Integrated metabolomics and transcriptomics reveal metabolite differences between wild and cultivated Angelica sinensis

Abstract

Background

Angelica sinensis (Oliv.) Diels, a traditional Chinese herb widely used in food and medicine, predominantly grows in high-altitude regions under wild conditions. With the scarcity of wild resources, cultivated Angelica sinensis has garnered growing attention in recent years. However, the metabolic disparities and regulatory mechanisms between wild and cultivated A. sinensis remain poorly understood.

Results

This study employed an integrated transcriptomic and metabolomic approach to compare their metabolic profiles and biosynthetic pathways. Metabolomic profiling revealed 131 differentially accumulated metabolites (DAMs) primarily enriched in fatty acid metabolism, amino acid biosynthesis, and secondary metabolite biosynthesis, with wild A. sinensis displaying significantly elevated levels of these compounds compared to the cultivated counterpart. Transcriptome analysis found 18,993 differentially expressed genes (DEGs), with 9 key structural genes involved in fatty acid, amino acid, and phenylpropanoid biosynthesis being differentially regulated.

Conclusions

9 core structural genes (tktA, metE, GOT1, GOT2, ADT, ACX, PAL, 4CL, and CCR) potentially underlie the chemical compositional disparities observed between wild and cultivated A. sinensis. These findings advance our understanding of the molecular basis for metabolic adaptation in this potential industrial crop.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12870-026-09082-9.

Introduction

Angelica sinensis (Oliv.) Diels, commonly known as “Danggui”, is a perennial herbaceous plant in the Apiaceae family genus Angelica. Its dried root serves as a key traditional Chinese medicine (TCM) for its blood-enriching and activating properties, and has been recognized as a core medicinal herb for gynecological, cardiovascular, and inflammatory disorders [1–2]. Modern pharmacological studies have revealed its multifaceted bioactivities, including antioxidant [3], anti-inflammatory [4], immunomodulatory [5], anti-tumor [6], and neuroprotective effects [7]. Among these, ferulic acid and Z-ligustilide are recognized as major bioactive constituents contributing to its hematopoietic and cardiovascular protective activities, while polysaccharides are primarily responsible for its immunomodulatory effects [8]. These bioactivities are attributed to its bioactive constituents, including phthalides, organic acids, and polysaccharides [9].

Wild A. sinensis thrives in specific microclimatic conditions at elevations of 1800–3000 m, favoring cool, humid environments with well-drained sandy loam soils [10]. Documented in the “Shennong Ben Cao Jing” as early as the Eastern Han Dynasty (25–220 AD), its medicinal use historically relied on wild populations harvested from natural habitats, but overexploitation and habitat fragmentation driven by increasing market demand have caused a sharp decline in wild resources. Consequently, large-scale cultivation has been initiated in regions such as Gansu, Yunnan, and Sichuan to satisfy commercial requirements [11]. Although cultivated A. sinensis ensures sufficient supply, significant chemical differences have been observed between wild and cultivated variants [12]. The molecular mechanisms underlying these chemotypic divergences remain poorly understood.

Metabolomics, a robust analytical approach, has been extensively applied to elucidate metabolic variations in Angelica sinensis across geographical origins, and cultivation regimes [13]. This approach enables systematic profiling of global metabolite dynamics. Concurrently, high-throughput transcriptome sequencing has been used to provide critical insights into functional genes governing primary and secondary metabolite biosynthesis. Recent studies have leveraged these technologies to compare transcriptional landscapes between wild and cultivated A. sinensis. Differential gene expression patterns linked to metabolic biosynthesis and environmental adaptation have been uncovered [14]. Analogously, a dual omics investigation on Ophiocordyceps sinensis has systematically uncovered the molecular underpinnings of metabolic diversity by deciphering metabolite accumulation profiles and gene regulatory networks, offering significant theoretical support for medicinal value evaluation and artificial cultivation optimization [15].

Environmental, genetic, and agronomic influences [16] underpin quality differences in wild and cultivated A. sinensis, with metabolic accumulation disparities further verified [12]. However, few studies have systematically elucidated the molecular regulatory mechanisms underlying these quality variations nor explored how the regulation of key gene expression dynamically influences metabolic synthesis pathways. This study aims to conduct parallel multi-omics comparisons between wild and cultivated A. sinensis, revealing underlying changes in gene regulatory networks. It is expected to provide a sound theoretical basis for understanding their quality differences and guide future breeding efforts.

