Characterization of Carbapenemase-Producing Klebsiella oxytoca in Spain, 2016–2017

There is little information about carbapenemase-producing (CP) Klebsiella oxytoca, an important nosocomial pathogen. We characterized CP K. oxytoca isolates collected from different Spanish hospitals between January 2016 and October 2017.

ABSTRACT

There is little information about carbapenemase-producing (CP) Klebsiella oxytoca, an important nosocomial pathogen. We characterized CP K. oxytoca isolates collected from different Spanish hospitals between January 2016 and October 2017. During the study period, 139 nonduplicate CP K. oxytoca isolates were identified; of these, 80 were studied in detail. Carbapenemase and extended-spectrum β-lactamase genes were identified by PCR and sequencing. Genetic relatedness was studied by pulsed-field gel electrophoresis (PFGE). Whole-genome sequencing (WGS), carried out on 12 representative isolates, was used to identify the resistome, to elucidate the phylogeny, and to determine the plasmids harboring carbapenemase genes. Forty-eight (60%) isolates produced VIM-1, 30 (37.5%) produced OXA-48, 3 (3.7%) produced KPC-2, 2 (2.5%) produced KPC-3, and 1 (1.2%) produced NDM-1; 4 isolates coproduced two carbapenemases. By PFGE, 69 patterns were obtained from the 80 CP K. oxytoca isolates, and four well-defined clusters were detected: cluster 1 consisted of 11 OXA-48-producing isolates, and the other three clusters included VIM-1-producing isolates (5, 3, and 3 isolates, respectively). In the 12 sequenced isolates, the average number of acquired resistance genes was significantly higher in VIM-1-producing isolates (10.8) than in OXA-48-producing isolates (2.3). All 12 isolates had chromosomally encoded genes of the bla_OXY-2 genotype, and by multilocus sequence typing, most belonged to sequence type 2 (ST2). Carbapenemase genes were carried by IncL, IncHI2, IncFII, IncN, IncC, and IncP6 plasmid types. The emergence of CP K. oxytoca was principally due to the spread of VIM-1- and OXA-48-producing isolates in which VIM-1- and OXA-48 were carried by IncL, IncHI2, IncFII, and IncN plasmids. ST2 and the genotype bla_OXY-2 predominated among the 12 sequenced isolates.

INTRODUCTION

The prevalence of multidrug-resistant carbapenemase-producing Enterobacteriaceae (CPE), a major public health threat, continues to increase on a global level and is associated with significant morbidity and mortality (1). Frequently, patients with CPE infections cannot be treated with effective antibiotics because of the dearth of alternative drugs (2). Currently, the epidemiological situation regarding CPE in Spain is considered endemic, mainly due to carbapenemase-producing (CP) Klebsiella pneumoniae. However, carbapenemase production is also increasing in other Enterobacteriaceae species, such as Escherichia coli, Citrobacter spp., Enterobacter spp., and Serratia marcescens (3–5).

Klebsiella oxytoca has been recognized as an opportunistic pathogen causing health care-associated infections, such as urinary and respiratory tract infections and sepsis, primarily in immunocompromised and debilitated patients admitted to intensive care units. However, little information about carbapenemase-producing (CP) K. oxytoca is available, since only a few individual cases or nosocomial outbreaks have been reported (6–9).

The aim of this study was to gain insight into the microbiological features and molecular epidemiology of CP K. oxytoca isolates submitted to the Spanish National Reference Laboratory for Antibiotic Resistance.

RESULTS AND DISCUSSION

Bacterial isolates and carbapenemase types.

Between January 2016 and September 2017, 139 nonduplicate CP K. oxytoca isolates, submitted from 22 hospitals located in 9 different Spanish provinces, were identified (5.2% of the 2,673 CPE identified by the reference laboratory during the same period of time). Of them, 80 (57.5%) isolates were selected for further studies (see Table S1 in the supplemental material); this selection excluded 59 isolates obtained from rectal swab samples collected in three hospitals with nosocomial outbreaks.

