Abstract
Soil metabolites are not only the integrated expression of soil biological activities but also the pivotal drivers of biogeochemical cycling, thereby significantly influencing the formation of crop quality. This study established an improved ultra-performance liquid chromatography (UPLC) fingerprint with chemometrics for characterizing the nonvolatile metabolite profiles of soils from geographical locations defined by different tobacco flavor styles in Guizhou. The homogenization extraction and then UPLC analysis were selected as the optimal system due to their superior repeatability and reproducibility, with intraday and interday RSD% of the common peaks (retention time and peak area) below 2.75%. The fingerprint profiling was established using 18 soil samples from three locations, namely the honey-sweet region I, II, and the fresh-sweet region. Thirty common peaks were identified, with similarities ranging from 0.809 to 0.956. Then, the common peaks were subjected to chemometrics analysis. These results indicated that significant differences were observed by principal component analysis (PCA), and 17 characteristic metabolites were viewed as primarily discriminatory by partial least squares discriminant analysis (PLS-DA). The total content of characteristic metabolites followed a trend of honey-sweet region II > honey-sweet region I > fresh-sweet region, with individual metabolites generally higher in the honey-sweet region. Finally, external validation using the hierarchical cluster analysis (HCA) and Fisher discriminant analysis (FDA) model accurately classified four soil samples, further confirming the representativeness of the characteristic metabolites. This study supplies a theoretical foundation to understand the relationship between tobacco flavor formation and soil metabolism, showing great potential applications in agricultural research.