With the untargeted metabolomics approach using the UPLC-MS/MS detection platform, we detected a total of 1,144 metabolites in three experimental groups (Supplementary Tables S1–3). The principal component analysis (PCA) showed a segregating trend between the three samples (Figure 1A). Overall, there was 61.94% and 30.55% variation present on PC1 and PC2, respectively. A relatively higher Pearson’s correlation coefficient (PCC) was recorded between the replicates of each tissue type, i.e., tuber, stem, and leaf. A higher PCC was observed between CL and CS (0.82), followed by CS and CR (0.77), and CL and CR (0.62) (Figure 1B). This indicates that the metabolites accumulated in leaves and stems might be similar to some extent and could be different from those of tubers. The diverse set of detected metabolites could be grouped into thirteen different classes (Figure 1C). The metabolites were screened by combining the p-value or the fold change values. We considered the compounds as differentially accumulated metabolites (DAMs) if the variable importance in projection (VIP) was ≥1. Overall, we observed 849, 799, and 535 DAMs between CR vs. CL, CR vs. CS, and CL vs. CS, respectively. Of these, 345 metabolites were common among all the tissues (Figure 1D).

The highest 10 accumulated metabolites in CR belonged to three classes, i.e., amino acids and derivatives, flavonoids, and organic acids. In contrast, eight of the highest 10 accumulated metabolites in CL were flavonoids, whereas the other two were phenolic acids (Figure 2). The DAMs between CR and CL were enriched in flavonoid biosynthesis, flavone and flavonol biosynthesis, and tryptophan metabolism pathways. When we compared the DAMs between CR and CS, the top 10 most accumulated metabolites belonged to amino acids and derivatives, free fatty acids, organic acids, phenolic acids, saccharides, alcohols, and vitamins. In contrast, alkaloids were the top accumulated group in CS. The DAMs between CR vs. CS were enriched in flavonoid biosynthesis, flavone and flavonol biosynthesis, and the purine metabolism pathway (Figure 2). Flavonoids, lignans and coumarins, organic acids, and phenolic acids were the most enriched metabolites in CL whereas, in CS, alkaloids, terpenoids, organic acids, amino acids and derivatives, and flavonoids were the top 10 accumulated metabolites as compared to CL. The DAMs were enriched in the same pathways as CR vs. CL. In addition, phenylpropanoid biosynthesis, isoflavonoid biosynthesis, and arginine biosynthesis pathways were enriched (Figure 2).

We searched for tissue-specific metabolites within every three comparisons, i.e., CR vs. CS, CR vs. CL, and CS vs. CL, and generated a Venn diagram (Supplementary Figure S1). Of these tissue-specific accumulated metabolites, 322, 8, and 27 were common between CS and CL, CL and CR, and CR and CS, respectively. There were 14, 33, and 4 metabolites specific to CS, CL, and CR, respectively. The 33 CL-specific metabolites were classified as lignins and coumarins, phenolic acids, and alkaloids. Apart from these metabolites, we studied the pathways to which DAMs were significantly enriched. In particular, we focused on the common and unique pathways within each three-tissue comparison. The details are provided as follows.

Nine RNAseq libraries generated clean reads for each tissue ranging from 39,278,448 to 40,712,104 (average 40,174,095 reads). The average Q20 and Q30 rates were 93.96% and 97.86%, respectively. The average GC% for the libraries was 46.11% (Supplementary Table S4). Using the high-quality reads, the transcripts were assembled leading to the identification of unigenes with an average length of 748.50 bp and N50 length of 1,197 bp. In total, 123,024 unigenes could be annotated in different databases (Figure 5A). Maximum PCC based on fragments per kilobase of transcript per million fragments mapped (FPKM) values was noted between the replicates for each tissue while there was ≤0.53 indicating that the sampling was reliable (Figure 5B). PCA showed the grouping of the same tissue replicates together, whereas the PC1 and PC2 explained 43% and 13.9% variations between the samples, respectively (Figure 5C). Overall, the FPKM for CS was higher followed by CR and CL (Figure 5D).

The comparative transcriptome analysis revealed that 68,605, 23,606, and 50,210 transcripts were differentially expressed between CR vs. CS, CR vs. CL, and CL vs. CS, respectively (Supplementary Figure S2; Figure 6). A relatively higher number of genes were upregulated in CS, CL, and CS as compared to CR, CR, and CL, respectively (Figure 6). The KEGG classification showed that the genes could be classified into five major categories, i.e., cellular processes, environmental information processing, genetic information processing, metabolism, and organismal systems (Supplementary Figure S3A). On the other hand, GO annotation classified the annotated transcripts into three major categories, i.e., biological process, cellular components, and molecular function. Most of the genes were categorized as molecular functions (Supplementary Figure S3B).

