Foodborne sources of antimicrobial-resistant Escherichia coli in community-acquired urinary tract infections in Bangladesh

Abstract

Background

Extended-spectrum β-lactamase-producing Escherichia coli (ESBL-Ec) causing urinary tract infections (UTI) are presumed to originate from poultry, but evidence from resource-limited settings is scarce.

Methods

We conducted a prospective surveillance of UTI patients alongside food sampling within the same temporospatial context in Dhaka, Bangladesh (2016-2018). A total of 2297 food samples, including 970 poultry meat, 504 fresh produce, 334 fish, 327 eggs, and 162 beef were collected from 24 wet markets and tested for E. coli and ESBL-Ec. During the same period, 110 ESBL-Ec isolates were obtained from 394 UTI patients residing near these markets. Phylogroups of ESBL-Ec from both sources were determined, and representative isolates were subjected to whole-genome sequencing. Additionally, seven-day dietary recall interviews were conducted with 100 ESBL-Ec UTI patients and 200 age- and sex-matched controls.

Results

Here we show that 87% (n = 1993) of food samples are positive for E. coli while 30% (n = 601) contain ESBL-Ec, predominantly in poultry meat (43%), followed by beef (36%), fish (24%), eggs (12%), and fresh produce (6%). Urinary ESBL-Ec isolates are dominated by phylogroup B2 (63%) and ST131 (56%), both absent in food. Poultry isolates show higher diversity compared to urinary isolates. ESBL-Ec ST648, ST38, ST10, ST457 occur in both sources, but core genome SNP analysis shows no clonal relatedness ( > 22 SNPs). Poultry consumption is not associated with ESBL-Ec UTI, although a seven-day recall may not capture prior colonization events.

Conclusion

Overall, UTI-causing ESBL-Ec are genetically distinct from poultry isolates, suggesting that poultry is unlikely to be a major source of UTI in this setting.

Plain language summary

Urinary tract infections (UTI) are frequently caused by a strain of the bacterium Escherichia coli which produce an antibiotic resistance molecule, making many antibiotics ineffective. Although food animals are suspected sources of these resistant bacteria, evidence from low-income settings is limited. We investigated whether food contributes to UTI-causing antibiotic-resistant E. coli in urban Bangladesh. We compared these E. coli isolated from UTI patients to E. coli from commonly consumed foods collected within the same place and time. Genomic analysis revealed no genetic relatedness between food and clinical isolates. Moreover, dietary consumption of specific foods was not associated with infection risk. Together, these findings indicate that foodborne transmission is unlikely to be a major contributor to UTI-causing antibiotic-resistant E. coli in this region, highlighting the need to explore alternative transmission pathways.

Amin et al. conduct surveillance in Bangladesh to investigate potential foodborne sources of UTI-causing, ESBL-producing E. coli. Whole genome sequencing E. coli from UTI and poultry meat shows no clonal overlap, suggesting poultry consumption is unlikely a major source of UTI-causing E. coli and highlighting the need to explore other reservoirs.

Introduction

Urinary tract infections (UTI) are common extraintestinal infections in both community and hospital settings globally. UTI cases increased by about 66% from 1990 to 2021, with an estimated 553 million cases, and 6.8 million disability-adjusted life years reported worldwide in 2021. Alarmingly, the number of deaths attributed to UTI reached 300,112 (95% UI: 267,926–324,216), globally representing a 190% increase from 1990 to 2021¹. Women are more prone to UTI than men, largely due to anatomical and hormonal differences, with nearly 50% experiencing a UTI during their lifetime². Escherichia coli (E. coli) is the leading cause of UTI, responsible for 75–95% of all uncomplicated cystitis and pyelonephritis cases in the United States³. In Bangladesh, E. coli was reported as the predominant pathogen accounting for 52% of 920 community-acquired UTI cases⁴. The bacterial strains responsible for these infections, known as extraintestinal pathogenic E. coli, are genetically and functionally distinct from both diarrhoeagenic and the non-pathogenic commensal E. coli^5,6. Despite differences, these E. coli strains can colonize the human gastrointestinal tract like other E. coli, with fecal-oral transmission considered the primary route. The global rise of antibiotic resistance has further complicated the clinical management of UTI, posing significant public health challenges^7–12.

The sources of multi-drug resistant (MDR) E. coli, particularly the extended-spectrum β-lactamase-producing E. coli (ESBL-Ec), causing UTI, remain a topic of considerable interest and ongoing debate among public health researchers. Previous studies have demonstrated that UTI-causing E. coli are genetically closely related to, and in some cases indistinguishable from, pathogenic E. coli found in poultry, also known as avian pathogenic E. coli^13–15. This genetic similarity supports the hypothesis that antibiotic-resistant E. coli are selectively enriched in poultry raised with antibiotics used for prophylaxis and growth promotion, and subsequently transferred to humans through contaminated meat. Nonetheless, establishing evidence of foodborne transmission for these bacterial strains remains challenging. These E. coli strains can colonize the human gut like commensals, and the undefined interval between colonization and the onset of extraintestinal infection further complicates efforts to establish a direct epidemiological link between foodborne sources and UTI-causing strains. Many studies have investigated the potential foodborne sources of UTI-causing E. coli using molecular and epidemiological approaches by comparing E. coli from UTI patients and poultry meats^16,17. However, a key limitation of the majority of these studies is the lack of a clearly defined temporal and spatial sampling framework, which is a prerequisite for establishing the epidemiological connection between human and poultry-derived isolates. Moreover, these investigations have largely been conducted in high-income countries, where stringent food safety regulations result in a significantly lower prevalence of ESBL-Ec in the food supply. In contrast, no such studies have been conducted in resource-limited settings, where the burden of ESBL-Ec is significantly higher across environmental, poultry, and human sources, including community-acquired UTI cases^4,18,19.

