Acquisition of genomic elements were pivotal for the success of Escherichia coli ST410

Abstract

Background

Escherichia coli ST410 is an emerging MDR clone linked to bla_CTX-M-15 and bla_OXA-181. Limited comprehensive data about the global distribution of ST410 clades and mobile genetic elements associated with different β-lactamases are available.

Methods

Short- and long-read WGS were performed on a collection of ST410 producing carbapenemases (n = 45) obtained from 11 countries. The evolutionary history of global E. coli ST410 was also investigated.

Results

OXA-181 and NDM-5 were the most frequent carbapenemases and used different underlying strategies to ensure their successful association with ST410 clades. Our phylogenetic analysis of publicly available ST410 genomes amended the previously published ST410 B subclades: ST410-B1 is identical to B1/H24, ST410-B2 includes B2/H24R and B3/H24Rx, while ST410-B3 corresponds to B4/H24RxC. Long-read WGS identified the following genomic events that likely shaped the evolution of ST410-B3: (i) gyrA and parC mutations were acquired via homologous recombination events; (ii) chromosomal integration of bla_CMY-2 among ST410-B3; (iii) the emergence of ST410-B3 from ST410-B2 was accompanied by the replacement of IncFII plasmids harbouring bla_CTX-M-15 (i.e. F36:31:A4:B1 in ST410-B2 with F1:A1:B49 plasmids in ST410-B3); and (iv) the NDM-5 gene was integrated within F1:A1:B49 plasmids over time.

Conclusions

The global ST410 population producing carbapenemases is dominated by the ST410-B2 and B3 subclades with varied geographical distribution that requires ongoing genomic surveillance. We provided an updated timeline of pivotal genomic events that have shaped the success of the ST410-B3 subclade.

Introduction

Antimicrobial resistance (AMR) is one of the greatest threats to human health.¹ The major burden of AMR is due to the global spread of certain ‘successful’ or ‘high-risk’ MDR clones, and/or the movement of AMR genes between diverse clones.^2,3

Extra-intestinal Escherichia coli (ExPEC) is the most common cause of community-acquired and healthcare-associated urinary and bloodstream infections worldwide.⁴ The population structure of human ExPEC is dominated by the following STs: ST131, ST69, ST10, ST405, ST38, ST95, ST648, ST73, ST410, ST393, ST354, ST12, ST127, ST167, ST58, ST617, ST88, ST23, ST117 and ST1193.⁵ The prevalence of these STs depends on geographical location, inclusion criteria, sources and time periods of studies.

Global surveillance studies showed that E. coli is the second most common Enterobacterales species with carbapenemases.^6,7 Tracking global MDR E. coli clones and associated mobile genetic elements is a public health priority.⁸ This will aid in designing management and prevention strategies.

A recent genomic survey of carbapenemase-producing E. coli from 36 countries identified ST410 as the most common clone among this population.⁹ The first published reports of ST410 appeared in 2011 among Brazilian E. coli with bla_CTX-M-15¹⁰ and British E. coli with bla_NDM-1.¹¹ ST410 with bla_OXA-181 was first reported in 2015¹² and bla_NDM-5 in 2018.¹³ ST410 played an important role in the global distribution of bla_OXA-181.^14,15

ST410 belongs to two clades, namely A/H53 and B/H24.¹⁵ Clade B is divided into the flowing subclades: B1/H24, B2/H24R, B3/H24Rx and B4/H24RxC.¹⁵ The B2/H24R subclade is associated with fluoroquinolone/ciprofloxacin resistance, B3/H24Rx with fluoroquinolone resistance and bla_CTX-M-15, while B4/H24RxC is linked with fluoroquinolone resistance, bla_CTX-M-15 and bla_OXA-181.¹⁵

Comprehensive data about the global distribution of ST410 clades and mobile genetic elements linked with different carbapenemases are limited. This study used short- and long-read WGS to characterize mobile genetic elements among a global collection of ST410 isolates (n = 45).⁹ We also investigated the evolutionary history of global E. coli ST410 with carbapenemases. Our results showed that specific mobile genetic elements and chromosomal insertions are crucial for the global dominance of the B4/H24RxC clade.

Materials and methods

Bacterial isolates

We included 45 clinical, non-repeat E. coli ST410 producing carbapenemases from two global surveillance programmes, namely SMART and INFORM (2015–17) (Table S1, Figure S1). The SMART and INFORM isolates initially underwent phenotypic identification and microdilution panel susceptibility testing.^6,7 Carbapenem (ertapenem, imipenem, meropenem) non-susceptible isolates (using CLSI guidelines) underwent molecular screening for bla_KPC, bla_VIM, bla_NDM, bla_OXA-48-like, bla_IMP and bla_GES as described previously.^6,7 All E. coli producing carbapenemases from 2015 to 2017 (n = 229) underwent short-read WGS; 45/229 (20%) were identified as ST410⁹ and included in this study.

