Abstract

Background

   Foxtail millet (Setaria italica L.), a traditional Chinese crop, is
   valued for its rich abundance of health-beneficial compounds (e.g.,
   flavonoids). Despite the nutritional significance of flavonoids, their
   biosynthetic pathways in foxtail millet remain largely uncharacterized.
   In this study, we integrated targeted metabolomic and transcriptomic
   analyses to comprehensively dissect the flavonoid biosynthesis pathway
   and identify genes encoding key enzymes.

Results

   Quantitative profiling of a foxtail millet recombinant inbred line
   (RIL) population revealed significant variations in grain flavonoid
   content, with levels in the high-flavonoid (HF) group exceeding those
   in the low-flavonoid (LF) group by five-fold. Quantitative analysis of
   a foxtail millet recombinant inbred line (RIL) population showed
   striking variations in grain flavonoid content. The high-flavonoid (HF)
   group exhibited flavonoid levels five-fold higher than those of the
   low-flavonoid (LF) group. Through integrated targeted metabolomic and
   transcriptomic analyses, we elucidated the pivotal regulatory networks
   governing flavonoid biosynthesis in foxtail millet. Comparative
   transcriptomic analysis demonstrated significant differences in the
   expression patterns of flavonoid biosynthesis-associated genes between
   the HF and LF groups. Targeted metabolomic quantification further
   identified ten specific flavonoids with significantly elevated levels
   in the HF group. Integrative omics analyses revealed that genes
   encoding shikimate O-hydroxycinnamoyl transferase (HCT), phenylalanine
   ammonia-lyase (PAL), and phenylalanine/tyrosine ammonia-lyase (PTAL)
   are key determinants of the differential flavonoid accumulation
   observed between genotypes.

Conclusions

   Our study provides novel insights into the genetic regulatory
   mechanisms governing flavonoid metabolism in foxtail millet. The
   identified candidate genes, including those encoding HCT, PAL, and
   PTAL, represent valuable molecular targets for breeding programs
   focused on improving the nutritional profile of foxtail millet
   cultivars.

Supplementary Information

   The online version contains supplementary material available at
   10.1186/s12864-025-11780-x.

   Keywords: Foxtail millet, Targeted metabolomics, Transcriptomics,
   Flavonoid biosynthesis

Background

   Foxtail millet (Setaria italica L.), a nutrient-rich cereal
   domesticated from its wild progenitor (Setaria viridis) in northern
   China more than 8,000 years ago, has emerged as a promising functional
   food crop because of its natural antioxidant properties [[44]1, [45]2].
   Nutritionally, this drought-tolerant species is similar to major
   cereals, such as wheat and rice, making it an important source of
   protein, dietary fiber, and essential minerals [[46]3, [47]4]. In
   addition, foxtail millet accumulates diverse phytochemicals, including
   flavonoids, phenolic acids, and phytosterols, which collectively
   contribute to its health-promoting effects [[48]5].

   Flavonoids, a major class of plant secondary metabolites, are
   structurally diverse compounds characterized by variations in
   hydroxylation patterns, oxidation states, and substituent positions,
   leading to their classification into distinct subclasses: anthocyanins,
   flavanols, flavanones, flavones, flavonols, and isoflavones [[49]6].
   These phytochemicals, particularly flavonoids and phenolic acids, have
   dual roles as chromatic determinants responsible for grain pigmentation
   (ranging from white to black) and bioactive agents with antioxidant and
   anti-inflammatory properties [[50]7]. Flavonoids are known to possess
   antioxidant, anti-inflammatory, anticancer, cardiovascular protective,
   and antiviral properties [[51]8–[52]11], and may alleviate
   inflammation-related conditions such as arthritis and sore throat.
   Their anticancer properties are also reported to enhance immune
   function [[53]12, [54]13]. To date, a total of 116 flavonoids have been
   identified in foxtail millet, including catechins, rutin, vitexin,
   luteolin, kaempferol, and quercetin [[55]14]. The biosynthetic pathways
   of flavonoids have been extensively characterized in model plants such
   as Malus pumila Mill [[56]15]., Salvia miltiorrhiza [[57]16],
   Arabidopsis thaliana, and Zea mays [[58]17]. However, compared to the
   comprehensive genomic and metabolomic studies on major crops (e.g.,
   Oryza sativa and Z. mays) and model organisms (e.g., A. thaliana),
   research on foxtail millet—an agronomically important drought-tolerant
   cereal—remains limited, yielding fragmented findings and lacking
   systemic integration. Previous research on foxtail millet focused on
   determining nutrient contents, analyzing gene expression via
   RNA-sequencing (RNA-seq), and cloning and validating functional genes
   [[59]18–[60]20]. Several recent studies combined transcriptomic and
   metabolomic analyses to explore regulatory mechanisms related to
   certain functional components. Such studies have been conducted for
   honeysuckle [[61]21], cucumber [[62]22], Dendrobium species [[63]23],
   and peanut [[64]24]. He et al. [[65]25] reported that carotenoid
   cleavage dioxygenase helps degrade lutein, affecting carotenoid
   accumulation and color development in foxtail millet. In this study, we
   employed integrated transcriptomic and metabolomic approaches to
   investigate flavonoid biosynthesis in foxtail millet grains. Our
   findings provide critical genetic resources and molecular markers for
   breeding nutrient-dense foxtail millet cultivars, which could enhance
   agricultural productivity and socioeconomic sustainability in
   millet-producing regions.

