Molecular and epidemiological characterization of carbapenemase-producing Enterobacteriaceae in Norway, 2007 to 2014

The prevalence of carbapenemase-producing Enterobacteriaceae (CPE) is increasing worldwide. Here we present associated patient data and molecular, epidemiological and phenotypic characteristics of all CPE isolates in Norway from 2007 to 2014 confirmed at the Norwegian National Advisory Unit on Detection of Antimicrobial Resistance. All confirmed CPE isolates were characterized pheno- and genotypically, including by whole genome sequencing (WGS). Patient data were reviewed retrospectively. In total 59 CPE isolates were identified from 53 patients. Urine was the dominant clinical sample source (37%) and only 15% of the isolates were obtained from faecal screening. The majority of cases (62%) were directly associated with travel or hospitalization abroad, but both intra-hospital transmission and one inter-hospital outbreak were observed. The number of CPE cases/year was low (2–14 cases/year), but an increasing trend was observed. Klebsiella spp. (n = 38) and E. coli (n = 14) were the dominant species and blaKPC (n = 20), blaNDM (n = 19), blaOXA-48-like (n = 12) and blaVIM (n = 7) were the dominant carbapenemase gene families. The CPE isolates were genetically diverse except for K. pneumoniae where clonal group 258 associated with blaKPC dominated. All isolates were multidrug-resistant and a significant proportion (21%) were resistant to colistin. Interestingly, all blaOXA-48-like, and a large proportion of blaNDM-positive Klebsiella spp. (89%) and E. coli (83%) isolates were susceptible in vitro to mecillinam. Thus, mecillinam could have a role in the treatment of uncomplicated urinary tract infections caused by OXA-48- or NDM-producing E. coli or K. pneumoniae. In conclusion, the impact of CPE in Norway is still limited and mainly associated with travel abroad, reflected in the diversity of clones and carbapenemase genes.

The main carbapenemases among Enterobacteriaceae include KPC (Ambler class A), the metallo-β-lactamases NDM, VIM and IMP (Ambler class B), and OXA-48-like enzymes (Ambler class D) [1]. Certain carbapenemases dominate in specific regions and countries, i.e. NDM in the Indian subcontinent, KPC in Italy, Portugal, Israel, Greece and the US, and OXA-48-like in many Mediterranean (e.g. Turkey and Malta) and North African countries as well as some other European countries (e.g. Belgium, France, Germany and Spain) [7,[18][19][20]. Specific clones or clonal groups (CG) are often associated with specific carbapenemases, while other carbapenemases show a more broad diversity with respect to host genetic backgrounds [2,21]. The global spread of KPC has mainly been associated with Klebsiella pneumoniae sequence type (ST) 258 or CG 258 [2,21,22]. In contrast, NDM and OXA-48-like enzymes are broadly distributed in various genetic backgrounds of K. pneumoniae and Escherichia coli and for bla NDM there is no clear link to a specific plasmid backbone [2,21]. For bla OXA-48-like there is molecular evidence supporting an association with a specific internationally epidemic IncL plasmid backbone [23][24][25].
The aim of this study was to analyse the epidemiological, phenotypic and molecular characteristics of CPE isolated in Norway from 2007 to 2014 to understand the molecular epidemiology associated with the emergence of CPE in Norway.

Bacterial strains and demographic data
The study collection consisted of 59 CPE isolates genetically-verified at the Norwegian National Advisory Unit on Detection of Antimicrobial Resistance from 2007-2014. The criteria for submitting isolates to the Unit included reduced susceptibility to carbapenems according to the Norwegian Working Group for Antibiotics (AFA, https://unn.no/fag-og-forskning/ arbeidsgruppen-for-antibiotikasporsmal-og-metoder-for-resistensbestemmelse-afa)/Nordic Committee on Antimicrobial Susceptibility Testing (NordicAST) guidelines (www.nordicast. org). In 2012 mandatory reporting of confirmed CPE cases to the Norwegian Surveillance System for Communicable Diseases (MSIS) was established. After confirmation at the Advisory Unit, MSIS and the primary lab are notified. The primary laboratory subsequently notifies the responsible clinician, who also reports data to MSIS. Clinical data were collected from the laboratory requisition. Multiple isolates from the same patient were included in the analysis if they were (i) of different species, (ii) the same species, but harboured a different carbapenemase gene or (iii) if the isolates were of the same species and harboured the same carbapenemase gene, but were identified >1 year apart.