Sixty-one isolates produced clinical infections, as follows: 28 (45.9%) urinary tract infections, 11 cases of bacteremia (18%), 10 (16.4%) respiratory tract infections, 6 (9.8%) wound infections, and 6 (9.8%) abscesses. The patients infected with CP K. oxytoca isolates were mainly males (65.6%) aged 18 to 65 years (59%).

Of the 80 CP K. oxytoca isolates, 48 (60%) produced VIM-1, 30 (37.5%) produced OXA-48, 3 (3.7%) produced KPC-2, 2 (2.5%) produced KPC-3, and 1 (1.2%) produced NDM-1. Four isolates coproduced two different carbapenemases: two produced VIM-1 plus KPC-2, one produced VIM-1 plus OXA-48, and one produced VIM-1 plus KPC-3. Carba NP tests yielded positive results for all 80 CP isolates.

As previously described in other Enterobacteriaceae species, such as Citrobacter freundii, Enterobacter spp., and Serratia marcescens (3–5), VIM-1 was the most frequent carbapenemase produced by CP K. oxytoca in this study. However, in K. pneumoniae and Escherichia coli, the most frequent carbapenemases detected in Europe are KPC, OXA-48, and NDM (2). Recently, the clonal spread of an IMP-8-producing K. oxytoca clone was described in a Spanish hospital (9).

Five (6.25%) CP K. oxytoca isolates also coproduced extended-spectrum β-lactamases (ESBLs), as follows: two VIM-1-producing isolates (2.5%) also produced CTX-M-9 plus SHV-12, one VIM-1- and OXA-48-producing isolate and one OXA-48-producing isolate (2.5%) also produced CTX-M-9, and one VIM-1-producing isolate (1.25%) also produced CTX-M-14. The NDM-1-producing K. oxytoca isolate coproduced the plasmid-mediated AmpC (p-AmpC) CMY-4.

Antibiotic susceptibility.

The antibiotic susceptibilities of all CP K. oxytoca isolates are detailed in Table 1 and Table S2. In comparison with the VIM-1-producing isolates, the OXA-48-producing isolates were significantly more susceptible to gentamicin, tobramycin, cotrimoxazole, cefotaxime, and ceftazidime (P < 0.05) (Table 1).

TABLE 1.

Antibiotic susceptibility of 80 carbapenemase-producing Klebsiella oxytoca isolates^a

Antibiotic	% of isolates S			MIC (mg/liter) for all isolates
	Total isolates (n = 80)	VIM-1-producing isolates (n = 44)	OXA-48-producing isolates (n = 29)	50%	90%	Range
Colistin	97.5	95.5	100	0.5	1	0.5 to >8
Amikacin	86.2	81.8	96.5	4	16	4 to >32
Gentamicin	60	34.1	96.5	0.5	>8	≤2 to >8
Aztreonam	40	47.7	37.9	>32	>32	≤0.5 to >32
Cef-Avi	38.7	0	100	>16	>16	≤0.5 to >16
Tobramycin	38.7	4.5	89.6	8	>8	≤2 to >8
Meropenem	36.2	25	62.1	4	>16	0.5 to >16
Co-trimoxazole	35	6.8	79.3	>8	>8	≤1 to >8
Tigecycline	21.2	22.7	17.2	1	4	0.5 to 4
Ciprofloxacin	18.7	22.7	17.2	>2	>2	≤0.06 to >2
Imipenem	11.2	4.5	24.1	8	>16	1 to >16
Ceftazidime	11.25	0	31	>16	>16	≤1 to >16
Ertapenem	8.7	15.9	0	>2	>2	≤0.5 to >2
Cefotaxime	3.7	0	10.3	>8	>8	≤1 to >8
Pip-Taz	0	0	0	>32/4	>32/4	>32/4
Amox-Clav	0	0	0	>64/2	>64/2	>64/2 to >64/2

^aAntibiotic susceptibility was determined by the microdilution method according to EUCAST clinical breakpoints. Data are from reference 12. S, susceptible; Cef-Avi, ceftazidime-avibactam; Pip-Taz, piperacillin-tazobactam; Amox-Clav, amoxicillin-clavulanic acid.