Genomic analysis

E. coli ST410 producing carbapenemases (n = 45) underwent short- (average length 141 nt and depth 146×) and long-read (average length 3.8 kb and depth 88×) WGS, using procedures described previously.^16–18 The Nextera XT DNA sample preparation kit (Illumina, San Diego, CA, USA) was used to prepare short-read libraries. Samples were multiplexed and sequenced on an Illumina NovaSeq for 300 cycles (151 bp paired end). Draft genomes were obtained using SPAdes version 3.15.0.¹⁹ MinION (Nanopore Technologies, Oxford, UK) long-read sequencing libraries were prepped on all the isolates using the 1D rapid kit (SQK-RBK 004) and R9.4 flowcells and run on MinKNOW v5.0.0. Hybrid assembly was performed using Unicycler v0.4.9.¹⁸ To define the presence of genes and mutations, BLAST,²⁰ in combination with the following databases or typing schemes were accessed: NCBI Bacterial Antimicrobial Resistance Reference Gene Database (https://www.ncbi.nlm.nih.gov/bioproject/PRJNA313047), ResFinder (4.0),²¹ PlasmidFinder (2.2.1)²² and MLST.²³ Plasmid MLST was characterized using PubMLST online tools (https://pubmlst.org/organisms/plasmid-mlst). Insertion elements and transposons were queried against ISfinder²⁴ and Tn Number Registry database,²⁵ while restriction modification and toxin/antitoxin systems were mined using REBASE,²⁶ TADB 2.0²⁷ and T1TAdb (Release 5) database.²⁸E. coli fimH alleles were determined by FimTyper v1.0.²⁹

E. coli genome assemblies from NCBI RefSeq database (n = 27 738, dated as of 21 March 2022) were downloaded and a total of 442 were identified as ST410 using MLST (2.22.0, https://github.com/tseemann/mlst). The available raw reads from the public ST410 genomes (n = 143) were downloaded from the NCBI SRA database (Table S1).

Core SNP phylogenetic analysis was conducted as previously described.^16,30 In brief, raw reads from each isolate were mapped to ST410 clade B3 strain 2766 (sequenced in current study) by Snippy 4.4 (https://github.com/tseemann/snippy) using default settings. If the raw reads were not available, 10 M 150 bp paired-end simulated reads generated by the wgsim (https://github.com/lh3/wgsim) algorithm from SAMtools v1.13, were used.³¹ SNPs in the prophages and repeated regions were removed as previously described.³² Recombination regions were determined using Gubbins v3.1.5.³³ Bayesian population structure analysis was conducted using rhierBAPS v1.0.1 in R4.1.³⁴ Bayesian dating analysis was conducted as previously described³² using the recombination-corrected tree from Gubbins v3.1.5³³ output and the isolation time as the inputs in BactDating v1.1 in R4.1.³⁵ The additive relaxed clock (ARC) model³⁶ was determined to be the best model using the model compare function of the BactDating package, and 10⁷ iterations were used to ensure that the Markov chain Monte Carlo (MCMC) was run for long enough for convergence (the effective sample size of the inferred parameters α, μ and σ were >200). The potential donor genomes for the recombination regions were detected using Mash v2.3³⁷ to screen the non-ST410 E. coli strains from the NCBI RefSeq database. The genomes with the lowest MinHash Jaccard similarity coefficients were used to compare the SNP density with ST410 strains. Plasmid query coverages were determined using the bla_CXT-M-15 IncF plasmids from strains AZ1352 (clade B3/H24Rx) and 2771 (clade B4/H24RxC) as the references, and the sequences of the assemblies or complete genomes were BLAST against the two reference plasmids at an identity ≥95% and E-value ≤10e−10 cut-offs to determine the plasmid query coverages.

The ST410 Roer study divided ST410 into two clades (A and B) with four B subclades (B1 to B4).¹⁵ Clade A differed from B by different fimH alleles (A with fimH53 and B with fimH24). The different B subclades were mainly based on the MDR genotypes: B1/H24 [no fluoroquinolone resistance (FQ-R) mutation and bla_CTX-M-15], B2/H24R (FQ-R mutations without bla_CTX-M-15), B3/H24Rx (FQ-R mutations with bla_CTX-M-15), and B4/H24RxC (FQ-R mutations with bla_CTX-M-15 and bla_OXA-181). In our phylogenetic analysis of 614 [569 publicly available ST410 genomes and the 45 genomes from our current study (Table S1, Table S2)], the correlation between phylogenetic clusters and MDR genotypes were blurred since we described B3 and B4 subclade isolates without bla_CTX-M-15 and bla_OXA-181 (Figure S1). Therefore, an updated phylogenetic clustering is more appropriate (details below).