To address this critical knowledge gap, we conducted a prospective surveillance study at the icddr,b (formerly known as International Centre for Diarrhoeal Disease Research, Bangladesh) diagnostic clinic in Dhaka, Bangladesh, enrolling patients with suspected UTI who had no recent hospital exposure within the past 24–48 h. Simultaneously, we collected a large number of food samples from retail markets located within a 10-mile radius of the clinic during the same period to capture both exposure and infection with the same ESBL-Ec clone. We focused on detecting ESBL-Ec from both clinical and food sources to perform comparative genomic analyses and explore possible transmission pathways. Additionally, we conducted patient interviews to assess the association between dietary habits, particularly poultry meat consumption and the risk of developing UTI caused by ESBL-Ec.

This study demonstrates that ESBL-Ec are highly prevalent in both poultry sources and UTI patients in Bangladesh. Whole-genome sequencing shows no clonal relationship between poultry and urinary isolates, and poultry consumption is not associated with UTI caused by ESBL-Ec.

Methods

Ethics approval declaration

All participants provided written informed consent before taking part in the study. Written informed consent was obtained from the parents/guardians of the minors (<16 years of age) included in this study. The research team explained the background and objectives of the study clearly to the participants. This study was approved by the Institutional Review Board of icddr,b comprising the Research Review Committee (RRC) and the Ethics Review Committee (ERC).

Urine sample collection, processing, and E. coli identification

During September 2016 to November 2018, study clinicians at icddr,b diagnostic center enrolled 4500 patients with suspected UTI out of 5675 who attended voluntarily or were referred by physicians across Dhaka. Enrollment occurred on working days between 9 AM and 4 PM. Urine sample collection and processing were described previously⁴. Briefly, a clean-catch midstream urine was collected in a sterile container. A loopful ( ~ 10 µL) of urine was cultured on MacConkey agar (Oxoid Ltd., UK) and blood agar (Oxoid Ltd., UK) within 1 h of collection, and incubated aerobically at 35 °C for 18–24 h, with an additional 24 h for negative cases before reporting. Out of 4500 urine samples, 920 (29%) showed significant growth of a single bacterial species (≥ 1.0 × 10⁵ colony forming units [CFU]/mL of urine), defined as UTI²⁰. Presumptive E. coli were confirmed by colony characteristics, Gram staining, biochemical properties, and VITEK-2 system (BioMérieux Inc., France)⁴. E. coli ATCC 25922 was used as a reference strain in biochemical identification.

Food sample collection, processing, and E. coli identification

From December 2017 to December 2018, we collected 2297 food samples, including poultry meat (n = 970), fresh produce (n = 504), raw fish (n = 334), egg (n = 327), and raw beef (n = 162) from 24 retail markets in Dhaka city (Fig. 1). We collected these samples from each of these markets monthly over a period of 12 months. Poultry meats included broiler/layer chicken (n = 250), free range chicken (n = 250), duck (n = 236), and pigeon (n = 234). Fish samples comprised two commonly consumed fish types, such as pangas (n = 168) and tilapia (n = 166). Various types of farmed and backyard chicken eggs (n = 327) were also collected. Fresh produce comprised spinach (n = 168), carrot (n = 168), and potato (n = 168), representing both leafy greens and root vegetables. All samples were transported to the icddr,b laboratory within 2 h of collection under cold chain (4–8 °C) and processed the same day following ISO (International Organization for Standardization) procedures¹⁹. Briefly, a total of 25 g of food samples (except eggs) was homogenized in 225 mL peptone water to make an initial suspension using a homogenizer (Stomacher® 400 Circulator, Seward Ltd., UK). For the egg samples, the surfaces of three eggs were rubbed in 100 mL peptone water to prepare a suspension. Ten-fold serial dilutions were made up to 10⁻⁴. Suspension from the last two dilutions was inoculated on Tryptone Bile X-glucuronide (TBX) agar plate (Oxoid, UK). The inoculation was performed using the pour plate technique (1 mL) for non-supplemented TBX agar, and the spread plate technique (0.1 mL) for TBX agar supplemented with cefotaxime (4 µg/L) (Oxoid, UK). After 18–24 h incubation at 37 °C, E. coli colonies were enumerated and expressed as log₁₀ CFU per g of food sample. For eggs, E. coli counts were expressed as log₁₀ CFU/cm² of eggshell surface, where the surface area (S) in cm² was calculated from average egg weight (P) in g using the formula (S = 4.68 P^2/3)²¹. Zero (non-detect) values in the count data were imputed using random numbers between zero and the detection limit (which is 10 CFU), following recommended methods for left-censored data²². The two suspected blue-green E. coli colonies from both TBX plates were confirmed as E. coli by biochemical tests using API 20E.

Fig. 1.

Field sites showing the locations of retail markets and icddr,b diagnostic center in Dhaka city, where samples were collected.

Antibiotic susceptibility testing, detection of ESBL genes, and phylogrouping of E. coli from urine and ESBL-Ec from food samples

E. coli strains from urine and cefotaxime-resistant E. coli from food samples were tested for minimum inhibitory concentration (MIC) against 21 antibiotics using the Sensititre™ GNX2F micro-dilution plate (Thermo Fisher Scientific, MA, USA). Briefly, 3–4 bacterial colonies from overnight cultures on Tryptic Soy agar plates were suspended in sterile saline (0.85% NaCl), and the turbidity was adjusted to a 0.5 McFarland standard (~10⁸ CFU/mL). A 50 µL aliquot was further diluted in 10 mL of Mueller–Hinton broth to obtain a final inoculum of approximately 5 × 10⁵ CFU/mL. Fifty microliters of this suspension were dispensed into each well of the GNX2F plate containing dried antibiotics. The plate was sealed, gently mixed, and incubated at 37 °C for 16–20 h. E. coli ATCC 25922 and Pseudomonas aeruginosa ATCC 27853 were used as quality controls to ensure testing validity. The MIC results were interpreted following Clinical and Laboratory Standards Institute (CLSI) guidelines²³. Results were categorized as sensitive or resistant, and isolates exhibiting resistance to at least one agent in three or more antibiotic classes were classified as MDR²⁴. All cefotaxime-resistant E. coli isolates were examined for ESBL production using the combination disc test (CDT)²³. All confirmed ESBL-Ec were screened for ESBL-encoding genes (bla_CTX-M-group, bla_TEM, and bla_SHV) by conventional PCR²⁵ and assigned to phylogroup following the ClermonTyping scheme²⁶.

