Free-living amoebae predation in nutrient-enriched treated wastewater: ecological potential and machine learning-assisted analysis.

Rapid population growth has intensified pressure on freshwater (FW) resources, significantly impacting agricultural practices and worsening food scarcity in many regions. The reuse of treated wastewater (TWW) for irrigation addresses water scarcity but carries the risk of spreading pathogenic microbes and antibiotic-resistant bacteria (ARB). Free-living amoebae (FLA) in soil prey on bacteria and have been proposed as biocontrol agents. Here, we suggest an eco-biological approach, investigating how key ions in TWW (phosphate, ammonium, and sulphate) affect predation by three FLA species (Vermamoeba vermiformis, Acanthamoeba castellanii and Heterolobosea sp.) on model bacterial prey GFP-tagged Escherichia coli and GFP-tagged Enterococcus mundtii. This approach aims to enhance soil microbial safety and reduce pathogen persistence in irrigation systems. We also develop a YOLOv5 deep-learning model to quantitatively track FLA behaviour in microscopy videos. Our results show that PO43- or NH₄⁺ (100 mg/L) enrichment greatly enhances Vermamoeba's grazing on E. coli (7 log10 reductions in 72 h) compared to controls or SO₄²- addition. Fluorescence microscopy images confirm extensive bacterial interaction and killing by Vermamoeba in PO43- and NH₄⁺ enrichment. In contrast, Acanthamoeba and Heterolobosea had weaker effects. A mixed FLA consortium, however, eliminated E. mundtii faster than any single FLA species. The YOLOv5 model trained on an annotated dataset and validated against manual viable-cell counts reliably tracked FLA movement. These findings suggest that tailoring nutrient levels in TWW can enhance the biocontrol function of FLAs. Our interdisciplinary approach lays groundwork for ecological and AI-informed strategies to make TWW reuse safer.