Introduction

Escherichia coli, a member of the Enterobacteriaceae, is a major human pathogen causing various infections ranging from intestinal disease and urinary tract infections to invasive bloodstream infections. Carbapenems such as ertapenem, imipenem, and meropenem are potent antimicrobial agents against the Enterobacteriaceae and have become the mainstream agents of choice to treat severe infections caused by E. coli. This is due to the near-ubiquitous carriage of extended spectrum β-lactamases (ESBL) in E. coli causing urine and bloodstream infections1. However, carbapenem-resistant E. coli (CREC) has emerged worldwide, representing a serious challenge for clinical management and public health2. Carbapenem resistance in E. coli is largely due to the production of carbapenem-hydrolyzing enzymes (carbapenemases)3. There are a variety of carbapenemases, and the most common ones observed in clinical bacterial strains include KPC (Klebsiella pneumoniae carbapenemase), NDM (New Delhi metallo-β-lactamase), OXA-48 (oxacillinase-48), IMP (imipenemase), and VIM (Verona integron-encoded metallo-β-lactamase)3,4. NDM appears to be particularly common in CREC5.

In contrast to the well-studied carbapenem-resistant Klebsiella pneumoniae (CRKP), the clonal background of CREC is less well characterized including the transmission of CREC within and between hospitals. Studies of global E. coli isolate collections have shown that carbapenemase gene carriage is focused in strains belonging to lineages within the phylogroups A and B1 E. coli6, classically considered to be non-pathogenic commensals7. Local genomic epidemiological studies, such as of CREC in China, have also led to the discovery of globally disseminated clones ST167 and ST617, both of which belong to phylogenetic group A8. A similar study in Scandinavia also resulted in the discovery of a globally disseminated CREC lineage, ST410 (ref. 9). This lineage was similar to the global pandemic ESBL E. coli lineage ST131, in that a specific clone (B4/H24RxC) had arisen from the background population via acquisition of a resistance plasmid, in this instance the ST410 lineage containing an IncX3 plasmid carrying the blaOXA-181 carbapenemase gene9.

The evolutionary steps leading to the emergence of the E. coli ST131 lineage have been extensively reported10,11,12,13. However, the lack of global concerted genomic analyses of CREC means that our understanding of how potentially dominant CREC clones are evolving and emerging is lacking. Analysis of the ST167 and ST617 lineages showed some clear overlaps in evolutionary trajectory between these CREC clones and ST131 including mutations involved in host colonization and in intergenic regions associated with emergence of multi-drug resistant (MDR) plasmid-bearing clones8, but there remains a need to determine if this pattern is common across emerging CREC clones. Here we utilize a province wide analysis of clinical CREC strains performed at West China Hospital to address this question. Between June 2016 and February 2017 all CREC collected from eight Sichuan hospitals were genome sequenced. The majority (60%) of strains belonged to ST410, ST167, and ST617. Analysis of the ST410 genomes and comparison against all publicly available ST410 genome sequences confirmed the presence of an MDR B4/H24RxC clone within ST410 globally disseminating either blaNDM-5 or blaOXA-181. Long-read sequencing revealed these carbapenemases are freely interchanging on an identical IncX3 plasmid. Genetic loci which discriminate the MDR clone from the rest of the ST410 lineage included anaerobic metabolism loci and intergenic regions, as shown for other MDR clones of E. coli, and unique sequence variants of the fhu iron acquisition operon, which confer an increased ability to scavenge iron. Together our data show a footprint for evolution of a CREC clone, whereby recombination drives new alleles into the clone which provide a competitive advantage in colonizing mammalian hosts. The importance of enhanced colonization capabilities in the evolution of MDR clones must be fully characterized and presents a possible new avenue for combatting CREC.

Results

ST410 is the most common circulating CREC lineage in Sichuan strains

A total of 25 CREC strains were collected from eight hospitals (Supplementary Table 1) in Sichuan province, China, between June 2016 and February 2017. The strains were recovered from blood, sputum, urine, wound secretion, bile, pleural fluid, and ascites, suggesting that CREC is associated with various types of infections such as bloodstream infection, pneumonia, and urinary tract infection (Table 1). All CREC strains were resistant to imipenem (minimum inhibitory concentrations [MIC], 8 to >256 mg/l), meropenem (MIC, 32 to >256 mg/l), piperacillin/tazobactam, ceftazidime, and ceftazidime/avibactam but were susceptible to tigecycline (Table 2 and Supplementary Dataset 1). Most strains were resistant to ciprofloxacin (resistance rate, 96%), trimethoprim/sulfamethoxazole (88%), aztreonam (76%), and gentamicin (68%), while most were susceptible to aztreonam/avibactam (susceptible rate, 92%), colistin (88%), and amikacin (72%) (Table 2). All of the 25 CREC were subjected to short read whole-genome sequencing and antimicrobial resistance genes were identified based on their draft genome sequences. blaNDM was the only carbapenemase-encoding gene identified and was found in all 25 CREC strains (Supplementary Dataset 1). Four blaNDM variants were identified including blaNDM-5 (the most common type, present in 21 strains), blaNDM-1 (in two strains), blaNDM-7, and blaNDM-21 (each in one strain) (Table 1). Three colistin-resistant CREC carried the plasmid-borne colistin-resistance gene mcr-1, one of which also had another colistin-resistance gene mcr-3 (Table 1). Seven amikacin-resistant strains, all of which exhibited high-level resistance to amikacin (MIC, >256 mg/l), had the 16S rRNA methylase gene rmtB, and one strain had another 16S rRNA methylase gene armA in addition to rmtB. The CREC strains belonged to 13 sequence types (STs), highlighting a heterogeneous clonal background. Three STs, ST167 (n = 4), ST617 (n = 5), and ST410 (n = 6), accounted for the majority (60%) of CREC, while there was only a single strain for the remaining 10 STs. The number of SNPs among ST167 and ST617 strains are shown in Supplementary Tables 2 and 3. As we have previously characterized ST167 and ST617 CREC8. we therefore focused on ST410, the common type, in this study.

