Klebsiella quasipneumoniae Provides a Window into Carbapenemase Gene Transfer, Plasmid Rearrangements, and Patient Interactions with the Hospital Environment

Several emerging pathogens have arisen as a result of selection pressures exerted by modern health care. Klebsiella quasipneumoniae was recently defined as a new species, yet its prevalence, niche, and propensity to acquire antimicrobial resistance genes are not fully described.

human pathogen. More recently, whole-genome sequencing has revealed that many isolates classified as K. pneumoniae actually encompass three related but distinct species: K. pneumoniae, Klebsiella variicola, and Klebsiella quasipneumoniae (1,2). K. quasipneumoniae was originally thought to be largely confined to agriculture and the environment; however, it appears that it may also be prominent in human disease (3), and several recent reports have demonstrated that it harbors virulence factors and acquires clinically relevant genes of antimicrobial resistance (4,5). Although there have been relatively few reports of K. quasipneumoniae to date, the true prevalence of this organism is likely underestimated as it is not generally distinguished from K. pneumoniae in routine testing of clinical laboratories (2).
Bacterial evolution via horizontal gene transfer is central to the ongoing crisis of antimicrobial resistance among clinically relevant bacteria. Hospital wastewater is being increasingly recognized as an ideal reservoir for resistance gene exchange and amplification, with ongoing antimicrobial selection pressure exerted through antimicrobials excreted in patient waste (6). Premise plumbing can be seeded by antimicrobial resistance genes in diverse bacterial strains and species and represents a difficult-totreat reservoir for ongoing gene exchange, creating successful drug-resistant bacteria that can thrive in both the environmental and human niches (7).
Whole-genome sequencing studies have demonstrated that our understanding of the interplay between antimicrobial resistance plasmids and their host strains/species is limited (8). The host range of a plasmid is critical for acquisition and persistence in specific species, but it appears that some bacterial strains are better equipped than others to prevent the acquisition of or destroy foreign plasmid DNA (9). The durability of plasmid acquisition events and the creation of new highly resistant strains reflect complex dynamics which depend on the characteristics of the plasmid in question as well as host strain tolerance (10,11). Seldom do we have the opportunity to witness strains acquiring plasmids in vivo or in the environment, and inferences about genetic rearrangements are often highly speculative. However, understanding the mechanisms and frequency of resistance gene transfer events occurring in real world contexts can provide important insights into the wider evolutionary landscape creating modern multidrug-resistant bacteria which cannot be effectively modeled in lab experiments (12).
Within our institution, we have seen ongoing transmission of diverse carbapenemase-producing organisms for the last decade, driven by genetic exchange of the Klebsiella pneumoniae carbapenemase (KPC) gene (bla KPC ) in patients and the environment (13,14). This has enabled us to understand specific pathways of genetic mobility involving numerous different mobile genetic elements and host bacterial species (13,15). Herein, we examine bla KPC acquisition and associated genetic rearrangements within K. quasipneumoniae as a real-life representation of an emerging pathogen associated with the hospital wastewater environment.

RESULTS
From our collection of bla KPC -positive isolates from patients (2007 to 2017) and the hospital environment (2013 to 2017), there were a total of 32 bla KPC -positive K. quasipneumoniae isolates, all of which were identified as K. pneumoniae in the clinical microbiology laboratory (Table 1). Twenty-three of these were K. quasipneumoniae subspecies quasipneumoniae (KpIIA) (10 patient isolates from four patients and 13 environmental isolates from seven rooms), and nine were K. quasipneumoniae subspecies similipneumoniae (KpIIB) (five patient isolates from four patients and four environmental isolates from four rooms). The KpIIA and KpIIB isolates were separated by Ͼ100,000 single nucleotide variants (SNVs). We identified a single strain of KpIIA and four strains of KpIIB differing from each other by Ͼ20,000 SNVs (Fig. 1). Many isolates have multiple virulence factors (see Data Set S1 in the supplemental material), including several genes involved in capsule production (16) and several fimbrial elements. A type VI secretion system was present in all KpIIA but not all KpIIB isolates. From a resistance gene standpoint, in addition to bla KPC , all isolates harbored fosA and bla OKP as well as a multidrug efflux transporter (oqxA-oqxB) (17).
Within the KpIIA strain, there were two subclades separated by ϳ150 SNVs (Fig. 1a). The first subclade contained two isolates separated by 10 SNVs (Fig. 1a). CAV1360 was from patient 1 in November 2009, and CAV2279 was identified in early 2014 (shortly after environmental sampling began) from room B that patient 1 had occupied in May 2009.
