Population structure, antimicrobial resistance, and virulence factors of diabetic foot-associated Escherichia coli

TL;DR

This study analyzes the genetic diversity and drug resistance of Escherichia coli strains from diabetic foot infections, revealing their potential to cause severe infections.

Contribution

The first multisite genomic characterization of diabetic foot-associated Escherichia coli (DFEC) and their antimicrobial resistance and virulence traits.

Findings

DFEC strains are phylogenetically diverse, spanning 7 phylogroups and 28 sequence types.

High prevalence of multidrug-resistant and extensively drug-resistant DFEC strains was observed.

DFEC strains possess virulence factors linked to extraintestinal pathogenic E. coli and show metabolic convergence.

Abstract

Diabetic foot infections (DFIs) are a major complication of diabetes, often leading to lower limb amputations. Escherichia coli is a predominant Gram-negative pathogen in DFI, yet its genomic and pathogenic features remain poorly characterized. Here, we present a whole-genome sequence-based analysis of diabetic foot-associated E. coli (DFEC) isolates from diverse geographical locations. Phylogenetic reconstruction revealed substantial diversity, with strains spanning 7 phylogroups and 28 sequence types. Capsule biosynthesis loci linked to invasive infections, such as K1, K2ab, and K5, were also detected. The DFEC pangenome comprised 18,263 gene clusters, indicating high genomic plasticity. The plasmid repertoire was also varied and contributed to the genomic diversity of the strains. Approximately 78% of isolates were multidrug-resistant or extensively drug-resistant, with resistance to last-resort antibiotics such as colistin and carbapenems also observed. High frequencies of virulence factors involved in host cell adherence, iron metabolism, serum survival, as well as toxins and type 3 secretion system genes were also detected. In contrast, metabolic modeling showed conserved biochemical profiles. Clustering based on accessory metabolic functions did not mirror phylogeny, suggesting metabolic convergence among distinct lineages. Collectively, these findings reveal that DFEC are versatile pathogens with a repertoire of antimicrobial resistance and virulence determinants. These traits make them functionally distinct from commensal E. coli strains and highlight the potential of DFEC to cause severe and invasive infections.

This study presents the first multisite genomic characterization of diabetic foot-associated Escherichia coli (DFEC). Our findings reveal that DFEC strains are phylogenetically diverse and span multiple lineages. The high prevalence of multidrug-resistant and extensively drug-resistant genotypes underscores the underestimated antimicrobial resistance (AMR) threat posed by DFEC. We detect high frequencies of virulence factors commonly associated with extraintestinal pathogenic E. coli, which indicates that DFEC might have the potential to cause severe complications, such as sepsis. The large accessory genome and evidence of metabolic convergence across distinct lineages highlight the adaptive versatility of DFEC in the polymicrobial and inflammatory environment of chronic wounds. These insights advance our understanding of DFEC pathobiology and support the development of targeted diagnostics, AMR surveillance, and therapeutic strategies to improve clinical outcomes for diabetic patients.

Figures

Phylogenetic reconstruction of DFEC strains. (A) Maximum-likelihood codon-based phylogenetic tree, showing the distribution of DFEC isolates across phylogroups. 2 landmark strains (black text) were used to represent E. coli phylogroups A, B1, B2, C, D, F, E, and G. Shigella strains (S. boydii [Sb], S. sonnei [Ss], S. flexneri [Sf], and S. dysenteriae [Sd]), as well as E. albertii, E. ruysiae, E. marmotae, and E. fergusonii, were also included. Phylogroups that encompass DFEC strains are highlighted in colors. DFEC strains that were predicted as S. dysenteriae-like and S. boydii-like strains are labeled as A* and B1*, respectively. Text in different colors refers to DFEC strains from the different countries. Bootstrap support values were calculated from 100 replicates. All nodes are supported by ≥95% bootstrap values, except the node uniting DFI_GH_15 and DFI_GH_31 (77%). (B–E) Pie charts presenting percentages of DFEC strains per phylogroup (B), O-antigen type (C), flagellar/H type (D), and capsular/K type (E). Phylogroups were determined by ClermontTyping. O and H serotypes were predicted using SeroTypeFinder. K serotypes were predicted using Kaptive.

Colicin and plasmid diversity in DFEC strains. For each strain, the number of colicins and plasmids of a certain type is reported. Total counts per strain, as well as total counts for each colicin and plasmid Inc type, are also included. Plasmid and colicin types were identified using ABRicate.

Pangenome analysis for DFEC isolates. (A) Gene presence/absence heatmap. Genomes were annotated using Prokka, and the pangenome was reconstructed using Roary. Strain clustering is based on patterns of presence/absence. Phylogroups and countries of origin are labeled in different colors. (B) Gene frequency histogram plot, representing the number of genomes sharing different numbers of genes. Singletons are displayed at the far left, while core genome genes are represented at the far right. The inset pie chart represents the overall number of genes within core, soft core, shell, and cloud genomes. (C) Pangenome and core genome accumulation curves. Data are represented as average ± SD of 100 reinteractions performed with different randomized orders of genomes. See also Table S2.

AMR pattern for DFEC strains. (A–H) Percentage of strains (for all strains and each phylogroup individually) predicted to be resistant to different antibiotics. Different antibiotic classes are shown in different colors. (I) Percentage of strains (for all strains and each phylogroup individually) that are non-drug-resistant (NDR), drug-resistant (DR), multidrug-resistant (MDR), and extensively drug-resistant (XDR). AMR phenotypes were predicted using ABRicate. See also Table S3 and S4.

Presence/absence of ExPEC-associated virulence factors in DFEC isolates. (A) Presence/absence of adhesins and invasins. (B) Presence/absence of serum survival factors/protectins, iron acquisition factors, UPEC-associated pathogenicity island markers (U), toxins, and T3SS factors. Isolates are arranged by phylogroups, and different functional groups of virulence factors are labeled with different colors. Instances where factors are plasmid-borne are highlighted with text in bold and marked with a bold outline. Virulence factors were predicted using ABRicate. Presence of factors on plasmids or chromosomes was predicted via PlasmidHunter. See also Tables S5 and S6.