Selection of ESBL-Ec from urine and food sources for whole genome sequencing (WGS)

While urine sampling spanned a longer period, comparative genomic analyses focused on isolates collected during overlapping months with food sampling whenever possible. Based on phylogroup distribution, we selected 213 ESBL-Ec for WGS, including 110 from urine samples of UTI patients with unique ID numbers (to avoid recurrent cases) and 103 from poultry meat samples (chicken, duck, and pigeon), which had the highest prevalence of ESBL-Ec. Selection reflected the proportional distribution of phylogroups in both sources. Poultry isolates belonged to phylogroup A (n = 32), B1 (n = 12), C (n = 9), D (n = 25), E (n = 9), and F (n = 16). Urine isolates, chosen to match phylogroups also found in poultry during the same period, belonged to phylogroup A (n = 11), B2 (n = 73), C (n = 1), D (n = 7), E (n = 7), and F (n = 11).

DNA extraction and WGS

We extracted bacterial DNA from the overnight cultures using the Maxwell culture DNA extraction kit and Maxwell automated nucleic acid extraction system (Promega, Madison, USA) following the manufacturer’s instructions. We used a NanoDrop spectrophotometer (Thermo Scientific, USA) and Qubit 2.0 fluorometer (Life Technologies, CA, USA) to measure the purity and concentration of the extracted DNA, respectively. We prepared DNA libraries using the Nextera XT kit and conducted 150 bp paired-end sequencing using the Illumina NextSeq500 platform (Illumina, San Diego, CA, USA). We submitted sequencing files in FASTQ format of all 213 ESBL-Ec isolates from clinical and food sources to NCBI under the BioProject accession PRJNA1300378.

Bioinformatics analysis

We used FastQC (v 0.11.4) to assess the quality of raw sequencing reads (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/). Briefly, we evaluated per-base sequence quality, GC content, sequence length distribution, duplication levels, and adapter contamination. Reads were trimmed and filtered to remove low-quality bases and residual adapter sequences prior to downstream analyses. High-quality reads were de novo assembled using SPAdes (v3.13.0) with default settings²⁷. Genome annotation was carried out using Prokka (v1.14.5)²⁸ and characterization of pangenome, including the core and accessory genomes, was performed using Roary (v3.11.2)²⁹. The resulting gene presence/absence matrix was visualized in R (v4.2.0) using the pheatmap package (https://github.com/raivokolde/pheatmap), allowing cluster-based visualization of gene content variation across isolates. Multilocus sequence typing (MLST) was performed in silico using the mlst tool (https://github.com/tseemann/mlst) based on the Achtman 7-locus scheme, and O:H serotyping was carried out using SerotypeFinder³⁰. To investigate strain-level relatedness, we used Snippy v4.6.0 to align reads to a reference genome and identify core genome single-nucleotide polymorphisms (SNP) (https://github.com/tseemann/snippy). We used snp-dists v0.8.2 to calculate pairwise SNP distances between isolates (https://github.com/tseemann/snp-dists). A maximum-likelihood phylogenetic tree was reconstructed using IQ-TREE v2.0.7 and visualized using iTOL v5^31,32. A threshold of ≤22 core genome SNPs was used to define clonal relatedness, consistent with prior genomic epidemiology studies of ESBL-Ec transmission²⁶.

We used MGEfinder v1.0.6³³ to identify mobile genetic elements (MGE), including insertion sequences, integrons, and transposons. Plasmid replicons and their mobility classification (conjugative, mobilizable, non-mobilizable) were determined using MOB-suite v3.1.0 with mob-recon module^34,35. The chord diagram was constructed in R using the circlize package³⁶. The stacked percent bar plot was created in R using ggplot2³⁷. We detected antimicrobial resistance genes using ABRicate v1.0.1 with ResFinder v4.1 database by setting a cutoff for resistance gene identity and coverage at ≥90%^28,38–40.

Interviews of patients with and without ESBL-Ec infections

We interviewed 300 patients, comprising 100 cases with ESBL-Ec and 200 age (±5 years) and sex matched controls. Among the controls, 100 had non-ESBL-Ec, and 100 had no bacterial growth in their urine cultures. A study physician conducted telephone interviews on food consumption patterns within 7 days of urine culture results using a structured questionnaire. For children and adolescents, parents, mainly mothers, were interviewed. Participants recalled food consumption over the past 3–7 days and described their usual dietary habits, from which consumption frequency was estimated. Additional data collected included demographics, occupation, morbidity, hospitalization history, sources of drinking water, exposure to environmental contaminants, frequency of diarrhea and other enteric infections, antimicrobial use (past and present), and post-urination hygiene practices.

Statistics and reproducibility

Data were entered into Microsoft Excel 2010 (Microsoft Corp., Redmond, WA, USA) and analyzed in Stata version 13.0 (StataCorp, College Station, TX, USA). Descriptive statistics, including frequency distributions, percentages, and means with standard deviations (SD) of E. coli counts, were calculated for all food samples. The percentage of antibiotic resistance was determined for ESBL-Ec isolates from both urinary and food sources.

Sociodemographic characteristics of case (n = 100) and age and sex matched control participants (n = 200) were compared. Associations between food consumption and ESBL-Ec-associated UTIs were evaluated using conditional logistic regression, with UTI status (ESBL-Ec vs. non-ESBL-Ec and no growth) as the dependent variable and food consumption frequency (past three days) as the independent variable. Statistical significance was set at p < 0.05 (two-sided).