Table 1 CREC strains in this study
Full size table
Table 2 In vitro susceptibility of the 25 CREC isolates
Full size table

Most Sichuan ST410 CREC belong to the globally spread B4/H24RxC clone

Strain 020001, the first ST410 strain isolated in our study, was also subjected to long-read whole-genome sequencing using MinION (the sequencing yield is listed in Supplementary Table 4) to obtain its complete genome sequence. A hybrid assembly of the genome sequence of 020001 revealed that the strain had a 4.9-Mb chromosome and six plasmids (Supplementary Table 5). The chromosome sequence of strain 020001 was then used as a reference for mapping. Two (strains 020026 and 020031) of the six strains were separated from each other by 17 core single-nucleotide polymorphisms (SNPs; Table 3), indicating a potential clonal spread. Given that these two strains were recovered from different hospitals, such an observation suggests recent inter-hospital movement of a common strain. Another two strains (strains 020129 and 020147) were 51 to 90 SNPs distant from the above two strains, suggesting relatively recent shared ancestry for the four strains (Table 3). The remaining two strains had >2500 SNPs between each other and any of the aforementioned four strains (Table 3). This suggests that the two remaining strains had no recent linkage with the other four. We investigated the clonal relatedness between ST410 strains in this study and other ST410 strains with genome sequences available in GenBank (Supplementary Dataset 2). From this we clearly identified the B4/H24RxC clone of ST410, which contains 37 strains including four from the present study (strains 020026, 020031, 020129, and 020147) and strains from Asia (the Philippines and Thailand), Europe (Denmark, Italy, Norway, Turkey, and UK), and North America (Canada and USA) (Fig. 1, the numbers of SNPs are shown in Supplementary Dataset 3).

Table 3 Pairwise SNPs between ST410 strains of this study with strain 020001 as the reference
Full size table
Fig. 1
figure 1

Phylogenomic tree of ST410 E. coli strains. A circular phylogenomic tree of ST410 E. coli strains (n = 327) was inferred using strain 020001 as the reference. Information on the strains is available in Supplementary Dataset 1 and the numbers of SNPs are shown in Supplementary Dataset 3. The six strains isolated in this study are indicated in red. The B4/H24RxC clone is also highlighted by a gray region. The four colored circles represent the source, location, resistance genes, and year of isolation from inside to outside, respectively. Bootstrap values are represented by gradient colors and a scale bar for the ST410 phylogeny is shown

Full size image

Emergence of the B4/H24RxC clone in the last decade driven by recombination

Strain 020026, the first strain of the B4/H24RxC MDR clone in this study, was also subjected to long-read whole-genome sequencing to obtain its complete genome sequences. A more precise phylogeny (Fig. 2) was inferred by using the complete chromosome of strain 020026 as the reference genome and using Gubbins to identify and remove recombination regions. The results showed several recombination hotspots (Supplementary Fig. 1), including genes involved in toxin–antitoxin system, flagellum, metabolism, and phage life cycle, and notably, a 13-kb region that was unique to the B4/H24RxC MDR clone. A smaller scale phylogeny re-construction was performed solely on the B4/H24RxC MDR clone, closely-related strains and the sister clade and rooted on strain E006910 (accession no. ERR1197948) of the sister clade (Supplementary Fig. 2). Recent recombination events were identified in four strains of the B4/H24RxC MDR clone. The four strains, 115102 (accession no. SRR7716572), N14-01320 (accession no. SRR5714046), EuSCAPE_SC020 (accession no. ERR1374952), and ECS1_14 (accession no. ERR2088799) had recombination regions of 94,045, 5076, 4164, and 46,701 bp, respectively. By excluding SNPs within these recent recombination regions, the average pairwise SNP distance among the 37 B4/H24RxC MDR clone strains was 31, ranging from 0 to 105 (Supplementary Dataset 4). Given known dates and origins of isolation, such a low level of core SNPs strongly suggests recent emergence and likely on-going global dissemination of this clone. A total of 2602 SNPs were identified on the branch separating the B4/H24RxC MDR clone from its most closely-related strain (Supplementary Fig. 2), of which 2510 (96.46%) were identified to be within regions of recombination, giving a per site r/m ratio (the relative likelihood that a polymorphism was introduced through recombination rather than point mutation) of 27.28. This suggests that the primary evolutionary events underpinning the emergence of the clone were driven by homologous recombination.

Fig. 2
figure 2

A refined Phylogenomic tree of ST410 E. coli strains and the emerging lineage. a A circular phylogenomic tree of ST410 E. coli strains (n = 327) was inferred using strain 020026 as the reference. The numbers of SNPs are shown in Supplementary Dataset 4. Several strains that were closely related to the B4/H24RxC clone are highlighted by a yellow region, while a sister clade is highlighted by a green region. Bootstrap values are represented by gradient colors and a scale bar for the ST410 phylogeny is shown. b A fivefold enlarged phylogenomic tree of the B4/H24RxC clone (the pink region), several closely related strains (the yellow region) and a sister clade (the green region). Strain names, sources, locations, accession numbers, carbapenemase genes, and CTX-M ESBL genes are shown

Full size image

Coalescent analysis with all dated strains failed to converge within an applicable time (see Methods for details) during the run using the tool BactDating v1.0.1 (ref. 14) (Supplementary Fig. 3). Four distant ST410 strains (strains KOEGE 131, MOD1-EC5419, KTE221, and NC_STEC121, see Methods for detail) were therefore excluded for dating. The refined analysis revealed an average clock rate of μ = 3.86 [3.05–4.57] substitutions per year and a root date of December 1899 (95% confidence interval [95% CI], October 1850–April 1928; Fig. 3, Point C), indicating that ST410 emerged sometime around the turn of the twentieth century. The most recent common ancestor of the B4/H24RxC MDR clone was estimated to emerge in June 2009 (95% CI, March 2007–December 2010; Fig. 3, Point A), suggesting the clone has emerged in the past 10 years.

