Mechanistic interpretation of plant nutrients biogeographic patterns is necessary to a predictive framework for global biogeochemical cycle. However, the existed three hypotheses predicting these patterns have not been verified (especially the causal pathways) using both theoretical analysis and empirical data. We tested the validities, efficiencies, and key drivers of three hypotheses for the first time by synthesizing 2344 observations of major nutrients (nitrogen (N), phosphorus (P), and potassium (K)) in leaf from 1073 species and of supporting indices (plant traits, climate and soil) at global scale. Structural equation models proved the validities of each hypothesis at global scale, except for the prediction of foliar N by Soil Substrate Age hypothesis (indirect effect via soil nutrient) as soil N affected leaf N by a path opposite to this hypothesis. Specifically, among three hypotheses, Species Composition hypothesis (indirect effect via plant functional type) is the best predictive model for leaf N, while Substrate Age hypothesis for both P and K in leaf. Soil, climate, and plant functional type together explained over half variations in leaf nutrient concentrations. Among all traits, soil nutrient was the strongest determinant for both P and K in leaf, while plant functional type for foliar N. Overall, we highlight the fundamentally different mechanisms and trait controls regarding global patterns of different nutrients in leaf.