A preview of this full-text is provided by American Society for Microbiology.
Content available from Applied and Environmental Microbiology
This content is subject to copyright. Terms and conditions apply.
| Spotlight Selection | Microbial Genetics | Full-Length Text
Evolutionary genomic analyses of canine E. coli infections
identify a relic capsular locus associated with resistance to
multiple classes of antimicrobials
Kristina Ceres,1 Jordan D. Zehr,1 Chloe Murrell,1 Jean K. Millet,2 Qi Sun,1 Holly C. McQueary,1 Alanna Horton,1 Casey Cazer,1 Kelly
Sams,1 Guillaume Reboul,1 William B. Andreopoulos,3 Patrick K. Mitchell,1 Renee Anderson,1 Rebecca Franklin-Guild,1 Brittany
D. Cronk,1 Bryce J. Stanhope,1 Claire R. Burbick,4 Rebecca Wolking,4 Laura Peak,5 Yan Zhang,6 Rebeccah McDowall,7 Aparna
Krishnamurthy,7 Durda Slavic,7 Prabhjot kaur Sekhon,8 Gregory H. Tyson,9 Olgica Ceric,9 Michael J. Stanhope,1 Laura B. Goodman1
AUTHOR AFFILIATIONS See aliation list on p. 17.
ABSTRACT Infections caused by antimicrobial-resistant Escherichia coli are the leading
cause of death attributed to antimicrobial resistance (AMR) worldwide, and the known
AMR mechanisms involve a range of functional proteins. Here, we employed a pan-
genome wide association study (GWAS) approach on over 1,000 E. coli isolates from sick
dogs collected across the US and Canada and identied a strong statistical association
(empirical P < 0.01) of AMR, involving a range of antibiotics to a group 1 capsular
(CPS) gene cluster. This cluster included genes under relaxed selection pressure, had
several loci missing, and had pseudogenes for other key loci. Furthermore, this cluster
is widespread in E. coli and Klebsiella clinical isolates across multiple host species. Earlier
studies demonstrated that the octameric CPS polysaccharide export protein Wza can
transmit macrolide antibiotics into the E. coli periplasm. We suggest that the CPS
in question, and its highly divergent Wza, functions as an antibiotic trap, preventing
antimicrobial penetration. We also highlight the high diversity of lineages circulating in
dogs across all regions studied, the overlap with human lineages, and regional preva
lence of resistance to multiple antimicrobial classes.
IMPORTANCE Much of the human genomic epidemiology data available for E. coli
mechanism discovery studies has been heavily biased toward shiga-toxin producing
strains from humans and livestock. E. coli occupies many niches and produces a wide
variety of other signicant pathotypes, including some implicated in chronic disease.
We hypothesized that since dogs tend to share similar strains with their owners and are
treated with similar antibiotics, their pathogenic isolates will harbor unexplored AMR
mechanisms of importance to humans as well as animals. By comparing over 1,000
genomes with in vitro antimicrobial susceptibility data from sick dogs across the US and
Canada, we identied a strong multidrug resistance association with an operon that
appears to have once conferred a type 1 capsule production system.
KEYWORDS antibiotic resistance
Resistance has now been documented for nearly every antimicrobial that has
ever been developed (1), and many genes and mutations from numerous patho
gens have been identied that confer such resistance. There are still, however, many
unidentied genes and variants present in all or most bacterial pathogens. One
explanation for this gap is that the databases of isolate genotypes and phenotypes
used to validate antimicrobial resistance (AMR) prediction algorithms are heavily biased
August 2024 Volume 90 Issue 8 10.1128/aem.00354-24 1
Editor Edward G. Dudley, The Pennsylvania State
University, University Park, Pennsylvania, USA
Address correspondence to Laura B. Goodman,
laura.goodman@cornell.edu.
The authors declare no conict of interest.
See the funding table on p. 18.
The views expressed in this article are those of the
authors and do not necessarily reect the ocial policy
of the Department of Health and Human Services,
the U.S. Food and Drug Administration, or the U.S.
Government. Reference to any commercial materials,
equipment, or process does not in any way constitute
approval, endorsement, or recommendation by the
Food and Drug Administration.
Received 28 February 2024
Accepted 8 June 2024
Published 16 July 2024
This is a work of the U.S. Government and is not
subject to copyright protection in the United States.
Foreign copyrights may apply.
toward bacterial isolates recovered from food production animals and human blood
stream infections.
Companion animals may be exposed to antimicrobial-resistant bacteria from their
owners, hospitals, food, or environment, allowing sentinel surveillance of zoonotic
infections that are not being captured in human- and livestock-focused approaches.
Now, after 6 years of surveillance conducted by the Veterinary Laboratory Investigation
and Response Network (Vet-LIRN) (2), an unprecedented set of high-quality complete
genomes and AMR phenotypes of pathogenic Escherichia coli in dogs is available. Recent
studies of extraintestinal pathogenic E. coli (ExPEC) in dogs from North America (3)
and Australia (4) found that dogs and humans are both colonized by a broad range of
the same genotypes, which the authors suggest is likely a consequence of shared gut
carriage of bacteria strains between dogs and humans because of close proximity and
possibly shared diets. Elankumaran et al. identied certain genotypes as likely “spill-
back” zoonotic infections. Harrison et al. found that the abundance of AMR determinants
in companion animal strains was higher than in those from food production animals.
However, neither study compared the genomes and phenotypes for the discovery of
novel AMR determinants.
GWAS methods for bacteria have been developed that use, as the variable for genetic
correlation, the gene presence/absence characteristics typical of the bacterial accessory
genome (pan-GWAS) (5). The size of the accessory genome of dierent species of
bacteria varies a great deal, with E. coli having one of the largest (6, 7). Pan-GWAS of E. coli
has been used to identify genetic associations for antibiotic resistance and bloodstream
infections (8, 9) and compare E. coli gene content between dog and human isolates
(10). Although AMR has been assessed and analyzed in dogs for a range of bacterial
species and in a variety of settings and geographic regions [e.g., (11–14)], we are not
aware of a publication that employs pan-GWAS to assess the association of genes with
resistance phenotypes in E. coli from dogs. The collection analyzed herein involves the
periods 2017–2020 and includes in vitro susceptibility data for 34 dierent antibiotics
and over 1,000 isolates from clinical cases or necropsies. Here, we employ a pan-GWAS
perspective on E. coli isolates from sick dogs collected as part of the US National Action
Plan and its surveillance for resistant bacteria in animals (2, 15). We then go on to explore
with evolutionary genetic and protein structural modeling approaches the molecular
character of one of the top ranked candidate AMR loci arising from the pan-GWAS
analysis that result in working hypotheses regarding how it might function as an AMR
mechanism in the cell.
MATERIALS AND METHODS
Samples and genome sequencing
No research animals were used for this study. Pathogenic E. coli isolates from dogs were
collected as part of a previously described surveillance program (2). Briey, 30 animal
diagnostics laboratories participating in the Vet-LIRN AMR monitoring program each
saved the rst four clinical (from a sick or dead animal) E. coli isolates from unique
canine cases submitted for diagnostic testing by veterinarians each month. Deidentied
isolates were cryopreserved in glycerol and shipped to one of the six assigned sequenc
ing reference laboratories. Whole genome shotgun sequencing of pure cultures was
performed using Illumina chemistry (DNA Prep or Nextera XT library preparation), and
sequencing was performed on the MiSeq platform (v2 or v3 chemistry with 2 × 250 bp
reads). Genomes meeting standardized benchmarks (16) were then uploaded to NCBI,
from which they were retrieved for this study. Sample sources included urine, wound,
tracheal wash, ear swab, and a variety of tissues. Isolates included in this study were
collected from 2017 to 2020 and covered 11 dierent geographical regions of Canada
and USA. All genomes are available in the following NCBI BioProjects: PRJNA318591,
PRJNA324565, PRJNA324573, PRJNA481346, PRJNA503851, and PRJNA318589, all under
the umbrella of BioProject PRJNA314609.
Full-Length Text Applied and Environmental Microbiology
August 2024 Volume 90 Issue 8 10.1128/aem.00354-24 2
Genome assembly, capsular polysaccharide gene cluster diagrams, phyloge
netic reconstruction, population genetics, and serotyping
All BioProjects were registered with the NCBI Pathogen Detection system (https://
www.ncbi.nlm.nih.gov/pathogens/), which performs assemblies from raw Illumina data
uploaded to Sequence Read Archive (SRA) and deposits those into GenBank. Any
missing assembly was assembled using SKESA (17). Any assembly with >200 contiguous
sequences (contigs) was excluded from the study. All genomes were veried to have a
canine host in the metadata and identied as E. coli. The nal compiled and curated set
of dog E. coli genome sequences was 1,778. All genome assemblies were annotated with
Prokka v1.14.5 (18). Core genome alignments were generated within Panaroo v. 1.2.9 (19)
using the MAFFT algorithm (20). Gaps were then eliminated using Gblocks v. 0.91b (21,
22), and Panaroo was used to generate gene presence/absence matrices for downstream
analyses. Trees were built using IQ-TREE (23) v. 2.0.3 and viewed in iTOL (24). Multilocus
sequence typing (MLST) was performed according to the Achtman scheme using the
MLST tool (25), which makes use of the PubMLST database (26). Serotype predictions
were performed with ECTyper (27) for O and H antigens and Kleborate (28) for K
antigens. Python package pyGenomeViz (https://github.com/moshi4/pyGenomeViz) was
used to plot the aligned capsular polysaccharide (CPS) clusters. The R package RhierBAPS
v1.0.1 (29) was used for hierBAPS clustering of the genomes, performed on the same
multiple sequence alignment used for the construction of the phylogenetic tree.
