The spatiotemporal regulations of epicatechin biosynthesis under normal flowering and the continuous inflorescence removal treatment in Fagopyrum dibotrys

Flowering is a critical physiological change that interferes with not only biomass yield but also secondary metabolism, such as the biosynthesis of flavonoids, in rhizome/root plants. The continuous inflorescence removal (CIR) treatment is frequently conducted to weaken this effect. Fagopyrum dibotrys (D.Don) H.Hara (Golden buckwheat) is a kind of rhizome medicinal plant rich in flavonoids and is widely used for the treatment of lung diseases. The CIR treatment is usually conducted in F. dibotrys because of its excessive reproductive growth. To uncover the molecular mechanisms, comprehensive analysis was performed using metabolome and transcriptome data obtained from normally bloomed and the CIR treated plants. Metabolome results demonstrated that in the rhizomes of F. dibotrys, its bioactive compound called epicatechin has higher amount than most of the detected precursors. Compared with the normally bloomed plants, the level of epicatechin in the rhizomes of the CIR group increased by 25% at the withering stage. Based on 96 samples of the control and the CIR groups at 4 flowering stages for 4 tissues, RNA-Seq results revealed a 3 ~ 5 times upregulations of all the key enzyme genes involved in the biosynthesis of epicatechin in both time (from the bud stage to the withering stage) and spatial dimensions (from the top of branch to rhizome) under the CIR treatment compared to normal flowering. Integrated analysis of LC–MS/MS and transcriptome revealed the key roles of several key enzyme genes besides anthocyanidin reductase (ANR). A total of 93 transcription factors were identified to co-expressed with the genes in epicatechin biosynthetic pathway. The flowering activator SQUAMOSA promoter-binding protein like (SPLs) exhibited opposite spatiotemporal expression patterns to that of the epicatechin pathway genes; SPL3 could significantly co-express with all the key enzyme genes rather than the flowering repressor DELLA. Weighted gene co-expression network analysis (WGCNA) further confirmed the correlations among chalcone synthases (CHSs), chalcone isomerases (CHIs), ANRs, SPLs and other transcription factors. SPL3 might dominantly mediate the effect of normal flowering and the CIR treatment on the biosynthesis of epicatechin in rhizomes mainly through the negative regulations of its key enzyme genes including CHS, CHI and ANR.


