|Home | About | Journals | Submit | Contact Us | Français|
International data on the molecular epidemiology of Enterobacteriaceae with IMP carbapenemases are lacking. We performed short-read (Illumina) whole-genome sequencing on a global collection of 38 IMP-producing clinical Enterobacteriaceae (2008 to 2014). IMP-producing Enterobacteriaceae (7 varieties within 11 class 1 integrons) were mainly present in the South Pacific and Asia. Specific blaIMP-containing integrons (In809 with blaIMP-4, In722 with blaIMP-6, and In687 with blaIMP-14) were circulating among different bacteria in countries such as Australia, Japan, and Thailand. In1312 with blaIMP-1 was present in Klebsiella pneumoniae from Japan and Citrobacter freundii from Brazil. Klebsiella pneumoniae (n = 22) was the most common species; clonal complex 14 (CC14) from Philippines and Japan was the most common clone and contained In1310 with blaIMP-26 and In1321 with blaIMP-6. The Enterobacter cloacae complex (n = 9) consisted of Enterobacter hormaechei and E. cloacae cluster III. CC78 (from Taiwan) containing In73 with blaIMP-8 was the most common clone among the E. cloacae complex. This study highlights the importance of surveillance programs using the latest molecular techniques for providing insight into the characteristics and global distribution of Enterobacteriaceae with blaIMP genes.
Carbapenems are often the last line of effective therapy available for the treatment of serious infections due to multidrug-resistant bacteria. The rapid evolution of carbapenem resistance in Enterobacteriaceae during the last decade is an emerging global threat (1, 2). Enzymes that hydrolyze the carbapenems, known as carbapenemases, are the most important causes of carbapenem resistance. Carbapenemase-producing Enterobacteriaceae (CPE) have acquired multiple resistance genes, making treatment of infections due to these bacteria challenging (1, 2).
The most common carbapenemases among CPE are Klebsiella pneumoniae carbapenemases (KPCs) (Amber class A), IMPs, VIMs, NDMs (class B or metallo-β-lactamases [MBLs]), and OXA-48-like (class D) enzymes (1). Metallo-β-lactamases hydrolyze all β-lactams except aztreonam, although resistance levels may vary according to different subtypes. After the initial discovery of IMP-1 in Japan in 1991, bacteria with IMP enzymes have been detected worldwide (1). IMPs are common among MBL-producing Pseudomonas aeruginosa but remain relatively rare among members of the Enterobacteriaceae (3). IMP-producing Enterobacteriaceae are mainly identified in Asia-Pacific (i.e., China, Japan, Taiwan, and Australia) (1, 4). The most common species associated with IMPs among the Enterobacteriaceae include Klebsiella pneumoniae, Escherichia coli, and Enterobacter spp. (2, 3). IMP genes are often situated within class 1 integrons harbored on broad-host-range plasmids, with the exception of some blaIMP-encoding class 2 and 3 integrons (2, 3). These mobile genetic elements play an important role in the interspecies distribution of IMP types of carbapenemases (5).
Comprehensive global data regarding the molecular epidemiology of CPE with blaIMP are currently lacking. We designed a study that utilized short-read whole-genome sequencing to describe the molecular characteristics and international distribution of blaIMP among Enterobacteriaceae obtained from the Study for Monitoring Antimicrobial Resistance Trends (SMART) global surveillance system.
The 38 IMP-producing Enterobacteriaceae isolates were obtained from eight countries, mainly from the South Pacific (n = 21), Asia (n = 15), and 1 each from Brazil and Spain (Fig. 1; see also Data Set S1 in the supplemental material). Common sources were intra-abdominal specimens and urine samples (n = 27 and 11, respectively). The isolates included the following species: K. pneumoniae (n = 22), Enterobacter cloacae complex (n = 9), Citrobacter spp. (n = 4), and E. coli (n = 3) (Fig. 1 and Table 1).
