1
Acquired fluoroquinolone resistance genes in corneal isolates of
Pseudomonas aeruginosa
Mahjabeen Khan1, Stephen Summers2, Scott A Rice2,3,4, Fiona Stapleton1, Mark D P
Willcox1, and Dinesh Subedi1,5*
1School of Optometry and Vision Science, University of New South Wales Sydney, Australia
2The Singapore Centre for Environment Life Sciences Engineering (SCELSE), Singapore
3The School of Biological Sciences, Nanyang Technological University, Singapore
4The ithree Institute, The University of Technology Sydney, Sydney, New South Wales,
Australia
5School of Biological Sciences, Monash University, Clayton, Victoria, Australia
Key words: Antimicrobial resistance, Fluroquinolones, Pseudomonas aeruginosa, eye,
corneal ulcer, microbial keratitis
* Corresponding Author: dinesh.subedi@monash.edu
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprintthis version posted May 21, 2020. ; https://doi.org/10.1101/2020.05.17.100396doi: bioRxiv preprint
2
Abstract
Fluroquinolones are widely used as an empirical therapy for pseudomonal ocular infections. Based on increasing
reports on acquired fluroquinolone resistance genes in clinical isolates of Pseudomonas aeruginosa, we
investigated 33 strains of P. aeruginosa isolated from the cornea of microbial keratitis patients in India and
Australia between 1992 and 2018 to understand the prevalence of acquired fluroquinolone resistance genes in
ocular isolates and to assess whether the possession of those genes was associated with fluoroquinolone
susceptibility. We obtained the whole genome sequence of 33 isolates using Illumina MiSeq platform and
investigated the prevalence of two fluoroquinolone resistance genes crpP and qnrVC1. To examine the associated
mobile genetic elements of qnrVC1 positive strains, we obtained long read sequences using Oxford Nanopore
MinION and performed hybrid assembly to combine long reads with Illumina short sequence reads. We further
assessed mutations in QRDRs and antibiotic susceptibilities to ciprofloxacin, levofloxacin and moxifloxacin to
examine the association between resistance genes and phenotype. Twenty strains possessed crpP in genetic islands
characterised by possession of integrative conjugative elements. The qnrVC1 gene was carried by four isolates on
class I integrons and Tn3 transposons along with aminoglycoside and beta-lactam resistance genes. We did not
observe any evidence of plasmids carrying fluroquinolone resistance genes. Resistance to fluroquinolones was
observed in those strains which possessed crpP, qnrVC1 and that had QRDRs mutations. The presence of crpP
was not a sole cause of fluroquinolone resistance.
Introduction
Pseudomonas aeruginosa is a highly adaptable opportunistic pathogen which is ubiquitously
present in the environment. This bacterium is naturally resistant to many antimicrobials and
can acquire antibiotic resistance through mutations in chromosomal genes and lateral gene
transfer [1, 2]. P. aeruginosa is associated with different types of human infections and because
of emerging multidrug-resistant strains, these infections are major global public health
concerns [3].
Fluoroquinolones are broad spectrum and widely prescribed antibiotics to treat pseudomonal
infections including ocular infections [4-6]. Fluoroquinolone resistance in various clinical
isolates is on the rise [7]. For example, a single centre study has shown that the prevalence of
fluoroquinolone resistance P. aeruginosa increased from 15% to 41% in ten years [8]. This
increase in fluoroquinolone resistance has been linked to the excessive use of the antibiotics
[9]. The rate of isolation of fluoroquinolone resistant strains also depends on the type of
infections; nosocomial isolates are more resistant than isolates from non-nosocomial sources
[10]. In general, fluoroquinolone resistance is relatively low in ocular isolates of P. aeruginosa
compared to other infections [11]. However, higher resistance rates have been reported in
ocular isolates in certain regions of the world and like other systemic infections this rate has
been increasing over time [12]. This has raised the concerned that the horizontal transfer of
fluoroquinolone resistance genes can be associated with spread of fluoroquinolone resistance
in ocular isolates.
Mutations that alter target sites (DNA gyrase [gyrA/gyrB] and topoisomerase IV [parC/parE])
and increased membrane permeability are common mechanisms of fluoroquinolone resistance
in P. aeruginosa [1, 13-15]. In other Gram-negative bacteria, fluoroquinolone resistance genes
such as qnr have been shown to be carried on plasmids [16, 17]. The gene encodes a
pentapeptide repeat protein which protects DNA gyrase and topoisomerase from the action of
fluroquinolones [18]. In contrast, carriage of qnr on plasmid is very rare in P. aeruginosa [19].
