Full scan data analysis
The data files raw acquired with the Masslynx 4.1 software (Masslynx 4.1, Waters) were transformed into .cdf files with Databridge tool. Chromatographic data files were processed using the software R (http://www.r‐project.org/). The XCMS algorithm (www.bioconductor.org; Smith et al., 2006) was used to obtain the peak peaking, grouping, and signal corrections. Metabolite amounts were analysed based on the normalized peak area units relative to the dry weight. To test the metabolomic differences between treatments, a nonparametric Kruskal–Wallis test (p < 0.01) was done. Partial least square discriminant analysis and heat map analysis were performed with the metaboAnalyst 4.0 (Chong et al., 2018). Adduct and isotope correction, filtering, clustering, exact mass mapping and metabolic pathway exploration was carried out with the packages MarVis filter, MarVis cluster and MarVis pathway that are integrated in the Marvis suit 2.0 (Kaever et al., 2014). Metabolite identification was carried out based on exact mass accuracy and fragmentation spectra matching with different online database. The database kegg (https://www.genome.jp/kegg/) was used for exact mass identity and for fragmentation spectrum analysis, the Massbank and the Metlin databases were used (www.massbank.jp; www.masspec.scripps.edu).