Materials and methods

Chemicals and reagents

HPLC-grade acetonitrile, methanol, and LC-MS-grade formic acid were sourced from Sigma-Aldrich Shanghai Trading Co., Ltd. (China), with water generated by a Merck Millipore (USA) system.

Sample

Cultivated A. sinensis were collected in November 2024 from Tanchang County (33.78°N, 104.38°E; elevation: 1780 m a.s.l.), Gansu Province, Northwest China. Wild A. sinensis originated from seeds collected from the Minshan Mountain Range (Tanchang County, Gansu Province) in August 2023, with necessary permissions from local forestry and agricultural authorities for establishing a germplasm nursery and conducting scientific research. The seeds were formally identified by Research Fellow Sun Lianhu from Agricultural Science Research Institute of Longnan. They were subsequently planted and cultivated in Tanchang County for the purpose of germplasm conservation and breeding studies at the institute. The plants harvested in November 2024 served as the wild-origin group in this study. All the samples were identified as the root of Angelica sinensis (Oliv.) Diels by Research Fellow Sun Lianhu from Agricultural Science Research Institute of Longnan. The samples were kept in a dry and cool area at Sun Yat-sen University.

Following soil removal, A. sinensis individuals of similar size and weight were selected, pooled (six/replicate), flash-frozen in dry ice, and stored at -80 °C, with wild A. sinensis (WA) and cultivated A. sinensis (CA) samples abbreviated accordingly. Morphological and internal transcribed spacer (ITS) sequence analyses were used to identify all samples.

Metabolite extraction and UPLC-Q-Exactive-MS analysis

Following 24-hour freeze-drying, WA and cultivated CA samples were ground into a fine powder. 50 µg sample was added to 1.5 mL tube, soaked with 800 µL -20℃ 70% methanol, ultrasonically extracted at 4℃ for 30 min. Subsequently, samples were centrifuged at 14,000 rpm for 15 min at 4℃, and the supernatant was filtered through a 0.22 μm membrane before analysis. Each group contained six biological replicates, with quality control (QC) samples generated via pooling 20 µL of supernatant from each sample.

Sample analysis was conducted on a UPLC Ultimate 3000 system coupled to a Thermo Q Exactive Focus mass spectrometer (Thermo Fisher Scientific, USA), with a Hypersil GOLD aQ column (100 × 2.1 mm, 1.9 μm, Thermo Fisher Scientific, USA) employed. Elution used mobile phases A (water + 0.1% formic acid) and B (acetonitrile + 0.1% formic acid) with a gradient: 0–2 min, 5% B; 2–22 min, 5%-95% B; 22–27 min, 95% B; 27–27.1 min, 95%-5% B; 27.1–30 min, 5% B. The flow rate was set at 0.3 mL/min, and the column was maintained at 40 ℃. The spray voltage was set at 3.8 kV in positive mode and 3.2 kV in negative mode. The sheath gas and auxiliary gas flow rate were set at 40 and 10, respectively. The capillary temperature was set at 320℃. The Orbitrap analyzer scanned over a mass range of m/z 100–1500 for full scan at a resolution of 70,000.

The raw MS datasets underwent peak annotation, filtration, and peak alignment via Compound Discoverer 3.3 (Thermo Fisher Scientific, USA) and metaX. After quality control (data with relative standard deviation (RSD) > 30% were rejected), a data matrix comprising mass-to-charge ratio (m/z), retention time (RT), and peak intensity values was constructed for downstream analysis.

RNA extraction, cDNA library construction and transcriptome analysis

Purification of total RNA aliquots utilizing CTAB extraction reagent (Bio Basic, Toronto, Canada) was followed by integrity validation via agarose gel electrophoresis and concentration determination using the Agilent 2100 bioanalyzer (Agilent, CA, USA). With oligo(dT) magnetic beads, poly(A) mRNA was enriched and fragmented to ~ 300 bp, after which cDNA libraries were constructed using the Optimal Dual-mode mRNA Library Prep Kit (BGI Genomics, China) following the manufacturer’s protocol. Finally, the paired-end (PE) reads of 100/150 bases were generated on the DEBSEQ platform (BGI-Shenzhen, China) for transcriptome analysis, with six biological replicates for each group.