The second subclade of KpIIA contained isolates from three patients (patients 2 to 4) and six rooms (rooms A and C to G). The earliest of these was from patient 2 in November 2013. Patient 2 was in the hospital with a prolonged stay in the surgical, trauma, and burn intensive care unit (STBICU) following complications of a liver transplant (Fig. 2). Patient 2 was noted to be first colonized with bla KPC -positive KpIIA in November 2013. KpIIA was not found in the STBICU environment prior to closure for remediation of KPC contamination of the drains in December 2013. Following drain exchange and unit reopening, patient 2 was immediately moved back into the STBICU and subsequently occupied rooms C, D, E, and G in the STBICU, suggesting that the KpIIA isolates in these rooms originated from patient 2 (Fig. 2). Patient 3 was admitted to the STBICU at the same time as patient 2 and thus could have acquired KPC-KpIIA directly from patient 2 without environmental transmission. Patient 4 was later admitted to STBICU room E for 28 days and discharged before he was found to have KpIIA. He was never on a ward at the same time as any other patients known to carry KpIIA, suggesting acquisition from the hospital environment. There were four patients (patients 5 to 8) carrying four distinct strains of bla KPC -KpIIB seen over a 5-year period (Fig. 1b, Table 1). For patient 7, the same KpIIB strain (ϳ80 SNV differences) was also seen in sinks from two rooms in the medical intensive care unit (MICU) (rooms H and I) and two rooms in the STBICU (rooms J and K) in December 2013 when environmental sampling first began; this preceded the detection in the patient in February 2014. Patient 7 was admitted to the MICU (location of rooms H and I), but did not stay any of the rooms where the isolates within the same KpIIB were identified. The other three patients with KpIIB each had unique bla KPC strains, none of which were identified in another patient or the environment. Patient 6 also with a unique strain had a prolonged hospital stay and was also colonized/infected with another bla KPC -positive species (K. pneumoniae). Three patients developed infections with KPC-KpIIA (Table 1). Patient 1 died of ventilator-associated pneumonia with KPC-KpIIA following a complicated heart transplant. Patient 2 had both ventilator-acquired pneumonia, which was successfully treated, and a subsequent untreatable intraabdominal infection with KPC-KpIIA bacteremia, which contributed to the patient's death after a long hospital stay with a complicated liver transplant. Patient 4 had a successfully treated complicated KPC-KpIIA urinary tract infection. Patient 3 did not develop an infection with KpIIA. None of the patients with KpIIB developed K. quasipneumoniae infections; however, two of the patients did develop infections with other species carrying bla KPC (K. pneumoniae for patient 6 and Serratia marcescens for patient 8) ( Table 2).
Genetic variation and rearrangements within KpIIA. All KpIIA isolates were closely related at the core chromosome level, with a maximum divergence of Ͻ180 SNVs. If bla KPC was acquired only once in this lineage, then any sequence variation within the 10-kb bla KPC transposon Tn4401 would be the result of mutational change, which is expected to be rare. Surprisingly, the Illumina sequence data revealed a great deal of sequence variation within Tn4401 (Fig. 1a). Two sites (positions 8015 and 9663 in the Tn4401b reference) showed variation at the single nucleotide level, and one isolate had a deletion at positions 7075 to 7153. Interestingly, several isolates showed mixtures at one or both of the variable sites, indicating two or more different versions of Tn4401 in the same isolate. This included mixtures at position 8015, which is located within the bla KPC gene and differentiates bla KPC-2 and bla KPC-3 , indicating that there were isolates with both bla KPC alleles.
Similarly, if a single bla KPC plasmid was acquired and stably maintained within KpIIA, then we would expect to see a single flanking sequence context for Tn4401. On the contrary, there was significant diversity in Tn4401 flanking regions, with eight and seven different 5-bp sequences on the left and right sides of Tn4401, respectively, suggesting active transposition of Tn4401 within KpIIA and/or multiple plasmid acquisitions.
To better understand the origin of the genetic diversity within and surrounding Tn4401, we performed long-read PacBio sequencing on three of the KpIIA isolates (CAV2013 from patient 2, CAV1947 from room C, and CAV2018 from room C), as well as a S. marcescens isolate from patient 2 (CAV1761). The room C isolates were chosen because this room only became positive after the admission of patient 2 following sink trap exchange in the STBICU; hence, they are expected to be descended from the patient 2 KpIIA. Both patient 2 isolates had a single bla KPC plasmid each ( Fig. 3a and b). The KpIIA isolate had a 447,095-bp "RepA_CP011611" bla KPC-3 plasmid, and the S. marcescens isolate had a 69,158-bp IncU/IncX5 bla KPC-2 plasmid (18). Both plasmids contained Tn4401b; however, there were two SNV differences within the Tn4401b sequence, one at position 8015 (differentiating bla KPC-2 and bla KPC-3 ) and one at position 9663.
