Phenotype–genotype discordance in antimicrobial resistance profiles of Gram-negative uropathogens recovered from catheter-associated urinary tract infections in Egypt

Abstract

Objectives

Catheter-associated urinary tract infections (CAUTIs) are among the most common healthcare-associated infections in low- and middle-income countries (LMICs), but there are few resistome data available for relevant uropathogens. The goal of this study was to characterize the antimicrobial resistance (AMR) phenotypes and genotypes of a large collection of Gram-negative bacteria recovered from CAUTIs in a hospital in Mansoura, Egypt.

Methods

Phenotypic AMR profiles and whole-genome sequence data were generated for 132 isolates. Resistomes were predicted using ResFinder, CARD and AMRFinder. Similarity of uropathogen genomic data was determined using sourmash (kmer signatures). Escherichia coli genomic data were subject to a pangenome analysis using Panaroo.

Results

Sixty-seven E. coli (Phylogroup B2; 53.7%, 36/67), 14 Pseudomonas aeruginosa, 11 Klebsiella pneumoniae, 9 Proteus mirabilis, 8 Providencia spp., 5 Enterobacter hormaechei and 18 rare CAUTI-associated isolates were identified. Several (22/132) isolates were multidrug-resistant, while almost half (62/132) were extensively drug-resistant. Phenotype–genotype discordance was found to be an important consideration in resistome studies in Egypt, with a total concordance of 91% (1115/1225), 85.7% (1273/1485) and 80.5% (1196/1485) for ResFinder, CARD and AMRFinder, respectively. Pseudomonas, at the species level, exhibited the greatest discordance. At the antimicrobial level, meropenem was subject to greatest discordance. New AMR variants were found for Egypt for Pseudomonas (bla_OXA-486, bla_OXA-488, bla_OXA-905, bla_IMP-43, bla_PDC-35, bla_PDC-45, bla_PDC-201) and E. coli (bla_TEM-176, bla_TEM-190).

Conclusions

This study shows that there is phenotype–genotype discordance in AMR profiling among CAUTI isolates, highlighting the need for comprehensive approaches in resistome studies. We also show the genomic diversity of Gram-negative uropathogens contributing to disease burden in a little-studied LMIC setting.

Introduction

Urinary tract infections (UTIs) are an increasing public health problem caused by a range of uropathogens. Uropathogenic Escherichia coli (UPEC) is the primary cause of UTIs, responsible for ∼75% of cases. Other common pathogens include Klebsiella pneumoniae (8%), Candida species (7%), Proteus mirabilis (2%) and Pseudomonas aeruginosa (2%), especially in catheter-associated UTIs (CAUTIs).¹ The frequency of healthcare-associated UTIs is 12.9%, 19.6% and 24%, respectively, in the USA, Europe and developing countries.² In 2019, antimicrobial resistance (AMR) contributed to 48 700 deaths associated with UTIs out of 541 000 cases of infectious syndromes.³ Resistance to almost all antibiotics in healthcare-associated UTIs is above 20% and there is a significant geographical variation.² However, meta-analysis and surveillance data are scarce from low- and middle-income countries (LMICs).^4,5 Egypt initiated a national surveillance programme for healthcare-associated infections (HAIs) to establish benchmarks to reduce AMR in clinical settings.⁶ Phase one of the National Action Plan revealed high levels of resistance to quinolones and cephalosporins in E. coli and K. pneumoniae isolates from urine.⁶ Across 27 countries, Egypt came in third place for total consumption of antibacterials (1355.4 tonnes) in 2020.⁷

In Egypt, CAUTIs are among the most common types of HAIs affecting patients in ICUs.^8–11 In Egypt between 2011 and 2012, 15% (69/472) of HAIs were UTIs, and the rate of CAUTIs was 1.2 per 1000 device-days.⁸ Another surveillance showed that UTIs represented 15% (406/2688) of HAIs, the rate of CAUTI development increased up to 1.9/1000 device-days, and CAUTIs represented 98.3% of UTIs.⁹ An epidemiological study from Egypt (2011–2017) of carbapenem-resistant Enterobacteriaceae showed that UTIs accounted for 14.7% (236/1598) of HAIs, with 63.8% of UTIs catheter-related.¹¹ The second most common HAIs were UTIs (12.2%) in Egyptian clinical settings.¹⁰