After adapter trimming, removal of reads with > 1% N bases, or > 40% low-quality bases (quality value < 15) using SOAPnuke v1.6.5, raw reads were assembled de novo with Trinity (https://github.com/trinityrnaseq/trinityrnaseq/wiki). Subsequent to mapping cleaned sequencing data to assembled unigenes with Bowtie2 v2.4.5, gene expression levels were quantified and normalized to reads per kilobase per million mapped reads (RPKM) using RSEM v1.3.3. The assembled unigenes were annotated with five major function databases (NR, KEGG, GO, SwissProt, and KOG).

Analysis of differentially accumulated metabolites and differentially expressed genes

Differentially accumulated metabolites (DAMs) were determined through principal component analysis (PCA) and Orthogonal partial least squares discriminant analysis (OPLS-DA). By comparing MS and MS/MS spectra against BGI Metabolome Database and mzCloud, we putatively identified metabolites and exported the processed data to SIMCA 13 for PCA and OPLS-DA. Prior to conducting these analyses, each dataset underwent mean centering and variance normalization. Features were defined as DAM if they met two criteria: OPLS-DA variable importance in projection (VIP) > 1.0 and Benjamini-Hochberg Q < 0.05 (Tab. S1).

Differentially expressed genes (DEGs) were determined through DESeq2 analysis, with significant expression changes defined as those exceeding a two-fold change (≥ 2) and Q value ≤ 0.05. Using the phyper function in R for KEGG enrichment and the TermFinder package for GO analysis, candidate genes were defined as significantly enriched at a Q value threshold of 0.05.

Given the non-normal distribution of certain datasets, Spearman’s rank correlation coefficient was employed between the DAMs and DEGs. Finally, DAMs and DEGs with absolute correlation values greater than 0.800 were systematically mapped to their corresponding KEGG pathways.

Transcriptomic data verification by Quantitative real-time polymerase chain reaction (qRT-PCR)

To validate the transcriptomic data, qRT-PCR was performed using the same RNA samples previously employed for sequencing. Nine key DEGs were selected for the test, and ACT was served as an internal control gene [17]. Gene-specific primers were designed for target genes and reference genes using PriExpress Software v2.0, and are provided in Tab. S5. The qRT-PCR mixture (10 µL) comprised 5 µL of 2×Master MIX (QIAGEN, Germany), 0.05 µL ROX, 0.2 µL of each primer (10 pM/µL), 1 µL cDNA, and 3.55 µL of RNase-free water. Quantitative PCR was carried out on a QuantStudio™ 6 Pro System (Applied Biosystems, USA) under the following cycling conditions: 95 °C for 2 min, followed by 40 cycles of 94 °C for 10 s and 59 °C for 10 s, with subsequent melt curve analysis. Six biological replicates for each treatment were analyzed. The relative expression levels were calculated using the 2^−ΔΔCt method, and statistical significance was determined with Student’s t-test. P-values under 0.05 were identified as statistically significant.

Results

Metabolic profiling of wild and cultivated A. sinensis

For method reproducibility and reliability validation, QC samples were constructed by merging equal aliquots of each A. sinensis sample followed by UPLC-Q-Exactive-MS analysis. Fig. S1A and B demonstrate excellent overlap in QC samples’ base peak chromatograms (BPCs), indicating consistent retention times and peak intensities, while Fig. S1C shows 81.0% of features with RSD < 30%, confirming data reliability. From all A. sinensis samples in positive and negative ESI modes, 7,910 ion features were obtained after removing low-repeatability peaks. Self-built databases, together with publicly available resources including the BGI HRAM-PMDB 1.0, mzCloud database, and the Human Metabolome Database (HMDB), were leveraged to preliminarily identify 571 metabolites, which were categorized into 56 classes (Fig. S1D), including alkaloids, amino acid related compounds, carbohydrates, fatty acyls, flavonoids, lignans and related compounds, organic acids, phenylpropanoids and related compounds, quinones, etc. (Tab. S1).

