Exploring the Global Spread of Klebsiella grimontii Isolates Possessing bla_VIM-1 and mcr-9

ABSTRACT

The spread of plasmid-mediated carbapenemases within Klebsiella oxytoca is well-documented. In contrast, data concerning the closely related species Klebsiella grimontii are scarce. In fact, despite the recent report of the first bla_KPC-2-producing K. grimontii, nothing is known about its clonality and antibiotic resistance patterns. In a retrospective search in our collection, we identified 2 bla_VIM-positive K. oxytoca strains. Whole-genome sequencing with both Illumina and Nanopore indicated that our strains actually belonged to K. grimontii and were of sequence type 172 (ST172) and ST189. Moreover, the two strains were associated with 297-kb IncHI2/HI2A-pST1 and 90.6-kb IncFII(Yp) plasmids carrying bla_VIM-1 together with mcr-9 and bla_VIM-1, respectively. In the IncHI2/HI2A plasmid, bla_VIM-1 was located in a class 1 integron (In110), while mcr-9 was associated with the qseC-qseB-like regulatory elements. Overall, this plasmid was shown to be very similar to those carried by other Enterobacterales isolated from food and animal sources (e.g., Salmonella and Enterobacter spp. detected in Germany and Egypt). The IncFII(Yp) plasmid was unique, and its bla_VIM-1 region was associated with a rare integron (In1373). Mapping of In1373 indicated a possible origin in Austria from an Enterobacter hormaechei carrying a highly similar plasmid. Core-genome phylogenies indicated that the ST172 K. grimontii belonged to a clone of identical Swedish and Swiss strains (≤15 single nucleotide variants [SNVs] to each other), whereas the ST189 strain was sporadic. Surveillance of carbapenemase-producing K. oxytoca strains should be reinforced to detect and prevent the dissemination of new species belonging to the Klebsiella genus.

INTRODUCTION

The worldwide spread of Klebsiella pneumoniae isolates producing KPC-, OXA-48-like-, NDM-, and less frequently VIM-type carbapenemases represent a public health concern (1, 2). Infections due to these pathogens are difficult to treat because last-generation cephalosporins and carbapenems are not useful, and resistance to other families of antibiotics is common. In particular, resistance to the last-resort antibiotics polymyxins is frequently observed under treatment in the hospital setting, and it is usually due to chromosomal mechanisms (3); more rarely, strains expressing the plasmid-mediated mcr genes have been described (4).

In this scenario, the carbapenemase-producing (CP) Klebsiella oxytoca isolates are also emerging as potential life-threatening pathogens. In the nosocomial setting, sporadic CP K. oxytoca can be encountered, and two outbreaks due to the KPC or OXA-48 producers have also been described (5–7). In contrast, only one mcr-1-possessing K. oxytoca strain has been isolated so far from humans (8). However, correct identification of K. oxytoca isolates can be difficult and, therefore, routine typing methods (e.g., the matrix-assisted laser desorption ionization–time of flight mass spectrometry [MALDI-TOF MS]) can be unreliable due to the lack of reference spectra (9).

To date, this species can be divided into 6 closely related ones as follows: K. oxytoca (sensu stricto), Klebsiella michiganensis, Klebsiella grimontii, Klebsiella pasteurii, Klebsiella huaxiensis, and Klebsiella spallanzanii. These species can be correctly differentiated by their gyrA, rpoB, or their chromosomal β-lactamase gene family bla_OXY (9). For Klebsiella grimontii, carrier of the bla_OXY-6 gene family, little is known about its potential to become an emerging pathogen (10). Yet, in 2017, the first bla_KPC-2-possessing K. grimontii strain recovered from human sputum was reported in China (11). Thus, this event highlighted the need to surveil this species more closely.

In this work, we describe two unique K. grimontii isolates found in Switzerland that possess bla_VIM-1 and bla_VIM-1 together with mcr-9, respectively. We performed short- and long-read whole-genome sequencing (WGS) approaches to generate complete de novo assemblies and characterize plasmids. Finally, to get an overview of the clonality and the spread of carbapenem resistance genes in K. grimontii, we performed core genome analyses using publicly available K. grimontii genomes.

