3.3 Root tissue metabolites
The Partial Least Squares Discrimination Analysis (PLS-DA) model separated the samples from the two planting treatments. PLS-DA is a supervised discriminant analysis statistical method, which uses partial least squares regression (Boulesteix et al., 2007) to establish the relationship model between metabolite expression and sample category. PC1 and PC2 were the scores of the test samples in the first and second principal components, respectively. From Fig. 7, it was clear that the two planting density treatments had a certain degree of differentiation; hence, the subsequent data analysis was reliable.
The threshold values of Variable Importance in the Projection (VIP) >1.0, Fold Change (FC) >1.5 or <0.667 and P value< 0.05 were set as the screening criteria for significantly different metabolites. A total of 51 metabolites with significant differences were found. Plotting all metabolites of the low-density (L) vs high-density (H) treatments in the volcano map can help us quickly find the differences in expression of root tissue metabolites. Among the 51 different metabolites, 28 were significantly up-regulated and 23 were significantly down-regulated, with a relatively small difference (Fig. 7b). We selected the top 10 metabolites with the largest differences for the annotation study. These top 10 differential metabolites were: schisandrin C (up), s7p (down), pelargonidin chloride (down), N-feruloylspermidine (up), tyrosol (down), acetylharpagide (down), KMH (up), kanamycin (up), 3-ureidopropionic acid (up), and eicosapentaenoic acid (down) (Fig. 7c).
All the differentially expressed metabolites in the low-density vs high-density treatments were put into the KEGG database for annotations. The 14 annotated differential metabolites and the involved metabolic pathways are listed in Table 3. There were five up-regulated and nine down-regulated metabolites.
It can be seen from Fig. 7c that the metabolic pathways with the lowest P values was biosynthesis of unsaturated fatty acids, and the differential metabolites within the pathways were eicosapentaenoic acid and docosahexaenoic acid. Both eicosapentaenoic and docosahexaenoic acids were down-regulated.
The metabolite S-adenosylmethionine was the most enriched metabolite in all metabolic pathways, so we focused on it. The mean relative content of S-adenosylmethionine was 161 × 105 in the low-density treatment (LD) and 71 × 105 in the high-density treatment (HD), indicating the content of S-adenosylmethionine in the LD was twice that in the HD. The mean relative content of docosahexaenoic acids was 2971 × 105 in the LD treatment and 7183 × 105 in the HD treatment, i.e., the content of docosahexaenoic acids in the HD treatment was more than twice that in the LD treatment.