Multi-Omics Reveals Stage-Specific Flavonoid Accumulation in Bupleurum chinense DC. Inflorescences

Abstract

The large-scale cultivation of medicinal plants generates substantial agricultural by-products that are often discarded. A notable example is the floral biomass of Bupleurum chinense DC. (B. chinense), which is routinely removed during cultivation to promote root yield. To explore the potential valorization of these discarded tissues, we performed an integrated transcriptomic and metabolomic analysis of flavonoid biosynthesis across three developmental stages: F1 (Initial Flowering Stage), F2 (Full Bloom Stage), and F3 (Late Flowering Stage). Our results revealed distinct stage-specific regulatory dynamics. Flavonoid biosynthesis was initiated at F1 through the activation of upstream structural genes, reached its peak at F2 with strong up-regulation of branch-specific genes and the accumulation of diverse flavonols and anthocyanins, and declined at F3, despite the sustained presence of several antioxidant metabolites. These findings indicate that F2 represents the optimal stage for harvesting B. chinense flowers to obtain a broad spectrum of bioactive flavonoids, while late-stage flowers may serve as a complementary source of stable antioxidant compounds. Collectively, this study highlights the potential for transforming discarded floral biomass into valuable phytochemical resources and provides a framework for exploring underutilized tissues in other medicinal plants.

1. Introduction

Bupleurum chinense DC. (B. chinense), a member of the genus Bupleurum in the Apiaceae family, is a core medicinal plant widely utilized in East Asian traditional medicine, including Traditional Chinese Medicine and Kampo medicine [1]. Historically, its roots have been incorporated into classical formulations for the treatment of fever [2] and inflammation [3], reflecting a broad spectrum of clinical applications. The roots contain diverse bioactive compounds, including terpenoids [4,5], flavonoids [6], essential oils [7,8], and phenolic compounds [9], which contribute to antioxidant [10], anti-inflammatory [11,12], and hepatoprotective effects [13]. During the COVID-19 pandemic, preparations containing B. chinense, such as Qingfei Paidu Decoction [14], highlighted its relevance in contemporary public health applications of Chinese herbal medicine. In response to increased public awareness of health and growing demand for natural remedies, the main source of B. chinense has shifted from wild populations to cultivated production, establishing it as a significant specialty crop. Despite extensive research and commercialization of the roots, the potential value of other plant organs remains largely unexplored.

In agriculture and horticulture, non-traditional plant parts and processing by-products have garnered increasing attention due to their rich content of bioactive compounds. Numerous food crops provide illustrative examples, as materials typically discarded often contain health-promoting substances. For instance, grape skins and seeds are rich in proanthocyanidins [15], while tomato pomace contains high levels of lycopene and carotenoids [16]. These observations highlight the hidden nutritional and functional potential of plant components traditionally regarded as waste.

A similar trend is emerging in the study of medicinal plants. Although the root of Saposhnikovia divaricata is traditionally considered the medicinal portion, its leaves actually contain higher levels of 4′-O-β-D-glucosyl-5-O-methylvisamminol—a compound whose relevance in treating ulcerative colitis has been validated through network pharmacology [17]. In Sanguisorba officinalis, extracts from stems, leaves, and flowers are rich in polyphenols and flavonoids, exhibiting pharmacological activities comparable to those of the root, including inhibitory effects against helicobacter pylori [18]. Similarly, in Dendrobium officinale, flowers—typically categorized as non-medicinal parts—contain anthocyanins with potent antioxidant activities, such as radical scavenging and enhanced superoxide dismutase activity observed in vivo [19]. Collectively, these examples demonstrate that non-traditional plant parts can harbor bioactive compounds of comparable value to the traditionally used portions, thereby justifying their systematic investigation.

In the agricultural production of B. chinense, cultivation practices primarily aim to maximize root yield, as the roots are the recognized medicinal organ. To achieve this, growers often remove inflorescences during flowering—a practice known as “topping and inflorescence removal” [20]—to redirect nutrients to the underground biomass. While this practice can moderately improve root size and quality, it also results in the loss of a substantial portion of aboveground biomass. Previous studies indicate that B. chinense inflorescences can account for nearly 40% of total plant biomass [21]. However, this significant resource has long been overlooked for potential applications in the medicinal and food sectors, leading to considerable waste. This underutilization is particularly striking given that the inflorescence is rich in valuable flavonoid components such as rutin, quercetin, isorhamnetin, isoquercitrin, and kaempferol [22,23]. These compounds are well known for their nutritional and health-promoting properties: rutin helps lower blood sugar levels and exhibits antiplatelet aggregation activity [24,25]; quercetin demonstrates antiviral effects, inhibits lipid peroxidation, and suppresses pathogens such as Salmonella enterica [26,27]; and kaempferol reduces reactive oxygen species production and has shown efficacy against atherosclerosis [28,29]. Collectively, these properties underscore the potential value of B. chinense inflorescences. Consequently, a large proportion of the plant is discarded, representing not only agricultural waste but also a missed opportunity for applications in medicine, nutrition, and health-related industries. Despite this, systematic studies evaluating the chemical composition and functional value of B. chinense flowers remain limited.

In flowering plants, developmental stage plays a critical role in shaping phytochemical composition. Studies in other species have shown that levels of bioactive compounds fluctuate markedly during flower maturation. For example, in Lonicera japonica, the total content of phenolic acids and iridoids declines progressively throughout flowering [30]. Similarly, in Hibiscus syriacus, the concentration of several anthocyanins decreases substantially—by up to 64%—as flowers open [31]. These findings highlight the dynamic nature of floral metabolites and indicate that harvest timing strongly influences functional quality. However, for B. chinense, no systematic analysis has been conducted to elucidate how flavonoid accumulation varies across distinct floral developmental stages, and the optimal harvest time for maximizing flavonoid content remains unknown.