Background
Flowering is a critical physiological change during reproductive growth that interferes with not only biomass yield but also secondary metabolism. Generally, the process of flowering is divided into 4 stages: the bud stage, the initial flowering stage, the flowering stage and the withering stage [1]. In Arabidopsis, numerous interwoven genetic pathways exemplified by vernalization, photoperiod/circadian clock, gibberellin, autonomous, and aging could regulate flowering process [2]. On physiological level, the above pathways can end at promotion of the flowering process by repression of vegetative growth, or by upregulation of reproductive growth [3]. The transcription factors (TFs) called SQUAMOSA PROMOTER-BIND-ING PROTEIN-LIKEs (SPLs), which is known as plant "aging genes", has been proved to play a dominate role during the whole process of vegetative-to-reproductive transition and flowering: as the plant gets mature and progresses toward flowering, the expressions of SPLs are programmed to increase till the withering stage [4]. The repressor of gibberellin signaling pathway called DELLA could interferes with SPL transcriptional activity and consequently delays floral transition [5].
Recently, vegetative and reproductive growth were found to have different effects on the biosynthesis of secondary metabolites in plants. The accumulation rate of many bioactive compounds decreased dramatically in roots when the plants stepped into reproductive stage [6,7]. For instance, the biosynthesis of anthocyanin, one of the products of flavonoid pathway, could be improved by vegetative growth but repressed by reproductive growth through the interaction network involving SPLs, DELLAs, ABIs, dihydroflavonol 4-reductases (DFRs) or MYBs [8][9][10][11][12]. The abundance of SPL9/10/13 in the inflorescences could repress the accumulation of anthocyanin by directly down-regulating its key enzyme genes. Notably, for the Arabidopsis plants at flowering stage, anthocyanin has been found to accumulate in an acropetal manner with the highest level at the junction between rosette and stem, and the lowest level at the inflorescence top, and this pattern was regulated by SPL genes through the opposite spatial expression patterns; the high levels of SPL9 in the inflorescences repress anthocyanin accumulation by directly preventing expression of anthocyanin biosynthetic genes, such as anthocyanidin synthase (ANS), flavonoid 3'-hydroxylase (F3'H), DFR and UGT75C1, through destabilization of the MYB-bHLH-WD40 transcriptional activation complex [12]. Therefore, it is indicated that spatially, the repressing gradient of the genes in flavonoid pathway by the differential expressions of SPLs might be one of the important molecular mechanisms for the effects of flowering on secondary metabolism in plant.
In root/rhizome medicinal plants, especially for the perennial herbs belonging to Polygonaceae, Asteraceae or Apiaceae, the genus with a large number of branched stems and inflorescences, the treatment of inflorescence removal/continuous inflorescence removal (CIR) is usually applied to weaken the negative effects of reproductive growth on the yield and quality of roots/rhizomes [13,14]. It was proved that these treatments have great effects on the biosynthesis of some secondary metabolites. In Bupleurum chinense and Panax notoginseng plants, these treatments significantly improved the biosynthesis of triterpenoid in the roots/rhizomes by activating the metabolomic acid (MVA) and methylerythritol phosphate (MEP) pathways, but decrease the accumulation of flavonoids in P. notoginseng [13,15]. These studies indicated that the regulation of normal flowering or the CIR treatment on the secondary metabolism in rhizomes/roots might depend on metabolites or plant species. However, the underlying molecular mechanism have been poorly studied.
Fagopyrum dibotrys (D. Don) H.Hara (Golden buckwheat) is one of important medicinal plants which mainly distributed in China, India, Thailand and Vietnam [16]. Its dried rhizomes which called Fagopyri Dibotryis Rhizoma is an important crude drug and functional food that widely used for the treatment of lung disease, dysentery, rheumatism and throat inflammation [17]. Flavonoids have been shown to be the major active ingredients in F. dibotrys. The content of flavonoids in F. dibotrys is higher than that found in other buckwheat cultivars [18]. As a member of the genus Polygonaceae, one of the biggest problems in the cultivation of F. dibotrys is the excessive reproductive growth due to the large number of branched stems and inflorescences [19]. The biomass of the reproductive organs account for approximately 30% of the whole plant [20]. Moreover, its bioactive compound called epicatechin, one of the products of flavonoid pathway, could also be synthesized in nonmedicinal organs such as flowers or leaves, resulting in serious competition in the biosynthesis of epicatechin between aboveground organs and rhizome [21]. To enhance the biosynthesis of epicatechin in rhizomes, the CIR treatment is usually conducted in F. dibotrys but with unclear molecular mechanism. Thus, we believe that F. dibotrys could be used as a model rhizome medicinal plant to investigate the molecular regulations of normal flowering and the CIR treatment on the biosynthesis of epicatechin or other flavonoids.
In this study, we aimed to investigate the spatiotemporal regulations of the biosynthesis of epicatechin by normal flowering and the CIR treatment in F. dibotry. Metabolomics and transcriptome were performed, and comprehensively analyzed. All the results would provide new insight into the molecular mechanism of normal flowering-or the CIR-regulated biosynthesis of epicatechin or other flavonoids in rhizome/root medicinal plants.
The morphologic characteristics of floral organs at the 4 sampling stages were shown in Fig. 1a. A total of 75 days of treatment were conducted after the appearance of floral bud. The levels of epicatechin in the dried rhizomes were detected by LC-MS/MS and the extracted ion chromatograms of the reference substance of epicatechin and the samples were shown in Fig. 1b. As expected, the average content of epicatechin in the CIR group was significantly higher than that of the normal flowering group in each stage, and finally reached to 0.73 ± 0.002%, which improved by 25% in comparison with the normally bloomed plants (Fig. 1c).
At the bud stage (BD), the difference between the two groups reached to the maximum of 7 times. We speculated the reason was: the CIR treatment started at Aug 21 when the buds firstly appeared; while the sampling time of the bud stage was 7 days after inflorescence appearance (Aug 28) when most of the buds appeared. Thus, this one-week CIR treatment might be critical for the accumulation of epicatechin in rhizomes. During the following 3 stages, the increasement of epicatechin slowed down in the normal flowering group, but kept increasing till the withering stage (W) in the CIR group, indicating the great effects of the CIR treatment on the biosynthesis of epicatechin in the rhizomes of F. dibotry.