Fig. 3
figure 3

The dated phylogenomic tree of ST410 E. coli strains. The tree was constructed using BactDating v1.0.1 and corrected for recombination using Gubbins v2.3.4. Four strains within the B4/H24RxC MDR clone, i.e. strain KOEGE 131 (358a) (accession no. SRR785629), MOD1-EC5419 (accession no. SRR6512532), KTE221 (accession no. SRR633754,) and NC_STEC121 (accession no. SRR5470036), were excluded due to their distant relationship to the remaining clonal strains (>5000 SNPs with almost all other ST410 strains). Point A, the common ancestor of the B4/H24RxC MDR clone was estimated to emerge in June 2009 (95% CI, March 2007–December 2010). Point B, the separation of B4/H24RxC MDR clone and its closely related strains from the sister clone was estimated in September 1994 (95% CI, August 1989–December 1998) but without any other identified intermediate strain. Point C, the emergence of ST410 E. coli was estimated in December 1899 (95% CI, October 1850–April 1928)

Full size image

Our data are consistent with the initial characterization of the B4/H24RxC MDR clone9, including its recent emergence from the ST410 lineage. All but three strains had either a blaNDM or a blaOXA-181-like carbapenemase (blaOXA-181 and blaOXA-232, two OXA-48-type carbapenemase-encoding genes; OXA-232 differs from OXA-181 by a single amino acid) or both. Although blaNDM and blaOXA-181 have also been seen in strains of other lineages, half (17/34) of the blaNDM-carrying ST410 strains sequenced, and almost all (21/23) blaOXA-181-carrying ST410 strains belonged to the clone. Previous work by our group has demonstrated that both blaNDM and blaOXA-181 are carried by IncX3 plasmids with identical backbones and swapping of the corresponding locus generates a plasmid carrying either blaNDM or blaOXA-181 (ref. 8). In addition, most (26/37) strains within the clone also carried blaCTX-M-15. A previous study found interspecies transmission of ST410 E. coli carrying a blaCTX-M gene between wildlife, humans, companion animals, and the environment15. Unlike blaNDM and blaOXA-181, blaCTX-M-15 is not largely restricted to the clone but is dispersed across the wider ST410 population.

bla NDM and bla OXA-181 are carried by IncX3 plasmids with a common backbone

Self-transmissible plasmids carrying blaNDM were obtained from all six of the Sichuan province ST410 strains by conjugation and all of the plasmids had IncX3 replicons. Among the 37 strains of the B4/H24RxC clone, 17 carried blaNDM and 21 had blaOXA-181 including 5 that had both blaNDM and blaOXA-181. Ten of 17 blaNDM-carrying strains had an IncX3 replicon and all 10 strains had contigs with 100% coverage and 99.93–100% identity to the reference IncX3 plasmid, pNDM5_020001, the blaNDM-5-carrying plasmid of strain 020001 (Table 4). It is therefore likely that blaNDM was carried by a common pNDM5_020001-like IncX3 plasmid in the 10 strains. The complete sequence of the blaNDM-carrying plasmid, pNDM5_020026, in strain 020026 was obtained using hybrid assembly of MinION long and Illumina short reads (Supplementary Table 5) and was indeed identical to that of pNDM5_020001. All of the 21 blaOXA-181-carrying strains had an IncX3 replicon. pOXA181, a blaOXA-181-carrying plasmid, was recovered and fully sequenced from one of the 21 strains by our group as reported previously16, and all of the remaining 20 strains had contigs with 100% coverage with pOXA181, suggesting that blaOXA-181 was located on a common IncX3 plasmid in these strains (Table 4). pNDM5_020001 and pOXA181 have an identical IncX3 backbone with the exception of several SNPs. Therefore, it appears that a common IncX3 plasmid is frequently interchanging blaOXA-181 and blaNDM genes, with both successfully co-circulating in the population. It is also possible that blaOXA-181- and blaNDM-5-carrying IncX3 plasmids arose independently and the plasmids are acquired interchangeably in the clone, but it is impossible to determine this from the data available.

Table 4 Plasmids in strains of the lineage
Full size table

For the five strains carrying both blaNDM and blaOXA-181 with genome sequences available in GenBank, their contigs had 100% coverage with both pNDM5_020001 and pOXA181. Due to the fact that only short reads are available for these five strains, we were unable to determine whether both blaNDM and blaOXA-181 were located on a single IncX3 plasmid or on different plasmids by mapping. Other approaches, such as tracking unique paths in assembly graphs from different assemblers, were attempted. However, ambiguous paths were associated with contigs containing IS26, and the largest contigs aligning with the references were not the exact size of either plasmid. Nonetheless, by comparing and contrasting the predicted plasmid replicons of the strains within the same clade, it could be deduced that blaNDM-1 was carried on an IncA/C plasmid, while blaOXA-181 was carried on an IncX3 plasmid, in the five strains carrying both blaNDM-1 and blaOXA-181. Seven blaNDM-carrying strains within the clone did not have an IncX3 replicon, suggesting that blaNDM was carried by plasmids of other replicon types and strains of the clone have acquired blaNDM more than once.

IncX3 plasmids were stably maintained in nutrient-rich media

We performed plasmid stability tests for representative IncX3 plasmids carrying blaNDM-5 or blaOXA-181 from strains of the B4/H24RxC MDR clone. In LB media (representing nutrient-rich settings), the stability frequency of pNDM5_020026 and pOXA-181 was 0.99 ± 0.01 and 1.00 ± 0.01 (mean ± standard derivation [SD]; Fig. 4a and Supplementary Table 7), respectively. The plasmid loss rate of pNDM5_020026 and pOXA-181 in LB media was 1.22% ± 0.77% and 1.11% ± 0.51%, respectively. The stability frequency in M9 minimal media (representing nutrient-restricted settings) of pNDM5_020026 and pOXA-181 was 0.89 ± 0.01 and 0.92 ± 0.00 and the plasmid loss rate of the two plasmids was 27.33 ± 3.18% and 24.11 ± 1.71%, respectively. The above findings suggest that IncX3 plasmids carrying blaNDM-5 or blaOXA-181 were stably maintained at equal frequencies in nutrient-rich settings. Our data also suggest that these plasmids were prone to loss at an elevated frequency in nutrient-restricted settings.