The 38 genomes were sequenced at an average depth of 120 (standard deviation [SD], 56.3) (Data Set S1). Assembled genomes had an average number of contigs of 104 (SD, 41.5) and an N50 value of 283,649 bp (SD, 107,564 bp). We confirmed the presence of blaIMP in the draft genomes of all of the isolates.
The presence of resistance genes, antibiotic resistance profiles, plasmid replicons, and plasmid addiction systems is shown in Fig. S1 in the supplemental material. Table 1 shows the geographical distribution of the different species, types of carbapenemases, and integrons. We identified 7 blaIMP variants, namely blaIMP-1 (n = 5), blaIMP-4 (n = 8), blaIMP-6 (n = 4), blaIMP-8 (n = 4), blaIMP-13 (n = 1), blaIMP-14 (n = 3), and blaIMP-26 (n = 13). The following genes were present in more than 1 species: blaIMP-1, blaIMP-4, blaIMP-6, and blaIMP-14 (Table 1). Enterobacteriaceae with blaIMP-14 (i.e., K. pneumoniae, Enterobacter spp., and E. coli) and blaIMP-13 (Enterobacter spp.) were obtained from urine samples in Thailand and Spain; the remainder of the blaIMP genes were present in both urine samples and intra-abdominal specimens. The distribution of the different blaIMP subtypes was similar to previously published data (i.e., blaIMP-1 and blaIMP-6 were present in Japan, blaIMP-4 in Australia, blaIMP-8 in Taiwan, blaIMP-14 in Thailand, and blaIMP-26 in Philippines (Table 1) (2, 6,–8).
All of the blaIMP genes were situated within class 1 integrons. We were unable to sequence the complete integron-associated gene cassettes in 13 isolates due to the limitations associated with short-read sequencing.
We identified 11 different integron types containing blaIMP, including 7 novel cassette combinations (Table 2). The novel cassette combinations included the following: blaIMP-26-qacG-aacA4 (In1310), aacA4-blaIMP-1-aadA2-tnpA (In1311), blaIMP-1-aacA31-blaOXA-142 (In1312), blaOXA-10, aacA4-blaIMP-1-qacG (In1313), blaIMP-14-blaOXA-10, aacA4 (In1314), blaIMP-13-aacA4-blaOXA-2 (In1319), and aacA4-blaIMP-6-aadA2-tnpA (In1321). In1310 containing blaIMP-26-qacG-aacA4 was the most common cassette. The novel integron In1312 with blaIMP-1-aacA31-blaOXA-142 had international, intercontinental, and intergenus distribution (present in K. pneumoniae from Japan and Citrobacter freundii from Brazil). Country-specific blaIMP subtypes corresponded to the specific integron types previously characterized in that country, i.e., blaIMP-4, In809 from Australia (9); blaIMP-6, In722 from Japan (10); blaIMP-8, In73 from Philippines (11); and blaIMP-14, In687 from Thailand (7).
The aacA variant (especially aacA4) aminoglycoside acetyltransferase genes were the most prevalent gene cassettes in the different integron cassette combinations. They were present in all 11 blaIMP-containing integrons. The second most common cassette was the aminoglycoside adenylyltransferase gene (aad variant), which was present in 3 blaIMP-containing integrons.
The blaOXA-142 cassette had previously been identified in class 1 integrons among P. aeruginosa from Bulgaria (without blaIMP) (12) and from Taiwan (with blaIMP) (13). This suggests that appropriate surveillance and control methods may need to extend beyond the Enterobacteriaceae.