Molecular analysis
The presence of carbapenemase genes was initially determined by various PCRs for bla KPC , bla IMI , bla VIM , bla NDM , bla IMP , bla GIM , bla SPM , bla SIM and bla OXA-48-like [41][42][43][44]. WGS was performed on all isolates using the MiSeq platform (Illumina, San Diego, CA, USA) according to the manufacturer's instructions. Briefly, genomic DNA was purified using the GenElute bacterial genomic DNA kit (Sigma-Aldrich, St. Louis, MO, USA). DNA libraries were prepared using Nextera/Nextera XT kits (Illumina) followed by paired-end sequencing. Contigs were assembled using SPAdes [45] through the iMetAMOS extension [46] of the MetAMOS package [47]. The presence of resistance genes/mutations, carbapenemase genes and single nucleotide polymorphisms (SNP) variations were determined using a customised algorithm that uses Bowtie 2 to map reads against a locally curated reference database and assembled from publically accessible databases. The database comprised sequences for all reported carbapenemase variants. Samtools was used to generate an mpileup file [48] which was then parsed based on read depth (> 10 reads per base) and base-call agreement (> 90%) to determine the base type at each nucleotide position relative to the closest reference sequence. Presence of reported carbapenemase variants were defined based on 100% identity across the whole length of the corresponding reference gene.

Genbank accession numbers
WGS data have been deposited at the National Center for Biotechnology Information (NCBI) under BioProject PRJNA295003.

Ethical considerations
The study was reviewed and approved by the Regional Committee for Medical and Health Research Ethics North (reference no. 2016/2122/REK Nord and 2017/146/REK Nord) and the Data Protection Officer at the University Hospital of North Norway (reference no. 2017/1562). The need for patient consent was waived by the Regional Committee for Medical and Health Research Ethics North (reference no. 2017/146/REK nord)

Bacterial isolates
In total 59 CPE were identified from 53 patients of which 44 were hospitalized patients. Samples from eight patients were taken at general practitioners or in other health care institutions (e.g. elderly care homes). For one patient no information was obtained. Of the 53 patients, four had multiple CPE isolates belonging to different species or different STs. One patient had four bla NDM-1 -positive strains of different species (Proteus mirabilis, Providencia stuartii, Citrobacter sp. and K. pneumoniae) isolated within a four-month period. Another had bla KPC-2 -positive K. pneumoniae and Enterobacter cloacae complex isolates in the same faecal screening sample. A third had bla NDM-1 -positive E. coli and E. cloacae complex isolates identified in two different specimens (wound secretion and urine, respectively) within a one-month period. The fourth patient yielded two bla NDM-1 -positive K. pneumoniae strains with unrelated STs from specimens taken 21 months apart.
Increasing number of CPE identified during the study period from a high proportion of clinical isolates CPE isolates were identified in 14 of 22 clinical microbiology laboratories representing all health regions in Norway. The number of CPE cases per year, diversity of carbapenemase variants and species increased during the study period (Table 1), but with a trend towards dominance of NDM and OXA-48-like carbapenemase variants and increasing number of carbapenemase-producing E. coli. Fifty-six percent of the patients were male. The patient age ranged from 3-96  years (mean 63 and median 66 years). The majority of CPE were isolated from urine (n = 22, 37%), blood culture (n = 9, 15%) and faecal screening (n = 9, 15%).

