Genomic characterization of ESBL-producing Escherichia coli isolates from patients in an orthopaedic ward in Mwanza, Tanzania

Study setting and design

This study was nested within a larger longitudinal investigation conducted over a 4-month period, from 3 January 2020 to 30 May 2020, assessing gastrointestinal colonization with ESBL-producing Enterobacteriaceae among orthopaedic patients admitted to a tertiary hospital in Mwanza, Tanzania.³⁰ The parent study included broader sampling activities within the ward. For this study, we analyzed ESBL-confirmed E. coli isolates recovered from stool or rectal swabs collected from patients admitted to the orthopaedic ward. By integrating whole-genome sequencing data with available clinical and ward-level metadata, the study aimed to investigate the population structure of ESBL-producing E. coli, define their AMR and associated genomic features, and assess patterns of clustering consistent with the circulation of multidrug-resistant lineages within the orthopaedic ward.

Bacterial isolation and ESBL confirmation

Clinical stool and rectal swab samples were processed using standard microbiological procedures at the Department of Microbiology and Immunology, Weill Bugando School of Medicine, Catholic University of Health and Allied Sciences, Mwanza, Tanzania. Samples were inoculated onto MacConkey agar (Oxoid, UK) and CHROMagar™ ESBL (CHROMagar, France), a selective medium containing cephalosporins for the isolation of ESBL-producing Gram-negative bacteria. Plates were incubated aerobically at 35–37 °C for 18–24 hours. Colonies with morphology consistent with E. coli were selected and subcultured to obtain pure isolates. ESBL production was then confirmed phenotypically using the combination disk method, comparing inhibition zones for cefotaxime and ceftazidime, alone and in combination with clavulanic acid, in accordance with CLSI recommendations for phenotypic ESBL confirmation.³¹ Only isolates confirmed as ESBL-producing E. coli were included in downstream genomic analyses.

DNA extraction, library preparation, and sequencing

Pure E. coli isolates were shipped to the Genomics Laboratory, Department of Immunology and Molecular Biology, College of Health Sciences, Makerere University, Kampala, Uganda, where genomic DNA was extracted using standardized protocols. DNA concentration was measured using the Qubit dsDNA High Sensitivity Assay (Thermo Fisher Scientific, USA), and DNA purity was assessed using a NanoDrop spectrophotometer (Thermo Fisher Scientific, USA) by evaluating A260/280 and A260/230 ratios. DNA integrity was further assessed by 1.0% agarose gel electrophoresis. Whole-genome sequencing was performed at the Earlham Institute (Norwich, UK) using the Illumina NovaSeq 6000 platform. Sequencing libraries were prepared using the LITE (Low Input Transposase-Enabled) protocol, and sequencing was carried out using paired-end 150-bp chemistry (2 × 150 bp reads).

Read processing and genome assembly

Raw sequencing reads were subjected to quality control using FastQC (v0.12.1),³² followed by adapter trimming and removal of low-quality bases using Trimmomatic (v0.40),³³ with default parameters. High-quality reads were retained for downstream analyses. De novo genome assemblies were generated using SKESA (Strategic K-mer Extension for Scrupulous Assemblies) assembler (v2.4.0),³⁴ with default settings and assembly quality was assessed using standard metrics, including genome size, contiguity, and N50 values, to ensure suitability for comparative genomic analysis.

Genomic species confirmation

Genomic species confirmation was performed using a multi-tool approach to ensure robust and accurate taxonomic classification. Draft genome assemblies were analyzed using GTDB-Tk (v2.6.0)³⁵ (Genome Taxonomy Database Toolkit), which assigns taxonomy based on genome-wide phylogenetic placement against the GTDB reference database.

To complement this, GAMBIT v1.1.0 (Genomic Approximation Method for Bacterial Identification and Tracking)³⁶ was used to provide k-mer–based taxonomic classification, enabling rapid and high-resolution identification of bacterial genomes. In addition, Basic Local Alignment Search Tool for nucleotides (BLASTn) searches were performed against the NCBI nucleotide database to confirm species identity based on sequence similarity to well-characterized reference genomes. Blast analyses were conducted using the Docker image gmboowa/blast-analysis:1.9.4.

Concordance across these three independent approaches (GTDB-Tk, GAMBIT, and BLAST) was used to validate species assignments, ensuring all isolates were accurately classified as E. coli prior to downstream comparative genomic analyses.