Fig. 4
figure 4

Plasmid stability, biofilm formation, virulence assay, and strain relative fitness. a Stability of pNDM5_020026 and pOXA181 in E. coli J53. The mean ± standard deviation (SD) of the stability frequency is shown. N = 3 biologically independent experiments. The results are also shown in Supplementary Table 7. b Biofilm formation of bacterial strains. Absorption values of strains 020001, 020026, and 020032 at both OD590 nm and OD595 nm are shown. Strain ATCC 19606 and H2O were used as the positive and negative control, respectively. The results are also shown in Supplementary Table 8. N = 3 biologically independent experiments. c Survival of G. mellonella after infection by bacterial strains. The effect of 1 × 104, 1 × 105, 1 × 106, and 1 × 107 CFU of each strain on survival of G. mellonella at 72 h after infection is shown. The exact survival rates are also shown in Supplementary Table 8. KP767, a hypervirulent K. pneumoniae, was used as a positive control, while E. coli DH5α was used as a negative control. N = 5 biologically independent experiments. The results are also shown in Supplementary Table 9. d Relative fitness of strain 020026 compared to strain 020001 and strain 020032. The competition between strain 020026 and strain 020001 was also performed in the presence of 375 DIP to create iron-deprived conditions. N = 3 biologically independent experiments. The results are also shown in Supplementary Table 10

Full size image

B4/H24RxC-specific genes encoding adherence and iron acquisition

All strains of the B4/H24RxC clone had three genes that had no orthologous genes with >90% nucleotide identity present in all other ST410 strains. The first gene is yadC, which encodes a fimbriae-like protein YadC (NCBI Reference Sequence accession no. WP_000848455.1). yadC was absent from strain 020001, while strain 020032, which is a ST410 CREC identified in the present study but is more phylogenetically distinct from B4/H24RxC than strain 020001 (Fig. 2), had a gene with 76% coverage and 56.5% identity. YadC has been purported to be involved in adhesion, internalization, and motility of E. coli and contribute to its pathogenicity17. The second gene is ybjI, which encodes a pentapeptide repeat-containing protein (NCBI Reference Sequence accession no. WP_000868898.1) but could be a pseudogene (https://www.uniprot.org/uniprot/P32690). ybjI was absent from both strain 020001 and strain 020032. The third gene, fhuA, encodes a ferrichrome porin FhuA (NCBI Reference Sequence accession no. WP_039023099.1) and constitutes the fhu operon with fhuB, fhuC, and fhuD. The fhu operon is essential for the utilization of ferric siderophores of the hydroxamate type18 and also contributes to bacterial virulence19. There are multiple types of FhuA-like proteins in E. coli. The remaining 290 ST410 strains including strain 020001 and strain 020032 contained a gene encoding an alternative allele of FhuA (NCBI Reference Sequence accession no. WP_000124383.1; Supplementary Fig. 4), which had <74% nucleotide identity with the clone-specific fhuA. Blast analysis of the clone-specific fhu operon showed a 100% nucleotide identity match with the fhu operon of numerous other E. coli strains, suggesting that the allelic replacement of the fhu operon was derived from recent recombination in the B4/H24RxC MDR clone. The three gene alleles (yadC, ybjI, and fhuA) were present in all strains of the B4/H24RxC MDR clone while absent from all other ST410 strains, suggesting that the alleles were acquired at the emergence of the B4/H24RxC MDR clone and then swept to fixation in the clone.

B4/H24RxC-specific SNPs associated with adherence and iron acquisition

We further identified SNPs unique to the B4/H24RxC clone. A total of 382 SNPs were present in all members of the clone but absent from other ST410 genomes (Supplementary Dataset 5). Among the 382 SNPs, 362 were in coding sequences but only 60 were non-synonymous substitutions, present in 48 genes (Fig. 5 and Supplementary Dataset 5). Most of the 48 genes encode products involved in metabolism including three dehydrogenases involved in anaerobic metabolism. In addition, we found 20 clone-specific SNPs in intergenic regions, which have been shown to be under strong evolutionary constraints20. Nine clone-specific SNPs in intergenic regions were located in the -10, -35 boxes of promoter such as those found in the upstream of ferrichrome porin gene fhuA, recombination-promoting family gene rpn, and aspartate decarboxylase gene panD, or within the 5′ UTR of downstream genes such as glucose uptake transporter regulator sgrR and inhibitor sgrT (Supplementary Dataset 5). Previously, we have also found unique intergenic SNPs and unique gene alleles encoding anaerobic metabolism in ST167 and ST617 (ref. 8), as well as showing that these are key evolutionary events in the emergence of the globally disseminated ST131 clone C21. Several genes encoding clone-specific SNPs (elfG, fhuC, fhuD, lptD, and emrK) may promote the survival of bacteria. The elfG gene is part of the elfADCG-ycbUVF fimbrial operon-encoding proteins which promote adhesion of bacterial cells to abiotic surfaces22 and may therefore facilitate bacteria to colonize the wider hospital environment. fhuC and fhuD are part of the fhu operon19 as described above. lptD (also known as imp) encodes the lipopolysaccharide-assembly protein LptD, which is essential for integrity of the membrane and is related to the sensitivity of bacteria to detergents, antibiotics, and dyes23,24. emrK encodes part of a tripartite efflux system named EmrYK-TolC, which confers stress-inducible functions including those imposed by antimicrobial agents to reduce the lethal effects25,26. Further analysis of the genomic location of these clone defining SNPs identified clear clustering of the SNPs in a defined region of the chromosome. Confirmatory analysis of the recombination detection tests using Gubbins27 on the reference-mapping based-SNP alignment demonstrated that over 90% (n = 346) clone-specific SNPs, 65% (n = 39) clone-specific non-synonymous SNPs had been introduced via a recombination event or events, facilitating a more rapid adaptation and evolutionary emergence of the MDR clone (Supplementary Fig. 1).