RESULTS AND DISCUSSION

With the aim to detect and characterize new CP Klebsiella spp. not belonging to the K. pneumoniae species, we retrospectively inspected our collection of strains spanning 2011 to 2020. As a result, three bla_VIM-possessing K. oxytoca clinical isolates (i.e., BD-50-Km, 2481359, and 5512.56) were retrieved. These isolates were initially identified by using MALDI-TOF MS and characterized with the CT103XL microarray (12). Strain BD-50-Km was actually a bla_VIM-1-possessing K. michiganensis and has been recently described (13). In the present work, we further characterized strains 2481359 and 5512.56 by implementing state-of-the-art WGS methods.

Isolation, species identification, and resistome.

Strain 2481359 was isolated on October 2015 from both urine and blood cultures of a 67-year-old Swiss male patient suffering lymphoma, while strain 5512.56 was detected in August 2014 from a rectal swab of a 58-year-old Romanian male patient coming to Switzerland from a medical center based in Austria. WGS confirmed that both strains belonged to the K. grimontii species rather than K. oxytoca and carried bla_VIM-1 along with other antimicrobial resistance genes (ARGs).

The complete genome assembly of 2481359 resulted in a 5.9 Mb chromosome that possessed bla_OXY-6-2 and oqxB; 3 plasmids were also found. The bla_VIM-1 was carried by a 297.3-kb IncHI2/HI2A plasmid (p2481359-1-VIM1-MCR9) together with bla_ACC-1, mcr-9, aac(6′)-Ib-cr [also named aac(6′)-Ib3], aadA1, aph(3′′)-Ib, aph(6)-Id, catA1, and 2 copies of sul1 (see details below). The second plasmid (p2481359-2; 75.7 kb) was of IncFII(pCRY) and harbored the bla_OXA-10, bla_LAP-2, aadA1, arr-2, cmlA1, dfrA14, and qnrS1 ARGs. Notably, bla_OXA-10 was associated with a class 1 integron called In467 (intI1-arr-2-cmlA1-bla_OXA-10-aadA1) first identified in Pseudomonas aeruginosa strain paeG18 (GenBank accession number EU886979). The last plasmid (p2481359-3; 17.6 kb, ColRNAI) did not possess ARGs.

The genome assembly of K. grimontii 5512.56 resulted in a chromosome (5.7 Mb) carrying bla_OXY-6-3 and oqxB; 6 plasmids were also observed. Most of the ARGs were associated with the novel IncFII(Yp) plasmid p5512.56-3-VIM1 (90.6 kb), which carried bla_VIM-1, bla_OXA-9, aac(6′)-Ib-cr, and sul1 (see details below). Another IncFII(Yp) plasmid (p5512.56-2-CTXM14; 92.9 kb) harbored the bla_CTX-M-14 extended-spectrum β-lactamase (ESBL) gene associated with ISEcp1 and IS903B flanking elements (i.e., ISEcp1-bla_CTX-M-14-IS903B). This transposable element is well-documented (14) and can be found in the plasmids of multiple Klebsiella spp. or the chromosomes of Serratia marcescens (data not shown). The rest of the following plasmids carried by strain 5512.56 did not contain ARGs: p5512.56-1 [232.1 kb, IncFIB(K)/FII(K)], p5512.56-4 (9.5 kb, Col440I), p5512.56-5 (3.3 kb, ColRNAI), p5512.56-6 [2.7 kb, Col(MGD2)], and p5512.56-7 (1.8 kb, not typeable).

IncHI2/HI2A bla_VIM-1- and mcr-9-carrying plasmid (p2481359-1-VIM1-MCR9).