Association with travel or hospitalization abroad
Thirty-three patients (62%) had a known history of travel and/or hospitalization abroad ( Table 2). Sixteen patients (30%) reported no travel or hospitalization abroad and for four patients (8%), no information was obtained. With respect to the non-direct import cases, eight cases were associated with secondary spread from imported cases. This included six cases associated with a previously described, small but long-term outbreak of bla KPC-2 -positive K. pneumoniae/E. cloacae complex in 2007-2010 [50]. In addition, two other intra-hospital transmissions of bla KPC-2 -positive K. pneumoniae [28] and bla VIM-27 -positive K. pneumoniae were observed involving one additional patient in each case.
c Six K. pneumoniae ST258, one K. pneumoniae ST461 and one E. cloacae complex ST484, all bla KPC-2 -positive, were associated with a long-term outbreak [50]. The first case (K. pneumoniae ST258 with bla KPC-2 ) of the outbreak were associated with import from Greece. d One bla KPC-2 -positive K. pneumoniae ST258 associated with intra-hospital transmission (first case associated with import from Greece) [28].
e The bla VIM-27 -positive isolate were associated with a case of intra-hospital transmission (first case associated with import from Greece).
f Both isolates identified from the same patient.  The four carbapenemase-producing E. cloacae complex isolates were all of different STs: ST456 and ST484 both with bla KPC-2 , ST92 with bla NDM-1 and ST635 with bla IMI-9 . All STs were defined as singletons (no SLVs) by BURST analysis of the E. cloacae MLST database (http://pubmlst.org/ecloacae/, last accessed 24.06.2016).
Antimicrobial susceptibility profile and performance of phenotypic methods for detection of CPE All isolates were multidrug-resistant (MDR) according to the definitions by Magiorakos et al. [51]. (Table 3 and S1 Table). One isolate, a bla NDM-1 -positive P. stuartii was non-susceptible to all relevant antimicrobial agents tested. Overall fosfomycin and colistin were the most active antimicrobial agents with 85% and 79% of the isolates being susceptible when excluding P. mirabilis and P. stuartii isolates which are intrinsically resistant to colistin [52] (Table 3). Seven of the twelve colistin-resistant isolates were K. pneumoniae ST258 with bla KPC-2 (n = 6) or bla KPC-3 (n = 1). The other colistin-resistant isolates included K. pneumoniae ST525 with bla NDM-1 + bla OXA-181 , K. pneumoniae ST147 with bla NDM-1 , K. pneumoniae ST336 with bla NDM-7 , E. cloacae complex ST635 with bla IMI-9 and E. cloacae complex ST456 with bla KPC-2 .
With respect to the carbapenems, 41% were susceptible to meropenem, 39% to imipenem and 3% to ertapenem. All isolates had meropenem and ertapenem MIC values above the EUCAST screening breakpoint for carbapenemase detection (http://www.eucast.org/ fileadmin/src/media/PDFs/EUCAST_files/Resistance_mechanisms/EUCAST_detection_of_   resistance_mechanisms_v1.0_20131211.pdf) (S1 Table). For imipenem nine isolates had MIC values below the screening breakpoint. There was no clear correlation between carbapenemase variant and susceptibility to meropenem and imipenem with the exception that among the isolates harbouring bla OXA-48-like (excluding the strain with both bla NDM-1 and bla OXA-181 ) 9/11 and 8/11 were susceptible to meropenem and imipenem, respectively. As expected, a high level of resistance was observed against other β-lactams (Table 3 and S1 Table). Three isolates: one K. pneumoniae (bla OXA-48 ), one K. variicola (bla OXA-48 ) and the bla IMI-9 -positive E. cloacae complex isolate were susceptible to extended-spectrum cephalosporins (cefotaxime, ceftazidime and cefuroxime) and aztreonam. Interestingly, all OXA-48-like-positive E. coli and Klebsiella spp. as well as 83% and 89% of NDM-positive E. coli and Klebsiella spp. isolates, respectively were susceptible to mecillinam. Nine (15%) of the isolates tested negative for carbapenemase-production with the in-house Carba NP test (S1 Table), including six bla NDM-1 -positive isolates (E. coli n = 2, P. stuartii, P. mirabilis, Citrobacter sp. and K. pneumoniae), two bla OXA-48-like -positive isolates (E. coli and K. pneumoniae) and one E. cloacae complex isolate (bla IMI-9 ). The KPC, MBL and OXA-48 confirm kit correctly identified the presence of either an MBL or KPC in all relevant isolates except for one bla NDM-1 -positive P. mirabilis strain (S1 Table). The single bla IMI-9 -positive E. cloacae complex isolate also showed significant synergy with boronic acid only. With the exception of the isolate harbouring both bla NDM-1 and bla OXA-181 , where synergy was observed between meropenem and dipicolinic acid, no synergy was observed with the β-lactamase inhibitors for all bla OXA-48 -likepositive isolates. Moreover, with the exception of two isolates, all bla OXA-48-like -positive isolates showed no zones of inhibition around the temocillin tablet, which may indicate the presence of OXA-48-like carbapenemases according to the manufacturer's guidelines. The meropenem-meropenem/EDTA gradient strip correctly identified all MBL-positive isolates, with the exception of the K. pneumoniae strain positive for both bla NDM-1 and bla  where the test was inconclusive (S1 Table).