There are few reports on the genomic diversity of uropathogens found in Egyptian clinical settings.^12–17 However, high rates of AMR are observed in Egypt besides documented transmission of carbapenem-resistant strains from Egypt to other countries.^18,19 We collected a range of Gram-negative uropathogens from an Egyptian hospital over a 9-month period and generated genotypic and phenotypic data for them. Our aims were to investigate the AMR patterns associated with UTIs in an Egyptian clinical setting and provide evidence that could inform the development of effective antimicrobial surveillance plans and targeted infection control policies. In addition, we assessed the agreement/concordance between phenotypic and genotypic data to evaluate the usefulness of such data from a diagnostic perspective.

Methods

Recovery of isolates

All Gram-negative bacterial isolates (n = 132), anonymized with no patient data, recovered from urinary catheters between December 2020 and August 2021 by staff at the Urology and Nephrology Center, Mansoura, Egypt, during routine diagnostic procedures were included in this study. Isolates were recovered as described previously.¹⁷ The study of anonymized clinical isolates beyond the diagnostic requirement was approved by the Urology and Nephrology Center, Mansoura, Egypt. No other ethical approval was required for the use of the clinical isolates.

Antimicrobial susceptibility testing

Antimicrobial susceptibility testing was performed using the disc diffusion test (DDT) as described previously.¹⁷Twelve different antimicrobial discs (Oxoid Ltd, UK) were used: penicillins–piperacillin/tazobactam (30/6 μg); cephalosporins–ceftazidime (10 μg) and cefepime (30 μg); monobactams–aztreonam (30 μg); carbapenems–doripenem (10 μg) and meropenem (10 μg); aminoglycosides–tobramycin (10 μg) and amikacin (30 μg); quinolones–ciprofloxacin (5 μg) and levofloxacin (5 μg); miscellaneous–nitrofurantoin (100 μg) and sulfamethoxazole/trimethoprim (1.25–23.75 μg).

Results were interpreted according to breakpoints of EUCAST guidelines version 12.0, 2022 (http://www.eucast.org/clinical_breakpoints/), with reference strains—E. coli ATCC 25922 and P. aeruginosa ATCC 27853—used for quality control purposes.

WGS and bioinformatic analyses

DNA extraction from isolates, sequencing and genome assembly have been described previously.^17,20

Bakta v1.5.1 (database v4.0)²¹ was used to annotate genomes. sourmash v4.8.4²² was used to compare and, where possible, identify (with Genome Taxonomy Database release 214) species for all genomic data (Table S1, available as Supplementary data at JAC Online). Genomes not assigned to species were identified using ribosomal MLST (https://pubmlst.org/species-id)²³ and by comparison with relevant reference sequences using FastANI v1.33.²⁴

Serotypes were determined using ECTyper v1.0.0 (database v1.0).²⁵ Phylogrouping of E. coli was achieved using EzClermont v0.7.0.²⁶ MLST of each isolate was determined using mlst v2.23.0 (https://github.com/tseemann/mlst) and relevant MLST schemes from https://pubmlst.org/.^27–33

AMR genes were identified using three AMR databases (accessed February 2025): the Resistance Gene Identifier v6.0.3 of the Comprehensive Antibiotic Resistance Database (CARD) v4.0.0,³⁴ ResFinder v4.6.0 (ResFinder database v2.4.0, PointFinder database v4.1.1)³⁵ and AMRFinder v4.0.3 (database version 2024-12-18.1).³⁶ Efflux-pump-associated genes were excluded from analyses.³⁷ Only AMR genes with identity ≥90% were reported.³⁸

