4.6.2. Metabolites Acquisition

An Agilent 6520 UPLC-QTOF MS/MS (Agilent Technologies, Santa Clara, CA, USA) was used in the study. Chromatographic separations were performed on an XSELECTTM HSS T3 column (2.1 mm × 100 mm, 2.5 μm, Waters, Milford, CT, USA) at a column oven temperature of 40 ◦C; 0.1% formic acid (A) and ACN (B) were used as the mobile phase. The gradient conditions were as follows: 0–2 min, 5% buffer B; 2–17 min, 5–95% buffer B; 17–20 min, 95% buffer B; 20–21 min, 95% buffer B. The post time was 6 min for column equilibration. The flow rate was maintained at 0.4 mL/min, the injection volume was 4 μL and the auto-sampler temperature was set at 4 ◦C. Electrospray ionization source (ESI) was set in both positive and negative-ion mode, and MS parameters were performed as follows: the scanning range was set 50~1100 *<sup>m</sup>*/*<sup>z</sup>*, electrospray capillary voltage with 4 kV in positive mode and 3.5 kV in negative mode were used, the nebulizer pressure was set at 45 psi, nitrogen was used as drying gas with a flow rate of 11 L/min and temperature was set as 350 ◦C, fragment voltage was maintained at 120 V, skimmer voltage was set at 60 V, Octopole RF Peake was set at 750 V, 121.0509 Da and 922.0098 Da were used at reference masses (*m*/*z*).

### 4.6.3. Data Reduction and Pattern Recognitio

All data were acquired using Agilent MassHunter workstation software version B.01.04 (Agilent, Santa Clara, CA, USA). Firstly, the UPLC-QTOF MS/MS raw data were converted to mzdata files. The isotope interferences were excluded and the threshold of the absolute peak height was set at 500. The R package "xcms" was employed to generate a data matrix through peak extraction, alignment and integration, and the forma visual table including sample name, and peak indexes (*m*/*z*-Rt pairs and peak area). XCMS parameters were default settings except for the following: fwhm = 8, bw = 10 and snthersh = 5. All the ions were filtered based on the 80% rule before all of the detected ions in each sample were normalized to the sum of the peak area to obtain the relative intensity of metabolites based on MATLAB7.1 (MathWorks, Natick, MA, USA). After being normalized, ion intensities were converted to CSV (Comma-Separated Values) data and imported into the SIMCA-P program (version 12.0, Umetrics, Umea, Sweden) for principal component analysis (PCA) and orthogonal partial least-squares-discriminant analysis (OPLS-DA) after mean-centering and Pareto scaling. The parameters (R2X, R2Y, and Q2Y) showing the goodness of fit and prediction were assessed by SIMCA-P for internal validation.

In order to verify the different metabolites, first ions based on the extracted ion chromatogram (EIC) and then the extracted molecular weight with the common metabolite databases, such as the Human Metabolome Database, (http://metlin.scripps.edu) were confirmed. KEGG website was also adopted to enrich the relative pathways for these metabolites.