Results

ESBL-Ec in urine and food samples

Among the 920 culture-positive urine specimens, 42% (n = 586) were positive for E. coli, of which 44% (n = 259) were identified as ESBL-Ec. Of 2297 food samples, around 87% (n = 1993) tested positive for E. coli, of which 32% (n = 643) were resistant to cefotaxime, and 30% (n = 601) were ESBL-Ec. The highest prevalence of ESBL-Ec was found in poultry meat samples (43%), followed by raw beef (36%), raw fish (24%), raw egg (12%), and fresh-produce samples (6%). Among the poultry meat samples, commercially produced broiler/layer chicken meats were predominantly positive for ESBL-Ec, with a prevalence of 55%, followed by duck meat (45%), free range chicken meat (40%), and pigeon meat (33%) (Table 1).

Table 1.

Prevalence of antibiotic-resistant E. coli in different types of food samples collected from retail markets in Dhaka city

Sample type	Total sample tested	No. (%) of samples “+” for E. coli	Mean log10 count (SD) of E. coli per g of sample	No (%) of samples “+” for cefotaxime-resistant E. coli	No (%) of samples “+” for ESBL-Ec
Fresh produce
Spinach	168	164 (97.62)	3.06 (0.93)	17 (10.37)	13 (7.93)
Potato	168	68 (40.48)	2.93 (1.60)	6 (8.82)	5 (7.35)
Carrot	168	149 (88.69)	3.28 (1.36)	9 (6.04)	7 (4.70)
Total	504	381 (75.60)		32 (8.40)	25 (6.56)
Raw meat
Beef	162	161 (99.38)	3.63 (1.01)	61 (37.89)	58 (36.02)
Total	162	161 (99.38)		61 (37.89)	58 (36.02)
Raw fish
Pangas	166	155 (93.37)	2.71 (1.18)	38 (24.52)	36 (23.22)
Tilapia	168	151 (89.88)	2.72 (1.29)	40 (26.49)	38 (25.17)
Total	334	306 (91.62)		78 (25.49)	74 (24.18)
Egga
Farmed	168	70 (41.67)	1.48 (1.10)	15 (21.43)	10 (14.29)
Deshi (Duck)	159	112 (70.44)	2.35 (1.23)	17 (15.18)	13 (11.61)
Total	327	182 (55.66)		32 (17.58)	23 (12.64)
Poultry
Broiler/layer	250	248 (99.20)	4.40 (0.93)	143 (57.66)	137 (55.24)
Free range	250	249 (99.60)	4.10 (0.90)	106 (42.57)	101 (40.56)
Duck	236	232 (98.31)	4.22 (1.07)	110 (47.41)	105 (45.26)
Pigeon	234	234 (100)	4.16 (0.79)	81 (34.62)	78 (33.33)
Total	970	963 (99.28)		440 (45.69)	421 (43.72)
Total	2297	1993 (86.76)		643 (32.26)	601 (30.15)

^aFor egg sample, the counts were expressed as per cm².

Antibiotic susceptibility of ESBL-Ec

Antibiotic susceptibility testing revealed that 98% (n = 255) of ESBL-Ec isolates from urine samples were MDR. Over 95% of these isolates were resistant to aztreonam (n = 249) and ciprofloxacin (n = 247), while 87% (n = 225) showed resistance to cefepime, 86% (n = 225) to levofloxacin, and 78% (n = 201) to ticarcillin-clavulanic acid (Fig. 2 and Supplementary Data 1, 2). The prevalence of resistance was lowest to ertapenem (4%, n = 10), followed by imipenem (3%, n = 9), meropenem (1%, n = 2), and colistin (1%, n = 3).

Fig. 2.

Antibiotic susceptibility profiles of ESBL-Ec isolated from urine and food samples.

Antimicrobial susceptibility testing of 601 ESBL-Ec isolates from food sources revealed that 82% (n = 495) were MDR. Over 90% of isolates (n = 556) were resistant to aztreonam, followed by 74% to doxycycline (n = 447), and 66% to ciprofloxacin (n = 396). Resistance to other antibiotics was low (<5% of isolates), including piperacillin-tazobactam (n = 13), imipenem (n = 8), ertapenem (n = 8), meropenem (n = 6), doripenem (n = 2), and tigecycline (n = 2) (Fig. 2 and Supplementary Data 1, 2).

Prevalence of ESBL genes

Test for ESBL-encoding genes revealed that around 60% (n = 155) of urine ESBL-Ec isolates harbored bla_CTX-M-gr-1, followed by bla_TEM (36%, n = 92), bla_CTX-M-gr-9 (3%, n = 8), bla_SHV (2%, n = 4), and bla_CTX-M-gr-2 (n = 1). Of 13 carbapenem-resistant ESBL-Ec, 85% (n = 11) were positive for bla_NDM-1. None of these isolates were positive for bla_IMP, bla_VIM, bla_OXA-48, and bla_KPC. In case of food isolates, the majority were positive for bla_CTX-M-gr-1 (66%, n = 400), followed by bla_TEM (53%, n = 317), bla_CTX-M-gr-9 (4%, n = 24), and bla_SHV (1.5%, n = 9). All carbapenem-resistant isolates (n = 10) from food samples were negative for bla_NDM, bla_IMP, bla_VIM, bla_OXA-48, and bla_KPC.

Phylogroup analysis of ESBL-Ec

We found significant differences in the phylogroup distribution of ESBL-Ec isolates from food sources compared to those from patients with UTI. Approximately 63% (n = 163) of urine isolates belonged to phylogroup B2, whereas none of the 601 food-derived ESBL-Ec belonged to this phylogroup. In contrast, 52% (n = 311) of food isolates belonged to phylogroup A, compared to only 4% (n = 11) of urine isolates. Phylogroups that were found at similar prevalence rates in both food and urine isolates included phylogroups C (5% and 4%), D (11% and 6%), E (2% and 3%), and F (12% and 11%), respectively. Additionally, 8% of food ESBL-Ec were classified into phylogroup B1, compared to only 1% of urine isolates. A small proportion of both food (8%) and urine (7%) isolates were untypeable.

