79.1%) of unique patient CTR urine isolates and 36/36 (100%) CTR sterile-site isolates. Initial E. coli identification was performed with MALDI-TOF (Vitek MS v3.0; bioMérieux Inc., Durham, NC). Antibiotic susceptibility (including ESBL production) was assessed with Vitek 2 (bioMérieux Inc.; AST-GN70 test card). Phenotypic ESBL production assessed by Vitek2 is indicated by "+" (Data S1 -Antibiogram). Cefazolin susceptibility was determined by Kirby-Bauer disk diffusion for isolates from sterile-sites and is indicated with "/KB" in Data S1 -Antibiogram. Kirby-Bauer zone of inhibition diameter is indicated in millimeters. Susceptibility was interpreted with CLSI standard M100 (CLSI, 2019).

Bacterial Growth Conditions and Genomic DNA Extraction
E. coli isolates were archived at −80°C in trypticase soy broth (TSB) with 20% glycerol and maintained at 35°C on blood agar (BD BBL trypticase soy agar with 5% Sheep Blood; BD). Bacterial DNA was extracted with the MagNA Pure Compact System (Roche, Indianapolis, IN). DNA was quantified with the QuantiFluor dsDNA system (Promega, Madison, WI).
Genomic DNA Sequencing DNA library preparation was performed according to the manufacturer's protocol (Nextera XT DNA Library Preparation Kit; Illumina, San Diego, CA), purified using Agencourt AMPure XP beads (Beckman Coulter Inc., Indianapolis, IN), and quality-checked using the Agilent 4,200 TapeStation System (Agilent; Santa Clara, CA). Purified PCR products were normalized using Nextera XT Library Normalization Beads (Illumina). Normalized samples were pooled and quantified using the Qubit ssDNA Assay kit (Invitrogen). Library pools were loaded with 2.45 ng ssDNA and 20 μl PhiX control DNA (20 pM). Paired-end sequencing was performed with MiSeq Reagent v3 600-cycle kits on the MiSeq instrument (Illumina).

Plasmid Purification and Long-Read Sequencing
Plasmids were sequenced on the MinION platform (Oxford Nanopore Technologies; Cambridge, MA). Briefly, for each of the 89 strains, 100 ml of Luria-Bertani (LB) broth was inoculated and incubated for ~18 h at 37°C, and plasmids were purified using the QIAfilter Plasmid Midi Kit (Qiagen, Germantown, MD). Plasmid DNA was purified (Agencourt AMPure XP beads), and quantified (QuantiFluor dsDNA), and adjusted to 400 ng in a total volume of 7.5 μl molecular biology-grade H 2 O. Sequencing library preparation (Rapid Barcoding Sequencing kit; Oxford Nanopore Technologies) was performed as indicated with the following alterations: Fragmentation Mix RB01-12 volume was reduced to 1.5 μl and was incubated for 20 s at 30°C. Base-calling and de-multiplexing was performed using Albacore v2.3.1 (Oxford Nanopore Technologies) and Illumina-Nanopore hybrid read assemblies were generated by Unicycler (Wick et al., 2017). Plasmids were denoted as "circular" (complete) or "uncircularized" (i.e., incomplete or fragments). BLASTn was used to identify plasmids, plasmid multi-locus sequence typing (pMLST), and virulence and antimicrobial resistance genes (ARGs), which were also assessed with Center for Genomic Epidemiology (CGE) tools (including ResFinder and FimTyper; Thomsen et al., 2016) and with ABRicate (Seemann, 2017). Plasmids were annotated using RAST (Aziz et al., 2008) and aligned with Mauve (Darling et al., 2004). In some instances, genomic sequences recovered from long-read plasmid sequencing were used to create hybrid assemblies with Illumina reads.
Statistical calculations were performed in Prism 8 (GraphPad Software, San Diego, CA). Multiple t-tests (Holm-Sidak method) were used for comparison of AMR frequencies. Otherwise, Fisher's exact test was used, including for gene content comparisons between ST131 and ST38. Unless otherwise indicated, data shown represent the mean ± SEM.