Despite this low carriage, recent studies have identified fluoroquinolone resistance genes in
certain mobile genetic elements. CrpP and qnrVC can be carried on the P. aeruginosa mega
plasmids pUM505 and pBM413, respectively [20, 21]. In addition, qnrVC was found to be
associated with a class I integron, which also carried beta-lactamase genes [20, 22]. These
studies were based on a single strain and the prevalence of acquired fluoroquinolone resistance
genes in P. aeruginosa remains unclear. This led us to examine mobile fluoroquinolone
resistance genes in P. aeruginosa. Given the concern of increasing fluoroquinolone resistance
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprintthis version posted May 21, 2020. ; https://doi.org/10.1101/2020.05.17.100396doi: bioRxiv preprint
3
in ocular isolates, we have undertaken sequence analysis of 33 ocular isolates of P. aeruginosa,
isolated from corneal ulcers in the last 25 years to assess whether the possession of
fluoroquinolone resistance genes was associated with fluoroquinolone susceptibility and
whether resistance genes carrying strains had any genetic commonalities.
Materials and Methods
Pseudomonas aeruginosa strains
Isolates in this study were collected from the cornea of microbial keratitis patients in India and Australia between 1992 and
2018. They comprise 20 isolates collected for this study and used for the antibiotic susceptibility testing and whole genome
sequencing. Information on antibiotic susceptibility and the whole genome data of 13 isolates was obtained from our previous
study [23] (Supplementary Table 1). All isolates were retrieved from the culture collection of School of Optometry and Vision
Science, the University of New South Wales, Australia without identifiable patients’ data.
Antibiotic susceptibility testing of Pseudomonas aeruginosa strains
Susceptibility of P. aeruginosa isolates to ciprofloxacin (Sigma-Aldrich, Inc., St. Louis Missouri, USA), levofloxacin (Sigma-
Aldrich, Inc.,) and moxifloxacin (European Pharmacopoeia, Strasbourg, Cedex France) was investigated using the broth
microdilution method following the protocol of Clinical and Laboratory Standard Institute [24] [25]. The MIC was taken as
the lowest concentration of an antibiotic in which no noticeable growth (turbidity) was observed and the break point was
established according to published standards [26]. Based on MICs, the isolates were categorised into four groups, susceptible
( resistance break point), resistant (> resistance break point – 32 g/mL), highly resistant (> 32 – 128 g/mL) and very highly
resistant (>128 g/mL) for the analysis.
DNA extraction and Illumina sequencing
Bacterial DNA was extracted from overnight cultures using a DNeasy Blood and Tissue Kit (Qiagen, Hilden, Germany)
following the manufacturer’s instructions. The extracted DNA was quantified and purity-checked using Nanodrop (NanoDrop
Technologies, Wilmington, DE, USA), Qubit fluorometer (Life Technologies, Carlsbad, CA, USA), and 1% agarose gel
electrophoresis. The extracted DNA was dried for transport to the sequencing facility at Singapore Centre for Environmental
Life Sciences Engineering, Singapore. DNA was sequenced on MiSeq (Illumina, San Diego, CA, USA) generating 300 bp
paired end reads. The paired-end library was prepared using Nextera XT DNA library preparation kit (Illumina, San Diego,
CA, USA). All the libraries were multiplexed on one MiSeq run.
Bioinformatics analysis of short read sequences and construction of phylogenies
The quality of raw reads was analysed using online tool FastQC v0.117 (https://www.bioinformatics.babraham.ac.uk/projects/fastqc).
Adaptor sequences were removed using Trimmomatic v0.38 with quality and length filtering (SLIDINGWINDOW:4: 15
MINLEN:36) [27]. The reads were de-novo assembled using Spades v3.13.0 with the programs’ default setting [28] followed
by annotation using Prokka v1.12 [29]. To examine acquired resistance genes, the genomes were uploaded into online database
ResFinder v3.1 of Centre for Genomic Epidemiology, DTU, Denmark [30]. The contigs carrying crpP and qnrVC1 genes
were selected and analysed for mobile genetic elements using BLAST, Integron finder v1.5.1 [31] and IS finder [32]. The
genome was visualised and manually modified using Geneious prime v2019.2 [33]. A figure of BLAST comparison was
generated using EasyFig v2.2.2 [34]. To identify mutations in the QRDRs (gyrA, gyrB, parC and parE), the genome sequences
were analysed using Snippy v4.2 with program’s default settings (https://github.com/tseemann/snippy). The genomes were
also examined for the presence of type IV secretion factors (exoU and exoS) using BLAST search, and for the presence of the
CRISPR cas system using the CRISPRcasFinder online tool [35]. Codon adaptation index (CAI) [36] was examined to
understand the possible expression of different orthologues of crpP using CAIcal [37, 38]. 50s ribosomal protein L19 (rplS),
which is a highly expressed chromosomal gene was included as the reference to show the difference in expression level
between chromosomal gene and acquired genes.