The sequencing data was deposited in the NCBI database (BioProject PRJNA780590).

Ethics

Ethics approval for this study was obtained through the University of Calgary Conjoint Health Research Ethics Board (REB17-1010).

Results

OXA-181 and NDM-5 were the most frequent carbapenemases

ST410 (n = 45) from this study was positive for OXA-181 (47%), NDM-5 (27%), OXA-181 + NDM-5 (9%), KPC-2 (7%), OXA-48 (4%), VIM-23 (4%) and NDM-7 (2%) (Table 1).⁹ ST410 belonged to two subclades,¹⁵ namely B3/H24Rx (n = 10) and B4/H24RxC (n = 35). The distribution of carbapenemases was linked to different subclades: e.g. KPC-2, OXA-48, VIM-23 were limited to B3/H24Rx; NDM-5, NDM-7 were limited to B4/H24RxC, and OXA-181 was identified in both clades (Table 1).⁹

Table 1.

Characteristics of dominant STs among E. coli with carbapenemases

	B3/H24Rx N = 10	B4/H24RxC N = 35	All ST410 N = 45
Geographical location	Global	Global	Global
QRDR mutations, n (%)
gyrA S83L	10 (100)	35 (100)	45 (100)
gyrA D87N	10 (100)	35 (100)	45 (100)
parC S80I	10 (100)	35 (100)	45 (100)
parE S458L	10 (100)	35 (100)	45 (100)
Carbapenemases, n (%)
KPC-2	3 (30)	0	3 (7)
NDM-5	0	12 (34)	12 (27)
NDM-7	0	1 (3)	1 (2)
OXA-48	2 (20)	0	2 (4)
OXA-181	3 (30)	18 (51)	21 (47)
VIM-23	2 (20)	0	2 (4)
NDM-5 + OXA-181	0	4 (11)	4 (9)
Other β-lactamases, n (%)
OXA-1	7 (70)	14 (40)	21 (47)
CMY-2	2 (20)	35 (100)	37 (82)
CMY-42	3 (30)	0	3 (7)
CTX-M-3	1 (10)	0	1 (2)
CTX-M-15	7 (70)	30 (86)	37 (82)
CTX-M-14	0	1 (3)	1 (2)
TEM-1	3 (30)	32 (91)	35 (78)
TEM-143	0	1 (3)	1 (2)
Aminoglycoside-modifying enzymes, n (%)
aadA2	1 (10)	14 (40)	15 (33)
aadA5	3 (10)	24 (67)	27 (60)
aac(3′)-IIa	2 (20)	0	2 (4)
acc(3′)-IId	2 (20)	29 (83)	31 (69)
aac(6′)-Ib-cr	7 (70)	34 (97)	41 (91)
aph(3′)-Ib	0	15 (43)	15 (33)
aph(6′)-1d	1 (10)	33 (94)	34 (76)
Other AMR determinants, n (%)
fosA	0	0	0
qnrS1	3 (30)	23 (66)	26 (58)
dfrA12	1 (10)	15 (43)	16 (36)
dfrA17	4 (40)	31 (89)	35 (78)
sul1	6 (60)	35 (100)	41 (91)
sul2	1 (10)	33 (94)	34 (76)
tet(A)	9 (90)	0	9 (20)
tet(B)	0	35 (100)	35 (78)

Global distribution of the ST410-B3/H24Rx/ST410-B4/H24RxC subclades

Different carbapenemases (i.e. KPC-2, OXA-48, OXA-181, VIM-23) from this study were evenly distributed among the B3/H24Rx subclade (Table 1) and showed a global distribution [Georgia (n = 1) with OXA-48, Kuwait (n = 1) with OXA-181, Mexico (n = 2) with VIM-23, Morocco (n = 1) with OXA-48, South Africa (n = 2) with OXA-181, USA (n = 2) with KPC-2 and Vietnam (n = 1) with KPC-2].

The B4/H24RxC subclade in this study was mainly found in Jordan (n = 15 with OXA-181), Egypt (n = 11 with NDM-5 and OXA-181) and Thailand (n = 6 with NDM-5 and OXA-181). The B4/H24RxC subclade was also detected in the Philippines (n = 1 with NDM-7), South Korea (n = 1 with NDM-5 and OXA-181) and Vietnam (n = 1 with NDM-5).