Fig. 5
figure 5

Location of all synonymous and non-synonymous SNPs unique to the emerging MDR clone compared with other ST410 strains. The circular chromosome diagram was generated using CGView62. The detailed information of the SNPs is available in Supplementary Dataset 5

Full size image

Decreased ability of the B4/H24RxC clone to form biofilms

Given the unique yadC gene and unique SNPs in elfG we performed assays of biofilm formation to plastic abiotic surfaces for a representative strain (020026) of the clone with two non-clone ST410 strains, 020001 and 020032, as control. Strain 020026 exhibited significantly less biofilm formation (e.g. absorption at OD590 nm, mean ± standard deviations, 0.35 ± 0.05 vs 0.46 ± 0.11, P < 0.001; Fig. 4b and Supplementary Table 8) than strain 020001. Therefore, the unique gene and SNPs seen in the clone significantly decrease adherence to abiotic surfaces and the ability to form a biofilm. Adherence is a key factor for bacteria to colonize hosts including humans28. Although the clone has a decreased ability to form biofilms on abiotic surfaces, it still has the ability to form biofilms and does not lose adherence, meaning that it is still capable of colonizing hosts.

The B4/H24RxC clone exhibits enhanced ability to utilize iron sources

Given our identification of clone-specific SNPs in the fhu operon, we performed an iron source growth assay and found that the presence of 250 µM DIP completely inhibited growth of strain 020032. By contrast, the MIC of 2′2-dipyridyl (DIP) for strain 020001 and strain 020026 was 500 µM DIP, suggesting that the two strains were more resistant to iron-deprived conditions than strain 020032. The addition of FeCl2 restored growth of all three strains, while the addition of lactoferrin only restored growth of strain 020001 and strain 020026 but not that of strain 020032 (Table 5). Lactoferrin is present on mucous membranes and is part of the human innate defense29. The ability to utilize iron from lactoferrin may therefore facilitate bacteria to colonize the human gut. The enhanced resistance to iron-deprived conditions and the ability to sequester iron from lactoferrin allow strain 020001 and strain 020026 to adapt to the human host better than the more distant strain 020032. The addition of hemin and bovine serum albumin restored growth of strain 020026 but not strain 020001 nor strain 020032, while the addition of hemoglobin only restored growth of strain 020001 (Table 5). Hemin (ferric chloride heme) is an oxidized form of heme and is produced during processing aged red blood cells30. Serum albumin is the most abundant blood protein in humans. The utilization of hemin and serum albumin as the sole iron source may therefore facilitate the survival of B4/H24RxC strains in human hosts.

Table 5 Iron source growth assay results in the presence of DIP at the MIC (500 μM for strains 02001 and 020026, and 250 μM for strain 020032)
Full size table

The B4/H24RxC clone does not have enhanced virulence

The 50% lethal dose (LD50) at 72 h of strains 020026, 020001, and 020032 against G. mellonella were identical at 1 × 105 CFU, also identical to that of the hypervirulent K. pneumoniae strain KP767 (Fig. 4c and Supplementary Table 9). Therefore, the B4/H24RxC strain 020026 displays virulence comparable to other members of the ST410 lineage, but no obvious enhancement of the virulence phenotype. It is well known that the fhu iron acquisition operon contributes to bacterial virulence19,31. However, the clone-specific fhuA and the clone-unique SNPs in fhuC and fhuD of the emerging MDR clone do not lead to enhanced virulence.

The B4/H24RxC clone shows a fitness advantage in iron-deprived conditions

Strain 020026 exhibited a fitness advantage compared to strain 020032 (relative fitness value [w], 1.28 ± 0.06; mean ± standard deviation; Fig. 4d and Supplementary Table 10) but was slightly outcompeted by strain 020001 (w, 0.93 ± 0.02; Fig. 4 and Supplementary Table 10) in LB media. As strain 020026 was able to utilize more iron sources than 020001, we also performed competition experiments between strain 020026 and 020001 under iron-deprived conditions. Strain 020026 outcompeted strain 020001 (w, 1.07 ± 0.02; Fig. 4d and Supplementary Table 10) in iron-deprived media. The difference of the w values of strain 020026 compared to strain 020001 between iron-deprived and non-iron-deprived conditions was statistically significant (t = 17.33, P < 0.001). The fitness advantage in iron-deprived environments seen in strain 020026 is therefore likely to promote the survival and persistence of B4/H24RxC strains in human hosts.

Discussion

Our data presented here stemmed from a genomic epidemiology and surveillance study of CREC in Sichuan Province, China. The identification of ST410, ST167, and ST617 as dominant CREC clones in the province led us to comprehensively characterize the ST410 lineage (having previously characterized the ST167 and ST617 lineages)8. Phylogenomics revealed that the majority of Chinese CREC ST410 belonged to a previously identified, globally disseminated clone of CREC ST410 labeled B4/H24RxC9. By using MinION sequencing in combination with the available Illumina genome data, we were able to additionally show that the clone is dominated by an IncX3 plasmid which has expanded with the clone, but which frequently interchanges the carbapenemase genes blaNDM-5 and blaOXA-181 without any impact on plasmid stability or fitness. In an effort to identify key evolutionary events in the emergence of the B4/H24RxC clone, we uncovered a number of SNPs and core-gene alleles unique to the clone in comparison to the remainder of the ST410 lineage. Our findings show unique SNPs in core anaerobic metabolism genes and intergenic regions within the B4/H24RxC clone. These have been shown to be key events in the emergence of MDR ST167 and ST617 lineages, as well as the MDR ST131 clone C21. Therefore, our data add further compelling evidence to the notion that evolution of MDR in E. coli is parallel in nature and as such predictable. In this study we also show for the first time how SNPs and gene alleles associated with increased colonization of mammalian hosts are associated with fundamental changes in important phenotypes. Indeed, our data are indicative of a scenario where emerging clones of MDR E. coli accumulate SNPs enhancing key phenotypes such as iron acquisition, while abrogating phenotypes are more associated with environmental survival such as adhesion to abiotic surfaces. This concept of adaptation to the human host is further supported by the fact that in laboratory media competition experiments, the MDR B4/H24RxC clone of ST410 is slightly outcompeted by strains of other clones within ST410 but has advantage under iron-deprived conditions, suggesting that the fitness associated with SNPs affecting colonization potential are advantageous within the human clinical environment. These findings present a vitally important new direction in our understanding of the emergence and dynamics of clones of MDR E. coli.