Pangenomics, plasmid identication, and Scoary
In vitro antimicrobial susceptibility data (either by a minimum inhibitory concentra
tion or disk diusion) and interpretations were collected from the veterinary diagnos
tic laboratories participating in the study, compiled and coded into binary data for
each antimicrobial reported (0 = susceptible; 1 = resistant). The laboratories provi
ded susceptibility interpretations according to their standard procedures for clinical
testing, and this interpretation was used for analysis. If no interpretation or an intermedi
ate susceptibility was provided, those results were omitted. Correlation of resistance
phenotypes with accessory gene presence/absence was performed with Scoary v. 1.6.16
(5). Short lists of candidate loci correlated with resistance were compiled from genes
passing both empirical P and Bonferroni correction, at P < 0.01. We were interested in
determining putative novel antimicrobial resistance genes, as well as evaluating which of
those genes were chromosomal; hence, we rst identied those that were known AMR
genes, as well as genes carried on plasmids. Signicant Scoary genes were compared
with known AMR determinants using the AMRFinderPlus (30) database, and genes likely
carried on plasmids were identied using Deeplasmid (31). Deeplasmid applies a deep
learning model to classify and separate plasmids from bacterial chromosomes and was
used to predict plasmids from all assembled genome contigs. Panaroo gene clusters
were designated as plasmid-borne only if >10% of the cluster members were located on
the predicted plasmids. This resulted in our grouping of the Scoary signicant genes into
the following four categories: (i) known AMR locus carried on a plasmid, (ii) on a plasmid
but not a known AMR gene, (iii) known AMR gene not on a plasmid, and (iv) not a known
AMR gene and not on a plasmid. Our eorts to identify putative novel antimicrobial
resistance genes focused on genes estimated to be chromosomally encoded because
signicant Scoary genes judged to be on plasmids may simply be correlated with a
resistance phenotype due to linkage on a plasmid carrying a known AMR locus. Plasmids
were identied from assemblies using ABRicate (Galaxy Version 1.0.1) with the plasmid
nder database version 2020-Apr-19.
Natural selection
We tested for gene-wide signatures of selection intensity dierences across two genes
arising from the pan-GWAS Scoary analysis that were correlated with resistance and
were components of a group 1 operon for CPS that lacked several critical genes. Our
Full-Length Text Applied and Environmental Microbiology
August 2024 Volume 90 Issue 8 10.1128/aem.00354-24 3
hypothesis was that the absence of critical genes could render this CPS locus non-func
tional, resulting in a signal of relaxed selection in these two individual CPS genes, and
that this in turn could be related to their role in AMR. The two genes in question were
allelic variants of wza (polysaccharide export protein) and wzi (capsule assembly protein).
The relaxation of selective strength can drive evolutionary innovation and/or foreshadow
loss of function; newly developed evolutionary genetic statistical approaches provide the
means to identify genes evolving under such relaxed selection pressure (32).
We generated codon-aware alignments following the procedure available at
the Github repository (github.com/veg/hyphy-analyses/tree/master/codon-msa). Briey,
in-frame nucleotide sequences were translated and then aligned using MAFFT v7.471
(20). Aligned protein sequences were then mapped back to the nucleotide sequence,
and a single copy of each unique sequence was retained. Alignments were screened for
the presence of genetic recombination using the genetic algorithm for recombination
detection (GARD) method (33). If GARD identied supported recombinant breakpoints,
then alignments were partitioned accordingly and a maximum likelihood phylogeny was
inferred for each partition using RaxML-NG v0.9.0git (34) under the GTR+Γ nucleotide
substitution model. Where necessary, we used phylotree.js (35) to label phylogenetic
branches (based on the gene they coded for) for downstream selection analyses. The
Hypothesis testing using Phylogenies (HyPhy) v.2.5.43 software package (36) was used
to infer signals of selection and gene-wide signals of selection intensity were compared
using RELAX (32).
Structural comparison
To generate predicted structural models of Wza2 proteins (ElI3491725.1, ElI6415193.1,
EFO5660681.1, and QDM04109.1), GfcE (P0A932), and Wza1 from Klebsiella michiga
nensis (WP_224378673.1), we used AlphaFold 2.0 (37) accessed with ColabFold (38).
The models were generated as monomers, without structural templates and with the
mmseqs2_uniref_env multiple sequence alignment (MSA) mode. Predicted structures
were then visualized using UCSF ChimeraX (39). AlphaFold 2 predicted structures for
Wza2, gfcE, and Wza1, which presented overall similar domain architecture and folding
compared with the X-ray crystal structure of Wza1 (PDB 2j58), and an overall high
predicted local distance dierence test (pLDDT) condence score (>84) was chosen
to map conservative and non-conservative amino acid substitutions. This mapping
was performed by identifying substituted positions from protein alignments (20) and
comparing the sequence derived from Wza1 reference structure (PDB 2j58) with the
corresponding sequences of the AlphaFold 2 predicted structures. Visualizations and
highlighting of substituted residues were performed using UCSF ChimeraX. The rst 20
residues of Wza1 (PDB 2j58) are absent in the mature protein structure as they constitute
the protein’s signal peptide (40). As such, the corresponding residues in the AlphaFold
2 predicted structures were not displayed. The MatchMaker structural alignment tool
in UCSF ChimeraX was used to generate octameric representations of the AlphaFold
2 predicted protein monomers, with the Wza1 reference structure (PDB 2j58) used as
an octameric template. The octameric representation allowed us to identify residues in
domain 2 of each monomer facing the inner pore cavity.
We computed a pairwise TM-score with the tmscoring method (41, 42) between
5 Wza1, 38 Wza2, and four gfcE predicted monomeric structures. All structures were
predicted using AlphaFold 2 as described above, and a heatmap for the matrix was
constructed to identify structural relationships between pairs.
Phenotypic characterization of strains harboring wza2
In vitro susceptibility of 36 strains identied as harboring wza2 was repeated by antibiotic
susceptibility testing (AST) on the automated Sensititre platform (Thermo Fisher,
COMPAN1F panel) and interpreted using the CLSI VET01S standard (43). Susceptibility for
cefovecin was also measured by disc diusion on Mueller Hinton agar using commercial
30 µg discs (Thermo Fisher).
Full-Length Text Applied and Environmental Microbiology
August 2024 Volume 90 Issue 8 10.1128/aem.00354-24 4
Transcriptomics on three of the strains was performed in LB medium with 0.5 µg/mL
reconstituted cefovecin (Zoetis) or an equivalent amount of sterile water (same diluent
as the antibiotic). Three fresh colonies from each strain were separately inoculated into
1 mL of the broth and incubated at 37°C with shaking for 5 h. Cultures were then
immediately pelleted, the supernatant was removed, and the pellet was resuspended
in an RNA stabilizer (DNA/RNA shield, Zymo Research). These suspensions were stored
at room temperature until processed by the Cornell Transcriptional Regulation and
Expression Facility, where RNA extraction and library preparation were performed using
Trizol and the NEB directional library preparation kit. Sequencing was performed with PE
2 × 150 bp read length on the NovaSeq 6000 instrument with 20M reads per sample.
Reads were trimmed for low quality and adaptor sequences with TrimGalore v0.6.0 (44), a
wrapper for cutadapt (45) and fastQC (46). Reads were mapped to the specic reference
genome for that strain using STAR v2.7.0e (47). SARTools and DESeq2 v1.26.0 were used
to generate normalized counts and statistical analysis of dierential gene expression (48,
49). All RNA-seq data have been deposited to NCBI SRA under the BioProject accession
PRJNA1053349.
RESULTS
Pan-genome, core-genome phylogeny, and population genetics
The pangenome content for 1,778 canine E. coli isolate genomes meeting all quality
thresholds was open, with a total of 23,956 genes (Fig. 1A) and a core genome size
nearing plateau at 3,179 genes (Fig. 1B). Additional break-down of components into core
(3,191 clusters), shell (16502 clusters), soft-core (235 clusters), and cloud (4028 clusters)
is represented in Fig. 1C. Core and soft-core are genes present in 99% and 95%–99%
of genomes. Cloud genes are dened as those clusters which contain a single gene
(singletons), plus those which have more than one gene, but its organisms are proba
bly clonal due to identical general gene content (colloquially dened as strain-specic
genes). Shell genes are dened as those clusters that neither belong to the core genome
nor to the cloud genome.
A core genome phylogeny with AMR data, phylogroup, and geographic region
mapped to it appears in Fig. 2. Eight dierent phylogroups were represented in that
phylogeny, with the majority of isolates falling within B1 and B2, including most of the
multidrug-resistant (MDR) clades, with the exception of one major MDR group represen
ted in each of phylogroups A, C, and F. Based on this core-genome phylogeny, there
was no evidence for clustering of separate E. coli genotypes in dierent geographic
regions, and there was also no clustering by sample type (Fig. S1) or year (not shown).
An examination of accessory gene content did support distinct clusters, not correlated
with geographic variation or sample type (urine versus non-urine) but was correlated
with phylogroup. HierBAPS analysis identied 20 distinct genetic clusters at hierarchical
FIG 1 Canine E. coli pangenome (N = 1,778). (A) Roary (50) matrix containing identied gene clusters. (B) Pangenome curve consisting of all genes in purple and
core genes in green. (C) Distribution of the pangenome into cloud, shell, and core components.
Full-Length Text Applied and Environmental Microbiology
August 2024 Volume 90 Issue 8 10.1128/aem.00354-24 5
level two clustering, with a minimum cluster size of N > 16 (Fig. S2). The genetic clusters
showed no correlation with geographic regions or year and were highly mixed with
isolates from dierent regions of USA and Canada.
The MLST data depicted a similar mixed picture (Fig. S3) and indicated that ST372
was the predominant ST in all regions of the USA and Canada studied. Sequence type
singletons (N = 152) appeared in each region as well. The predominant predicted serovar
was O83:H31, followed by O4:H5 and O6:H31 (Fig. S4). No O antigen type 157 was
predicted, but several H7 combinations with dierent O types were found. Sixteen of
the canine ST372 strains carried a pUTI89 plasmid (IncFII (29)_1_pUTI89) that has been
described as a marker of human infection (51). These were distributed across dierent
serotypes including O83:H31, O4:H31, and O15:H31, which correspond to the ST372
M, G, and L clusters identied by the pan-GWAS analysis performed Elankumuran et
al. comparing globally distributed human and dog isolates (10). One strain from our
study (ECOL-18-VL-OH-ON-0048) was closely related (SNP distance of 36) to human strain
MVAST4963 characterized by Elankumuran et al. as belonging to cluster K. A majority of
the dog-associated propanediol operon components highlighted by Elankumuran et al.
in clusters G and M1/2 were annotated in the accessory content of 331 of our dog strains,
which harbored pduA, pduB, pduC, pduE, pduF,pduL,pduN pduO, pduP, and pduU. The
complete gene presence/absence matrix from our set is available in the data repository
(https://doi.org/10.7298/pdzk-rq02.2).