Results

Analysis of flavonoid content variations in a recombinant inbred line (RIL)
population

   We determined the total flavonoid content in mature grains obtained
   from a population comprising 180 foxtail millet RILs. The generated
   dataset revealed a continuous phenotypic variation in flavonoid
   accumulation, which followed a normal distribution (Fig. [66]1A). These
   results indicate that grain flavonoid content is a quantitative trait
   influenced by the environment. We hypothesized that multiple genes
   co-regulate this trait. Because of the complexity of quantitative
   traits in cereals and the difficulty in localizing related genes, we
   constructed an RIL population and then conducted targeted metabolomic
   and transcriptomic analyses to map foxtail millet flavonoid
   biosynthesis-related genes. The RILs were ranked on the basis of the
   grain flavonoid content (from highest to lowest). The 30 lines with the
   highest flavonoid contents and the 30 lines with the lowest flavonoid
   contents were selected as a mixed pool. The flavonoid content was
   approximately five-fold higher in the HF RIL group than in the LF RIL
   group (Fig. [67]1B).

Fig. 1.

   [68]Fig. 1
   [69]Open in a new tab

   Population flavonoid content. A Frequency distribution of the grain
   flavonoid content in the recombinant inbred line (RIL) population. B
   Grain flavonoid contents in the high-flavonoid (HF) and low-flavonoid
   (LF) groups. Data are presented as the mean ± SD (n = 30). Asterisks
   indicate statistically significant differences according to Student’s
   t-test (P < 0.01)

Identification of metabolites

   Metabolomic profiles were compared between the HF and LF foxtail millet
   groups. On the basis of a principal component analysis (PCA), PC1 and
   PC2 explained 69.30% and 21.50% of the variance in grain metabolomic
   profiles, respectively (Fig. [70]2A). An orthogonal partial least
   squares discriminant analysis (OPLS-DA) of the foxtail millet grain
   metabolite data resulted in R^2Y and Q^2 values of 0.940 and 0.997,
   respectively (Fig. [71]2B), indicating high model stability and
   reliability. Both PCA and OPLS-DA score plots revealed the significant
   separation between the LF and HF groups, indicative of the differences
   in metabolism between the two groups.

Fig. 2.

   [72]Fig. 2
   [73]Open in a new tab

   PCA and OPLS-DA score plots of foxtail millet grain metabolite
   profiles. A PCA score plot of LF and HF samples; Principal component 1
   (PC1) and principal component 2 (PC2) are presented on the x- and
   y-axes, respectively. B Permutation test for OPLS-DA of LF vs HF. LF,
   low-flavonoid lines; HF, high-flavonoid lines

Identification of differentially accumulated flavonoid metabolites

   A total of 40 flavonoid metabolites were detected in the LF and HF
   groups (three biological replicates per line). Supplementary Table S1
   provides detailed information about the detected flavonoid metabolites,
   including their molecular weight, formula, class, and Kyoto
   Encyclopedia of Genes and Genomes (KEGG) ID. Flavonoids (32.5%), acids
   (22.5%), flavanols (12.5%), flavonols (10%), dihydroisoflavones (7.5%),
   and chalcones (7.5%) were the main flavonoid metabolites (Fig. [74]3A).

Fig. 3.

