Multiple β-Lactam Resistance Gene-Carrying Plasmid Harbored by Klebsiella quasipneumoniae Isolated from Urban Sewage in Japan

In our investigation of urban wastewater in Japan, carbapenem-resistant Klebsiella quasipneumoniae subsp. quasipneumoniae was isolated that carried the pTMSNI47-1 plasmid, which carries four β-lactamase genes and has transferability among Enterobacteriaceae. pTMSNI47-1 was found to encode a rarely reported carbapenemase, KHM-1. Cooperative effects of β-lactamases encoded by pTMSNI47-1 appeared to have broad-spectrum resistance to β-lactams. The detection of the KHM-1 gene in urban wastewater suggests that such a rare antimicrobial resistance (AMR) gene can be pooled in the environment, potentially emerging as an AMR determinant in a pathogen. When the number of β-lactamase resistance genes is increased in one plasmid, the transfer of this plasmid can confer broad-spectrum resistance to β-lactams, even if the individual gene confers narrow-spectrum resistance. The present study adds important information about the potential risk of sewage treatment plants as reservoirs and environmental suppliers of AMR genes, contributing to the public health from a One Health perspective.

in the environment, particularly from a One Health perspective. Nevertheless, CPE environmental contamination has not been investigated fully. Sewage treatment plants (STPs) are one of the most important interfaces between the human population and the aquatic environment. Several previous studies (3)(4)(5) have proposed STPs and wastewater to be the hot spots for horizontal gene transfer, facilitating the spread of antimicrobial resistance (AMR) genes, including carbapenemase genes, between different bacterial species. Additionally, sewerage system diffusion has improved in recent years, leading to an increased proportion of sewage effluent in the environmental water. These facts suggest that STPs and wastewater can act as anthropogenic sources, reservoirs, and environmental suppliers of AMR genes.
A large variety of carbapenemases have been reported, including those belonging to Ambler class A (e.g., KPC and IMI), class B (e.g., IMP and NDM), and class D (e.g., OXA-48 and OXA-162) (6). In clinical fields, highly carbapenem-resistant strains harboring KPC-or NDM-producing Enterobacteriaceae have been spreading rapidly between countries (7,8). In Japan, the most prevalent carbapenemase in Enterobacteriaceae is an IMP-type enzyme (9,10). On the other hand, NDM-, VIM-, KPC-, or OXA-48-producing Enterobacteriaceae have been rarely isolated from sporadic cases (e.g., patients with carbapenem-resistant infections who have travelled abroad) (8,11). Ambler class B carbapenemases are metallo-␤-lactamases (MBLs) and classified into various types according to their amino acid sequences. In general, MBLs harbor hydrolytic activity against broad-spectrum ␤-lactams except monobactams and demonstrate reduced carbapenem susceptibility (12). Kyorin Health Science MBL-1 (KHM-1) was identified in 1997 in a multidrug-resistant Citrobacter freundii isolate from a patient with a catheterassociated urinary tract infection, in Japan. The bla KHM-1 gene was carried in a plasmid of approximately 200 kbp, designated pCF243 (13). However, reappearance of this enzyme has not been identified in clinical settings or the natural environment since its first report (14). Thus, the extent of spread and whether this MBL can contribute to CPE infections are unknown.
We reported the isolation of a novel multidrug-resistant IncA/C2 plasmid, pTMSNI47-1, containing carbapenemase gene bla KHM-1 , in Klebsiella quasipneumoniae SNI47 isolated from a municipal STP in Japan. It was suggested that CPE, harboring a highly transferable and broad-spectrum resistance plasmid, had been disseminated and deposited into the sewage.