Pandemic ExPEC lineages (as reviewed by LW Riley) (52) ST131 (53), ST127 (54),
ST73, ST69, and ST95 were also prominent across multiple regions. Canine-derived
ST131 strains identied included both O25:H4 and O16:H5 predicted serotypes. Thirty
of 55 had two parC point mutations in common with Clade C reference strain EC958
(GCA_000285655.3), and 7 of those also harbored a blaCTX-M beta-lactamase gene. All
canine ST131 strains had a pmrB E123D point mutation predicting Colistin resistance,
although this phenotype was not assessed. The NCBI Pathogen detection browser
identied 11 matched clusters from this set, and those were primarily comprised of
FIG 2 Core genome phylogeny of E. coli sampled from dog hosts across US and Canadian regions participating in the surveillance program, with geographic
regions, phylogroups, and antibiotic resistance phenotypes annotated on that phylogeny. The map was created using the ne_states function within the
rnaturalearth R package, the geom_sf function within the sf R package, and ggplot2.
Full-Length Text Applied and Environmental Microbiology
August 2024 Volume 90 Issue 8 10.1128/aem.00354-24 6
clinical strains presumed to be from humans. The canine strains were from urine,
wounds, and one lung. The K1 capsule gene neuC, a hallmark of neonatal meningitis
causing E. coli (NMEC) (55), was detected in 70 isolates, and kpsF was detected in 598
isolates.
In vitro antimicrobial resistance characterization
In vitro AST data for 50 antimicrobials (or combinations) were obtained from the
participating veterinary diagnostic laboratories. Of these, 12 antimicrobials (Bacitra
cin, Danooxacin, Florfenicol, Moxiacin, Oxytetracycline, Rifampin, Spectinomycin,
Sulphamethoxine, Tiamulin, Tilmicosin, Tulathromycin, and Tylosin) did not include
interpretations and were excluded from further downstream analysis. Four additional
antimicrobials (Ciprooxacin, Levooxacin, Ooxacin, and Oxacillin) had only one value
and were also excluded. Thirty-four antimicrobials with sucient interpretation data
were used for pan-GWAS analysis on the corresponding 1,133 genomes.
In vitro MDR was relatively common and varied depending on region (Fig. 2; Fig. S5).
Resistance to >5 classes of antimicrobials was seen at low levels (<5%) in 3/4 Canadian
regions and 4/7 US regions. Resistance to 3–5 antimicrobial classes occurred in every
region and was highest in central Canada (24%), followed by west south-central US
(17%). In contrast, the proportion of isolates harboring known antimicrobial resistance
genes for multiple classes (based on AMR Finder annotation) was similar for either 3–5
or >5 classes, with the exception of two Canadian regions, which had a higher proportion
of strains with >5 AMR gene classes.
AMR gene correlation with in vitro susceptibility
Accessory gene content was correlated with resistant/susceptible coding of the AMR
data using Scoary, and the genes judged to be signicantly correlated (P < 0.01 for both
empirical P and Bonferroni) to resistance were divided into our four genomic categories
for each antibiotic (Fig. 3).
The total number of signicantly correlated genes varied widely depending on the
antibiotic, with as few as two (ticarcillin) and as many as 288 (tetracycline). The majority
of Scoary-signicant genes fell within the category of plasmid-borne but not known AMR
—in other words, plasmid cargo genes—and this ranged from 44%–91%. The smallest
category was “known AMR not on a plasmid,” so likely chromosomal. The category we
focused particular attention on, with regard to identifying putatively novel AMR loci, was
the “putative chromosomal AMR gene.” Due to the large number of genes identied, we
focused on those associated with MDR. Protein annotations for the top six genes within
this category that were Scoary-signicant for resistance to multiple antibiotics were
(number of antibiotics resistant in brackets) the following: Wzi, capsule assembly protein
(10); transcription regulator (9); hypothetical relaxase (8); Wza, polysaccharide export
protein (8); transcription regulator (8); and IucC, aerobactin synthase (8). (Table S1). Due
to two of these being adjacent to each other on the chromosome, and the strong
association with susceptibility of the colanic acid wza, we chose to focus on those for this
study and not discuss the other ndings.
Wza and Wzi and their associated group 1 CPS
Two group 1 CPS genes—wza and wzi—were correlated with resistance to cefovecin,
other cephalosporins, and several other antibiotics (Table S1). Wza is a polysaccharide
export protein, and Wzi is a capsule assembly protein involved in group 1 capsules in
both E. coli and Klebsiella (56). Wzi is only present in group 1 CPS (57). The wza sequences
in our analysis (Excel Table S2; Fig. 4) that were correlated with MDR are wza group 1
allelic variants with protein sequence divergence ranging from 0%–10%. A similar level of
divergence of 0%–13% was apparent when our set of 38 Wza was compared with other
Wza group 1 sequences randomly chosen from NCBI to reect a range of sequence
divergence. For the purposes of discussion in this manuscript, hereinafter, all Wza
proteins from the group 1 CPS arising from these canine E. coli genomes are referred to
Full-Length Text Applied and Environmental Microbiology
August 2024 Volume 90 Issue 8 10.1128/aem.00354-24 7
as Wza2 and other Wza group 1 chosen from GenBank for comparison are referred to as
Wza1. Three distinct types of wza genes are present in E. coli: those within group 1
capsules, colanic acid operon, and the group four capsule operon (referred to as gfcE)
(57). All three form separate monophyletic clusters (Fig. S6). The most abundant form in
our E. coli isolates was wza-colanic acid (hereinafter referred to as wzaCA) (n = 1089),
followed by gfcE (group four wza) (n = 289) and then wza2 (n = 38). wzaCA is a soft-core
gene in our analysis, which is to be expected, because colanic acid production in E. coli is
common (56). No E. coli genomes with wza2 had wzaCA. Those genomes with wzi always
also had wza2 and gfcE (38/38); six genomes had gfcE only; 245 genomes had wzaCA and
gfcE and 884 genomes had wzaCA only. wzaCA was associated with susceptibility to a
variety of antibiotics, including cefovecin and other cephalosporins (Table S1). Phyloge
nies of solely wza2 and wzi from canine strains had a similar pattern consisting of two
distinct clades, one of high divergence and the other a group of 18 identical, plus three
nearly identical, sequences. There were 11 dierent MLST sequence types represented in
these wza2/wzi carrying isolates, with 18 isolates belonging to ST162 (Fig. 4). Phylogenies
of our group 1 Wza2 with the inclusion of homologous sequences of Wza1 from GenBank
did not support the monophyly of the dog derived sequences (Fig. S7). The closest
relatives of our Wza2 and Wzi arose from a mixture of E. coli and Klebsiella pneumoniae
(Fig. S8). Resistant and susceptible isolates to cefovecin and the other seven (in the case
of Wza2) or nine (Wzi) antibiotics came from both the divergent and non-divergent
clades of these gene-specic trees. Upon searching the NCBI database for the wza2
FIG 3 Number of candidate genes from Scoary analysis in four genomic categories with signicant correlation to resistance (P
< 0.01 with both empirical P and Bonferroni correction) by the antibiotic tested.
Full-Length Text Applied and Environmental Microbiology
August 2024 Volume 90 Issue 8 10.1128/aem.00354-24 8
operon, strains from food production animals, as well as humans, were close matches
(Fig. S8). Active transcription of the wza2 operon was veried for 3 of the dog strains (Fig.
S9), with and without cefovecin added at the MIC50 level established in the original
publication (58). The levels of normalized transcripts with and without the treatment
were nearly identical.
A typical E. coli group 1 CPS operon consists of about 12–14 genes between galF
and gnd on the E. coli chromosome (Fig. 4), comprising two separate groups. The 5’
part of the locus contains four genes (wzi, wza, wzb, and wzc), always present, and
the 3’ region is serotype-specic, variable in gene content and includes enzymes for
producing sugar nucleotide precursors, glycosyltransferases (GTs), and two integral inner
membrane proteins (Wzy and Wzx) (57). All but two of the CPS depicted in Fig. 4 do not
possess the essential gene content for group 1 capsule production. Two essential loci
are wzx and wzy, which typically exist together in the same operon. The two sequences
that do appear to have intact group 1 capsule gene content are GCA_007012305.1
and GCA_013094695.1. Much of the evolutionary history behind wza2 and its group
1 accompanying wzi is likely linked to lateral gene transfer (LGT) events mediated
by the transposons that are prevalent in the CPS of these particular genomes (Fig.
4). One prominent transposon evident in our group 1 CPS sequences is “DDE-type
integrase/transposase/recombinase” and is present in the majority of these 38 genomes,
residing one or two genes upstream of galF. This same transposon is found in the
majority of wzaCA containing soft-core genomes in our data set, immediately upstream
of wzaCA, suggesting a common LGT history at some point between these dierent
wza types. Other transposons are also present, including in some instances, as many
as four immediately adjacent to one another, residing between galF and wzi (Fig. 4).
Two common transposons were Insertion Sequences (IS) IS3 and IS30 elements; IS30
occurring in 8 of 38 genomes and in all but one occasion anked by IS3, suggesting
FIG 4 Group 1 CPS gene content from the 38 isolates carrying wza2 and wzi genes that were correlated with resistance to
multiple antibiotics. The phylogeny is based on wza2 with the respective genome IDs as labels. ST of each genome is listed on
the right. Color-coding of the isolate genome accessions denotes resistant (red) and susceptible (blue). Group 1 capsules that
do not extend to gnd are because of truncated contigs.
Full-Length Text Applied and Environmental Microbiology
August 2024 Volume 90 Issue 8 10.1128/aem.00354-24 9
these paired IS elements represent a possible composite transposon. Both these IS
elements share a high sequence identity with other E. coli transposons of the same type.
Relaxed selection
Discerning a signal of selection intensity in targeted sequences (test) requires a
comparative group (reference)—in our case, the wzi reference group was comprised of
other homologous wzi sequences obtained from NCBI, from outside our study collection.
Because E. coli carries dierent homologous versions of wza, our reference groups for
wza2 included E. coli wza from wild-type group 1 capsules (wza1), the colanic acid
operon (wzaCA), and the group 4 capsule operon (gfcE). These dierent forms of wza are,
respectively, monophyletic (Fig. S7). In analyses of selection pressure intensity (Table 1),
both wza2 and wzi were judged to be evolving under relaxed selection—the entire wza2
gene relative to wzaCA and in the case of wzi, GARD fragment 6 relative to reference
wzi sequences. wza2 comparisons to wza1 were signicant for relaxed selection for a 5’
region that included primarily the sequence encoding the signal peptide (nucleotides
1–135; GARD fragment 1: k = 0.24, P-value = 0.006). To get an additional comparative
perspective on how wza2 might dier from wza1, we compared wza1 with the same
outgroup set of wzaCA, and in this case, wza1 was signicant for relaxed selection for a
3’ ~57% of the gene (contained within GARD fragment 2: k = 0.67, P-value = 0.002), but
not for the 5’ ~43%. Comparisons of both wza2 and wza1 to gfcE were not signicant for
relaxed selection. gfcE compared with wzaCA was signicant for relaxed selection in the
3’ ~55% of the gene. Taken collectively, this suggests that there is a history of relaxed
selection in wza2, and what distinguishes it from wza1 in this regard is the 5’ ~12% of the
gene, which is only under relaxed selection in wza2. Test results for each of the relevant
comparisons can be found in Table 1.