For the E. coli genomes, a phylogenetic tree was constructed using core genes identified using Panaroo v1.3.4 (strict mode, identity threshold 98%, alignment option ‘core’).³⁹ Core genes present in all 67 E. coli were aligned with prank (v170427).⁴⁰ Gene alignments were concatenated with FASconCAT-G (v1.02) into a single alignment.⁴¹ A phylogenetic analysis was performed using the concatenated sequences and RaxML v1.2.0 (GTR + G8 + IO).⁴² Node support was determined using 1000 non-parametric bootstrap replicates. The phylogenetic tree was visualized using iTOL v6.9.⁴³ Virulence traits encoded in genomes were detected via BLASTP (≥90% coverage,  ≥ 70% identity) with the core protein sequence dataset from the Virulence Factors Database (VFDB).⁴⁴

Comparison of AMR genotypes and phenotypes

According to EUCAST guidelines, ‘susceptible, standard dosing regimen’ (S) and ‘susceptible, increased exposure’ (I) are grouped as ‘susceptible’. Hits with AMR genes from different databases (CARD, ResFinder, AMRFinder) were compared with DDT results for 132 isolates, except when breakpoints were unavailable (e.g. nitrofurantoin was used only for E. coli, while sulfamethoxazole/trimethoprim was not used for Pseudomonas). Concordance was positive when (i) AMR genes were predicted and isolates were phenotypically resistant (WGS-R/DDT-R) or (ii) no AMR genes were predicted and isolates were phenotypically susceptible (WGS-S/DDT-S). Discordance [major errors (MEs) or very major errors (VMEs)] was noted when genotypic and phenotypic results disagreed: MEs = false-positives (WGS-R/DDT-S); VMEs = false-negatives (WGS-S/DDT-R).^45,46 PointFinder (ResFinder) and Point mutations (AMRFinder) tools were included to detect AMR gene mutations.

Results and discussion

Genome characterization

Summary statistics for the 132 high-quality genomes assembled in this study can be found in Table S1. Sourmash analyses (Figure 1a) assigned all genomes to the class Gammaproteobacteria, with the exception of Me47, which was found to represent a novel member of the family Alcaligenaceae. Among our isolates, UPEC dominated (50.7%, 67/132), followed by P. aeruginosa (10.6%, 14/132), K. pneumoniae (8.3%, 11/132), other uropathogens (6.8%, 9/132), P. mirabilis (6.8%, 9/132), Providencia spp. (6%, 8/132), Enterobacter hormaechei (3.7%, 5/132), Alcaligenes spp. (3%, 4/132), Serratia spp. (2.2%, 3/132) and Pseudomonas vlassakiae (1.51%, 2/132). UPEC was previously reported as the main cause of UTIs for adults and neonates in Egypt, followed by other Enterobacterales or Gram-positive bacteria.^49–52 Once considered a rare cause of UTIs,⁵³  Providencia was among the main taxa represented in our study.

Figure 1.

AMR profiles of uropathogens (n = 132 isolates) included in this study. (a) Dendrogram generated from comparison of sourmash signatures for uropathogens showing their species affiliations and whether they are MDR or XDR. Phylogroup affiliations are given for E. coli isolates. (b) Resistance summary for the 132 uropathogens using seven different antimicrobial classes according to EUCAST guidelines. (c) Data presented using the category designations suggested previously.^47,48 Penicillins: piperacillin/tazobactam. Monobactams: aztreonam. Cephalosporins: ceftazidime and cefepime. Carbapenems: doripenem and meropenem. Quinolones: ciprofloxacin and levofloxacin. Aminoglycosides: amikacin and tobramycin. Miscellaneous: sulfamethoxazole/trimethoprim and nitrofurantoin.