Antimicrobial susceptibilities of five ␤-lactamases. To detect which ␤-lactamases harbored hydrolytic activity against each ␤-lactam, we transformed five recombinant plasmids to DH5␣ cells and measured the MICs of 10 ␤-lactams. As Table 1 shows, both CTX-M-2-producing and DHA-1-producing strains showed increased MICs of penicillin, cephem, or monobactam derivatives. In particular, the MICs of cefotaxime (Ͼ32 g/ml), cefepime (8 g/ml), and aztreonam (8 g/ml) for the former were the largest among the five tested ␤-lactamases. The production of CTX-M-2 had a small effect on the hydrolytic activity against ceftazidime (0.5 g/ml). On the other hand, the DHA-1- producing strain showed a higher MIC increase for ceftazidime (4 g/ml) and lower MIC increases for piperacillin (32 g/ml), cefotaxime (2 g/ml), and aztreonam (1 g/ml) than those of the CTX-M-2-producing strain. Regarding the three cephem derivatives, clavulanic acid could inhibit CTX-M-2 activities but could not inhibit DHA-1. The KHM-1-producing strain showed generally increased MICs of carbapenems and cephems. The MICs of carbapenems were 0.5 g/ml for imipenem, 0.5 g/ml for meropenem, 0.25 g/ml for ertapenem, and 0.25 g/ml for doripenem, corresponding to 4-,

DISCUSSION
STPs are considered major anthropogenic sources, reservoirs, and environmental suppliers of AMR genes, including carbapenemase genes (3,4). In fact, recent studies of Japanese STPs identified GES-5, GES-24, IMP-19, and KPC-2-type CPE isolates from wastewater (11,15). Our study identified a CPE isolate, SNI47, from the influent water of an STP in Japan. The phylogenetic tree constructed using MiSeq data illustrated that SNI47 belongs to K. quasipneumoniae subsp. quasipneumoniae (Fig. 1). In clinical settings, a higher prevalence of K. pneumoniae than K. quasipneumoniae has been reported (16,17). Gomi et al. (15) reported that the prevalence among the carbapenemaseproducing Klebsiella isolates from STPs or hospital wastewater was different from those observed in clinical isolates, with K. quasipneumoniae isolated more frequently than K. pneumoniae from wastewater. Our results support their suggestion that the difference may primarily be because K. quasipneumoniae is associated more frequently with carriage, whereas K. pneumoniae is associated with human infection.
K. quasipneumoniae subsp. quasipneumoniae SNI47 harbored four plasmids: IncA/C2 plasmid pTMSNI47-1, IncFIB(K) plasmid pTMSNI47-2, IncFII(K) plasmid pTMSNI47-3, and IncR plasmid pTMSNI47-4. Of these, only pTMSNI47-1 harbored drug resistance genes (Table 3). IncA/C plasmids have demonstrated a wide Enterobacteriaceae host range and are one of the main plasmid families that mediate AMR dissemination (18). In addition, our whole-genome sequencing results indicated that the genes contributing to conjugal transfer, such as traA, were annotated in pTMSNI47-1 (Fig. 4A). It was demonstrated experimentally that pTMSNI47-1 could transfer among Enterobacteriaceae through conjugation. The MICs of all tested ␤-lactams, including carbapenems, for both transconjugant J53/pTMSNI47-1 and transformants DH5␣/pTMSNI47-1 were higher than those for the recipient strain, indicating that this plasmid conferred resistant properties for carbapenems. The broad host range and the high self-transferability of plasmid pTMSNI47-1 can lead to drug resistance acquisition in one horizontal gene transfer event.
Gene module transpositions are facilitated by transposons and ISs. These elements are frequently detected in plasmids, including pTMSNI47-1, that involve AMR genes, non-AMR genes, and transposable genetic elements. For example, analysis of the genetic context showed ISEc68 and recombinase/integrase-encoding genes in the vicinity of bla KHM-1 . The ISEc68 transposase gene was flanked by 17-bp inverted repeats IRL (5=-GGAAGGTGCGAATAAGT-3=) and IRR (5=-ACTTAATCGCAGCTTCC-3=); however, the repeat regions did not encompass the KHM-1-encoding gene. On the other hand, the region from bla KHM-1 to int, encoding tyrosine-type integrase, was conserved (Fig. 4B). It is possible that both plasmids acquired this region from a common ancestor (i.e., bla KHM-1 -harboring mobile genetic element) via homologous recombination in the evolution process. Further studies are required to conclude which elements mediate a transposition mechanism involving bla KHM-1 . In addition, bla OXA10 was located in a class 1 integron, which includes a site-specific recombination system capable of integrating and expressing open reading frames contained in structures called mobile gene cassettes. This integron gene cassette array, aadA1-bla OXA-10-cmlA5-arr2-dfrA14, has been identified in other plasmids harbored by several Enterobacteriaceae, such as Providencia stuartii (19), demonstrating that gene module transpositions are a vital factor in the successful spread of bla genes among various plasmids.