   [75]Fig. 3
   [76]Open in a new tab

   Putatively annotated metabolites and enriched pathways among
   differentially accumulated metabolites (DAMs). A Putatively annotated
   metabolites in the LF and HF groups. B Classes of DAMs in the LF and HF
   groups and their associated KEGG pathways. LF, low-flavonoid lines; HF,
   high-flavonoid lines

   After pre-treatments, such as filtering individual metabolites and
   simulating missing values in the original data, 36 differentially
   accumulated metabolites (DAMs) were detected between the HF and LF
   groups. To functionally characterize flavonoid DAMs and identify their
   associated metabolic pathways, a KEGG pathway enrichment analysis was
   conducted, which revealed the flavonoid pathway was significantly
   enriched among the DAMs in the HF and LF groups (Fig. [77]3B). These
   results further confirmed the substantial differences in flavonoid
   metabolism between the HF and LF foxtail millet groups.

   A comparison between the HF and LF groups indicated that among the
   identified DAMs, 10 were up-regulated and 26 were down-regulated in the
   HF group. Notably, flavonoid DAMs constituted the majority of
   up-regulated metabolites in the LF group (Supplementary Table S2).
   Moreover, flavonoid DAMs, including vanillin, syringaldehyde,
   (+)-dihydrokaempferol, naringenin chalcone, naringenin, apigenin,
   protocatechualdehyde, and trans-ferulic acid, were the main
   up-regulated DAMs (approximately 50%) revealed by the HF vs. LF
   comparison.

Transcriptome analysis

   An RNA-seq analysis was performed to examine the gene expression
   profiles of HF and LF foxtail millet grains. The raw data were
   deposited in the National Center for Biotechnology Information (NCBI)
   database ([78]https://www.ncbi.nlm.nih.gov/sra/PRJNA1183448). The
   sequencing generated 42.81 Gb of clean data per sample, with GC content
   exceeding 57.3% and Q30 Phred scores ranging from 91.58 to 92.95%
   (Supplementary Table S3).

   According to a comparison with the LF group (control), 582 genes were
   differentially expressed in the HF group (146 up-regulated genes and
   436 down-regulated genes) (Fig. [79]4A). To functionally annotate the
   DEGs, Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes
   (KEGG) analyses were conducted. On the basis of a GO analysis, DEGs
   were annotated with 399 GO terms from the three main categories
   (molecular function, cellular component, and biological process). The
   enriched GO terms assigned to DEGs in grains included the following:
   “movement of cell or subcellular component” (GO:0006928, nine DEGs),
   “microtubule-based movement” (GO:0007018, nine DEGs), and “chromosome
   organization” (GO:0051276, eight DEGs) from the biological process
   category; “nutrient reservoir activity” (GO:0045735, seven DEGs),
   “protein heterodimerization activity” (GO:0046982, 11 DEGs),
   “microtubule motor activity” (GO:0003777, nine DEGs), and “microtubule
   binding” (GO:0008017, nine DEGs) from the molecular function category;
   and “chromosome” (GO:0005694, three DEGs), “chromosomal part”
   (GO:0044427, two DEGs), and “extracellular region” (GO:0005576, two
   DEGs) from the cellular component category (Supplementary Fig. S1).
   Hence, the DEGs were predominantly associated with subcellular
   component motility and microtubule dynamics regulation, which is in
   accordance with the transport mechanisms and antioxidant properties of
   flavonoid compounds in plants.

Fig. 4.

   [80]Fig. 4
   [81]Open in a new tab

   Transcriptomic Profiling and Pathway Enrichment Analysis. A Volcano
   plot of up- and down-regulated differentially expressed genes in the HF
   and LF groups. LF, low-flavonoid lines; HF, high-flavonoid lines. B
   KEGG pathway classification of differentially expressed genes in
   high-flavonoid and low-flavonoid foxtail millet grains

   KEGG pathway enrichment analysis of the DEGs was conducted [[82]26];
   the top 20 enriched KEGG pathways among the DEGs in the HF and LF
   groups are presented in Fig. [83]4B. Notably, four and two DEGs were
   involved in phenylpropanoid biosynthesis and phenylalanine metabolism,
   respectively. Among the DEGs involved in the phenylpropanoid
   biosynthesis pathway, two PAL genes (phenylalanine ammonia-lyase) and
   SPC4 (cationic peroxidase) were up-regulated in the HF group, whereas
   POD (peroxidase), ACT (agmatine coumaroyltransferase), and Met
   (methyltransferase) were down-regulated in the HF group (Table [84]1).

