KPC and NDM-1 Genes in Related Enterobacteriaceae Strains and Plasmids from Pakistan and the United States

To characterize the genomic context of New Delhi metallo-β-lactamase-1 (NDM-1) and Klebsiella pneumoniae carbapenemase (KPC), we sequenced 78 Enterobacteriaceae isolates from Pakistan and the United States encoding KPC, NDM-1, or no carbapenemase. High similarities of the results indicate rapid spread of carbapenem resistance between strains, including globally disseminated pathogens.

slices corresponding to DNA of 250-300bp using a QIAGEN MinElute Gel Extraction kit per the manufacturer's protocol. The purified DNA was enriched by PCR using 12.5µL 2X Phusion HF Master Mix and 1µL of 10µM Illumina PCR Primer Mix in a 25µL reaction using 1µL of purified DNA as template. DNA was amplified at 98°C for 30 seconds followed by 18 cycles of 98°C for 10 seconds, 65°C for 30 seconds, 72°C for 30 seconds with a final extension of 5min. at 72°C. Afterwards, the DNA concentration was measured using the Qubit fluorometer and

Genome Sequence Assembly
All sequencing reads were de-multiplexed by barcode into separate genome bins. Reads were quality trimmed to remove adapter sequence and bases on either end with a quality score below 19. Any reads shorter than 31bp after quality trimming were not used in further analysis.
The best reference sequence was chosen for each isolate or plasmid by mapping 10000 reads chosen randomly from that isolate against all reference genomes (from NCBI Genome, downloaded July 14 th 2014) of the same species as the isolate (in the case of genomic DNA assembly) or against all plasmid sequences containing NDM-1, KPC, or CTX-M (in the case of plasmid DNA assembly). Reads were mapped using Bowtie 2 (4) (command: bowtie2 -x <reference_genome_index_name> -1 <forward_read_file> -2 <reverse_read_file> -q --phred33 --very-fast -I 100 -X 600 --no-discordant --no-mixed --no-unal --no-hd --no-sq --omit-secstrand). The genome or plasmid against which the highest percentage of reads mapped was used as the reference sequence for that assembly. It was empirically determined that if this first mapping included fewer than 60% of the reads, then the assembly would be best done completely de novo. For isolates with >60% of reads matching a reference sequence, all reads were mapped to that sequence (command: bowtie2 -x <reference_genome_index_name> -1 <forward_read_file> -2 <reverse_read_file> -q --phred33 --very-sensitive-local -I 200 -X 1000 -S <sam_output>). Variants from the reference were called using samtools (commands: samtools view -buS <sam_file> | samtools sort -m 4000000000 -<sample_prefix> ### samtools index <bam_file> ### samtools mpileup -uD -f <reference_genome> <bam_file> | bcftools view -bcv -> <bcf_file> ### bcftools view <bcf_file>). The variant call format file was then filtered to remove SNPs with a quality score lower than 70 or coverage greater than twice the average coverage expected per base. Custom scripts were then used to extract DNA sequences from the reference genome with > three independent reads, to create a fragment file of regions in the sample genome matching the reference genome modified with high-quality variant information.

ORF Prediction and Annotation
ORF prediction for each genome was performed separately using GeneMark (6) Figure 1, panels A,B).

Specific β-lactamase Identification
A BLAST database was constructed from the amino acid sequences of all β-lactamases cataloged in the Bush and Jacoby (13) database at www.lahey.org/Studies/ (accessed March 25 th , 2014). Genes from our genomes annotated as β-lactamases were extracted and compared against this database by BLAST. Exact matches were then re-annotated with their specific β-lactamase name, while inexact matches were recorded as their closest hit plus an asterisk.

Plasmid Comparisons
We compared plasmid sequences by an all-against-all pairwise nucleotide BLAST alignment. For each pair of plasmids, we calculated the percentage of each plasmid that aligned at >99% identity. We then binned the percentages from each pairwise alignment into groups based on the defining β-lactamase of their query and subject plasmids. We also generated network diagrams from the pairwise BLASTs using custom Python scripts and Cytoscape (14), only including regions above 99% identity and over 500 bp.

Isolate Characteristics
The sampled Enterobacteriaceae isolates are phylogenetically diverse and include multiple examples of known pathogenic sequence types. We performed WGS of each isolate, totaling 33 Escherichia coli isolates, 30 Klebsiella pneumoniae isolates, 9 Enterobacter cloacae complex isolates, and 6 Enterobacter aerogenes isolates. We then used whole genome alignment of the core genomes of each species to reconstruct the phylogenetic relationships of each isolate at high resolution (Technical Appendix Figure 2). The species trees demonstrate that we sampled genomes from a variety of evolutionary clades as well as from multiple members of specific clades. They also demonstrate that clades could include isolates from both the United States and ST131 is noted for its virulence as well as for its frequent association with the CTX-M βlactamases and fluoroquinolone resistance (15)(16)(17). Previous reports have found ST258, and closely related K. pneumoniae, to have relatively high rates of carbapenemase carriage (18,19).
Despite the utility of MLST-based classification for large-scale epidemiological purposes, binning clinical isolates into sequence types masks genotypic and phenotypic variation due to HGT or single nucleotide polymorphisms, and therefore MLST cannot be used for fine-grained epidemiology or as an accurate predictor of antibiotic susceptibility. For example, two previous studies have shown that ST131 can be subdivided into three distinct lineages with different rates of antibiotic resistance (15,16). One of these studies found that the rapid global expansion of ST131 has been driven by the success of a specific subclone of ST131 that encodes fluoroquinolone resistant gyrA and parC alleles and CTX-M-15 (16), a characterization which fits 7 of our 11 ST131 isolates. We also identified a single ST131 isolate carrying KPC-2, which was resistant to all β-lactams tested. We also observed K. pneumoniae ST11 isolates carrying KPC-3, and others carrying NDM-1, which fits with reports characterizing ST11 as being highly common worldwide and frequently encoding carbapenemases (18,19).

Antibiotic Resistance Phenotypes
To establish the overall susceptibility profiles of each of our strains, we performed phenotypic tests using Kirby Bauer Disk diffusion in accordance with CLSI guidelines on all 78 clinical isolates against 12 antibiotics including 7 β-lactams (Technical Appendix Table 1). We found that 63% of all isolates were resistant to ciprofloxacin, a fluoroquinolone commonly used to treat urinary tract infections. We also found resistance to trimethoprim-sulfamethoxazole in 65% of isolates, and gentamicin, doxycycline, and chloramphenicol exhibited in vitro resistance in 45%, 54%, and 56% of isolates, respectively. In the β-lactams, we saw near universal resistance to ampicillin (96% of isolates) and variable resistance to the cephalosporins. A high rate of resistance to meropenem was observed (31% of isolates), but this finding was not surprising since this was the property on which many of the isolates had been selected.