Genome-based metabolic modeling of DFEC isolates. Metabolic reactions were predicted using CarveMe. Strain clustering is based on patterns of presence/absence. Phylogroups and countries of origin are labeled in different colors. See also Table S7.

Tables

TABLE 1. Summary of the characteristics of the DFEC strains used in this studya

	Name	Country	Phylogroup	O type	H type	K type (K locus)	MLST (Enterobase)	MLST (Pasteur)
1	DFI_NG_015	Nigeria	A	Unknown	H10	Unknown (Unmatched)	2705	532
2	DFI_NG_016	Nigeria	A	Unknown	H19	Unknown (KL214)	4447	446
3	DFI_GH_30A	Ghana	A	O101	H10	Unknown (KL139)	10	2
4	DFI_GH_31	Ghana	A	O101	H10	Unknown (KL139)	10	2
5	DFI_NG_007	Nigeria	A	O101	H10	Unknown (KL139)	10	2
6	DFI_GH_15	Ghana	A	O101	H10	Unknown (KL139)	10	*64d4
7	DFI_GH_13	Ghana	A	O101	H10	Unknown (KL139)	617	2
8	DFI_MY_GCF_018095305.1	Malaysia	A	O101	H21	Unknown (KL59)	167	2
9	DFI_CN_GCF_019660995.1	China	A	O101	H9	Unknown (KL214)	167	2
10	DFI_NG_019	Nigeria	A	O128	H53	Unknown (KL214)	*d091	*59a5
11	DFI_NG_033	Nigeria	A	O25	H16	K5 (KL6)	450	132
12	DFI_NG_020	Nigeria	A	O32	H21	K96 (KL36)	*afd6	*9e26
13	DFI_NG_011	Nigeria	A	O45	H16	K5 (KL6)	450	132
14	DFI_NG_034	Nigeria	A	O45	H16	Unknown (Atypical)	450	132
15	DFI_NG_013	Nigeria	B1	Unknown	H8	Unknown (KL214)	448	58
16	DFI_NG_014	Nigeria	B1	Unknown	H8	Unknown (KL214)	448	58
17	DFI_SE_GCF_021498165.1	Sweden	B1	Unknown	H5	Unknown (KL139)	940	*695a
18	DFI_BR_GCF_001660545.2	Brazil	B1	O82	H8	Unknown (KL214)	101	88
19	DFI_UK_039MTS00172	UK	B1	O9	H9	Unknown (KL58)	*038c	355
20	DFI_UK_045MTS00209	UK	B2	O1	H7	K1 (KL8A1)	95	1
21	DFI_GH_34	Ghana	B2	O25	H4	K2ab (KL3)	131	9
22	DFI_UK_084MTS00019	UK	B2	O25	H4	K16 (KL47)	131	9
23	DFI_UK_229MTS6	UK	B2	O25	H4	Unknown (KL74)	131	43
24	DFI_UK_294MTS000723	UK	B2	O25	H4	K5 (KL6)	131	43
25	DFI_UK_295MTS00211	UK	B2	O25	H4	K5 (KL6)	131	43
26	DFI_KR_GCF_029992095.1	South Korea	B2	O25	H4	K2ab (KL3)	131	43
27	DFI_NG_009	Nigeria	B2	O25	H4	K2ab (KL3)	131	43
28	DFI_SE_GCA_021559635.1	Sweden	B2	O25	H4	K5 (KL6)	*0a27	*4ad2
29	DFI_SE_GCA_021562195.1	Sweden	B2	O25	H4	K5 (KL6)	*55cb4	*da89
30	DFI_IN_GCF_014893515.1	India	B2	O6	H31	K2ab (KL3)	127	22
31	DFI_GH_10	Ghana	B2	O6	H31	K14 (KL16)	127	33
32	DFI_GH_17	Ghana	B2	O6	H31	K14 (KL16)	127	33
33	DFI_GH_30B	Ghana	B2	O6	H31	K14 (KL16)	127	33
34	DFI_UK_234MTS000612	UK	B2	O75	H5	K5 (KL6)	404	6
35	DFI_SE_GCF_021498145.1	Sweden	D	O102	H6	Unknown (KL59)	405	44
36	DFI_US_GCA_032338595.2	USA	D	O153	H30	K2ab (KL3)	38	8
37	DFI_UK_177MFL00079	UK	D	O17	H18	Unknown (Atypical)	69	3
38	DFI_UK_046MTS00052	UK	F	O1	H6	Unknown (Atypical)	648	*5253
39	DFI_MY_GCF_018095165.1	Malaysia	F	O1	H6	K7 (KL10)	648	*5253
40	DFI_NG_008	Nigeria	B1*	O9	H21	Unknown (KL139)	120	*9623
41	DFI_NG_024	Nigeria	A*	O9	H10	Unknown (KL139)	46	*bd9e
42	DFI_NG_010	Nigeria	A*	O9	H4	Unknown (KL214)	609	679

Full text

INTRODUCTION

Diabetic foot infections (DFIs) are a common complication of diabetes and a leading cause of lower limb amputation worldwide (1). It is estimated that 19%–34% of diabetic patients will develop diabetic foot ulcers (DFUs) during their lifetime, 80% of which will progress to DFIs. DFIs are frequently caused by Gram-positive cocci, particularly Staphylococcus aureus (2). However, recent studies also report frequent isolation of Gram-negative bacilli, including Escherichia coli, Klebsiella pneumoniae, and Pseudomonas aeruginosa (3–6). Notably, studies conducted in low- and middle-income countries report significantly higher frequency of E. coli isolation from DFIs compared to those from high-income countries. This disparity may reflect differences in healthcare infrastructure, diagnostic practices, or environmental factors (5, 6). Although E. coli strains are frequently isolated from DFIs, their role in disease pathogenesis remains poorly characterized. E. coli is a major cause of extraintestinal infections and is responsible for over 2 million deaths each year (7). Extraintestinal pathogenic E. coli (ExPEC) exhibits considerable diversity in both its phylogenetic background and its repertoire of virulence factors. These strains commonly carry virulence genes located either on plasmids or within chromosomal pathogenicity islands, enabling adherence, iron acquisition, immune evasion, and host cell damage (8). Individuals with diabetes are at elevated risk of ExPEC infections compared to the general population, with urinary tract infections and DFIs representing 2 of the most common clinical manifestations (9). Host factors prevalent in diabetic patients, such as immunosuppression, poor glycemic control, and advanced age, may also predispose individuals to infection by E. coli strains with lower inherent virulence (10). Therefore, understanding the genomic features of diabetic foot-associated E. coli (DFEC) is essential to resolve their pathogenic potential and inform both surveillance and treatment strategies.