Antimicrobial resistance determinants were linked with ST410 clades

The presence of other β-lactamases in this study is shown in Table 1. CTX-M-15 was common among both subclades (70% in B3/H24Rx and 86% in B4/H24RxC); CMY-2 (100%) and TEM-1 (91%) were frequent in the B4/H24RxC clade, while OXA-1 (70%) was frequent in B3/H24Rx (Table 1).

Certain antimicrobial resistance determinants were common among both subclades (i.e. aac(6′)-Ib-cr and sul1) (Table 1). Other antimicrobial resistance determinants were more frequent among the ST410-B4/H24RxC subclade [i.e. aadA5, acc(3′)-IId, aph(6′)-1d, qnrS1, dfrA17, sul2 and tet(B)]. The tet(A) gene was more common among ST410-B3/H24Rx isolates (Table 1).

ST410 clades contained identical gyrA, parC, parE mutations

The combination of mutations in gyrase A (i.e. gyrA S83L, gyrA D87N), DNA topoisomerase IV genes (parC S80I) and parE S458L in the QRDRs were present in all the E. coli ST410 isolates from this study (Table 1).

Mobile genetic elements associated with bla_KPC-2, bla_VIM-23, bla_OXA-48, bla_OXA-181 and bla_NDM-7 were similar within the same carbapenemase type

The KPC-2 genes (n = 3) from this study were situated within Tn4401b. Those obtained from the USA (n = 2) were situated on identical 81 kb IncY plasmids and the Vietnamese KPC-2 was carried on a 63 kb IncN plasmid. BLAST analysis failed to identify similar IncY plasmids containing bla_KPC-2 in GenBank. The IncN plasmid showed ∼90% query coverage and >99.9 nucleotide identities to another IncN plasmid identified in 2010 from Vietnam (accession no. CP018945). The VIM-23 genes (n = 2 from Mexico) from this study were situated within a class I integron (i.e. In1374) and were carried on nearly identical 44 kb IncN plasmids. The Mexican strains differed by two SNPs. The IncN plasmid backbones showed highest identities (>99.9%) to plasmid pTRE-131, recovered from an environmental sample (accession no. KX863569).

Both of the OXA-48 genes from this study were situated on highly similar (99.9%–100% similarities) to global 64 kb IncL plasmids.^38,39 The bla_OXA-48 from Georgia was situated in Tn1999.3 and from Morocco in Tn1999.2. The bla_OXA-181 from Jordan (n = 15), Egypt (n = 4), Kuwait (n = 1), Thailand (n = 2), South Africa (n = 2) and South Korea (n = 1) were situated in Tn2013 harboured on nearly identical 51 kb IncX3 plasmids with 99.9%–100% similarities to the previously published plasmid p72_X3_OXA181.³⁰ These plasmids also contained qnrS1 and truncated ColKp3 replicons.³⁰

The bla_NDM-7 (n = 1) from this study (the Philippines) was carried on a 45 kb IncX3 plasmid⁴⁰ situated within Tn3 element with IS26 downstream, i.e. Tn3-IS5-bla_NDM-7-ble_MBL-trpF-desD-IS26. Details on bla_NDM-5 (n = 16) are provided below.

bla_CMY-2 was integrated within the chromosomes of B4/H24RxC

The bla_CMY-2 from B3/H24Rx isolates (this study) obtained from the USA (n = 2) was located on identical 109 795 bp IncC plasmids.⁴¹ However, the bla_CMY-2 from all the B4/H24RxC isolates was integrated within an ∼73 kb tRNA-Ser site chromosomal genomic island as previously described.⁴² The bla_CMY-2-harbouring IncC plasmids from B3/H24Rx and the B4/H24RxC chromosomal regions shared an ∼3.1 kb element consisting of ISEcp1-bla_CMY-2.

The bla_CMY-42 genes (n = 3) from B3/H24Rx (this study) obtained from South Africa (n = 2) and Kuwait (n = 1) were inserted downstream of ISEcp1 and located on ∼38 kb IncI(Gamma) plasmids identical to pCMY42_020032 obtained from China (accession no. CP034963).

The IncFII plasmids with bla_CTX-M-15 were ST410 clade specific

The majority of the B3/H24Rx subclade (7/10, this study) were positive for CTX-M-15. We were able to fully assemble six bla_CTX-M-15-containing plasmids obtained from Vietnam, USA, Mexico, Kuwait and South Africa. The bla_CTX-M-15 was situated on IncFII-A-B plasmids ranging from 93 to 133 kb (Figure 1). These plasmids showed mosaic structures and contained four different IncF replicons, namely two IncFII (i.e. FII36 and FII31), IncFIA and IncFIB and translated to F36:31:A4:B1 with the plasmid MLST scheme. Additionally, three of these plasmids (i.e. p3090-CTX, p1115-CTX and p1114-CTX) also harboured a Col156-like replicon gene (Figure 1). The CTX-M-15 gene was flanked by IS26 and ISEcp-1 upstream and Tn3, cat, bla_OXA-1, aac(6′)-Ib-cr and IS26 downstream, i.e. IS26-ISEcp-1-bla_CTX-M-15-Tn3-cat-bla_OXA-1-aac(6′)-Ib-cr-IS26 (Figure 1). The following AMR genes were also situated within all the plasmids: aadA2, aac(3′)-II, aadA5, aph(6′)-I, dfrA12, dfrA17, tet(A) and sul1.

