The NICU of V. Monaldi Hospital in Naples is a tertiary care level NICU and consists of three rooms and 16 cot spaces with a 1:2–1:3 nurses/neonates ratio. The NICU serves approximately 260 admissions per year and admits exclusively babies from the regional Newborn Emergency Transport Service or Territorial Emergency Service and through the transfer from the internal departments of Pediatric Cardiology and Pediatric Heart Surgery. In case of necessity, the department performs repeated hospitalizations for some particular types of newborns (lower birth weight, heart disease, etc.). Active patient-based surveillance of healthcare-associated infections on neonates with >2 days NICU stay is performed as previously described (Horan et al., 2008; Crivaro et al., 2015). Surveillance of CRE in V. Monaldi Hospital, based on rectal swabs at hospital admission, is performed according to the guidelines of European Centre for Disease Prevention and Control [ECDC] (2012) and the Ministero della Salute (2013), Italy.

Klebsiella pneumoniae isolates were identified using the Vitek 2 automatic system and the ID-GNB card according to the manufacturer’s instructions (bioMérieux, Marcy l’Etoile, France) as previously described (Del Franco et al., 2015).

Carbapenem resistance of Enterobacteriaceae was screened using the meropenem disk alone as previously described (Pournaras et al., 2013). Identification of MBL activity was performed using imipenem + EDTA 15 + 750 μg combined disk method (ROSCO Diagnostica A/S, Taastrup, Denmark). Susceptibility tests were performed using the Vitek 2 system and the AST-GN card (bioMérieux, Marcy l’Etoile, France); carbapenem and colistin susceptibility were evaluated by broth microdilution in Mueller–Hinton broth II (MHBII) according to Clinical and Laboratory Standards Institute guidelines (CLSI, 2015). Breakpoints values were those recommended by the EUCAST (2016).

Characterization of β-lactamase genes was performed as previously described (Poirel et al., 2011). Two multiplex PCRs were set up, the reaction no. 1 including detection of bla_KPC and bla_OXA-48-like and the reaction no. 2 including detection of bla_IMP, bla_{V IM}, and bla_NDM. The following thermal cycling conditions were used: 3 min at 94°C and 35 cycles of amplification consisting of 1 min at 94°C, 1 min at 57°C, and 1 min at 72°c with 5 min at 72°C for the final extension. PCR products were analyzed by electrophoresis in a 1.8% agarose gel stained with ethidium bromide. The following strains were used as positive quality controls: Acinetobacter baumannii AC 54/97 (Riccio et al., 2000), A. baumannii 161/07 (Bonnin et al., 2012), K. pneumoniae D001 (Pournaras et al., 2013), K. pneumoniae 1, K. pneumoniae 2, and E. coli 6 (Del Franco et al., 2015) for bla_IMP-2, bla_NDM-1, bla_OXA-48, bla_KPC-2, bla_KPC-3, and bla_{V IM-1} carbapenemase genes, respectively. The full-length alleles of bla_{V IM} were amplified using primers 5′VIM1 and 3′VIM1 shown in Supplementary Table S1. Sanger DNA sequencing and identification of deduced amino acid sequences were performed as previously described (Del Franco et al., 2015).

Klebsiella pneumoniae isolates were genotyped by XbaI DNA macrorestriction, pulsed-field gel electrophoresis (PFGE) and dendrogram analysis as described previously (Del Franco et al., 2015).

Klebsiella pneumoniae isolates were typed using the Institut Pasteur’s MLST (multi-locus sequence typing) scheme (Diancourt et al., 2005) and primes and PCR conditions available at http://bigsdb.pasteur.fr/klebsiella/primers_used.html eBURST analysis of profiles available at http://bigsdb.pasteur.fr/perl/bigsdb/bigsdb.pl?db=pubmlst_klebsiella_seqdef_public&page=profiles was performed as described previously (Feil et al., 2004). Minimum spanning trees of sequence types (STs) were built by Phyloviz using the goeBURST algorithm (Francisco et al., 2012).

Filter mating experiments were performed using sodium azide resistant E. coli J53 (Jacoby and Han, 1996) as recipient strain in the presence of either 5 μg/ml imipenem and 100 mg/l sodium azide or 50 μg/ml ampicillin and 100 μg/ml sodium azide. The frequency of transfer was calculated as the number of transconjugants divided by the number of surviving recipients as previously described (Zarrilli et al., 2005). Plasmids typing was performed by PCR-based replicon typing (PBRT) kit (Diatheva s.r.l., Fano, Italy) using previously described primers and conditions (Carattoli et al., 2005, 2006).

