Astragali Radix (AR), derived from *Astragalus membranaceus*, is a widely used Traditional Chinese Medicine (TCM). The market primarily features two forms: wild AR (WAR) and cultivated AR (CAR). However, dwindling WAR resources and increased CAR cultivation have led to market mixing. Chemical composition differences between WAR and CAR could affect clinical efficacy and dose-effect relationships, highlighting the need for a rapid and effective identification method. Existing techniques like spectroscopy and chromatography-mass spectrometry (LC-MS) offer improved authentication, but often require extensive sample preparation and are time-consuming. This study aimed to develop a faster, less-destructive method to address this challenge.

Various spectroscopic techniques (Raman, infrared, UV-Vis) and chromatographic-MS methods (HPLC, TLC, LC-MS, GC-MS) have been used for plant authentication. LC-MS, combining liquid chromatography with mass spectrometry, is particularly useful due to its speed and detailed information. Targeted and non-targeted MS strategies have also been employed in plant metabolomics for phenotype distinction. However, these methods can still be limited by cumbersome sample preparation and lack of comprehensive sample characterization, leading to lower discrimination accuracy. REIMS, a recent ambient ionization MS technique, offers rapid analysis with minimal sample preparation, showing promise in real-time phenotyping and species differentiation. Previous studies have utilized MS to analyze AR for classification and quality control, but single methods may not fully characterize complex components. This necessitates an integrated strategy combining different MS instruments for comprehensive analysis.

This study used a three-part approach. First, REIMS was used to create a discriminant model for real-time identification of WAR and CAR. The REIMS experimental conditions (heater bias and cone voltage) were optimized using quality control (QC) samples. Forty AR samples (20 WAR, 20 CAR) were analyzed, generating REIMS fingerprints. Thirty-six compounds were tentatively characterized from the mass spectra. Second, 2DLC-MS (combining HILIC and RPLC) was used to qualitatively analyze 20 batches of WAR and CAR samples. A total of 124 compounds were characterized using molecular formula prediction, database retrieval, and software analysis. Third, UHPLC-QTRAP-MS in multiple ion monitoring (MIM) mode was used to semi-quantify potential markers screened in the previous steps. Multivariate statistical analysis, including PCA and LDA, were used to process the REIMS and 2DLC-MS data to identify differentiating compounds. The REIMS model was validated using a stratified 5-fold cross-validation. Finally, the REIMS model was used for real-time identification of eight unknown AR samples.

REIMS analysis showed clear separation between WAR and CAR samples in PCA and LDA score plots. The model achieved 99.04% accuracy in cross-validation, successfully identifying eight unknown samples in real-time with >99% confidence. 2DLC-MS analysis identified 124 compounds, including flavonoids, organic acids, saponins, and other compounds. Multivariate analysis of 2DLC-MS data (PCA and OPLS-DA) also showed clear separation between WAR and CAR. The study identified several potential markers that distinguish WAR and CAR based on their different chemical profiles. The relative content of most compounds was higher in WAR than in CAR.

The combined REIMS and 2DLC-MS approach provides a rapid, accurate, and almost nondestructive method for distinguishing WAR and CAR. REIMS eliminates the need for sample pretreatment and offers real-time identification, making it superior to traditional LC-MS methods. The integration of REIMS with 2DLC-MS provides comprehensive chemical characterization and enhances the accuracy of identification. The findings address the critical need for efficient quality control in the AR market. The identification of potential markers allows for improved authentication and quality control of AR products.

This study successfully developed a novel method using REIMS combined with 2DLC-MS for the rapid and accurate identification of WAR and CAR. The method is efficient, requiring minimal sample preparation, and highly accurate. The identified potential markers can be used for quality control purposes. Future research could focus on expanding the compound database and exploring the application of this method to other easily confused TCMs.

The study's sample size, while substantial, could be further expanded for enhanced statistical power. The tentative identification of some compounds based on mass spectrometry data could be confirmed with additional techniques like NMR or reference standards. The generalizability of the REIMS model might need to be tested across a wider range of geographical origins and cultivation practices of AR.