Chromosomal Integration of the Klebsiella pneumoniae Carbapenemase Gene, blaKPC, in Klebsiella Species Is Elusive but Not Rare

ABSTRACT Carbapenemase genes in Enterobacteriaceae are mostly described as being plasmid associated. However, the genetic context of carbapenemase genes is not always confirmed in epidemiological surveys, and the frequency of their chromosomal integration therefore is unknown. A previously sequenced collection of blaKPC-positive Enterobacteriaceae from a single U.S. institution (2007 to 2012; n = 281 isolates from 182 patients) was analyzed to identify chromosomal insertions of Tn4401, the transposon most frequently harboring blaKPC. Using a combination of short- and long-read sequencing, we confirmed five independent chromosomal integration events from 6/182 (3%) patients, corresponding to 15/281 (5%) isolates. Three patients had isolates identified by perirectal screening, and three had infections which were all successfully treated. When a single copy of blaKPC was in the chromosome, one or both of the phenotypic carbapenemase tests were negative. All chromosomally integrated blaKPC genes were from Klebsiella spp., predominantly K. pneumoniae clonal group 258 (CG258), even though these represented only a small proportion of the isolates. Integration occurred via IS15-ΔI-mediated transposition of a larger, composite region encompassing Tn4401 at one locus of chromosomal integration, seen in the same strain (K. pneumoniae ST340) in two patients. In summary, we identified five independent chromosomal integrations of blaKPC in a large outbreak, demonstrating that this is not a rare event. blaKPC was more frequently integrated into the chromosome of epidemic CG258 K. pneumoniae lineages (ST11, ST258, and ST340) and was more difficult to detect by routine phenotypic methods in this context. The presence of chromosomally integrated blaKPC within successful, globally disseminated K. pneumoniae strains therefore is likely underestimated.

chromosomally, it will be maintained through replication without selective pressure, and gene loss from the bacterial population becomes less likely (4)(5)(6).
The Klebsiella pneumoniae carbapenemase gene (bla KPC ) is maintained within the self-mobilizing 10-kb transposon Tn4401 (7). The frequency of transposition in vitro is relatively high (4.4 ϫ 10 Ϫ6 /recipient cell) and without site specificity, but the rate of movement outside laboratory settings is largely unknown (8). Nevertheless, Tn4401 has been described in several different genetic environments (9,10).
Historically, most descriptions of bla KPC have been in a globally successful lineage of K. pneumoniae, namely, multilocus sequence type 258 (ST258) and the associated clonal group 258 (CG258), which includes ST11, ST258, and ST340 (11). CG258 isolates are widespread, albeit concentrated in geographic hotspots, with the bla KPC gene being plasmid associated in the majority of reports (9). Interestingly, the earliest observed chromosomal Tn4401 integration events have been sporadic and in non-Klebsiella spp.: Pseudomonas aeruginosa in 2006 (12), Raoultella spp. in 2008 (13) and Acinetobacter baumannii in 2009 (14). More recently, chromosomal integration among K. pneumoniae ST258 (n ϭ 4) isolates has been described in the United States by three separate groups (15)(16)(17).
The rate of chromosomal integration within clinical settings for many genetic elements is largely unknown. Chromosomal integration of bla KPC may be underreported, as its investigation often requires detailed genetic analysis, and if a plasmid copy is present then a chromosomal copy may be overlooked. Even high-resolution genetic methods, such as whole-genome sequencing (WGS) using short-read technologies, can be confounded by the presence of multiple copies of Tn4401 and/or repetitive flanking sequences, which limit the ability to accurately reconstruct the genetic context(s) of bla KPC (10,18). Here, we use a combination of short-and long-read WGS to investigate chromosomal integration of bla KPC among 281 isolates (62 distinct strains) of bla KPC -positive Enterobacteriaceae isolated from 182 patients in a single hospital.
As the de novo assembly approach was only able to assess 44% of isolates for possible chromosomal integrations, we also used a mapping approach, which is unaffected by multiple Tn4401 copies. Reassuringly, all seven isolates identified by the de novo assembly approach described above were also identified by this method as having a likely chromosomal integration of Tn4401. In addition, the mapping approach identified a further 11 isolates as having putative chromosomal integrations, all of which were unevaluable by the de novo assembly approach. Six of these were K. pneumoniae ST11 isolates from a single patient (the earliest isolate was CAV1351). Three were K. pneumoniae ST340 isolates from two additional patients (earliest isolates were CAV1217 and CAV1518). One was a non-CG258 K. pneumoniae isolate (ST244; CAV1042) from an additional patient, and one was a K. oxytoca isolate (CAV1755) unrelated to CAV1752 (Ͼ40,000 single-nucleotide variant [SNV] differences), also from an additional patient.