The KEGG pathways enrichment analysis showed that the DEGs could be mapped onto different pathways. The DEGs between CR vs. CL and CL vs. CS were enriched in multiple pathways involved in signaling, growth and development, and biosynthesis of important secondary metabolites. The DEGs between CR vs. CS were enriched in the biosynthesis of unsaturated fatty acids, betalain biosynthesis, cutin submarine, and wax biosynthesis. Apart from these three pathways, DEGs were also enriched in multiple other pathways (Figure 6). The pathway-specific results are presented as follows.

To further understand the link between the DAMs and DEGs, we performed a co-joint analysis on both omics data. The combined KEGG pathway enrichment analysis showed that the DEGs and DAMs were enriched in multiple pathways. In particular, we noted that the DEGs and DAMs were enriched in the flavone and flavanol biosynthesis pathways (p-value < 0.05 and <0.01). Further, other pathways related to flavonoid biosynthesis were also enriched, further confirming the individual metabolome and transcriptome KEGG pathway enrichment results for all three comparisons (Figures 7A–C). Log2 conversion data for both metabolites and genes having PCC > 0.8 were selected and nine-quadrant diagrams were generated for each comparison (Figures 7D–F). The DEGs and DAMs in the 3^rd and 7^th quadrants for each comparison indicate that the differential expression pattern of the genes and differential accumulation of metabolites are consistent. This indicates that the changes in metabolite accumulation may be positively regulated by the genes in these quadrants. Since there are tens of pathways to which DEGs and DAMs were enriched, for brevity we have presented the results related to the flavonoid biosynthesis pathway as follows. We chose the flavonoid biosynthesis pathway due to its significant enrichment in all the tissue comparisons. Another reason for choosing this pathway for elaboration was the higher accumulation of flavonoids in leaves and stems as compared to tubers.

We found that ten DEGs (F3H, CYP73A, CCOAOMT, CHS, CHI, FLS, CYP75B1, CYP98A (C3′H), HCT, and DFR) were either positively or negatively correlated with 24 DAMs in the three comparisons (Figure 7; Supplementary Table S6). This observation indicates that the expression of multiple transcripts (or the same) of the same gene affects the accumulation of different metabolites. For example, CYP73A showed a positive correlation with 16 metabolites that were enriched in the flavonoid biosynthesis pathway. This means that the higher expression of CYP73A can lead to a higher accumulation of the positively correlated DAMs. PCC between DAMs and DEGs enriched in the flavonoid biosynthesis pathway ranging from −1 to 1 (Supplementary Table S6).

Limited metabolome research on the Codonopsis species offers unique challenges as well as opportunities for researchers to characterize the useful compounds. Although Tibetan herbal medicine and Chinese traditional medicine have reported on the use of C. convolvulacea in different treatments (Guang-Xuan et al., 2001; Tang et al., 2017), it is not known which specific metabolites are present in different plant parts. Our metabolome profiling of the tuber, stem, and leaves presents interesting results that will lead toward specialized usage of these plant organs in traditional medicine. In particular, our results that the leaves and stems are a rich source of flavonoids as compared to roots are interesting because previously the above-ground parts of some Codonopsis species have been considered as waste. These results are consistent with the earlier report on the metabolome profile of leaves and stems of C. pilosula (Zeng et al., 2021). Furthermore, the usefulness of flavonoids in curing a range of diseases (Yao et al., 2004) due to their bioactivity (Robak and Gryglewski, 1996) indicates that C. convolvulacea plants as a whole and their parts are of interest to both the traditional and modern medicine industry. In addition to flavonoids, the stems and leaves were also rich in L-tryptophan (or its derivative compounds, e.g., tryptamine), indole, anthranilic acid, and L-phenylalanine. L-Tryptophan is an essential amino acid and is used to treat insomnia, depression, anxiety, premenstrual dysphoric disorder, and teeth grinding and is used to improve athletic performance (Friedman, 2018). This means that either the leaves and stems as a whole or their extracts can be valuable plant parts and should not be discarded. The tubers of C. convolvulacea, as found in our results, were rich in amino acids and derivatives, free fatty acids, organic acids, phenolic acids, saccharides, alcohols, and vitamins. This is consistent with the earlier reports on C. lanceolata and C. peninsula (Du et al., 2018; Zheng et al., 2018). The compositional analyses of the roots/tubers of these two species have also shown the higher accumulation (or the presence) of amino acids, vitamins, phenolic acids, saccharides, and alcohols. The accumulation of these compounds in excess in C. convolvulacea tubers indicates their potential for extraction and medicinal uses. This is probably the reason for the higher use of tubers instead of stems and leaves of C. convolvulacea in traditional Tibetan medicine and the northwestern United States to treat the Kashin–Beck disease (Malaisse and Mathieu, 2008). Overall, these datasets not only are useful to the public and researchers but also add resources for the policymakers to consider these traditional sources from a newer perspective within the framework of the “Healthy China 2030” (http://en.nhc.gov.cn/HealthyChinaActionPlan.html) plan of the Chinese government.