The bla_VIM-1- and mcr-9-carrying p2481359-1-VIM1-MCR9 plasmid from K. grimontii 2481359 was of plasmid sequence type 1 (pST1) and shared the highly conserved backbone of the archetype R478 plasmid from S. marcescens (Fig. 1A). The bla_VIM-1 was located in a class 1 integron named In110 (intI1-bla_VIM-1-aac(6′)-Ib-cr-aadA1) that was first described in Italy from a P. aeruginosa isolate 67MG (GenBank accession number AJ969234). The In110 was controlled by a hybrid Pc (P1) promoter (−35, TTGACA [strong]; −10, TAAGCT [weak]) that explains the observation of low MICs for carbapenems (Table 1) (15, 16). Moreover, In110 was flanked by a complex region of insertion elements (IS). In particular, IS1326 was located upstream of the In110. This IS codes for a Tn21-like transposon, which was previously identified as an important element for the dissemination of bla_VIM-1 within livestock and food in Germany (17).

FIG 1.

Circular BLASTn comparisons of K. grimontii bla_VIM-1-carrying plasmids against most similar plasmids. (A) Comparison of p2481359-1-VIM1-MCR9 against the 5 most similar plasmid sequences and the IncHI2/HI2A (pST1) reference plasmid R478. (B) Comparison of p5512.56-3-VIM1 against the 4 most similar plasmid sequences, the In1373 full integron sequence, and the reference IncFII(Yp) plasmid MT. The CDS/genes and IS elements of interest are represented by colored arrows (red, ARGs; light green, intI1; yellow, qacEΔ1; dark green, IS elements; light blue, replicon(s); dark blue, other), with corresponding annotations (red, bla genes; blue, mcr-9); IS annotations are listed in Table S4 in the supplemental material. The annotation tra corresponds to the bp position as follows: (A) tra 1 (200 to 39,500) and tra 2 (198,000 to 213,357); (B) tra (54,148 to 84,344]. In the key below the plasmids, we show GenBank accession number, plasmid name (or integron name and description), pST, plasmid size, bla, mcr-9 (if present), year of isolation, country of origin, isolation source, and the species of isolation.

TABLE 1.

Phenotypic characterization of the bla_VIM-1-possessing K. grimontii strains 2481359 and 5512.56 found in Switzerland

Antibiotic	MIC value (μg/ml)a
	K. grimontii 2481359	K. grimontii 5512.56
Piperacillin-tazobactam	>64/4 (R)	>64/4 (R)
Ticarcillin-clavulanate	>128/2 (R)	>128/2 (R)
Cefotaxime	32 (R)	>32 (R)
Cefotaxime-clavulanate	32/4 (NA)	>64/4 (NA)
Ceftazidime	>16 (R)	>16 (R)
Ceftazidime-clavulanate	128/4 (NA)	>128/4 (NA)
Cefepime	8 (R)	>16 (R)
Aztreonam	≤2 (S)	16 (R)
Imipenem	1 (S)	>8 (R)
Meropenem	≤1 (S)	>8 (R)
Doripenem	0.5 (NA)	>2 (NA)
Ertapenem	≤0.25 (S)	>4 (R)
Gentamicin	≤1 (S)	≤1 (S)
Tobramycin	2 (S)	4 (R)
Amikacin	≤4 (S)	≤4 (S)
Ciprofloxacin	>2 (R)	2 (R)
Levofloxacin	2 (R)	2 (R)
Doxycycline	≤2 (NA)	16 (NA)
Minocycline	≤2 (NA)	16 (NA)
Tigecycline	≤0.25 (S)	1 (R)
Trimethoprim/sulfamethoxazole	>4/76 (R)	1/19 (S)
Colistin	≤0.25 (S)	≤0.25 (S)
Polymyxin B	≤0.25 (S)	≤0.25 (S)

^aMICs were obtained with microdilution Sensititre panels ESB1F and GNX2F and interpreted according to the EUCAST 2019 criteria (v9.0) for Enterobacterales. R, resistant; S, susceptible; NA, not available.