RNA Sequencing, assembly and annotation
To explore the spatiotemporal regulation of the biosynthesis of epicatechin by normal flowering and the CIR treatment in F. dibotrys, 96 RNA-Seq libraries (2 groups, 4 stages, 4 different tissues, 3 biological replicates) were constructed. After sequencing, a total of 20.13 ~ 29.96 million raw reads were generated. After removing the reads that contained adapter or poly-N and the low quality, 19 Table S5).

PCA and cluster analysis of the relationships among the sequencing samples
PCA analysis was performed to study the relationships among the 32 F. dibotrys samples (after taking the average coverage of the three biological replicates). As shown in Fig. 2a, all the samples were grouped into three clusters: the first group was the branch samples from the first 3 stages (BD, IF, F); the second was the branch samples from the W stage; the third was all the rhizome samples, which indicated the significant difference between branch and rhizome samples. The H-cluster analysis further showed that the overall expression profiles of rhizome samples were opposite to that of the branch samples; and the global expression levels in the rhizome samples were apparently higher or lower than that of the branch samples, indicating the higher sensitivity of rhizome to flowering or the CIR treatment (Fig. 2b).

GO and KEGG enrichment of DEGs
Using the threshold of the adjusted P value < 0.05 and |log 2 FoldChange|> 1, a total of 614 ~ 8,222 DEGs were finally identified in each comparison, based on the average coverage of three biological replicates (Supplementary Table S6). In comparison with the normal flowering group, GO enrichment analysis showed that the most abundant DEGs in the CIR group were identified in "biological process" and the "secondary metabolic process" pathways ( Supplementary Fig. S1).
Taken the group of normal flowering as the control, the top 10 KEGG pathways with the most significant enrichment DEGs in the CIR groups were assigned in plant hormone signal transduction, phenylpropanoid biosynthesis, metabolism of sucrose and sugar, etc., flavonoid biosynthesis was ranked the fifth (Fig. 2c, Control & CIR). Moreover, the DEG numbers from the top 10 KEGG pathways gradually decreased from the top of the branch to the rhizome, which indicated that the closer to the rhizome, the more similar for the genes that being responsive to normal flowering and the CIR treatment.
Within the group of normal flowering, when taking the sample T (the top part of branch) as the control, the number of DEGs in flavonoid pathway increased from the top of the branch to the rhizome (Fig. 2c, the Control group), which confirmed the higher sensitivity of rhizome to normal flowering. However, in the CIR group, except the bud stage, there were no significant differences on DEG numbers among different tissues (Fig. 2c, the CIR group), which indicated that the response of F. dibotrys to the CIR treatment tend to be the whole plant.

The different spatiotemporal regulations of the epicatechin biosynthetic pathway genes under normal flowering and the CIR treatment
According to our previous study in F. dibotrys, epicatechin is the product of flavonoid pathway and is synthesized from anthocyanin by the catalysis of ANR [17]. To uncover the spatiotemporal regulation of normal flowering and the CIR treatment on the biosynthesis of epicatechin, we analyzed the expression changes of all the key enzyme genes in the epicatechin biosynthesis pathway. For each group, taking the sample of the top part of the branch (T) as the control, we calculated whether there are any spatial expression changes in other tissues.
As shown in Fig. 3, except ANS, all the key enzyme genes of epicatechin biosynthetic pathway including PAL, C4H, 4CL, CHS, CHI, F3'H, F3H, DFR and ANR were identified. All of them exhibited different spatial expression patterns in response to normal flowering and the CIR treatment. Under normal condition, all the key enzyme genes were down-regulated spatially at all stages, with the sharpest decline in the rhizomes. Although the expressions of CHS, CHI and ANR were improved at the flowering (F) stage, the increasement was 1 ~ 3 times lower than that of the CIR group (Supplementary Table  S7).
On the contrary, under the CIR treatment, all the key enzyme genes including cinnamate 4-hydroxylase (C4H), CHS, ANR, PAL and flavanone 3-hydroxylase (F3H) were expressed highly during the whole process. In spatial dimension, C4H, CHS and ANR exhibited clearly gradient of up-regulations which increased 3 ~ 5 times from the top of branch to the rhizome, especially at the initial flowering (IF) and the flowering (F) stages. Therefore, in comparison with the widely repressions under normal flowering, the expressions of all the key enzyme genes involved in epicatechin biosynthetic pathway were significantly enhanced by the CIR treatment in both time (from bud stage to harvest) and spatial dimensions (from the top of branch to rhizome).