To confirm the chromosomal locations of Tn4401, we used long-read sequencing. For 6/8 patients identified above, the earliest isolate from each was sequenced using PacBio. In 5/6 cases, the presence of Tn4401 on the chromosome was confirmed, and in one case (CAV1042) it was refuted. For one of the two remaining patients (the earliest isolate was CAV1351), we did not perform additional PacBio sequencing, as a subsequent isolate from the same patient (CAV1392; 1 SNV difference from CAV1351) was previously PacBio sequenced as part of another study and shown to contain two copies of Tn4401, one chromosomal and one plasmid (18). CAV1518 from the final patient was not PacBio sequenced, as it was demonstrated using other methods (see below) to be a false positive from the mapping approach, since Tn4401 was not present on the chromosome.
Altogether, we found five distinct loci of chromosomal integration of bla KPC verified by long-read sequencing from six patients, with two additional patients having isolates that were falsely identified by the mapping approach ( Table 2). All chromosomal integrations were in Klebsiella spp., mostly K. pneumoniae CG258, even though CG258 represented only a small proportion of the outbreak (4/18 versus 2/164 non-CG258 patients, P ϭ 0.0009; 13/34 versus 2/247 isolates, P Ͻ 0.0001; Fisher's exact test).
Mechanism of chromosomal integration in K. pneumoniae ST340. There were three patients with K. pneumoniae ST340 isolates where the mapping approach identified Tn4401 as having a possible chromosomal location (CAV1417, CAV1217, and CAV1518) ( Table 2). To assess the molecular basis of chromosomal integration, the chromosomal region encompassing Tn4401 in the closed PacBio assembly of the ST340 CAV1417 isolate was aligned to the NJST258_2 reference (GenBank accession number CP006918), which belongs to the closely related ST258. Relative to the reference, CAV1417 had a 24-kb inversion and a 16-kb insertion, which included Tn4401 and several other resistance genes (Fig. 1A).
At one end of the 16-kb region was an IS15-ΔI element, highly genetically similar (3 nucleotide differences) to IS26, both of which undergo replicative transposition with 8-bp target site duplication (TSD) (19). IS26 has been shown to undergo frequent intramolecular transposition, which can result in inversion of the sequence between the original and duplicated elements. This process disrupts the pairing of TSD sequences, which can be used to trace the history of transposition events (20).
The other end of the 16-kb region terminated in a sequence with high similarity to the 14-bp IS15-ΔI inverted repeat (IR) sequence (IS15-ΔI, GGCACTGTTGCAAA; other, GGCTTTGTTGAATA; 10/14 nucleotide identities), suggesting that this acted as a cryptic recognition site mediating transposition of the entire 16-kb region, similar to a com-  posite transposon. The 8-bp flanking sequences of the 16-kb region were identical to the 8-bp flanking sequences of a neighboring IS15-ΔI element located at the other end of the 24-kb inversion relative to NJST258_2 (Fig. 1A). This indicates that the 16-kb region initially integrated into the chromosome by intermolecular transposition, with an 8-bp TSD (GTGATGGC). Subsequently, IS15-ΔI underwent intramolecular transposition, resulting in a 24-kb inversion, with duplication of a second 8-bp sequence (TTCAGATG), with one copy flanking each of the duplicated elements. In CAV1217, long-read sequencing demonstrated two copies of Tn4401, one chromosomal and one plasmid. For the chromosomal copy, the 8-bp sequence adjacent to the non-IS15-ΔI side of the 16-kb region (Fig. 1, right side) was identical to CAV1417 (GTGATGGC), while the plasmid copy had a different 8-bp flanking sequence (CGTAG-GCA).
In CAV1518, the above-described signature of chromosomal integration (GTGATGGC sequence adjacent to the 16-kb region) is not present (Fig. 1B). Instead, the flanking sequence is identical to the plasmid copy in CAV1217 (CGTAGGCA). Therefore, the 16-kb region is most likely plasmid located in CAV1518, indicating that this isolate was a false positive from the mapping method.