Interestingly, though p2481359-1-VIM1-MCR9 carried mcr-9, its bacterial host remained susceptible to colistin (MIC, ≤0.25 μg/ml) (Table 1). The upstream genetic environment of mcr-9 was composed of IS903B-like elements and downstream of wbuC followed by the qseC-like and qseB-like regulatory system (Fig. 1A). This two-component regulatory system has been shown to be important to mediate the expression of mcr-9 with subinhibitory concentrations of colistin (18, 19). When grown under colistin exposure, K. grimontii 2481359 showed a robust resistant phenotype to polymyxins (e.g., colistin; MIC, >4 μg/ml) (see Table S1 in the supplemental material). However, this phenotype was not reversible and not due to the induction of mcr-9. In fact, using a PCR/sequencing approach, we noted that the mgrB gene was inactivated by the insertion of an ISEc36 (see Fig. S5 in the supplemental material), an element harbored by both chromosome and p2481359-1-VIM1-MCR9. Such mechanisms of colistin resistance are well described in other Klebsiella spp. but not yet in K. grimontii (3).

Because in strain 2481359 the mcr-9 is not expressed, detection of this ARG can be challenging (20). For instance, the National Reference Laboratory for Salmonella in Germany recently identified bla_VIM-1-positive Salmonella spp. and Enterobacter cloacae strains in pigs and pork meat that actually also harbored mcr-9. Nevertheless, the presence of mcr genes was neither acknowledged nor was reduced susceptibility to colistin shown (17, 21). Therefore, mcr-9 may silently disseminate worldwide among different Enterobacterales, especially via the spread of similar IncHI2 plasmids (22). However, unlike mcr-1 where reports have shown that removal of an IS disrupting mcr-1 can rescue its colistin resistance phenotype (23, 24), it is not known if an mcr-9 lacking a functional qseC-qseB system can become expressed (e.g., via IS or recombination). Therefore, further studies are needed to address this concerning issue.

Remarkably, p2481359-1-VIM1-MCR9 was highly similar (100% identity, 99% query coverage) to the bla_VIM-1-/mcr-9-carrying plasmid pSE15-SA01028 from a German Salmonella enterica serovar Infantis found in 2015 and mentioned above (21). Moreover, in 2017, this plasmid was linked to bla_VIM-1-producing Enterobacter cloacae and S. enterica serovar Goldcoast/Infantis isolates from a German pig farm (17). Other bla_VIM-1-possessing plasmids highly similar to ours were pEC17-AB02384 from an Escherichia coli strain isolated from the colon content of a fattening pig at the slaughterhouse in Germany and pMS-37a from an Enterobacter hormaechei strain that was recovered from a beef burger in Egypt (100 to 99.97% identity; 93 to 82% query coverage) (20, 25) (Fig. 1A). We also note that other non-bla_VIM-1-carrying plasmids of human origin very similar to p2481359-1-VIM1-MCR9 have also been reported, such as the bla_ACC-1-/mcr-9-carrying pEcl4-1 and the bla_SHV-12-/mcr-9-possessing pMCR-SCNJ07 from E. hormaechei (100% to 99.99% identity; 98% to 81% query coverage) (Fig. 1A). Therefore, our results confirm that mcr-9 is frequently carried by similar IncHI2/HI2A-ST1 plasmids; such mobile genetic elements (MGEs) may also acquire further clinically important ARGs, such as the carbapenemase genes (22).

IncFII(Yp) bla_VIM-1-carrying plasmid (p5512.56-3-VIM1).

Strain 5512.56 carried its bla_VIM-1 on a 90.6-kb IncFII(Yp) plasmid named p5512.56-3-VIM1. Its replicon sequence showed 86.2% identity with the reference FII(Yp) replicon of the plasmid MT of Yersinia pestis Pestoides F. p5512.56-3-VIM1 was unique and shared mostly the partial backbone to other plasmids (none carrying the bla_VIM-1) containing the FII(Yp)/FIB(pB171) or FIB(pHCM2) replicons. The two most similar plasmids—the bla_NDM-1 carriers pKp_Goe_629-2 and pNDM_22ES—shared with p5512.56-3-VIM1 > 99% identity but only 60% query coverage. These two plasmids, and two less like ours (p10164-NDM and unitig_2_pilon), originated from various species hosts mainly from human sources (Fig. 1B).