Determination of the flavonoid content and the mechanisms governing their biosynthesis in dietary sources is an active topic for the last few decades. This is due to their health benefits that have been revealed through epidemiological studies (Yao et al., 2004). Their health beneficiary effects and human dietary intake as antioxidants stress the need to find novel flavonoid sources and evaluate them. Our results that the leaves, stems, and tubers have flavonoids are important from the applied and basic research perspective. Transcriptome sequencing is an effective tool to possibly understand the differential accumulation of flavonoids and identify novel candidate genes for gene manipulation. The higher accumulation of chrysin, galangin, pinobanksin, betein, phlorizin, p-coumaroyl quinic acid, luteolin, luteoforol, quercetin, vitexin, naringenin chalcone, naringenin, and sakuranetin in CL and CS as compared to CR (Supplementary Table S5) can be linked to the higher expressions of the respective genes (Figures 8 and 9). In particular, we can relate the observed changes in the isoflavone content to the expressions of CHI, CYP73A, CYP98A (C3′H), F3H, CYP75B1, ANS, and FLS. Moreover, the higher isoflavone content in CL can be related to the higher expression of CHS since it converts p-coumaroyl-CoA to isoliquiritigenin (Feinbaum and Ausubel, 1988), which is subsequently converted to different isoflavones (Du et al., 2010). In a similar way, its expression also correlates with the accumulation of naringenin chalcone and naringenin (Yahyaa et al., 2017). The higher contents of quercetin in CS and CL as compared to CR are due to the increased expressions of the ANS, FLS, CYP75B1, and the upstream genes mentioned above (Figure 9). Quercetin is synthesized from kaemferol or dihydroquercetin by the action of FLS or ANS, respectively. Functional characterization of CYP75B1 in Camellia sinensis has shown its role in flavonoid biosynthesis (Wang et al., 2014). On the other hand, the characterization of FLS in Citrus unshiu revealed its role in the conversion of dihydrokaempferol to kaempferol and dihydroquercetin to quercetin (Wellmann et al., 2002). Based on these results, we can propose that the expression of flavonoid biosynthesis-related genes, i.e., CHI, CHS, CYP73A, CYP98A, C3′H, F3H, CYP75B1, ANS, and FLS is important and these genes should be characterized under different growing conditions and at different plant stages to determine the most suitable harvesting time to achieve higher flavonoid production.

Tryptophan, being an essential plant-derived amino acid, has significant value in human nutrition. Tryptophan and the bioactive metabolites produced from it have the potential to contribute to a broad range of health conditions, e.g., autism, cardiovascular problems, depression, multiple sclerosis, kidney diseases, and cognitive function (Friedman, 2018). Apart from these functions, tryptophan also shows antimicrobial, antidiabetic, and anti-inflammatory functions (Nayak and Buttar, 2015). Thus, finding novel plant sources for tryptophan is an ongoing task. Our findings that CL had higher indole accumulation followed by CS and CR are interesting and provide information on the use of specific tissues. It is a product of the action of tryptophanase (DDC) on L-tryptophan (Snell, 1975) and this could be due to the upregulation of one DDC (TRINITY_DN5862_c0_g1_i10) in leaves as compared to CS. Tryptophan is also converted into tryptamine by the action of DDCs. The highest accumulation of tryptamine in CS can be linked to the sole expression of five TDC/DDCs (aromatic-L-amino-acid/L-tryptophan decarboxylase; TRINITY_DN50728_c0_g1_i2, TRINITY_DN68814_c0_g2_i1, TRINITY_DN68814_c0_g1_i1, TRINITY_DN178280_c0_g1_i1, and TRINITY_DN79687_c0_g2_i2). This is consistent with the long-known role of TDC/DDC when expressed in tobacco (Songstad et al., 1990). Not only these but also the genes present in the upstream pathways, i.e., photosynthesis and phenylpropanoid biosynthesis pathways, are also a possible possibility to reason for the higher accumulation of most of the metabolites enriched in the tryptophan metabolism pathway. Research on the exogenous application of tryptophan (and other amino acids) improves photosynthetic assimilation (Khan et al., 2019). However, understandably, higher photosynthesis would result in higher biosynthesis of tryptophan or other related metabolites but specific research on C. convolvulacea would reveal further details. The genes enriched in photosynthesis and phenylpropanoid biosynthesis pathways are ideal candidates for higher tryptophan accumulation. Furthermore, it would be possible to avoid/reduce tryptophan conversion to indole and tryptamine by downregulating DDCs and TDC/DDCs in leaves and stems.

