Label-free Raman microspectroscopic cellular imaging and untargeted UPLC-HRMS metabolomics of apple branch canker progression based on chemometrics and machine learning

Apple branch canker, caused by Botryosphaeria dothidea , poses a serious threat to apple production worldwide. Despite its economic importance, current understanding and analysis techniques of the infection dynamics and host defense responses remains limited. Here, an in situ , label-free, high-throughput, and precise imaging method based on confocal Raman microspectroscopic imaging coupled with chemometrics was developed to characterize key defensive constituents, including lipids, lignin, phenols, cellulose, hemicellulose, and pectin, within the cuticle and cell wall, and to monitor their spatiotemporal dynamics across the epidermis and cortex. During infection, lignin of the cuticle in epidermis initially decreased while lipids and phenols increased; Defensive components of cell wall in the cortex decreased and then increased by 3 days; All these constituents decreased by 5 days, leading to disease symptoms. Furthermore, to gain a systems-level understanding, untargeted metabolomics using UPLC–Q–Orbitrap–MS and machine learning screened 2090 differential metabolites (including phenylpropanoids and polyketides, organic heterocyclic compounds, lipids, and lipid-like molecules), and revealed significantly enriched metabolic pathways associated with defense activation ( e.g. , isoquinoline alkaloid biosynthesis, flavonoid biosynthesis). Additionally, a rapid, non-destructive diagnostic platform integrating Raman spectroscopy with machine learning achieved early detection of apple branch canker with an overall accuracy of 97.2%. Collectively, this integrated multi-technology strategy provides new insights into the pathogenicity and defense responses of fruit trees and establishes a non-destructive tool for early disease diagnosis.