An increase in MICs of ␤-lactams was detected with transformants of E. coli DH5␣ with the recombinant plasmids pHSG-bla CTX-M-2 , pHSG-bla DHA-1 , pHSG-bla KHM-1 , pHSGbla OXA-10 , and pHSG-bla OKP-A-4 . Of these, bla CTX-M-2 enhanced amoxicillin, piperacillin, cefotaxime, cefepime, and aztreonam resistance, while the bla DHA-1 strain showed a higher MIC increase for ceftazidime than the bla CTX-M-2 stain. Clavulanic acid, a mechanism-based ␤-lactamase inhibitor, could inhibit CTX-M-2 activities but could not inhibit DHA-1 ( Table 1). The production of OXA-10 or OKP-A-4 strongly increased the MICs of amoxicillin and piperacillin (Table 1). CTX-M-type enzymes are classified as Ambler class A extended-spectrum ␤-lactamases (ESBLs) and exhibit a striking substrate preference for cefotaxime and ceftriaxone over ceftazidime because of the unique geometry of the ␤-lactam-binding site (20). DHA-1 is an AmpC-type ␤-lactamase belonging to Ambler class C and generally not inhibited by clavulanic acid. AmpC-type ␤-lactamases harbor hydrolytic activity against cephalosporin and cefamycin and variably to aztreonam, but they remain sensitive to cefepime and carbapenems (21,22). OXA-10-type enzyme, classified as Ambler class D, was known to have narrow-spectrum ␤-lactamase activity, although variants in this enzyme family (such as OXA-11 and OXA-16) exhibit expanded-spectrum activity (23). The class A ␤-lactamase OKP enzymes, which are similar to SHV and LEN enzymes, exist in almost all K. quasipneumoniae chromosomes and are penicillinases (24,25). Our antimicrobial susceptibility results are completely consistent with those of these previous studies.
Only bla KHM-1 enhanced carbapenem resistance (Table 1). KHM-1 is an acquired Ambler class B MBL and harbors the hydrolytic efficiencies of carbapenems. So far, this enzyme has been recognized as a rare form of MBL because there is only one report of its isolation (14). However, our study has isolated KHM-1 in a transferable plasmid (pTMSNI47-1) harbored by an efficient carrier (K. quasipneumoniae) from a hot spot for horizontal gene transfer (an STP), and this enzyme might be widely disseminated in the near future in Japan. Carbapenem resistance has been documented by several mechanisms besides carbapenemase production. One of them is the loss of outer membrane porins combined with ESBLs or AmpC ␤-lactamases (26,27). We did not examine the expression of porins, CTX-M-2, or DHA-1 in this study; therefore, these may also have influenced carbapenem resistance. However, regardless of the mechanism of resistance, horizontal transfer of pTMSNI47-1, harboring four different Ambler class ␤-lactamases, leads to a very-wide-spectrum ␤-lactam resistance acquisition with mutual compensation.
Except for ␤-lactams, SNI47 was fully resistant to SXT (Fig. 3). Previous reports have stated that sul contributes to SXT resistance, and the sul and dfrA genes could synergistically lead to high-level SXT resistance (28,29). Both genes exist in pTMSNI47-1. Additionally, it was interesting to find that the sul1 gene was triplicated in this plasmid (Fig. 4A), suggesting that these genes, especially triplicated sul1 genes, might play an important role in high-level SXT resistance. On the other hand, SNI47 was sensitive to kanamycin, gentamicin, and amikacin (Fig. 3). Aminoglycoside resistance may occur due to several mechanisms. Of them, aadA1, which is located in the class 1 integron of pTMSNI47-1, encodes an aminoglycoside adenylyltransferase that influences enzymatic modification and inactivation (30). This enzyme imparts streptomycin and spectinomycin resistance by modifying the 3-hydroxyl position of streptomycin and the 9-hydroxyl position of spectinomycin, but it has marginal effects on the resistance to other aminoglycosides (30). Our results are consistent with those of these previous studies (Fig. 3 and Table 2).