Despite its economic and medicinal importance, research on B. chinense has remained heavily focused on the roots, with limited attention paid to aerial tissues. Although preliminary reports indicate the presence of pharmacologically relevant flavonoids in the flowers, no integrated study has explored their metabolic dynamics or genetic regulation. Moreover, the developmental trajectory of flavonoid biosynthesis during flowering and its underlying molecular mechanisms remain unexplored.

To address these knowledge gaps, this study combines metabolomic and transcriptomic approaches to characterize flavonoid accumulation and gene expression profiles in B. chinense flowers across different developmental stages. By correlating metabolite variation with transcript-level regulation, we aim to elucidate the molecular mechanisms governing flavonoid biosynthesis during floral development. Specifically, the study focuses on three objectives: (1) profiling flavonoid composition throughout developmental stages; (2) identifying key regulatory genes in the flavonoid biosynthetic pathway; and (3) determining the optimal harvest time for B. chinense flowers to maximize flavonoid content and quality. The results provide both theoretical and experimental guidance for determining the optimal harvest period of B. chinense inflorescences, thereby enhancing their nutritional and medicinal value. Practically, these findings support the valorization of currently discarded aboveground biomass, improving the overall utilization efficiency of the plant. More broadly, the approaches developed here can serve as a model for other root-based medicinal species, where non-traditional plant parts are often neglected. By revealing the hidden potential of floral tissues, this research contributes to more sustainable and efficient strategies for medicinal plant production, in line with global efforts toward resource optimization and green health innovation.

2. Materials and Methods

2.1. Plant Materials and Experimental Treatments

Plant materials were collected from Beiyu Village, Jinan City, Shandong Province, China, which has a mean annual temperature of approximately 15 °C and an average annual precipitation of 600–700 mm. Inflorescences of B. chinense at different developmental stages were collected during the flowering period (June to September) in 2023 for transcriptomic and metabolomic analyses. Specifically, flowers at the F1 stage were sampled on 15 June 2023, while those at the F2 and F3 stages were collected 10 and 20 days after the initial sampling, respectively. Additional inflorescence samples were harvested from the same location between June and September 2024 to validate the expression levels of target genes via quantitative PCR (qPCR).

Flowering stages of B. chinense inflorescences were determined under a stereomicroscope based on morphological features, including the color transition of petals and flower discs from green to yellow, and the enlargement of ovaries from small to swollen. Based on these criteria, three distinct developmental stages were defined: F1 (Initial Flowering Stage): Flower buds remain closed, with petals and discs predominantly green; ovaries are small and immature. F2 (Full Bloom Stage): Flowers are fully open, displaying bright yellow petals and visible stamens; ovaries begin to swell, corresponding to the peak of anthesis. F3 (Late Flowering Stage): Petals gradually wilt and fade in color, while ovaries become fully swollen; most flowers have completed pollination and entered senescence (Figure 1A–C).

Collected inflorescences were immediately wrapped in aluminum foil, frozen in liquid nitrogen, and stored at –80 °C. All experiments were conducted with three biological replicates.

2.2. RNA Extraction, Sequencing, and Transcriptome Data Analysis

Total RNA was extracted from B. chinense flowers using the TIANEGNRNAsimple Total RNA Kit (TIANGEN Biotech Co., Ltd., Beijing, China), including DNase I treatment, following the manufacturer’s protocol. RNA concentration was quantified with a Qubit 2.0 fluorometer, and RNA integrity was assessed to ensure suitability for library construction. High-quality RNA was used to generate cDNA libraries with the NEBNext^® Ultra™ RNA Library Prep Kit (New England Biolabs, Ipswich, MA, USA) for Illumina, and paired-end sequencing was conducted on the Illumina NovaSeq 6000 platform by Metware Biotechnology Co., Ltd. (Wuhan, China).

Raw sequencing reads were processed with fastp v0.19.3 to remove adaptor sequences, reads containing poly-N, and low-quality reads. Data quality was evaluated using FastQC v0.11.9. Clean reads were aligned to the B. chinense reference genome [32] using HISAT2 v2.1.0, and transcript assembly and quantification were performed with StringTie v1.3.4. Gene expression levels were calculated as fragments per kilobase of transcript per million mapped reads (FPKM), and raw read counts were extracted for differential expression analysis.

Differentially expressed genes (DEGs) were identified using DESeq2 v1.26.0 with thresholds of |log2(fold change)| > 1 and a false discovery rate (FDR; Benjamini–Hochberg adjusted p-value) < 0.05. DEGs were annotated by BLAST v2.12.0+ searches against the NR, Swiss-Prot, TrEMBL, KOG, GO, and KEGG databases. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses were subsequently performed to investigate biological functions and metabolic pathways associated with the DEGs.

2.3. Quantitative Real-Time PCR Validation

Six differentially expressed genes (DEGs) were selected for qRT-PCR to validate the reliability of the high-throughput sequencing data. Quantitative real-time PCR was performed using the QuantStudio 5 Flex real-time PCR system (ABI, Foster City, CA, USA) with TB Green Premix Ex Taq™ II (TaKaRa Bio Inc., Kusatsu, Shiga, Japan). Actin from B. chinense was used as the internal reference gene for qRT-PCR. Relative gene expression levels were calculated using the 2^−∆∆Ct method. All primer sequences are provided in Supplementary Table S1.