Integrated analysis and co-expression analysis of the gene expression pattern and the epicatechin level
To verify the correlations between the gene expression pattern and the epicatechin level, the integrated analysis of LC-MS/MS and the rhizome transcriptome was further performed. Since the levels of epicatechin increased continuously during the whole process, while the expression patterns of most enzyme genes were upregulated until the flowering stage, no significant correlation was found between the two datasets. However, when we only focused on the first 3 stages, the expression patterns of 3 PALs, 2 C4Hs, 2 4CLs, 4 CHSs and 3 ANRs showed significant positive correlations with the level of epicatechin under both normal flowering and the CIR treatment (Table 1), which suggested the important roles of these PALs, C4Hs, 4CLs, CHSs and ANRs in the biosynthesis of epicatechin.
To identify the key TFs that potentially function in the biosynthesis of epicatechin, we also performed the co-expression analysis between all the differentially expressed TFs and the key enzyme genes in epicatechin biosynthetic pathway. Using the threshold of |r|> 0.8, a total of 93 TFs were identified to correlated with the epicatechin biosynthetic pathway genes. Among them, a total of 77, 10, 6, 5 and 5 TFs were related to CHS, ANR, F3H, DFR and 4CL, respectively (Supplementary Table S8). The TFs that positively co-expressed

The different spatiotemporal regulations of the flowering regulators under normal flowering and the CIR treatment
In Arabidopsis, it was revealed that, spatially, the expression levels of SPL decreased gradually from the top to the basal part of stem, while the flowering repressor DELLA could interfere with SPL's transcriptional activity. With the involvement of some stress-responsive genes or TFs, SPL and DELLA could inhibit or promote the biosynthesis of anthocyanin, respectively [8][9][10][11][12]. Epicatechin is also the product of flavonoid pathway, so we further investigated whether the effects of flowering or the CIR treatment on the biosynthesis of epicatechin is mediated by the above genes or TFs. For each group, the spatial expression changes were analyzed by the control of the sample T (the top part of branch). In total, we identified 26 differentially expressed SPLs belonging to SPL1/2/3/6/7/8/9/12/13, which were proved to promote phase change and accelerate flowering in Arabidopsis or switchgrass [25][26][27]. Most of them were up-regulated at the bud stage (BD) and the Fig. 3 The spatiotemporal expression patterns of the key enzyme DEGs involved in the biosynthesis of epicatechin. The left pattern indicated the key enzyme genes involved in the epicatechin biosynthesis pathway in F. dibotrys according to Chen et al. (2016) [17]; the right pattern showed the spatiotemporal expression changes of the above genes. The abbreviations of the stages, groups and tissues were consistent with Fig. 1 and Fig. 2. For each group at each stage, the expression level of the sample T (the top part of the branch) was set as the control, thus the expression changes of the sample M/B/R were indicated as M/T, B/T and R/T. In the heatmap, the different members of each enzyme genes were randomly numbered from ①, which was indicated in the heatmaps withering stage (W) and down-regulated at the initial flowering stage (IF) and the flowering stages (F), with the strongest response in the rhizome. In comparison with the normally bloomed plants at the same stage, the expression levels of SPL2/3/8/9 were down-regulated by 2 ~ 3 times in the rhizomes under the CIR treatment (Fig. 4, Supplementary Table S7). The expression patterns of SPL9 (Cluster-13504.18693) were also confirmed by qRT-PCR (Fig. 5).
Besides, four differentially expressed DELLAs belonging to GAI and DELLA1 were identified. In time dimension, DELLAs were down-regulated at the early stages (BD and IF) and up-regulated at the late stages (IF and W). However, the spatial expression gradient of all DELLAs were not as obvious as that of SPLs. Under the CIR treatment, the expression levels of DELLAs only decreased by 1 ~ 2 times than that of the normally bloomed plants. No significant differential expressions were found for the suppressor of ABI3, ABI5 and MYB33 (Fig. 4, Supplementary  Table S7). The expression patterns of DELLA3 as well as ANR were confirmed by qRT-PCR (Fig. 5).