Phylogenetic analysis of ESBL-Ec

Multi-locus sequence typing (MLST) analysis of 110 ESBL-Ec isolates from urine samples identified 24 distinct sequence types (ST). ST131 was the most prevalent, representing 56% (n = 61) of isolates, followed by ST1193 (7%), ST648 (6%), ST38 (6%), ST8881 (3%), and ST69, ST410, ST827, ST12, and ST405 (each 2%). Each of the remaining STs accounted for 1% of the urine isolates (Fig. 3A and Supplementary Data 1, 2). In contrast, analysis of the 103 poultry meat isolates showed greater genetic diversity, encompassing 50 distinct STs with no predominant clustering. ST10 was the most frequent (~8%), followed by untypeable STs (7%), ST117 (5%), and ST1011, ST155, ST354, and ST648 (each 4%) (Fig. 3B and Supplementary Data 1, 2).

Fig. 3. Distribution of sequence types (ST) among ESBL-Ec isolates.

A shows proportional distribution of ST of ESBL-Ec isolated from urine samples and B shows the proportional distribution of ST of ESBL-Ec isolated from poultry meat samples.

Of the 74 STs identified in ESBL-Ec isolates from both poultry meat and urine sources, only four STs were commonly found in both sources, representing a total of 34 isolates (15%; 34/213): ST648 (n = 11), ST38 (n = 10), ST10 (n = 9), and ST457 (n = 4). Serotype analysis revealed that isolates within each ST exhibited multiple O:H serotypes. Identical O:H serotypes were observed only in three of the four overlapping STs, such as ST648 (n = 5), ST38 (n = 2), and ST457 (n = 3), representing only 10 isolates (5%; 10/213) (Table 2). Core genome SNP analysis showed that isolates with matching O:H serotypes differed by 30–73 SNPs, with the smallest difference (30 SNPs) observed between isolates belonging to O11:H25 isolated from a urine sample of a child (Isolate ID: CH-089) and a meat sample from a free-ranged chicken (Isolate ID: 13-7-ChFE1) (Table 2).

Table 2.

Serotype and SNP distribution among ESBL-Ec isolates belonged to the same sequence type, overlapping between poultry meat and urine sources

ST (no. of isolate)	Human (serotype)		Chicken (serotype)		Pigeon (serotype)	Duck (serotype)	No. of SNP difference
	Adult	Child	Free-ranged	Broiler
ST648 (N = 11)	O102:H6	O102:H6	O102:H6	O102:H6	O102:H6		67–73
	O1:H6
		O15:H1
		O153:H6
					O83:H42
ST38 (N = 10)	O1:H15
	O86:H18
		O1:H18				O1:H18	53
		O11:H30
					O153:H30
					O153:H30
ST10 (N = 9)	O12:H4
			ONT:H2
			O16:H32
			O8:H17
			O89:H9		O89:H9	O89:H9
			ONT:H12
				O141:H2
ST457 (N = 4)		O11:H25	O11:H25	O11:H25			30
						O25:H25

Clonal overlaps among ESBL-Ec isolates within each source

Core genome SNP analysis of isolates within the same STs identified 21 cases of clonal overlap between isolates from the two sources (Fig. 4 and Table S1). Among these, 9 cases were identified in meat isolates, while the remaining 12 cases were detected in urine isolates. The highest number of clonal overlaps was identified among urine isolates belonging to ST131 and ST1193. Among meat isolates, clonal overlap was observed between isolates from different meat types and markets in different locations. For example, an isolate from pigeon meat was clonally related to isolates from free-range chicken meat and duck meat, all collected from different markets on separate dates. Similarly, clonal relationships were identified among isolates from different patients across various locations and time points. Surprisingly, no clonal overlaps were detected between isolates from urine specimens and poultry meat sources, suggesting no direct transmission between these sources. Overall, ESBL-Ec causing UTI are primarily clustered within a few sequence types, particularly ST131, ST1193, ST648, and ST38, which together account for nearly 75% of all urine isolates tested. While ST648 and ST38 were also identified in poultry meat sources, they were not clonally identical to urine isolates as indicated by substantial SNP differences.

Fig. 4. Maximum-likelihood SNP-based phylogenetic tree of 213 ESBL-Ec isolated from urine specimens of UTI patients (n = 110), and poultry meat (n = 103).

Blue and green highlights indicate a clonal relationship between isolates (≤22 SNPs). Colored rings represent source (inner ring), study site (second ring from the center), phylogroup (third ring), MLST (fourth ring), and serotype (fifth ring).

Characterization of ESBL-Ec ST131 isolates

Among the 61 ST131 isolates from urine samples, 53 belonged to sub-clone C, specifically C1 (n = 10) and C2 (n = 43). All C1 isolates carried fimH41, and 8 out of 10 were positive for bla_CTX-M-15. In contrast, all C2 isolates harbored both bla_CTX-M-15 and fimH24. Additionally, six ST131 isolates were classified into sub-clone A, all carried fimH30, and all but one harbored bla_CTX-M-15. Notably, all ST131 isolates were resistant to ciprofloxacin, except for one isolate that belonged to subtype A (AD-621) (Fig. S1).

Pangenome analysis of ESBL-Ec isolated from urine and poultry meat samples

Pangenome analysis of 213 ESBL-Ec isolates was performed using hierarchical clustering of the Roary-derived gene presence/absence matrix to assess core and accessory genome diversity between the two sources (Fig. 5). The analysis revealed distinct clustering patterns between ESBL-Ec isolates from urine and meat sources. ST131 ESBL-Ec isolates from urine samples formed a tight and well-defined cluster, indicating high conservation of core and accessory gene content within this dominant clonal lineage. In contrast, isolates from poultry meat samples displayed dispersed and heterogeneous clustering patterns, reflecting their diverse sequence types and gene content. Notably, ESBL-Ec isolates of ST10, ST457, ST38, and ST648 were found in both meat and urine sources (Fig. S2). However, they formed separate clusters, suggesting a lack of clonal overlap between the two sources based on accessory genome profiles.

Fig. 5. Heatmap of gene presence and absence from pangenome analysis of 213 ESBL-Ec obtained from poultry meat and urinary sources.