Integrons with weak promoters (i.e., PcW and PcH1) were common, whereas strong promoters (i.e., PcS and PcH2) were rare (Table 2; see also Table S1 in the supplemental material). There was no correlation between the type of promoter (weak versus strong) and MICs to ertapenem and imipenem. It seems that carbapenem MICs are influenced by various factors, such as type of IMP, porin deficiency, presence of other extended-spectrum β-lactamases (ESBLs), and efflux pumps. We were able to characterize the downstream structures in 9 blaIMP-containing integrons (Table 2; see also Table S2 in the supplemental material). All integrons contained 3′-conserved segment (CS) structures immediately downstream of the gene cassettes except for blaIMP-1-containing In1313 with a Tn402-like tniRQBA structure (14) (Table 2). The majority contained In4-like structures consisting of 3′-CS-IS6100 (with or without partial deletion of 3′-CS and chrA-padR insertion).
The integron diversity observed in this study most likely represents the following: (i) the sequential evolution over time of structurally similar cassettes (i.e., blaIMP-1, blaIMP-6 with only one single nucleotide variant [SNV] difference) situated within homogenous integrons (i.e., variants of In722), and (ii) the multiple acquisition of the same IMP variant within more genetically divergent integron structures (e.g., blaIMP-14 situated within In687 and In1314).
The phylogenetic relationships in Fig. 1 identified all of the 22 K. pneumoniae isolates as K. pneumoniae subsp. pneumoniae. K. pneumoniae isolates consisted of 6 clonal complexes (CCs) and 2 sequence types (STs) (Fig. 2). The most prevalent CCs (with ≥4 isolates) included CC14 (n = 11; from Japan and Philippines) and CC37 (n = 4; from Japan). In1310 with blaIMP-26 from Philippines and In1321 with blaIMP-6 from Japan were present in CC14. CC14 is the only clone with international distribution. ST14 and ST37 are global multidrug-resistant clones and have been associated with the production of AmpC β-lactamases, extended-spectrum β-lactamases (ESBLs), and carbapenemases (15, 16).
OmpK35 and OmpK36 deficiencies and variants are responsible for alterations in porins that contribute to increased MICs to carbapenems (15). The majority of the study isolates had OmpK35 deficiency due to premature stop codons and wild-type OmpK36 (Fig. 2).
K1, K2, K5, K20, K54, and K57 capsular types are associated with community-acquired invasive infections due to K. pneumoniae (17). CC14 isolates from this study were positive for K2 and present in Japan and Philippines (Fig. 2). Brisse et al. reported that CC14-K2 was not associated with rmpA (i.e., regulator of mucoid phenotype) and mouse lethality compared to CC65-K2 (18).
Hypervirulent K. pneumoniae strains often possess siderophore clusters (i.e., yersiniabactin, aerobactin, colibactin, and salmochelin) as well as rmpA or rmpA2 (16). Yersiniabactin, which is encoded by a pathogenicity island that includes ybt, irp12, and fyuA genes (16), was present in isolates from this study belonging to CC14 and ST626 (Fig. 2).
The E. cloacae complex is made up of 13 groups that are difficult to distinguish using phenotypic methods (19). Recent studies showed that E. hormaechei and E. cloacae cluster III are the most prevalent clinical species among the E. cloacae complex (20, 21). E. hormaechei subsp. steigerwaltii is the most prevalent subspecies, followed by E. hormaechei subsp. oharae, while E. hormaechei subsp. hormaechei is generally rare (20, 21).
The E. cloacae complex (n = 9) was the second most common species in our study and consisted of E. hormaechei subsp. steigerwaltii, E. hormaechei subsp. oharae, and E. cloacae cluster III (Fig. 3). In silico multilocus sequence type (MLST) analysis identified 7 CCs among the E. cloacae complex (Fig. 3). E. cloacae cluster III CC78 (with blaIMP-8 from Taiwan) was the most common CC among the E. cloacae complex. Previous molecular epidemiology studies have shown that CC78 includes global clones associated with blaCTX-M-15 or blaVIM-1, especially among European countries (22).