Co-expression analysis and WGCNA of flowering regulators and epicatechin biosynthetic genes
To investigate the correlations between normal flowering or the CIR treatment and the biosynthesis of epicatechin, co-expression analysis was performed between DELLAs, SPLs and the genes involved in epicatechin biosynthetic pathway. According to transcriptome profiles (Fig. 4), two representative members that showed the most significant spatial differential expressions in the normal flowering group and the CIR group were selected for each family, respectively. As shown in Fig. 6a, SPL3, which has been known to strongly expressed in response to floral induction compared with other members [4], were found to negatively co-expressed with almost all the enzyme genes in epicatechin pathway as expected. SPL2/12 could negatively co-express with PALs, CHSs, F3Hs under normal flowering as expected, but positively co-expressed with those enzyme genes under the CIR treatment; opposite situations were found for C4Hs and 4CLs; for CHIs, DFRs and ANRs, the two groups showed identical results. No significant regular patterns were found for DELLAs.
We further performed WGCNA based on all the genes or TFs included in correlation analysis (Supplementary Table S8) and co-expression analysis (Fig. 6a). Consistent with Fig. 3 and Fig. 6a, SPLs played a much more important role in regulating the biosynthesis of epicatechin than DELLAs under either normal flowering or the CIR treatment. SPL3 was highly correlated with several enzyme genes that showed significant spatial expression changes in response to the CIR treatment, such as 4CL-4, CHI-1, CHS-7 and C4H-4. SPL3 was also directly correlated with the AUX/IAAinvolved auxin pathway; SNF2, MYB and bHLH were also engaged in the above regulations. While, SPL8 and SPL12 were mainly correlated with TFs, such as TRAF and bHLH, and some enzyme genes that mainly response to normal flowering, such as F3H-5 and 4CL-1.

Correlation of flowering to the spatiotemporal gene regulations of SPL-DELLA and epicatechin biosynthetic pathway
In this study, we identified 26 SPLs belonging to SPL1/2/3/6/7/8/9/12/13 in F. dibotrys (Fig. 4). With the cooperation of all the members, SPLs formed an expression gradient that increased from the top of branch to the rhizome under normal flowering, while the CIR treatment could eliminate or largely weaken this gradient ( Fig. 2c, Fig. 4, Fig. 6a). These results extended the range of the spatial expression gradient of SPLs from aboveground to underground [12], and indicated the strong effects of the flowering process and the CIR treatments on the secondary metabolism in the roots/rhizomes of plants.
Moreover, the spatiotemporal expression patterns of SPLs were converse to that of the epicatechin biosynthetic genes as expected (Fig. 3). However, considering the excessive number of members that involved, few SPLs could significantly co-express with the enzyme genes, which indicated that most of SPLs might involve in the regulation of epicatechin pathway only in particular stage or tissue. Besides SPL9/10/13 which was reported to inhibit the biosynthesis of some flavonoids like anthocyanin [10,12], we found that SPL3 showed the most significant negative correlations with all the enzyme genes in epicatechin biosynthetic pathway (Fig. 6). In Arabidopsis, SPL3 was proved to promote floral meristem identity transition [25]. In all, the above results might be an important reason why the biosynthesis of epicatechin is inhibited during normal flowering and enhanced after the CIR treatment (Fig. 7). DELLA didn't exhibit significant correlations with any enzyme genes of epicatechin biosynthetic pathway (Fig. 4, Fig. 6). We consider that although plant will bloom after the completion of phase change, there is still a lagging process of phase change in spatial dimension after flowering; SPL might still be the dominant regulator of this process, and further affects the biosynthesis of epicatechin, rather than DELLA or GA pathway. Consistently, in rice, it was revealed that the GA3ox2 loss-offunction mutation didn't affect the expression levels of miR156, which targets SPLs [3,28].
Notably, at whole-plant level, we also firstly revealed the spatiotemporal expression patterns of all the key enzyme genes in the epicatechin biosynthetic pathway in response to normal flowering and the CIR treatment. Among the key enzyme genes, CHS, CHI, 4CL and C4H could directly co-expressed with SPL3, which indicated that SPL3 might mediate the effect of normal flowering and the CIR treatment on the biosynthesis of epicatechin  Fig. 1 and Fig. 2. For each group at each stage, the expression level of the sample T (the top part of the branch) was set as the control. Three biological replications and three technical replications were conducted for each sample Fig. 6 Co-expression analysis and WGCNA of the genes/TFs involved in epicatechin pathway and flowering regulation. a Co-expression analysis between the flowering process (the flowering activator SPLs and the flowering repressor DELLAs) and the biosynthesis of epicatechin. The heatmap was generated based on the r value. The genes or TFs that showed spatial differential expressions at more than three stages were selected for analysis. The naming rules was identical to that of Fig. 3 and Fig. 4. b Weighted gene co-expression network analysis (WGCNA) of the SPLs, DELLAs and the epicatechin biosynthetic pathway genes. Low value of clustering coefficient was indicated by dark blue. The naming rules was identical to that of Fig. 3   in the rhizomes mainly through the negative regulations of the above enzyme genes. Moreover, it is also suggested that the levels of their products such as p-coumarate, chalcone or other flavonoids might also be improved by the CIR treatment. However, in P. notoginseng, the contents of flavonoids were decreased in roots after one-time inflorescence removal [15]. We considered the one-time treatment might not enough to trigger the SPL-dominated spatial phase change in plant. Therefore, it is indicated that the biosynthesis of flavonoids in the rhizome/ root was highly correlated to the state of vegetative or reproductive growth of plant.
The stress responsive genes/TFs, such as ABI5, ABI3 and MYB33, were proved involve in the interactions with SPLs and the accumulation of anthocyanin [8,9]; while they didn't show significant differential expression in this study (Fig. 4), which indicated that in F. dibotrys, the CIR treatment might not regulate the biosynthesis of epicatechin through abiotic stress pathways.