Antimicrobial susceptibility of isolates

According to Egyptian Urological guidelines, symptomatic CAUTIs are treated like complicated UTIs with second- or third-generation cephalosporins, while uncomplicated pyelonephritis is treated with a short course of fluoroquinolones as the first-line therapy.⁵⁴ AMR for quinolones (ciprofloxacin and levofloxacin) and cephalosporins (cefepime and ceftazidime) was >64% among MDR uropathogens in Egyptian clinical settings.⁵⁵ More than half of UPEC isolates (57.8%) showed high-level ciprofloxacin resistance, and gyrA mutations were detected in 76.7% of isolates in a previous study, leading to recommendations for revision of empirical antibiotic treatment of UTIs.⁵⁶

Many of our isolates were resistant to ceftazidime (69.2%, 90/130) and ciprofloxacin (63%, 82/130). Few isolates were resistant to nitrofurantoin (5.9%, 4/67) and meropenem (18.3%, 24/131) (Figure 1b; Table S2). Based on resistance profile classification,^47,48 the prevalence of susceptible, R1 ‘resistant to one antimicrobial category’ and R2 ‘resistant to two antimicrobial categories’ was 21.2%, 6% and 9.1% respectively; 16.7% (22/132) of isolates were MDR, while 47% of isolates (62/132) were XDR (Figure 1c). E. hormaechei (60%, 3/5) and K. pneumoniae (100%, 11/11) were associated with the highest prevalence of MDR and XDR, respectively.

AMR determinants

A wide range of AMR determinants for β-lactams, quinolones and aminoglycosides have been reported in different uropathogens.^57–60 The main AMR genes encoded in our genomes were identified using ResFinder, CARD and AMRFinder (Table S3). New AMR variants are reported here for uropathogens in Egypt: bla_OXA-486, bla_OXA-488, bla_OXA-905, bla_IMP-43, bla_PDC-35, bla_PDC-45 and bla_PDC-201 for P. aeruginosa, and bla_TEM-176 and bla_TEM-190 for E. coli.

WGS-based diagnostics may soon take over phenotypic testing for surveillance purposes, particularly in situations where the low error rate has minimal consequences.⁶¹ Comparison of WGS data and DDT results (with respect to predicted AMR genes and empirical resistance phenotypes) yielded a total concordance of 91% (1115/1225), 85.7% (1273/1485) and 80.5% (1196/1485) for ResFinder, CARD and AMRFinder, respectively. The highest concordance was recorded for Alcaligenes (n = 4) species: 97.7%, AMRFinder; 97.2%, ResFinder; 100% CARD (Table S4).

Variations in discordance were detected for different databases, where the MEs (WGS-R/DDT-S) were 4%, 12% and 15.5%, while VMEs (WGS-S/DDT-R) were 4.8%, 2.2% and 3.9% for ResFinder, CARD and AMRFinder, respectively (Figure 2). The use of an ineffective therapeutic agent in treatment due to a VME could result in treatment failure, while an ME may restrict therapeutic choices and create complications in the treatment process. The more severe impact of VMEs is evident in FDA regulations for the approval of diagnostic tests, where the FDA mandates that VMEs be kept below 1.5% and MEs below 3% for the approval of a new AMR diagnostic test or device.⁶² Although the broth microdilution method for determining the minimum inhibitory concentration (MIC) is considered the gold standard, DDT was used here as a standard, cost-effective approach commonly employed in Egyptian clinical settings. This approach enables clinicians to develop practical knowledge that can inform future studies.

Figure 2.

AMR genotype–phenotype discordance summarized for the 132 isolates included in this study. (a, b) Percentage and number of genotype–phenotype concordances and discordances are shown, when comparing WGS data and DDT results using ResFinder, CARD and AMRFinder. The bars are coloured according to Concordance or Discordance (refer to colour legend at the top of the figure). White numbers on top of the bars correspond to the respective number of results. (a) Species-level data, (b) antimicrobial-level data. Numbers in parentheses in (a) correspond to the number of isolates per species/the number of isolate-antimicrobial combinations. (a) Others: Paenalcaligenes suwonensis, Klebsiella (Raoultella) ornithinolytica, Citrobacter portucalensis, Leclercia adecarboxylata, Morganella morganii, Stenotrophomonas maltophilia, Acinetobacter baumannii and Achromobacter xylosoxidans. (b) Levofloxacin and doripenem were not analysed with ResFinder/PointFinder. TZP, piperacillin/tazobactam; ATM, aztreonam; CAZ, ceftazidime; FEP, cefepime; DOR, doripenem; MER, meropenem; CIP, ciprofloxacin; LEV, levofloxacin; AK, amikacin; TOB, tobramycin; SXT, sulfamethoxazole/trimethoprim; NIF, nitrofurantoin. Major errors are WGS-R/DDT-S, while very major errors are WGS-S/DDT-R.