The rates of susceptibility to carbapenems were 36.2% to meropenem, 11.2% to imipenem, and 8.7% to ertapenem (Table 1); all isolates were nonsusceptible to at least one carbapenem antibiotic. These data suggest that ertapenem has the highest sensitivity for the detection of CP K. oxytoca strains, although specificity could not be determined. However, susceptibility to carbapenems varied highly in relation to the carbapenemase types, as follows: the rates of ertapenem, imipenem, and meropenem susceptibility were 0%, 24.1%, and 62.1%, respectively, in OXA-48 producers and 15.9%, 4.5%, and 25%, respectively, in VIM-1 producers. These differences between carbapenemases have been previously described in other species, such as Citrobacter freundii and Escherichia coli (3, 4).

Of the 28 OXA-48-producing isolates that were negative for other carbapenemases, ESBLs, or p-AmpCs, 11 were cefotaxime susceptible but 17 (60.7%) were cefotaxime resistant. Overproduction of chromosomal OXY enzymes can result in reduced susceptibility or resistance to penicillin-inhibitor combinations, cefuroxime, cefotaxime, and aztreonam and has been classically observed in 10% to 20% of clinical isolates of K. oxytoca (10).

Forty-nine isolates (61.3%) were resistant to ceftazidime-avibactam, but all of them produced metallo-β-lactamases. Only two isolates were colistin resistant but were negative for mcr genes. The most frequently described mechanisms of chromosomal resistance to colistin in K. pneumoniae are mutations caused by insertional inactivation or deletion of the mgrB gene and upregulated transcription of phoP, phoQ, and pmrK (which is part of the pmrHFIJKLM operon) (11).

The percentages of susceptibility to ertapenem, imipenem, and meropenem, according to EUCAST breakpoints (12), varied significantly between VIM-1-producing isolates (15.9%, 4.5%, and 25%, respectively) and OXA-48-producing isolates (0%, 24.1%, and 62.1%, respectively) (P < 0.05) (Table 1). However, all isolates had meropenem MICs of >0.25 mg/liter and therefore would be suspected of producing carbapenemases according to the screening cutoff values proposed by EUCAST (13).

Molecular epidemiology.

Pulsed-field gel electrophoresis (PFGE) analysis revealed a high degree of genetic diversity, as 69 different PFGE patterns were obtained from the 80 CP K. oxytoca isolates analyzed (simple diversity index, 86.2%). However, four well-defined clusters with more than two isolates each were detected when a genetic linkage of ≥85% was considered (Fig. S1). The largest cluster (cluster 1) consisted of 11 OXA-48-producing isolates submitted from two hospitals (which sent 9 and 2 isolates, respectively) located in the same province. The other three clusters consisted of five (cluster 2), three (cluster 3), and three (cluster 4) VIM-1-producing isolates, and each of these clusters came from individual hospitals. In addition, five clusters of two isolates each were also detected (Fig. S1).

These PFGE results suggest polyclonal dissemination, since 46 of the 80 CP K. oxytoca isolates showed nonrelated PFGE patterns. However, specific clonal disseminations of CP K. oxytoca either within the same hospital or between different close hospitals were detected.

Resistome of CP K. oxytoca determined by WGS analysis.

Whole-genome sequencing (WGS) was performed on 12 selected isolates representing (i) the four main clusters detected by PFGE (five isolates), (ii) the clusters formed by pairs of isolates (four isolates), and (iii) three single isolates: one producing NDM-1, one producing VIM-1 plus KPC-2, and one producing VIM-1 plus KPC-3. According to the PFGE results, these 12 isolates are representative of 33 of the 80 isolates studied (11 isolates of cluster 1, 5 of cluster 2, 3 of cluster 3, 3 of cluster 4, 4 pairs of isolates, and 3 individual isolates).