Structural genes and transcription factors that involved in epicatechin biosynthesis
In this study, we only focused on the level of epicatechin under normal flowering and the CIR treatment (Fig. 1). On the one hand, only epicatechin has been listed as the bioactive compound of F. dibotrys in Chinese Pharmacopoeia; on the other hand, our quantified metabolome data showed that, the amount of epicatechin was dozens or even hundreds of times higher than that of the most detected precursors (Supplementary  Table S2); moreover, epicatechin showed good stability during the 24 h stability test, and the modified products of epicatechin such as (-)-Epigallocatechin (EGC), (-)-Epicatechin gallate (ECG) or (-)-Epigallocatechin gallate (EGCG) cannot be detected because of the very limited contents (data not shown). These results confirmed that these products are not the main products in the rhizome of F. dibotrys [16,29].
In this study, the epicatechin level was most positively correlated with the expressions of ANRs, followed by CHSs (Table 1), which was consistent with our previous study that the ectopically high expressions of ANR and CHS were contributed for the accumulation of epicatechin in another mutant line of F. dibotrys [17], and confirmed again the pivotal roles of ANR and CHS in the biosynthesis of epicatechin in F. dibotrys. CHI has also been reported to reduced level of anthocyanin by up-regulating ANR [30,31]. However, in our integrated analysis, CHI didn't show significant correlation with epicatechin level (Table 1). Instead, as shown by integrated analysis, co-expression analysis and WGCNA, we uncovered the potential important roles of 4CL and DFR in the biosynthesis of epicatechin for the first time; the expression of 4CL was highly correlated with SPL and DELLA, while DFR was only correlated with ANR (Table 1, Fig. 5, 6).
A total of 93 TFs including SPL, MYB, bHLH, SNF2 were identified to be significantly correlated with ANR, CHS and 4CL, respectively (Supplementary Table S8). According to WGCNA, some identified SPLs showed significant correlations with CHS, CHI, 4CL and C4H (Fig. 6b), which was different with the previous studies that SPL could repress the biosynthesis of anthocyanin by prevent the expression of F3'H and DFR [12,32]. Thus, it was indicated that the regulations of the flavonoid pathway genes by SPL might depend on metabolites. AUX/IAAs were found to only correlated with ANR, C4H and SPL3 (Fig. 6b), which suggested the notable potential interactions between root growth and the biosynthesis of epicatechin, and provided new clue to our understanding of the importance of growth years in the quality control of medicinal plant. Other TFs such as MYB, SNF2 and bHLH could also directly correlated with ANR, CHS and CHI (Fig. 6b). In many model and horticultural plants, MYB and bHLH were widely demonstrated to involved in the regulations of ANR, LAR, FLS, etc. [33][34][35]. SWI/SNF complex was known to mediate ATP-dependent chromatin remodelling to regulate gene expression; several subunits of SWI/SNF were demonstrated to involve in the control of flowering and potentially correlated with DNA methylation [36]. Other TFs, such as BBR-BPC and Trihelix, might be the novel regulators in the epicatechin biosynthetic pathway.

