*2.2. Qualitative and Quantitative Analyses of Metabolites and Quality Control (QC) Analysis of Sample*

Anthocyanin is the cause of red leaf color. From the pigment contents, it can be seen that the contents of total flavonoids and total anthocyanins change significantly in the process of leaf color change. To compare the differences of flavonoid-metabolites, the UPLC (Shim-pack UFLC SHIMADZU CBM30A)−MS/MS (Applied Biosystems 4500 QTRAP) technique was used to detect the flavonoid-related metabolites in RL, YL and GL. The total ion flow map of the mixed sample quality control (QC) sample (Total ions current, TIC, is the map obtained by adding the intensities of all ions in each time point mass spectrum) is shown in Figure S1a. The multi-peak map of MRM metabolite detection of multi-substance extraction is illustrated in Figure S1b. Based on the local metabolic database, the metabolites of the samples were qualitatively and quantitatively analyzed by mass spectrometry. The multi-peak map of MRM metabolite detection in the multi-reaction monitoring mode shows the substances that can be detected in the sample, and each mass peak of different color represents a detected metabolite. A total of 196 flavonoid-related metabolites were detected (Table S1), including 11 anthocyanins, 3 Chalcone, 9 Dihydroflavonoid, 5 Dihydroflavonol, 12 flavanols, 11 Flavone, 59 Flavonoid, 20 Flavonoidcarbonoside, 3 Flavonoids, 59 Flavonols and 4 Isoflavones. In the process of instrumental analysis, one QC sample is inserted into every 10 test and analysis samples to monitor the repeatability of the analysis process. Through the overlap display analysis of the TIC map of different quality control QC samples (Figure S1c), the results showed that the curve overlap of metabolite detection of total ion current was high, that is, the retention time and peak intensity were the same, indicating that the signal stability was good when the same sample was detected by mass spectrometry at different time points. All the detected metabolite content data were normalized by range method, and the accumulation patterns of metabolites among different samples were analyzed by cluster analysis (Hierarchical cluster analysis, HCA) by R software (www.r-project.org/) (Figure 3).