The KpIIA isolates from room C (CAV1947 and CAV2018) had three and two bla KPC plasmids, respectively ( Fig. 3c and d). Both isolates harbored the IncU/IncX5 bla KPC plasmid from the patient 2 S. marcescens isolate, indicating likely bla KPC plasmid transfer from S. marcescens to K. quasipneumoniae (Fig. 3e). In CAV1947, the plasmid sequence was identical to that from the patient isolate, CAV1761, with the exception of two large indels (Fig. 4a). One of these was a 16,315-bp deletion immediately adjacent to Tn4401, presumably as a result of intramolecular transposition in cis, that converted the left flanking sequence from TTTTT to ACAAT and removed the IncU replicon sequence (Fig. 3g). In CAV2018, the plasmid sequence was identical to that in CAV1761, except for a single 5,923-bp deletion that truncated part of the Tn4401 sequence ( Fig.  3h and 4a).
Both isolates also harbored the ancestral RepA_CP011611 bla KPC plasmid from the patient 2 KpIIA isolate, with several SNVs and large indels (Fig. 4b). Interestingly, in CAV2018, one of the SNVs was located within Tn4401, such that the CAV2018 RepA_CP011611 plasmid contained bla KPC-2 rather than bla KPC-3 . Given that there was plasmid transfer of the IncU/IncX5 bla KPC-2 plasmid from S. marcescens, we infer that the bla KPC-2 -containing RepA_CP011611 plasmid most likely arose as a result of homologous recombination between these two different plasmids flanking the bla KPC region ( Fig. 3f and k). The Illumina data also revealed similar patterns of homologous recombination in other isolates (notably CAV2983, CAV2984, CAV3444, CAVp64 and CAVp275, which all have the TTTTT IncU/IncX5 plasmid flanking sequences, but with the C8015T bla KPC-3 mutation and without the T9663C mutation), suggesting frequent exchange of Tn4401 variants between different bla KPC plasmids within the same host bacterium ( Fig.  1 and 3k). CAV1947 also harbored a third bla KPC plasmid, representing transposition of Tn4401 into a 4,095-bp nontypeable plasmid that was present in the CAV2013 ancestor from patient 2 ( Fig. 3i and 4c).
K. quasipneumoniae has acquired bla KPC on multiple occasions. The average unique plasmid Inc types per isolate was more than four according to PlasmidFinder Nosocomial KPC-K. quasipneumoniae Antimicrobial Agents and Chemotherapy (Data Set S1). Within KpIIB, there were four divergent strains separated by Ͼ20,000 SNVs, suggesting four separate acquisitions of bla KPC in this subspecies. Within KpIIA, there were two subclades separated by ϳ180 SNVs. Given that Tn4401 variation and flanking sequences were different between the two subclades (apart from the GTTCT flanking sequence which is known to be present in many different bla KPC plasmids) (13) and that there was no epidemiological overlap, it is most likely that the subclades acquired bla KPC independently. Additionally, as described above, the second subclade likely acquired bla KPC on two occasions, with the second acquisition originating from S. marcescens. Therefore, overall, there were likely seven acquisitions of bla KPC by K. quasipneumoniae: three in KpIIA and four in KpIIB. Interestingly, there was evidence that one of the acquisitions in KpIIB also originated from S. marcescens, indicating the compatibility of these two species in exchanging plasmids. This was in the patient 8 KpIIB lineage. Patient 8 was first colonized with bla KPC -S. marcescens carrying Tn4401b with a T9663C mutation and TTTTT/TTTTT flanking sequences. Four months later, bla KPC -KpIIB was identified with the same Tn4401 mutation and flanking sequences, suggesting plasmid transfer from S. marcescens to K. quasipneumoniae within this patient.

DISCUSSION
We describe the behavior of nosocomial bla KPC -positive K. quasipneumoniae strains within a single hospital setting, observing their propensity to take up multiple carbapenemase plasmids from other species and disseminate between patients and sink drains. Our study also suggests that rapid genetic rearrangement occurs in the mobile genetic elements carrying bla KPC in KpIIA.