Clinical microbiologic characteristics. Susceptibility results and phenotypic testing for these isolates varied substantially (  Pink and blue shading indicate high sequence similarity (Ͼ99.8%) in the same or opposite orientations, respectively. Flanking sequences for mobile elements present in the complete (PacBio) CAV1417 genome are indicated (an asterisk indicates that the sequence shown has been reverse complemented). (B) Coverage of Illumina reads for CAV1417, CAV1217, or CAV1518 mapped to the complete CAV1417 genome across the region shown in panel A. Flanking sequences for Tn4401 and the right side of the 16-kb region were determined from the mapped Illumina reads. and CAV1417 were high-level carbapenem resistant but with weak to negative carbapenemase phenotypes. All of these isolates had a sequence disruption in major K. pneumoniae porin channels (ompK35 and/or ompK36) ( Table 2). CAV1453 had only a single Tn4401 copy in the chromosome; however, it was Tn4401a, which generally results in increased KPC expression compared to Tn4401b, as was seen here with a higher level of expression compared to the other isolates with a single copy of Tn4401b in the chromosome (21,22). It was negative by indirect carbapenemase testing (23) but weakly positive by the modified Hodge test. CAV1417 had a single chromosomal copy of Tn4401b, and phenotypic testing for carbapenemase activity was consistently negative, with low-level bla KPC expression by quantitative PCR. Despite negative carbapenemase production by phenotypic tests, CAV1417 was highly carbapenem resistant and had disruptions in both ompK35 and ompK36. The functional consequences of these disruptions have not been confirmed, but loss of function of these genes is expected to increase carbapenem resistance above that conferred by KPC alone (11,21,22,24).
K. oxytoca isolate CAV1755 was carbapenem resistant and phenotypically positive by carbapenemase testing. This was also true for the K. pneumoniae isolate CAV1042, which was ultimately rejected as having a chromosomal copy. Both of these isolates had multiple loci of Tn4401 integration (including a plasmid) with increased coverage of Tn4401b relative to the chromosome and also demonstrated increased bla KPC expression and intact porin channel sequences.
CAV1217 was not noted to be bla KPC positive until the patient was found to be colonized through perirectal screening (25) with a K. pneumoniae isolate with a weakly positive indirect carbapenemase test. This prompted further testing of the phenotypically carbapenem-susceptible K. pneumoniae urine isolate (CAV1217), which was then found to be bla KPC PCR positive. This isolate had no disruption in the ompK35 and ompK36 coding regions and had two copies of Tn4401 (one plasmid and one chromosomal) but with relatively low coverage.
K. oxytoca CAV1752, which had a single chromosomal copy of Tn4401b with low-level expression by quantitative PCR (qPCR), was detected on perirectal screening, again with conflicting phenotypic results (Table 2) (23). On automated susceptibility testing (VITEK 2), it was predicted to be susceptible to ceftriaxone and cefepime (MIC of Յ1 g/ml) and carbapenems, with an aztreonam MIC of 4 g/ml. Analysis of porin genes demonstrated an insertion disrupting ompK35.
Clinical course and epidemiology. All six patients confirmed to have chromosomal integration of bla KPC had multiple risk factors and prolonged hospital exposure either within our health system or at outside hospitals prior to transfer (Table 3). Several patients with extensive outside health care exposure had isolates identified relatively early during hospitalization, raising the possibility of isolates circulating elsewhere in the region at outside hospitals. Interestingly, the patient with CAV1518 (the ST430 K. pneumoniae without chromosomal integration but with an ancestor common to CAV1217 and CAV1417) presented at an outside hospital prior to exposure to our health system, consistent with persistence of this lineage elsewhere in the region.
There were three patients who had infections (two with ventilator-associated pneumonia and one complicated urinary tract infection) with a K. pneumoniae isolate with chromosomal integration of bla KPC . The outcomes were complete microbiologic resolution and clinical cure at 30 days without any patients having known relapse within our health system.