Rows represent individual isolates, and columns represent gene clusters. Black and white bars indicate gene presence and absence, respectively. Urine isolates (orange) and food isolates (blue) are annotated in the side panel under “Source.” MLST types are represented with distinct colors; ST comprising less than 3% of the food or urine isolates were grouped as “Other STs” (ST12, 29, 43, 46, 48, 57, 69, 73, 141, 156, 162, 189, 191, 205, 206, 219,224, 226, 295, 359, 361, 405, 410, 542, 602, 609, 617, 683, 711, 744, 827, 976, 1072, 1140, 1158, 1290, 1312, 1585, 1588, 1638, 1716, 1727, 2141, 2161, 2659, 2732, 3052, 3171, 3268, 3489, 4684, 6793, 9660, 10340, and 11613) shown in grey.

Characterization of isolates for mobile genetic elements (MGE)

Detection of MGE indicated that they were more frequent in urine isolates compared to meat isolates. Among the top 20 MGEs identified in both meat and urine isolates, insertion sequence (IS) associated transposable element MITEEc1 (48% food vs. 52% urine), ISEc1 (37% vs. 45%), and ISSen9 (36% vs. 48%) were the most prevalent in both sample types. IS2, IS629, IS30, ISEc12, and IS6100 were among the top 10 MGEs in urine isolates but did not appear in the top 10 MGEs of poultry meat isolates (Fig. 6A). Overall, only five MGEs, including MITEEc1, ISSen9, ISEc1, ISEc38, and ISKpn8, were commonly detected in ESBL-Ec isolates from both urine and meat sources (Fig. 6B).

Fig. 6. Proportional distribution of the top 20 MGEs identified in ESBL-Ec isolates obtained from urine and poultry meat samples using normalized counts.

A shows the stacked percent bar plot representing the proportional distribution of each of 20 MGEs. B shows the chord diagram illustrating connections between shared MGEs and sources of isolates based on normalized counts to adjust for sample size differences. Each chord represents the normalized frequency of an MGE in a given source. Link widths reflect how often each MGE appears in each source, adjusted for sample size. Category segment sizes (urine and meat) represent total normalized MGE counts, allowing comparison across sources regardless of sample number.

Further analysis of these MGEs for antibiotic resistance genes (ARGs) revealed that none of the top 10 MGEs in poultry or urine isolates carried ARGs. However, among 110 urine isolates, four distinct MGEs were found to carry ARGs. Two isolates carried IS26 with multiple plasmid-borne ARGs (Table S2), one isolate carried Tn7 on the chromosome encoding trimethoprim and aminoglycoside resistance genes (dfrA1_8, and ant(3”)-Ia_1), and one isolate contained a genomic island carrying sul2 gene conferring resistance to sulfamethoxazole (Table S3). Among the 103 poultry meat isolates, 20 harbored ARGs associated with MGEs, including 14 with plasmid-linked MGEs, and six with chromosome-associated MGEs. Among the 14 plasmid-linked MGEs, nine were positive for Tn2, all carrying the bla_TEM-1B encoding cephalosporin resistance. The remaining isolates carried IS26 harboring trimethoprim and aminoglycoside resistance genes, and ISVsa3 carrying a florfenicol resistance gene. One isolate from pigeon meat was positive for Tn6114 carrying bla_CMY-2, conferring resistance to third-generation cephalosporins, and ISApl1 carrying mcr1.1, conferring colistin resistance; these genes were located on plasmids IncIγ/K1 and IncP, respectively. Of the 5 chromosome-associated MGEs, three carried Tn7, harboring either aminoglycoside or trimethoprim resistance genes. The remaining isolates harbored Tn1721 with a tetracycline resistance gene, ISVsa3 with a florfenicol resistance gene, and ISSen9 with a tetracycline resistance gene (Table S3).

Of the 213 isolates analyzed, a single chromosome-associated Tn7 element carrying both aminoglycoside and trimethoprim resistance genes was detected in ESBL-Ec strains from a broiler meat (isolate ID: 1-11-ChBE1) and a child’s urine sample (Isolate ID: CH-119). However, these two isolates belonged to different serogroups and STs and were phylogenetically unrelated to each other and demonstrated no epidemiological linkage.

Association between dietary habits and the incidence of community-acquired UTI

The characteristics of UTI patients with ESBL-Ec and their age- and sex-matched controls included in the study for interview were homogeneous in terms of their socioeconomic status, marital status, recurrent episodes of UTI, self-reported disease morbidity, antibiotic use, incidence of diarrhea and hospitalization history. There was no significant difference between the participants other than the incidence of UTI caused by ESBL-Ec (Table 3).

Table 3.

Sociodemographic characteristics of patients with community-acquired UTI

Variables	Cases, n (%)	Controls, n (%)
Age (years), Mean (SD)	41.7 (19.8)	41.1 (19.6)
Sex (Female)	81 (81)	162 (81)
Monthly household income in thousand Taka (US$)
<30 ($349)	17 (17)	29 (15)
30-50 ($349–$581)	17 (17)	25 (13)
>50 (>$581)	51 (51)	122 (61)
Unwilling to answer/missing data	15 (15)	24 (11)
Marital status
Married	75 (75)	130 (65)
Widow	9 (9)	34 (17)
Single	5 (5)	13 (7)
Unwilling to answer/missing data	11 (11)	23 (11)
Duration of marital status
<6 months	3 (4)	0
6 months–1 year	2 (3)	3 (2)
>1 year	70 (93)	127 (98)
Recurrent UTI (n = 268)
Yes	85 (85)	135 (80)
No	15 (15)	33 (20)
Treatment for the last time UTI
Antibiotic	79 (93)	119 (88)
Do not remember	1 (1)	4 (3)
No treatment	5 (6)	12 (9)
Self-reported comorbidities
Stone in kidney	3 (3)	2 (1)
Hypertension	41 (41)	74 (37)
Diabetes	39 (39)	67 (34)
Diarrhoea/enteric disease within last 6 months
Yes	22 (22)	37 (19)
No. of episodes of Diarrhoea
One	8 (36)	20 (54)
Two	4 (18)	1 (3)
Three	3 (14)	7 (19)
>Three	7 (32)	7 (19)
Do not remember	0	2 (5)
Hospitalized in the last 3 months
Yes	11 (11)	10 (5)

Cases: UTI patients with urine culture positive for ESBL-Ec

Controls included: (1) UTI patients with urine cultures positive for bacteria other than ESBL-Ec, and (2) participants with urine samples negative for any bacterial growth.