As an unsupervised pattern recognition method, PCA visualized distinct inter-group clustering and intra-group variability, with UPLC-Q Exactive-MS data grouping all samples into two clusters (Fig. 1A) and indicating metabolic differences between CA and WA. PCA analysis of UPLC-Q Exactive-MS datasets revealed that all samples were substantially categorized into two cohorts, reflecting distinct metabolic phenotypes for CA and WA. The total variance explained by the first two principal components was 86.3%.

Fig. 1.

Metabolomics analysis of cultivated and wild A. sinensis. A Metabolomics profiles of CA and WA PCA score plots. B Number of up-regulated and down-regulated DAMs in WA vs. CA. C Cluster heatmap analysis of DAMs between WA vs. CA. D KEGG enrichment maps of differently accumulated metabolites in WA vs. CA

Identification of differentially accumulated metabolites (DAMs) in CA and WA

To investigate the metabolites responsible for differentiating samples, orthogonal partial least squares discriminant analysis (OPLS-DA) was conducted. As shown in Fig. S2, the score plots of pairwise comparisons from OPLS-DA models revealed a clear separation between wild and cultivated groups. Significant metabolites were identified using variable importance in projection (VIP) values greater than 1.0 and Q-values less than 0.05 as criteria.

CA and WA exhibited 131 DAMs (76 up-regulated and 55 down-regulated, Fig. 1B), with hierarchical cluster analysis (HCA) performed to illustrate DAM relative abundance and inter-group correlations (Fig. 1C). KEGG enrichment analysis was conducted to explore DAM functions, and Fig. 1D shows the top 15 enriched KEGG terms, with “Biosynthesis of secondary metabolites” being the most significant term, including 32 DAMs. Other prominently enriched pathways included “alpha-Linolenic acid metabolism”, “Biosynthesis of amino acids”, “Carbon metabolism”, “Phenylalanine, tyrosine and tryptophan biosynthesis”, and “Pentose phosphate pathway”.

Transcriptomic profiling of CA and WA

CA and WA samples generated an average of 6.56 Gb raw data per sample through DEBSEQ platform sequencing. Tab. S2 shows FastQC analysis results of Q20 (98.59%) and Q30 (95.90%) scores, confirming high-quality sequencing data that was filtered and mapped to assembled unigenes. Tab. S2 shows the total mapping rate (84.61%-92.36%) and unique mapping rate (33.32%-56.05%), with 88,582 unigenes obtained.

Mapping 22,674 unigenes to five identified plant species (83.35%) and other plant taxa (16.65%) via the NCBI NR database (Fig. 2C), the KEGG database concurrently annotated 8,111 unigenes to 19 biochemically enriched pathways (Fig. S3D). Categorizing 19,810 unigenes into 56 functional groups, the GO database annotated these under three distinct functional divisions: biological process, cellular component and molecular function (Fig. S3C). 16,891 unigenes were annotated via the SwissProt database, with annotations verified against the UniProt Knowledgebase. Inferred from identified RNA sequences, proteins were matched to 14,946 unigenes via the KOG database.

Fig. 2.

Transcriptomics analysis of cultivated and wild A. sinensis. A Spearman correlation analysis among six biological replicates. B Principal component analysis score plot of gene expression for the two groups using transcriptome expression. C Species distribution of the total homologous sequences in NR database. D Volcano plot of DEGs in WA vs. CA. GO and KEGG enrichment analysis of differentially expression genes (E) and correlation analysis (F) in the WA vs. CA comparison

Spearman’s correlation (Fig. 2A) and PCA analyses (Fig. 2B) showed tight clustering and strong correlation (r = 0.810 to 0.990) among six biological replicates per group, confirming the high reproducibility of transcriptome data.

Differently expressed genes (DEGs) in CA and WA

DEGs in the two groups were classified by GO and KEGG analyses, with 18,993 DEGs (11,607 upregulated, 7,386 downregulated) identified from WA vs. CA comparison (Fig. 2D). KEGG analysis showed 2,470 DEGs in five categories (cellular processes, environmental information processing, genetic information processing, metabolism, organismal systems), mostly in metabolism (Fig. S4A). According to the dot plot (Fig. 2E), DEGs were significantly enriched in top 15 pathways, such as biosynthesis of secondary metabolites, carbon metabolism, mRNA spliceosome, and biosynthesis of amino acids and fatty acid metabolism.