2.4. UPLC-MS/MS Determination and Data Analysis

The B. chinense samples were freeze-dried using a vacuum freeze dryer (Scientz-100F; Scientz Biotechnology Co., Ltd., Ningbo, China). The lyophilized materials were then pulverized using a zirconia bead mixer (MM 400; Retsch GmbH, Haan, Germany) for 1.5 min at a frequency of 30 Hz. Subsequently, 100 mg of the powdered sample was extracted with 1.2 mL of 70% methanol by vortexing for 30 s every 30 min, for a total of six cycles. After centrifugation at 12,000 rpm for 10 min, the supernatants were filtered through a 0.22 μm membrane (SCAA-104; ANPEL, Shanghai, China).

Metabolite extraction, detection, and identification were conducted by Metware Biotechnology Co., Ltd. (Wuhan, China) using a plant-wide targeted metabolomics approach. Qualitative analysis was first performed on pooled samples using a high-resolution mass spectrometer (AB Sciex TripleTOF 6600, Framingham, MA, USA) coupled to a UPLC system (Shimadzu Nexera X2; Shimadzu Corporation, Kyoto, Japan). Quantitative analysis was subsequently carried out using an AB Sciex 4500 QTRAP system (AB Sciex, Framingham, MA, USA) equipped with an electrospray ionization (ESI) source, combining the advantages of non-targeted and targeted metabolomics. Chromatographic separation was achieved on an Agilent SB-C18 column (1.8 µm, 2.1 mm × 100 mm) with the same gradient program and parameters as described below. The mobile phase consisted of solvent A (water containing 0.1% formic acid) and solvent B (acetonitrile containing 0.1% formic acid). The gradient elution program was set as follows: from 0 to 9 min, linear increase from 5% B to 95% B; from 9 to 10 min, maintained at 95% B; from 10 to 11.1 min, returned to 5% B; and from 11.1 to 14 min, equilibrated at 5% B. The flow rate was set at 0.35 mL/min, the column temperature was maintained at 40 °C, and the injection volume was 4 μL. To assess the stability of the UPLC-MS/MS system throughout the acquisition process, a QC sample, consisting of a pooled mixture of all samples, was analyzed after every 10 injections.

Metabolite identification was performed based on accurate mass, MS² fragmentation patterns, retention time, and isotope distribution, matched against the MetWare database (MWDB). Metabolites were identified and annotated using the KEGG Compound database, and the annotated metabolites were subsequently mapped to the KEGG pathway database. Differentially expressed metabolites (DEMs) among different floral developmental stages were screened based on the criteria of fold change ≥2 or ≤0.5 and VIP ≥ 1. Metabolite Set Enrichment Analysis (MSEA) was conducted to identify metabolic pathways significantly enriched in the DEMs, with pathway significance determined using p-values derived from a hypergeometric test.

Unsupervised principal component analysis (PCA) was performed using the prcomp function in R v4.3.1 (www.r-project.org). Prior to PCA, missing values were imputed with the constant 9, and the data were mean-centered and unit variance (UV, auto/scale) standardized. Pooled QC (Mix) samples were included only to monitor instrument stability and analytical reproducibility.

For heatmap visualization, metabolite/gene expression data were first standardized by Z-score transformation across samples. Hierarchical clustering of both samples and variables was performed using Euclidean distance and complete linkage. Heatmaps were generated using TBtools-II v2.310 or MetWare Cloud Platform (https://cloud.metware.cn) (accessed on 20 July 2025), with row and column dendrograms displayed to illustrate clustering patterns.

2.5. Correlation Network Analysis of Differentially Expressed Genes and Metabolites

Pearson correlation analysis was performed between differentially expressed genes (DEGs) and differentially accumulated metabolites (DEMs) identified in B. chinense inflorescences at different flowering stages. Only strong (|cor| > 0.9) and statistically significant (p < 0.01) correlations were retained. A correlation network was subsequently constructed, in which genes were represented as circles and metabolites as triangles, with yellow and blue edges indicating positive and negative correlations, respectively. The visualization was generated using the Metware Cloud Platform (https://cloud.metware.cn) (accessed on 24 October 2025), providing an intuitive overview of the potential regulatory interactions between gene expression and metabolite accumulation.

2.6. Determination of Total Flavonoids and Total Anthocyanins

The total flavonoid content (TFC) in B. chinense was determined following the method described by Yang [33]. Rutin was used as the reference standard, and standard solutions at concentrations of 0.5, 0.25, 0.1, 0.05, 0.025, 0.0125, and 0.00625 mg·mL⁻¹ were prepared to construct a calibration curve. The relationship between absorbance (Y) and rutin concentration (X) was described by the regression equation Y = 29.702X − 0.025 (r = 0.9994). TFC was calculated based on this equation and expressed as milligrams of rutin equivalents (RE) per 100 g of dry weight (DW).

The total anthocyanin content was determined using a Total Anthocyanin Assay Kit (Geruisi, Suzhou, China) according to the manufacturer’s protocol. All measurements were performed in triplicate using independent biological samples.

2.7. Data Statistics and Analysis

WPS Excel was used to organize the data. Significance analysis was performed using LSD in SPSS Statistics 26 software, where p < 0.05 indicated a significant difference and p < 0.01 indicated an extremely significant difference. MetWare Cloud (https://cloud.metware.cn) (accessed on 20 July 2025), Prism10.2 and TBtools-II.v2.310 software were used for plotting.