The Citrobacter species isolates (n = 4) included in our study belonged to STs 97 to 100 (Fig. 4). Two isolates (SMART316 and SMART314) were classified as Citrobacter spp. based on the phylogenetic tree constructed with type strains (Fig. 4) (23). The average nucleotide identity (ANI) value between these 2 isolates and the 5 most closely related Citrobacter species (i.e., C. freundii, Citrobacter braakii, Citrobacter werkmanii, Citrobacter youngae, and Citrobacter pasteurii) was <95% (i.e., the cutoff value of species definition) (see Table S3 in the supplemental material). ANI is a promising method of defining species using whole-genome sequencing replacing DNA-DNA hybridization (24).
The phylogenetic relationship of E. coli isolates (n = 3) is shown in Fig. S2 in the supplemental material. Two of the E. coli isolates belonged to the multidrug-resistant ST131 pandemic clone, which has been associated with fluoroquinolone resistance and ESBLs, including the recent acquisition of carbapenemases (25). The blaIMP-14 in one of the ST131 isolates (SMART640) was nested within a 54-kb multidrug resistance region located on an epidemic IncA/C2 plasmid (26).
This study has several limitations. Our collection may not represent the global prevalence of IMP and integron subtypes. We were unable to determine all of the integron structures due to the limitation of short-read sequencing. Long-read sequencing techniques, including the detailed analysis of plasmids, would provide more knowledge on the location, mobile elements, and plasmid backbones of these carbapenemases.
We used whole-genome sequencing with comprehensive molecular analysis to elucidate the global epidemiology at a large scale of blaIMP-containing Enterobacteriaceae and showed that some blaIMP subtypes with associated integrons were present in certain countries within multiple species. Examples include the following: (i) In809 was present in E. hormaechei subsp. steigerwaltii, C. freundii, and a Citrobacter sp. from Australia; (ii) In722 was present in K. pneumoniae and E. coli from Japan; (iii) In687 was identified in K. pneumoniae and E. coli from Thailand; and (iv) In1312 was present in K. pneumoniae from Japan and C. freundii from Brazil. This study identified certain high-risk global clones with specific IMP integrons (i.e., K. pneumoniae ST14 from Philippines contained In1310, K. pneumoniae ST37 from Japan contained In722, In1311, and In1312, and E. cloacae ST78 from Taiwan contained In73).
This study highlights the importance of surveillance programs using the latest molecular techniques for providing insight into the characteristics and the global distribution of CCs and their association with integrons on containing blaIMP genes. Our results suggest that specific blaIMP-containing integrons are circulating locally among different bacteria in countries such as Australia, Japan, and Thailand, while the identification of high-risk clones has the potential to expand the global distribution of CPE. This emphasizes the importance of identifying global types of IMPs among different Enterobacteriaceae species and clones.
We included 38 IMP-producing clinical, nonrepeat Enterobacteriaceae collected from a global surveillance program, namely the Merck Study for Monitoring Antimicrobial Resistance Trends (SMART) (2008 to 2014) (see Data Set S1 in the supplemental material). The SMART program included isolates from intra-abdominal and urinary tract infections. The program collected consecutive clinically relevant Gram-negative aerobes at each institution. These isolates initially underwent microdilution panel susceptibility testing and molecular screening for blaIMP as described previously (25). Overall, 107,366 isolates were obtained from 2008 to 2014; of these, 755 were positive for blaKPC, 281 for blaOXA-48-like, 271 for blaNDM, 89 for blaVIM, and 38 for blaIMP.
We used the Nextera XT DNA sample preparation kit (Illumina, San Diego, CA, USA) to prepare libraries for sequencing. Samples were multiplexed and sequenced on an Illumina NextSeq 500 for 300 cycles (151 bp paired end).
Draft genomes were obtained using SPAdes version 3.8.1 (27). Species identification was performed using SILVA 16S rRNA gene database release 123 (28). In addition, we used the hsp60 gene for identification of Enterobacter spp. (19) and a whole-genome-based phylogenetic tree, including type strains for identification of Klebsiella spp. and Citrobacter spp. Average nucleotide identity (ANI) was calculated using JSpecies (24).