Integrated analysis of DAMs and DEGs

Through KEGG database mapping, integrated analysis of DAMs and DEGs in CA and WA revealed 34 co-enriched pathways (Fig. S4B), and the top 10 of which are presented in Fig. 2F. A concerted evaluation of the P-value (Fig. S4B), rich ratio, and bubble size (Fig. 2F) revealed conspicuous enrichment in the amino acid metabolism, alpha-linolenic acid metabolism, and secondary metabolite pathways. To identify key pathways with coordinated changes between metabolites and genes, Spearman’s rank correlation analysis was performed between pathway-associated metabolites and DEGs using the cor.test function in R. DAMs and DEGs showing strong correlations (Spearman’s r > 0.8) were selected for further in-depth analysis. These pathways, characterized by a concomitant abundance of differential metabolites and associated genes (Fig. 3), were prioritized for subsequent mechanistic investigation.

Fig. 3.

DEGs and DAMs of amino-acid-relevant biosynthesis (A), α-Linolenic acid metabolism (B), and phenylpropanoid biosynthesis (C) in WA vs. CA based on the KEGG pathway database. The red colors represent upregulation in WA, green colors represent downregulation, blue colors represent both up and down regulation. Full details of metabolites and genes are provided in Supplementary Table S3 and S4, respectively

In the amino-acids metabolic pathway, a total of 7 DAMs were enriched, including 1 amino acid, 2 amino acid and derivatives, and 4 other compounds. In the “Biosynthesis of amino acids” pathway (map01230), asparagine, S-Adenosylmethionine, D-ribulose 5-phosphate, and phenylpyruvate showed high correlation with 58 unigenes. Phenylpyruvate, quinic acid, 6-Deoxy-5-ketofructose 1-phosphate, and 3-Benzylmalic acid were highly correlated with 11 unigenes in the “Phenylalanine, tyrosine and tryptophan biosynthesis” pathway (map00400) (Tab. S4).

The α-Linolenic acid metabolism pathway is one of the significant lipid metabolic pathways. Based on the findings of DEGs and DAMs in the “α-Linolenic acid metabolism” pathway (map00592), 4 fatty acids, all markedly upregulated in WA, were highly correlated with 10 unigenes. (Tab. S4).

Secondary metabolites are classified into pathways like terpenoid backbone, isoquinoline alkaloid, and phenylpropanoid biosynthesis based on initial metabolite differences. Ferulic acid, a principal bioactive compound in A. sinensis, has its biosynthetic origin in the phenylpropanoid pathway. Integrated analysis of DAMs and DEGs in phenylpropanoid biosynthesis (map00940) identified 11 unigenes strongly correlated with WA-increased coniferin: among these, five downregulated and six upregulated unigenes exhibited negative and positive correlations, respectively (Tab. S4).

To validate the transcriptomic data, 9 key genes were selected for qRT-PCR confirmation. These included five genes in amino-acids metabolism (tktA, Unigene54639-S2, metE, Unigene3787-S3, GOT1, Unigene29402-S2, GOT2, Unigene58052-S5, and ADT, Unigene21044-S4), one associated with α-Linolenic acid metabolism (ACX, Unigene16811-S4), and three linked to phenylpropanoid biosynthesis (4CL, Unigene1220-S6, PAL, Unigene17875-S6, CCR, Unigene10986-S2). The qRT-PCR results showed expression patterns consistent with the transcriptomic data (Fig. 4), supporting the reliability of the RNA-seq analysis.

Fig. 4.

Comparison of nine differentially expressed genes related to amino acid, fatty acid, and phenylpropanoid metabolism in WA vs. CA by qRT-PCR. The data represent the means from six biological repeats. Error bars indicate standard errors (SEs). Different asterisks (ns P > 0.05, * P < 0.05, ** P < 0.01, *** P < 0.001) above each column indicate significant differences between two groups

Discussion

The metabolites of A. sinensis serve as crucial defense against biotic, abiotic stresses and important traditional Chinese medicine research targets owing to their significant pharmacological activities [1]. Notably, this study compared plants derived from wild-sourced and cultivated-sourced seeds grown under identical agricultural conditions in Tanchang County. This design reveals that the significant enrichment of amino acids, fatty acids, and phenylpropanoids in wild-sourced plants, underpinned by distinct transcriptional regulation, is primarily attributed to their inherent genetic background rather than transient environmental influences. These core metabolic classes—each with specific biosynthetic and functional significance—are discussed separately below.