3. Results

3.1. Enrichment Analysis of DEGs and KEGG Pathways at Different Flowering Stages

DEGs were identified using the criteria of |log₂Fold Change| ≥ 1 and a false discovery rate (FDR) < 0.05 across the comparisons. A total of 8112 genes were differentially expressed when comparing F1 with F2 and F1 with F3. Specifically, 3934 differentially expressed genes were identified in the F1 vs. F2 comparison, of which 1799 were up-regulated and 2135 were down-regulated. In contrast, the F1 vs. F3 comparison revealed a greater number of differentially expressed genes, totaling 4178, which included 2117 up-regulated genes and 2061 down-regulated genes (Figure 2A). Figure 2B illustrates the expression patterns of differentially expressed genes (DEGs) across the various comparison groups. The Venn diagram in Figure 2C indicates that a total of 230 differentially expressed genes were common to all three comparison groups (Table S2). Additionally, 959 common genes were identified between the F1 vs. F2 and F1 vs. F3 groups, 1490 common genes between the F1 vs. F3 and F2 vs. F3 groups, and 1203 common genes between the F1 vs. F2 and F2 vs. F3 groups.

KEGG pathway enrichment analysis revealed that the most significantly enriched pathways during B. chinense flowering were biosynthesis of secondary metabolites and phenylpropanoid biosynthesis, which were prominent in both the F1 vs. F2 (Figure 2D) and F1 vs. F3 (Figure 2E) comparisons. Given that phenylpropanoids are key precursors for flavonoids, this enrichment strongly suggests active flavonoid biosynthesis. Furthermore, the significant enrichment of MAPK and plant hormone signaling pathways indicates their potential involvement in indirectly regulating flavonoid production, likely through transcription factor-mediated mechanisms.

3.2. Expression of Genes Related to the Flavonoid Synthesis Pathway

Analysis of the expression patterns of key structural genes revealed stage-specific regulation of the flavonoid biosynthesis pathway during flower development (Figure 3).

The F1 stage was characterized by the initiation of the pathway. High expression levels of upstream genes were observed, including the PAL gene (Hap0_BC04_G12486), the C4H gene (Hap0_BC04_G10296), and key 4CL genes such as Hap0_BC01_G01344. Crucially, the core genes CHS (Hap0_BC03_G29062) and CHI (Hap0_BC05_G13084) reached their peak expression in this stage, indicating that F1 serves as a preparatory phase for flavonoid skeleton formation.

In the F2 stage, the pathway was fully activated and began to diverge. The expression of specific genes was significantly up-regulated compared to F1. For instance, the expression of the 4CL gene (Hap0_BC04_G12603) was significantly up-regulated in F2 (Log₂FC = 3.05), and the PAL gene (Hap0_BC04_G11019) exhibited a similar up-regulation (Log₂FC = 1.82). A marked shift in metabolic flux was evident: the expression of the FLS gene (Hap0_BC05_G17527) was strongly induced (Log₂FC = 7.72), implying a potential shift in metabolic flux favoring flavonol production, while the ANS gene (Hap0_BC05_G13995) maintained high activity, supporting anthocyanin production. Additionally, the FLS gene (Hap0_BC02_G03390) showed high expression in F1 and F2 but was significantly down-regulated in F3 (Log₂FC F1 vs. F3 = −3.14), highlighting the decline in flavonol synthesis in later stages.

The F3 stage featured a distinct expression profile dominated by late-pathway genes. While upstream gene expression generally declined, key downstream genes exhibited sustained or maximal activity. Most notably, two FNS genes (Hap0_BC06_G18155 and Hap0_BC05_G14309) were sharply up-regulated compared to F1 vs. F2, indicating active flavone biosynthesis. Concurrently, a critical ANS gene (Hap0_BC05_G16058) reached its highest expression level in F3 (Log₂FC = 1.96). This was accompanied by high expression of specific DFR genes, including Hap0_BC03_G06363 and Hap0_BC04_G12454. The ANS gene (Hap0_BC06_G17901) exhibited moderate expression across all stages without significant fold-change, suggesting a stable but secondary role in anthocyanin synthesis. The concerted up-regulation of these genes in F3 provides a clear molecular explanation for the substantial accumulation of anthocyanins observed in mature petals.

Several genes exhibited consistently high expression levels across all three developmental stages, indicating their fundamental role in maintaining basal flux through the flavonoid pathway. The 4CL gene (Hap0_BC05_G15994) showed high expression in three stages, reflecting its continuous involvement in generating CoA esters for downstream branches. Similarly, the C4H gene (Hap0_BC05_G34388) maintained elevated expression in three stages, with a significant up-regulation from F1 to F3 (Log₂FC = 1.40), underscoring its persistent activity in early phenylpropanoid metabolism. The sustained expression of these genes suggests a constitutive supply of precursors, ensuring steady substrate availability for both early and late pathway enzymes across flower development.

3.3. qRT-PCR Validation

To corroborate the RNA-seq results, the relative expression of six genes (Hap0_BC05_G15994, Hap0_BC05_G34388, Hap0_BC05_G13084, Hap0_BC06_G18155, Hap0_BC02_G03390, Hap0_BC05_G16058) identified via transcriptome analysis was quantified using qRT-PCR. RNA-seq and qRT-PCR results correlated strongly (Figure 4), which underscores the reliability of the transcriptome analysis.

3.4. Metabolic Analysis Based on UPLC-MS Analysis

To evaluate metabolic variation among samples, principal component analysis (PCA) was conducted based on metabolite profiles from the three groups. The PCA score plot (Figure 5A) showed minimal variation among biological replicates within each group, while clear separation was observed between groups. These distinct clustering patterns indicate that different flowering stages exert a pronounced influence on metabolite accumulation.