In the 12 sequenced isolates, the average number of acquired resistance genes was 9.2 (range, 1 to 16), being much higher in the 5 VIM-1-producing isolates (average, 10.8; range, 6 to 15) than in the 3 OXA-48-producing isolates (average, 2.3; range, 1 to 3). A summary of the main antibiotic resistance genes detected is provided in Table 2. The isolate with the highest number of acquired resistance genes (n = 16) was the NDM-1-producing K. oxytoca isolate, which had several genes encoding resistance to β-lactams, aminoglycosides, sulfonamides, quinolones, chloramphenicol, macrolides, rifampin, tetracycline, and trimethoprim; in addition, the genome of this isolate encoded the methyltransferase gene armA, which is of special relevance since it confers high levels of resistance to all clinically important aminoglycosides (14). ArmA has been associated with K. oxytoca in Japan and China (15, 16); however, ArmA has only sporadically been recovered from Enterobacteriaceae isolates in Spain, and these were mainly obtained from animals (17, 18), but it has not been recovered from K. oxytoca. The association of the bla_NDM-1 and armA genes in the same plasmid has been described previously (19).

TABLE 2.

Main antibiotic resistance genes and other molecular markers detected by WGS of 12 representative CP Klebsiella oxytoca isolates^a

Isolate	PFGE/ST	Carbapenemase gene(s)	Plasmid(s) carrying carbapenemases genes (size [kb])	Gene(s) detected
				blaOXY	Other bla genes	aadA	aac(6′)Ib	arm	ant(2ʺ)-Ia	sul	dhfr	qnr	cat	tet	msr and mph
K7533	C1/ST2	blaOXA-48	IncL (∼63)	blaOXY-2-8	N	aadA1	ND	N	N	ND	dfrA1	N	N	N	N
K0100	C1/ST2	blaOXA-48	IncL (∼63)	blaOXY-2-8	N	aadA1	ND	N	N	ND	dfrA1	N	N	N	N
K9103	C2/ST36	blaVIM-1	IncHI2 (∼250)	blaOXY-2-2	N	aadA1	aac(6′)Ib	N	N	sul1, sul2	dfrA1, dfrB1	N	catA1, catB2	N	msrE, mphE
K9635	C3/ST2	blaVIM-1	IncL (∼70)	blaOXY-2-8	N	aadA1	aac(6′)Ib	N	N	sul1	dfrA1 dfrB1	N	catA1, catB2	N	N
K9599	C4/ST36	blaVIM-1	IncFII (∼70)	blaOXY-2-2	blaTEM-1A	aadA1	aac(6')Ib	N	N	sul1	dfrB1	N	catB2	N	N
K8025	Couple/ST20	blaVIM-1	IncL (∼70)	blaOXY-2-5	N	aadA1	aac(6′)Ib	N	N	sul1	dfrB1	N	catA1, catB2	N	msrE mphE
K8021	Couple/ST2	blaVIM-1	IncHI2 (∼250)	blaOXY-2-8	blaCTX-M-9, blaSHV-12	aadA1, aadA2	aac(6′)Ib	N	ant(2ʺ)-Ia	sul1	dfrA1, dfrA16, dfrB1	qnrA1	catA1, catB2	tetA	N
K7929	Couple/ST141	blaOXA-48	IncL (∼48)	blaOXY-2-2	N	ND	ND	N	N	N	ND	N	N	N	msrE, mphE
K8792	Couple/ST2	blaVIM-1	ND	blaOXY-2-8	N	aadA1	aac(6′)Ib	N	N	sul1	dfrA1, dfrB1	N	catB2	N	N
K9534	NR/ST36	blaVIM-1, blaKPC-3	IncN (∼50) for blaVIM-1; IncFIB (∼103) for blaKPC-3	blaOXY-2-2	blaTEM-1A, blaOXA-9	aadA1	aac(6′)Ib	N	N	sul1	dfrA14, dfrB1	qnrS1	catA1, catB2	N	N
K9455	NR/ST176	blaNDM-1	IncC (∼154)	blaOXY-2-5	blaOXA-10, blaCMY-4	aadA1	aac(6′)Ib	armA	N	sul1, sul2	dfrA14	qnrB32	N	tetA	msrE, mphE
K9682	NR/ST176	blaVIM-1, blaKPC-2	IncL (∼70) for blaVIM-1; IncP6 (∼39) for blaKPC-2	blaOXY-2-5	blaTEM-1A	aadA1	aac(6′)Ib	N	N	sul1	dfrB1	N	catA1, catB2	N	N