Wza protein structure comparisons
Previous studies have suggested a possible role of Wza group 1 from E. coli K30 isolates
in translocating antibiotic into the bacterial cell, via the central pore that forms in the
octameric structure of the protein [summarized by Sun et al. (59)]. This prompted us to
TABLE 1 RELAX results of wzi, wzaCA, wza1, wza2, and gfcEa
Gene(s) analyzed GARD fragment RELAX
kP value
wzi vs. NCBI wzi sequences (test: wzi,
reference: NCBI wzi sequences)
1 xx xx
2 xx xx
3 xx xx
4 xx xx
5 xx xx
6 0.24 0.006
7 xx xx
wza2 vs. wzaCA
(test: wza2, reference: wzaCA)
1 0.8 0.048
2 0.5 0
wza1 vs. wzaCA
(test: wza1, reference: wzaCA)
1 xx xx
2 0.67 0.002
gfcE vs. wzaCA
(test: gfcE,reference: wzaCA)
1 xx xx
2 0.2 0.014
wza2 vs. wza1
(test: wza2, reference: wza1)
1 0.54 0.011
2 xx xx
3 xx xx
4 xx xx
5 xx xx
aThe k parameter measures the selection intensity. Intensied selection is inferred when k > 1 and relaxed when k
< 1 along test branches.
Full-Length Text Applied and Environmental Microbiology
August 2024 Volume 90 Issue 8 10.1128/aem.00354-24 10
investigate structural comparisons between resistant and susceptible isolates carrying
Wza2 as well as between Wza2 and Wza1 and WzaCA and GfcE. Structural models of
octamers constructed using AlphaFold 2 and compared using Tm-scores (Fig. 5 and 6)
suggest that Wza2 has a similar overall structure to Wza1. Comparisons to WzaCA and
group 4 GfcE suggest that group 1 Wza (Wza2 and Wza1) was more similar to GfcE than
WzaCA. Structural dierences between Wza2 from cefovecin-resistant and susceptible
isolates were overall not apparent.
We experimented with the computational docking of cefovecin with Wza2, Wza, and
GfcE using SeamDock (60). These data are located in the supplemental data repository.
If one considers the entire search space of the protein, cefovecin docked with all three
of these proteins but tended to be in dierent places and could be inside or outside the
inner cavity. For Wza2, cefovecin docking was outside either ring 2 or 3. However, if one
considers just the internal cavity as a search space (a logical premise for the transmission
of an extracellular antibiotic), for both Wza1 and Wza2, cefovecin docks in the D2 domain
of the protein in the region where there were a number of non-conservative substitu
tions. Wza2 was predicted to bind with a slightly better anity to the antimicrobial
(−6.5 kcal/mol vs −5.7 kcal/mol) and included ionic interaction, hydrophobic contact,
hydrogen bonds, and weak hydrogen bonds; Wza1 (2j58) did not include any ionic
interaction or weak hydrogen bonds. Overall, we consider that these docking results
FIG 5 Heatmap generated from computed pairwise TM scores between WzaCA, Wza1, Wza2, and GfcE protein structures. TM
scores are between 0 and 1, the closer to 1 (red), the more similar the structures being compared; the further from 1 (blue), the
less similar. All sequences used in this study (WzaCA, Wza2, and GfcE) are also annotated based on either their resistance (RES)
or susceptibility (SUS) status for cefovecin.
Full-Length Text Applied and Environmental Microbiology
August 2024 Volume 90 Issue 8 10.1128/aem.00354-24 11
should be regarded with a degree of circumspection since the results are quite sensitive
to the initial parameters. Therefore, we consider that although it is likely that cefovecin
could indeed dock within the internal pore of Wza1 and Wza2, we cannot condently
conclude that a greater binding anity of the antibiotic within the internal pore would
be expected with Wza2.
An ESPript (61) alignment comparison of a resistant representative of Wza2 and the
original 2j58 crystal structure (40) protein sequence identied several non-conservative
amino acid changes in Wza2 that mapped to domain D2 in the monomer and the R2
ring in the octamer and tended to cluster along the inner pore of the protein (Fig. 7A).
Mapping of non-conservative changes involving a susceptible Wza2 with a seemingly
intact group 1 capsule (GCA_013094695.1) was overall similar, but with two notable
dierences, and these were N181 an K184 compared with K181 and Q184 in the resistant
representative. These changes at 181 and 184 were, however, not consistent across other
comparisons of resistant and susceptible isolates.
FIG 6 Inferred dendrogram of WzaCA, Wza1, Wza2, and GfcE protein pairwise TM scores. The scores were hierarchically
clustered with the unweighted pair group method with arithmetic mean (UPGMA). All sequences used in this study (WzaCA,
Wza2, and GfcE) are also annotated based on either their resistance (RES) (highlighted in orange) or susceptibility (SUS) status.
Full-Length Text Applied and Environmental Microbiology
August 2024 Volume 90 Issue 8 10.1128/aem.00354-24 12
DISCUSSION
We found wide variation for both in vitro and predicted resistance proles in this set
of >1,000 E. coli isolates from sick dogs, supporting that dogs may be a reservoir of
pathogenic and MDR E. coli with unexplored mechanisms of importance to humans. This
could be due to shared exposures to strains from food (62), humans (63), or hospital
environments (64, 65). Our population genetics ndings indicate recurring sequence
types, and predicted serotypes, across every region studied. The most predominant
serotype predicted (n = 110) was O83:H31, which has been highlighted (10, 66) as a
subset of ST372 shared between human and canine ExPEC strains worldwide but is also
highly similar to a strain considered to be a probiotic (67). The next most common
serotype, O4:H5 (n = 93) is considered to be highly uropathogenic in humans and several
animal hosts (68). The high number of strains harboring ExPEC-associated factors may
originate from enteric strains of animal owners that could easily be passed back and
forth in households. Food is also an important consideration, as Liu et al. (69) attributed
8.4% of extraintestinal E. coli infections in humans to strains isolated from commercial
meat, based on accessory gene content. Our ndings are consistent with Elankumuran et
al. in the prevalence of pdu propanediol metabolism genes in canine strains, which they
hypothesize could be a host adaptation due to diet (10). This is a promising direction for
future analyses.
The E. coli capsule is a cell surface structure composed of long-chain polysaccharides
involved in protecting the cell, as well as acting as a major virulence component, playing
an important role in the bacteria evading or counteracting the host immune system. It
has also been suggested that E. coli capsules may act as a physical barrier to impede
FIG 7 Changes between Wza1 (crystal structure 2j58) and Wza2 mapped on a model of the monomer and the octamer of
a representative resistant (panel A) and susceptible (panel B) Wza2, distinguished by conservative (gray) vs non-conservative
(red) mutations. A conservative mutation is considered one that does not dramatically change the biochemical properties of
the residue (e.g., L to I or R to K) and the opposite for non-conservative (e.g., D to R). Within the R2 ring (D2 domain of the
monomer), there are six non-conservative mutations at positions facing the inner pore cavity, and on the outer “lip” within the
R4 ring, there are two mutations in the D4 domain from basic residues to asparagines (N).
Full-Length Text Applied and Environmental Microbiology
August 2024 Volume 90 Issue 8 10.1128/aem.00354-24 13
antibiotics entering the cell (70), resulting in antibiotic resistance. Our goal was to try
and identify novel chromosomal loci correlated with AMR. Our Scoary analysis supported
two chromosomal genes, wza and wzi, from group 1 cluster as strongly correlated with
resistance to cefovecin, other cephalosporins, and a variety of other antibiotics. Wza
is an integral outer membrane protein, essential for capsule export and exists as an
octamer, with a large central cavity, or pore, 100 Å long and 30 Å wide (40), through
which polysaccharides are translocated across the periplasm and outer membrane (71).
Earlier studies of E. coli wza1 deletion mutants indicated that wza1 deletion strains
were resistant to a number of macrolides (72) and that complementation of wza1 in
a strain bearing a deletion of this gene could restore the wild-type susceptibility to
erythromycin (73). Su et al. measured the diameter of the Wza1 octamer protein pore
as approximately 17 Å and estimated the size of erythromycin as 12 Å, suggesting that
the pore may be large enough to allow transmission of the antibiotic into the periplasm
[summarized by (59)]. The strong correlation of the wza2/wzi from the group 1 cluster
with an array of antibiotics suggests that an opposite eect of inhibition of antibiotic
transmission occurs in these widely dispersed strains. This in turn led us to hypothesize
that structural dierences might be apparent between Wza2-resistant and susceptible
strains and possibly this could be related to an inhibition of antibiotic transmission by
Wza2.
Based on AlphaFold 2 models compared using Template Modeling score (Tm-score),
there were no discernable dierences between Wza2 from cefovecin-resistant and
susceptible isolates, and there were no dierences in comparisons of the protein
structure of Wza2 compared with Wza1. Dierences were apparent between the two
group 1 Wza (Wza2 and Wza1) and WzaCA and GfcE, although the Tm scores were
still relatively high, indicating very similar folds (42). The colanic acid operon and the
associated wzaCA gene from that operon are absent from the wza2-carrying isolates,
but gfcE was present. In isolates where WzaCA is present (which is the vast majority
of isolates), it is signicantly associated with susceptibility to cephalosporins, as well as
a number of other antibiotics, and the other genes in that colanic acid operon share
the same pattern. It is possible that WzaCA is a portal for antibiotic transmission, but
to our knowledge, this is an untested idea. It is the case that wza1, wzb, and wzc from
colanic acid have been reciprocally interchanged with those same three homologous
genes from group 1 and function normally (74) in both operons. In comparisons between
Wza1 (2j58 PDB crystal structure sequence) and Wza2 sequences, there were a number
of non-conservative amino acid substitutions (changing either the polarity or the charge
and two instances of A to the cyclic amino acid P) lining the inner pore of Wza2 (Fig.
