Abstract Consumer-driven demand for premium foxtail millet necessitates systematic identification of quality-determining metabolites. This investigation employed integrated metabolomic and transcriptomic analyses to compare two contrasting cultivars - the elite variety Jingu 21 (JG21) and traditional landrace Niumaobai (NMB) - during two critical grain-filling stages. 552 metabolites were identified in both cultivars, with 144 showing differential abundance. Comparative analysis of early (S2) to late (S4) grain-filling stage revealed 108 co-regulated metabolites alongside 78 JG21-exclusive and 72 NMB-specific metabolites, exhibiting differential accumulation patterns likely governing quality divergence. Differential metabolites were predominantly enriched in flavonoid and phenylpropanoid pathways between cultivars, suggesting potential roles in modulating color, nutritional quality, and grain texture. Co-expression network predictions revealed cultivar-specific regulatory associations, with 10 candidate genes potentially governing six pigmentation/nutrition-related flavonoid metabolites, while another four genes showed tentative correlations with five lignin pathway intermediates that may contribute to grain texture variations between JG21 and NMB cultivars. These findings provide mechanistic insights into metabolic determinants of millet quality while establishing a framework for targeted breeding strategies to enhance both nutritional value and sensory characteristics in foxtail millet. Keywords: Foxtail millet, Grain quality, Grain-filling stages, Metabolome, Transcriptome Highlights * • Metabolic profiling shows variations between divergent foxtail millet cultivars. * • Two cultivars exhibit shared and unique metabolic pathways during grain-filling. * • Differential metabolites enriched in flavonoid and phenylpropanoid pathways. * • A gene-metabolite regulatory network was constructed in key metabolic pathways. 1. Introduction Foxtail millet (Setaria italica), domesticated in northern China, is now a climate-resilient crop cultivated across arid/semi-arid regions for sustainable agriculture under water scarcity ([35]He, Lu, Zhang, & Wang, 2022). Through long-term adaptation to harsh environments such as drought, poor soil, and high salinity, foxtail millet has developed the ability to accumulate a diverse array of metabolites that support its growth and survival ([36]Arrivault et al., 2019). These metabolites not only benefit the plant but also offer health advantages for humans, including regulating blood glucose levels, reducing inflammation, and promoting digestive health ([37]Zhang et al., 2020). Unlike rice and wheat, foxtail millet is naturally gluten-free ([38]Abdollahzadeh, Vazifedoost, Didar, Haddadkhodaprast, & Armin, 2023) and exhibit superior nutritional profiles, containing higher levels of nutraceuticals including dietary fiber, phenolic antioxidants, and essential minerals (e.g., iron, zinc), alongside balanced macronutrient composition ([39]Arora, Aggarwal, Dhaliwal, Gupta, & Kaushik, 2023). As a potential C4 model plant, foxtail millet has attracted considerable attention from researchers in functional genomics due to its short stature ([40]Yang et al., 2020), high photosynthetic efficiency ([41]Chen et al., 2023), and exceptional drought tolerance ([42]Zhang, Xiao, Yi, & Yu, 2023). However, understanding of the metabolites and their regulatory genes involved in foxtail millet growth and development remains limited. Metabolites, low-molecular-weight organic compounds, are classified into primary metabolites, secondary metabolites, and plant hormones based on their functions in plant growth, pigmentation, and responses to biotic and abiotic stresses ([43]Erb & Kliebenstein, 2020). Utilizing metabolomics to dissect crop metabolic dynamics has become an effective strategy for identifying key functional metabolites and elucidating the mechanisms underlying agronomic trait formation, with applications across diverse crop species ([44]Xu et al., 2024; [45]Yang et al., 2024). Over 300 metabolites have been identified in foxtail millet, including flavonoids, phenolamides, hydrocinnamoyl derivatives, vitamins, and lysophosphatidylcholines (LPCs) across various tissues ([46]Li et al., 2018). During the grain-filling stage, key metabolic pathways such as flavonoid biosynthesis, glutathione metabolism, linoleic acid metabolism, starch and sucrose metabolism, and the biosynthesis of valine, leucine, and isoleucine play critical roles ([47]Wang et al., 2023). Genes involved in photosynthesis and phenylpropanoid biosynthesis are downregulated during this stage, while those related to peroxisome function, purine metabolism, and zeatin biosynthesis are upregulated ([48]Song et al., 2022). Additionally, 54 carotenoid metabolism-related genes have been identified in foxtail millet panicles, with their promoters primarily associated with plant hormones, drought stress resistance, and endosperm- and seed-specific cis-acting elements ([49]Li et al., 2022). Studies have shown that SiADCL1 and SiGGH are highly expressed during early panicle development, playing roles in folate synthesis and degradation ([50]Hou et al., 2022). Furthermore, genes linked to starch biosynthesis, cell wall invertase, hormone signal transduction, and polyamine metabolism pathways are believed to significantly influence grain-filling quality in foxtail millet ([51]Wang et al., 2020). Millets can be processed into diverse forms, including porridge, rusk, bars, cookies, and papad ([52]Amadou, Gounga, & Li, 2013). However, their products often experience low market sales due to a suboptimal tasting experience. Thus, it is essential to preserve their high nutritional value while simultaneously enhancing