DNA was extracted from the K. pneumoniae parental strain, and from the E. coli strain before and after transconjugation using a DNeasy Blood & Tissue Kit according to the manufacturer’s instructions (Qiagen, Milan, Italy). Whole genome sequencing was performed using an Illumina Miseq platform with a 2 by 250 paired-end run after Nextera XT paired-end library preparation. Sequencing reads from the K. pneumoniae parental strain and from the E. coli strain before the transconjugation obtained in this study were assembled using the software SPAdes (Bankevich et al., 2012) with accurate settings.

The BIGSdb genome database software (Jolley and Maiden, 2010) was used in the BIGSdb-Kp database¹ to identify genes associated with virulence, heavy metal and drug resistance in K. pneumoniae 4898 genome sequence. In Silico detection of plasmids was performed using PlasmidFinder Web tool at https://cge.cbs.dtu.dk/services/PlasmidFinder/ as previously described (Carattoli et al., 2014). Sequencing reads of the E. coli transconjugant strain containing the plasmid were mapped to the assembled genome of the same E. coli sequenced prior to transconjugation, using the mapping software Bowtie2 (Langmead and Salzberg, 2012). All non-mapping reads were extracted and assembled using SPAdes (Bankevich et al., 2012) with accurate settings. Presence of the obtained contigs in the original K. pneumoniae genome was verified by blast-searching followed by manual analysis, i.e., all the newly obtained contigs that aligned with the K. pneumoniae assembly with an identity over 95% were kept. The co-linearity of the contigs was assessed using Bandage tool for visualizing assembly graphs (Wick et al., 2015). The connections indicated by Bandage were used as starting point for finishing experiments by PCR and Sanger sequencing to bridge the ends of contigs (see Supplementary Table S1 for a complete list of PCR primers used).

Plasmid MLST (PMLST) and core genome PMLST (cgPMLST) analysis of Inc A/C plasmid profiles available at https://pubmlst.org/bigsdb?db=pubmlst_plasmid_seqdef was performed as previously described (Hancock et al., 2017). The 28 IncA/C conserved genes from each PMLST profile were aligned using Muscle (Edgar, 2004). Unreliable positions were removed from each alignment using Gblocks (Castresana, 2000). All alignments were concatenated and used as input for a maximum likelihood phylogenetic analysis, which was performed with the software PhyML 3.0 (Guindon et al., 2010) using the GTR substitution model. PMLST minimum spanning trees were built by Phyloviz (Francisco et al., 2012). The gene annotation of the plasmid was performed using the software Prokka (Seemann, 2014) followed by accurate manual control, based on blast-searches against the nr-protein database. Inverted repeats were identified manually, based on the sequence and patterns found in Hancock (AAC 2017).

The genome sequences of K. pneumoniae 4898 and plasmid pIncAC-KP4898 have been deposited in the GenBank nucleotide database under accession numbers FWYI00000000.1 and KY882285, respectively.

The study has been evaluated by the local Ethics committee (Comitato Etico Università degli Studi della Campania “Luigi Vanvitelli” AOU “Luigi Vanvitelli” – AORN “Ospedali dei Colli”) (protocol number 421/2017). Because the patients included in the study were anonymized, no written informed consent was required.

The emergence of carbapenem-resistant (CR) K. pneumoniae was observed in the NICU of V. Monaldi Hospital from October 2015 to March 2016, when CR K. pneumoniae were isolated from rectal swabs of 19 neonates into two consecutive clusters. Only one sporadic CR K. pneumoniae isolate was obtained from a rectal swab of a neonate in the NICU of the hospital during the previous 22 months, while CRE were occasionally isolated from rectal swabs and other clinical specimens of adult patients in other wards of the hospital (Figure 1 and data not shown). Immediately following the first isolation of CR K. pneumoniae, a multimodal infection control program was implemented in the NICU, which included: weekly or biweekly screening for CRE at rectal swab of hospitalized neonates, increased frequency of environmental cleaning using chloride derivatives at 1100 ppm, reinforce handwashing compliance before and after patient contact, and daily visits to the ward by the hospital Infection Control Nurse. Colonized neonates were isolated either structurally or following strict adherence to contact precautions and staff cohorting was performed. During the two clusters, the ward was temporarily closed to external admissions. Environmental microbiological investigation of room surfaces, equipment, and staff hands failed to identify sources or reservoirs of CR K. pneumoniae. Neonates’ mothers were not screened for the presence of CR K. pneumoniae. Because all neonates with CR K. pneumoniae isolation did not show signs of infection, they did not receive antimicrobial therapy against CR K. pneumoniae and were not eradicated before discharge. The cluster ended on March 2016 when the last CR K. pneumoniae intestinal carrier neonate was discharged from the ward (Figure 1).