Together with other Enterobacteriales, E. coli is recognized by the World Health Organization as a critical-priority pathogen for the development of new antimicrobial agents (11). This designation reflects growing concerns over E. coli’s resistance to multiple drug classes. Of particular concern is the emergence of extended-spectrum β-lactamase-producing and carbapenemase-producing E. coli strains, resistant to third-generation cephalosporins and carbapenems, which severely limit therapeutic options (12–15). While extensive characterization of pathogenic E. coli from various infection sites has been performed, the population structure, antimicrobial resistance (AMR) patterns, and virulence arsenal of DFEC strains remain largely unexplored.

In this study, we performed whole-genome sequence (WGS) analysis of 42 E. coli strains isolated from DFIs (14 from Nigeria, 9 from the UK, 8 from Ghana, 4 from Sweden, 2 from Malaysia, 1 from China, 1 from South Korea, 1 from Brazil, 1 from India, and 1 from the USA). We describe, for the first time, their phylogenetic diversity, AMR profiles, plasmid content, and virulence gene repertoire. Our findings highlight the clinical significance of E. coli in DFIs and its potential to cause severe systemic disease in the vulnerable diabetic population.

RESULTS

DFEC isolates are highly diverse

To investigate the genomic diversity of DFEC, we collected isolates from 3 geographically distinct sites (Nigeria, UK, and Ghana) and performed WGS. In addition to newly sequenced isolates, we incorporated publicly available genomes from previous studies, resulting in a total data set of 42 E. coli isolates (14 from Nigeria, 9 from the UK, 8 from Ghana, 4 from Sweden, 2 from Malaysia, 1 from China, 1 from South Korea, 1 from Brazil, 1 from India, and 1 from the USA).

Genome sizes ranged from 4,675,442 to 5,466,448 base pairs, with coding sequence counts between 4,567 and 6,385, as predicted by BV-BRC. GC content was consistent across isolates, ranging from 50.28% to 51.07% (Table 1; Table S1).

Core genome phylogenetic reconstruction was conducted using BV-BRC tools. Reference strains of known phylogroups were incorporated in this analysis to contextualize the DFEC isolates (Fig. 1A). DFEC strains were distributed across 7 phylogroups: A, B1, B2, D, F, A*, and B1*. Phylogroups A (33.33%) and B2 (35.71%) were the most prevalent (Fig. 1B). Clusters A, B1, B2, D, and F were supported by multiple isolates from different geographic origins. Clusters A* and B1* consisted of 2 isolates and 1 isolate, respectively, all from Nigeria. Although related to phylogroup A, A* was considered a distinct lineage, since species prediction using the tool Pathogenwatch classified these strains as S. dysenteriae. The 2 A* strains also had the smallest genomes, consistent with the process of genomic streamlining that occurred in S. dysenteriae (16). Similarly, the single isolate in the lineage B1* was predicted as belonging to phylogroup B1 according to Clermont typing but was classified as S. boydii according to Pathogenwatch, and this strain clustered together with other Shigella lineages (S. boydii and S. sonnei). Despite these genetic similarities to Shigella lineages, A* and B1* strains were all devoid of the typical plasmid of invasion (pINV) of Shigella and Shiga toxins and were able to ferment lactose on MacConkey plates (Fig. S1), as opposed to Shigella strains, which are typically unable to do so (17).

Phylogenetic reconstruction of DFEC strains. (A) Maximum-likelihood codon-based phylogenetic tree, showing the distribution of DFEC isolates across phylogroups. 2 landmark strains (black text) were used to represent E. coli phylogroups A, B1, B2, C, D, F, E, and G. Shigella strains (S. boydii [Sb], S. sonnei [Ss], S. flexneri [Sf], and S. dysenteriae [Sd]), as well as E. albertii, E. ruysiae, E. marmotae, and E. fergusonii, were also included. Phylogroups that encompass DFEC strains are highlighted in colors. DFEC strains that were predicted as S. dysenteriae-like and S. boydii-like strains are labeled as A and B1*, respectively. Text in different colors refers to DFEC strains from the different countries. Bootstrap support values were calculated from 100 replicates. All nodes are supported by ≥95% bootstrap values, except the node uniting DFI_GH_15 and DFI_GH_31 (77%). (B–E) Pie charts presenting percentages of DFEC strains per phylogroup (B), O-antigen type (C), flagellar/H type (D), and capsular/K type (E). Phylogroups were determined by ClermontTyping. O and H serotypes were predicted using SeroTypeFinder. K serotypes were predicted using Kaptive.*

Multilocus sequence typing (MLST) revealed substantial diversity, with 25 and 26 distinct sequence types identified by the Achtman and Pasteur schemes respectively, resulting in a total of 28 unique sequence types when both schemes were considered simultaneously (Table 1; Table S1). Serotyping predicted 13 O-antigen types (an additional 5 strains were untypeable) and 14 H-antigen types (Fig. 1C and D). The most common O/H combinations were O25:H4 (21.43%), O101:H10 (11.90%), and O6:H31 (9.52%), all lineages previously associated with ExPEC (18). K-antigen typing identified 12 distinct K loci (KL, 4 additional strains were atypical or unmatched), 7 of which could be attributed to known capsule types (Fig. 1E). These included K1, K2 (subtype K2ab), and K5, which are commonly associated with ExPEC pathotypes causing sepsis and urinary tract infections. Capsule loci KL214 and KL139 (for which the precise capsule serotype is unknown) were particularly abundant, and these loci were also previously detected in other bioinformatic screens on ExPEC strains (19–21).