Genome characterization and typing

Sequence types (STs) were assigned from draft genome assemblies using the MLST module in rMAP-2.0, implemented with the mlst tool (Torsten Seemann; Docker image staphb/mlst:2.19.0). This approach scans assembled contigs against PubMLST-curated schemes to determine the sequence type and allele profile for each isolate, enabling classification into globally recognized E. coli lineages.³⁷

Read mapping and variant calling

Quality-controlled reads were mapped to the clinical E. coli reference genome GCF_000285655.3_EC958.v1 using the Snippy pipeline (v4.6.0),³⁸ which integrates BWA-MEM for read alignment and bcftools for variant calling. This clinically relevant reference was selected to improve mapping accuracy and reduce reference bias in the analysis of hospital-associated isolates. Core genome single nucleotide polymorphisms (SNPs) were identified across all isolates, and a multiple sequence alignment of core SNPs was generated for downstream phylogenetic analysis.

Recombination filtering and phylogenetic reconstruction

To account for homologous recombination, SNP alignments were processed using Gubbins v3.4.1 (Genealogies Unbiased By recomBinations In Nucleotide Sequences),³⁹ which identifies and masks regions of elevated SNP density consistent with recombination. The resulting recombination-filtered alignment represents clonal variation across the dataset. Maximum-likelihood phylogenetic trees were inferred using IQ-TREE v3.1.1,⁴⁰ based on the recombination-filtered core genome SNP alignment. Branch support was assessed using standard bootstrap approaches. Pairwise SNP distances between isolates were calculated using snp-dists, providing a quantitative measure of genomic relatedness. The final phylogenetic tree was visualized in iTOL v7.0 (Interactive Tree of Life),⁴¹ where isolate metadata were incorporated to annotate the tree and facilitate interpretation of clustering patterns, epidemiological relationships, and potential transmission events.

Transmission network analysis

To investigate potential transmission events among ESBL-producing E. coli isolates, pairwise SNP distances were calculated from the recombination-filtered core genome alignment generated using Snippy and processed with Gubbins. SNP distances were computed using snp-dists, providing a matrix of pairwise genomic distances among the 39 clinical isolates and the included reference genome. A transmission-linked network was constructed using a predefined SNP threshold of ≤5 SNPs, consistent with thresholds commonly applied to infer recent transmission or shared-source acquisition in bacterial genomic epidemiology. Isolates and the reference genome were represented as nodes, and edges were drawn between nodes with pairwise SNP distances at or below this threshold. Edge labels indicate the number of SNP differences between connected nodes. The network was generated using Python (version 3.14) with the NetworkX library for graph construction and Matplotlib for visualization. A force-directed layout algorithm (spring layout) was applied to position nodes based on their connectivity, enabling clear visualization of clusters of closely related isolates and facilitating identification of putative transmission events within the orthopaedic ward.

Antimicrobial resistance, plasmid, and virulence analysis

Antimicrobial resistance genes were identified using the ResFinder v4.5.0 database,⁴² enabling classification of resistance determinants into antimicrobial classes. MDR was defined as the presence of resistance determinants spanning three or more antimicrobial classes.⁴³ Plasmid replicons were identified within rMAP-2.0 using PlasmidFinder⁴⁴ to screen assembled genomes for known plasmids, while virulence-associated genes were detected using ABRicate,⁴⁵ implemented through the Docker image staphb/abricate:1.0.0 against curated E. coli virulence factor databases. These analyses were executed through the containerized rMAP-2.0 workflow, ensuring reproducibility, portability, and consistent processing across all samples.³⁷

Data integration and comparative analysis

Genomic data were integrated with available epidemiological and clinical metadata, including non-identifiable clinical and epidemiological metadata. This enabled an integrated analysis of population structure, antimicrobial resistance burden, plasmid content, and virulence-associated profiles. Phylogenetic clustering was interpreted in the context of ward-level and patient-level metadata to assess patterns consistent with the circulation and local persistence of closely related multidrug-resistant lineages within the orthopaedic ward.