Table 1.

   Enriched flavonoid biosynthesis-related KEGG pathways among DEGs
   Expression patern Gene ID Gene name Descrition Log[2](FoldChange)
   Up regulated Seita.1G240600 LOC101781925 Phenylalanine ammonia-lyase
   (PAL) 2.96087934
   Seita.5G462500 LOC111256430 Cationic peroxidase SPC4 (CAT) 3.161535713
   Seita.6G181000 LOC101766005) Phenylalanine tyrosine/ammonia-lyase
   (PTAL) 1.461248461
   Down regulated Seita.2G431300 LOC111256430 Peroxidase 70 (POD)
   −1.935260709
   Seita.8G188600 LOC101758480 Agmatine coumaroyltransferase-2
   −5.291043079
   Seita.8G210100 LOC101772386 Probable inactive methyltransferase Os04
   g0175900 −2.516508672
   [85]Open in a new tab

Verification of RNA-seq data

   To validate RNA-seq reliability, eight DEGs were analyzed via
   quantitative real-time PCR (qRT-PCR). The strong concordance between
   qRT-PCR and RNA-seq results confirmed the robustness of the
   transcriptomic data for molecular analyses and gene discovery (Fig.
   [86]5).

Fig. 5.

   [87]Fig. 5
   [88]Open in a new tab

   Validation of the RNA-seq data of eight DEGs in foxtail millet via
   qRT-PCR. Data are presented as the mean± SD of three biological
   replicates. LF, low-flavonoid lines; HF, high-flavonoid lines

Correlation analysis of metabolomic and transcriptomic data

   Correlations between DAMs and DEGs were assessed on the basis of a
   combined analysis of metabolomic and transcriptomic data. We
   constructed a nine-quadrant diagram to visualize the relationships
   between DAMs and their associated genes (Spearman’s rank correlation
   coefficient ≥ 0.8). Most of the DAMs and DEGs in foxtail millet grains
   were clustered in quadrant 1 (those with positive correlations) and
   quadrant 7 (those with negative correlations) (Fig. [89]6).

Fig. 6.

   [90]Fig. 6
   [91]Open in a new tab

   Associations between differentially accumulated metabolites (DAMs) and
   differentially expressed genes (DEGs) in high-flavonoid (HF) and
   low-flavonoid (LF) foxtail millet grains. The x-axis presents the
   log[2]ratio of metabolite abundance, whereas the y-axis presents the
   log[2] ratio of transcript abundance. The dashed line on the horizontal
   coordinate represents the fold difference threshold for metabolites,
   whereas the dashed line on the vertical coordinate represents the fold
   difference threshold for genes. Metabolites/genes with significant
   differences in abundance/expression between the HF and LF groups are
   outside of the threshold line, while those without significant
   differences are inside the threshold line. Each point represents one
   metabolite/gene. Blue dots represent DAMs associated with DEGs
   (up-regulated or down-regulated), while green dots represent DAMs whose
   associated genes had expression levels that did not differ between the
   HF and LF groups

   To further characterize flavonoid biosynthetic pathways in foxtail
   millet grains, we conducted pathway and regulatory network analyses of
   DAMs and DEGs related to phenylpropanoid biosynthesis and phenylalanine
   metabolism (Fig. [92]7). The following three genes, which are the main
   genes involved in flavonoid biosynthesis, were highly correlated (r >
   0.8 or r < − 0.8) with 15 DAMs (Supplementary Table S4): Seita.8G188600
   (HCT), Seita.1G240600 (PAL), and Seita.6G181000 (PTAL). Similar
   correlation analyses of DEGs and DAMs associated with flavonoid
   biosynthesis and phenylalanine metabolism indicated that Seita.8G188600
   (HCT) was highly correlated (r > 0.8 or r < − 0.8) with 14 flavonoids
   related to the flavonoid biosynthesis pathway, whereas Seita.1G240600
   (PAL) and Seita.6G181000 (PTAL) were highly correlated (r > 0.8 or r <
   − 0.8) with six flavonoids related to phenylalanine metabolism (Fig.
   [93]7 and Supplementary Table S4). PAL activity drives phenylpropanoid
   metabolism, converting phenylalanine to cinnamic acid, which is
   hydroxylated to p-coumaric acid and further to p-coumaroyl-CoA—a
   critical precursor for flavonoid biosynthesis.

