Exploring metabolic responses upon harvesting via untargeted metabolomics coupled with machine learning algorithm in rubber tree

Rubber tree (Hevea brasiliensis) is important for modern life due to its commercial product, natural rubber, which is indispensable for many important sectors. Unlike cereal crops, rubber trees can be harvested by tapping every several days for up to 40 years. In unharvested virgin trees, tapping stimulates rubber production. However, the underlying molecular basis of tapping-induced rubber production still remains elusive. To identify core tapping-responsive metabolic pathways and key marker metabolites for optimal rubber yield, we used three field-grown elite rubber tree cultivars with contrasting yields, and investigated their latex chemistry among 15 successive tappings by untargeted metabolomics coupled with random forest machine learning algorithm. A total of 367 metabolites were annotated. The metabolite profiles clearly mirrored the yielding traits of the three cultivars prior to and post the tapping treatment. Exploration of tapping-responsive differentially accumulated metabolites (DAMs) underlined the roles of sugar metabolism, amino acid and nucleotide metabolism and stress responses in tapping-induced rubber production in a cultivar-independent manner. Further analysis revealed the significance of energy production primed by sugar metabolism, stress-associated amino acid synthesis, and oxidative homeostasis in cultivar-dependent rubber production. Finally, 19 potential marker compounds were screened using the random forest regression model, in which 6 metabolites including fructose-6-phosphate and raffinose were identified as positive markers, whilst 13 metabolites including citric acid and fumaric acid were revealed as negative markers to predict high-yield cultivars. The specialized metabolites and pathways identified in this work will aid targeted rubber tree breeding in the future.