Concordance for aminoglycosides, quinolones and miscellaneous antibiotics exceeded 87.6% in all databases. Meropenem was associated with the highest discordance (48%) in both AMRFinder and CARD, but showed 10.6% discordance in ResFinder. The range of total discordance of β-lactam antibiotics for the three databases was 10.6%–20.6% ResFinder, 9.1%–48% CARD and 22.1%–48% AMRFinder (Table S5).

In a comparison with CARD and ResFinder across 2587 isolates representing five pathogens of medical importance, ME rates were higher in CARD (42%) than ResFinder (25%). However, CARD showed almost no VMEs (1.1%) compared with ResFinder (4.4%).³⁷ Another report showed an overall concordance of 91% for ResFinder while ME and VME rates were 6.2% and 2.1%, respectively, for 488 different isolates, where P. aeruginosa had the highest percentage of discordance (44.4%).⁴⁵ In agreement with previous studies,^17,45,61  P. aeruginosa showed the highest discordance in our study; 31.2% for AMRFinder, 28.9% for ResFinder and 21.2% for CARD.

In this study, 34% of total MEs tested phenotypically ‘susceptible, increased exposure’ and were therefore considered as not resistant due to the EUCAST update of susceptibility definitions 2019 ‘by the pooling of S and I isolates together into one category’.⁶³ Notably, in another study, 22.7% of MEs tested phenotypically as ‘susceptible at higher exposure’ in 234 E. coli strains.⁴⁶ Discordant bioinformatic predictions of AMR have previously been observed using three databases (ResFinder, CARD, ARG-ANNOT), with overall ME (false-positive) and VME (false-negative) rates of 20% and 14%, respectively.³⁸ In the current study, the overall MEs and VMEs in three databases were 10.9% and 3.5%, respectively. The reduction in MEs and VMEs compared with an earlier study³⁸ may in part reflect improvements in curation and coverage of AMR genes through regular database updates. The high MEs rate may be also due to low expression levels of enzymes, insufficient understanding of gene regulation, mutations in intergenic regions and de novo rRNA mutations.

ResFinder/PointFinder (overall concordance, 91%) showed better in silico AMR prediction than CARD (85.7%) and AMRFinder with Point mutations (80.5%) (Figure 2). PointFinder tackles chromosome mutation-mediated resistance only in E. coli and K. pneumoniae. A combined AMR prediction from ResFinder and PointFinder for ciprofloxacin was shown to result in improved prediction performance and a reduced VME rate.³⁷

For accurate in silico AMR predictions, our understanding of phenotypic–genotypic correlations must be improved to reach FDA requirements for clinical microbiology diagnostic testing. Excluding genes associated with efflux-pump mechanisms from results is recommended.³⁷ Identity thresholds for AMR gene detection should be ≥90% for different databases.³⁸ Point mutations leading to AMR should be considered alongside acquired resistance determinants, especially if they are reported from specific clinical sources such as urine, as the most common site for discordance was urine for both Enterobacterales and P. aeruginosa, and mutations in gyrA are associated with quinolone resistance of Enterobacteriaceae in urine.^64–66 Databases should replace general terminology (β-lactams, quinolones, etc.) of antimicrobial class as a prediction to confer resistance to all members of the same antimicrobial class, so in silico AMR prediction should be tailored to antibiotics themselves rather than antibiotic classes to avoid MEs. For example, the aminoglycoside gene aac(6′)-1 leads to amikacin and tobramycin resistance, while aac(3)-IIa leads to gentamicin and tobramycin resistance.⁶⁷ Poor correlation (85.9%) between EUCAST and WGS data was reported for aminoglycosides using ARG-ANNOT and CARD databases, with a phenotype-based algorithm developed to reflect mechanisms responsible for the corresponding aminoglycoside resistance phenotypes.⁶⁸