Sample characteristics

A total of 45 ESBL-producing clinical E. coli isolates were initially collected from patient-derived stool and rectal swab samples obtained from individuals admitted to the orthopaedic ward. Of these, 39 isolates passed quality control and were included in the whole-genome sequencing and downstream analyses ( Table 1). Core genome phylogenetic analysis revealed multiple distinct clusters, indicating genomic diversity among isolates circulating within the ward. Several isolates formed tight phylogenetic clusters, suggestive of potential shared sources or localized transmission events, while others appeared more genetically distinct. Notably, isolate A55728 formed a long independent branch, indicating substantial divergence from other isolates. The distribution of isolates across phylogenetic clusters, together with metadata on sample source and collection site, provided a framework for investigating potential transmission dynamics between patients and the hospital environment. Detailed isolate-level characteristics are presented in Extended data 1 (Table S1).

Genomic features

Whole-genome sequencing of 39 ESBL-producing E. coli isolates generated high-quality draft assemblies suitable for downstream comparative genomic analysis. Species identity was genomically confirmed using a combination of GTDB-Tk, GAMBIT, Extended data 1 (Table S1) and BLAST-based approaches (https://gmboowa.github.io/rMAP-2.0/reports/esbl_ecoli.html#blast), ensuring accurate taxonomic assignment and minimizing misclassification within the E. coli species complex.

The assemblies had a median genome size of approximately 5.1 Mb, consistent with typical E. coli genomes. Assembly fragmentation varied across isolates, with a median of 90 contigs (range: 50–3339), indicating overall good assembly quality, although a small number of genomes were highly fragmented. The GC content was highly conserved across isolates, with a median of approximately 50.6%, reflecting the expected genomic composition of E. coli.

Sequence typing revealed a genetically diverse population comprising several globally recognized high-risk and community-associated lineages, including ST131, ST1193, ST69, ST617, and ST38, alongside additional less frequent sequence types. This diversity indicates the co-circulation of internationally disseminated high-risk clones and locally established lineages within the orthopaedic ward population.

Core genome SNP analysis using Snippy identified substantial genomic diversity across the isolates. The reference-based alignment revealed a wide range of variant sites, highlighting the presence of both closely related and highly divergent isolates within the same clinical setting. The distribution of SNPs per isolate demonstrated marked heterogeneity, with some isolates showing minimal divergence consistent with recent transmission or shared ancestry, while others exhibited extensive polymorphism, reflecting the presence of multiple independent lineages.

Antimicrobial resistance, plasmids, and virulence determinants

All 39 isolates were classified as MDR, carrying resistance determinants to three or more antimicrobial classes. ESBL production was predominantly associated with bla_CTX-M-15, which was detected in most isolates and is consistent with its recognized role as a globally disseminated ESBL determinant.

Beyond β-lactam resistance, the isolates harbored genes associated with resistance to several additional antimicrobial classes, including aminoglycosides, fluoroquinolones, sulfonamides, trimethoprim, and tetracyclines. This broad resistome highlights the accumulation of multiple resistance determinants within individual isolates and reflects the complex AMR burden present in the orthopaedic ward setting.

Plasmid analysis showed a predominance of IncF-family replicons, which are frequently implicated in the dissemination of ESBL genes in E. coli. Additional plasmid types, including IncY, were also identified in a subset of isolates, indicating the involvement of diverse mobile genetic elements in the acquisition and spread of resistance determinants.

Virulence-associated genes were also widely distributed across the isolate collection, including factors linked to adhesion, iron acquisition, and extraintestinal survival, such as fdeC and ybtP/ybtQ. The co-occurrence of antimicrobial resistance and virulence-associated determinants underscores the clinical relevance of these isolates and raises concern about the circulation of lineages with both enhanced pathogenic potential and limited treatment options.

Pairwise SNP distances and transmission inference

Pairwise SNP distances were calculated from the recombination-filtered core genome alignment to assess genomic relatedness among E. coli clinical isolates. SNP distances ranged widely across the dataset, from 0 to 20,176 SNPs, with a mean distance of approximately 2,506 SNPs, indicating substantial genetic diversity within the orthopaedic ward population. Despite this overall diversity, a subset of isolates exhibited very low SNP distances (0–5 SNPs), consistent with recent transmission events or a shared source within the hospital environment.⁴⁶ In contrast, the majority of isolate pairs showed large SNP distances (>1000 SNPs), suggesting the presence of genetically unrelated strains and multiple independent introductions into the hospital ward.

Identification of genetically indistinguishable isolates (0 SNP cluster)

A cluster of isolates with zero SNP differences was identified, comprising isolates A55724, A55793, and A55798. These isolates were genetically indistinguishable at the core genome level ( Table 2), strongly indicating a recent transmission event or a common source. Such zero-distance clustering provides high-confidence evidence of direct or near-direct transmission within the clinical setting.