The two patients with distinct strains of K. oxytoca (CAV1752 and CAV1755) did not demonstrate signs of infection with these isolates, thus the clinical consequences remain unclear. Interestingly, both patients had been colonized months before with other bla KPC -positive Enterobacteriaceae. As new acquisition at our institution is relatively rare and there were no other patients with similar isolates, this finding suggests separate within-patient horizontal transfer and chromosomal integration of Tn4401.

DISCUSSION
We demonstrate, for the first time, multiple instances of chromosomal integration of Tn4401/bla KPC within a single-center outbreak. Although this is not the first description of chromosomal integration of Tn4401, the frequency of integration in a clinical setting is largely unknown and likely overlooked. With a focused effort on identifying all instances of chromosomal integration, this event does not appear to be rare. In addition, we found that phenotypic carbapenemase testing was variable in these contexts, and bla KPC could be missed by nonmolecular methods.
Illumina sequencing is increasingly used to undertake epidemiological investigation but is limited in investigating multicopy regions, which are of particular relevance to assessing resistance genes and mobile genetic elements. We utilized two bioinformatic approaches to characterize the genetic contexts of Tn4401. First, we used a de novo assembly approach, which identified three distinct chromosomal integrations. These were all verified by PacBio sequencing, demonstrating the robustness of this method. However, the method cannot identify chromosomal integration if multiple copies of Tn4401 are present or if Tn4401 is flanked by repetitive sequences, and only 44% of isolates were evaluable using this method. Second, we used a mapping approach which was able to assess all isolates but which identified several false positives. Therefore, results from the mapping approach on its own should be interpreted with caution, and a combination of methods may be most appropriate for other studies investigating sequence-based identification of resistance gene integration into the chromosome.
bla KPC chromosomal integration was overrepresented among CG258 strains in our outbreak, suggesting that it is more widespread in high-risk clones such as K. pneumoniae ST258. This phenomenon has been seen in the high-risk Escherichia coli ST-131 clone, with chromosomal integration of the CTX-M beta-lactamase gene followed by vertical transmission (26,27). It remains unclear if there is a propensity for chromosomal integration among certain K. pneumoniae strains or, more likely, whether integration is associated with the extended timespan over which these strains have been associated with Tn4401, representing greater opportunity for random integration.
Within K. pneumoniae ST340, we identified three patients with putative chromosomal integrations of Tn4401, confirmed by PacBio sequencing in two isolates, CAV1217 and CAV1417. Comparison of the chromosomal region surrounding Tn4401 in CAV1417 to a related strain lacking Tn4401, combined with the analysis of TSD sequences, allowed us to reconstruct the history of transposition events. This revealed that chromosomal integration of Tn4401 occurred via IS15-ΔI-mediated transposition of a 16-kb region encompassing Tn4401. Further analysis of TSD sequences from Illumina data for CAV1518 revealed that the 16-kb region was not integrated into the chromosome in this isolate but rather shared flanking sequence with the plasmid copy from CAV1217. Taken together, this indicates that Tn4401 most likely was acquired once in this lineage, in a common ancestor of all three isolates, with chromosomal integration occurring after the divergence of CAV1518 (in the ancestor of CAV1217 and CAV1417) and subsequent loss of the plasmid copy in CAV1417 (Fig. 2).
Another important feature of this study was the variability in phenotypic carbapenem resistance. A prior evaluation of phenotypic carbapenemase testing in Enterobacteriaceae from our institution (May 2010 to December 2011), using an ertapenem MIC of Ն1 g/ml by VITEK 2 for inclusion, demonstrated that the indirect carbapenemase test had 90% sensitivity for detecting bla KPC -producing isolates (23). Isolates with chromosomally integrated bla KPC accounted for 3 of the 5 false-negative bla KPC Enterobacteriaceae identified in that study (total of 56 isolates tested). ␤-Lactamase carriage on high-copy-number plasmids compared to low-copy-number plasmids is known to alter the degree of resistance seen phenotypically (22,28). Here, we found preliminary evidence that a single chromosomal copy of Tn4401 was associated with a more subtle carbapenemase phenotype and lower expression of bla KPC . For example, CAV1752 would not have been detected with our current methods. This also could contribute to the misclassification of these isolates in a typical clinical microbiology laboratory, where the cost of molecular screening is often prohibitive (25).The differential impact on expression of resistance of bla KPC located in the chromosome alone remains an interesting finding but will require additional in vitro studies to confirm this effect.
