Comparison of typing by whole genome sequencing (WGS) and pulsed field gel electrophoresis (PFGE) of isolates from a hospital outbreak with a CTX-M-15 producing Klebsiella pneumoniae

Objective: Klebsiella pneumoniae (KP) is the archetype of a Gram negative outbreak organism with acquired antibiotic resistance. Reference epidemiology typing method have for many years been PFGE. The ultimate typing tool must however be knowing the entire variation within the genomes of the involved organisms.

Recent technological development and pricing have made this possible. Our objective was to compare typing by WGS with PFGE on nosocomial KP outbreak isolates. Methods: 44 KP isolates from 2006 (pre-outbreak) through 2007/8 (outbreak) to 2011 (endemic) from 33 patients and 5 KP reference strains, were investigated.

The 44 isolates were phenotypic similar; ESBL-producers with reduced susceptibility to gentamicin and ciprofloxacin. PFGE was performed using XbaI; data handling in BioNumeric. For WGS a reference genome (outbreak isolate 2006-1-264) was assembled to a single scaffold using data from two platforms, Illumina (200x) and 454 (10x) and the program Consed (v23).

Reads were trimmed (AdapterRemoval v1.1.) and mapped to the reference genome using Bowtie (v2.0). Single Nucleotide Polymorphisms (SNPs) were called using Samtools. To be considered valid the depth of each SNP position was ≥ 10, and mapping quality was ≥ 20. Pruning of all SNPs within 10bp of each other were done.

All SNPs were concatenated into a multiple alignment and phylogeny was inferred using FastTree; tree was visualized using FigTree (v1.4.0). Results: Both WGS and PFGE divided the 44 KP isolates in a group of 37 outbreak isolates and 7 singletons. Number of SNPs between outbreak isolates and reference strains, and within the outbreak isolates, were up to 23096 and 64 (range: 2-64), respectively.

Band differences within outbreak group were 0-5. For single PFGE event no clear correlation to number of SNPs were seen; one pair of isolates differed by 4 band and 2 SNPs, while another pair differed by 0 band and 11 SNPs. The relation over time for consecutive isolates from the same patients were complex, but with a trend towards more SNPs over time.

Conclusion: In general full agreement between WGS and PFGE was seen. WGS has higher resolution and are able to discriminate between isolates of same PFGE type.