To define the presence of resistance genes other than β-lactamases, plasmid replicons, and virulence genes of Klebsiella spp. and E. coli (detailed below), we used raw sequence data and SRST2 (29) (default settings: thresholds for detection of 90% identity and 90% coverage) in combination with the ARG-ANNOT (30) and PlasmidFinder (31) databases. To perform MLST and to define the presence of β-lactamases and other genes of interest, we used draft genomes and BLAST+ (32) in combination with the following databases or typing schemes: NCBI BLAST database (https://blast.ncbi.nlm.nih.gov/Blast.cgi), NCBI Beta-Lactamase Data Resources (http://www.ncbi.nlm.nih.gov/pathogens/beta-lactamase-data-resources/), plasmid addiction systems (33), and MLST (http://bigsdb.pasteur.fr/klebsiella/; http://pubmlst.org/ecloacae/; http://pubmlst.org/cfreundii/; http://mlst.ucc.ie/mlst/dbs/Ecoli/). We used BLAST+ thresholds of 90% nucleotide identity and 90% coverage, except for the detection of gene alleles in which we used 100% identity (β-lactamases and ompK35 and ompK36 at protein level and the others at nucleotide level).
The goeBURST algorithm implemented in PHYLOViZ software (34) was used to demonstrate relationships between STs and to define the founder of a clonal complex (CC). We defined CCs at the single-locus variant level. We used the assembled contigs with blaIMP to determine their genetic environments. Integrons were classified according to INTEGRALL (http://integrall.bio.ua.pt/), and promoters of gene cassettes were characterized according to a previous study (35). For Klebsiella isolates, we performed in silico detection of the K capsular type based on wzi alleles (17), virulence genes (http://bigsdb.pasteur.fr/klebsiella), and promoters and coding sequences of ompK35 and ompK36 (36, 37). For E. coli isolates, we performed in silico phylogenetic grouping (38), virulence genotyping (39), O:H typing (40), fimH typing (41), and detection of H30Rx-status (42).
We used a core genome single nucleotide polymorphism (SNP)-based approach to create a phylogenetic tree for each Enterobacteriaceae genus. First, we made the reference genome-like pseudochromosomes that contained only SNPs. For study isolates and 6 reference strains downloaded from the NCBI database (see Data Set S2 in the supplemental material) for which complete genomes were not available, SNPs were identified using trimmed reads mapping to a genus-specific reference genome (Data Set S2) followed by GATK Best Practices workflow (43) and SAMtools (44) (depth of sequencing, >10; Phred score, >20). Complete genomes and draft genomes for which raw reads were not available on the NCBI database (Data Set S2) were aligned against the reference genome of the genus using progressiveMauve to obtain pseudochromosomes that contained only SNPs (45). The SNP-only core genome was identified as the common block of >500 bp to all of the study isolates. A maximum-likelihood tree was built using these core genomes and RAxML (46) and visualized using FigTree (http://tree.bio.ed.ac.uk/software/figtree/).
We thank the team of curators of the Institut Pasteur MLST and whole-genome MLST databases for curating the data and making them publicly available at http://bigsdb.web.pasteur.fr/. We thank Thomas Jové from INTEGRALL for curating integrons.
J.D.D.P had previously received research funds from Merck and Astra Zeneca. M.R.M. is an employee of Merck.
This work was supported by the John Mung Program from Kyoto University, Japan (to Y.M.), a research grant from the Calgary Laboratory Services (10015169; to J.D.D.P), federal funds from the National Institute of Allergy and Infectious Diseases, National Institutes of Health, Department of Health and Human Services under award number U19AI110819 (to M.D.A.), and National Institute of Allergy and Infectious Diseases, National Institutes of Health, Department of Health and Human Services award number R01AI090155 (to B.K.).
The funding organizations had no role in study design, data collection and interpretation, or the decision to submit the work for publication.
Supplemental material for this article may be found at https://doi.org/10.1128/AAC.02729-16.