Detection of closely related transmission clusters (≤3 SNPs)

A second cluster of closely related isolates was identified, with pairwise SNP distances ranging from 0 to 3 SNPs. This cluster included isolates A55769, A55939, A55941, and A55966 ( Table 3). The low level of genomic variation within this group is consistent with a recent transmission chain or ongoing circulation of a closely related lineage within the ward.

Interpretation of transmission dynamics

Together, these findings indicate a mixed epidemiological scenario within the orthopaedic ward. The presence of both genetically indistinguishable isolates and closely related clusters supports the occurrence of localized transmission events, while the broader distribution of high SNP distances across the dataset reflects a diverse background population structure, likely driven by multiple introductions of unrelated E. coli lineages into the hospital environment. These results demonstrate the coexistence of recent transmission clusters and genetically diverse strains, underscoring the importance of integrating genomic data with epidemiological context to better understand transmission pathways and inform infection prevention strategies.

Two distinct transmission clusters were identified based on pairwise SNP distances of ≤5 SNPs. The first cluster, comprising isolates A55724, A55793, and A55798, showed 0 SNP differences, indicating genetically indistinguishable isolates and providing strong evidence of recent transmission or a common source. The second cluster, comprising A55769, A55939, A55941, and A55966, showed pairwise SNP distances ranging from 0 to 3 SNPs, consistent with a very closely related group and suggestive of an ongoing transmission chain within the ward.

Shared variant analysis

Analysis of shared SNPs identified groups of isolates with common variant profiles, further supporting the phylogenetic structure observed in the core genome SNP analysis. Isolates that were closely related in the phylogeny shared a higher proportion of SNPs, forming distinct genomic clusters consistent with possible transmission chains or recent common ancestry. In contrast, isolates with few or no shared variants were more genetically distant, supporting the presence of multiple unrelated lineages within the study population.

Phylogenetic relationships

Core genome phylogenetic analysis based on the recombination-filtered SNP alignment demonstrated a clearly polyclonal population structure among the 39 ESBL-producing E. coli isolates ( Figure 2). The isolates were distributed across multiple distinct lineages, with eight phylogenetic clusters identified alongside one genetically distinct outlier, rather than forming a single outbreak clone. This overall structure indicates substantial genomic diversity within the orthopaedic ward population.

Despite this broad diversity, several isolates formed tight phylogenetic clusters characterized by short branch lengths, consistent with the low pairwise SNP distances observed and suggestive of recent transmission or shared-source exposure within the ward ( Figure 2). In contrast, a subset of isolates displayed long branches, indicating marked genetic divergence and supporting the presence of unrelated strains likely introduced independently into the hospital setting. Notably, isolate A55728 formed a long independent branch and remained clearly separated from the main clusters, consistent with the high SNP distances observed and arguing against its involvement in recent local transmission events.

Taken together, the recombination-filtered phylogeny, pairwise SNP distance analysis, and shared variant patterns support a mixed epidemiological scenario in which localized transmission occurs within a background of multiple independent introductions. The coexistence of closely related isolate clusters and more widely separated lineages suggests that ESBL-producing E. coli in this orthopaedic ward are shaped by both local clonal expansion and the repeated introduction of diverse MDR strains. High-risk sequence types, including ST131 and ST1193, were represented across this population, further underscoring their contribution to the dissemination of ESBL-producing E. coli in this setting.

Comparative analysis

Comparative genomic analysis revealed both heterogeneity and shared features across the ESBL-producing E. coli population. Although the isolates were distributed across multiple sequence types and phylogenetic clusters, many shared a common repertoire of antimicrobial resistance determinants, most notably bla_CTX-M-15, indicating the widespread distribution of key ESBL-associated resistance genes across genetically distinct lineages.

Comparison of plasmid content further showed that diverse sequence types frequently carried similar IncF-family replicons, supporting a common role for these plasmids in the dissemination of resistance determinants. At the same time, variation in accessory resistance and virulence-associated genes across isolates highlighted the presence of lineage-specific genomic features superimposed on a shared multidrug-resistant background.

Taken together, these findings suggest that the ESBL-producing E. coli population in the orthopaedic ward is shaped by both the persistence of successful lineages and the circulation of shared mobile genetic elements, resulting in isolates that are phylogenetically diverse but convergent in their resistance profiles.