^aC1, C2, C3, and C4, representative isolates from PFGE clusters 1, 2, 3, and 4, respectively; couple, representative isolates of PFGE clusters of only two isolates; NR, isolates not related by PFGE to any other isolate; ND, not determined; N, negative.

In addition to carbapenemase genes, the most predominant acquired resistance genes were aadA1 (91.7%), aac(6′)-Ib (75%), sul1 (75%), and catB2 (66.7%).

Concerning the chromosomal bla_OXY genes, all 12 isolates had the bla_OXY-2 genotype, with three variants detected: bla_OXY-2-8 (n = 5), bla_OXY-2-2 (n = 4), and bla_OXY-2-5 (n = 3) (Table 2). Eight different types of OXY enzymes have been described (10, 20). Of the 68 isolates demonstrating nonsusceptibility to 3rd-generation cephalosporins obtained from 14 hospitals across Europe (10), 42 (61.8%) belonged to phylogroup KoII, which includes isolates of the OXY-2 type (20). Interestingly, in this study, KoII isolates were frequently OXY hyperproducers.

Phylogenetic analysis of carbapenem-producing Klebsiella oxytoca by MLST and core genome MLST.

The 12 sequenced isolates belonged to the following types determined by multilocus sequence typing (MLST): sequence type 2 (ST2; n = 5), ST36 (n = 3), ST176 (n = 2), ST20 (n = 1), and ST141 (n = 1) (Table 2; Fig. 1). Of the 12 isolates, 8 (66.7%) had STs belonging to clonal complex 2 (CC2): ST2, ST176 (a double-locus variant [DLV] of ST2), and ST141 (a DLV of ST2). CC2 has been previously described to be the most frequent clonal complex in K. oxytoca (10, 21).

FIG 1.

Minimum-spanning tree of 12 representative carbapenemase-producing K. oxytoca isolates (colored circles). Distance is based on an ad hoc cgMLST of 3,201 genes using the parameters pairwise ignoring missing values and all the K. oxytoca reference sequences available in NCBI. Each circle is named with the MLST type, and colors indicate carbapenemase types; white circles represent control isolates available in NCBI (n = 7) or representative of OXA-48-producing K. oxytoca outbreak isolates from Tunisia (*, n = 3 [unpublished data]).

The genome assemblies of the 12 K. oxytoca isolates were analyzed using a gene-by-gene approach (22), and the allelic distance from the core genome MLST (cgMLST) was visualized in a minimum-spanning tree (Fig. 1). Additionally, the study of single nucleotide polymorphisms (SNPs) was performed as described in Materials and Methods (Fig. S2). ST2 isolates differed by an average of 56 alleles (range, 15 to 88 alleles) and 288 SNPs (range, 35 to 448 SNPs); in ST36 isolates, these differences were 139 alleles (range, 48 to 188 alleles) and 1,013 SNPs (range, 120 to 1,495), and in ST176 isolates, these differences were 134 alleles and 403 SNPs. The most prevalent cgMLST clusters, ST2, ST136, and ST176, included isolates producing different carbapenemase types (Fig. 1).

Characterization of plasmids harboring carbapenemases genes.

In 11 of the 12 K. oxytoca isolates (Table 2), the PlasmidID mapping tool was used for the identification and reconstruction of 13 plasmids harboring carbapenemase genes (7 carrying bla_VIM-1, 3 carrying bla_OXA-48, 2 carrying bla_KPC, and 1 carrying bla_NDM-1). In the remaining isolate, plasmid reconstruction failed.