Fig. 7.

   [94]Fig. 7
   [95]Open in a new tab

   Network of DEGs related to flavonoid biosynthesis, phenylpropanoid
   biosynthesis, and phenylalanine metabolism and flavonoid DAMs. The
   pathways indicated by red nodes include sita00941 (flavonoid
   biosynthesis), sita00940 (phenylpropanoid biosynthesis), and sita00360
   (phenylalanine metabolism)

Flavonoid biosynthesis-related gene expression

   A qRT-PCR analysis was completed to verify the relative expression of
   three genes encoding key enzymes involved in flavonoid biosynthesis in
   grains (Fig. [96]8). The results showed that PAL (Seita.1G240600) and
   PTAL (Seita.6G181000) expression levels were significantly up-regulated
   in the HF group, whereas HCT (Seita.8G188600) expression was
   up-regulated in the LF group. These findings indicate that the
   expression of flavonoid biosynthesis-related genes, especially PAL, is
   induced to varying degrees in foxtail millet grains.

Fig. 8.

   [97]Fig. 8
   [98]Open in a new tab

   Comparison of flavonoid biosynthesis-related gene expression levels in
   HF and LF foxtail millet groups.Seita.1G240600, Seita.6G181000, and
   Seita.8G188600 are PAL,PTAL, and HCT genes, respectively. The y-axis
   presents relative gene expression levels. Each value is the mean of
   three replicates, and error bars indicate standard deviations. All
   gene-specific primer sequences are provided in Supplementary Table S5

Discussion

   Previous studies on critical metabolic genes in plants have employed
   integrated metabolomic and transcriptomic analyses across diverse plant
   organs [[99]27–[100]32]. Flavonoids, a major class of phytochemicals in
   crops, regulate multiple essential biological processes. A total of 14
   transcription factor genes and 17 flavonoid biosynthesis-related genes
   have been characterized in peanut testa [[101]33]. However, flavonoid
   composition and its regulatory mechanisms in foxtail millet grains
   remain poorly understood. Earlier research demonstrated that red-grain
   foxtail millet exhibits substantially higher total flavonoid content
   than yellow-grain varieties [[102]34]. In the current study, the total
   flavonoid content was five-fold higher in the HF group than in the LF
   group (Fig. [103]1B). Candidate genes involved in the accumulation of
   flavonoids in grains were preliminarily identified following a
   comparison between the HF and LF groups.

   Transcriptome libraries were constructed for HF and LF foxtail millet
   grains for the subsequent high-throughput sequencing analysis conducted
   to identify nutrients and their biosynthetic mechanisms. Moreover,
   flavonoid metabolites in foxtail millet grains were identified using an
   ultra-performance liquid chromatography-electrospray ionization-tandem
   mass spectrometry (UPLC-ESI-MS/MS) system. After sequences were
   assembled, 582 DEGs in the HF and LF groups were identified.
   Additionally, 36 flavonoid metabolites in foxtail millet grains were
   identified using metabolomics techniques. These transcriptomic and
   metabolomic datasets establish a critical foundation for exploring
   metabolic pathways in foxtail millet. As key nutritional components,
   flavonoids in foxtail millet grains showed differential accumulation
   across multiple subclasses, including flavonoids, phenylpropanoids,
   dihydroflavones, isoflavones, and flavonoid glycosides. KEGG pathway
   analysis confirmed significant enrichment of associated biosynthetic
   pathways, providing crucial insights into molecular mechanisms
   governing flavonoid accumulation.

   Metabolomic analysis of foxtail millet grains detected 10 flavonoid
   metabolites significantly enriched in the HF group, including
   naringenin, naringenin chalcone, apigenin, vanillin, syringaldehyde,
   trans-ferulic acid, and phthalic acid. Naringenin has demonstrated
   diverse bioactivities, such as antioxidant, anti-inflammatory, and
   metabolic regulatory effects). Several flavonoids isolated from the
   peanut testa, such as cyanidin, peonidin, and pelargonidin, reportedly
   have significant anti-inflammatory, immune-enhancing, and anticancer
   properties [[104]35]. Apigenin has some medicinal properties. More
   specifically, it can decrease the risk of cancer [[105]36], has
   neuroprotective properties [[106]37], and lowers blood glucose and
   lipid levels [[107]38]. While foxtail millet consumption confers health
   benefits, our metabolomic data reveal differential accumulation of
   flavonoid-related metabolites between HF and LF groups, suggesting that
   targeted modification of grain groups could optimize flavonoid
   utilization and efficacy. These findings are relevant to future
   evaluations of crop plants, with potential implications for the
   applicability of different foxtail millet groups.