7). Water interaction with the polarity of the inner pore of Wza1 has been suggested
as important in facilitating the export of polysaccharides (40). It is possible that these
Wza2 substitutions could involve electrostatic interactions with antibiotic molecules
that might aect antibiotic transmission; however, there was no consistent correlation
between resistant and susceptible isolates carrying Wza2 regarding the specic character
of such substitutions. In our relaxed selection analysis, wza2 was under relaxed selection
relative to wzaCA across the entire gene. wza1 was also under relaxed selection relative
to wzaCA, but only in the 3’ ~57% of the gene (GARD fragment 2); it was not under
relaxed selection in the 5’ ~43% of the gene (GARD fragment 1). The two GARD
fragments are judged to have distinct evolutionary histories, with a putative recombina
tion break-point separating the two halves. GARD fragment 1 in wza2 is the region where
these non-conservative changes reside. This suggests that relaxed selection pressure
on this ~12% of the gene resulted in random nonsynonymous mutations yielding the
non-conservative amino acid changes that happen to line the inner pore of the octamer.
We suggest that this signal of relaxed selection is a consequence of missing essential
gene content in these group 1 capsule sequences, rendering the group 1 CPS non-func
tional.
It is important to consider that Wza is only one protein in the group 1 CPS assembly
pathway. Of particular importance is the interaction that occurs between Wza and Wzc.
Full-Length Text Applied and Environmental Microbiology
August 2024 Volume 90 Issue 8 10.1128/aem.00354-24 14
This second protein is thought to regulate the opening of the PES (polysaccharide export
sequence) domain of Wza within the periplasm, which is in a closed conformation in the
absence of Wzc interaction. Collins et al. proposed that the polysaccharide enters the
central pore of Wza through the opened PES domain and, then moving along the central
cavity, exits to the extracellular space through the alpha-helical domain of Wza (75). The
interaction with Wzc is thought to trigger the active or open conformation of Wza. Thus,
if this association of Wza and Wzc is disrupted because of sequence changes arising from
a relaxation of selective constraints on Wza2, then an antibiotic entering the pore at the
alpha-helical domain of Wza may not be able to exit into the periplasm or cytoplasm
because of a closed conformation of Wza, essentially functioning as an antibiotic trap.
Additional genes within the Wza2 CPS cluster sequence were also correlated with
AMR (Table S1). The clearest example of this was wzi; however, other genes do show
a bias toward resistance but do not pass our strict statistical criteria because of the
reduced sample number for those particular genes, due to their partitioning into
dierent variants. As an example, our data set has two variants of wzc; hence, the Scoary
analysis treats them as separate loci, and one of them is signicant for empirical P (P <
0.01) but does not pass the Bonferroni correction. Similarly, wbaP (UDP-Gal:phosphoryl-
polyprenol Gal-1-phosphate transferase), involved in the rst step in biosynthesis of
group 1 CPS, is present in several of our CPS, is signicant for empirical P (P < 0.01), but
does not pass Bonferroni correction. The interactive necessities of the Wzx/Wzy pathway
in capsule production suggest that focusing on a single locus as the potential AMR
mechanism may be misguided. Failure of the pathway could aect the transmission of
antibiotics through the Wza2 pore, ultimately into the periplasm and the cytoplasm, thus
rendering the isolates resistant.
These Wza2 capsule clusters carry many of the genes necessary and typical of a
functional capsule one assembly locus; however, the gene content is variable between
isolates, except in the members of the low divergence Wza2 clade (ST162), and in the
majority of instances, they are missing some key loci. Although dierent serotypes carry
gene-specic loci, genes encoding Wzx and Wzy, dening the Wzx/Wzy pathway, are
always present in a functional group 1 CPS locus (57). Wzx translocates the O-antigens
across the inner membrane, where the O units are polymerized by Wzy. This latter
protein is known to exhibit low sequence conservation both between and within
bacterial species (76). Our Wza2 genomes carry examples of ve dierent putative
Wzy proteins, with low sequence homology. Some form of Wzy, and its representation
within this CPS cluster sequence, is essential for group 1 capsule production in E. coli.
Certain Salmonella serotypes employ a Wzy polymerase that is located elsewhere in
the genome, outside the CPS cluster sequence (77), but this is not reported for E. coli.
Wzx, the O-antigen ippase protein, is also missing from the majority of these CPS
sequences. Only two of these genomes appear to have both Wzx and Wzy in the CPS
sequence, although it is possible this number could be increased by a few, since several
contigs are prematurely truncated upstream of gnd (Fig. 5). Thus, key components of the
Wzx/Wzy pathway are missing from these putative group 1 CPS cluster sequences. That
fact, combined with the relaxed selection evident for wza2 and wzi, suggests this CPS
locus is no longer functional as a group 1 capsule assembly system. More specically,
we suggest that a recombinant history, at least partially mediated by the transposons
associated with these capsule sequences, resulted in shuing the gene content in these
CPS, rendering it non-functional for capsule production, and this in turn is reected
in the relaxed selection for wza and wzi. Although our examination of Wza protein
structure proved inconclusive with regard to distinct patterns associated with resistance,
our proposed model would still be that the antibiotic may get trapped in the pore of
Wza2, but this could be due to a host of interactive anomalies either involving the core
components of Wzi, Wza2, Wzb, and Wzc or involving other missing components in the
biosynthetic region of this operon.
All 38 of our isolates carrying the CPS locus depicted in Fig. 5 carry the group 4 cluster
sequence, with its wza homolog, gfcE. Our working hypothesis is that these isolates
Full-Length Text Applied and Environmental Microbiology
August 2024 Volume 90 Issue 8 10.1128/aem.00354-24 15
originally carried the group 4 CPS capability and then independently acquired the group
1 CPS via LGT from various lineages. The surrounding gene content of the group 4 CPS
is conserved in Wza2 isolates, with the seemingly requisite operon composition for a
functioning group 4 capsule (78, 79) and is similarly conserved in non-Wza2 isolates
from our collection, suggesting an ancient syntenic inheritance. In contrast, independent
LGT events are suggested by the gene content variability of the group 1 CPS between
Wza2 isolates, and the wide diversity in STs of this group. Homologous recombination
events are undoubtedly part of the history of this group 1 CPS locus, and some of
this recombination likely involved the colanic acid operon. All E. coli group 1-producing
isolates are unable to produce colanic acid, and it has been suggested that this locus may
have been lost during chromosomal rearrangements involving group 1 CPS loci (57).
In our case, the same transposon is present just upstream of galF in our group
1 sequences (Fig. 4) and is just upstream of wza in the vast majority of colanic acid
operons in our set of sequences from dogs. IS elements have been implicated previously
in the history of group 1 CPS and suggested as a likely factor continuing to mediate
recombination between strains (80). Colanic acid and group 1 capsules have similar
chemical structure (57) and share similarities in the operon structure, and the wza, wzb,
and wzc genes can be interchanged (74). Our wza2 gene tree phylogeny identies
an identical sequence clade of wza2 sequences, all from ST162, that include in their
respective cluster sequence, at least one gene locus common in colanic acid operons,
gmd, involved in GDP-fucose biosynthesis from mannose-6-phosphate (Man6P). All of
this points to an extensive, and likely ongoing, history of LGT between colanic acid
and group 1. ST162 is a globally distributed MDR pandemic lineage that has been
isolated from clinical, environmental, domestic, and wild animal samples (81), including
an endangered species, the Andean condor (82), and, most recently, a pygmy sperm
whale (83). Our results suggest the possibility that wza2 and/or the group 1 CPS from
ST162 could be a contributing factor to its MDR phenotype.
Our hypothesis is that the majority of our Wza2 group 1 CPS have lost their previ
ous role in capsule production, due to extensive recombination and the consequent
loss of certain critical genes. This loss of function is further supported by the signal
of relaxed selection we see in genes from this CPS locus that were correlated with
resistance. Ultimately, we suggest this has had the unexpected eect of these G1C
serendipitously developing into an MDR AMR mechanism. The majority of known AMR
mechanisms involve functional loci, and mutations therein, resulting in (i) changes
in membrane permeability, which limit the uptake, (ii) modication of a target, (iii)
enzymatic inactivation, and (iv) active eux. There are, however, other examples in
Gram-negative bacteria where an analogous loss of function to our proposal for the
Wza2 operon results in AMR. For example, colistin resistance in Acinetobacter bauman
nii is due to mutations within genes essential for lipid A biosynthesis (lpx operon),
eliminating the ability to produce lipid A and therefore lipopolysaccharide (LPS) (84).
An important dierence in this A. baumannii study compared with our proposal is that
colistin resistance due to loss of LPS has the corollary eect of increased sensitivity to
other clinically relevant antibiotics; we see no such evidence of that in the wza2/group 1
isolates examined here.
The Wza2 cluster sequences are very widely distributed, nearly intact, across a
number of bacteria species (Fig. S8), particularly, E. coli and Klebsiella pneumoniae,
isolated from dierent host species (including other mammals, birds, plants, and y).
Whether one considers Wza2 as the primary resistance mechanism, or the probable
inactive group 1 CPS cluster, the wide dispersal range of these sequences suggests that
this resistant CPS may be a much more important source of AMR than just in this set
of dog isolates. Although Wzi and Wza2 were two of only a small handful of putatively
novel chromosomal loci correlated with resistance to multiple antimicrobials, it is also
true that not all of these 38 genomes carrying these two genes were devoid of any
possible confounding inuence of known resistance mechanisms, of either plasmid or
chromosomal origin. We focused exclusively on gene presence/absence here, and our
Full-Length Text Applied and Environmental Microbiology
August 2024 Volume 90 Issue 8 10.1128/aem.00354-24 16
study did not include mutational resistance. Given the large number of known AMR
mechanisms in E. coli, and in our specic data set (Fig. S5), this is to be expected.
However, in further attempts to specically associate a cause-eect relationship of these
loci to AMR, it will be important to eliminate the confounding inuence of known AMR
mechanisms on the role of these putative novel antimicrobial resistance genes. Ongoing
and future work in our laboratories regarding Wzi/Wza2 will involve knockouts of these
genes and others in the Wza2 CPS cluster, within several strains of the germane genetic
background that are devoid of known resistance mechanisms. Another future direction
needed is to characterize any capsules that may be formed by these strains.
As genomic surveillance expands to additional hosts, microbes, and geographic
regions, the importance of collecting paired AST data cannot be understated. The
merging and binary coding of this data is challenging, particularly for microbe-antimi
crobial combinations that do not have breakpoints available, or when the dilutions on
commercial panels do not have a sucient range for an accurate MIC determination.