Analysis of food consumption data among UTI patients with and without ESBL-Ec revealed no significant associations between regular consumption of any particular type of food and prevalence of UTI caused by ESBL-Ec (Table 4). Around 41% of patients with ESBL-Ec UTI reported consuming broiler meats five times or more in a week, while 32% reported consuming free-range chicken meats at the same frequency. However, when compared with the control groups, consumption of any type of chicken meat was not associated with an increased risk of ESBL-Ec UTI (OR: 1.19, 95% CI: 0.72, 1.99). Similar results were observed for consumption of other foods, including fish, beef, vegetables, eggs, bread, rice and bakery products.

Table 4.

Association between food consumption patterns and the prevalence of ESBL-Ec associated UTI

Dietary factors	Prevalence of UTI associated with ESBL-Ec	OR (95% CI)
Chicken		1.19 (0.72, 1.99)
Free-range chicken	32 (7/22)	0.93 (0.39, 2.25)
Pakistani cock	34 (13/38)	0.92 (0.46, 1.85)
Sonali	29 (8/28)	1.12 (0.50, 2.52)
Broiler	41 (15/37)	0.65 (0.32, 1.31)
Beef	25 (13/53)	1.42 (0.74, 2.72)
Eggs	35 (56/162)	0.89 (0.55, 1.44)
Fish		0.55 (0.32, 0.93)
Small fish	31 (23/75)	1.18 (0.70, 2.02)
Catfish	33 (8/24)	1.00 (0.43, 2.34)
Carp	36 (26/72)	0.83 (0.48, 1.44)
Hilsha	26 (11/43)	1.40 (0.69, 2.87)
Vegetables		1.26 (0.60, 2.65)
Spinach	42 (10/24)	0.95 (0.43, 2.10)
Pointed gourd	36 (14/39)	1.09 (0.54, 2.19)
Gourd	28 (23/83)	1.45 (0.84, 2.49)
Pumpkin	27 (17/62)	1.66 (0.88, 3.11)
Papaya	27 (16/42)	1.88 (1.12, 3.14)
Beans	38 (16/42)	0.96 (0.48, 1.91)
Egg plant	26 (12/46)	1.81 (0.88, 3.75)
Tomato	31 (17/54)	1.28 (0.67, 2.44)
Cucumber	44 (34/78)	0.50 (0.28, 0.91)
Cauliflower	34 (20/58)	1.07 (0.57, 2.00)
Cabbage	25 (8/32)	2.08 (0.87, 4.97)
Carrot	27 (9/33)	1.86 (0.82, 4.24)
Potato	34 (67/199)	0.95 (0.55, 1.63)
Dal	30 (47/159)	1.39 (0.88, 2.22)
Rice	33 (97/290)	0.84 (0.20, 3.49)
Bread		0.89 (0.50, 1.59)
Bakery	27 (6/22)	1.44 (0.51, 4.10)
Chapati	34 (62/185)	1.06 (0.60, 1.87)

Discussion

This study provides a comprehensive analysis of potential foodborne sources of ESBL-Ec associated with community-acquired UTI in a high-burden setting of Bangladesh. Our results demonstrated a substantial burden of ESBL-Ec in both urinary and food samples. Nearly half of all E. coli isolates from urine samples were ESBL producers, and a similar trend was observed in poultry meat, with a notably high prevalence in commercially produced broiler and layer chicken. By integrating clinical surveillance, extensive food sampling, genomic analyses, and patients’ dietary assessments, our findings suggest that ESBL-Ec from poultry meat are genetically distinct from strains causing UTI in the community. This evidence challenges the hypothesis that these infections originate from foodborne sources⁴¹. Contrary to our results, multiple studies demonstrated that food-producing animals, especially poultry, may act as reservoirs for extraintestinal pathogenic E. coli (ExPEC), potentially transmitting these strains to humans via the food chain⁴². This hypothesis was supported by the close genetic relatedness, overlapping virulence genes and comparable pathogenic potential between avian pathogenic E. coli and ExPEC strains⁴². For instance, several studies have reported that E. coli clonal lineages ST69, ST95, ST117, ST648, and ST131, commonly associated with UTIs worldwide, have also been identified in poultry-meat, although the directionality of transmission remains unidentified⁴². A major limitation of these studies was the absence of a clear epidemiological linkage between human and food sources, as isolates were often collected from different times and locations without direct association. Consequently, even when clonal similarities were identified, it remained difficult to definitively attribute human infections to foodborne sources. The only study that directly addressed this issue was carried out by Liu et al.⁴³. While the study estimated that approximately 8% of E. coli strains causing human UTI infections may originate from poultry meat, the WGS–based core-genome analysis found no clonal overlap between meat and human isolates, consistent with our study. Liu et al. extended their analysis to accessory genomes of isolates and identified MGEs that varied by source. By analyzing these MGEs from both meat and human isolates and by applying Bayesian latent class models, they identified the probable host origins of these UTI isolates. Notably, they found significant overlap in several clinically important sequence types such as ST10, ST38, ST69, ST117, and ST131 across meat and human isolates, prompting further analysis of the accessory genome within these overlapping sequence types. In our study, we identified four shared STs (ST10, ST38, ST457 and ST648) between meat and urine isolates, encompassing 15% (n = 34) strains. Only 5% (n = 10) of these isolates belonged to three STs that had matching O:H types across sources, but all exhibited SNP differences higher than the clonality threshold (≤22 SNPs considered as clonal)²⁶. Additionally, accessory genome analysis did not show any overlapping MGE among these isolates, attesting to the non-clonal nature of the isolates. Therefore, the presence of shared STs likely reflects global dissemination of successful lineages rather than recent local transmission.