The impact of porin channel alterations may also affect the degree of in vitro resistance, highlighted here when comparing the related isolates CAV1417 and CAV1217. Although CAV1217 had two Tn4401 copies and CAV1417 a single copy, CAV1417 had disruptions in both porin channel genes and was more phenotypically resistant. Nevertheless, porin channel loss does not always result in high resistance (e.g., CAV1752 has a disruption in ompK35 but remained carbapenem susceptible), and the prediction of phenotype from genotype in these cases is challenging.
Our small number of cases is insufficient to reliably assess the clinical impact of low MICs and subtle phenotypes associated with chromosomal integration. However, these cases are not notably distinct from other descriptions of infection with bla KPC -positive K. pneumoniae, with all infected patients ultimately responding to therapy (29). Interestingly, CAV1217 was phenotypically susceptible; however, despite 48 h of meropenem treatment with an initial Foley exchange, the patient had persistent symptoms and no change in bacterial burden but was ultimately treated successfully with tigecycline (Table 3).
This work also highlights that the frequency of chromosomal integration of genes of drug resistance is largely unknown and would be missed by most methods. Even with WGS, sophisticated focused analysis was required to identify chromosomal integration when a plasmid copy was also present. Currently, bla KPC is carried within a replicative transposon, and although the in vitro transposition frequency of this element is relatively high, the real-world rate is unknown (8). Although the frequency of chromosomal integration cannot be fully assessed without additional study, this remains an intriguing preliminary finding from prospectively collected isolates in a clinical setting.
In summary, we present five independent examples of chromosomal integration of Tn4401-bla KPC in Klebsiella spp. from six patients among a comprehensively characterized outbreak, indicating that chromosomal integration is not an infrequent event. We also demonstrate that chromosomal integration often may be overlooked even with WGS, especially when multiple Tn4401 copies are present. In the clinical microbiology laboratory, phenotypic tests for carbapenem resistance/carbapenemase production were variable and could also miss these cases. Thus, chromosomal integration of bla KPC was relatively common, was difficult to detect, and may facilitate stable inheritance of bla KPC , representing an evolutionary adaptation with potential implications for surveillance and treatment.

MATERIALS AND METHODS
Isolate collection, characterization, and selection for whole-genome sequencing. We previously described short-read (Illumina) sequencing of 281 bla KPC -positive Enterobacteriaceae isolates, from 182 patients, collected between August 2007 and December 2012 in the University of Virginia Health System (18). These isolates comprised 62 distinct strains, defined on the basis of phylogenetic clustering, using cutoffs of ϳ500 single-nucleotide variants (SNVs). Long-read (PacBio) sequencing was carried out on a random subset of 17 isolates (18). For the work presented here, an additional six isolates underwent long-read sequencing with PacBio technology as previously described (30).
Sequence analysis. We used three methods to identify chromosomally integrated Tn4401/bla KPC . First, we used a de novo assembly approach. Illumina reads from each isolate were assembled using Velvet and VelvetOptimiser (31). We then queried these with BLASTn, using 400 bp of sequence from each end of Tn4401 in order to identify contigs containing possible Tn4401 junction sequences. Isolates were evaluable if Ն400 bp of Tn4401 sequence plus Ն400 bp of flanking sequence were present on a single contig for at least one of the two Tn4401 ends. For evaluable isolates, Tn4401 flanking sequences were first compared to known plasmid flanking sequences from previously performed long-read sequencing of a subset of the 281 isolates (18). For isolates that did not show a match to any of these plasmid references, Tn4401 flanking sequences were used as BLASTn queries against NCBI's nucleotide database, with homology to chromosomal sequences taken as evidence for a likely chromosomal location.
Second, we used a mapping approach. For each of the 281 isolates, Illumina reads were mapped to a reference consisting of a species-specific chromosome (Table 1) plus the pKPC_UVA01 Tn4401b-1 sequence (18). For five isolates, there was no species-specific reference available, and we used a closely related species instead (Table 1). Mapping was performed using bwa mem version 0.7.12-r1039 (32) with default parameters, and the output was filtered to remove supplementary alignments. For read pairs where one read mapped within 1 kb of either end of the Tn4401 sequence and the other read mapped to the chromosome, the corresponding position of mapping in the chromosome was extracted from the bam file. The length of the chromosome was divided into 1-kb nonoverlapping windows, and Ն10 reads within a single window were taken as evidence for chromosomal integration of Tn4401.