To develop a highly diversified medicine portfolio for human and animal use, plant-based natural products offer a useful resource (Ganesan, 2008). In this regard, alkaloids play a dual role, i.e., in-planta they protect plants from invading pathogens as well as regulate growth, whereas they offer a range of health benefits (Kurek, 2019). Our findings that both the C. convolvulacea leaves and stems had a higher accumulation of alkaloids are interesting from the usage point of view. Besides, these two plant parts appear a viable raw material for the extraction of alkaloids and other metabolites accumulated in them for the commercial medicine industry. Stem-specific expression of a large number of DDCs indicates that L-tyrosine, in the stem, is being converted to tyramine and dopamine. It is known that DDCs are also responsible for these conversions apart from their role in tryptamine biosynthesis as reported in the above section (Facchini et al., 2000). The expression of DDCs and higher concentrations of tyramine in CS are consistent (Table 1; Supplementary Table S5). The higher expressions of ASP5 (De la Torre et al., 2014), ARO8 (Hirata et al., 2012), GOT, (Murooka et al., 2002), and AOC3 further indicate the conversion of L-tryptophan and L-tyrosine to downstream metabolites (Tipping and Mcpherson, 1995). Relative to lower expressions of these genes in CL are consistent with the respective metabolite accumulation as compared to CS. From these observations, we can understand that leaves could be a usable source for alkaloids in addition to the compounds discussed in the above sections, i.e., flavonoids. The stem also may act as an additional source of alkaloids and flavonoids. But it is also important to keep in mind that in the stem, these compounds might have been detected as they were being transported from tuber to leaves or vice versa. In contrast, CS specific expression of a large number of DEGs, e.g., HPD, 4CL, wrbA, HPDs, CYP73A, and ARO8 (Supplementary Table S5), excludes this assumption and suggests biosynthesis or interconversion of these compounds in the stem. However, these assumptions need further research.

C. convolvulacea and other species in this genus have long been used in traditional Chinese and Tibetan medicine (Hong, 2015). Our metabolome results highlight the significance of C. convolvulacea from traditional medicine’s point of view as well as from its utility as a raw material in the plant-based commercial production of flavonoids, tryptophan, amino acids, and alkaloids. However, since this research is a pioneering attempt in this species, many questions should be answered for its utility. In particular, future studies should focus on the age-specific detection of metabolites in different plant parts and how these metabolites accumulate with the aging of the plants. Furthermore, differential accumulation of these metabolites would need studies focusing on gene-specific characterization experiments to elucidate the roles of individual genes. Since commercial use of this species as raw material would need its cultivation on farmland, it would be interesting to know how the accumulation of different compounds will be affected under different growing conditions and biotic and abiotic stresses. An interesting research avenue could be finding ways to enhance the biosynthesis of compounds of interest. One strategy could be the application of abiotic stress at specific plant growth stages. One approach, which has been investigated in different plant species, could be the increasing endogenous plant hormone signaling and/or exogenous application of specific hormones (Ji et al., 2019). Likewise, it is known that abiotic stresses result in the increased accumulation of alkaloids and other metabolites (Kumar and Sharma, 2018; Jan et al., 2021). Can we use a similar strategy in C. convolvulacea? From these perspectives, our combined metabolome and transcriptome study provides an initial but detailed understanding of the pathways related to alkaloids, flavonoids, tryptophan, and other metabolites.

To determine the links between transcriptome and metabolome datasets, we also analyzed both results jointly. First, we performed a combined KEGG pathway enrichment analysis to determine the degree of enrichment of KEGG pathways. We then used the Corson program in R to calculate PCC between DEGs and DAMs. The PCC was then used to construct the transcript–metabolite network.