Dierent laboratories also use dierent platforms, methods, and criteria for interpreta
tion. Additionally, some breakpoints in veterinary medicine are intentionally set low
to discourage use. Here, we conservatively chose to include only the denitive interpre
tations provided to clinicians, omitting the intermediate phenotypes and the MIC or
disk diusion diameters where no interpretation was provided to the clinician. Due to
the limitations imposed by the aforementioned factors, we chose to focus on genes
associated with eight or more antimicrobials after correction for repeated measures.
Interestingly, with this strict criterion, we found dierent versions of the same gene
(wza) to be associated with either susceptibility or resistance (Table S1). The strong
association with cefovecin observed may have been facilitated by heavy demand for
susceptibility testing of this third-generation cephalosporin in dogs (and cats). As we also
saw resistance to antimicrobials of human medical importance in these same dog strains,
stewardship of these treatments in our companion animals is of utmost importance.
ACKNOWLEDGMENTS
The authors are grateful to Chris Whiteld and Michael Feldgarden for helpful discus
sions. Trevor L. Alexander, Melanie Prarat, Ashley Johnson, Katherine Shiplett, Dominika
Jurkovic,and Jing Cui provided excellent technical support.
This work was supported (FOA PAR-18–604) and performed in collaboration with the
US Food and Drug Administration’s Veterinary Laboratory Investigation and Response
Network under grants 1U18FD006993 to Cornell; 1U18FD006442 to Louisiana State;
1U18FD006712 to Ohio ADDL; 1U18FD006862 and 1U18FD007244-to University of
Guelph; 1U18FD006558 to South Dakota, 1U18FD006453- to Washington State.
AUTHOR AFFILIATIONS
1Cornell University, Ithaca, New York, USA
2Université Paris-Saclay, INRAE, UVSQ, Virologie et Immunologie Moléculaires, Jouy-en-
Josas, Paris, France
3San José State University, San José, California, USA
4Washington Animal Disease Diagnostic Laboratory, Department of Veterinary Microbiol
ogy and Pathology, Washington State University, Pullman, Washington, USA
5Louisiana Animal Disease Diagnostic Laboratory, School of Veterinary Medicine,
Louisiana State University, Baton Rouge, Louisiana, USA
6Ohio Department of Agriculture Animal Disease Diagnostic Laboratory, Reynoldsburg,
Ohio, USA
7University of Guelph, Animal Health Laboratory, Guelph, Ontario, Canada
8South Dakota State University, Brookings, South Dakota, USA
9US Food and Drug Administration, Veterinary Laboratory Investigation and Response
Network, Laurel, Maryland, USA
Full-Length Text Applied and Environmental Microbiology
August 2024 Volume 90 Issue 8 10.1128/aem.00354-24 17
PRESENT ADDRESS
Yan Zhang, Arizona Veterinary Diagnostic Laboratory, Tucson, Arizona, USA
Prabhjot kaur Sekhon, Oklahoma state University, Stillwater, Oklahoma, USA
AUTHOR ORCIDs
Kristina Ceres http://orcid.org/0000-0002-6682-0031
Jordan D. Zehr http://orcid.org/0000-0003-2099-4172
Gregory H. Tyson http://orcid.org/0000-0002-2729-5035
Michael J. Stanhope http://orcid.org/0000-0002-4590-1529
Laura B. Goodman http://orcid.org/0000-0002-8327-3092
FUNDING
Funder Grant(s) Author(s)
HHS | U.S. Food and Drug
Administration (FDA) 1U18FD006993, 5U18FD006716 Michael J.
Stanhope
Laura B. Goodman
HHS | U.S. Food and Drug
Administration (FDA) 1U18FD006442 Laura Peak
HHS | U.S. Food and Drug
Administration (FDA) 1U18FD006712 Yan Zhang
HHS | U.S. Food and Drug
Administration (FDA) 1U18FD006862,1U18FD007244 Durda Slavic
HHS | U.S. Food and Drug
Administration (FDA) 1U18FD006558 Prabhjot kaur
Sekhon
HHS | U.S. Food and Drug
Administration (FDA) 1U18FD006453 Claire R. Burbick
DATA AVAILABILITY
All genomes are available in the following NCBI BioProjects: PRJNA318591,
PRJNA324565, PRJNA324573, PRJNA481346, PRJNA503851, and PRJNA318589, all under
the umbrella BioProject PRJNA314609. All RNA-seq data have been deposited to
NCBI Sequence Read Archive (SRA) under the BioProject accession PRJNA1053349. All
alignments, inferred phylogenies, and raw output les are available at the Cornell
University eCommons Repository under https://doi.org/10.7298/pdzk-rq02.2.
ADDITIONAL FILES
The following material is available online.
Supplemental Material
Supplemental gures (AEM00354-24-s0001.docx). Figures S1 to S9.
Table S1 (AEM00354-24-s0002.xlsx). Summary of gene groups associated with
resistance or susceptibility to 8 or more antibiotics.
Table S2 (AEM00354-24-s0003.xlsx). Summary of strains carrying wza2 including
metadata and antibiotic susceptibility phenotypes based on MIC testing with VET01S
interpretations. Where dog breakpoints were not available, the human breakpoint was
applied.
REFERENCES
1. Ventola CL. 2015. The antibiotic resistance crisis: part 1: causes and
threats. P T 40:277–283.
2. Ceric O, Tyson GH, Goodman LB, Mitchell PK, Zhang Y, Prarat M, Cui J,
Peak L, Scaria J, Antony L, etal. 2019. Enhancing the one health initiative
by using whole genome sequencing to monitor antimicrobial resistance
of animal pathogens: Vet-LIRN collaborative project with veterinary
Full-Length Text Applied and Environmental Microbiology
August 2024 Volume 90 Issue 8 10.1128/aem.00354-24 18
diagnostic laboratories in United States and Canada. BMC Vet Res
15:130. https://doi.org/10.1186/s12917-019-1864-2
3. Harrison L, Tyson GH, Strain E, Lindsey RL, Strockbine N, Ceric O,
Fortenberry GZ, Harris B, Shaw S, Tillman G, Zhao S, Dessai U. 2022. Use
of large-scale genomics to identify the role of animals and foods as
potential sources of extraintestinal pathogenic Escherichia coli that
cause human illness. Foods 11:1975. https://doi.org/10.3390/
foods11131975
4. Elankumaran P, Cummins ML, Browning GF, Marenda MS, Reid CJ,
Djordjevic SP. 2022. Genomic and temporal trends in canine ExPEC
reect those of human ExPEC. Microbiol Spectr 10:e0129122. https://doi.
org/10.1128/spectrum.01291-22
5. Brynildsrud O, Bohlin J, Scheer L, Eldholm V. 2016. Rapid scoring of
genes in microbial pan-genome-wide association studies with Scoary.
Genome Biol 17:238. https://doi.org/10.1186/s13059-016-1108-8
6. Beavan AJS, Domingo-Sananes MR, McInerney JO. 2024. Contingency,
repeatability, and predictability in the evolution of a prokaryotic
pangenome. Proc Natl Acad Sci U S A 121:e2304934120. https://doi.org/
10.1073/pnas.2304934120
7. Tantoso E, Eisenhaber B, Kirsch M, Shitov V, Zhao Z, Eisenhaber F. 2022.
To kill or to be killed: pangenome analysis of Escherichia coli strains
reveals a tailocin specic for pandemic ST131. BMC Biol 20:146. https://
doi.org/10.1186/s12915-022-01347-7
8. Denamur E, Condamine B, Esposito-Farèse M, Royer G, Clermont O,
Laouenan C, Lefort A, de Lastours V, Galardini M, COLIBAFI, SEPTICOLI
groups. 2022. Genome wide association study of Escherichia coli
bloodstream infection isolates identies genetic determinants for the
portal of entry but not fatal outcome. PLoS Genet 18:e1010112. https://
doi.org/10.1371/journal.pgen.1010112
9. Li X, Lin J, Hu Y, Zhou J. 2020. PARMAP: a pan-genome-based computa
tional framework for predicting antimicrobial resistance. Front Microbiol
11:578795. https://doi.org/10.3389/fmicb.2020.578795
10. Elankumuran P, Browning GF, Marenda MS, Kidsley A, Osman M, Haenni
M, Johnson JR, Trott DJ, Reid CJ, Djordjevic SP. 2023. Identication of
genes inuencing the evolution of Escherichia coli ST372 in dogs and
humans. Microb Genom 9:mgen000930. https://doi.org/10.1099/mgen.
0.000930
11. Bourély C, Cazeau G, Jarrige N, Leblond A, Madec JY, Haenni M, Gay E.
2019. Antimicrobial resistance patterns of bacteria isolated from dogs
with otitis. Epidemiol Infect 147:e121. https://doi.org/10.1017/
S0950268818003278
12. Hata A, Fujitani N, Ono F, Yoshikawa Y. 2022. Surveillance of antimicro
bial-resistant Escherichia coli in sheltered dogs in the Kanto Region of
Japan. Sci Rep 12:773. https://doi.org/10.1038/s41598-021-04435-w
13. Li Y, Fernández R, Durán I , Molina-López RA, Darwich L. 2020. Antimicro
bial resistance in bacteria isolated from cats and dogs from the Iberian
Peninsula. Front Microbiol 11:621597. https://doi.org/10.3389/fmicb.
2020.621597
14. Osman M, Albarracin B, Altier C, Gröhn YT, Cazer C. 2022. Antimicrobial
resistance trends among canine Escherichia coli isolated at a New York
veterinary diagnostic laboratory between 2007 and 2020. Prev Vet Med
208:105767. https://doi.org/10.1016/j.prevetmed.2022.105767
15. CDC. 2022. Antibiotic resistance is a national priority. Centers for Disease
Control and Prevention. Available from: https://www.cdc.gov/
drugresistance/us-activities.html. Retrieved 23 Sep 2023.