The oqxA and oqxB genes, encoding the OqxAB efflux pump, and plasmid-mediated quinolone resistance (PMQR) gene qnrB4 were found in the chromosomes of SNI47 and pTMSNI47-1, respectively. However, contrary to our expectations, SNI47 was sensitive to all tested quinolones (Fig. 3) and did not reach the Clinical and Laboratory Standards Institute (CLSI) breakpoint (Table 2). At least three distinct mechanisms of quinolone resistance in bacteria have been identified: (i) mutations in target enzymes, (ii) alteration in membrane permeability, and (iii) protection of target enzymes from quinolone inhibition (31). Of them, resistance is mediated mainly by the accumulation of point mutations in the quinolone resistance-determining region (QRDR) of DNA gyrase (gyrA) and DNA topoisomerase IV (parC) (3,31). It was reported that differences in the expression level of OqxAB influenced reduced susceptibility to quinolones; however, most fell short of susceptibility breakpoints (32). PMQR genes contribute to the low level of resistance to fluoroquinolones, and these genes exert their influence by widening the mutant selection window and elevating mutant prevention (33). Our present Etest results for fluoroquinolones support the abovementioned contention. The MICs for SNI47, which harbored the oqxA, oqxB, and qnrB4 genes and did not harbor mutations in QRDR, were 0.5 g/ml for ciprofloxacin and 1 g/ml for levofloxacin. On the other hand, those for transconjugant J53/pTMSNI47-1 (harboring only qnrB4) were 0.25 g/ml for ciprofloxacin and 0.5 g/ml for levofloxacin, corresponding to Ͼ4and Ͼ8-fold increases compared with those for the recipient strain, respectively (Table 2). To some extent, these genes contribute to fluoroquinolone resistance, although both SNI47 and transconjugants J53/pTMSNI47-1 did not reach the CLSI breakpoint, even for intermediate resistance.
The present study uncovered evidence that the horizontal transfer of pTMSNI47-1 was a causative agent of a very-wide-spectrum ␤-lactam resistance. Whole-genome sequencing illustrated that isolate SNI47 was K. quasipneumoniae subsp. quasipneumoniae and harbored four plasmids. Of these, the IncA/C2 plasmid, pTMSNI47-1, could transfer to E. coli J53 through conjugation and carried resistance genes, including those for four ␤-lactamases (bla CTX-M-2 , bla DHA-1 , bla KHM-1 , and bla OXA-10 ). The MICs of ␤-lactams for both transconjugant J53/pTMSNI47-1 and transformant DH5␣/pTMSNI47-1 were higher than those for the recipient strain. Furthermore, bla KHM-1 was mainly responsible for reducing carbapenem susceptibility. Our findings indicate that this plasmid conferred properties for wide-spectrum ␤-lactam resistance, including carbapenems, with mutual compensation. The highly transferable plasmid pTMSNI47-1 was detected in efficient carriers isolated from hot spots for horizontal gene transfer. Therefore, it might be widely disseminated and become a concerning clinical issue in the near future.