   Limited genetic resources have hindered research on foxtail millet
   metabolic pathways. To further elucidate the regulation of flavonoid
   accumulation in different foxtail millet groups, DEGs in HF and LF
   grains identified following an RNA-seq analysis were subjected to a
   KEGG pathway enrichment analysis. The results revealed the enrichment
   of flavonoid biosynthesis, flavone and flavonol biosynthesis,
   isoflavonoid biosynthesis, and upstream phenylpropanoid biosynthesis
   pathways in all samples. The basic flavonoid metabolic pathways in
   plants have been characterized, with various genes (e.g., CHS, CHI,
   CYP75 A, ANR, FLS, and DFR) encoding proteins with key roles in the
   synthesis of bioactive components in plants [[108]39–[109]47]. In the
   present study, PAL and PTAL expression levels were higher in HF grains
   than in LF grains (Fig. [110]8), which led to the increased
   accumulation of phenylalanine ammonia-lyase. The results of the
   UPLC-ESI-MS/MS-based analysis were also confirmed (Supplementary Table
   S4). Phenylalanine ammonia-lyase is a key enzyme in the flavonoid
   biosynthesis pathway, but the enzymes encoded by different PAL genes
   may have the same or different functions [[111]48]. In the current
   study, PAL transcript levels positively correlated with apigenin,
   vanillin, and naringenin, but PTAL expression negatively correlated
   with salicylic acid and catechin (Supplementary Table S4). In A.
   thaliana, PAL and PTAL genes encode proteins with important functions
   related to flavonoid biosynthesis [[112]49]. The main flavonoids that
   accumulated more in the HF group than in the LF group were naringenin
   and aldehyde derivatives, such as naringenin chalcone, vanillin, and
   syringaldehyde, suggesting that PAL catalyzes the conversion of
   chalcone to naringenin. Other substrates that are specifically bound by
   PAL are ferulic acid, vanillic acid, and syringic acid [[113]50].

   HCT, a member of the plant acyltransferase family, catalyzes reactions
   with diverse substrates, such as quercetin, kaempferol, quinic acid,
   and gentianic acid, yielding ester or amide derivatives. HCT plays
   pivotal roles in chlorogenic acid and lignin biosynthesis
   [[114]51–[115]53]. The roles of HCT in lignin biosynthesis were
   recently validated, while its involvement in the synthesis of
   chlorogenic acid and flavonoids has also been determined. In A.
   thaliana, inhibited AtHCT expression is accompanied by a decrease in
   lignin accumulation, but an increase in flavonoid levels [[116]54]. In
   foxtail millet, HCT catalyzes the conversion of p-coumaroyl-CoA to
   caffeoyl-CoA, which is further processed by transferases into lignin
   precursors. In this study, significant downregulation of HCT in HF
   grains disrupted lignin precursor synthesis, redirecting metabolic flux
   toward flavonoid biosynthesis via an alternative pathway. This shift
   correlated with marked upregulation of naringenin chalcone production,
   demonstrating a trade-off between lignin and flavonoid metabolism.

   Integrated transcriptomic and targeted metabolomic analyses in this
   study reveal novel insights into flavonoid biosynthesis in foxtail
   millet. However, additional research is required to comprehensively
   clarify the specific mechanisms regulating flavonoid compositions and
   contents in foxtail millet. In general, flavonoid synthesis in foxtail
   millet involves dynamic interactions among multiple regulatory
   networks. Ultimately, flavonoid compositions and contents in foxtail
   millet grains depend on the coordinated effects of the overall
   regulatory network.

Conclusion

   The total flavonoid content in foxtail millet grains was significantly
   higher in the HF group than in the LF group. Comparative transcriptomic
   and metabolomic profiling of HF and LF groups identified critical
   factors and candidate genes governing flavonoid accumulation. Notably,
   PAL, PTAL, and HCT were identified as genes encoding key regulators of
   flavonoid biosynthesis in foxtail millet. Both PAL and PTAL positively
   regulate flavonoid accumulation, while HCT negatively regulates
   flavonoid synthesis and accumulation. These study results provide new
   insights relevant to breeding and selecting foxtail millet lines with
   improved nutrient profiles.