Core genome single nucleotide polymorphisms (SNPs) were identified using Parsnp v from Harvest Suite [39] using settings
to exclude SNPs identified in regions that had arisen by recombination. The core genome SNPs were used to construct a
maximum likelihood phylogenetic tree. All genomes were examined for multi-locus sequence type (MLST) using the MLST
database. Nucleotide sequences of all MLST locus were extracted and concatenated to use in Bayesian phylogenetic analyses
using BEAST2 v2.4.7 with the following parameters: gamma site heterogeneity model, Hasegawa-Kishino-Yano (HKY)
substitution model and relaxed-clock log-normal [40]. BEAST 2 output was summarized using TreeAnnotator with a 5% burn
in. The phylogenetic trees were visualised using iTol v4 [41].
MinION sequencing and analysis
Wizard Promega DNA extraction kits (Promega, Madison, WI) was used to extract DNA from overnight broth culture. DNA
was quantified and transported as mentioned above. Long reads libraries were prepared using a rapid sequencing kit (RAD-
SQK004, ONT, Oxford, UK) and subsequent sequencing was conducted using the MinIon flowcell (R9.4.1) for 48 hours.
Long reads were basecalled using Guppy v3.3.0 and adapters removed using Porechop v0.2.4. Assembly of both long and
short reads into a hybrid genome assembly was achieved with Unicycler v0.4.3 [42], opting for default parameters and using
only short reads with merging pairs. Following assembly, all assemblies were assessed for quality using Quast v5.0.2.
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprintthis version posted May 21, 2020. ; https://doi.org/10.1101/2020.05.17.100396doi: bioRxiv preprint
4
Results
Population structure, phylogeny and fluoroquinolone resistance of P. aeruginosa strains
Whole genomes were analysed from 33 corneal isolates of P. aeruginosa, 20 of which were
sequenced as a part of this study and 13 genome that were published previously [23]. All strains
were isolated between 1992 and 2018 in India (19 isolates) or Australia (14 isolates)
(Supplementary Table 1). Draft genomes were mapped against the reference genome P.
aeruginosa PAO1 and a total of 202,232 SNPs were observed among the 33 isolates, which
were used to construct core genome phylogeny using Parsnp v1.2 [43]. Similar to previous
reports [23, 36, 44, 45], the phylogenetic tree based on the core genome revealed two major
clades (Fig 1). These clades followed a similar pattern as for other previously described strains
of P. aeruginosa, where exoU carrying strains clustered together in a single clade (Phylogroup
2) [23]. Our results showed that 18 (of 33) isolates carried exoU, of which 16 were clustered
together in phylogroup 2, which contained predominantly Indian isolates. The detection rate of
exoU was 68% in Indian isolates and 36% in Australian isolates.
Figure 1. Maximum likelihood phylogenetic tree based on core genome SNPs analysis using
Pseudomonas aeruginosa PAO1 as the reference, excluding SNPs identified in regions that had arisen by
recombination, using the default parameters of Parsnp v1.2 [43]. Isolates from India are labelled red and
Australian isolates are labelled blue. Numbers given at the nodes represent bootstrap values. The presence
of crpP, exoU, qnrVC1, and CRISPR cas are represented by red squares. Orange squares represent the
presence of mutations in the quinolone resistance determining regions (QRDRs) and fluoroquinolone (CIP
= Ciprofloxacin; LEVO = Levofloxacin; and MOX = Moxifloxacin) susceptibilities are shown as a heat
map with the ranges indicated in the figure. The figure was drawn using iTol v4 [41].
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprintthis version posted May 21, 2020. ; https://doi.org/10.1101/2020.05.17.100396doi: bioRxiv preprint
5
We further examined acquired fluoroquinolone resistance genes, mutations in quinolone
resistance determining region (QRDRs), the presence of CRISPR genes and susceptibility to
three different fluoroquinolones (Fig 1). None of the Australian isolates regardless of
phylogenic grouping were resistant to fluoroquinolones. In contrast, 75% of isolates of
phylogroup 2 were resistant to at least one fluoroquinolone. Of the 33 strains, 73.7 % (14 out
of 19) of Indian and 42.8% (6 out of 14) of Australian isolates possessed crpP, which has been
recently shown to be on a plasmid (pUM505) and associated with ciprofloxacin resistance [20].
However, eight (40%) crpP carrying strains in the current study regardless of country of
isolation were not resistant to the fluoroquinolones, including ciprofloxacin (Fig 1). Eleven out
of 14 fluoroquinolone resistance strains had mutations in both gyrA and parC and all except
one carried the exoU gene. In addition, four strains from the latter cohort of 11 strains carried
another fluoroquinolone resistance gene, qnrVC1, in combination with mutations in gyrA and
parC, and this was associated with a very high MIC (>128 µg/mL) to all three fluoroquinolones
(Fig 1).