MATERIALS AND METHODS
Bacterial strains used in this study. A 10-ml aliquot of influent water from an STP in Japan was cultured overnight in 2ϫ brilliant green lactose bile (BGLB) broth (Eiken Chemical, Tokyo, Japan) at 37°C. A 1-ml aliquot of the gas-producing BGLB broth culture was spread on a Pro-media Tricolor agar (ELMEX, Tokyo, Japan) plate to perform antimicrobial susceptibility testing with imipenem and meropenem using BD Sensi-Discs (Becton, Dickinson and Company, Franklin Lakes, NJ). Red colonies (␤-galactosidase produced by coliforms hydrolyzes the Magenta-GAL complex) formed inside the inhibition rings were picked, and we determined the MICs of imipenem and meropenem using the dry-strip technique (Etest; bioMérieux, La Balme-les-Grottes, France) according to CLSI criteria. MICs of Ͼ32 mg/liter were identified for each drug. We designated this coliform isolate strain SNI47. S1 pulsed-field gel electrophoresis and gel extraction. S1-PFGE, for separating plasmids from chromosomes, was performed according to the method of Barton et al. (34), with modifications. Briefly, DNA plugs digested with S1 nuclease (Takara Bio, Shiga, Japan) were electrophoresed on a CHEF Mapper XA PFGE system (Bio-Rad, Hercules, CA) with autoalgorithms for 5 to 250 kbp. A lambda ladder (Promega, Fitchburg, WI) was used as the size marker. Extraction of the chromosome or plasmid DNA fragments from the electrophoresed gel for MiSeq analysis was performed using a Zymoclean large-fragment DNA recovery kit (Funakoshi Co., Ltd., Tokyo, Japan).
Preparation of genomic DNA. The genomic DNA of SNI47 was extracted for MiSeq analysis using a QIAamp DNA minikit (Qiagen GmbH, Hilden, Germany). The concentration of the extracted genomic DNA was determined using a QuantiFluor ONE double-stranded DNA (dsDNA) system (Promega), and its purity was assessed using a NanoDrop 2000c spectrophotometer (Thermo Fisher Scientific, Waltham, MA).
DNA library preparation and MiSeq analysis. An index-tagged library was prepared using the Nextera XT DNA library preparation kit (Illumina, San Diego, CA), and 300-bp paired-end reads were sequenced on an Illumina MiSeq instrument according to the manufacturer's instructions. To ensure that only high-quality data were used for assembly, reads were trimmed and filtered using the CLC Genomics Workbench 11.0 (Qiagen) set to a minimum length of 100 bp and a quality score threshold of 30. These trimmed reads were assembled de novo using the CLC Genomics Workbench with default settings. Species identification was performed by ANI analysis (http://enve-omics.ce.gatech.edu/ani/). Phylogenetic analysis of k-mer diversity was performed according to a previously reported method (35). The k-mer length was set at 16 bases for this analysis. The MLST web server (https://cge.cbs.dtu.dk/services/ MLST/) was used to determine sequence types for chromosome. The PubMLST web server (https:// pubmlst.org/) was used to determine sequence types or allele ID for plasmid. The ResFinder web server (https://cge.cbs.dtu.dk/services/ResFinder/) was used to identify AMR genes. The PlasmidFinder web server (https://cge.cbs.dtu.dk/services/PlasmidFinder/) was used to identify the Inc type. The threshold for minimum coverage was set at 60% of the length of the gene sequence in the database, with a minimum sequence identity of 80%.
Conjugation experiment. The SNI47 strain containing a ␤-lactamase-encoding plasmid was used as a donor, and the spontaneous-sodium azide-resistant E. coli J53 strain was used as a recipient in our conjugation experiment. The recipient was susceptible to all antibiotics tested and did not harbor a ␤-lactamase-encoding plasmid. SNI47 and E. coli J53 were mixed in a 1:1 ratio, and then the mixture was cultured at 37°C for 5 h. Transconjugants were selected on Pro-media tricolor agar supplemented with sodium azide (100 g/ml) and ampicillin (50 g/ml). The blue colonies (␤-glucuronidase produced by E. coli hydrolyzes the 5-bromo-4-chloro-3-indolyl-␤-D-glucuronic acid [X-Gluc] complex, indicating successful plasmid transformation) were picked and subjected to S1-PFGE. Then plasmid DNA was extracted from the transconjugant colonies using a HiSpeed plasmid maxikit (Qiagen). E. coli DH5␣ cells were transformed with the extracted plasmid. The Etest dry-strip technique was used to determine the hydrolytic activities of the J53 transconjugants and DH5␣ transformants.