Genetic Context of the ESBL bla CTX-M-15 in ST131 and ST38 E. coli
Among the 30 isolates with the bla CTX-M-15 gene, 15/22 ST131 isolates had the gene integrated into the chromosome. The bla CTX-M-15 gene was chromosomal in 4/8 isolates of other STs, one of which was ST38 (Figure 2A and Data S1 -Genomes). For most chromosomal insertions, alignment (with 100% identity) of the immediate bla CTX-M-15 context revealed the presence of the well-known ISEcp1-flanked bla CTX-M-15 -orf477 arrangement (Rodríguez et al., 2004). The bla CTX-M-15 gene insertions were also typically flanked by other IS element insertions and scars ( Figure 2C). Among the ST131 isolates with chromosomal bla CTX-M-15 , the gene was frequently located in one of two sites, either: (1) between genes encoding shikimate kinase (aro) and pyrroline-5-carboxylate-reductase (proC; 6/15 isolates); or (2) inserted adjacent to a molybdate metabolism regulator gene (molR; 4/15 isolates). In the remaining 5/15 ST131 isolates with chromosomal integration, the bla CTX-M-15 gene was inserted in unique sites. In two instances, long-read hybrid assembly using chromosomal reads obtained during plasmid sequencing also identified one additional chromosomal copy of the bla CTX-M-15 gene inserted into unique locations (in URMC_7 and URMC_59; Figure 2C).
In 6/11 isolates, the plasmid-encoded bla CTX-M-15 genes were carried on IncF-type plasmids of a variety of pMLSTs ( Figure 3B). Two additional plasmids (URMC_62_p_96678 and URMC_112_p_99275) were typed as IncY (Data S1, Plasmids). In general, bla CTX-M-15 plasmids shared little synteny with each other or with bla CTX-M-27 plasmids (Figures 3A,B). The plasmids varied in their carriage and arrangement of ARGs.

