Figure 2. Comparative metabolomics using LCM combined GC-MS reveals metabolites associated with SC or non-SCs across naturalA. lancea accessions. (A) Clustering of metabolites identified in both SC and non-SCs across three natural accessions. (B) Clustering of SC metabolites identified in different tissues across three natural accessions. The heat-maps were generated based on a min-max normalization of metabolite content. (C) Result of PCA analysis of SC metabolites in four different tissues across three natural accessions.
Calling differential metabolitesby multivariate data analysis
To fish out the differential metabolites between red and non-red SCs, we carried out orthogonal partial least squares-discriminant analysis (OPLS-DA), a regression approach commonly applied to a two-group comparison of omics data (Trygg and Wold, 2002; Worley and Powers, 2016). Pairwise comparisons were conducted among the cortex SCs (n=6) of HAR (non-red SC), JAR (red SC) and SAR (red-SC). The cortex SCs in JAR (red SC) were also compared with phloem SCs in JAR (non-red SC). The resultant OPLS-DA models showed that the red SCs could be clearly separated from the compared non-red SCs (Figures 3A-D), suggesting potential differential metabolites underlying the red colour of AR. We next used the Variable Importance Parameter (VIP) to rank the differential metabolites in each comparison and identified putative differential compounds between the compared samples (Figures 3E-H). For example, in the comparison between cortex SCs in JAR and HAR, 11 differential metabolites (VIP >1) correlated to the red SCs (JAR), and 4 were associated with the non-red SCs (HAR). The full comparison results are summarized and shown as a Venn diagram to reveal the shared differential metabolites (Figure 3I). Three polyacetylenes (atractylodin, 1Z-atractylodin & 3E,5E,11E-tridecatriene-7,9-diyne-1,2-diacetate) are shared in all red versus non-red SC comparisons, but not in red (JAR-cortex) versus red (SAR-cortex) SC comparisons. We further verified this result by quantifying the atractylodin (standard was available, see figure S3) content among red and non-red SCs. Indeed, atractylodin was highly accumulated in JAR and SAR SCs in comparison to HAR SCs (Figure 3J). Notably, the pure atractylodin powder shows in red (Figure S3), similar to the cinnabar-like colour in AR SCs. To test the reliability of the strategy above, we collected SCs after LCM, extracted metabolites with 80% methanol and analyzed by LC-MS which is commonly used for analysis of polar compounds. In total, 43 compounds were detected. Multivariate data analysis revealed no correlation of metabolites associated with red SCs (Figure S4). Altogether, our data indicated that the three non-polar polyacetylenes are likely the causal compounds related to the red SCs inA. lancea natural accessions.