Conclusions
In this study we found that the classic model of the spatial expression gradient of SPL appears to apply to flowering process in F. dibotrys. Moreover, the range of such spatial expression gradient could be extended from aboveground to underground, and further apply to all the key enzyme genes of epicatechin biosynthetic pathway in a reverse pattern. The CIR treatment could enhance the epicatechin level by sharply altering the spatiotemporal expression patterns of the above genes/TFs. SPL3 appears to mediate the effect of normal flowering and the CIR treatment on the biosynthesis of epicatechin in the rhizomes mainly through the negative regulations of its key enzyme genes including CHS, CHI, and C4H. Other TFs such as AUX/IAA, SNF2, MYB and bHLH were potentially engaged in the biosynthesis of epicatechin. Taken together, this study provides new insights into the molecular mechanism of flowering-regulated secondary metabolism in the roots/rhizomes of medicinal plants or other crops.

Plant materials and treatments
The F. dibotrys used in this study was the new line separated from M 2 generation of the γ-Co 60 irradiated Jiangsu germplasm. This new line exhibited red leaves and was authenticated by Dr. Ailian Li of the Institute of Medicinal Plant Development (IMPLAD). It was reproduced asexually and had stable characters [17,19].
The plants of both the control and the CIR groups were obtained by propagating from the one-year-old rhizome of this new line, a total of 30 plants were randomly selected for each group. All the plants were planted at April 16 of 2020 in the same plot of the experimental base of IMPLAD (lat. 39°47' N, long. 116°25' E, alt. 50 m).
Since August of 2020, we observed every day to see whether the flower bud appears. For the CIR group, as soon as the flower bud firstly appeared, the CIR treatment by removing the inflorescence with scissors was started. The treatment was carried out every day throughout the whole flowering period until the withering period, which continued two and a half months in total. All the plants in the CIR group were treated since August 21 (average temperature 22.5 ℃) and were sampled at different date: August 28 (average temperature 24.5 ℃) for the bud stage (BD), when the most branches are budding and the most buds are still green; October 7 (average temperature 15 ℃) for the initial flowering stage (IF), when the most buds are preparing to bloom (turns to white); October 20 (average temperature 12.5 ℃) for the full flowering stage (F), when the most buds bloom completely; and November 9 (average temperature 9.5 ℃) for the withering stage (W), when all the flowers withered and the rhizome was harvested. The inflorescence of the normal flowering group was saved.
For each stage, three plants from the control and the CIR groups were randomly selected, respectively. For each plant, the top part (T, the top of the branch, approximately 5 cm in length) of 3 branches were mixed as one sample; the middle part (M, at 1/2 of plant height, approximately 5 cm in length) and the basal part (B, approximately 5 cm in length) were sampled in the same way. The rhizome (R) of each plant was seen as one sample. For each rhizome sample, some part was used as the fresh sample for transcriptome sequencing, the other part was dried for the chemical analysis. In all, at each stage, there were 4 samples for each plant and 3 biological repeats for each group. A total of 96 fresh samples were used for transcriptome sequencing.

Metabolomics analysis
The harvested rhizomes from the normal flowering group were used as materials. After freeze-drying, the rhizomes were grinded into power and ultrasonicly extracted; the supernatants were taken after centrifugation (12,000 rpm, 15 min) and filtration. The target compounds were separated by EXON LC System (SCIEX) UPLC. Mass spectrometry analysis was performed by SCIEX 6500 Qtrap with MRM mode. Peak detection and identification were conducted with the database constructed by ourselves [15].