DISCUSSION
Our data indicated that the bla CTX-M-15 gene was the predominant ESBL in our region, but also that the bla CTX-M-27 gene constituted a large minority, being highly represented in ST131 (26.8%) and in ST38 (64.2%) isolates. Prior studies conducted in the US have indicated that the bla CTX-M-15 gene is the predominant ESBL, and is frequently carried by ST131, the most widely established extraintestinal clone (Johnson et al., 2012). In our clade C2 (H30-Rx) ST131 isolates, two integration sites accounted for 10/15 chromosomally integrated bla CTX-M-15 genes. Both of these integration sites have been reported for ESBL E. coli collected in other studies (i.e., Genbank: NZ_CP018979 and NZ_CP018991.1). Indicating that these groups did not represent recent local clonal transmission, they were separated by >50 SNPs. The bla CTX-M-15 gene was carried on a diverse group of plasmids in the clade C2 group, all with unique pMLSTs (Figure 3B).
Others have shown the increasing prevalence of bla CTX-M-27 in ST131 (Matsumura et al., 2015). In these studies and others, the bla CTX-M-27 gene has been associated with fimH30 and fluoroquinolone resistance as part of clade C1-M27. Here, we found the same, as well as noting that the bla CTX-M-27 gene was often embedded in plasmids of pMLST IncF[F1:A2:B20] which was almost exclusively restricted to the C1-M27 clade, as reported by others (Ghosh et al., 2017;Kondratyeva et al., 2020). Interestingly, An IncF[F1:A2:B20] plasmid was also found in a single ST38 strain (URMC_96; Figure 3C).
Among ST38 strains in our study set, the bla CTX-M-27 gene was borne on plasmids with few close homologs in Genbank, and which were distinct from those in ST131. These were among the most novel of all the plasmids described in this study. The closest homologue was the IncF plasmid p7_2.1 (Genbank: CP023821), which shared >99% identity over a query coverage range of 33-61% with the ST38 plasmids that carried the bla CTX-M-27 gene (Data S1 -Plasmids). The p7_2.1 plasmid was identified in a Swedish study of stool isolates but does not harbor the bla CTX-M-27 gene. Another ST38 bla CTX-M-27 -carrying plasmid (URMC_96_p153061) matched closely (99% identity and 80% query coverage) to plasmid p4_4.1 from the same study (GenBank: CP023827.1), but also had 100% identity (34% query coverage) to pDA33137-178 from a ST44 isolate (Nicoloff et al., 2019). In our study, IncF plasmids in ST38 were commonly (8/9) IncF[F2:A-:B10], where the bla CTX-M-27 gene was closely associated with IS903B and IS26. The latter has been reported to help drive the dissemination of some CTX-M β-lactamase genes (e.g., bla CTX-M-14 ; Zhao and Hu, 2013). FAB pMLST assessment of ST38 plasmids has not been widely done, though IncF[F2:A-:B10] has been shown in one study to account for 1/12 isolates of a collection of ST38, where IncF[F1:A-:B33] was more commonly observed (4/12; Shaik et al., 2017).
Published genomic comparisons of ST131 with ST38 suggest that the latter harbors fewer UPEC-associated virulence genes, though the two have similar in vitro adhesion, invasion, and serum resistance phenotypes (Shaik et al., 2017). Elsewhere, ST38 has been described as an ExPEC or a UPEC/EAEC hybrid (Chattaway et al., 2014;Phan et al., 2015;Muenzner et al., 2016). Others have suggested that EAEC attributes may increase the potential of such strains to cause UTIs (Boll et al., 2013). The ST38 isolates identified in this study did not harbor genes encoding for aggregative adherence fimbriae (AAF) nor for AggR, the transcription factor that regulates AAF biogenesis (Boll et al., 2012). While profiling virulence factors in silico is limited by the quality and quantity of available databases, the ST38 isolates in this study did harbor some putative virulence factors that may be associated with aggregation and dispersion in EAEC. For example, ST38 harbored afaF-III (Afa/Dr. adhesin family; Muenzner et al., 2016). The pathogenicity and clinical pertinence of E. coli expressing Afa/Dr. adhesins in UTIs are well established (Servin, 2014).
In this study, ST131 exhibited increased frequency of resistance to fluoroquinolones and ampicillin-sulbactam compared to ST38, which was often non-susceptible to nitrofurantoin. Increased frequency of fluoroquinolone resistance in ST131 vs. ST38 has been previously described (Alghoribi et al., 2015;Gauthier et al., 2018;Guiral et al., 2019) and is a hallmark of clade C strains. Only 3/41 ST131 isolates in this study were susceptible (S or I) to fluoroquinolones, all of which were clade A. To the best of our knowledge, the increased Frontiers in Microbiology | www.frontiersin.org frequency of nitrofurantoin resistance in ST38 has not been reported. Nonsense mutations in the nitroreductase genes nfsA and nfsB are associated with nitrofurantoin resistance and were found in all nonsusceptible strains of both ST38 and ST131 (Sandegren et al., 2008;Shanmugam et al., 2016).
If ST38 is emerging as a prominent ESBL lineage, then concurrent resistance to nitrofurantoin is concerning because this drug has thus far remained useful for fluoroquinolone-resistant and ESBL-producing organisms (Hertz et al., 2016;Tulara, 2018). Vice-versa, the emergence of this lineage may be influenced by the reduction of fluoroquinolone use. With respect to the observed differences for ampicillin-sulbactam, others have noted that bla CTX-M-15 genes (or isolates harboring these determinants) are associated with increased frequency of resistance compared to isolates carrying the bla CTX-M-27 gene (Faheem et al., 2013;Matsumura et al., 2015). This is consistent with our results given the respective preponderance of these enzymes in ST131 vs. ST38. This observation may also be related to the presence of other β-lactamases, as isolates of both ST131 and ST38 were more often resistant to ampicillin-sulbactam if they also harbored broad spectrum β-lactamase genes such as bla TEM and bla OXA .
Limitations of this study include the regional nature and narrow timeframe of isolate collection. Furthermore, while our sequenced E. coli may have similar phenotypic susceptibility to both current and past isolates, they may not be representative of the ARGs, STs, and mobile genetic elements found in the community across time. The findings in this study raise several questions. For example, is there a fitness advantage for isolates carrying the bla CTX-M-27 gene? This gene may confer greater resistance to ceftazidime (Kuroda et al., 2012). What explains the almost exclusive association between ST38 and the bla CTX-M-27 gene, while ST131 is associated with both the bla CTX-M-27 and bla CTX-M-15 genes? Is the spread of bla CTX-M-27 in our region associated with human carriage from areas of high prevalence or an isolated clonal outbreak? Recent surveillance of foodproducing animals in the US showed that cattle and turkey E. coli frequently carried the bla CTX-M-27 gene (Tadesse et al., 2018). Salmonella spp. from food-producing animals have also been shown to carry the bla CTX-M-27 gene (Zhang et al., 2016), though the plasmids in our study bore little resemblance to publically available sequences from Salmonella (data not shown). Establishing a link between these observations highlights the need for more extensive and longitudinal "One Health" surveillance studies. In conclusion, although this work may serve as a window through which to view national epidemiological trends, additional surveillance is needed to confirm the emergence of ST38 and its association with the bla CTX-M-27 gene.

DATA AVAILABILITY STATEMENT
The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/ Supplementary Material.

AUTHOR CONTRIBUTIONS
HM and NP designed the study. HM and NP selected clinical isolates. HM isolated genomic DNA. JW performed sequencing. ST performed and managed bioinformatics analyses and pipelines. HM, AM, and AC analyzed sequence data. HM and AC performed statistical analyses. HM, AC, and NP analyzed overall data/results and wrote the first draft of the manuscript. NP provided funding and resources. NP, GD, and DH provided technical expertise. All authors participated in editing and reviewing the manuscript and approved the final manuscript. All authors contributed to the article and approved the submitted version.

FUNDING
Funding from the Department of Pathology and Laboratory Medicine, University of Rochester Medical Center supported this study.