Given the role of plasmids in the dissemination of AMR and virulence genes, we also characterized plasmid incompatibility (Inc) types (Fig. 2). All but 4 isolates harbored between 2 and 6 Inc types. The most prevalent were IncFIB (AP001918) (61.90% of strains) and IncFIA (45.24% of strains). We also screened for colicin-associated plasmids, identifying Col156 (26.19% of strains) and ColRNAI (23.81% of strains) as the most common types (Fig. 2). Plasmids and colicins may contribute significantly to niche adaptation in the polymicrobial environment of DFUs, for example, by mediating interbacterial competition or by encoding systems that enhance nutrient acquisition and metabolic versatility (22, 23).

Colicin and plasmid diversity in DFEC strains. For each strain, the number of colicins and plasmids of a certain type is reported. Total counts per strain, as well as total counts for each colicin and plasmid Inc type, are also included. Plasmid and colicin types were identified using ABRicate.

Taken together, these findings underscore the high phylogenetic diversity of E. coli in the diabetic foot.

DFEC isolates possess an open pangenome and a large accessory genome

To characterize the genomic diversity of DFEC, we performed a pangenome analysis. Across the 42 genomes, we identified a total of 18,263 gene clusters, reflecting substantial genomic variability (Fig. 3A and B; Table S2). As expected, clustering of isolates according to their pattern of gene presence/absence recapitulated phylogenetic reconstruction and was in agreement with the strains’ phylogroups (Fig. 3A).

Pangenome analysis for DFEC isolates. (A) Gene presence/absence heatmap. Genomes were annotated using Prokka, and the pangenome was reconstructed using Roary. Strain clustering is based on patterns of presence/absence. Phylogroups and countries of origin are labeled in different colors. (B) Gene frequency histogram plot, representing the number of genomes sharing different numbers of genes. Singletons are displayed at the far left, while core genome genes are represented at the far right. The inset pie chart represents the overall number of genes within core, soft core, shell, and cloud genomes. (C) Pangenome and core genome accumulation curves. Data are represented as average ± SD of 100 reinteractions performed with different randomized orders of genomes. See also Table S2.

The core genome, defined as genes present in at least 99% of isolates, comprised 2,413 genes, aligning with previous estimates for other E. coli pathotypes. The soft core genome (genes present in more than 95% and less than 99% of isolates) was relatively small, consisting of only 411 genes. In contrast, the shell genome (genes present in more than 15% and less than 95% of isolates) and cloud genome (genes present in less than 15% of isolates) were extensive, comprising 3,566 and 11,873 genes, respectively (Fig. 3B). Notably, nearly 7,000 genes were singletons (genes found in only 1 genome), highlighting the remarkable accessory genome diversity within this data set.

To further assess the openness of the DFEC pangenome, we generated pangenome and core genome accumulation curves by iteratively sampling genomes in random order (100 permutations) and averaging the results (Fig. 3C). The pangenome curve exhibited a steep initial increase followed by a gradual rise, without reaching a plateau, indicating an open pangenome. In contrast, the core genome size rapidly declined with the addition of the first few genomes and stabilized at 2,413 core genes (Fig. 3C).

Taken together, these findings underscore the extensive genomic plasticity of DFEC isolates and support the notion that E. coli from diabetic foot maintain an open and dynamic pangenome, likely shaped by horizontal gene transfer and adaptation to diverse ecological niches.

DFEC isolates exhibit widespread AMR

To assess the AMR profiles of the DFEC isolates, we analyzed further the genome of the isolates to detect resistance determinants across 12 antibiotic functional classes: macrolides, tetracyclines, polymyxins, phosphonic acid antibacterials, fluoroquinolones, folate synthesis inhibitors (sulfamethoxazoles and trimethoprim), monobactams, carbapenems, cephalosporins, penicillins, amphenicols, and aminoglycosides (Fig. 4; Tables S3 and S4). Resistance determinants were detected across over 85.71% isolates. Overall, resistance to all antibiotic classes was detected (Fig. 4A). However, resistances to colistin, fosfomycin, meropenem, cefixime, and tigecycline were relatively uncommon, each observed in fewer than 5% of isolates. Specifically, colistin resistance was identified in a single isolate of phylogroup B1 and attributed to the presence of mcr-1.1. Fosfomycin resistance was detected in 2 isolates (phylogroups B1 and F), both carrying fosA3. Meropenem resistance was observed in 2 isolates (phylogroups A and B2) harboring either blaNDM-5 or blaKPC-2. Cefixime resistance was detected in 1 isolate (phylogroups A) harboring blaNDM-5. Tigecycline resistance was also found in 1 isolate (phylogroup A) and was conferred by tet(X) (Table S4).

AMR pattern for DFEC strains. (A–H) Percentage of strains (for all strains and each phylogroup individually) predicted to be resistant to different antibiotics. Different antibiotic classes are shown in different colors. (I) Percentage of strains (for all strains and each phylogroup individually) that are non-drug-resistant (NDR), drug-resistant (DR), multidrug-resistant (MDR), and extensively drug-resistant (XDR). AMR phenotypes were predicted using ABRicate. See also Table S3 and S4.

In contrast, resistance to tetracycline, sulfamethoxazole, trimethoprim, cephalothin, ticarcillin, ampicillin, and streptomycin was highly prevalent, with cognate resistance genes detected in over 50% of isolates. The factor mdf(A) (a transporter involved in resistance to a wide spectrum of antibiotics) was present in 100% of isolates, including those not predicted to exceed susceptibility/resistance thresholds for any antibiotic classes (Table S4). Highly represented were also blaTEM-1B (contributing to resistance against β-lactam antibiotics, 64.29% of isolates), sul1, sul2, dfrA17 (contributing to resistance against folate pathway inhibitors, 45.24%, 42.86%, and 40.48% of isolates respectively), aph(6)-Id (contributing to resistance against aminoglycosides, 40.48% of isolates), and tetB (contributing to resistance against tetracyclines, 35.71% of isolates) (Table S4).