Materials and methods

Plant materials and sample collection

   The RIL population included in this study was derived from a cross
   between the HF variety ‘Jinmiaohongjiugu’ (maternal parent) and the LF
   variety ‘Yugu28’ (paternal parent). The RIL population consisted of 325
   lines, with 180 lines selected for the screening of HF and LF lines
   based on the grain flavonoid content at maturity. The 30 lines with the
   highest grain flavonoid contents were included in the HF group, whereas
   the 30 lines with the lowest grain flavonoid contents were included in
   the LF group. Three biological replicates of mature grain samples were
   collected and immediately frozen in liquid nitrogen before being stored
   at − 80 °C until the subsequent transcriptomic and metabolomic
   analyses.

Determination of grain flavonoid contents

   Grain flavonoid contents were determined for the 180 selected RILs
   using a plant flavonoid content determination kit (Beijing Solarbio
   Science & Technology Co., Ltd., Beijing, China). Millet grain samples
   were dried and pulverized and then passed through a 50-mesh sieve.
   Approximately 0.1 g powdered sample was added to 1 mL extraction
   solution, after which the mixture was sonicated (300 W for 30 min at 60
   °C). The mixture was centrifuged (12,000 rpm for 10 min at 25 °C) and
   the supernatant was retained. Test reagents were mixed thoroughly and
   then incubated in a water bath (37 °C) for 45 min. The mixture was
   centrifuged (1,000 rpm for 10 min at room temperature) and the
   supernatant was retained. The absorbance of the supernatant was
   measured at 470 nm, after which a rutin standard curve was used to
   determine the flavonoid content. This analysis was completed three
   times per sample. The rutin standard curve was prepared by diluting a
   rutin standard solution (10 mg/mL) to 1.5, 1.25, 0.625, 0.3125,
   0.15625, 0.078, 0.039, and 0.02 mg/mL using a standard diluent. SPSS
   software was used to analyze the grain flavonoid contents of 180 RILs,
   whereas Excel software was used to select 30 lines with low grain
   flavonoid contents and 30 lines with high grain flavonoid contents to
   construct a mixed pool.

Targeted metabolomic profiling and analysis

   Sample preparation and extraction: Frozen plant samples (100 mg each)
   were sonicated for 30 min using 500 µL 80% (v/v) methanol containing
   0.2% (w/v) vitamin C. The mixture was centrifuged (12,000 rpm for 10
   min) and the supernatant was retained. These steps were repeated twice
   and the resulting supernatants were combined.

   UPLC conditions: Sample extracts were analyzed using a UPLC-Orbitrap-MS
   system (UPLC, Vanquish; MS, QE; Thermo Fisher Scientific, Waltham, MA,
   USA). Analytical conditions were as follows: UPLC: column, Waters HSS
   T3 (50 × 2.1 mm, 1.8 μm); column temperature, 40 °C; flow rate, 0.3
   mL/min; injection volume, 2 µL; solvent system, water (0.1% v/v acetic
   acid): acetonitrile (0.1% v/v acetic acid); gradient program, 90:10
   (v/v) at 0 min, 90:10 (v/v) at 2.0 min, 40:60 (v/v) at 6.0 min, 40:60
   (v/v) at 8.0 min, 90:10 (v/v) at 8.1 min, and 90:10 (v/v) at 12.0 min
   [[117]55].

   LC-MS/MS analysis: HRMS data were recorded using a Q Exactive hybrid
   Q-Orbitrap mass spectrometer equipped with a heated ESI source (Thermo
   Fisher Scientific) according to fullms-ms2 acquisition methods. ESI
   source parameters were as follows: sheath gas, 40 arb; auxiliary gas,
   10 arb; spray voltage, − 2,800 V; temperature, 350 °C; and ion transfer
   tube temperature, 320 °C. The single ion monitor mode (negative ion
   mode) was applied [[118]56].

   Public metabolite databases (e.g., MassBank, KNAPSAcK, HMDB, and
   METLIN) and the MVDB v2.0 database from Smart Biotechnology Co., Ltd.
   (Tianjin, China) were used for the qualitative analysis of primary and
   secondary mass spectrometry data. Accordingly, structural
   characteristics of the metabolites were determined. For the
   quantitative analysis of metabolites, MRM was used and PCA and OPLS-DA
   were performed to identify DAMs. We divided foxtail millet lines into
   LF and HF groups for a comparative analysis. The following thresholds
   were used to identify significant DAMs: VIP > 1 and|log[2](HF group/LF
   group)| ≥ 1.

