Machine learning-powered activatable NIR-II fluorescent nanosensor for in vivo monitoring of plant stress responses

Real-time monitoring of plant stress signaling molecules is crucial for early disease diagnosis and prevention. However, existing methods are often invasive and lack sensitivity, rendering them inadequate for continuous monitoring of subtle plant stress responses. In this study, we develop a non-destructive near-infrared-II (NIR-II) fluorescent nanosensor for real-time detection of stress-related H2O2 signaling in living plants. This nanosensor effectively avoids interference from plant autofluorescence and specifically responds to trace amounts of endogenous H2O2, thereby providing a reliable means to real-time report stress information. We validate that it is a species-independent nanosensor by effectively monitoring the stress responses of different plant species. Additionally, with the aid of a machine learning model, we demonstrate that the nanosensor can accurately differentiate between four types of stress with an accuracy of more than 96.67%. Our study enhances the understanding of plant stress signaling mechanisms and offers reliable optical tools for precision agriculture. Hu and colleagues present a machine learning-powered NIR-II fluorescent nanosensor that enables real-time, non-invasive monitoring of plant stress by visualising internal signalling, thereby allowing for the identification of different types of stress.