As shown in Figure 5B, qualitative and quantitative analyses identified a total of 496 metabolites across three flowering stages of B. chinense (Table S3). These metabolites were categorized as follows: phenolic acids (189, 38%), flavonoids (185, 37%), terpenoids (29, 6%), lignans and coumarins (65, 13%), tannins (5, 1%), alkaloids (13, 3%), and others (9, 2%). Heatmaps of the accumulation patterns of other differentially expressed metabolites, including terpenoids, phenolic acids, and alkaloids, are provided in Supplementary Figure S1.

The data demonstrate that flavonoids represent the predominant metabolite category. The composition of the 185 flavonoid compounds includes: 121 flavonoids and flavonol compounds, 24 flavanones and flavanonols, 13 anthocyanins, 8 isoflavones, 8 chalcones, 7 flavanols and 4 flavonoid carbonosides (Figure 5C).

3.5. Differentially Expressed Metabolites (DEMs) of Different Flowering Stages of B. chinense and KEGG Classification

A total of 53 DEMs were detected between F1 vs. F2, including 41 up-regulated and 12 down-regulated. Between F1 vs. F3, a total of 92 DEMs were detected. including 58 up-regulated and 34 down-regulated (Figure 6A,B).

Figure 6C,D shows the KEGG categorical metabolites between F1 vs. F2 and between F1 vs. F3. 40.74% of KEGGs in the F1 vs. F2 group were synthesized from flavonoids, 22.22% were synthesized from flavonoids and flavonols, 11.11% were synthesized from isoflavonoids, and 11.11% were synthesized from anthocyanins. In the F1 vs. F3 group, 51.72% were synthesized from flavonoids, 24.14% from flavonoids and flavonols, and 3.45% each were synthesized from anthocyanins and isoflavone.

3.6. Analysis of KEGG Pathway Differential Abundance (DA) Scores and Differential Flavonoid Metabolites at Different Flowering Stages

To further investigate the dynamic features of flavonoid metabolism across flowering stages, we integrated differential metabolite trends with changes in KEGG pathway expression and applied the differential abundance score (DA Score) for pathway-level assessment. The DA Score quantifies overall metabolic changes within a pathway by evaluating the balance between up-regulated and down-regulated metabolites and is calculated as follows:

DA Score = (Number of up-regulated metabolites in the pathway − Number of down-regulated metabolites in the pathway)/Total number of metabolites annotated to the pathway

DA Score analysis for the F1 vs. F2 comparison (Figure 7A) showed a marked up-regulation of the flavonoid biosynthesis pathway, indicating increased flavonoid production during the transition from the initial flowering stage. Conversely, the isoflavone biosynthesis pathway exhibited down-regulation. Although the flavone and flavonol biosynthesis pathways also demonstrated upward trends, the changes were not statistically significant. Collectively, these metabolic patterns indicate that, at this developmental phase, metabolic flux was largely concentrated within the core flavonoid biosynthetic pathway, with relatively limited activity in specific branch pathways.

A comparative analysis between the F1 and F2 stages identified a total of 20 flavonoid DEMs. Among these, 12 metabolites were significantly up-regulated, while 8 were down-regulated (

Table 1. Flavonoids DEMs in F1 vs. F2 stages.

Compounds	Class II	Log2FC	Type
Kaempferol-3-O-(2″-p-Coumaroyl)glucoside	Flavonols	13.85	up
Naringenin-4′,7-dimethyl ether	Flavanones	9.24	up
5,7,4′-Trihydroxy-6,8-dimethoxyisoflavone-7-O-galactoside	Isoflavones	−1.23	down
Cyanidin-3-O-(6″-O-malonyl)glucoside	Anthocyanidins	−3.77	down
Kaempferol-3-O-(3″-O-p-Coumaroyl)rhamnoside	Flavonols	5.61	up
Kaempferol-3-O-(4″-p-coumaroyl)rhamnoside	Flavonols	5.58	up
Kaempferol-3-O-(4″-O-p-Coumaroyl)rhamnoside	Flavonols	5.43	up
Tricin-7-O-Glucoside	Flavones	−1.17	down
Phloretin-4′-O-glucoside (Trilobatin)	Chalcones	−1.38	down
Iristectorin A	Isoflavones	−1.06	down
Isorhamnetin-3-O-rhamnoside	Flavonols	−1.23	down
Prunetin (5,4′-Dihydroxy-7-methoxyisoflavone)	Isoflavones	−1.46	down
Aromadendrin (Dihydrokaempferol)	Flavanonols	2.81	up
Kaempferol-3-O-rhamnoside (Afzelin)(Kaempferin)	Flavonols	2.74	up
Rosinidin-3-O-glucoside	Anthocyanidins	−1.35	down
Fisetin	Flavones	2.69	up
Kaempferol-3-O-sophorotrioside	Flavonols	2.69	up
Kaempferol-7-O-rhamnoside	Flavonols	2.67	up
Delphinidin-3-O-rutinoside-7-O-glucoside	Anthocyanidins	2.21	up
Apigenin-7-O-glucoside(Cosmosiin)	Flavones	2.12	up

). Notably, several kaempferol derivatives exhibited marked accumulation, with Kaempferol-3-O-(2″-p-Coumaroyl)glucoside showing the most pronounced up-regulation (Log₂FC = 13.85). Other kaempferol glycosides, including Kaempferol-3-O-(4″-p-coumaroyl)rhamnoside (Log₂FC = 5.58) and Kaempferol-3-O-(3″-O-p-Coumaroyl)rhamnoside (Log₂FC = 5.61), were also highly up-regulated. The flavanone Naringenin-4′,7-dimethyl ether was significantly accumulated (Log₂FC = 9.24). Furthermore, the anthocyanins Delphinidin-3-O-rutinoside-7-O-glucoside (Log₂FC = 2.21) and the flavones Apigenin-7-O-glucoside (Cosmosiin, Log₂FC = 2.12) were up-regulated.Conversely, some anthocyanins were down-regulated, including Cyanidin-3-O-(6″-O-malonyl)glucoside (Log₂FC = −3.77) and Rosinidin-3-O-glucoside (Log₂FC = −1.35).