Phylogroup-specific analysis revealed notable differences in resistance profiles (Fig. 4). All isolates from phylogroups B1, F, and A* were multidrug-resistant (MDR), defined as resistance to agents from at least 3 distinct antibiotic classes (24, 25). Furthermore, within phylogroup A, 7.14% of isolates met the criteria to be considered extensively drug-resistant (XDR) strains (Table S3), defined as resistance to antimicrobials in all but 2 antibiotic classes (26).

These findings underscore the significant AMR burden among DFEC isolates and highlight their potential role as reservoirs of multidrug resistance in chronic wound infections.

DFEC isolates exhibit virulence profiles characteristic of ExPEC lineages

To characterize the virulence potential of DFEC isolates, we performed in silico prediction of virulence-associated genes. 469 distinct virulence factors were identified across the data set (Tables S5 and S6). To determine whether these isolates exhibit traits typical of ExPEC, we focused on adhesins, invasins, serum resistance factors, iron acquisition factors, toxins, and type 3 secretion system (T3SS)-associated factors, since these functional groups have previously been associated with ExPEC pathogenesis (27–30) (Fig. 5; Table S5). A broad repertoire of adhesins was identified, including multiple fimbrial and pili operons (Fig. 5A). Curli subunit genes (csgA, csgB, csgC, csgD, csgE, csgF, csgG) and type 1 fimbriae genes (fimA, fimB, fimC, fimD, fimE, fimF, fimG, fimH, fimI) were nearly ubiquitous. In contrast, pyelonephritis-associated pili genes (papA, papB, papC, papD, papE, papF, papG, papH, papI, papJ, papK) were detected only in a subset of isolates across multiple phylogroups (16.67%–45.24% of isolates for each pap gene), and S fimbrial adhesins (sfaA, sfaB, sfaC, sfaD, sfaE, sfaF, sfaG, sfaH, sfaS) were detected only in a subset of B2 phylogroup strains (9.52% of isolates). The autotransporter gene agn43 was highly prevalent (52.38% of isolates) and frequently found in multiple copies, suggesting potential amplification events. Invasins ibeB and ibeC were commonly detected (100% and 78.57%, respectively) (Fig. 5A). Genes associated with serum resistance, including kpsD, kpsM, traT, ompA, ompT, and iss2 2 (42.86%–57.14% of isolates), were widely distributed across phylogroups. Iron acquisition systems were well represented (Fig. 5B). The chromosomal heme uptake genes (chuA, chuS, chuT, chuU, chuV, chuW, chuX, chuY) were present in isolates from phylogroups B2, D, and F (47.62%–50.00% of isolates for each chu gene). The ferric enterobactin transport ATP-binding protein fepC was present in all isolates. The siderophore receptor gene fyuA was highly common (73.81% of isolates). Interestingly, hlyE (encoding a hemolysin) was present in all isolates, except those from phylogroup B2. Genes encoding the aerobactin system (iucA, iucB, iucC, iucD, iutA) and the sit siderophore system (sitA, sitB, sitC, sitD) (50.00%–66.67% of isolates) were detected across all phylogroups. The malX gene, often associated with UPEC pathogenicity islands, was consistently present in all B2 and F isolates and was largely exclusive to these groups (Fig. 5B). Toxin-encoding genes, such as sat and usp (30.95%–38.10% of isolates), were enriched in B2 isolates and elements of the T3SS (14.29%–35.71% of isolates) were identified in several isolates, except those from phylogroups B2 and B1* (Fig. 5B).

Presence/absence of ExPEC-associated virulence factors in DFEC isolates. (A) Presence/absence of adhesins and invasins. (B) Presence/absence of serum survival factors/protectins, iron acquisition factors, UPEC-associated pathogenicity island markers (U), toxins, and T3SS factors. Isolates are arranged by phylogroups, and different functional groups of virulence factors are labeled with different colors. Instances where factors are plasmid-borne are highlighted with text in bold and marked with a bold outline. Virulence factors were predicted using ABRicate. Presence of factors on plasmids or chromosomes was predicted via PlasmidHunter. See also Tables S5 and S6.

Next, we sought to determine whether the virulence genes classically associated with ExPEC were localized on the chromosome or plasmids. The factor traT was the only factor to be always present on plasmids, in agreement with its function in conjugative transfer of DNA. A few adhesion factors (afaA, afaB, afaC-I, afaD, draP, papX), serum survival factors (cvaC, iss2, ompT), hemolysins (hlyA, hlyB, hlyC, hlyD), iron acquisition systems (iroB, iroC, iroD, iroE, iroN, iucA, iucB, iucC, iucD, iutA, sitA, sitB, sitC, sitD), and toxins (cnf1, sat) could be either chromosomally encoded or plasmid encoded. Other adhesins, invasins, iron acquisition systems, toxins, and T3SS-associated factors were always chromosomal.

Taken together, these findings indicate that DFEC strains display virulence profiles similar to those described for other ExPEC pathotypes.

DFEC isolates possess broad metabolic abilities

To investigate whether the genomic diversity among DFEC isolates corresponds to diversity in metabolic capacity, we built genome-scale metabolic models for each strain. The number of metabolic reactions predicted spanned between 2,450 and 3,366 (Fig. 6; Table S7). A total of 3,921 metabolic reactions were identified. Of these, 2,106 (53.71%) were shared between all strains. However, only a few reactions were specific to a single phylogroup (Table S7). A single reaction was predicted uniquely for phylogroup B1* (PTRCTA2: Putrescine Transaminase pyruvate, providing the ability to catabolize putrescine with pyruvate), and 4 were predicted only in phylogroup F (FGLU: Formimidoylglutamase [involved in the histidine degradation pathway], GNNUC: Ribosylpyrimidine nucleosidase [involved in nucleotide recycling and nucleoside turnover], NMNN: Nicotinamide mononucleotide nucleosidase [involved in the regulation of NAD^+^ levels], and URAt2: Uracil transport via proton symport [allowing the import of uracil for nucleotide synthesis]). Notably, the clustering according to metabolic abilities did not entirely recapitulate the phylogenetic clustering (Fig. 6). In fact, all the phylogroups with 2 or more isolates segregated into 2 or more clusters with other strains, indicating that metabolic abilities may have been acquired and lost multiple times as a result of adaptive processes, for example, via gene inactivation and horizontal gene transfer.