Amino acids, key bioactive components of A. sinensis, contribute significantly to its nutritional and medicinal value [18]. Metabolomic analysis identified 8 differentially accumulated amino acids, with 5 compounds showing higher abundance in WA than CA, while 3 compounds were downregulated. Integrated metabolomic and transcriptomic analyses revealed 1 amino acid (Asparagine) and 6 metabolic intermediates (D-Ribulose 5-phosphate, S-Adenosylmethionine, Phenylpyruvic acid, D-(-)-Quinic acid, 6-Deoxy-5-ketofructose 1-phosphate, and 3-Benzylmalic acid) enriched in amino acid synthesis pathways (map01230, map00400), all markedly upregulated in WA. Furthermore, 5 key structural genes (including tktA, metE, GOT1 and GOT2, ADT) were strongly correlated with amino acid metabolism (Fig. 3A).

Notably, the accumulation of D-Ribulose 5-phosphate, linked to tktA downregulation, may support oxidative stress adaptation via NADPH generation [19]. Increased S-Adenosylmethionine (SAM), associated with metE upregulation, likely enhances methylation capacity under stress [20]. As a nitrogen storage and transport amino acid, asparagine balances C-N ratios to regulate protein synthesis and stress adaptation [21]. Although GOT1 and GOT2 are not directly involved in asparagine synthesis, their expression correlates with asparagine accumulation, possibly reflecting a coordinated metabolic adjustment in nitrogen metabolism under stress conditions [22]. Phenylpyruvic acid, a key intermediate in phenylalanine metabolism, modulates phenylpropanoid flux to participate in stress response and secondary metabolism regulation [23]. Phenylpyruvic acid levels, modulated by GOT1, GOT2, and ADT, help direct metabolic flux toward antioxidant production. These results indicate that WA achieves optimized amino acid metabolism through coordinated expression of key genes, underscoring an adaptive mechanism to environmental challenges [24].

Fatty acids are essential for human health, playing a critical role in maintaining physiological functions [25]. Metabolomic analysis identified 20 differentially accumulated fatty acids, with 16 compounds showing higher abundance in WA than CA, while 4 compounds were downregulated. As a polyunsaturated fatty acid, α-linolenic acid maintains cell membrane stability and promotes secondary metabolite accumulation [26]. As shown in Fig. 3B, four DAMs associated with the α-linolenic acid metabolic pathway—α-linolenic acid, traumatic acid, 3,6-nonadienal, and 9-hydroxy-12-oxo-10(E),15(Z)-octadecadienoic acid—exhibited upregulated trends. Among the DEGs strongly correlated with these DAMs, ACX is particularly noteworthy. Its significant upregulation enhances the activity of ACOX, a key enzyme in the β-oxidation pathway. ACOX activity influences jasmonic acid biosynthesis, and together with upregulated traumatic acid, these function as trauma-activated repair compounds [27, 28]. Wild A. sinensis with increased fatty acid content shows improved stress tolerance (insect damage, drought, low temperature), volatile compound accumulation, and medicinal quality, as reported [29, 30].

Phenylpropanoids, a major class of secondary metabolites in A. sinensis, underpin its traditional medicinal functions of “tonifying blood and promoting circulation” [1]. Phenolic acids, particularly ferulic acid, exhibit dominant bioactivity due to their antioxidant and cardioprotective properties [31]. Among all kinds of up-regulated phenylpropanoids in WA vs. CA, comprising two phenolic acids, two coumarins, and one monolignol. Notably, key phenolic acids (ferulic and caffeic acid) remained stable, whereas coniferin (a monolignol) showed significant WA-specific upregulation. Transcriptomic profiling highlights the significant overexpression of structural genes in the phenylpropanoid pathway—PAL, 4CL, and CCR—in WA (Fig. 3C). PAL catalyzes the initial step of phenylalanine conversion to cinnamic acid, which is subsequently hydroxylated to p-coumaric acid and activated by 4CL to p-coumaroyl-CoA, directing precursors into downstream metabolic branches. CCR further reduces feruloyl-CoA to coniferyl aldehyde, a critical step that channels metabolites into lignin biosynthesis and ultimately coniferin production [32–34]. Coniferin’s upregulation in WA enhances stress resistance via its antioxidant activity and physical damage tolerance [24]. Together, these findings highlight that the enhanced accumulation of ferulic acid and its glycosylated derivatives in WA not only contributes to stress resilience but also reinforces the traditional medicinal value of wild-sourced A. sinensis.

