The OPLS-DA score scatter plots (left) and loading scatter plots (right) of shoot samples of (A) kappa-ox and kappa-KO (B) chi-ox and chi-KO. Wild type samples were used as controls. Each point represents an independent plant in the score scatter plots and an individual metabolite peak in the loading plots. (A) The OPLS-DA model of kappa samples shows two significant components, with R2X, R2Y and Q2Y values of 0.53, 0.95 and 0.65, respectively. (B) The OPLS-DA model of chi samples shows three significant components, with R2X, R2Y and Q2Y values of 0.31, 0.67 and 0.41, respectively. These models were validated using analysis of variance of cross-validated predictive residuals (CV-ANOVA) (pCV < 0.01). (Left) Black square, wild type; blue diamond, kappa-ox; pale-green star, kappa-KO; yellow triangle, chi-ox; green circle, psi-ox; pink-inverted triangle, chi-ox. (Right) Pale-green circle, amino acids; orange diamond, TCA intermediates; blue star, metabolites that consists of CHON-elemental composition; pink square, metabolites that consists of CHO-elemental composition; gray triangle, unclassified peaks. Number of biological replicates: wild type, n = 6; kappa-ox, n = 16; kappa-KO, n = 6; chi-ox, n = 16; chi-KO, n = 6; psi-ox, n = 17; psi-RNAi, n = 6. pCV, p-value of the probability level of the F-test in each model.