Genome-based metabolic modeling of DFEC isolates. Metabolic reactions were predicted using CarveMe. Strain clustering is based on patterns of presence/absence. Phylogroups and countries of origin are labeled in different colors. See also Table S7.

DISCUSSION

This study provides the first genomic characterization of E. coli isolates from DFUs across multiple geographic regions, revealing high diversity in phylogeny, resistance, and virulence content. Our findings reveal that DFEC has a heterogeneous population structure, spanning several phylogenetic lineages, sequence types, and serotypes. This diversity underscores the adaptive versatility of E. coli in chronic wound environments and highlights the complexity of managing DFIs in clinical settings (5, 31, 32).

The phylogenetic analysis revealed representation across 7 phylogroups (A, B1, B2, D, F, A*, and B1*), with A and B2 being the most prevalent. The frequent detection of O25:H4 (phylogroup B2), O101:H10 (phylogroup A), and O6:H31 (phylogroup B2) serotypes, which are lineages associated with ExPEC pathotypes (33), supports the hypothesis that DFEC infections may be established by strains with extraintestinal virulence potential. In agreement, ExPEC-associated capsular types (K1, K2ab, K5) were also identified.

Plasmid analysis revealed a high prevalence of IncF-type plasmids, particularly IncFIB and IncFIA, which are known to carry both AMR and virulence genes. The detection of colicin-associated plasmids (Col156 and ColRNAI) in a subset of isolates suggests a potential role for interbacterial competition and niche adaptation in the polymicrobial environment of DFUs. These plasmids may enhance bacterial fitness by promoting survival under nutrient-limited or competitive conditions.

Pangenome analysis revealed that DFEC strains have diversified genomes and a substantial accessory gene content. The identification of 2,413 core genes alongside over 10,608 cloud genes (of which nearly 7,000 singletons) aligns with other large-scale studies and reflects the ecological versatility of E. coli as a species (4). This diversity is likely driven by ongoing horizontal gene transfer and pathoadaptations. Genomic plasticity may also be particularly relevant in the context of polymicrobial infections, where interspecies interactions and selective pressures are intense.

AMR genotypic profiling predicted a concerning burden of multidrug resistance, with 78.57% of all isolates classified as MDR or XDR based on genotypic predictions. β-Lactamase genes, particularly bla_TEM-1B, were widespread, and determinants involved in the resistance to front-line and last-resort antibiotics were detected (Table S4). The consistent presence of the efflux pump gene mdf(A), even in isolates not phenotypically resistant, suggests a baseline potential for broad-spectrum resistance and general recalcitrance to antimicrobial therapy (34, 35). Phylogroup-specific analysis revealed that resistance burden was not uniformly distributed. Isolates from phylogroups B1, F, and A* were all MDR. XDR strains were only detected within phylogroup A, and strains from phylogroups B2 and D had the largest proportions of non-drug-resistant strains. This pattern may reflect differences in plasmid content and horizontal gene transfer dynamics. Importantly, none of the antibiotics addressed was predicted to be ubiquitously effective. However, resistance to tigecycline, colistin, meropenem, and cefixime was rare, and therefore, these antibiotics could still be applied in clinical practice. Although AMR was inferred through genomic data rather than phenotypic testing, genotypic predictions have been shown to correlate well with resistance profiles in E. coli (36). Nonetheless, further phenotypic validation by gold-standard antimicrobial susceptibility testing (i.e., disk diffusion assay, MIC determination) will be valuable to support treatment guidelines.

In agreement with the phylogenetic evidence, virulence-associated genes in DFEC isolates revealed a profile consistent with ExPEC strains. We detected high frequencies of genes associated with adhesion, iron acquisition, serum resistance, and invasins (27–30). The widespread detection of adhesins, such as curli (csg operon) and type 1 fimbriae (fim operon), suggests potential for biofilm formation and epithelial attachment, both of which are critical in chronic wound colonization (37–41). The detection of pyelonephritis-associated pili genes and S fimbrial adhesins in a subset of isolates (particularly those from phylogroup B2) aligns with their established roles in urinary tract and bloodstream infections (38). The detection of multiple autotransporters, invasins (ibeB, ibeC), and serum resistance genes (traT, iss2) across phylogroups suggests that DFEC strains might also be equipped to evade host immune responses and persist in the inflammatory milieu of diabetic wounds. Iron acquisition systems, including the chu, sit, and iuc operons, as well as the siderophore receptors fyuA and iutA, were also prevalent, reflecting the importance of iron scavenging in nutrient-limited environments such as chronic ulcers. Of note, our study also identified T3SS components and toxins in several isolates, which further exemplifies the functional diversity of DFEC virulence mechanisms. Although T3SS factors have been described in ExPEC lineages (42–44), they are more commonly associated with enteropathogenic E. coli (EPEC) (45). The combination of both ExPEC and EPEC virulence factors might contribute to unique mechanisms of host cell manipulation and immune modulation.

Metabolic modeling highlighted the presence of a large shared metabolic core (53% of all reactions were consistent across strains). The lack of strict concordance between accessory metabolic functions and phylogenetic clustering implies that metabolic traits are fluid and may have been secondarily acquired or lost during the process of environmental adaptation.