In the F1 vs. F3 comparison, KEGG annotation revealed significant down-regulation of the flavonoid biosynthesis pathway (Figure 7B). The flavonol and isoflavone sub-pathways were also mapped. This collective down-regulation suggests a marked decline in flavonoid synthesis after full flowering, which may reflect floral organ maturation or a shift in metabolic resources toward other physiological processes.

F1 vs. F3 stages revealed more extensive changes in the flavonoid metabolome, with 33 DEMs identified. A clear trend of down-regulation was observed, as 25 metabolites decreased in abundance while only 8 increased (

Table 2. Flavonoids DEMs in F1 vs. F3 stages.

Compounds	Class II	Log2FC	Type
Phloretin-4′-O-glucoside (Trilobatin)	Chalcones	−2.34	down
Quercetin-3-O-apiosyl(1→2)galactoside	Flavonols	−1.69	down
Quercetin-3-O-sambubioside	Flavonols	−1.66	down
Apigenin-7-O-neohesperidoside (Rhoifolin)	Flavones	−2.07	down
Petunidin-3-O-(6″-O-Acetyl)glucoside	Anthocyanidins	10.71	up
Diosmetin-7-O-glucoside	Flavanones	−1.55	down
Diosmetin-7-O-galactoside	Flavanones	−1.43	down
Isohyperoside	Flavonols	−1.41	down
6-C-MethylKaempferol-3-glucoside	Flavones	−1.41	down
5,7,4′-Trihydroxy-6,8-dimethoxyisoflavone-7-O-galactoside	Isoflavones	−2.28	down
Hispidulin-7-O-Glucoside	Flavones	−1.46	down
Catechin	Flavanols	5.87	up
Iristectorin A	Isoflavones	−2.13	down
Gallocatechin 3-O-gallate	Flavanols	3.65	up
Tricin-7-O-Glucoside	Flavones	−2.15	down
Quercetin-7-O-rutinoside-4′-O-glucoside	Flavonols	−1.57	down
Isorhamnetin-3-O-rhamnoside	Flavonols	−1.57	down
Quercetin-5-O-β-D-glucoside	Flavonols	−1.29	down
Tricin-7-O-Glucuronide	Flavones	−1.83	down
Malvidin-3-O-rutinoside	Anthocyanidins	−1.64	down
Eriodictyol-8-C-glucoside	Flavanones	−1.34	down
Azaleatin (5-O-Methylquercetin)	Flavonols	2.19	up
Eriodictyol-7-O-glucoside	Flavanones	−1.76	down
Hesperetin-7-O-glucoside	Flavanones	−1.46	down
Isorhamnetin-3-O-sophoroside-7-O-rhamnoside	Flavonols	1.97	up
Kaempferol-3-O-sophorotrioside	Flavonols	1.55	up
Isorhamnetin-3-O-glucoside-7-O-rhamnoside	Flavonols	1.27	up
Quercetin-3-O-xyloside (Reynoutrin)	Flavonols	−1.33	down
Jaceosidin-7-O-Glucoside	Flavones	−2.19	down
Delphinidin-3-O-rutinoside-7-O-glucoside	Anthocyanidins	1.13	up
Apigenin-7-O-rutinoside (Isorhoifolin)	Flavones	−1.89	down
3-O-Methylquercetin	Flavonols	−1.88	down
Avicularin(Quercetin-3-O-α-L-arabinofuranoside)	Flavonols	−1.28	down

). Widespread down-regulation affected numerous compounds across various subclasses, including the chalcone phloretin-4′-O-glucoside (trilobatin; Log₂FC = −2.34), the flavone rhoifolin (apigenin-7-O-neohesperidoside; Log₂FC = −2.07), and a range of quercetin glycosides such as quercetin-3-O-apiosyl(1→2)galactoside (Log₂FC = −1.69) and quercetin-3-O-sambubioside (Log₂FC = −1.66).

Despite this overall down-regulation, several compounds remained significantly up-regulated in the F3 stage and may perform key physiological defense functions late in flowering. Most notably, these included the anthocyanin petunidin-3-O-(6″-O-acetyl)glucoside (Log₂FC = 10.71) and the flavanols catechin (Log₂FC = 5.87) and gallocatechin 3-O-gallate (Log₂FC = 3.65).

Total flavonoid and anthocyanin contents were measured in B. chinense flowers at different developmental stages. The results are shown in Supplementary Figure S2A,B. The highest total flavonoid content was observed at the F2 stage, while the F3 stage exhibited the highest anthocyanin content. These patterns are in agreement with the accumulation trends of differentially expressed metabolites revealed by our metabolomic analysis.

3.7. Integrated Transcriptome and Metabolome Analyses of B. chinense Flower Stages

We annotated the differentially expressed genes in F1 and F2, and F2 and F3, and identified 330 and 339 transcription factors, respectively (Table S4). Several transcription factors (TFs), mainly from the MYB (Hap0_BC03_G08886, Hap0_BC01_G02450, Hap0_BC01_G02450), bHLH (Hap0_BC03_G05824, Hap0_BC01_G02453), and bZIP (Hap0_BC04_G30855, Hap0_BC01_G02453, Hap0_BC01_G02450) families, exhibited expression trends consistent with key structural genes during floral development. The observed co-expression patterns suggest that transcriptional regulation may contribute to the stage-specific accumulation of flavonoids in B. chinense, although detailed regulatory mechanisms remain to be elucidated.