Given that exoU, crpP and qnrVC1 are components of the accessory genome which is mostly
shaped by the CRISPR-Cas system, a bacterial defence system against foreign DNA [46], we
searched isolates for CRISPR-Cas genes using the CRISPRCasFinder database [35] following
software default parameters. Only one of the exoU, and none of the qnrVC1 carrying strains,
were positive for CRISPR-Cas genes. However, CRISPR-Cas was observed in seven crpP
positive strains, which did not possess exoU and/or qnrVC1 (Fig 1).
Figure 2. Consensus tree of 33 P. aeruginosa isolates, based on Bayesian evolutionary analysis by
sampling trees (BEAST) of concatenated multi-locus sequence type (MLST) under strict clock analysis
[40]. The tip of the tree was constrained by date of isolation. The time scale is shown in years at the top
and each internal node is labelled with posterior probability limit. Isolates from India are labelled red and
Australian isolates are labelled blue. The presence of genes crpP, exoU, qnrVC1, and CRISPRcas are
represented by red squares. Orange square represents presence of mutations in quinolone resistance
determining region (QRDRs), Fluoroquinolone (CIP = Ciprofloxacin; LEVO = Levofloxacin; and MOX
= Moxifloxacin) susceptibilities are shown as heat maps in the grey scale indicated in the figure. The
figure was drawn using iTol v4 [41].
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprintthis version posted May 21, 2020. ; https://doi.org/10.1101/2020.05.17.100396doi: bioRxiv preprint
![Figure 5. Comparison of qnrVC1-associated genomic islands of P. aeruginosa strains. Protein-coding regions are represented by the arrows and common key features/associated genes among all strains are shown in various coloured arrows. The gradient blue and red shading represent regions of nucleotide sequence identity (100% to 65%) in forward and reverse directions, respectively, determined by BLASTn analysis. The sequences are from strains top to bottom; PA198, PA219, PA202 and PA221. Figures are drawn to scale using Easyfig [34]. (Tn3-tnp = Tn3-transposase HP= hypothetical protein, ydhC= Inner membrane transport protein, tet(R)-tet(g) = tetracycline resistance genes, qnrVC1=quinolone resistance gene, VOC = VOC family protein, aph(6)-Id = aminoglycoside resistance protein, DNA-inv = DNA invertase, aadA10= aminoglycoside resistance protein, IS110= IS110 family transposase, qacEdealta1= quaternary ammonium compound-resistance protein, sul1= Dihydropteroate synthase, blaLCR-1=betalactamase gene, attC = recombination sites of gene cassette, attI = integron recombination site, IR=invert repeat).](/figures/figure-5-comparison-of-qnrvc1-associated-genomic-islands-of-1uogchkk.webp)
![Figure 1. Maximum likelihood phylogenetic tree based on core genome SNPs analysis using Pseudomonas aeruginosa PAO1 as the reference, excluding SNPs identified in regions that had arisen by recombination, using the default parameters of Parsnp v1.2 [43]. Isolates from India are labelled red and Australian isolates are labelled blue. Numbers given at the nodes represent bootstrap values. The presence of crpP, exoU, qnrVC1, and CRISPR cas are represented by red squares. Orange squares represent the presence of mutations in the quinolone resistance determining regions (QRDRs) and fluoroquinolone (CIP = Ciprofloxacin; LEVO = Levofloxacin; and MOX = Moxifloxacin) susceptibilities are shown as a heat map with the ranges indicated in the figure. The figure was drawn using iTol v4 [41].](/figures/figure-1-maximum-likelihood-phylogenetic-tree-based-on-core-twtxvxs9.webp)
![Figure 2. Consensus tree of 33 P. aeruginosa isolates, based on Bayesian evolutionary analysis by sampling trees (BEAST) of concatenated multi-locus sequence type (MLST) under strict clock analysis [40]. The tip of the tree was constrained by date of isolation. The time scale is shown in years at the top and each internal node is labelled with posterior probability limit. Isolates from India are labelled red and Australian isolates are labelled blue. The presence of genes crpP, exoU, qnrVC1, and CRISPRcas are represented by red squares. Orange square represents presence of mutations in quinolone resistance determining region (QRDRs), Fluoroquinolone (CIP = Ciprofloxacin; LEVO = Levofloxacin; and MOX = Moxifloxacin) susceptibilities are shown as heat maps in the grey scale indicated in the figure. The figure was drawn using iTol v4 [41].](/figures/figure-2-consensus-tree-of-33-p-aeruginosa-isolates-based-on-3rwqoqu2.webp)