Taken together, this study underscores the genomic, metabolic, and virulence versatility of E. coli strains associated with DFUs. The high prevalence of ExPEC-associated virulence traits, widespread AMR, and broad metabolic capabilities in DFEC contribute to explaining how E. coli persists in chronic wounds and why DFIs are frequently associated with failure of antimicrobial therapies. Our findings offer valuable insights for both clinical management and epidemiological surveillance. Specifically, the characterization of AMR profiles across isolates informs empirical treatment decisions, particularly in poor-resource settings where inappropriate use of antibiotics is frequent and antimicrobial susceptibility testing is not readily available. The identification of phylogenetic clusters with wider representation across multiple collection sites suggests the presence of dominant lineages that may warrant targeted monitoring in clinical environments. These findings also emphasize the importance of integrating genomic surveillance into routine diagnostics workflows and public health frameworks to enable real-time tracking of resistance trends and the dissemination of virulence determinants.

Future research should prioritize the functional validation of key virulence factors to elucidate their precise roles in DFEC pathogenesis. Longitudinal studies in patients or relevant animal models will also be instrumental in identifying the genetic and phenotypic changes that drive pathoadaptation over time (46–48). Importantly, future work should also aim to overcome current limitations in clinical sampling. This includes the collection of deep tissue biopsies, microscopy observation of clinical samples, and access to detailed patient metadata to distinguish colonizing flora from invasive pathogens. These approaches will be critical for validating the pathogenic role of DFEC isolates and for contextualizing genomic predictions with phenotypic evidence. Additionally, exploration of host-microbe and microbe-microbe interactions within the polymicrobial communities of DFUs is essential. Such studies could reveal how DFEC strains exploit or compete within this environment. Ultimately, a more comprehensive understanding of DFEC’s adaptive strategies will be vital for the development of targeted and effective interventions and therapies aimed at improving outcomes for diabetic patients suffering from chronic foot wounds.

MATERIALS AND METHODS

Sample collection

Strains were isolated from surface swabs of infected DFUs. Samples were collected by expert healthcare practitioners from anonymized patients at 3 different locations: Kumasi Hospital (Ghana, 8 strains, collected between 2011 and 2014; a kind gift from Dr. Patrick Kimmitt), the GSTT Synnovis Center at St Thomas’ Hospital (UK, 9 strains, collected between 2023 and 2024), the diabetes clinic of Benue State University Teaching Hospital (Nigeria, 14 strains, collected in 2024). Collection typically occurred during debridement procedures or as a result of diagnostic pipelines aimed at pathogen identification. A single isolate was selected from each sample and subjected to preliminary species confirmation using either matrix-assisted laser desorption/ionization time-of-flight mass spectrometry, molecular testing via multilocus PCR (49), or conventional biochemical methods (API 20E test, bioMérieux). All isolates were ultimately confirmed as Escherichia coli (or Shigella, which is genetically indistinguishable from E. coli) based on WGS.

Preparation of bacteria for WGS

All DFEC strains were prepared and outsourced to MicrobesNG (Birmingham, UK, https://microbesng.com/) for DNA sequencing. For each strain, 2 mL of Tryptic Soy Broth was inoculated with a single colony of DFEC strains. Bacteria were grown overnight at 37°C with agitation. The following day, OD_600_ was measured, and a volume corresponding to 6 × 10⁹ cells per sample was spun in a centrifuge at 4,000 rpm for 5 minutes at room temperature. Pellets were washed in 1 mL sterile Milli-Q water, centrifuged again, resuspended in 500 µL of inactivation buffer (DNA/RNA Shield, Zymo Research, provided by MicrobesNG), and outsourced for DNA extraction, quality control, and WGS.

Genome sequencing and assembly

DFEC genome sequences were obtained via Illumina short reads for the Ghanaian strains and Oxford Nanopore long reads for the UK and Nigerian strains. Summaries of the strain details are reported in Table 1 and Table S1. Short-read sequencing libraries were constructed using Illumina Nextera XT or equivalent kits according to MicrobesNG’s standardized protocols. Sequencing was performed on the Illumina platform, generating 2 × 250 bp paired-end reads, targeting a minimum genome coverage of 30× per sample to ensure high-quality assemblies. Raw sequencing data were provided as paired FASTQ files. Assemblies were generated in SPAdes (50) and delivered in FASTA format for downstream analyses. For long-read sequencing, libraries were generated using the rapid sequencing DNA v.14 – barcoding SQK-RBK114.96 (Oxford Nanopore Technologies). Sequencing was performed on a GridION using an R10.4.1 flow cell (Oxford Nanopore Technologies). Raw nanopore reads were basecalled with Guppy (51). Reads were randomly subsampled to 50× coverage using Rasusa (v.0.7.1), and assembled using Flye (v.2.9.2-b1786) (52). Assemblies were polished using Medaka (v.1.8.0) (https://github.com/nanoporetech/medaka) to improve consensus sequence accuracy. Taxonomic classification of the reads and assemblies was performed using Kraken (v.2.0.9) (53), and 16S identification was performed using Barrnap and SINA (54, 55). Also, in this case, the assemblies in FASTA format were utilized for all downstream analyses.

Use of publicly available sequences

Existing genomic sequences of E. coli strains from diabetic foot were identified from publicly available databases. The pathogen detection database (https://www.ncbi.nlm.nih.gov/pathogens/) was searched for strains meeting the following criteria: Organism group “E. coli and Shigella,” Host “Homo sapiens,” and Host disease “Diabetic foot ulcer” or “Diabetic Foot” or “Diabetic foot” or “Diabetic foot infection” or “Foot infection” or “Diabetes mellitus type 2 with foot-gangrene.”

Non-redundant isolates were identified in the Bacterial and Viral Bioinformatics Resource Center (BV-BRC) database (https://www.bv-brc.org/), filtering for Species “Escherichia coli,” Genome quality “Good,” Host name “Homo sapiens,” and Isolation Source Isolates “bone infection of foot.”

Overall, 11 strains were identified. Isolates were collected from Sweden (4 strains), Malaysia (2 strains), China, Brazil, South Korea, India, and the USA (1 strain each). The sequences of the isolates were downloaded in FASTA format and utilized for all downstream analyses alongside the sequences of the newly collected samples. The accession numbers of these sequences are reported in Table S1. For these strains, the accession number prefixed with the “DFI_country” identifier was adopted as the strain name.