In the three isolates producing class D carbapenemases, bla_OXA-48 was carried on IncL plasmids, one of ∼63 kb (two isolates) almost identical to plasmid pOXA-48 (GenBank accession number NZ_CP018342.1; average identity, >95%; average coverage percentage, 99.8%) (Fig. S3) and another of ∼48 kb (one isolate) highly similar to the plasmid with GenBank accession number NZ_CP018694.1 (average identity, >95%; average coverage percentage, 94.3%), reconstructed with a low number of contigs (8, 7, and 3 respectively) (Table 2). In all three isolates, bla_OXA-48 was flanked by IS1999 sequences. In the two isolates within cluster 1/ST2 (Fig. 1), bla_OXA-48 was located in a Tn1999.2 transposon (Fig. S3), and in the ST141 isolate, bla_OXA-48 was located in a Tn1999.3 transposon.

In VIM-1-producing isolates, four different plasmid types carrying bla_VIM-1 were identified, as follows (Table 2): one IncL plasmid of ∼70 kb identical to the plasmid with GenBank accession number NZ_CP023419.1 in three isolates (Fig. S4); one IncHI2 plasmid of ∼250 kb very similar to the plasmid with GenBank accession number NZ_CP026661.1 in two isolates (average identity, >95%; average coverage percentage, 81.8%) (Fig. S5), one IncN plasmid of ∼50 kb very similar to the plasmid with GenBank accession number NC_014208.1 (average identity, >95%; average coverage percentage, 90.96%) in one isolate, and one IncFII plasmid of ∼70 kb similar to the plasmid with GenBank accession number NZ_CP023419.1 (average identity, >95%; average coverage percentage, 86.4%) in one isolate.

In Greece, a multiclonal epidemic of VIM-1-producing K. pneumoniae isolates mainly due to the spread IncN plasmids of between 50 and 70 kb was described (23). A conjugative IncHI2 plasmid of ∼300 kb has previously been associated with the dissemination of bla_VIM-1 among genetically diverse Enterobacter cloacae isolates (24). We identified two different genetic environments for bla_VIM-1; in IncN, IncL, and IncFII plasmids, bla_VIM-1 was located in the In624 class 1 integron (24, 25) (Fig. S4), and in IncHI2 plasmids, VIM-1 was linked to another similar class 1 integron that was previously described (26) (Fig. S5). In Spain, In624 has been described in VIM-producing E. cloacae (24) and C. freundii (25) isolates, suggesting an important role in the interspecies transfer of bla_VIM-1.

Two plasmids carrying bla_KPC were identified (Table 2). One of them was an IncP6 plasmid of ∼39 kb identical to the plasmid with GenBank accession number NZ_CP026224.1 with bla_KPC-2 as the sole antimicrobial resistance gene in the plasmid (Fig. S6). The genetic environment of the bla_KPC-2 gene included ISKpn27 and a truncated bla_TEM-1 (Fig. S6) and was the same as that previously detected in the first IncP6 plasmid carrying bla_KPC-2 in K. oxytoca (6). The other one was an IncFIB plasmid of ∼103 kb that was similar to the plasmid with GenBank accession number NZ_CP015395.1 (average identity, >95%; average coverage percentage, 88.36%) and that harbored bla_KPC-3 in the classical transposon Tn4401 isoform (27, 28), bla_TEM-1A, and bla_OXA-9 (Fig. S7).

In the NDM-1-producing isolate, bla_NDM-1 was harbored by an IncC plasmid of ∼154 kb similar to the plasmid with GenBank accession number NZ_CP029118.1 (average identity, >95%; average coverage percentage, 82.18%) (Table 2; Fig. 2); the genetic environment of bla_NDM-1 included the ISAba125 composite transposon Tn125 (Fig. 2), previously described in pM214_A/C2 (29). Additionally, a class 1 integron located adjacent to Tn125 contained several genes associated with resistance to trimethoprim (dfrA14), rifamycins (arr-2), chloramphenicol (clmA5 encoding the efflux MFS transporter), β-lactams (bla_OXA-10), aminoglycosides (aadA1), and sulfonamides (sul1) (Table 2, Fig. 2). This IncC plasmid also contained the armA, bla_CMY-4, and msrE and mphE genes, encoding resistance to aminoglycosides, β-lactams, and macrolides, respectively. The bla_CMY-4 gene was included in an ISEcp1 transposition unit (Fig. 2), which is usually found in A/C plasmids and which is thought to be responsible for the dissemination of the bla_CMY cephalosporinase genes (30).