In p5512.56-3-VIM1, the bla_VIM-1 was associated with a class 1 integron (In1373, bla_VIM-1-aac(6′)-Ib-cr-qacF-cmlB-bla_OXA-9) carrying a strong Pc (P1) promoter (−35, TTGACA [strong]; −10, TAAACT [strong]) that is consistent with the high MICs recorded for the carbapenems (Table 1) (16). Moreover, because the In1373 sequence of p5512.56-3-VIM1 showed 100% identity to the reference described in E. hormaechei subsp. xiangfangensis (GenBank accession number LC224311), we identified the NCBI Sequence Read Archive (SRA) experiment (SRA accession number SRR5939945; strain AZ 597) corresponding to the sequence’s source. Interestingly, a mapping analysis of strain AZ 597 against p5512.56-3-VIM1 revealed a highly related plasmid structure possessing the same ARGs and the same FII(Yp) replicon sequence identity (86.2%) (see Fig. S1 in the supplemental material). We note that strain AZ 597 was identified during a global study of CP Enterobacter spp. in a human urine sample in Austria (26). Therefore, it is possible that other similar plasmids to p5512.56-3-VIM1 exist in Central Europe.

Conjugation.

Despite multiple attempts, conjugation experiments for both of our bla_VIM-1-possessing strains were unsuccessful. This was not expected since both p2481359-1-VIM1-MCR9 and p5512.56-3-VIM1 plasmids possess the propagation-related genes (tra) (Fig. 1). However, our results are, in part, consistent with previous analyses focusing on similar MGEs. In fact, several IncHI2/HI2A plasmids similar to p2481359-1-VIM1-MCR9 and found in S. enterica and E. coli were reported to be nonconjugative (27–29), whereas others were conjugative (25, 30). Other less similar IncHI2/HI2A plasmids suggest transferability by conjugation from E. hormaechei to E. coli: the bla_VIM-1-/mcr-9-possessing pMS-37a and the mcr-9-harboring pMCR-SCNJ07 (Fig. 1A) (19, 20).

With regard to the novel IncFII(Yp) p5512.56-3-VIM1, we note that a similar non-bla_VIM-1-possessing plasmid [pP10164-NDM; also possessing an additional FIB(pB171) replicon] (Fig. 1B) from Leclercia adecarboxylata was conjugative (31). Likewise, in the first carbapenemase producer K. grimontii, the bla_KPC-2-carrying plasmid [possessing two FII(Yp) replicons] was also not self-transmissible (11).

Core-genome analyses.

Our bla_VIM-1-possessing K. grimontii strains 2481359 and 5512.56 belonged to ST172 and ST189, respectively. Among the 155 available genomes of this species (including 26 from Switzerland), the initial core-genome analysis (61% core genome alignment corresponding to 103,122 single nucleotide variants [SNVs] across 155 genomes) identified 8 further K. grimontii isolates that carried carbapenemase genes originating in the United States (n = 1), China (n = 2), Sweden (n = 2), and Switzerland (n = 3). These belonged primarily to ST172, ST215, ST262, and to two undefined STs (ST184/225-like and ST214/261-like) carrying multiple novel alleles (see Fig. S2 in the supplemental material).

To find all closely related genomes to each carbapenemase carrier (independently of the ST), we selected those with >95% identical single nucleotide variants (SNVs) and conducted a high-resolution core-genome analysis (Fig. 2; see also Table S2 in the supplemental material) (76% core genome alignment corresponding to 49,606 SNVs across 29 genomes). We identified 3 families of carbapenemase genes (number of strains) as follows: bla_VIM-1 (n = 7, including ours), bla_KPC-2 (n = 2), and bla_IMP-38 (n = 1).

FIG 2.