Third, we used long-read PacBio sequencing (18,30), fully resolving the genetic flanking structures around Tn4401, in order to validate the findings of putative chromosomal integration from the short-read analyses described above. In all cases where multiple isolates were identified from the same patient using the short-read approaches described above, these represented the same strain and chromosomal locus. Therefore, we performed long-read sequencing on the earliest isolate identified from each of the eight patients, with two exceptions (i.e., six isolates in total). We did not perform long-read sequencing on CAV1351, as a later isolate from the same patient (CAV1392) was previously PacBio sequenced as part of another study (18), or on CAV1518, as this isolate was demonstrated not to have a chromosomal integration of bla KPC by further analysis of Illumina data (see Results).
For Tn4401 relative coverage estimation, we mapped reads to a two-contig reference (Tn4401 contig plus chromosome contig) as described above for the mapping approach. The average coverage of Tn4401 relative to the chromosome was calculated as (number of reads mapping to Tn4401/length of Tn4401 contig) ϫ (length of chromosome contig/number of reads mapping to chromosome).
For determination of 5-bp sequences flanking Tn4401 in ST340 isolates, reads were mapped to the Tn4401b-1 reference (18) using bwa mem as described above. For each read mapped to the start/end of Tn4401b-1 with Ն5 bp flanking the start/end position, the 5-bp region immediately adjacent to the start/end position was extracted. Five-base-pair sequences with Ͻ10 counts were excluded to account for sequencing errors. Determination of 8-bp sequences flanking the 16-kb composite region was performed similarly, except that the CAV1417 genome was used as a reference, with reads mapping to Ն30 bp of the end of the 16-kb region within this extracted, and 8-bp flanking sequences were determined. Because the IS15-ΔI sequence exists as multiple copies in the ST340 isolates, it is not possible to determine flanking sequences using Illumina data for this end of the 16-kb region (Fig. 1, left side). Therefore, this analysis was only performed for the other (non-IS15-ΔI) end (Fig. 1, right side).
Analysis of porin genes was performed using tBLASTn comparisons against the de novo assembly of each patient's earliest isolate, with translated ompK35 and ompK36 K. pneumoniae reference sequences (AJ011501 and JX291114, respectively) as queries. Reading frame disruptions were reported as likely loss-of-function mutations.
Expression of bla KPC . Total RNA was extracted using the RNeasy minikit (Qiagen, GmBH, Hilden, Germany) and underwent column DNase I digestion (Qiagen) as recommended by the manufacturer. cDNA was synthesized from RNA in accordance with the manufacturer's instructions (qScript cDNA supermix; Quanta Biosciences, Gaithersburg, MD). Twenty nanograms of RNA equivalent DNA was used in triplicate. Quantitative reverse transcription-PCR (qRT-PCR) was performed using SsoFast EvaGreen supermix (Bio-Rad Laboratories, Hercules, CA), template DNA, and 500 nM each primer RT-KPC-F/RT-KPC-R, K. oxytoca rpoB-F/rpob-R, or K. pneumoniae rpoB-F/rpoB-R. Relative quantification of gene expression was determined using the averaged cycle threshold (C T ) values for each isolate using the Pfaffl method (33). This equation uses an expression ratio to normalize the expression levels of bla KPC to the transcriptional level of the constitutively expressed rpoB gene.
Clinical microbiology. Enterobacteriaceae cultured from clinical (August 2007 to December 2012) and surveillance (starting in April 2009) specimens prospectively underwent carbapenemase phenotypic testing using the modified Hodge test (August 2007 to June 2008) or the indirect carbapenemase test (July 2008 to December 2012). These tests were repeated for all isolates from frozen subculture (23). The earliest available isolate with chromosomal integration from each patient underwent multiple modes of phenotypic testing, including the VITEK 2 system using a GN70 card (bioMérieux, Durham, NC), disc diffusion, Etest (bioMérieux, Durham, NC), and broth microdilution of meropenem to determine susceptibilities by following the CLSI's and manufacturers' guidelines (34).
Clinical characteristics. Electronic medical records were reviewed for treatment, clinical outcome, location and timing of transfer, and exposure to other patients with bla KPC -producing Enterobacteriaceae (approved by IRB 13558).