Materials and methods

Strain isolation and in vitro susceptibility testing

All non-duplicate CREC clinical strains (n = 25) were collected from one referral and seven municipal hospitals in Sichuan Province, China, between June 2016 and February 2017 (information about the hospitals is available in Supplementary Table 1). This study was approved by the ethical committee of West China Hospital and informed consents were waived. All of the strains were initially identified as E. coli using Vitek II (bioMérieux; Marcy-l'Étoile, France). The strains were isolated from various clinical samples including blood (n = 8), sputum (n = 6), and urine (n = 5) (Table 1 and Supplementary Dataset 1). MICs of antimicrobial agents were determined using the microdilution method of the Clinical and Laboratory Standards Institute (CLSI)32. For ceftazidime/avibactam, colistin, and tigecycline, the breakpoints defined by the European Committee on Antimicrobial Susceptibility Testing (EUCAST) were used, while the breakpoints of aztreonam were applied for aztreonam–avibactam.

Whole-genome sequencing and analysis

All Chinese strains isolated in this study (n = 25) were subjected to whole-genome sequencing using the HiSeq X10 (Illumina; San Diego, CA, USA) according to the manufacturer’s instructions. Genomic DNA was prepared using the QIAamp DNA Mini Kit (Qiagen, Hilden, Germany). Generated reads were subjected to strict quality-control filtering including trimming 10 bases from each end and bases with quality below Q15, removing adaptor sequences and discarding reads with an average quality below Q20 using Cutadapt v1.16 (ref. 33) and BBTools v37.92. The reads were then de novo assembled into contigs using SPAdes v3.13.0 (ref. 34) applying the careful and auto-cutoff modes. Strain 020001 (the first ST410 strain recovered in this study), 020026 (the first strain of the B4/H24RxC MDR clone identified in this study, see below), and 020032 (a phylogenetically distant strain within ST410 isolated in this study) were also subjected to whole-genome sequencing using the long-read MinION Sequencer (Nanopore; Oxford, UK). Libraries were constructed using the SQK-LSK109 kit and were multiplexed using native barcodes from the EXP-NBD104 kit, according to protocols of the manufacturer (Nanopore). Sequencing was performed in an R9.4.1 Flow Cell for 48 h and the yield of MinION sequencing data of the three strains is shown in Supplementary Table 4. The MinION reads were base-called and demultiplexed using Guppy v3.0.3. Reads with adaptors at the ends were trimmed and those with adapters in the middle were discarded using Porechop v0.2.4. A de novo hybrid assembly of both short Illumina reads and long MinION reads was performed using Unicycler v0.4.7 (ref. 35) under conservative mode for increased accuracy. Complete circular contigs were then corrected and polished using Pilon v1.22 (ref. 36), in addition to the integrated polishing steps in Unicycler, with Illumina reads for several rounds (7 on average) until no further improvements were reported. Prokka v1.13 (ref. 37) was used to annotate the genome sequence. Acquired antimicrobial resistance genes were identified using ResFinder v3.1 (http://genomicepidemiology.org/). STs were assigned by querying the multi-locus sequence typing database of E. coli (http://enterobase.warwick.ac.uk/species/index/ecoli).

Determining clonal relatedness by SNPs analysis

The chromosomal sequences of strain 020001 and 020026 obtained from the hybrid assembly of MinION/Illumina sequencing reads were used as the reference for mapping. Reads passing the quality-control thresholds aligned to the reference using Snippy v4.3.6 with default settings. Aligned pseudo-genomes were created and cleaned using the integrated scripts provided by Snippy v4.3.6. Phage regions (Supplementary Table 6) were identified using the PHASTER server38 and intact phage regions were masked using “N” with other settings as default. Recombination regions and a recombination-corrected phylogenetic tree were identified and inferred using Gubbins v2.3.4 (ref. 27) with the GTRGAMMA model and a maximum of 50 iterations. Matrix representing pairwise SNP distance was calculated using snp-dists v0.6.3.

Determining the population structure of ST410

All ST410 genome sequences with short reads available in GenBank (n = 327; Supplementary Dataset 2, accessed 1 August 2018) were retrieved from either the EnteroBase collection or from NCBI SRA database. Strict quality-control, de novo assembly and SNP calling were also performed on these reads as described above. Six strains (Supplementary Dataset 2) had >15% undetermined sites, which are shown as “N” in their genome sequences, in the pseudo-genome and were therefore excluded from all further analyses. A precise phylogeny was obtained by masking SNP sites residing in recombination regions using Gubbins v2.3.4 (ref. 27) as described above and the output phylogenetic tree was tested using bootstrapping (n = 1000) in RAxML v8.2.12 (ref. 39) under the GTRGAMMA model. The phylogenetic tree of ST410 genomes was visualized and annotated using iTOL v3 (ref. 40) and Phandango v1.3.0 (ref. 41).

Coalescent analysis of dated ST410 strains

Dated strains (Supplementary Dataset 2) with either a specific or an interval of time in unit of years were fed into Gubbins v2.3.4 (ref. 27) as described above to obtain a recombination-corrected tree, which was then used as the input in BactDating v1.0.1 (ref. 14) under mixed model with 108 iterations to ensure that the Markov chain Monte Carlo (MCMC) was run for long enough to converge (the effective sample size of the inferred parameters α, μ, and σ were >200). In addition, a coalescent analysis without the most distant ST410 strains including KOEGE 131 (358a) (accession no. SRR785629), MOD1-EC5419 (accession no. SRR6512532), KTE221 (accession no. SRR633754), and NC_STEC121 (accession no. SRR5470036) to accelerate convergence was performed using the same settings. The four strains do not belong to the B4/H24RxC MDR clone and are >5000 SNPs distant from all other ST410 strains (Supplementary Dataset 4).