Core-genome phylogeny of carbapenemase-carrying K. grimontii and closely related strains. The alignment consisted of 29 strains, of which 10 were positive for carbapenemase genes, 18 closely related strains, and the reference K. grimontii type strain 06D021. For each strain, we show name, source, country, year of isolation, antibiotic resistance genes, plasmid replicons, and STs. Strains carrying carbapenemase genes are shown with a black circle; the strains in this study are underlined; in bold, the carbapenemase genes, mcr-9, and the replicon sequences associated with the carbapenemase genes. The tree scale represents the average number of nucleotide substitutions per site. Shimodaire-Hasegawa-like support values are shown in the tree branches. See Table S3 in the supplemental material for SNVs and identity matrices corresponding to this analysis. (a) Strain names corresponding to the genome assembly or SRA accessions listed in Data Set S1 in the supplemental material. (b, c, d) Annotated as described in the corresponding BioSample metadata (accessed via the genome assembly or SRA accession); NA, not available. (e) A plus (+) superscript sign corresponds to genes that were only identified in the following strains: aadA1 in 2481359, SN1015-66, 402620-12; and ant(3′)-Ia in 804720-15. (f) An asterisk (*) is shown when more than one replicon of the same type was detected. (g) Strains with a nearest ST are represented by an ST-like; for the strains that showed multiple novel allele hits, the nearest STs are shown as “ST/ST-like.”

Strain 5512.56 was a sporadic clone (ST189) and did not share >95% identical SNVs with another strain(s). In contrast, strain 2481359 belonged to a clonal complex (ST172) of strains originating in Sweden and Switzerland. In particular, strains H6 and SN1015-66, originally identified as K. oxytoca during a study in Sweden focusing on the clonal dissemination of CP Enterobacterales in aquatic environments and humans (32), were both only 14 SNVs apart from 2481359 (see Table S3 in the supplemental material). Similarly, strains 402620-12 and 804720-15, identified in the SRA database as originating from Switzerland, were 15 and 11 SNVs different from 2481359, respectively (see Data Set S1, section D, in the supplemental material).

We found that the high similarity between the core genomes of the ST172 Swedish and Swiss strains likely meant that some ARGs (e.g., bla_VIM-1 and mcr-9) and replicon types (e.g., HI2/HI2A) corresponded to a very similar plasmid as in 2481359 (Fig. 2). In fact, a mapping analysis of the ST172 strains against p2481359-1-VIM1-MCR9 revealed identical plasmid structures (see Fig. S3 in the supplemental material). Thus, because of the lack of identical plasmids in other STs or ST172 strains without the HI2/HI2A replicon(s) (Fig. 1 and 2), we speculate that these bla_VIM-1- and/or mcr-9-carrying K. grimontii strains arose from a clonal event rather than through MGE acquisition. This is the opposite situation that we can observe for a non-CP ST184/225-like clone mainly spreading in the United Kingdom (657 to 695 ΔSNVs); one Swiss K. grimontii strain detected in 2017 acquired bla_VIM-1 and other ARGs (Fig. 2). Interestingly, the bla_VIM-1 was associated with the same In1373 described above, but its genetic location (e.g., plasmid) could not be determined based on contig mapping (see Fig. S4 in the supplemental material).

Conclusions.

In this work, we described for the first time bla_VIM-1-possessing K. grimontii isolates. Remarkably, one of the strains (2481359) belonged to a clone (ST172) recovered from human and environmental sources from Sweden and Switzerland. This clone has the potential to spread undetected due to its very low MICs for both carbapenems and colistin. Moreover, 2481359 co-carried bla_VIM-1 and mcr-9 in the same IncHI2/HI2A pST1 plasmid (p24813591-VIM1-MCR9). Recently, very similar plasmids have also been reported, mainly in E. coli, Salmonella, and Enterobacter species strains found in food and food-producing animals in Germany (17, 20, 21, 25). This observation could drive the conclusion that such plasmids may spread via conjugation among different Enterobacterales. However, our conjugation experiments failed to demonstrate this ability and only few similar plasmids demonstrated to be self-conjugative in vitro (25, 30). In this context, we note that IncHI2/HI2A plasmids similar to p24813591-VIM1-MCR9, but possessing only mcr-9, have been reported in multiple hosts and in many countries (22, 33). Therefore, it can be speculated that these MGEs may serve as scaffold for the integration of further carbapenemase-carrying elements, such as the In110 class 1 integron possessing bla_VIM-1.