To further explore the coordinated regulation between gene expression and metabolite accumulation during the floral development of B. chinense, we performed an integrated analysis using the annotated DEGs and DEMs. The joint pathway analyses of F1 vs. F2 (Figure 8A) and F2 vs. F3 (Figure 8B) both indicated significant enrichment in pathways related to secondary metabolite biosynthesis.

In the F1 vs. F2 comparison, although flavonoid biosynthesis and flavone and flavonol biosynthesis pathways were enriched, they did not reach statistical significance. However, several genes involved in the precursor pathway phenylpropanoid biosynthesis were significantly up-regulated at this stage, suggesting the initiation of flavonoid precursor formation. In contrast, in F2 vs. F3, genes in the phenylpropanoid biosynthesis pathway remained significantly expressed, while DEMs were significantly enriched in the flavonoid biosynthesis pathway. Moreover, DEGs were also significantly enriched in the isoflavonoid biosynthesis pathway, and the flavone and flavonol biosynthesis pathway was still detected.

To elucidate potential regulatory mechanisms between these molecular layers, a Pearson correlation network was constructed based on the DEGs and DEMs (Figure 8C). The resulting network revealed several genes showing strong correlations with flavonoid-related metabolites, including Hap0_BC02_G03390, Hap0_BC05_G14728, and Hap0_BC06_G18155. These genes are likely key regulators of flavonoid biosynthesis and will be further investigated in future studies.

4. Discussion

In recent decades, rapidly expanding agricultural cultivation has substantially increased the generation of agricultural by-products [34]. However, a long-standing lack of systematic strategies for their high-value utilization means these materials are often treated as waste, resulting in not only low resource-use efficiency but also potential pressures on ecosystem stability [35]. A growing body of evidence reveals that these by-products are often abundant in bioactive compounds with significant potential for valorization. Nevertheless, current research remains predominantly focused on staple food crops, including wheat [36], rice [37], corn [38], soybean [39], and common fruits [40].

Similarly, in the field of medicinal plants, large-scale cultivation has led to the development of specialized commercial production systems. As with conventional crops, post-harvest processing of Chinese medicinal herbs typically utilizes only the legally defined medicinal parts, while the remaining portions are often discarded, leading to substantial resource waste. To address this gap, we employed an integrated transcriptomic and metabolomic approach to delineate the dynamic patterns and molecular regulatory mechanisms of flavonoid biosynthesis across three developmental stages (F1, F2, F3) in B. chinense flowers.

4.1. Initial Flowering Stage (F1): Initial Activation of the Flavonoid Skeleton Biosynthesis Pathway

During the initial flowering stage (F1), key upstream structural genes in the flavonoid biosynthetic pathway—including PAL, C4H, 4CL, and the core genes CHS and CHI—were highly expressed. This expression pattern suggests that F1 represents a critical preparatory phase for flavonoid skeleton formation, during which the plant actively accumulates precursors for the synthesis of diverse downstream flavonoid compounds. This interpretation is further supported at the pathway level by the significant enrichment of the phenylpropanoid biosynthesis pathway observed in the F1 vs. F2 KEGG enrichment analysis.

4.2. Full Bloom Stage (F2): Significant Diversion of Flavonoid Metabolism and Active Flavonol Synthesis

During the transition to the F2 stage, flavonoid metabolism in B. chinense flowers exhibited distinct branch-specific patterns. Transcriptomic analysis showed strong stage-dependent expression of key flavonoid biosynthetic genes. Compared with the F1 stage, several 4CL and PAL genes were up-regulated (Log₂FC > 1), indicating enhanced flux through the phenylpropanoid pathway. Notably, core structural genes (CHS, CHI, F3H) and downstream branching genes (FLS, DFR, ANS) were significantly induced in F2. The pronounced up-regulation of the FLS gene (Hap0_BC05_G17527) (Log₂FC = 7.72) suggests a shift in metabolic flux toward flavonol synthesis.

Metabolomic data strongly supported these transcriptional changes. Among the 20 DEMs identified in the F1 vs. F2 comparison, 12 were significantly up-regulated, including six kaempferol glycoside derivatives and multiple anthocyanins. Pathway-level DA Score analysis confirmed an overall up-regulation of the flavonoid biosynthesis pathway. Kaempferol derivatives, such as kaempferol-3-O-(2″-p-coumaroyl)glucoside (Log₂FC = 13.85), accumulated to high levels. Other up-regulated metabolites of interest included naringenin-4′,7-dimethyl ether—a methylated derivative with reported anti-epileptic and analgesic activity [41]—and apigenin-7-O-glucoside (cosmosiin), known for its anti-inflammatory properties [42]. The accumulation of anthocyanins, such as delphinidin-3-O-rutinoside-7-O-glucoside, not only contributes to flower pigmentation but also enhances antioxidant and anti-inflammatory potential, potentially supporting chronic disease prevention [43].

4.3. Late Flowering Stage (F3): Anthocyanin Synthesis Peaks While Overall Flavonoid Metabolism Declines

At the full-bloom stage (F3), the transcriptional landscape shifted markedly. Although the expression of most upstream genes declined, key genes within specific downstream branches were distinctly activated. Two FNS genes were significantly up-regulated. Notably, an ANS gene (Hap0_BC05_G16058) exhibited peak expression (Log₂FC F1 vs. F3 = 1.96), accompanied by elevated expression of select DFR genes, providing molecular evidence for the enhanced anthocyanin biosynthesis observed at this stage.