Identifying loci specific to the B4/H24RxC MDR clone

The genome sequences of all ST410 strains (n = 327) were annotated using Prokka v1.13 (ref. 37). A pan-genome matrix of the 327 ST410 genomes was obtained using Roary v3.12.0 (ref. 42) with 90% used as the minimum amino acid identity, followed by a clone-association analysis using Scoary v1.6.16 (ref. 43) with a maximum adjusted p-value of 1e-30. fimH typing of ST410 strains was performed with FimTyper 1.0 (ref. 44). Genes present in the B4/H24RxC MDR clone (n = 37) but absent from all other ST410 genomes (n = 290), and vice versa, were considered as clone-specific. SNPs specific to the B4/H24RxC MDR clone were identified by feeding the entire SNP matrix of all ST410 strains into Scoary v1.6.16 (ref. 43) using the same settings as described above. SNPs over-represented in strains of the clone but absent from all other ST410 strains were considered as specific for the B4/H24RxC MDR clone. Gubbins v2.3.4 (ref. 27) was used to determine whether the clone-specific genes and SNPs were due to recombination. SNPs in the -10, -35 boxes of promoter or in the 5′ UTR regions of downstream genes were predicted using the online tool BPROM (http://www.softberry.com/)45.

Analysis of IncX3 plasmids within the B4/H24RxC MDR clone

All strains within the B4/H24RxC MDR clone of ST410 (n = 37) were screened for plasmid replicons using PlasmidFinder v2.0 and antimicrobial resistance genes identified using ResFinder v3.1. For strains containing an IncX3 replicon and blaNDM or blaOXA-181, their assembly graphs generated from SPAdes v3.13.0 and Unicycler v0.4.7 were investigated using Bandage v0.8.1 (ref. 46) by tracking paths flanking the resistance gene-containing and the replicon-containing contigs. Filtered reads of these strains were mapped against either plasmid pNDM5_020001 (accession no. CP032424; for blaNDM-carrying strains) or plasmid pOXA181 (accession no. KP400525; for blaOXA-181-carrying strains)16 or both plasmids (for strains carrying both blaNDM and blaOXA-181) using BWA v0.7.17 (ref. 47) with default settings. pNDM5_020001 is a blaNDM-carrying IncX3 plasmid of strain 020001, while plasmid pOXA181 (accession no. KP400525) is a blaOXA-181-carrying IncX3 plasmid of strain 005828 (ref. 16), both of which are within the B4/H24RxC MDR clone. Contigs representing part of the complete plasmids were retrieved from entire genomes using standalone nucleotide BLAST v2.7.1 (ref. 48).

Plasmid mobility testing

Conjugation experiments were carried out in broth and on filters with the azide-resistant E. coli strain J53 as the recipient as described previously49,50,51. For filter-based mating, overnight donor cultures (1 ml) were harvested by centrifugation, washed twice with 1 ml saline and re-suspended in 100 µl saline. Recipient cells were harvested from plates using a bent Pasteur pipette, washed, and suspended in 500 µl saline. Donor and recipient suspensions were mixed (50 μl each). The mixture was placed on a 0.45 μM cellulose-ester filter (Xinya; Shanghai, China) and then incubated on a blood agar plate at 37 °C for 4 h. Subsequently, the mixture of cultures was harvested in 1 ml saline, centrifuged, and re-suspended in 200 μl saline. For broth-based mating, overnight cultures of donor (25 µl) and recipient strains (250 µl) were added to 3 ml fresh BHI broth. The mixture was incubated for 18 h at 37 °C without shaking. Potential transconjugants were selected on LB agar plates containing 4 μg/ml meropenem and 150 μg/ml azide. The presence of blaNDM and the IncX3 replicon was verified by PCR with primers NDM-up (5′-TCGCCCCATATTTTTGCTAC)/NDM-dw (5′-CTGGGTCGAGGTCAGGATAG) for blaNDM52 and self-designed primers IncX3_forward (5′-GTTTTCTCCACGCCCTTGTTCA)/IncX3_reverse (5′-CTTTGTGCTTGGCTATCATAA) for the IncX3 replicon.

Plasmid stability testing

Plasmid pNDM5_020026, which was the blaNDM-5-carrying self-transmissible plasmid from strain 020026, and blaOXA-181-carrying plasmid pOXA-181 from strain 00582816 were selected as representatives for testing plasmid stability. Both pNDM5_020026 and pOXA-181 are IncX3 replicon type plasmids. Plasmid stability was tested in LB broth and minimal media (M9 with 0.2% glucose, which is simply referred as M9 broth in this study) to reflect both nutrient-rich and nutrient-restricted settings as described previously53. Briefly, E. coli J53 containing pNDM5_020026 or pOXA-181 was inoculated at 37 °C overnight in 15 ml LB or M9 broth in the presence of 150 μg/ml sodium azide plus 2 μg/ml meropenem (for strain with pNDM5_020026) or plus 0.5 μg/ml imipenem (for strain with pOXA-181). These cultures were washed with saline (0.9% NaCl) several times to remove carbapenems and sodium azide. Aliquots (15 μl) were added to 15 ml LB or M9 broth correspondingly. After incubation at 37 °C for 24 h with shaking at 200 r.p.m., 100 μl samples were collected, diluted 1:104 with LB or M9 broth after being measured to the 0.5 McFarland standard, and were then streaked onto LB or M9 agar plates with and without 2 μg/ml meropenem (for strain with pNDM5_020026) or 0.5 μg/ml imipenem (for strain with pOXA-181). The stability frequency of plasmids was calculated by log10(Ng)/log10(Nw), where Ng and Nw represent number of bacterial cells containing the plasmid and all bacterial cells in the media, respectively. In addition, 100 colonies from the agar plate without meropenem were randomly selected and streaked onto plates containing 2 μg/ml meropenem to calculate the percentage of plasmid loss. For each of the two plasmids in either LB or M9 medium, 300 colonies were selected in total from triplicate repeats of each experiment.