In contrast to the first isolate, K. grimontii 5512.56 represented a sporadic strain. However, we showed that its class 1 bla_VIM-1-containing In1373 integron could be associated with other similar plasmids, such as the one from Enterobacter hormaechei strain AZ 597 recovered from a urine sample in Austria (the same country where 5512.56 was probably acquired by our patient). Moreover, In1373 was also integrated in a Swiss K. grimontii strain closely related to a non-CP ST184/225-like UK clone cluster.

In conclusion, we showed that K. grimontii can be involved in human and animal infections and colonization along with contamination of food chain and environment; moreover, this species can carry clinically important ARGs. However, information regarding antimicrobial resistance patterns, clonality, and the plasmid(s) background of K. grimontii is still limited, mainly due to its frequent misidentification as K. oxytoca (9). Therefore, further surveys to understand the role of this emerging pathogen are needed. Furthermore, an improvement on the ability to routinely identify K. grimontii is also advised.

MATERIALS AND METHODS

Species identification, microarray, and antimicrobial susceptibility tests.

The initial species identification (ID) was routinely obtained using MALDI-TOF MS (Bruker). Strains were screened with the CT103XL microarray (Check-Points) that is able to detect mcr-1, mcr-2, and the main ESBL, plasmidic AmpCs, and carbapenemase bla genes (12). ID was confirmed by WGS data and the implementation of the Type (Strain) Genome Server (TYGS) (https://tygs.dsmz.de/). Likewise, for all of the genome assemblies used in the final core-genome alignment, the ID was also confirmed with TYGS. Antimicrobial susceptibility tests (ASTs) were performed using the broth microdilution ESB1F and GNX2F Sensititre panels (Thermo Fisher). MICs for antibiotics were interpreted according to the 2019 European Committee on Antimicrobial Susceptibility Testing (EUCAST) criteria (v9.0). Clinical and demographic data were obtained from the laboratory database information system.

Induction of colistin resistance.

Induction experiments were performed as described by Kieffer et al. with modifications (18). Single colonies of strain 2481359 grown in MacConkey agar were inoculated in 5 ml LB under the following conditions: nonsupplemented and supplemented with 0.25 μg/ml, 0.5 μg/ml, or 1 μg/ml of colistin. After overnight incubation at 37°C, each inoculum was diluted to 0.5 McFarland, and both microbroth (cation-adjusted Mueller-Hinton broth II) and disk-diffusion (Mueller-Hinton agar) susceptibility tests were performed in parallel and interpreted as described above. The mgrB gene was analyzed by PCR and Sanger sequencing, implementing primers mgrB_Kg_F (5′-CTTTCGTATTACAGTTAGCCGC-'3) and mgrB_Kg_R (5′-CACCTCAAAGAGAAGGCG-3′) that were designed with Geneious Prime (v2021.1.1).

Conjugation experiments.

Conjugation experiments were performed by filter mating and liquid broth with modifications as previously described (34, 35). The rifampin-resistant E. coli J53 and rifampin- and sodium azide-resistant E. coli J53d-R1 were used as recipient strains for the filter and liquid method, respectively. For both experiments, incubations were done at 25°C and 37°C; filter and liquid matings were done for 5 h and overnight, respectively. Transconjugants were selected on MacConkey agar plates supplemented with ampicillin (100 μg/ml) plus rifampin (100 μg/ml) and the same plates with a disk of imipenem (10 μg).

Database search, curation, and screening.

A comprehensive search for K. grimontii genomes in the NCBI database (https://ncbi.nlm.nih.gov/labs/gquery/; download date, 6 January 2021) resulted in 116 available assemblies. To avoid missing misplaced K. grimontii genomes in other databases due to incorrect ID, the K. oxytoca and K. michiganensis databases were also included, each resulting in 227 and 256 assemblies, respectively. In addition, sequencing experiments of Swiss origin (as per BioSample description; query search, “Klebsiella grimontii Switzerland”) were retrieved from the NCBI SRA and assembled independently (see WGS and analyses section).