RNA isolation and transcriptomic analysis

   Total RNA was isolated and purified from foxtail millet grains using a
   Complex Plant RNA Rapid Extraction Kit (Beijing Labhelper Biotechnology
   Co., Ltd., Beijing, China). Samples were treated with an RNase-free
   DNase I digestion kit (Aidlab, Beijing, China) to remove genomic DNA
   contaminants. RNA degradation was assessed using 1% agarose gels, while
   RNA concentrations were determined using a NanoDrop 2000
   spectrophotometer (Thermo Scientific, Wilmington, DE, USA).

   After verifying the quality of the constructed libraries using a 2100
   Bioanalyzer (Agilent, Palo Alto, CA, USA), they were sequenced using an
   Illumina HiSeq™ 2000 system (Illumina, San Diego, CA, USA). Raw data
   were filtered by removing reads with adapter sequences and low-quality
   reads using fastp software [[119]57]. The remaining clean reads were
   assembled to obtain unigenes, which were functionally annotated on the
   basis of NCBI non-redundant protein sequences (Nr;
   [120]https://fp.ncbi.nlm.nih.gov/blast/db/FASTA/) as well as sequences
   in the KEGG ([121]https://www.genome.jp/kegg), Clusters of Orthologous
   Groups of proteins (COG; [122]https://www.ncbi.nlm.nih.gov/COG/), GO
   ([123]https://www.geneontology.org), Swiss-Prot
   ([124]http://www.ebi.ac.uk/uniprot/), and translation of EMBL (TrEMBL)
   databases. Similarly, we conducted a comparative analysis of the HF and
   LF groups. DEGs were screened using the following criteria:
   |log[2](fold-change)| > 1 and P < 0.05. DEG biological functions and
   associated metabolic pathways were determined using GO
   ([125]http://geneontology.org/) and KEGG
   ([126]http://www.genome.jp/kegg/) databases [[127]58–[128]60].

Integrated analysis of metabolomic and transcriptomic data

   Transcriptomic and metabolomic data for the grains in the HF and LF
   foxtail millet groups were analyzed. DEGs related to flavonoid
   biosynthesis and flavonoid DAMs in each comparison group (three
   replicates each) were subjected to a correlation analysis. Spearman’s
   correlation coefficients were used to determine differences between
   flavonoid biosynthesis-related genes and transcription factor genes
   (correlation coefficient ≥ 0.90 and P ≤ 0.05).

qRT-PCR analysis

   On the basis of metabolomic, transcriptomic, and combined analyses,
   eight candidate DEGs identified by the RNA-seq analysis were selected
   for a qRT-PCR analysis. RNA was reverse transcribed to cDNA using a
   PrimeScript™ RT reagent kit (Tolo Biotech, Shanghai, China). For the
   qRT-PCR analysis, 2× Q3 SYBR qPCR Master Mix (Tolo Biotech) and a CFX96
   Real-Time PCR Detection System (Bio-Rad, USA) were used. Gene-specific
   qRT-PCR primers (Supplementary Table S5) were designed using Beacon
   Designer 8.0. The qRT-PCR conditions were as follows: pre-denaturation
   at 95 °C for 30 s; 40 cycles of 95 °C for 30 s, 59 °C for 30 s, and 72
   °C for 30 s. Relative gene expression levels were calculated according
   to the 2^−ΔΔCT method, with three biological replicates per gene
   [[129]61].

Statistical analysis

   Transcriptomic and metabolomic analyses were completed using three
   biological replicates. Statistical analyses were performed using SPSS
   22.0 software (IBM, Chicago, IL, USA). An independent samples t-test
   was used for the statistical comparison of HF and LF groups.

Supplementary Information

   [130]Supplementary Material 1.^ (118.1KB, jpg)
   [131]Supplementary Material 2.^ (14.4KB, xlsx)
   [132]Supplementary Material 3.^ (17.3KB, xlsx)
   [133]Supplementary Material 4.^ (9.8KB, xlsx)
   [134]Supplementary Material 5.^ (48KB, xlsx)
   [135]Supplementary Material 6.^ (10.8KB, xlsx)

Acknowledgements