As mentioned in the literature review25,26, saponins have high medicinal value. However, QSs have two sides. Despite medicinal value, on the other hand, its bitterness reduces the taste of food, which brings a lot of trouble to the production of quinoa food. To solve this dilemma, many studies have focused on the desaponification treatment of quinoa seeds in food production. With the development of high-throughput sequencing technology, particularly completion of quinoa genome annotation23, genomics-assisted breeding has gradually been put on the agenda for the improvement of quinoa varieties. This study was designed to mine the key genes related to QSs metabolism by RNA-seq technologies and several statistical methods. These findings might help others to better understand the metabolic process and regulatory mechanism of QSs. The screened key genes can be used as promising targets for the molecular breeding of improved quinoa, which, in turn, would lay the foundation for the rational utilization of QSs.
The results of differential analysis revealed that there were more DEGs between developmental stages than between breeds. There were 19181 differential genes between the Yr group and Yl group (Fig. 2a); 20265 genes between the Yr group and Wl group (Fig. 2c); 4013 genes between the Yr group and Wr group (Fig. 2b). This may be because in addition to the differential expression of metabolic genes, the differences in the expression of structural genes were also highly significant than that between breeds. Through the intersection of three differentially expressed gene sets, most of the differentially expressed structural genes were eliminated, while the differentially expressed metabolic genes were maximally retained. We also carried out KEGG pathway analysis of the screened DEGs and found these genes were significantly enriched in metabolism-related pathways (Fig. S2), which confirmed our speculation.
Gene set enrichment analysis is useful for identifying genes closely related to saponin synthesis. We obtained 116 key candidate genes by performing GSEA. Among 116 genes, 56 genes were up-regulated and 60 were down-regulated between Yr group and Wr group (Table S2). Remarkably, we found that the majority of saponin metabolic pathway genes in Yr group compared to Wr group were down-regulated (Table S1), which was contrary to the up-regulation of saponin content. Consequently, we speculate that there are new metabolic pathways for saponin in yellow quinoa seeds in addition to the metabolic pathway to our knowledge. In 57 key genes analyzed by PCA, we found 22 genes encoding uncharacterized proteins closely related to the metabolism of quinoa saponin (marked by asterisks in Fig. 5), which partially corroborates our speculation.
In order to select key genes from 116 genes, principal component analysis was carried out (Fig. 4a). We filtered 57 key genes using PCA. Details about the expression of these genes are displayed in Fig. 4b and Fig. 5. As can be seen, of the 57 selected genes, 51 genes were up-regulated and 6 genes were down-regulated between Yr and Wr group. We speculated that these up-regulated genes were positive regulators of saponin biosynthesis, while down-regulated genes may play a role in negative regulators.
To identify key genes, hierarchical cluster analysis and PCA analysis were performed based on 57 genes. Hierarchical clustering of 12 quinoa samples according to the expression patterns of 57 key genes suggested that these genes could effectively discriminate four groups (Fig. 4b). PCA analysis showed that the expression of 57 key genes explained 84.19% of the variable (Fig. S3), which better explained the differences of the four groups of samples than that based on 116 genes (Fig. 4a).
There are still some weaknesses of this work. Firstly, we have not identified the 57 key genes in an experimental way. Secondly, in this study, one interesting finding was we also identified 10 lncRNAs. Previous studies have been suggested that, in plants, lncRNAs play important roles as regulators in gene silencing27, flowering time control28, organogenesis in roots29, photomorphogenesis in seedlings30, abiotic stress responses31,32, and reproduction33. What role these 10 lncRNAs play in the regulation of saponin metabolism that we need to explore further. Thirdly, in this study, due to incomplete information on genome annotation, we found some uncharacterized genes (Table S1 and Fig. 5), whose specific functions need to be further verified in the saponin biosynthesis process.