Curation of in silico databases is required. Some AMR genes listed in databases are not truly responsible for AMR. For example, crpP—identified by ResFinder and AMRFinder—was previously known as a novel ciprofloxacin-modifying enzyme.⁶⁹ However, recent studies showed that crpP presence is not always associated with ciprofloxacin resistance.^70–72 Variation between databases was also observed for isolate Me17, where different variants of bla_TEM were identified exclusively by ResFinder, thereby increasing the total number of genes detected by this database up to 21 hits (Table S3). Another example, the AMR phenotype associated with armA by ResFinder was amikacin and tobramycin resistance, while AMRFinder predicted resistance to gentamicin. Moreover, the phenotypic prediction for aac(6′)-Ic in ResFinder and CARD is resistance to amikacin and tobramycin, while AMRFinder only provides aminoglycoside as subclass. Although both ResFinder and AMRFinder identified the gene from the exact same contig region with identical coverage, the predicted gene names and associated resistance prediction differed in isolate Me21. ResFinder annotated the gene as aac(6′)-Ib3, predicting resistance to amikacin and tobramycin, while AMRFinder identified it as aac(6′)-Ib, predicting resistance to gentamicin.

Detailed analyses of UPEC isolates

Pangenomic approaches highlight genetic diversity, revealing adaptations, virulence factors and resistance traits among collections of genomes. Several pangenome studies of UPEC strains have been performed for global collections,⁷³ and individual countries, including Ireland and Bangladesh,^74,75 with the smallest pangenome consisting of at least 16 797 genes in total, the number of core genes ranging from 2926 to 2945 and the number of unique genes ranging from 2900 to 15 266. The UPEC pangenome (n = 67) here consisted of 10 453 genes: 3136 (30%) core; 6303 (60.3%) accessory; 1014 (9.7%) unique.

UPEC isolates from Phylogroup B2 were responsible for most of the reported UTIs (53.7%, 36/67) (Figure 3), in agreement with a previous Egyptian study.⁷⁶ We identified 29 different serotypes (Table S1), with O25:H4 (17.9%, 12/67) and O75:H5 (16.4%, 11/67) being the most common. Although Phylogroup B2 strains are uncommon among the commensal microbiota, they are highly virulent when present; they persist in the intestinal microbiota of infants,⁷⁷ while also being associated with a range of animals.^78,79 A previous Egyptian study showed the most prevalent phylogroup in the country was A followed by B2.⁸⁰ Other phylogroups detected by us were B1 (19.4%), A (11.9%), D (7.4%), F (5.9%) and C (1.4%).

Figure 3.

Phylogenetic tree for 67 E. coli isolates constructed using pangenome analysis of core genes, via Panaroo and RaxML. UPEC strains are grouped into clusters (A, B1, B2, C, D, F) depending on EzClermont phylogroups. The maximum-likelihood tree was annotated using iTOL, with bootstrap values expressed as a percentage of 1000 replicates. Scale bar, nucleotide substitutions per site. Virulence genes were identified through BLASTp with VFDB based on coverage ≥ 90% and identity ≥ 70%.

The principal virulence genes associated with all UPEC strains were Type 1 fimbriae (fimH), outer membrane protein (ompA) and enterobactin (entS). Curli fibres (csgA 98.5%, 66/67), common pilus (ecp 97%, 65/67), nutritional and metabolic factors (fyuA 80.5%, 54/67; chuA 67.1%, 45/67), K1 capsule type (kpsM 65.6%, 44/67), aerobactin (iutA 59.7%, 40/67), P fimbriae (papC 52.2%, 35/67; papG 32.8%, 22/67), α-hemolysin and iron siderophores (hlyA and iroN 23.8%, 16/67), exotoxins (cnf1 13.4%, 9/67) and S fimbriae and F1C fimbriae (sfa 11.9%, 8/67; foc 5.9%, 4/67) were also represented (Figure 3; Table S6). fimH and sfaX were present in all phylogroups; all other sfa genes were associated with Phylogroup B2 only. entD was only in Phylogroup B1. cnf-1 was present in Phylogroup B2 isolates and only one isolate (Me87) from B1.