All 3 databases and SRA assemblies (overall, n = 153; see Data Set S1 in the supplemental material) were screened in silico for ARGs with AMRfinder v3.9.3 (https://github.com/ncbi/amr) to include only bla_OXY-6 family-positive genomes (the K. grimontii-specific marker [10]) and exclude other closely related species within the K. oxytoca complex (9). Similarly, the databases were screened for ST clones with the mlst v2.19.0 (https://github.com/tseemann/mlst) program using the most recent K. oxytoca MLST alleles (https://pubmlst.org/; download date, 6 February 2021).

WGS and analyses.

WGS was performed with short- and long-read sequencing technologies as previously described with some modifications (36). In short, Illumina sequencing was conducted using a NovaSeq 6000 sequencer with 2 × 150-bp paired-end reads output (for strains 2481359 and 5512.56, the average coverage was 250.98× and 262.28×, while the average read length was of 149.19 bp and 149.15 bp, respectively). Long-reads were generated with a MinION (Oxford Nanopore) sequencer using a rapid barcoding library prep (SQK-RBK004) and a FLO-MIN 106D R9 flow cell (for strains 2481359 and 5512.56, the average coverage was 185.48× and 132.46×, while the average read length was of 6,615.8 bp and 3,668.2 bp, respectively). Short and long reads were preprocessed with Trimmomatic (v0.36) and Porechop (v0.2.4) with default parameters, respectively. The complete genome assemblies were generated with the Unicycler (v0.4.8) hybrid assembly pipeline with default parameters. For the strain 5512.56, an external long-read assembly was initially generated with Flye (v2.8-b1674; with the following parameters: plasmids, trestle, and 5 polish iterations), and then used as input scaffold for the Unicycler assembly. The additional sequencing data retrieved from the NCBI SRA (Data Set S1, section D) and of strain AZ 597 (SRA accession number SRR5939945; see also Fig. S1 in the supplemental material) were preprocessed with Trimmomatic (v0.36) and assembled with Spades (v3.14) with default parameters, respectively.

Gene annotation was performed with the NCBI Prokaryotic Genome Annotation Pipeline, and ISs were manually curated with ISfinder (https://isfinder.biotoul.fr/) (see Table S4 in the supplemental material). The final genomes were analyzed using the following tools of the Center for Genomic Epidemiology (www.genomicepidemiology.org/): PlasmidFinder (v2.1; 50% threshold for minimum percentage identity), ResFinder (v4.1), MLST (v2.0; K. oxytoca scheme), and pMLST (v2.0). The bla_VIM-1-harboring integrons were classified according to INTEGRALL (http://integrall.bio.ua.pt/). Plasmid database searches were done with NCBI BLASTn (default parameters) and with PLSDB (v2020_11_19; https://ccb-microbe.cs.uni-saarland.de/plsdb/) using “Mash distribution” and default parameters. Plasmid alignments were generated with BLAST Ring Image Generator (BRIG; v0.95). Contigs from WGS assemblies were mapped to either p2481359-1-VIM1-MCR9 or p5512.56-3-VIM1 with minimap2 (v2.17) set to 5% max divergence; their consensus sequences were generated with Geneious (v11.1.5).

Core-genome analyses.

Two independent core-genome analyses were performed. The first one, an initial “global” core-genome analysis, was done to identify closely related strains to ours that shared the same ST (or not) and to identify all carbapenemase gene-carrying strains, and a final one, which included all strains that shared >95% identical SNVs relative to all strains carrying a carbapenemase gene.

In short, the first core-genome alignment consisted of all K. grimontii genomes (n = 155, including our two strains), and the second consisted of 10 carbapenemase gene carriers and their 18 closely related strains (n = 28). Both core-genome alignments were performed with Parsnp v1.2 (https://github.com/marbl/parsnp) with recombination filter enabled, and the K. grimontii type strain 06D021 as reference genome, as previously described with modifications (36). Trees were visualized with FigTree (v1.4.4) and annotated with Inkscape (v1.0).

Data availability.

The complete genome assemblies of strains 2481359 and 5512.56 are deposited in GenBank (CP067380 to CP067383 and CP067384 to CP067391, respectively) under BioProject PRJNA689715.