Biofilm formation assays

Strain 020026 was subjected to a biofilm formation assay with strains 020001 and 020032 used as controls as described previously54,55. Briefly, bacterial cells were harvested from overnight cultures in LB broth by centrifugation at 2500 r.p.m. for 10 min, were re-suspended with saline, and were adjusted to 0.5 McFarland standard. Aliquots (100 μl) were then pipetted into 96-well polystyrene culture plates and were incubated for 3 h at 37 °C to allow the formation of biofilms. The plates were washed twice with distilled water. Biofilms in the wells were fixed with 100 μl methanol per well for 15 min and were stained with 100 μl staining buffer containing 1% crystal violet for 5 min. The stained biofilms were washed again to remove the unbound stain and allowed to dry at room temperature. Biofilms were detected with 110 μl 33% glacial acetic acid by ELX800 Universal Microplate Reader (Bio-Tek, Winooski, VT, USA) at OD595 nm and OD595 nm. Acinetobacter baumannii strain ATCC 19606 was used as a positive control, while distilled water was used as the negative control. Absorption values of strain 020026 were compared with those of strain 020001 and 020032 separately using analysis of variance with the least significant difference method.

Virulence assay

Wax moth (Galleria mellonella) larvae weighing 250–350 mg (Tianjin Huiyude Biotech Company, Tianjin, China) were used to assess the virulence of strains 020026, 020001, and 020032. A hypervirulent K. pneumoniae strain, KP767 (ref. 56), was used as a positive control, while E. coli DH5α was used as a negative control. Overnight bacterial cultures were washed using phosphate-buffered saline (PBS) and were further adjusted with PBS to concentrations of 1 × 106 CFU/ml, 1 × 107 CFU/ml, 1 × 108 CFU/ml, and 1 × 109 CFU/ml. Larvae (n = 16) were injected with 10 μl of inoculum into hemocoel via the last left proleg using a 25-μl Hamilton syringe57. The infected larvae were then incubated in plastic containers at 37 °C. The number of live larvae was counted every 12 h for 3 days.

Iron source growth assays

The growth of strains 020026, 020001, and 020032 in the presence of 200, 250, 375, and 500 µM 2′2-dipyridyl (DIP; Sigma, St. Louis, MO, USA) in LB agar plates was examined to determine the MIC of DIP. Growth assays of strain 020026 under different iron sources were performed as described previously58 with strain 020001 and strain 020032 used as controls. Briefly, prior to inoculation, bacterial strains were cultured in LB broth containing 200 µM DIP, which was lower than MIC, for 6 h to limit growth of the strains and were then washed in PBS. Approximately 105 CFU of each strain were streaked onto LB agar plates in the presence of DIP at the MIC (500 µM for strains 020001 and 020026, and 250 µM for strain 020032). Iron sources (10 µl) including 10 mg/ml bovine serum albumin (BSA), 1 mM FeCl2, 10 µM hemin, 1 mg/ml hemoglobin, 10 mg/ml holo-transferrin, and 10 mg/ml lactoferrin (Sigma) were spotted directly onto the plate and were incubated 48–72 h at 37 °C. The growth of bacteria was detected by visual inspection.

Head to head competitions and relative fitness determination

The relative fitness (w) of strain 020026 compared with strains 020001 and 020032 was determined in 24-h head to head competitions in LB broth as described previously59. Briefly, the competitors were preconditioned in prewarmed LB broth for 24 h. After that, each strain cultures were measured to the 0.5 McFarland standard and a 10-μl aliquot of each competitor was mixed at a 1:1 ratio. The initial inoculum density of each competitor was approximately 2 × 103 cfu/ml. The mixture was then inoculated in 10 ml LB broth for 24 h at 37 °C and 200 r.p.m. Strain 020026 was resistant to aztreonam (MIC, >256 μg/ml), while strain 020001 was intermediate to aztreonam (MIC, 8 μg/ml; Supplementary Dataset 1). Therefore, strain 020026 could be differentiated from strain 020001 on agar plates containing 16 μg/ml aztreonam. Strains 020026 and 020032 could be differentiated on agar plates containing 2/4 μg/ml aztreonam–avibactam as strain 020026 was susceptible to aztreonam–avibactam (MIC, 1/4 μg/ml) and strain 020032 was intermediate (MIC, 8/4 μg/ml; Supplementary Dataset 1). Initial (N0) and final (N24) densities of each competitor were measured by selective (with 8 μg/ml aztreonam for strains 020026 and 020001 or 2/4 μg/ml aztreonam–avibactam for strains 020026 and 020032) and non-selective (without aztreonam or aztreonam–avibactam) plating on LB agar plates. The w value was calculated using the equation, w = log10(Ng24/Ng0)/log10(Nw24/Nw0), where Ng and Nw are bacterial densities of strain 020026 and the competitor strain 020001 or 020032, respectively. A <1w value suggests a fitness disadvantage, while w > 1 suggests a fitness advantage60. Strain 020026 was also competed with strain 020001 in the presence of 375 µM DIP in LB broth and LB agar plates and the competition was performed as described above to determine the relative fitness of strain 020026 under iron-deprived conditions.

Statistics and reproducibility

For biofilm formation assays, the differences of absorption values at both OD590 nm and OD595 nm among strains 020026, 020001, and 020032 were compared with one-way ANOVA, which were calculated using SPSS version 21.0 (IBM Analytics; Armonk, NY, USA). For head to head competition, two-tailed t-test was used to compare the relative fitness of strain 020026 compared to the competitor strain in non-iron-deprived and iron-deprived conditions, which was calculated using SPSS. All P values were two-tailed, and P < 0.05 was considered statistically significant.

For biofilm formation assays, all experiments were performed in triplicate (biological replicates) and for each replicate, experiments were repeated nine times (technical replicates). For plasmid stability testing, the experiments for both pNDM5_020026 and pOXA-181 in LB broth and M9 media were all performed in triplicate (biological replicates). All experiments of virulence assay using wax moth were performed with five biological replicates, while iron source growth assay was performed in triplicate (biological replicates). For head to head competition, all experiments were performed in triplicate (biological replicates) and each biological replicate was repeated three times (technical replicates)61.

Reporting Summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.