Association with other antibiotic resistance determinants
Bla CTX-M and specifically bla CTX-M-15 were the most common ESBL variants identified and were mainly associated with K. pneumoniae and E. coli isolates with bla NDM (10/15 isolates) or bla OXA-48-like (8/11 isolates) and E. coli isolates with bla VIM (2/2 isolates) (S1 Table). Bla CTX-M were not identified in bla KPC -positive K. pneumoniae isolates. One E. coli isolate with bla OXA-48 harboured both bla CTX-M-14 and bla CTX-M-15 . bla CTX-M-15 was also identified in one bla KPC-2and one bla NDM-1 -positive E. cloacae complex. Bla CMY (n = 12) were the most common plasmid-mediated AmpC variants identified with bla CMY-6 particularly associated with bla NDM (n = 9). The two bla OXA-48-like -positive Klebsiella spp. isolates that were susceptible to extendedspectrum cephalosporins and aztreonam were negative for ESBL and plasmid-mediated AmpC genes.
In addition to various genes encoding aminoglycoside-modifying enzymes, the 16S rRNA methylase genes rmtC and armA, were identified in eight and five isolates, respectively (S1 Table). With the exception of the single isolate of E. coli with bla IMP-26 , armA and rmtC were only associated with isolates harbouring bla NDM-1 . In Klebsiella spp. insertional disruption of mgrB [53] associated with colistin resistance was identified in seven K. pneumoniae isolates (S1 Table). Insertional disruption of mgrB was also observed in two clinically colistin susceptible (MIC = 1 mg/L) K. pneumoniae isolates. One K. pneumoniae isolate with a disrupted mgrB also carried a nonsense mutation in pmrB leading to a truncated PmrB. Two colistin-resistant K. pneumoniae isolates had mutations in pmrA resulting in amino acid substitutions of G53C and D86E in one, and G53C in the other. In one colistin-resistant Klebsiella spp. isolate (MIC >8 mg/L) no previously described colistin resistance determinants were identified. The strain had mutations in pmrA (PmrA E57G) and pmrB (PmrB T246A) compared with the colistin-susceptible K. pneumoniae strain MGH 78578 [54], but neither mutation has been linked with colistin resistance and PmrB T246A is commonly found in K. pneumoniae [54]. No mutations were identified in phoP, phoQ or the mgrB promoter for this isolate. The plasmid-mediated colistin resistance genes mcr-1 [9], mcr-2 [10], mcr-3 [12], mcr-4 [13] and mcr-5 [14] were not detected.