fim, ecp and csg gene families have previously been found to be core to the UPEC pangenome.⁷⁵ In contrast, a PCR-based study from Mansoura reported a much higher prevalence of afimbrial adhesins such as afa/dra (14%) and S fimbriae (sfaS 60.6%),⁸¹ which were largely absent from our dataset. Additionally, our data show a higher prevalence of P fimbriae genes (papG 33%; papC 52%) compared with⁸¹ who found these genes in only 21.3% and 49.3% of isolates, respectively. Similarly, our detection of iron acquisition genes iutA (60%) and chuA (67%) exceeds previously reported values of 34.6% and 54%.⁸¹ Notably, cnf1 (cytotoxic necrotizing factor 1) was much rarer in our dataset (13%) compared with 42%⁸¹ suggesting potential differences in strain pathogenicity. Overall, our results highlight a strong presence of core adhesion and iron acquisition systems in UPEC strains, with notable variations in toxin-associated genes compared with previous studies. These findings emphasize the genetic diversity of predicted UPEC virulence factors and the importance of comparing detection methods when assessing pathogenic potential.

The β-lactam resistance genes bla_OXA-1 and bla_TEM-1 were present in all phylogroups, bla_TEM-176 was restricted to Phylogroup B1, while other bla_TEM variants were present in only Phylogroup A. bla_DHA variants were restricted to Phylogroup B2, while bla_NDM-5 was absent from B2. bla_CTX-M-15 was present in all phylogroups, while bla_CTX-M-14 and bla_CTX-M-16 were present in Phylogroup F only. In Egypt, bla_CTX-M-15 is the most predominant ESBL-producing type among UPEC and diarrheagenic E. coli strains.⁸² For quinolone resistance, qnrB4 was detected only in Phylogroup B2, whereas mutations in gyrA and parC were present across all phylogroups. Aminoglycoside-modifying enzyme genes were likewise detected in all phylogroups.

UPEC isolates showed the highest susceptibility to nitrofurantoin (5.9% resistance; 4/67). Nitrofurantoin was reported as a drug of choice for treating UTIs in previous Egyptian studies,^83,84 but it has therapeutic limitations due to side-effects in certain patient groups.⁸⁵  In silico AMR prediction across different databases showed >94% concordance with phenotypic results (Figure 2). In a recent Egyptian study, the resistance level to nitrofurantoin was 51% in Klebsiella spp., but it remains effective for E. coli UTIs in males and females (92% sensitivity).⁸⁶ Conversely, E. coli and Klebsiella spp. showed elevated resistance rates to cephalosporins: 75% and 81%, respectively.⁸⁶

Conclusions

Our study provides crucial insights into the genomic and AMR landscape of CAUTI-associated uropathogens in Egypt, highlighting the importance of understanding AMR phenotype–genotype discordance in this setting. To improve the accuracy of in silico AMR prediction, our results emphasize the need for better-curated and harmonized resistance gene databases. These findings reinforce the need for integrated phenotypic and genomic surveillance strategies to guide treatment decisions and inform infection control policies in resource-limited settings.

Funding

This work was supported by the Egyptian Ministry of Higher Education & Scientific Research represented by the Egyptian Bureau for Cultural & Educational Affairs in London. N.H. was funded through a placement studentship provided by the Department of Biosciences, Nottingham Trent University. Computing resources used in this study were funded through the Research Contingency Fund of Nottingham Trent University.

Transparency declarations

The authors declare that there are no conflicts of interest.

Data availability

The sequence data included in the study are available under BioProject PRJNA1208035.

Supplementary data

Tables S1–S6 are available as Supplementary data at JAC Online.