All CR K. pneumoniae isolates from neonates in the NICU showed a MDR phenotype. In fact, they exhibited resistance or intermediate resistance to carbapenem (imipenem, meropenem, ertapenem), resistance to aminopenicillins, ureidopenicillins, third and fourth generation of cephalosporins (ceftazidime, cefotaxime, cefepime), gentamicin, fosfomycin, and trimethoprim–sulfamethoxazole, but were susceptible to amikacin, ciprofloxacin, and colistin (Table 1). All K. pneumoniae isolates gave a positive result in the MBL-assay performed using the imipenem–EDTA combined disk method. PCR and sequence analysis identified the presence of bla_{V IM-1} but not any of the other carbapenemase genes tested in all CR K. pneumoniae isolates from NICU (Figure 2).

To assess whether the increase of CR K. pneumoniae isolation in the NICU was caused by the spread of epidemic strains, all 20 CR K. pneumoniae isolates from 20 neonates were genotyped by XbaI digestion, PFGE, and dendrogram analysis. Molecular analysis identified an identical macrorestriction pattern in 17 isolates, which we named PFGE type A, whereas two isolates showing difference in the migration of one band and one isolate in the migration of four bands and a similarity of >85% at dendrogram analysis were classified into subtypes A1 and A2, respectively (Figure 2). The above data indicated that the increase of CR K. pneumoniae isolation in the NICU was caused by the spread of a single VIM-1-producing K. pneumoniae epidemic genotype. MLST analysis assigned CR K. pneumoniae isolates with PFGE type A and subtypes A1 and A2 to ST104 (Figure 2). K. pneumoniae ST104 isolate from milk during bovine mastitis and K. pneumoniae ST1923 and ST1942 isolates from human blood and human feces, respectively, which were single-locus variants of ST104, and 21 other STs, which were double-locus variants of ST104, were reported worldwide and found in Klebsiella PubMLST isolates database² (Figure 3).

Additional epidemiological information was provided by genome sequence of representative K. pneumoniae KP4898 isolate. Genome sequence confirmed MLST assignment of K. pneumoniae KP4898 isolate to ST104. Capsular typing through sequencing of CD1–VR2–CD2 region of wzc and outer membrane protein wzi genes of locus identified wzc-32 and wzi-102 alleles, respectively, which are associated with K31 capsular type. In the KP4898, an array of virulence-associated genes was found, which includes the type-3 fimbriae cluster mrkABCDF and transcription regulators mrkHIJ, the yersiniabactin siderophore cluster ybtAPQSTUX, the yersiniabactin receptor fyuA, yersiniabactin biosynthetic protein genes irp1 and irp2 and allB and allD genes of allantoinase cluster. The analysis of drug-associated resistance genes in KP4898 genome identified bla_{SHV -5} and bla_{SHV -12} extended-spectrum β-lactamases, bla_{V IM-1} carbapenemase, aadA1, aphA15, and aacA4 aminoglycoside resistance genes, a macrolide 2′-phosphotransferase cluster, other antimicrobial resistance genes, heavy metal resistance genes, efflux system, and regulators genes (Supplementary Table S2). The PlasmidFinder web tool identified A/C, FII(K) and FIB(K) replicons, thus suggesting the presence of at least three plasmids in KP4898 genome.