Discussion
The main objective of this study was to gain a better understanding of the molecular epidemiology associated with the emergence of CPE in Norway. As observed in other Nordic countries [26,27,[32][33][34][35][36] the emergence of CPE in Norway is also mainly associated with importation, highlighting the importance of targeted screening of patients hospitalized abroad and patients with a recent travel history to a country with a high prevalence of CPE. A relatively low number of cases (15%) were identified through faecal screening in contrast to Sweden (74,5%) and France (59.8%) [26,55]. This difference is most likely due to dissimilarities in the use of targeted screening and that CPE screening in Norway was not fully implemented in the study period. This could also explain why a higher proportion of CPE cases in Sweden (81%) were associated with import [26]. Revised recommendations for infection prevention and control, including indications for screening for CPE, were introduced in Norway in August 2015 and in the first six months of 2016, 63% of CPE cases were identified through faecal screening. The occurrence of one long-term outbreak and two separate incidences of secondary transmission further highlights the importance of rapid implementation of infection prevention and control measures before confirmation of CPE if patients have risk factors (e.g. hospitalization abroad) or when an MDR isolate is identified.
The diversity of species and genetic backgrounds observed is probably due to the high degree of importation from a variety of countries (Table 2). Several studies have shown that the dissemination of resistance genes among clinical strains of Enterobacteriaceae is often associated with high-risk clones and the linkage between specific genetic backgrounds and resistance genes [2,21,56]. The cgMLST analysis of K. pneumoniae isolates showed that the observed epidemiology reflects the current global epidemiology (Fig 1), where bla KPC-2/-3 spread is primarily driven by strains associated with CG258 (and more specifically, ST258). In contrast, ST11 (a member of CG258, and a single locus variant of ST258) has been shown to be associated with a diversity of carbapenemase genes including bla KPC , bla NDM , bla VIM and bla OXA-48-like in different geographical regions [2,57,58]. Accordingly, the four ST11 strains in this study harboured either bla NDM-1 (n = 2) or bla OXA-245 (n = 2). Notably, cgMLST has shown that ST11 and ST340 represent a genetic sublineage within CG258 [22]. Isolates with bla NDM and bla VIM belonging to two other globally dispersed high-risk CGs like CG17 and CG147 [2] were also identified. The identification of bla VIM-1 and bla OXA-48 in K. quasipneumoniae and K. variicola, respectively shows that these Klebsiella species also contribute to the dissemination of carbapenemase genes and infections as both isolates were associated with infection. K. variicola have been shown to be frequently associated with bloodstream infections and associated with higher mortality than K. pneumoniae [59].
All three E. coli ST38 isolates harboured bla OXA-48 , which is consistent with previous observations showing a prevalent linkage of ST38 to bla OXA-48 in a large collection of clinical isolates from European and North-African countries [23]. In contrast, the three E. coli isolates belonging to ST410 were associated with different carbapenemase genes (bla NDM-1 , bla VIM-4 or bla OXA-181 ) indicating the ability of this genetic background to maintain different plasmids and resistance genes. ST410 E. coli isolates have also previously been identified harbouring bla KPC-2 [60]. The global dissemination of bla NDM has so far not been linked to specific highrisk clones or epidemic plasmids [21] and this is also reflected among the five bla NDM -positive E. coli isolates, which belonged to five different genetic backgrounds. However, one strain belonged to the international high-risk clone ST131 [21] and another to ST101, which has previously been found to be associated with bla NDM and other carbapenemases in several countries (e.g. Bangladesh [61], USA [62], Canada [63,64] and Bulgaria [65]). CPE frequently exhibit MDR or XDR phenotypes, limiting treatment options [1,4]. This was also observed in our strain collection ( Table 2 and S1 Table) due to the association with a wide variety of other acquired resistance genes, including 16S rRNA methylase genes conferring high-level broad-spectrum aminoglycoside resistance [66] and chromosomal mutations/ insertions resulting in ciprofloxacin and colistin resistance (S1 Table). The mechanism(s) behind colistin resistance in one K. pneumoniae strain and the colistin-resistant E. cloacae isolates remains to be determined. Interestingly, a high prevalence of susceptibility to mecillinam among OXA-48-and NDM-producing E. coli and K. pneumoniae isolates was observed. Marrs et al. also showed high levels of in vitro susceptibility to mecillinam among NDM-producing E. coli and K. pneumoniae isolates from Pakistan [67], suggesting that mecillinam could have a role in the treatment of uncomplicated urinary tract infections caused by OXA-48-or NDMproducing E. coli or K. pneumoniae [68].
Rapid identification of CPE is essential for timely implementation of enhanced infection control measures to reduce transmission of CPE and prevent infections [3]. As observed in previous studies [69,70] false-negative results (15%) for carbapenemase production were observed with the in-house version of the Carba NP test, particularly with NDM-and OXA-48-like-producing isolates. Identification of OXA-48-like-producers can be particularly challenging due to their relatively low level of activity against carbapenems and the lack of specific inhibitors [71]. The relatively high number of false-negative Carba NP results could also be due to the media used. In our study, colonies for the Carba NP test were harvested from MH agar and Literacka et al have recently reported that MH agar from different companies were associated with false-negative results for MBL-producers [72]. High-level resistance to temocillin is a sensitive and specific indicator for the presence of OXA-48-like enzymes [73]. All bla OXA-48-like -positive isolates in our collection showed high-level resistance (MIC>128 mg/L) to temocillin, but several isolates harboring bla VIM and bla NDM also had temocillin MIC >128mg/L showing that testing for synergy with metal chelators (e.g. EDTA or dipicolinic acid) is necessary to discriminate between isolates with OXA-48 and MBLs.

Conclusions
The low prevalence of clinical CPE in Norway is consistent with the general low level of antimicrobial resistance compared with other countries. The relatively low level of antibiotic consumption and the use of narrow spectrum antibiotics [74] have probably contributed to this situation. The low prevalence is also reflected in the epidemiology of Norwegian CPE; mainly associated with importation, exhibiting a broad diversity of genetic backgrounds and carbapenemase variants that mirror the global epidemiology. Only a few cases of secondary spread also support this notion. In order to limit the infection pressure brought by increasing travel and globalization, continued emphasis must be put on diagnostic capabilities, surveillance and infection control.
Supporting information S1