16. Timme RE, Wolfgang WJ, Balkey M, Venkata SLG, Randolph R, Allard M,
Strain E. 2020. Optimizing open data to support one health: best
practices to ensure interoperability of genomic data from bacterial
pathogens. One Health Outlook 2:20. https://doi.org/10.1186/s42522-
020-00026-3
17. Souvorov A, Agarwala R, Lipman DJ. 2018. SKESA: strategic k-mer
extension for scrupulous assemblies. Genome Biol 19:153. https://doi.
org/10.1186/s13059-018-1540-z
18. Seemann T. 2014. Prokka: rapid prokaryotic genome annotation.
Bioinformatics 30:2068–2069. https://doi.org/10.1093/bioinformatics/
btu153
19. Tonkin-Hill G, MacAlasdair N, Ruis C, Weimann A, Horesh G, Lees JA,
Gladstone RA, Lo S, Beaudoin C, Floto RA, Frost SDW, Corander J, Bentley
SD, Parkhill J. 2020. Producing polished prokaryotic pangenomes with
the Panaroo pipeline. Genome Biol 21:180. https://doi.org/10.1186/
s13059-020-02090-4
20. Katoh K, Standley DM. 2013. MAFFT multiple sequence alignment
software version 7: improvements in performance and usability. Mol Biol
Evol 30:772–780. https://doi.org/10.1093/molbev/mst010
21. Castresana J. 2000. Selection of conserved blocks from multiple
alignments for their use in phylogenetic analysis. Mol Biol Evol 17:540–
552. https://doi.org/10.1093/oxfordjournals.molbev.a026334
22. Talavera G, Castresana J. 2007. Improvement of phylogenies after
removing divergent and ambiguously aligned blocks from protein
sequence alignments. Syst Biol 56:564–577. https://doi.org/10.1080/
10635150701472164
23. Nguyen L-T, Schmidt HA, von Haeseler A, Minh BQ. 2015. IQ-TREE: a fast
and eective stochastic algorithm for estimating maximum-likelihood
phylogenies. Mol Biol Evol 32:268–274. https://doi.org/10.1093/molbev/
msu300
24. Letunic I, Bork P. 2021. Interactive Tree Of Life (iTOL) v5: an online tool
for phylogenetic tree display and annotation. Nucleic Acids Res
49:W293–W296. https://doi.org/10.1093/nar/gkab301
25. Seemann T. 2023. Mlst tool. https://github.com/Tseemann/Mlst.
26. Jolley KA, Maiden MCJ. 2010. BIGSdb: scalable analysis of bacterial
genome variation at the population level. BMC Bioinformatics 11:595.
https://doi.org/10.1186/1471-2105-11-595
27. Bessonov K, Laing C, Robertson J, Yong I, Ziebell K, Gannon VPJ, Nichani
A, Arya G, Nash JHE, Christianson S. 2021. ECTyper: in silico Escherichia
coli serotype and species prediction from raw and assembled whole-
genome sequence data. Microb Genom 7:000728. https://doi.org/10.
1099/mgen.0.000728
28. Lam MMC, Wick RR, Watts SC, Cerdeira LT, Wyres KL, Holt KE. 2021. A
genomic surveillance framework and genotyping tool for Klebsiella
pneumoniae and its related species complex. 1. Nat Commun 12:4188.
https://doi.org/10.1038/s41467-021-24448-3
29. Tonkin-Hill G, Lees JA, Bentley SD, Frost SDW, Corander J. 2018.
RhierBAPS: an R implementation of the population clustering algorithm
hierBAPS. Wellcome Open Res 3:93. https://doi.org/10.12688/
wellcomeopenres.14694.1
30. Feldgarden M, Brover V, Gonzalez-Escalona N, Frye JG, Haendiges J, Haft
DH, Homann M, Pettengill JB, Prasad AB, Tillman GE, Tyson GH, Klimke
W. 2021. AMRFinderPlus and the Reference Gene Catalog facilitate
examination of the genomic links among antimicrobial resistance, stress
response, and virulence. 1. Sci Rep 11:12728. https://doi.org/10.1038/
s41598-021-91456-0
31. Andreopoulos WB, Geller AM, Lucke M, Balewski J, Clum A, Ivanova NN,
Levy A. 2022. Deeplasmid: deep learning accurately separates plasmids
from bacterial chromosomes. Nucleic Acids Res 50:e17. https://doi.org/
10.1093/nar/gkab1115
32. Wertheim JO, Murrell B, Smith MD, Kosakovsky Pond SL, Scheer K.
2015. RELAX: detecting relaxed selection in a phylogenetic framework.
Mol Biol Evol 32:820–832. https://doi.org/10.1093/molbev/msu400
33. Kosakovsky Pond SL, Posada D, Gravenor MB, Woelk CH, Frost SDW.
2006. GARD: a genetic algorithm for recombination detection.
Bioinformatics 22:3096–3098. https://doi.org/10.1093/bioinformatics/
btl474
34. Kozlov AM, Darriba D, Flouri T, Morel B, Stamatakis A. 2019. RAxML-NG: a
fast, scalable and user-friendly tool for maximum likelihood phyloge
netic inference. Bioinformatics 35:4453–4455. https://doi.org/10.1093/
bioinformatics/btz305
35. Shank SD, Weaver S, Kosakovsky Pond SL. 2018. phylotree.js - a
JavaScript library for application development and interactive data
visualization in phylogenetics. BMC Bioinformatics 19:276. https://doi.
org/10.1186/s12859-018-2283-2
36. Kosakovsky Pond SL, Poon AFY, Velazquez R, Weaver S, Hepler NL,
Murrell B, Shank SD, Magalis BR, Bouvier D, Nekrutenko A, Wisotsky S,
Spielman SJ, Frost SDW, Muse SV. 2020. HyPhy 2.5—a customizable
platform for evolutionary hypothesis testing using phylogenies. Mol Biol
Evol 37:295–299. https://doi.org/10.1093/molbev/msz197
37. Jumper J, Evans R, Pritzel A, Green T, Figurnov M, Ronneberger O,
Tunyasuvunakool K, Bates R, Žídek A, Potapenko A, et al. 2021. Highly
accurate protein structure prediction with AlphaFold. Nature 596:583–
589. https://doi.org/10.1038/s41586-021-03819-2
38. Mirdita M, Schütze K, Moriwaki Y, Heo L, Ovchinnikov S, Steinegger M.
2022. ColabFold: making protein folding accessible to all. Nat Methods
19:679–682. https://doi.org/10.1038/s41592-022-01488-1
Full-Length Text Applied and Environmental Microbiology
August 2024 Volume 90 Issue 8 10.1128/aem.00354-24 19
39. Pettersen EF, Goddard TD, Huang CC, Meng EC, Couch GS, Croll TI, Morris
JH, Ferrin TE. 2021. UCSF ChimeraX: structure visualization for
researchers, educators, and developers. Protein Sci 30:70–82. https://doi.
org/10.1002/pro.3943
40. Dong C, Beis K, Nesper J, Brunkan-Lamontagne AL, Clarke BR, Whiteld
C, Naismith JH. 2006. Wza the translocon for E. coli capsular polysacchar
ides denes a new class of membrane protein. Nature 444:226–229.
https://doi.org/10.1038/nature05267
41. Xu J, Zhang Y. 2010. How signicant is a protein structure similarity with
TM-score = 0.5?. Bioinformatics 26:889–895. https://doi.org/10.1093/
bioinformatics/btq066
42. Zhang Y, Skolnick J. 2004. Scoring function for automated assessment of
protein structure template quality. Proteins: Structure, Function, and
Bioinformatics 57:702–710. https://doi.org/10.1002/prot.20264
43. CLSI. 2020. Performance standards for antimicrobial disk and dilution
susceptibility tests for bacteria isolated from animals. 5th ed. Clinical and
Laboratory Standards Institute, USA.
44. Krueger F. 2024. Trimgalore tool. https://github.com/FelixKrueger/
TrimGalore.
45. Martin M. 2011. Cutadapt removes adapter sequences from high-
throughput sequencing reads. EMBnet j 17:10. https://doi.org/10.14806/
ej.17.1.200
46. Andrews S. 2010. FastQC: a quality control tool for high throughput
sequence data. Available from: https://www.bioinformatics.babraham.
ac.uk/projects/fastqc
47. Dobin A, Davis CA, Schlesinger F, Drenkow J, Zaleski C, Jha S, Batut P,
Chaisson M, Gingeras TR. 2013. STAR: ultrafast universal RNA-seq aligner.
Bioinformatics 29:15–21. https://doi.org/10.1093/bioinformatics/bts635
48. Love MI, Huber W, Anders S. 2014. Moderated estimation of fold change
and dispersion for RNA-seq data with DESeq2. Genome Biol 15:550.
https://doi.org/10.1186/s13059-014-0550-8
49. Varet H, Brillet-Guéguen L, Coppée J-Y, Dillies M-A. 2016. SARTools: a
DESeq2- and EdgeR-based R pipeline for comprehensive dierential
analysis of RNA-Seq data. PLoS ONE 11:e0157022. https://doi.org/10.
1371/journal.pone.0157022
50. Page AJ, Cummins CA, Hunt M, Wong VK, Reuter S, Holden MTG, Fookes
M, Falush D, Keane JA, Parkhill J. 2015. Roary: Rapid large-scale
Prokaryote pan genome analysis. Bioinformatics 31:3691–3693. https://
doi.org/10.1093/bioinformatics/btv421
51. Cusumano CK, Hung CS, Chen SL, Hultgren SJ. 2010. Virulence plasmid
harbored by uropathogenic Escherichia coli functions in acute stages of
pathogenesis. Infect Immun 78:1457–1467. https://doi.org/10.1128/IAI.
01260-09
52. Riley LW. 2014. Pandemic lineages of extraintestinal pathogenic
Escherichia coli. Clin Microbiol Infect 20:380–390. https://doi.org/10.
1111/1469-0691.12646
53. Forde BM, Roberts LW, Phan M-D, Peters KM, Fleming BA, Russell CW,
Lenherr SM, Myers JB, Barker AP, Fisher MA, Chong T-M, Yin W-F, Chan K-
G, Schembri MA, Mulvey MA, Beatson SA. 2019. Population dynamics of
an Escherichia coli ST131 lineage during recurrent urinary tract infection.
Nat Commun 10:3643. https://doi.org/10.1038/s41467-019-11571-5
54. Elankumaran P, Browning GF, Marenda MS, Reid CJ, Djordjevic SP. 2022.
Close genetic linkage between human and companion animal
extraintestinal pathogenic Escherichia coli ST127. Curr Res Microb Sci
3:100106. https://doi.org/10.1016/j.crmicr.2022.100106
55. Wijetunge DSS, Gongati S, DebRoy C, Kim KS, Couraud PO, Romero IA,
Weksler B, Kariyawasam S. 2015. Characterizing the pathotype of
neonatal meningitis causing Escherichia coli (NMEC). BMC Microbiol
15:211. https://doi.org/10.1186/s12866-015-0547-9
56. Whiteld C, Wear SS, Sande C. 2020. Assembly of bacterial capsular
polysaccharides and exopolysaccharides. Annu Rev Microbiol 74:521–
543. https://doi.org/10.1146/annurev-micro-011420-075607
57. Whiteld C. 2006. Biosynthesis and assembly of capsular polysacchar
ides in Escherichia coli. Annu Rev Biochem 75:39–68. https://doi.org/10.