Metabolomic profiling revealed an overall reduction in flavonoid biosynthesis during F3, reflected by a significantly decreased DA score and reduced levels of 25 out of 33 flavonoid DEMs. Several bioactive compounds decreased substantially, including multiple quercetin glycosides, apigenin disaccharide glycosides, and phloretin-4′-O-glucoside. This decline suggests a diminished accumulation of metabolites associated with anti-inflammatory [44] and antithrombotic activities [45] in late flowering. In particular, phloretin-4′-O-glucoside—recognized for hypoglycemic [46] and hepatoprotective effects [47]—was consistently down-regulated, supporting the notion that its primary accumulation occurs during earlier floral stages. This trend indicates a developmental shift in metabolic priorities toward terminal floral maturation.

Despite the general down-regulation, several metabolites remained abundant in F3, including petunidin-3-O-(6″-O-acetyl)glucoside (Log₂FC = 10.71), catechin (Log₂FC = 5.87), and gallocatechin-3-O-gallate. As potent antioxidants, their sustained accumulation implies ongoing antioxidant defense activity in senescing petals. These metabolites are also associated with cardiovascular protection [48] and anti-influenza properties [49], suggesting that late-stage flowers may retain value for the targeted extraction of highly stable antioxidant constituents.

Correlation network analysis provided deeper insights into the transcriptional regulation underlying flavonoid biosynthesis during B. chinense flower development. Several genes—including Hap0_BC02_G03390, Hap0_BC05_G14728, and Hap0_BC06_G18155—displayed strong correlations with multiple differential flavonoid metabolites (|r| > 0.9, p < 0.01), suggesting their potential involvement in fine-tuned control of flavonoid accumulation, particularly during the highly active transition from F2 to F3. These findings indicate that flavonoid biosynthesis in B. chinense flowers is temporally regulated and coordinated at the transcriptional level, and the identified genes represent promising candidates for future functional validation.

Co-expression patterns further revealed that MYB, bHLH, and bZIP transcription factors were associated with key structural genes in the flavonoid pathway, implying their participation in the developmental regulation of this metabolic process. Similar transcriptional modules have been reported in other plant species [50,51], supporting the possibility that B. chinense employs conserved mechanisms for flavonoid pathway regulation.

In addition, dynamic metabolic transitions across floral developmental stages have been observed in other medicinal and ornamental species, although the specific accumulation profiles and associated regulatory networks are generally species-dependent.

In Hibiscus syriacus, metabolic variation is closely associated with petal color changes, primarily driven by fluctuations in anthocyanin content. Transcriptomic and metabolomic analyses revealed that petal color gradually shifts from purplish-red to violet as anthocyanin levels (such as cyanidin, delphinidin, and peonidin derivatives) decline. Early biosynthetic genes, including PAL, CHS, and CHI, are activated during the initial flowering stages, facilitating flavonoid skeleton formation. Subsequently, downstream genes such as DFR, ANS, and UFGT drive anthocyanin biosynthesis, leading to peak pigment accumulation in the mid-flowering stage before a gradual decrease. Flavones and flavonols are synthesized throughout the flowering process, primarily serving as co-pigments rather than direct color determinants [31].

In Lonicera japonica, floral color transition is primarily attributed to the coordinated regulation of carotenoid and flavonoid metabolism. Multi-omics studies have revealed that genes involved in carotenoid biosynthesis (PSY1, PDS, ZDS, LCYB) are significantly up-regulated during the transition from white to yellow flowers, accompanied by a marked accumulation of α-carotene and lutein. Meanwhile, the down-regulation of chlorophyll biosynthetic genes and up-regulation of degradation-related genes lead to petal de-greening. Moreover, L. japonica flowers are already recognized as medicinal parts and widely consumed as functional teas, which distinguishes them from B. chinense inflorescences that remain underutilized despite their comparable bioactive potential [52].

In summary, the flavonoid metabolism of B. chinense flowers is characterized by precise temporal regulation. The Full Bloom Stage (F2) is a critical period for the synthesis of various highly active flavonoid components (such as kaempferol glycosides and anthocyanins). In contrast, the Late Flowering Stage (F3) is enriched with specific antioxidant substances. The findings of this study not only provide an important theoretical basis and practical guidance for the high-value development and utilization of B. chinense flowers as an agricultural by-product but also offer an analyzable paradigm for researching the accumulation patterns of secondary metabolites in non-traditional parts (flowers, leaves) of other medicinal plants.

5. Conclusions

Based on integrated transcriptomic and metabolomic analyses, this study systematically delineates the stage-specific regulatory patterns of flavonoid biosynthesis in B. chinense inflorescences. The results reveal that flavonoid metabolism in B. chinense flowers is tightly regulated in a temporal manner: during the Initial Flowering Stage (F1), the phenylpropanoid pathway is activated to supply precursors for flavonoid biosynthesis; at the Full Bloom Stage (F2), metabolic flux is markedly redirected toward flavonol and anthocyanin branches, accompanied by pronounced upregulation of key genes (FLS, DFR, ANS) and substantial accumulation of multiple bioactive metabolites, including quercetin glycosides and apigenin glycosides; during the Late Flowering Stage (F3), overall pathway activity declines, yet specific antioxidant compounds—such as anthocyanins and catechins—continue to accumulate.

Collectively, these results indicate that the Full Bloom Stage (F2) represents the optimal harvest period for maximizing the production of diverse highly active flavonoids. This work establishes a theoretical basis for the high-value utilization of B. chinense inflorescences as an agricultural by-product and provides a methodological framework for resource-focused studies on non-traditional tissues in other medicinal plant species.