Figure 1.

F36:31:A4:B1 and F1:A1:B49 plasmids with bla_CTX-M-15 and bla_NDM-5. F36:31:A4:B1 plasmids p3090-CTX, p1115-CTX, p1352-CTX contained bla_CTX-M-15 and were present in ST410-B2 subclades. F1:A1:B49 plasmids p2463-CTX and p2771-CTX contained bla_CTX-M-15 and were present in ST410-B3 subclades. F1:A1:B49 plasmid p2955-CTXNDM contained bla_CTX-M-15 and bla_NDM-5 and was present in ST410-B2 subclades. F1:A1:B49 plasmid p2768-NDM contained bla_NDM-5 and was present in ST410-B2 subclades. Highlighted regions are either identical and inverted. This figure appears in colour in the online version of JAC and in black and white in the print version of JAC.

The F36:31:A4:B1 plasmids contained several toxin/antitoxin systems (i.e. type II system pemI-pemK, ccdA-ccdB, vapC-vapB and type I hok-sok system) and putative virulence genes (i.e. aerobactin iuc operon, the arginine deaminase arc cluster, the iron transport sitABCD system and a CPBP family intramembrane metalloprotease gene). Complete gene transfer modules were absent. A previous study showed such plasmids could have been transferred through plasmid fusion with a self-conjugative helper plasmid.⁴³ The F36:31:A4:B1 plasmid backbones (with >90% blast query coverage and >99% nucleotide identities) were similar to CTX-M-15 plasmids initially described in E. coli ST167 isolates.^43,44

Most of the B4/H24RxC clade (30/35, this study) contained bla_CTX-M-15. We were able to fully assemble 20 of the CTX-M-15-containing plasmids. Among 19/20 isolates (obtained from Jordan, Egypt, Thailand, Vietnam, the Philippines and South Korea), the bla_CTX-M-15 were situated within IncFII-A-B-IncQ plasmids ranging from 72 to 113 kb (Figure 1). This translated to F1:A1:B49 with the plasmid MLST scheme. The remaining isolate (2771 from Egypt) contained an 88 kb IncFII-A-B (F1:A1:B49) plasmid that lacked the IncQ replicon (Figure 1). The following AMR genes were situated within these plasmids: bla_CTX-M-15, bla_NDM-5, bla_OXA-1, bla_TEM-1 (two copies), aac(6′)-Ib-cr, aadA2, aadA5, aph(3′)-Ib, aph(6′)-1d, acc(3′)-IId, strA, strB, mph(A), catB4, dfrA12, dfrA17, sul1 (two copies), sul2 and tet(B). Plasmids p2463-CTX (with the IncQ replicon) and p2771-CTX (without the IncQ replicon) did not contain bla_NDM-5 (Figure 1).

The overall structures of the F1:A1:B49 plasmids were similar (with >90% blast query coverage and >99% nucleotide identities) to other E. coli IncFII plasmids from Denmark,¹³ South Korea,⁴⁵ China (Genbank submission CP034956), Switzerland⁴⁶ and Egypt.⁴⁷ Some of these published IncFII plasmids were also obtained from ST410 isolates and included bla_NDM-5 and the truncated IncQ replicon.^13,45 The F1:A1:B49 plasmids contained three toxin/antitoxin pemI-pemK, ccdA-ccdB, vapC-vapB and ficT-ficA systems, and a type I restriction–modification system, namely hsdRMS. Complete gene transfer modules were also absent (similar to F36:31:A4:B1 plasmids described before).