There is increasing recognition that the hospital environment is an important reservoir in the transmission of carbapenemase-producing Enterobacteriaceae (CPE), but delineating transmission chains is often challenging (19,20). Through our K. quasipneumoniae example, we provide compelling evidence for patient-to-drain and drain-to-patient transmission, as has been observed in other studies (7). We also provide evidence supporting the ability of K. quasipneumoniae to be maintained in the environment for a long period of time, with the first subclade of KpIIA detected in the environment on initial sampling, even though it had not been seen in a patient nor had that patient been in the room for more than 3 years. The costly closure of the STBICU and exchange of all the sink drain plumbing pipes had a limited effect on environmental contamination with CPE; instead, it appears to have provided an environment for immediate new seeding and establishment of previously unobserved carbapenemresistant strains. There are potential other reservoirs to consider, but health care workers have not been identified as a source of CPE. We have a fairly robust screening program in place and have sequenced all patient isolates and included all K. quasipneumoniae in this series, making silent colonization less likely (21,22). We were not sampling the toilets or hoppers during most of the study, and we have only sequenced a portion of environmental isolates which could provide an unidentified environmental source of K. quasipneumoniae (14). Understanding the dynamics and natural history of colonization of premise plumbing with CPE will be important in designing effective interventions to limit transmission (23).
Although there have only been a few reports of K. quasipneumoniae since its definition as a species in 2014, it appears that this organism is widespread (2,5,24,25). As seen here, it is not readily distinguished from K. pneumoniae with current clinical microbiology techniques; thus, the true prevalence is unknown (2,26). On the evolutionary time scale, modern medicine has provided a novel ecology with immunocompromised patients, widespread antimicrobial use, newly circulating antimicrobial resistance genes, and the design of the modern hospital providing new microbiologic niches for organisms to emerge (7,27). We found several virulence factors in our collection, some of which have been identified in other K. pneumoniae or K. quasipneumoniae: capsule, fimbrial adhesion proteins, and a type VI secretion system (5,16). As seen here, we provide evidence for K. quasipneumoniae to be sustained in both a human host and the environment, encountering several different species which may be relatively new in the evolutionary tree of Klebsiella sp. (1). As a consequence of these encounters, the transfer of mobile DNA occurs via traceable carbapenemase plasmids. We found evidence for seven independent acquisitions of bla KPC by K. quasipneumoniae, suggesting that this species is amenable to take up plasmids from other species of Enterobacteriaceae. Given the difficulties in accurately identifying K. quasipneumoniae, this species may therefore be more significant in the context of bla KPC dissemination than has previously been recognized.
Within K. quasipneumoniae, there was surprising variability in mobile elements carrying bla KPC , which was the result of several different processes observed among a limited number of highly related isolates (n ϭ 23). We also found multiple acquired antimicrobial resistance genes, and every isolate had more than one plasmid incompatibility type (18). Specifically, there were multiple independent bla KPC plasmid acquisitions: homologous recombination between different bla KPC plasmids, transposition of Tn4401 into new plasmids, intramolecular transposition in cis of Tn4401, a deletion within Tn4401, and a deletion truncating Tn4401. This high degree of genetic mobility has been similarly observed in other small studies (28,29) and highlights the difficulty in developing an accurate understanding of the transmission epidemiology of important drug resistance genes which can be rapidly mobilized by multiple independent genetic modalities.
Within KpIIA, there were multiple acquisitions of bla KPC within the same lineage, such that a bla KPC -positive KpIIA strain acquired a second unrelated bla KPC plasmid from S. marcescens. Consequently, there were then two different bla KPC plasmids, with different Tn4401 sequences and different bla KPC alleles, within the same host bacterium. This situation facilitated multiple rearrangements via homologous recombination between the different plasmids, resulting in the generation of new combinations of Tn4401 SNVs and host plasmids. Multiple acquisition of resistance plasmids followed by rearrangements between those plasmids is likely to be important in the generation of adaptive allelic combinations which contribute to the amplification of cross-class antimicrobial resistance within strains. High-risk clones with a propensity to take up antimicrobial resistance plasmids may represent important targets for intervention (30).
This study has several limitations. Most notably, it is a small retrospective series, preventing a full understanding of the role of the environment. Also, the order of genetic rearrangements is also not completely known, given the limited number of long-read sequenced isolates and inability to capture all isolates from the environment over time. We offer, however, that this is higher resolution than seen in many studies, and the analysis contributes to the greater understanding of rapid rearrangement and mechanisms at play regarding the mobility of genetic elements harboring genes of antibiotic resistance in Enterobacteriaceae.