The transfer of carbapenem resistance from VIM-1-producing ST104 K. pneumoniae isolates with PFGE type A and subtypes A1 and A2 was evaluated by filter mating experiments. Resistance or intermediate resistance to aminopenicillins, ureidopenicillins, third and fourth generation cephems, imipenem, but not meropenem and ertapenem, was transferred from ST104 K. pneumoniae isolates with PFGE type A and subtypes A1 and A2 to E. coli J53 aziR at frequency ranging from 6.5 × 10^-3 to 1.5 × 10^-3 cfu/recipient cells. The frequency of transfer of imipenem resistance from K. pneumoniae ST104/PFGE A, A1, and A2 isolates to E. coli J53 aziR did not change if transconjugants were selected in the presence of imipenem and sodium azide or ampicillin and sodium azide. All transconjugants showed identical antimicrobial susceptibility profile and resistance to sodium azide (Table 2). Moreover, all transconjugants showed a PFGE profile identical with that of the recipient strain. The presence of bla_{V IM-1} and bla_{SHV -12} genes was demonstrated in all transconjugants. Furthermore, PBRT identified A/C replicon and IncA/C incompatibility group plasmid/s in ESBL positive and CR K. pneumoniae ST104 donor isolates and E. coli transconjugants expressing bla_{V IM-1} and bla_{SHV -12} genes. The above data suggested that conjugative plasmid/s mediated the horizontal transfer of carbapenem resistance.

In the genome sequence of E. coli transconjugants, one single plasmid was identified, which was present in K. pneumoniae KP4898 donor strain, but not in E. coli J53 recipient strain, and was designated pIncAC-KP4898, for plasmid of IncA/C incompatibility group from K. pneumoniae 4898 isolate. The pIncAC-KP4898 plasmid was 156,252 bp in size, with an average G + C content of 52.6%. Genome annotation identified 190 open reading frames (ORFs), of which 145 were transcribed in a clockwise orientation, while the remaining 45 were transcribed counterclockwise. Of these ORFs, 14 were associated with plasmid DNA replication and partition, 16 with DNA transfer, 13 with DNA-restriction and site-specific DNA methylation, 25 with DNA transposition, 19 with antimicrobial resistance, and 103 with unknown functions (Figure 4 and Supplementary Table S3). The pIncAC-KP4898 scaffold included the repA, parA, parB, and 053 genes, parM, kfrA, and a putative toxin–antitoxin system, which were demonstrated to be important for maintenance and replication of IncA/C plasmid (Hancock et al., 2017). Based on RepA similarity (Carattoli et al., 2006), the pIncAC-KP4898 plasmid belongs to IncA/C1 group. The IncA/C PMLST scheme based on the repA, parA, parB, and 053 genes and the cgPMLST scheme that extend the 4-gene PMLST to 28 conserved genes (Hancock et al., 2017) assigned pIncAC-KP4898 to novel profiles, which corresponded to ST12 and ST12.1, respectively. The phylogeny of IncA/C 28-gene profiles showed that ST12.1 was more closely related to ST11.1, to which IncA/C1 group pRA1 plasmid (GenBank accession number FJ705807.1) has been assigned, than to the remaining 10 profiles (ST1 to ST10), which corresponded to IncA/C2 group plasmids (Figure 5).

Plasmid pIncAC-KP4898 contained a composite transposon of 25,716 bp (residues 139,120–156,252 and 1–8,583, G + C content: 57.7%) with four IS6 family transposase and 14 bp inverted repeats (TTTGCAACAGTGCC) at residues 152,072–152,086 and 8,569–8,583 (3′-flanking region) (Figure 4). Within this transposon lied a class 1 integron containing a 5′-conserved structure (CS) with int1 site-specific integrase, and five head-to-tail arranged gene cassettes consisting of the genes bla_{V IM-1}, aacA4 family aminoglycoside N(6′)-acetyltransferase gene, aminoglycoside O-phosphotransferase APH(3′)-XV encoding gene, ANT(3″) family aminoglycoside nucleotidyl transferase gene, type B-2 chloramphenicol O-acetyltransferase catB2 gene. The 3′-CS showed the qacED1-encoding gene, which confers resistance to quaternary ammonium compounds. The Tn3-like composite transposon included also the bla_{SHV -12} gene flanked by IS6 family transposases in inverted orientation, a macrolide 2′-phosphotransferase gene cluster consisting of macrolide 2′-phosphotransferase mph(A), transporter mfs and macrolide 2′-phosphotransferase I repressor A mphR genes, quinolone resistance pentapeptide repeat protein qnrA1 gene, and sulfonamide-resistant dihydropteroate synthase sul1 genes. A mercury resistance gene clusters and trimethoprim-resistant dihydrofolate reductase dfrA14 gene were carried by pIncAC-KP4898 apart from the identified transposon region. In addition, 16 genes encoding transfer functions and a conjugative apparatus were found in pIncAC-KP4898, which indicated that the plasmid was self-conjugative (Figure 4 and Supplementary Table S3).