Multiomics and machine learning unveil root exudate-microbiota interactions for cadmium control in rice

Human health and the safety of rice are threatened by cadmium (Cd) contamination in paddy soils. Although root exudates affect Cd bioavailability, the impact of complex mixtures of exudates and their interactions with rhizosphere bacteria on Cd speciation remain poorly understood. This study pioneers the "metabolome - microbiome - Cd speciation" coupling mechanism using a machine learning-assisted multi-omics analysis. Specifically, we characterized root-exudate composition, microbiome, and Cd speciation in rhizosphere soils of seven rice genotypes spanning low-to high-Cd accumulation. Of the 2659 root exudates identified, XGBoost-SHAP pinpointed ten key exudates that significantly influenced acid-extractable Cd (ACD-Cd; R² = 0.72 with grain Cd). Mantel testing revealed strong relationships (P < 0.01) between these key exudates and specific microbial taxa, including Candidatus Sulfobium mesophilum, Geotalea uraniireducens, and Hypericibacter terrae. Key exudates exhibited direct detrimental effects on ACD-Cd (λ = -0.727 to -0.486) and indirect effects through microbial recruitment (e.g., λ = -0.282 for hydroxysuberic acid via Candidatus Sulfobium mesophilum), as determined by partial least squares-structural equation modeling (PLS-SEM) quantification of interaction pathways. Overall regulation was primarily influenced by direct effects, primarily through chelation or precipitation with Cd ions, thereby reducing their bioavailability. Additionally, certain exudates such as L-theanine and Pimelic acid attracted Candidatus Sulfobium mesophilum, which participates in the sulfur cycle and forms insoluble metal sulfide precipitates, reducing the solubility and bioavailability of Cd in soils. This insight offers new targets for rhizosphere engineering to reduce Cd accumulation in rice grains.