Genetic Basis of Emerging Vancomycin, Linezolid, and Daptomycin Heteroresistance in a Case of Persistent Enterococcus faecium Bacteremia

ABSTRACT Whole-genome sequencing was used to examine a persistent Enterococcus faecium bacteremia that acquired heteroresistance to three antibiotics in response to prolonged multidrug therapy. A comparison of the complete genomes before and after each change revealed the emergence of known resistance determinants for vancomycin and linezolid and suggested that a novel mutation in fabF, encoding a fatty acid synthase, was responsible for daptomycin nonsusceptibility. Plasmid recombination contributed to the progressive loss of vancomycin resistance after withdrawal of the drug.


Bacterial Isolates, Species Identification and Antibiotic Susceptibility Testing
Bacterial isolates were recovered by sampling from the original blood cultures in the Mount Sinai Hospital Clinical Microbiology Laboratory (CML), and stored in tryptic soy broth with 15% glycerol at -80°C. VITEK 2 (bioMérieux) automated broth microdilution antibiotic susceptibility profiles were obtained for each isolate in the CML according to Clinical and Laboratory Standards Institute (CLSI) 2015 guidelines and reported according to CLSI guidelines (1). Species confirmation was performed with MALDI-TOF (Bruker Biotyper, Bruker Daltonics). For further processing, all isolates were cultured on tryptic soy agar (TSA) plates with 5% sheep blood (blood agar) (ThermoFisher Scientific) under nonselective conditions. Minimum inhibitory concentrations (MIC) were determined in duplicate for vancomycin and linezolid using E-tests (bioMérieux) performed on Mueller Hinton Agar plates (Sigma-Aldrich) and a 0.5 McFarland standard, and for daptomycin using Sensititre GPN3F susceptibility plates (ThermoFisher Scientific) and a cation-adjusted Mueller-Hinton broth.

DNA Preparation and sequencing
Seven strains derived from single colonies of six isolates were selected for sequencing based on their susceptibility profiles. For the hetero-resistant patient isolate E, four single colonies were selected and grown separately on Blood Agar plates under nonselective conditions. E-tests were then performed, and a susceptible and a resistant strain were selected for sequencing. All seven strains were grown overnight on Blood Agar plates, followed by DNA extraction using the Qiagen DNeasy Blood & Tissue Kit (Qiagen, 69504) according to the manufacturer's instructions, with modified lysis conditions as follows. Bacterial cells were lysed by suspending cells in 3 µL of 100mg/ml RNase A (Ambion, AM2286) and 10 µL of 100mg/ml lysozyme (Sigma, L1667-1G) for 30 minutes at 37°C, followed by incubation with Proteinase K for one hour at 56°C and two rounds of bead beating of 1 min each using 0.1mm silica beads (MP Bio).
Quality control, DNA quantification, and gDNA library preparation for single molecule real-time (SMRT) sequencing was performed as previously described (2). Briefly, DNA was gently sheared into ~20,000 bp fragments using Covaris G-tube spin columns, and end-repaired before ligating SMRTbell adapters (Pacific Biosciences). The resulting library was treated with an exonuclease cocktail to remove un-ligated DNA fragments, followed by two additional (Ampure XP) purifications steps and Sage Science BluePippin size selection to deplete SMRTbells < 7,000 bp. Libraries were then sequenced on the Pacific Biosciences (PacBio) RS II platform using the P5-C3 sequencing enzyme and chemistry.
Illumina sequencing was performed for assembly finishing. Genomic DNA (1 µg) was sheared to an average fragment size of 200 bp using a Bioruptor Pico sonicator (Diagenode). Amplicon sequence libraries were prepared using the end repair, A-tailing, and adaptor ligation NEBNext DNA library prep modules for Illumina from New England Biolabs, according to the manufacturer's protocol. Following final purification with Ampure XP beads and secondary PCR (8 cycles) to introduce barcoded primers, multiplexed libraries were sequenced on the Illumina HiSeq 2500 platform in a single-end 100 nt-run format to >30x genomic coverage.

Complete genome assembly and finishing
PacBio sequencing data were assembled using HGAP3 version 2.2.0, (3). Illumina reads were then mapped to the curated PacBio assemblies and consensus calling was performed using the mpileup function of samtools (4) to correct errors in the PacBio assembly, which mainly consisted of insertion/deletion errors in homopolymer regions (Table S1). Genome circularization, curation and annotation were performed using a custom post-assembly pipeline (https://github.com/powerpak/pathogendb-pipeline) (2). Replicon and origin of replication were identified with PlasmidFinder 1.3 (5) and insertion elements were mapped using ISfinder (6).

Pairwise Genome Comparisons and Phylogenetic Analyses
Multi-locus sequence types (MLST) of each strain genome were determined using the E. faecium PubMLST database (http://pubmlst.org/efaecium/). Single nucleotide and structural variants between strains were identified by comparing each genome to the first isolate strain using NUCmer (Version 3.1) (7). Nonsynonymous sequence changes in predicted ORFs were confirmed by multiple sequence alignments using MUSCLE (8).
For phylogenetic analyses, 761 complete genome, chromosome, scaffold and contig E. faecium sequences were downloaded from GenBank on 1 July, 2017. Phylogenetic analysis of all ST736 genomes was performed using Parsnp, part of the Harvest bioinformatics tool suite (Version 1.1.2), using default MUMi (9) <0.01 cutoff settings and filtering SNPs located in regions of recombination (10). For plasmid recombination analyses, shared short homologous regions were identified by BLAST+ (11) and Contiguity (12).

Antibiotic Resistance Gene Annotation
Antibiotic resistance gene and variants were annotated by comparing to a manually curated database of known E. faecium resistance determinants from literature. BLASTP was used to identify the presence of genes in each strain genome, with sequence identity cutoff ≥90% and an e-value cut-off ≤ 1e-10. Resistance variants were identified by BLASTP alignment relative to the wild-type gene protein sequence. Only exact matches to variants identified in literature were considered. An analogous approach with BLASTN was used to identify 23S rRNA variants on nucleotide sequences.

Accession numbers
Genome sequences have been deposited in Genbank, Bioproject accession number PRJNA407447 (http://www.ncbi.nlm.nih.gov/bioproject/407447).  Figure S1. Maximum likelihood phylogenetic tree based on core genome SNVs for case strain genomes and 34 draft or complete ST736 E. faecium genomes present in Genbank. The 34 ST736 genomes had a calculated 75% shared core genome. SNV distances are indicated relative to the first patient strain A. Figure S2. Circos plots of whole-genome alignments for all case strains.