Overall, the distribution of phylogroups varied significantly between isolates from food and human sources. About 63% of urine isolates belonged to phylogroup B2, which is typical of ExPEC, and half of these ExPEC strains were part of the globally disseminated ST131 clone. In contrast, none of the 601 food isolates belonged to phylogroup B2, and no ST131 strains were detected, despite sampling a broad range of food types from all major retail markets in Dhaka city over a year-long period. Further characterization of ST131 isolates revealed that most belonged to clade C. Core-genome SNP analysis identified several clonal overlaps among ST131 strains from different patients, collected at different times, with no apparent epidemiological links. The high level of clonality among patient-derived ST131 isolates across densely populated Dhaka, dwelling over 20 million people is striking. Further investigation into the virulence traits, host adaptability, resistance mechanisms and temporal dynamics of ST131 is needed to limit their spread in densely populated urban settings. Assessment of MGE-mediated transmission of ARGs identified only a single MGE (Tn7) shared between a human and a meat isolate. However, these isolates belonged to different sequence types and were genetically unrelated. These findings suggest that MGE-based predictive models may have limited utility in high-background, environmentally saturated settings.

In our previous study, we observed a similar pattern that is no clonal overlap of ESBL-Ec isolates between poultry and humans living in close proximity, whether within shared households or on the same farms⁴⁴. Even among individuals with occupational exposure to poultry, there was no clear evidence of direct transmission. A likely explanation is the widespread environmental contamination with ESBL-Ec in low- and middle-income countries, where inadequate water, sanitation and hygiene (WASH) practices contribute to frequent human exposure. This is evident in the alarmingly high prevalence of ESBL-Ec colonization among healthy individuals in these regions compared to high-income countries. Additionally, it is probable that ESBL-Ec colonization in humans in such settings does not require a fitness advantage over ESBL-negative strains⁴⁵. A meta-analysis compiled evidence from 17 studies demonstrating no linkage of clonal transmission of ESBL-Ec between humans and animals in several countries of Asia, Europe and North America⁴¹. Even MGE-mediated transmission was not evident between ESBL-Ec from human and animal origins reported by two previous studies^46,47. However, the meta-analysis identified six other studies from the Netherlands, Spain and North America suggesting whole bacterium transmission of ESBL-Ec between poultry and humans⁴¹. The analysis also indicated the transmission of MGE-mediated genes encoding cephalosporin resistance from food-producing animals to human extraintestinal infections, including UTI cases. Further, 21% of E. coli from suspected UTI cases had the E. coli ST similar to that from poultry origins. All these findings are reinforced by a recent separate study, which revealed that vegetarian diets have a protective association against UTI, particularly in females and in cases of uncomplicated UTI⁴⁸. Contrary to the hypothesis that frequent consumption of poultry meat increases the risk of antibiotic-resistant UTI⁴⁹, our patient interviews and matched control analysis found no significant association between ESBL-Ec UTI and frequent chicken consumption (≥5 times per week). A comprehensive 7-day dietary recall also revealed no linkage of ESBL-Ec UTI with the consumption of any specific food item. This result aligns with the genomic evidence, further supporting the conclusion that, at least in this population, foodborne transmission did not appear to be a major driver of ESBL-Ec associated UTI. However, because ExPEC colonization may precede infection by weeks or months, the dietary analysis should be interpreted as supportive rather than definitive evidence.

A major strength of this study is the temporally and spatially matched sampling framework, which enhances the validity of comparisons between human and food sources. The use of high-resolution genomic tools provides a detailed understanding of transmission dynamics, while the dietary survey complements the genomic findings by adding an epidemiological perspective. However, the study is limited by its focus on a single urban area and selected food types, especially targeting major wet markets while excluding other grocery shops in Dhaka city, which may not capture transmission patterns in rural settings. Moreover, the possibility of indirect or past transmission events cannot be ruled out, which limited our ability to trace the foodborne sources of UTI isolates. Additionally, the cross-sectional study design did not allow us to follow up with patients to assess their exposure to contaminated meat.

In summary, this study demonstrates that while ESBL-Ec is highly prevalent in both food and clinical samples in a low-income setting, there is no genomic or dietary evidence supporting direct foodborne transmission as a major pathway for community-acquired UTI. The clinical burden is driven by a small number of successful clonal lineages, emphasizing the need for targeted surveillance, infection control, and antibiotic stewardship. Future research should prioritize investigating alternative pathways, such as household water, sanitation practices, and human-to-human transmission, using similar robust and temporally linked sampling approaches.

Author contributions

Conceptualization: M.A.I.; Data curation: M.A.I., M.B.A., B.M.F., R.K., M.R., and M.R.I.; Formal analysis: B.M.F., M.R.I., K.I.H., and M.B.A.; Funding acquisition: M.A.I.; Investigation: M.A.I., M.B.A., B.M.F. and R.K.; Methodology: M.R., B.M.F., M.A.I., M.B.A. and K.I.H.; Project administration: M.A.I. and M.B.A.; Resources: M.A.I. and M.B.A.; Software: B.M.F., M.A.I. and K.I.H.; Supervision: M.A.I. and M.B.A.; Visualization: B.M.F. and K.I.H.; Writing—original draft: M.B.A., B.M.F., and M.A.I.; Writing—review & editing: M.B.A., B.M.F., E.R., K.I.H., M.R., M.R.I., and M.A.I.

Peer review

Peer review information

Communications Medicine thanks Lee Heine and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

The whole genome sequencing data (FASTQ) generated in this study have been deposited in the NCBI under the following BioProject accession number: PRJNA1300378. Source Data underlying Figs. 2 and 3 are available in the Supplementary Data 1, 2. Source Data for Figs. 4, 5 and 6 are available in the deposited sequencing files in NCBI. Other relevant data supporting the findings of this study are available from the corresponding author upon reasonable request.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s43856-026-01649-9.