LC-MS/MS analysis of the contents of epicatechin
The reference substance of epicatechin was purchased from Beijing Jiachen Technology Co., Ltd. (Beijing, China) with purity ≥ 97% as determined by HPLC. HPLC-grade methanol and acetonitrile were purchased from Fisher (Emerson, IA, USA). Other reagents and chemicals were supplied by Sinopharm Chemical Reagent Beijing Co. Ltd (Shanghai, China). A MilliQ ® purification system (Millipore, Billerica, MA, USA) was used to prepare deionized water for all experiments.
According to Qiao et al. [37], approximately 2 g of powder from each rhizome sample were accurately weighed and were extracted by methanol. LC-MS/ms system consisting of an Agilent 1260 Infinity HPLC system (Agilent Technologies, USA) and an AB SCIEX 4500 QTRAP MS system (AB SCIEX, Canada) were performed for sample analysis. An Agilent Poroshell 120 SB C18 column (100 mm × 2.1 mm, 2.7 µ m) was used for the HPLC separations, and the mobile phase for gradient elution consisted of (A) water (containing 0.1% formic acid) and (B) acetonitrile with the following gradient procedure: 0 ~ 4.0 min 10 ~ 25% B, 4.5 ~ 7.0 min 90% B and 7.1 ~ 10.0 min 10% B, with a flow rate of 0.30 mL/min. The injection volume was 2.0 µL. The column effluent was monitored by 4500 QTRAP MS system. The electrospray ionization (ESI) in the negative-ion mode and multiple reaction monitoring (MRM) scanning was employed for quantification. The related source settings and instrument parameters were listed in Table 2. Chromatographic workstation (v1.6.2) was used for the data analysis.

RNA isolation and transcriptome library construction
Total RNA of each sample was extracted using the RNeasy plant mini kit (Qiagen, Germany) according to the manufacturer's instructions and was examined and quantified using a Nano Photometer spectrophotometer (IMPLEN, USA). A total amount of 1.5 μg of high-quality total RNA was used for transcriptome libraries construction, as described previously [38]. Then the libraries were sequenced using the Illumina HiSeq 2100 instrument (Novogene, China).

Transcriptome assembly and annotation
After sequencing, the clean reads were obtained by removing reads containing adapter, reads containing N base and low-quality reads from raw data. The value of Q20, Q30 and GC content of the clean reads were calculated. Then, the clean reads were De novo assembled using Trinity with the default parameters [39]. Corset hierarchical clustering was used to remove redundant sequences, and the longest cluster sequences were obtained after Corset hierarchical clustering for subsequent analysis. We used BUSCO software to test the quality of the assembled TRINITY.fasta, unigene.fa and cluster.fasta files, and evaluated the accuracy and integrity of the assembled results. Subsequently, using BLAST, the longest cluster sequences were annotated by Nr, Nt, KO, SwissProt, Pfam, GO and KOG databases with E-value cutoff of 10-5, respectively.

Differential expression analysis
Gene expression levels were calculated by FPKM (fragments per kilobase of transcript per million fragments mapped). Differential expression analysis of any two representative samples (three biological replicates per representative sample) was performed using the DESeq2 R package (v1.20.0). For each group in each stage, the expression level of the sample T (the top part of the branch) was set as the control; for KEGG analysis, we also took the normal flowering group as the control and analyzed the DEG enrhichment of the CIR group in each stage. The resulting P-values were adjusted using the Benjamini and Hochberg's approach from controlling the false discovery rate. Genes with ana adjusted P-value < 0.05 and |log 2 FoldChange|> 1 were assigned as differentially expressed [40].

GO and KEGG enrichment analysis of DEGs
GO enrichment analysis of DEGs was implemented by the clusterProfiler R package, in which gene length bias was corrected. GO terms with adjusted P-value < 0.05 were considered as significantly enriched DEGs. We also used cluster Profiler R package (v3.16) to test the statistical enrichment of DEGs assigned in KEGG pathways [41].

Corelation analysis between the transcriptome data and metabolome
For each rhizome sample, the data of epicatechin contents and the data of expression levels (FPKM) of genes involved in the biosynthesis of epicatechin (showed differentially expression at 3 stages or more) were used for correlation analysis, the Person correlation coefficient (r) was calculated.

Co-expression analysis and WGCNA
Among all the 96 samples, the expression levels (FPKM) of enzyme genes involved in the biosynthesis of epicatechin and TFs were calculated. The threshold of |r|> 0.6 was identified as significantly correlated.
Then, data processing was performed by using the WGCNA package in R-Studio 3.6.0 software. The average coverage of three biological replicates for 32 samples were used for analysis. An adjacency matrix was constructed using a soft threshold power of 12. The scale-free network was constructed using the blockwise Modules function, the procedure of module partition analysis and modules definition were according to Greenham et al. (2017) [42]. Based on the data of network node correlation profile of each module, the molecular interaction networks were visualized by Cytoscape software (v3.5) [43].