1146/annurev.biochem.75.103004.142545
58. Stegemann MR, Passmore CA, Sherington J, Lindeman CJ, Papp G,
Weigel DJ, Skogerboe TL. 2006. Antimicrobial activity and spectrum of
cefovecin, a new extended- spectrum cephalosporin, against pathogens
collected from dogs and cats in Europe and North America. Antimicrob
Agents Chemother 50:2286–2292. https://doi.org/10.1128/AAC.00077-
06
59. Sun E, Graves ML, Oliver DC. 2020. Propelling a course-based undergrad
uate research experience using an open-access online undergraduate.
Front Microbiol 11:589025. https://doi.org/10.3389/fmicb.2020.589025
60. Murail S, de Vries SJ, Rey J, Moroy G, Tuéry P. 2021. SeamDock: an
interactive and collaborative online docking resource to assist small
compound molecular docking. Front Mol Biosci 8:716466. https://doi.
org/10.3389/fmolb.2021.716466
61. Robert X, Gouet P. 2014. Deciphering key features in protein structures
with the new ENDscript server. Nucleic Acids Res 42:W320–W324. https:/
/doi.org/10.1093/nar/gku316
62. Nemser SM, Doran T, Grabenstein M, McConnell T, McGrath T,
Pamboukian R, Smith AC, Achen M, Danzeisen G, Kim S, Liu Y, Robeson S,
Rosario G, McWilliams Wilson K, Reimschuessel R. 2014. Investigation of
Listeria, Salmonella, and toxigenic Escherichia coli in various pet foods.
Foodborne Pathog Dis 11:706–709. https://doi.org/10.1089/fpd.2014.
1748
63. Martins LRL, Pina SMR, Simões RLR, de Matos AJF, Rodrigues P, da Costa
PMR. 2013. Common phenotypic and genotypic antimicrobial resistance
patterns found in a case study of multiresistant E. coli from cohabitant
pets, humans, and household surfaces. J Environ Health 75:74–81.
64. Cole SD, Perez-Bonilla D, Hallowell A, Redding LE. 2022. Carbapenem
prescribing at a veterinary teaching hospital before an outbreak of
carbapenem-resistant Escherichia coli. J Small Anim Pract 63:442–446.
https://doi.org/10.1111/jsap.13481
65. Menard J, Goggs R, Mitchell P, Yang Y, Robbins S, Franklin-Guild RJ,
Thachil AJ, Altier C, Anderson R, Putzel GG, McQueary H, Goodman LB.
2022. Eect of antimicrobial administration on fecal microbiota of
critically ill dogs: dynamics of antimicrobial resistance over time. Anim
Microbiome 4:36. https://doi.org/10.1186/s42523-022-00178-9
66. Flament-Simon S-C, Toro M de, García V, Blanco JE, Blanco M, Alonso MP,
Goicoa A, Díaz-González J, Nicolas-Chanoine M-H, Blanco J. 2020.
Molecular characteristics of extraintestinal pathogenic E. coli (ExPEC),
uropathogenic E. coli (UPEC), and multidrug resistant E. coli isolated from
healthy dogs in Spain. whole genome sequencing of canine ST372
isolates and comparison with human isolates causing extraintestinal
infections. Microorganisms 8:1712. https://doi.org/10.3390/microorgan
isms8111712
67. Wassenaar TM. 2016. Insights from 100 years of research with probiotic
E. coli. EuJMI 6:147–161. https://doi.org/10.1556/1886.2016.00029
68. Rivera-Betancourt M, Keen JE. 2001. Murine monoclonal antibodies
against Escherichia coli O4 lipopolysaccharide and H5 agellin. J Clin
Microbiol 39:3409–3413. https://doi.org/10.1128/JCM.39.9.3409-3413.
2001
69. Liu CM, Aziz M, Park DE, Wu Z, Stegger M, Li M, Wang Y, Schmidlin K,
Johnson TJ, Koch BJ, Hungate BA, Nordstrom L, Gauld L, Weaver B,
Rolland D, Statham S, Hall B, Sariya S, Davis GS, Keim PS, Johnson JR,
Price LB. 2023. Using source-associated mobile genetic elements to
identify zoonotic extraintestinal E. coli infections. One Health 16:100518.
https://doi.org/10.1016/j.onehlt.2023.100518
70. Slack MP, Nichols WW. 1982. Antibiotic penetration through bacterial
capsules and exopolysaccharides. J Antimicrob Chemother 10:368–372.
https://doi.org/10.1093/jac/10.5.368
71. Nickerson NN, Mainprize IL, Hampton L, Jones ML, Naismith JH, Whiteld
C. 2014. Trapped translocation intermediates establish the route for
export of capsular polysaccharides across Escherichia coli outer
membranes. Proc Natl Acad Sci U S A 111:8203–8208. https://doi.org/10.
1073/pnas.1400341111
72. Su A, Wang A, Yeo L. 2017. Deletion of the group 1 capsular gene Wza in
Escherichia coli E69 confers resistance to the antibiotic erythromycin on
solid media but not in liquid media | jemi.microbiology.ubc.ca. JEMI 3:1–
8.
73. Yuen B, Ting J, Kang K, Wong T. 2017. Investigation of Wza in erythromy
cin sensitivity of Escherichia coli K30 E69 by genetic complementation.
JEMI 21:52–57.
74. Reid AN, Whiteld C. 2005. Functional analysis of conserved gene
products involved in assembly of Escherichia coli capsules and
exopolysaccharides: evidence for molecular recognition between Wza
and Wzc for colanic acid biosynthesis. J Bacteriol 187:5470–5481. https://
doi.org/10.1128/JB.187.15.5470-5481.2005
75. Collins RF, Beis K, Dong C, Botting CH, McDonnell C, Ford RC, Clarke BR,
Whiteld C, Naismith JH. 2007. The 3D structure of a periplasm-spanning
Full-Length Text Applied and Environmental Microbiology
August 2024 Volume 90 Issue 8 10.1128/aem.00354-24 20
platform required for assembly of group 1 capsular polysaccharides in
Escherichia coli. Proc Natl Acad Sci U S A 104:2390–2395. https://doi.org/
10.1073/pnas.0607763104
76. Liu B, Furevi A, Perepelov AV, Guo X, Cao H, Wang Q, Reeves PR, Knirel
YA, Wang L, Widmalm G. 2020. Structure and genetics of Escherichia coli
O antigens. FEMS Microbiol Rev 44:655–683. https://doi.org/10.1093/
femsre/fuz028
77. Liu B, Knirel YA, Feng L, Perepelov AV, Senchenkova SN, Reeves PR, Wang
L. 2014. Structural diversity in Salmonella O antigens and its genetic
basis. FEMS Microbiol Rev 38:56–89. https://doi.org/10.1111/1574-6976.
12034
78. Larson MR, Biddle K, Gorman A, Boutom S, Rosenshine I, Saper MA. 2021.
Escherichia coli O127 group 4 capsule proteins assemble at the outer
membrane. PLoS One 16:e0259900. https://doi.org/10.1371/journal.
pone.0259900
79. Peleg A, Shifrin Y, Ilan O, Nadler-Yona C, Nov S, Koby S, Baruch K, Altuvia
S, Elgrably-Weiss M, Abe CM, Knutton S, Saper MA, Rosenshine I. 2005.
Identication of an Escherichia coli operon required for formation of the
O-antigen capsule. J Bacteriol 187:5259–5266. https://doi.org/10.1128/
JB.187.15.5259-5266.2005
80. Rahn A, Drummelsmith J, Whiteld C. 1999. Conserved organization in
the cps gene clusters for expression of Escherichia coli group 1 K
antigens: relationship to the colanic acid biosynthesis locus and the cps
genes from Klebsiella pneumoniae. J Bacteriol 181:2307–2313. https://
doi.org/10.1128/JB.181.7.2307-2313.1999
81. Fuentes-Castillo D, Navas-Suárez PE, Gondim MF, Esposito F, Sacristán C ,
Fontana H, Fuga B, Piovani C, Kooij R, Lincopan N, Catão-Dias JL. 2021.
Genomic characterization of multidrug-resistant ESBL-producing
Escherichia coli ST58 causing fatal colibacillosis in critically endangered
Brazilian merganser (Mergus octosetaceus). Transbound Emerg Dis
68:258–266. https://doi.org/10.1111/tbed.13686
82. Fuentes-Castillo D, Esposito F, Cardoso B, Dalazen G, Moura Q, Fuga B,
Fontana H, Cerdeira L, Dropa M, Rottmann J, González-Acuña D, Catão-
Dias JL, Lincopan N. 2020. Genomic data reveal international lineages of
critical priority Escherichia coli harbouring wide resistome in Andean
condors (Vultur gryphus Linnaeus, 1758). Mol Ecol 29:1919–1935. https://
doi.org/10.1111/mec.15455
83. Sellera FP, Cardoso B, Fuentes-Castillo D, Esposito F, Sano E, Fontana H,
Fuga B, Goldberg DW, Seabra LAV, Antonelli M, Sandri S, Kolesnikovas
CKM, Lincopan N. 2022. Genomic analysis of a highly virulent NDM-1-
producing Escherichia coli ST162 infecting a pygmy sperm whale (Kogia
breviceps) in South America. Front Microbiol 13:915375. https://doi.org/
10.3389/fmicb.2022.915375
84. Moatt JH, Harper M, Harrison P, Hale JDF, Vinogradov E, Seemann T,
Henry R, Crane B, St Michael F, Cox AD, Adler B, Nation RL, Li J, Boyce JD.
2010. Colistin resistance in Acinetobacter baumannii is mediated by
complete loss of lipopolysaccharide production. Antimicrob Agents
Chemother 54:4971–4977. https://doi.org/10.1128/AAC.00834-10
Full-Length Text Applied and Environmental Microbiology
August 2024 Volume 90 Issue 8 10.1128/aem.00354-24 21
Content uploaded by Jean Kaoru Millet
Author content
All content in this area was uploaded by Jean Kaoru Millet on Jul 23, 2024
Content may be subject to copyright.