The F36:31:A4:B1 plasmids from B3/H24Rx isolates and F1:A1:B49 plasmids from B4/H24RxC isolates (this study) contained different IncF plasmid replicon alleles, and had <45% blast query coverages, indicating that they are different IncF plasmids (Figure 1). Shared regions between the two plasmid groups mainly included the Tn3-bla_TEM-1-IS26-ISEcp-1-bla_CTX-M-15-Tn3-cat-bla_OXA-1-aac(6′)-Ib-cr-IS26 resistance module (Figure 1).

bla_NDM-5 was situated within IncFII-A-B-Q (F1:A1:B49) plasmids

Just under half (16/35) of the B4/H24RxC clade (this study) isolates contained bla_NDM-5. We were able to fully assemble 12 of the bla_NDM-5-containing plasmids obtained from Egypt, the Philippines, Vietnam and South Korea. The NDM-5 genes were situated on IncFII-A-B-Q (i.e. F1:A1:B49) plasmids ranging from 81 to 111 kb (Figure 1). These plasmids were similar (with >90% blast query coverage and >99% nucleotide identities) to other IncFII plasmids published before.^13,45 The bla_NDM-5 flanking region consisted of IS26-bla_NDM-5-ble_MBL-trpF-desD-ISCR1 (Figure 1).

A single NDM-5-containing F1:A1:B49 plasmid [p2768-CTX (77 kb)] from Egypt lacked the IncQ replicon and did not harbour bla_CTX-M-15 (Figure 1). The 21 kb MDR region consisting of IS6100-IS26-aac(3′)-IId-IS4-IS26-bla_TEM-1-IS26- ISEcp-1-bla_CTX-M-15-Tn3-IS26-cat-bla_OXA-1- aac(6′)-Ib-cr-IS26 was likely deleted by IS26-mediated excision.

Genomic acquisitions were pivotal for the evolution of E. coli ST410

We downloaded all global ST410 sequences from the NCBI RefSeq database (n = 442) and from a previous publication (n = 127).¹⁵ Core SNP phylogenetic analysis of the 569 publicly available genomes and the 45 genomes from our current study divided ST410 into two clades (Figure 2) [i.e. ST410-A (n = 12) and ST410-B (n = 602)]. ST410-B consisted of three subclades [i.e. ST410-B1 (n = 26), ST410-B2 (n = 283) and ST410-B3 (n = 293)]. With our SNP analysis, the previously published Roer subclades B2/H24R and B3/H24Rx¹⁵ belonged to a single subclade, namely ST410-B2 (Figure 2). The global phylogenetic analysis showed that the correlations between Roer subclade B3/B4 and bla_CTX-M-15 or bla_OXA-181 genes were less apparent (Figure 2). We propose to update the ST410 scheme based on the global population structures. Our revised clades corresponded to the following Roer clades:¹⁵ ST410-A was identical to A/H53, ST410-B1 was identical to B1/H24, ST410-B2 included B2/H24R, and B3/H24Rx and ST410-B3 corresponded to B4/H24RxC (Figure 2). Clade ST410-A and ST410-B clades differ with an average of 405 core SNPs (373 with B1, 402 with B2 and 411 with B3). Clade B1 strains have an average of 165 and 176 core SNPs differences in comparison to Clade B2 and B3 strains, respectively, whereas Clade B2 and B3 strains differ with an average of 57 core SNPs.

Figure 2.

Phylogenetic analysis of 172 E. coli ST410 from our current (n = 45) and previously published data (n = 127).¹⁵ Multi-allelic homologous recombination was responsible for QRDR mutations (gyrA S83L, gyrA D87N, parC S80I and parE S458L). The green shading illustrates the ∼180 kb recombination regions between ST410 clade B1 and B2/B3 strains, and the donors were likely from E. coli ST940 or ST694 strains. The purple shading denotes the recombination regions between clade B2 and B3 strains, and the donors were likely from E. coli ST10 complex strains (e.g. ST617). This figure appears in colour in the online version of JAC and in black and white in the print version of JAC.

We used Bayesian inference (BactDating) to deduce the temporal phylogenetic signal and molecular evolution of ST410 over time by using genomes (n = 172) from our current and previously published data (n = 127).¹⁵ The estimated mutation rate is 3.1 (95% CI, 2.4–3.9) (core SNP/genome/year). BactDating estimated that the most recent common ancestor (MRCA) of ST410 isolates, belonging to clade ST410-A, appeared around 1855 (95% CI, 1808–1893), approximately 170 years ago. ST410-B1 likely emerged in 1935 (95% CI, 1920–1950), which was associated with a type I pili switch from fimH53 to fimH24. The fluoroquinolone-resistant ST410-B2 subclade emerged in 1990 (95% CI, 1983–1996), which correlated with the clinical introduction of fluoroquinolones (i.e. norfloxacin and ciprofloxacin) in the late 1980s. The divergence of ST410-B2 from ST410-B1 correlated with the acquisition of bla_CTX-M-15 situated within F36:31:A4:B1 plasmids (Figure 2). The ST410-B3 subclade evolved from ST410-B2 in 2006 (95% CI, 2004–2009), which was accompanied by the chromosomal integration of bla_CMY-2 and the replacement of CTX-M-15-containing F36:31:A4:B1 plasmids with F1:A1:B49 plasmids harbouring bla_CTX-M-15 (Figure 2). The NDM-5 gene was incorporated into the F1:A1:B49 plasmids likely via IS26-mediated insertion in ∼2010 (95% CI, 2009–2013). IncX3 plasmids with bla_OXA-181 were acquired by both ST410-B2 and ST410-B3 subclades, but interestingly only became dominant over time within ST410-B3 (Figure 2).