FIG 2.

Overview of the IncC plasmid harboring bla_NDM-1 detected in Klebsiella oxytoca in this study. The figure represents the plasmid according to the homology with a highly similar one from the GenBank database (blue outer ring). The graph represents the reads mapped against this reference sequence with a depth of coverage ranging from 0 (red) to 500, with orange indicating values of 1 to 20 reads and green indicating values higher than 200 reads. Gray boxes represent the coding sequence from automatic annotation, with dark and light colors being used when they were found on the forward or the reverse strand, respectively. Colored stripes represent a more detailed annotation that includes antibiotic resistance genes in red, insertion sequences (IS) in blue, and Rep genes in yellow. The homology between the reference plasmid and the assembled contigs is represented in the inner ring, with each contig being colored according to its number.

Conclusions.

We observed the progressive emergence of CP K. oxytoca strains in Spain. Our data reveal several remarkable findings of concern that deserve active surveillance: (i) the clonal and mainly polyclonal spread of CP K. oxytoca across different geographical areas and hospitals; (ii) the production of five different carbapenemase types, mainly VIM-1 and OXA-48, but also KPC and NDM; (iii) the identification of a NDM-1-producing isolate presenting a high load of antibiotic resistance genes, including amrA; (iv) the predominance of ST2 and OXY-2 among the studied isolates; and (v) the detection of the plasmid types IncL, IncHI2, IncFIB, IncN, IncC, and IncP6, likely responsible for the dissemination of carbapenemase genes in K. oxytoca isolates in Spain.

MATERIALS AND METHODS

Study design and bacterial isolates.

This study was performed by the unrestricted and nonmandatory national Spanish Antibiotic Resistance Surveillance Program, operated by the official public health institute of Spain (Instituto de Salud Carlos III) (3, 4). The number of carbapenemase-producing K. oxytoca isolates referred to this program experienced a constant increase from 2012 (n = 6) (31) to 2015 (n = 35), although their proportion with respect to the total number of carbapenemase-producing Enterobacteriaceae isolates remained stable at about 2% to 3%. The observation of an unexpected increase in this proportion in 2016 (5.2%) prompted the present investigation. In this study, we included all CP K. oxytoca isolates submitted to the Instituto de Salud Carlos III between January 2016 and October 2017; during this period a total of 119 public hospitals (about 35% of Spanish public hospitals) voluntarily participated in the surveillance program.

Only the first isolate per patient was analyzed. Bacterial isolate identification was performed using an API 20E system (bioMérieux) and matrix-assisted laser desorption ionization–time of flight mass spectrometry (MALDI-TOF MS; Bruker Daltonik GmbH, Leipzig, Germany).

Antibiotic susceptibility of CP K. oxytoca isolates and phenotypic characterization of carbapenemase production.

Antibiotic susceptibility testing was performed by microdilution (DKMGN panel; Thermo Fisher Scientific, East Grinstead, UK), and the results were interpreted according to EUCAST breakpoints (12, 13). Susceptibility to colistin was carried out by an in-house microdilution method following the recommendations of the CLSI-EUCAST Joint Polymyxins Working Group (32). According to EUCAST (13), the inhibition of carbapenemase activity was determined by using EDTA, phenyl-boronic acid, and cloxacillin. In addition, all the isolates were tested by the Carba NP method (33).

Characterization of resistance mechanisms.