Phylogenetic reconstruction

A phylogenetic tree was obtained using the dedicated tree reconstruction tool made available by BV-BRC (v.3.54.6) (56). The tree was generated using a maximum likelihood approach, and the parameters for the tree construction were set as follows: Taxonomy = Escherichia, maximum allowed deletions and duplications = 1, number of genes considered = 1,000. The BV-BRC phylogenetic tree tool uses the RAxML algorithm with the rapid bootstrapping option, performing 100 bootstrap replicates (57). 2 representative strains of phylogroups A (strain B, MG1655), B1 (SE11, E24377A), B2 (E234849, CFT073), C (ECC325, F1072_12), D (E42, UMN026), F (SMS-3-5, IAI39), E (Sakai, CB9615), G (921A, H299), and Shigella lineages (Sd197, SWHEFF_49, Sb227, CDC_3083-94, 53G, SS046, 2457T, M90T) were included alongside the DFI E. coli strain sequences to confirm their expected distribution and clustering with novel strains according to their genotype. Phylogroup predictions were also confirmed using ClermonTyping (v.23.06) (http://clermontyping.iame-research.center). The tree also included E. ruysia (S1-IND-07-A, C61-1), E. marmotae (W49-2, H1-003-0086-C-F), E. albertii (KF1, 2011C-4180), and E. fergusonii (30038, EF21023765) strains. The FigTree (v.1.4.4) software (http://tree.bio.ed.ac.uk/software/figtree/) was used to render and annotate the phylogenetic tree (58).

Serotyping and multilocus sequence typing

O and H serotype predictions were performed using SerotypeFinder (v2.0) (https://cge.food.dtu.dk/services/SerotypeFinder/), provided by the Center for Genomic Epidemiology (59). Assembled genome FASTA files were uploaded, and the tool was run using default parameters, with an identity threshold of 90% and a minimum gene coverage of 60%. The tool predicted the most likely O and H antigen combinations for each strain, based on the presence of serotype-specific genes, as previously described (60). The O serotype of 5 isolates could not be typed using this method, and therefore, these strains were classified as non-typeable. K type predictions were performed using Kaptive (v.3) (https://kaptive.readthedocs.io/en/latest/) (61) and an ad hoc database for E. coli K types (https://github.com/rgladstone/EC-K-typing) (61, 62).

MLST was conducted using Pathogenwatch (v.23.5.0) (https://pathogen.watch/). Uploaded FASTA files were automatically analyzed by the tool, and sequence types were assigned based on the widely used Achtman and Pasteur MLST schemes (63).

Growth on MacConkey agar

To test for lactose fermentation, strains were grown on MacConkey agar plates overnight at 37°C.

S. sonnei 53G strain (47, 64) was used as a lactose non-fermenter control.

Pangenome analysis

Assembled genomic sequences of DFEC strains were uploaded to the Galaxy platform (https://usegalaxy.org/) for pangenome analysis (65). Genomes were first annotated using Prokka (v.1.14.6) (66), which generated annotated assemblies in GFF3 format. The resulting GFF3 files were then processed using Roary (v.3.13.0), a rapid large-scale pangenome pipeline (67), also run via Galaxy. Roary was used to compare annotated genomes, generate a core gene alignment, and assess gene presence and absence across isolates. Both Prokka and Roary were kept at default settings, except that “GenBank compliance” was forced in Prokka and the non-default option “Improve gene predictions for highly fragmented genomes” was selected. Genes were classified based on their frequency across the data set as follows: core genes, present in more than 99.0% of isolates; soft core genes, present in 95.0%–98.9% of isolates; shell genes, found in 15.0%–94.9% of isolates; cloud genes, present in fewer than 15.0% of isolates. All the genes identified in the pangenome of DFEC strains are reported in the data supplements (Table S2).

For visualization, Roary output files—including the summary statistics and gene presence/absence matrix—were downloaded and further processed in R Studio (R version 4.5.0) (68). A heatmap displaying gene presence and absence across isolates was generated using the ComplexHeatmap package (v.2.24.1) (69). Strains were clustered using a heuristic clustering method, while genes were ordered according to their frequency across the data set. Pangenome and core genome accumulation curves were computed in R Studio by randomly permuting the genome orders 100 times.

Prediction of colicins, plasmids, AMR, and virulence factors

Colicins, plasmids, AMR determinants, and virulence factors were predicted using ABRicate (v.1.0.1) via the Galaxy platform (70). Assembled genome FASTA files were analyzed using default parameters (minimum DNA % identity and % coverage both set at 80%). For colicins and plasmid detection, the PlasmidFinder database (v.2.1) was used. For AMR gene detection, the ResFinder database (v.3.2) was employed. For the detection of virulence factors, the dedicated E. coli virulence factor database Ecoli_VF (v.2019) was used. For ease of representation, colicins and plasmids were grouped according to the different types and Inc types. AMR genes were classified according to the antibiotic classes they confer resistance to, including β-lactams, aminoglycosides, tetracyclines, sulfonamides, fluoroquinolones, and others. Exact genotypic AMR determinants are reported in the supplementary data (Table S4). For visualization of virulence genes, selected genes commonly identified in ExPEC isolates were displayed, grouped into categories according to functions (such as adherence, iron acquisition systems, serum resistance, toxins, and secretion systems). All virulence factors identified are provided in the supplementary data (Table S6). For the prediction of virulence genes encoded by plasmids, PlasmidHunter (v.1.2) was used (71). Briefly, the tool allowed for the classification of contigs as plasmid-derived or chromosome-derived. These coordinates were cross-referenced to the ABRicate results to infer whether a virulence factor was encoded on a plasmid-derived or chromosome-derived contig.

Genome-scale metabolic modeling

Draft genome-scale metabolic models were generated for each isolate using CarveMe (v.1.5.1) (72, 73). Following model reconstruction, the presence/absence of individual metabolic reactions was retrieved for all isolates using the script reaction_presence_absence_generate.py (https://github.com/bananabenana/Metabolic_modelling_scripts/tree/main/reaction_pres_abs). All the metabolic reactions identified in the DFEC strains are reported in the data supplements (Table S7).