The E. coli ftsI gene encodes PBP3.⁴⁸ Certain PBP3 amino acid insertions (i.e. YRIN, YRIK or TIPY) confer reduced susceptibilities to several β-lactams, including aztreonam, ceftazidime, cefepime, ceftazidime/avibactam and ceftolozane/tazobactam.⁴⁹ Our global genomic analysis showed that 292/293 of ST410-B3 isolates contained the ftsI YRIN N337N insertion while 16/283 of ST410-B2 isolates contained the ftsI KYRI I336I insertion (examples in Figure 2). Our Gubbins recombination analysis showed that ST410-B3 likely acquired a ∼130 kb recombination region that included fstI and LPS synthesis genes (e.g. lpxC) from E. coli ST167 (belonging to ST10 complex) as described previously.¹⁴ In addition, ST410-B3 contained an additional ∼60 kb recombination region, encompassing the DD-carboxypeptidase gene dacB (pbp6B), likely from the same ST10 complex background of strains (Figure 2).

ST410 QRDR mutations were likely acquired via recombination

The majority of ST410-B2 and ST410-B3 isolates (i.e. 99.65% of global collection) contained identical QRDR mutations, namely gyrA S83L, gyrA D87N, parC S80I and parE S458L. These QRDR mutations were absent in ST410-A and ST410-B1 (Figure 2). Gubbins recombination analysis identified eight discontinuous recombination regions (∼180 kb) between ST410-B2/ST410-B3 and ST410-A/ST410-B1 clades, and interestingly, these recombination regions included the parE, parC and gyrA genes. Our analysis showed that the acquisition of fluoroquinolone resistance in ST410-B2 and B3 subclades was likely due to a single multi-allelic homologous recombination event (Figure 2). This was previously described in a different MDR E. coli clone, namely ST1193.⁵⁰ Additional sequence comparison analysis showed that the possible donors of the ∼180 kb recombination regions were likely E.coli ST940 (e.g. GCF_002257685.1) and E. coli ST694 (GCF_001911225.1) (Figure 2, Figure S2).

Discussion

Our phylogenetic analysis of publicly available ST410 genomes amended the previously published Roer B/H24 subclades.¹⁵ We renamed the subclades as follows: ST410-B1 is identical to B1/H24, ST410-B2 includes both B2/H24R and B3/H24Rx, while ST410-B3 corresponds to B4/H24RxC. Our Bayesian inference analysis confirmed the ST410 molecular evolutionary timelines and specific clade genomic acquisitions as previously published by Roer et al.¹⁵ and Patino-Navarrete et al.¹⁴ These included fimH transition from fimH53 to fimH24 (that defined ST410-B1), QRDR mutations (i.e. gyrA S83L, gyrA D87N, parC S80I, parE S458L) that defined ST410-B2, and the acquisition of ftsI YRIN insertions (that defined ST410-B3). These genomic events were also accompanied by gaining IncF plasmids containing bla_CTX-M-15 (by ST410-B2 and B3), bla_NDM-5 (by ST410-B3) and IncX3 plasmids containing bla_OXA-181 (mainly by ST410-B3 but also B2 to a lesser extent). This stepwise acquisition by ST410 subclades of initial fluoroquinolone resistance (in the 1980s), then CTX-M genes (in the 1990s), followed by carbapenemase genes (in the 2000s), is shared with other high-risk clones such as E. coli ST131⁵¹ and Klebsiella pneumoniae ST307 and ST147.⁵²

Long-read WGS of the ST410 isolates from this study identified the following additional genomic events that likely shaped the evolution of ST410 clades. Firstly, ST410-B2/B3 QRDR mutations were possibly acquired via homologous recombination events. The donors were likely E. coli ST940 and ST694. A similar QRDR homologous recombination event occurred in E. coli ST1193.^50,53 Secondly, the emergence of ST410-B3 from ST410-B2 was accompanied by the chromosomal integration of bla_CMY-2. The chromosomal integration of AMR genes offsets the fitness costs associated with plasmid carriage.⁵⁴ Thirdly, the emergence of ST410-B3 from ST410-B2 was accompanied by the replacement of IncFII plasmids harbouring bla_CTX-M-15 (i.e. F36:31:A4:B1 plasmids were replaced with F1:A1:B49 plasmids). A similar ‘plasmid-replacement’ scenario occurred during the evolution of ST131 when F2:A1:B− plasmids were replaced with F1:A2:B20 plasmids in the ST131-C1 subclade.⁵⁵ Fourthly, the NDM-5 gene was integrated within F1:A1:B49 plasmids over time. This did not happen with F36:31:A4:B1 plasmids. A similar situation occurred during the evolution of ST131 when F2:A1:B− plasmids in ST131-C2 acquired bla_CTX-M-15.⁵⁵ An updated timeline of the pivotal genomic events that shaped the success of the ST410-B3 subclade is illustrated in Figure 3.