The presence of genes encoding carbapenemases (bla_OXA-48, bla_KPC, bla_VIM, bla_IMP, and bla_NDM), extended-spectrum β-lactamases (bla_CTX-M, bla_SHV, and bla_TEM), plasmid-mediated AmpC β-lactamases (p-AmpC) (bla_CMY, bla_DHA, bla_MOX, bla_ACC, bla_EBC, and bla_FOX), and mcr genes coding for colistin resistance was identified using PCR and DNA sequencing, as previously described (31, 34–36).

Pulsed-field gel electrophoresis.

The genetic relationship between the CP K. oxytoca isolates was elucidated by PFGE after total chromosomal DNA digestion with XbaI (34). A simple diversity index (SDI) was calculated in order to determine the population diversity, as follows: (number of different PFGE profiles/total number of isolates) × 100 (37).

WGS, resistome, and core genome MLST analysis.

Whole-genome sequencing (WGS) was performed on 12 isolates representative of the main CP K. oxytoca clusters previously detected by PFGE and also on 2 additional isolates coproducing two carbapenemase types each.

DNA was extracted using a QIAamp DNA minikit (Qiagen, Hilden, Germany). Genomic DNA paired-end libraries were generated using a Nextera XT DNA sample preparation kit (Illumina Inc., San Diego, CA, USA). The libraries were sequenced using an Illumina NextSeq 500 sequencer system with 2 × 150-bp paired-end reads (Illumina Inc.).

The quality of the high-throughput sequence data was assessed by the use of FastQC software, and short reads were subsequently assembled de novo into contigs using the SPAdes (version 3.9.0) program (38), testing 5 different kmers under parameters optimized to give the best assembly using the QUAST program (http://quast.bioinf.spbau.ru/) in order to analyze the most relevant statistics (largest contig, N₅₀, and NG₅₀). Scaffolding was performed with the SSPACE program (39), and the GapFiller program was used to close sequence gaps (40). Automatic de novo annotation of draft genomes was performed using the Prokka (version 1.12-beta) program (41).

Antimicrobial resistance genes were analyzed using the ResFinder tool (CGE server, https://cge.cbs.dtu.dk; date last accessed, June 2018) with an identification threshold of 98%, with the exception of β-lactamase variants, which were determined with 100% identity. Additionally, the SRST2 program (42) was used to detect resistance genes and alleles with the ARGannot database (43).

Core genome multilocus sequence typing (cgMLST) was applied; an ad hoc scheme was created using the MLST+ target definer with the default parameters and a reference sequence (K. oxytoca KONIH1, GenBank accession number NZ_CP008788.1). A total of 13 NCBI RefSeq genomes were used as query genomes for validation in a pairwise comparison using the BLAST program. The final cgMLST scheme consisted of 3,201 target genes and 2,380 accessory genes. The percentage of good targets included for distance calculation was 94% (see Table S3 in the supplemental material). A distance matrix among all isolates was calculated with 3,013 target genes to analyze the homology within each cluster; genes with missing values were excluded (Table S4).

Curated Illumina sequence reads were mapped onto the reference sequence (K. oxytoca KONIH1, GenBank accession number NZ_CP008788.1), and putative SNPs were obtained using Lasergene Genomics Suite software (DNAStar Inc., Madison, WI, USA), as described previously (44).

Characterization of plasmids carrying carbapenemases genes.

In order to reconstruct the plasmids carrying the carbapenemase genes, an in-house script (PlasmidID, https://github.com/BU-ISCIII/plasmidID) was used to (i) map reads over a plasmid curated database to find those with the higher coverage and de novo assembly of these reads, (ii) make local alignments to localize resistance and replicative genes, and (iii) make a graphic representation of the plasmids identified (45).

Statistical analysis.

Differences in the prevalence of antibiotic susceptibility of OXA-48- and VIM-1-producing isolates were assessed using Fisher’s exact test. The null hypothesis was rejected when a P value of ≤0.05 was calculated. Statistical analysis was performed using GraphPad Prism software (version 3.02; GraphPad Software, Inc., San Diego, CA, USA).

Accession number(s).

The sequences are available on the ENA website under accession number PRJEB30102.