In summary, we demonstrate the relevance of K. quasipneumoniae as a species fit for nosocomial transmission in the modern era that is capable of acquiring and maintaining relevant resistance elements.

Setting.
Isolates were collected at the University of Virginia, a 619-bed tertiary care hospital, from August 2007 to May 2017. A robust K. pneumoniae carbapenemase-producing organism (KPCO) prevention program existed throughout the study period, as previously described (31), and included perirectal screening beginning in April 2009 in the medical intensive care unit (MICU) and surgical intensive care unit (STBICU) and weekly screening of all patients in the MICU and STBICU as well as units where any known KPCO-colonized patient was present (32). Screening was performed as previously described (32). Clinical Enterobacterales and Aeromonadaceae isolates, as identified by matrix-assisted laser desorption ionization time of flight mass spectrometry (MALDI-TOF MS) or VITEK2 (bioMérieux, Durham, NC), with an elevated ertapenem or meropenem MIC by VITEK2 (bioMérieux, Durham, NC), immediately underwent CarbaR (Cepheid, Sunnyvale, CA) carbapenemase PCR testing. All species identification was performed using a combination of VITEK2 and Vitek-MS (bioMérieux, Durham, NC). Clinical data were gathered by retrospective electronic medical record review under University of Virginia Health Sciences institutional review board (IRB) number 13558 with waiver of consent.
In September 2013, sink trap sampling for KPCO began using previously described techniques (14) with a swab for drain collection and P-trap water. Following identification of KPCO in the hospital environment, the STBICU was closed to patient care in December 2013. Over the following 9 weeks, all sink drain pipes were removed and replaced with sink traps that eliminated overflows in the sink bowl. Patients were readmitted to the surgical intensive care unit in February 2014. Bleach, hydrogen peroxide, and ozone-impregnated water (2 ppm) were applied weekly from February to May 2014 in the STBICU (following drain exchange and sink bowl overflow closure and removal) and from March to May 2014 in the MICU (without drain exchange or sink bowl overflow removal).
Whole-genome sequencing and bioinformatics analysis. Illumina sequencing was performed as described previously (33). PacBio long-read sequencing and assembly were performed as previously described (13).
Broad-level species classification was performed using Kraken (34). To identify K. quasipneumoniae isolates, we queried all isolates initially classified as K. pneumoniae against reference sequences representing each of the four clades described by Holt et al. (1). We arbitrarily selected a single reference sequence for each clade; these were ERR025521 (KpI), ERR025986 (KpIIA), ERR025528 (KpIIB), and ERR025573 (KpIII). We used mash v1.1.1 (35) with parameters "-r -m 5" to compare Illumina data for each of our isolates to these reference sequences. Each isolate was then assigned to one of the four K. pneumoniae clades according to the reference with the lowest distance value. All isolates assigned to KpIIA or KpIIB were included in the analysis. In addition, we also included any other KPCO isolates from patients carrying K. quasipneumoniae.
To identify chromosomal single nucleotide variants (SNVs), Illumina reads for each K. quasipneumoniae isolate were mapped to the CAV2013 chromosome sequence (derived from long-read sequencing), with mapping and variant calling performed as described previously (36). A phylogeny was generated using IQ-TREE v1.3.13 (37) from an alignment of variable sites where at least 70% of samples had a high-quality reference/variant call (i.e., we excluded sites where Ͼ30% of samples had an "N" call). This was run with parameters "-blmin 0.000000001 -nt 4 -m GTR," with -fconst used to specify the number of invariant sites.
To identify Tn4401 variation and flanking sequences from Illumina data, we used TETyper with published parameters (38).
The Illumina paired-end short reads were de novo assembled using SPAdes assembler v 3.10.1 (35). Assembly statistics were evaluated using QUAST v4.0. (36) Plasmid Inc typing was performed using PlasmidFinder v2.0.1 against the Feb 2018 version of the Enterobacteriaceae database (16), with minimum identity of 80% and minimum coverage of 50%. Acquired antimicrobial resistance genes were screened from the assemblies using NCBI's AMRFinder tool v1.0, which relies on a curated AMR protein database and a collection of hidden Markov models, with 90% minimum identity to translated amino acid residues and 50% minimum coverage of reference protein sequence. (37) Identification of bacterial virulence genes was performed using ABRicate v0.8.11 (https://github.com/tseemann/abricate), against the Virulence Factors Database (accessed on Feb 2019), with 80% minimum identity and 50% minimum coverage.

SUPPLEMENTAL MATERIAL
Supplemental material for this article may be found at https://doi.org/10.1128/AAC .02513-18.