The research team collected 87 surface water samples from 19 sites across the Huzhou watershed, covering all four seasons, and supplemented these with 31 comparison samples from other Zhejiang cities, freshwater sites elsewhere in China, and coastal marine locations. Using mTEC culture followed by uidA PCR confirmation, the researchers isolated between roughly 10 and 100 E. coli colonies per sample. Genomic DNA from all confirmed isolates in each sample was pooled, allowing each water sample to be represented by a single pooled DNA preparation for screening against a panel of 92 targeted antibiotic resistance genes.

E. coli proved to be a persistent and highly variable component of the region’s surface water microbiome. The bacterium was recovered from 70 percent of all sampled sites, with concentrations ranging from below the detection limit of 1 colony-forming unit per 100 milliliters to 9.3 × 10⁴ CFU/100 mL. Urban-impacted sites consistently showed the highest abundances, with a median concentration of 4.0 × 10⁴ CFU/100 mL, roughly two orders of magnitude above the suburban median of 348.5 CFU/100 mL. Seasonal dynamics were pronounced: summer peaks, with a median of 4.5 × 10⁴ CFU/100 mL, coincided with elevated rainfall and temperatures, while winter recoveries dropped to a median of 265 CFU/100 mL.

The resistance gene landscape in Huzhou was dominated by aminoglycoside, sulfonamide, and glycopeptide resistance determinants. Gentamicin resistance genes led the ranking with an average detection rate of 58.2 percent, exemplified by ant(2″)-I, followed by sulfamethoxazole genes such as sul1 and sul2 at 41.0 percent, and glycopeptide genes at 39.5 percent. In a striking and unexpected result, vanA, a vancomycin resistance gene typically associated with Gram-positive bacteria, was detected at a rate of 71.3 percent in the E. coli isolates. The authors caution, however, that PCR positivity does not confirm functional resistance, and such detections may reflect primer cross-reactivity or rare environmental strains rather than genuine vancomycin-resistant E. coli.

Hierarchical clustering of the 92 targeted genes in the Huzhou region revealed five distinct genetic clusters. Cluster 1, containing 13 genes including blaCTX-M-1, showed the highest average detection rate at 66.2 percent, while cluster 2, with 16 genes including blaFOX, followed at 52.9 percent. These dominant clusters were enriched with clinically significant β-lactamase genes and fluoroquinolone resistance markers, indicating that these resistance traits are widely integrated into local E. coli populations. Cluster 3, comprising 45 genes with a mean detection rate of just 7.9 percent, represented a low-abundance group that the researchers interpret as potentially emerging threats requiring long-term monitoring. Polymyxin resistance genes, including mcr variants, were rarely detected at 4.4 percent, suggesting limited regional selection pressure from colistin.

Spatial analysis uncovered a clear anthropogenic gradient in resistance gene burden. University sites recorded the highest average ARG detection rate at 33.5 percent, followed closely by urban parks at 29.1 percent, while suburban parks showed the lowest burden at 19.6 percent. Urban rivers and urban lakes showed intermediate rates of about 27 percent. The researchers attribute the urban enrichment to dense human populations, wastewater inputs, and complex waste streams. Site-specific gene signatures emerged as well: urban rivers were enriched in β-lactamase genes such as blaCTX-M-1 and blaFOX, consistent with wastewater or agricultural runoff contamination, while university sites showed elevated sulfonamide resistance genes and plasmid markers pointing to fecal pollution from concentrated human communities.

Co-occurrence network analysis revealed that resistance genes in Huzhou’s waters do not exist in isolation but form tightly linked assemblages. Of the 92 genes screened across 87 samples, 89 were detected at least once, and the most prevalent gene, ant(2″)-I, appeared in 68 samples. The strongest co-occurrence relationship was between gyrA and parC, two quinolone resistance markers in E. coli, which co-occurred 61 times with a mutual detection rate of 93.8 percent. Aminoglycoside resistance genes ant(2″)-I, aac(3)-IV, and aph(3′)-III each showed strong associations with gyrA. The researchers suggest these patterns may indicate shared horizontal gene transfer mechanisms, such as plasmid-mediated conjugation, or common selective pressures, though the study did not perform genomic sequencing to confirm physical linkage.

To quantify risk, the team applied a three-component framework combining a per-sample risk index, detection frequency analysis of individual genes, and cluster-based source apportionment. Sample risk indices, defined as the number of detected resistance genes per sample, ranged from 4 to 57 with a mean of 26.6. Twenty-two samples, or 25.3 percent of the total, met or exceeded the 75th percentile threshold of 32.5 genes and were classified as high risk. High-frequency genes, defined as those detected in more than 70 percent of samples, included ant(2″)-I at 78.2 percent, gyrA and parC at 74.7 percent each, and vanA at 71.3 percent. Principal component analysis offered only limited explanatory power, with the first five components accounting for 29.2 percent of variance, underscoring that resistance distribution in this system is shaped by multiple interacting factors rather than a single source.

K-means clustering, optimized by silhouette analysis, grouped the 87 Huzhou samples into three clusters with distinct risk profiles. Cluster 0, containing 16 samples with an average risk index of 32.98, and cluster 2, with 48 samples averaging 32.68, both represented high-risk groups carrying multiple resistance genes, while cluster 1’s 23 samples averaged 18.05 and marked a lower-risk baseline. The authors propose this clustering as a framework for prioritizing management, with high-risk clusters requiring immediate intervention and low-risk clusters serving as monitoring baselines. Urban-associated sites, particularly university campuses and city parks, emerge as multidrug resistance hotspots where dense human activity and waste inputs likely drive the accumulation of clinically relevant genes.

The study’s authors emphasize important limitations that temper interpretation of these findings. The characterization relied on targeted PCR for a pre-selected panel of genes, a method that measures prevalence but not gene abundance, expression, or functional resistance, and which carries risks of cross-reactivity, particularly for Gram-positive-associated genes like vanA and mecC. No genomic sequencing of the isolated strains was performed, leaving unresolved questions about whether the genes are chromosomal or plasmid-borne, whether co-occurrence patterns reflect true genetic linkage or PCR artifacts, and whether the genes can actually transfer to human pathogens. The study also lacked quantification of antibiotic residues in water and sediment and experimental tests of horizontal transfer. The researchers call for metagenomics, transcriptomics, and gene mobility experiments to validate these preliminary findings and to inform integrated One Health surveillance strategies for the Taihu Lake watershed.

Subject of Research: Antibiotic resistance genes in E. coli from surface waters of the Taihu Lake watershed, China

Article Title: High-risk ARGs in environmental E. coli from the Taihu Lake watershed

Article References: Wang, A., Xu, X., Jiang, W., Li, B., Zhu, W., Xiao, Q., Xu, L., Sun, J., Chen, K., Yi, Y., Guo, Y., Qiang, X., He, J., Zhou, H., Lin, Y., & Wang, X. (2026). High-risk ARGs in environmental E. coli from the Taihu Lake watershed. iScience, 29(11), Article 117986. https://doi.org/10.1016/j.isci.2026.117986

Image Credits: AI Generated

DOI: 10.1016/j.isci.2026.117986

Keywords: antimicrobial resistance, E. coli, antibiotic resistance genes, Taihu Lake, surface water, Huzhou, One Health, horizontal gene transfer, PCR screening, urban watersheds, vanA, risk assessment