These findings also offer significant potential for industrial application. The key genes regulating amino acid, fatty acid, and phenylpropanoid metabolism provide actionable targets for molecular breeding programs aimed at enhancing bioactive compound production in cultivated varieties. Besides, designing precision cultivation practices, such as applying targeted abiotic elicitors to imitate the wild environment, may help to boost the production of desired metabolites.

Despite these promising implications, certain technical limitations should be acknowledged: First, the untargeted metabolomics approach employed in this study, while comprehensive, is inherently limited by database coverage, especially for specialized metabolites unique to medicinal plants. Second, structural identification was further constrained by insufficient reference standards and challenges in resolving isomeric compounds. Additionally, the relative quantification provided by untargeted analysis necessitates further validation through targeted methods to absolutely confirm metabolite abundance differences. Beyond these technical considerations, it should also be noted that the “metabolite-gene” regulatory network proposed in this study is currently hypothesis‑generating in nature. Although the expression trends of key genes were validated by qRT-PCR, the causal relationships between these genes and the corresponding metabolites require further functional validation through approaches such as gene knockout, overexpression or transgenic assays. Future work will build on the present findings to functionally characterize the roles of these key genes in metabolic regulation in A. sinensis.

Future work should focus on overcoming these limitations through the integration of complementary analytical techniques such as NMR, the development of a species-specific metabolite database for A. sinensis, and the implementation of targeted metabolomics to validate and quantify key metabolites. Such efforts will not only refine our understanding of its metabolic network but also facilitate the translation of omics data into practical industrial applications.

Conclusion

This study employs an integrated multi-omics approach to elucidate the molecular basis of quality differentiation between wild and cultivated Angelica sinensis. Through combined transcriptomic and metabolomic analyses, we systematically identified 9 key structural genes (tktA, metE, GOT1, GOT2, ADT, 4CL, PAL, CCR, and ACX) primarily involved in amino acid, fatty acid, and phenylpropanoid metabolic pathways. The established “metabolite-gene” regulatory network provides mechanistic insights into the biochemical divergence. Our approach combines comprehensive metabolic profiling with targeted analysis to validate key metabolic changes at the molecular level. The identified genetic markers establish a scientific foundation for molecular marker-assisted breeding strategies, ultimately contributing to improved crop quality and nutritional value for agricultural applications.

Authors’ contributions

**Qiman Zeng: ** Investigation, Validation, Data curation, Formal analysis, Software, Writing – original draft, Writing – review & editing. **Meihui Gong: ** Writing – original draft, Methodology, Formal analysis, Investigation. **Rulan Jiang: ** Investigation, Data curation, Formal analysis Writing – review & editing. **Jieyu Lei: ** Data curation, Visualization, Validation. **Zifeng Chen: ** Data curation, Formal analysis. **Huimin Wu**: Data curation, Formal analysis. **Lianhu Sun: ** Formal analysis, Resources. **Xinjun Xu: ** Conceptualization, Funding acquisition, Writing – review & editing. **Wenli Chen: ** Conceptualization, Writing – review & editing, Project administration.

Funding

This work was supported by the National Key Research and Development Program for Young Scientists of China (no. 2023YFD1601400).

Data availability

The raw RNA sequencing data generated during the current study are available in the NCBI Sequence Read Archive (SRA) repository, under BioProject accession number PRJNA1418440 (https://dataview.ncbi.nlm.nih.gov/object/PRJNA1418440).

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

All authors listed have read the complete manuscript and have approved submission of the paper.

Competing interests

The authors declare no competing interests.