Figure 3.

The acquisition of genomic elements in E. coli ST410 and updated evolution timeline. This figure appears in colour in the online version of JAC and in black and white in the print version of JAC.

Surveys of ST410 clades are currently rare and show that certain clades are scarce (i.e. ST410-A, ST410-B1) while the ST410-B3 subclade is common in China⁵⁶ and Denmark.¹⁵ Results from our global survey that include E. coli ST410 with different carbapenemases (2015–17) obtained from 11 countries, showed a dominance of the ST410-B3 subclade, consisting of nearly 80% of the ST410 population. We did not find ST410-A or -B1 among this collection. The ST410-B2 and -B3 subclades were obtained from different geographical regions and were linked with distinct AMR resistance determinants. ST410-B2 isolates with different carbapenemases (KPC-2, VIM-23, OXA-48, OXA-181) were obtained from various countries while ST410-B3 isolates with OXA-181 and NDM-5 were mainly acquired from Jordan, Egypt and Thailand. The results highlight the successful local expansion of specific ST410 subclades in certain countries, especially ST410-B3 in Jordan and Egypt.

OXA-181 and NDM-5 were the most common carbapenemases in our study. Long-read WGS showed that bla_OXA-181 and bla_NDM-5 used different underlying strategies to ensure their successful association with ST410. The OXA-181 genes were located on near identical broad-host-range IncX3 plasmids (irrespective of the ST410 subclade or geographical location). Identical IncX3 plasmids with bla_OXA-181 were also detected among other global E. coli clones (ST167, ST131, ST648 etc.) obtained from 10 countries during the same survey.⁹ Highly similar IncX3 plasmids with OXA-181 were previously described from different Enterobacterales species worldwide.³⁹ The global success of OXA-181 genes is a combination of carriage on promiscuous IncX3 plasmids and the linkage with a successful clone such as ST410.

NDM-5 genes from this study were located on mosaic narrow-host-range IncFII plasmids that contained various AMR genes including bla_CTX-M-15. These plasmids were limited to ST410-B3 and contained several toxin/antitoxin systems and truncated gene transfer modules. The global success of NDM-5 genes is mainly due to their incorporation into narrow-host-range IncFII plasmids situated in successful high-risk clones such as ST410-B3 and ST167.⁴⁶ The control of IncX3 and IncFII plasmids producing carbapenemases should be public health priorities.

This study has several strengths. It included a global collection from 11 countries of recent ST410 isolates representing various lower- and middle-income countries (LMICs). LMICs bear a considerable AMR burden but lack adequate genomic surveillance systems.⁵⁷ We characterized isolates using long- and short-read WGS and provided novel information regarding the geographical distribution (especially from LMICs) and genomics of ST410 clades/subclades. Long-read WGS also identified additional genomic events that shaped the success of the ST410-B3 subclade.

Limitations of this study included that some IncFII plasmids harbouring CTX-M-15 and NDM-5 genes were not fully assembled even after several attempts. Publicly available genomes are often not closed genomes and the transferability of the results from our long-read-based study to short-genome-based results is not complete. Furthermore, some countries included few E. coli isolates for the survey and, therefore, may not be fully representative of which carbapenemase-producing ST410 clades dominate in those regions.

In summary, our geographical and genomic data suggested that distinct evolutionary events have shaped the population structure of ST410. The global carbapenemase-producing E. coli ST410 population is dominated by the ST41-B2 and -B3 subclades with different characteristics and varied geographical distributions. This requires ongoing genomic surveillance using methodologies that characterize high-risk clones and their respective clades. This study identified additional genomic events that shaped the success of the ST410-B3 subclade. We also described different underlying molecular epidemiology of OXA-181 and NDM-5 genes.

Funding

This work was supported by research grants from the JPIAMR/Canadian Institute Health Research program (#10016015) and National Institute of Health (#10028552). The study was in part supported by NIAID grant R01AI090155.

Transparency declarations

All authors have none to declare. The funding agencies have not played any decision-making role in the research.

Supplementary data

Figures S1 and S2 and Tables S1 and S2 are available as Supplementary data at JAC Online.