their taste quality. Previous studies have demonstrated that kernel structure, along with the composition of starch, protein, endosperm cell wall, fat, and amino acids, significantly influences taste quality ([53]Hou et al., 2022, [54]Zhang et al., 2023). Additionally, secondary metabolites, such as flavonoids and carotenoids, serve as key differentiators between high- and low-quality foxtail millet cultivars, significantly influencing both nutritional and tasting qualities ([55]Zhang et al., 2021). However, there is still limited information regarding grain quality-related metabolites and their formation mechanisms during grain-filling development, particularly those influencing taste quality, which remain unidentified and require further exploration. To address this, the study compared Jingu 21 (JG21), a high-quality foxtail millet cultivar from Shanxi Province valued for its nutritional and taste attributes, with Niumaobai (NMB), a local cultivar with relatively lower taste quality. Using widely targeted metabolomics and transcriptomics, JG21 and NMB were analyzed at two key developmental stages: the early (S2) and late (S4) grain-filling stages, aiming to identify differential metabolites, key metabolic pathways, and regulatory genes linked to quality-related metabolites. This research establishes a foundation for further investigation into the quality-related functional components of foxtail millet and supports the breeding and development of high-quality cultivars. 2. Materials and methods 2.1. Plant materials Previous studies have shown that Jingu 21 (JG21), an elite yellow-kernel cultivar from Shanxi Province, outperforms Niumaobai (NMB), a white-kernel landrace from the same region, in comprehensive cooking quality. While both cultivars share similar fatty acid profiles, JG21 is distinguished by its unique pigmentation and superior cooking aroma. In contrast, NMB exhibits a rougher and more astringent taste ([56]Zhang, Guo, Zhang, Han, & Li, 2021). Given these differences, these two representative cultivars of foxtail millet were chosen for preliminary research. A randomized complete block design with three replications was conducted at the experimental base of Shanxi Agricultural University (37°12′N, 112°28′E) during mid-May 2022. The environmental parameters for foxtail millet cultivation were as follows: soil pH of 8.36, daily average temperature of 21.24 °C, relative humidity of 58.07 %, accumulated effective temperature of 1722.22 °C, precipitation of 396.56 mm, and sunshine duration of 851.21 h. Samples were collected from the middle section of panicles from each foxtail millet cultivar at two critical grain-filling stages (S2 and S4), which were defined by their distinct physiological and metabolic profiles during grain development (Supplimentary Fig. S1): early grain-filling stage (S2, 35 days after heading, characterized by embryo formation and liquid-state endosperm). Late grain-filling stage (S4, 55 days after heading, marked by solidified embryos and fully mature endosperm). For each stage, spikelets from the same position of 5 individual panicles were pooled as test samples and stored at −80 °C for metabolomic and transcriptomic analyses. 2.2. Sample preparation The freeze-dried grains were milled into powder using a grinder (MM 400, Retsch, Germany) with zirconia beads for 1.5 min at 30 Hz. The 100 mg sample was weighed and dissolved in 1.0 mL of extraction solution (70 % methanol aqueous solution containing 0.1 mg/L lidocaine as an internal standard) and stored overnight at 4 °C. The dissolved samples were then centrifuged at 10000 rpm for 10 min, and the supernatant was collected and filtered using a microporous filter membrane (0.22 μm pore size; ANPEL, Shanghai, China) for LC-MS/MS analysis. 2.3. LC-MS conditions Ultra-high-performance liquid chromatography (UPLC; Shim-pack UFLC SHIMADZU CBM30A, Japan) coupled with tandem mass spectrometry (MS/MS; Applied Biosystems 6500 QTRAP) was used for analysis. HPLC conditions: mobile phase composition A: ultrapure water +0.04 % acetic acid; B: acetonitrile +0.04 % acetic acid. Gradient elution program: 95:5 V/V at 0 min, 5:95 V/V at 11.0 min, 5:95 V/V at 12.0 min, 95:5 V/V at 12.1 min, 95:5 v/v at 15.0 min. Flow rate, 0.4 mL/min. The column temperature was 40 °C and the injection volume was 2 μL. Mass spectrometry conditions: The electrospray ion source (ESI) temperature was set to 500 °C, with a mass spectrometry voltage of 5500 V. The curtain gas (CUR) pressure was set to 25 psi, and the collision-activated ionization (CAD) parameter was set to high. In the triple quadrupole mass spectrometer, each ion pair was scanned and detected using the optimized declustering potential (DP) and collision energy (CE). Metabolite identification and quantification: based on the MWDB (MetWare Database) and public databases of metabolite information, substances were identified using secondary spectral data. Metabolite quantification was performed using the Multiple Reaction Monitoring (MRM) mode of the triple quadrupole mass spectrometer. Metabolite detection was performed by Wuhan Metwell Biotechnology Co., Ltd. ([57]http://www.metware.cn/). 2.4. cDNA library construction and sequencing The millet samples were processed for cDNA library construction and sequencing by Novogene Technology Co., LTD ([58]https://www.novogene.com). This involved total RNA extraction, cDNA library preparation using the NEBNext®UltraTM RNA Library Prep Kit, and high-throughput sequencing on the HiSeq XTen platform. 2.5. Statistical analysis Differential metabolites were screened employing fold change (FC) and OPLS-DA VIP values with combined criteria: (1) primary screening was conducted using FC thresholds (FC ≥2 or ≤ 0.5), reflecting ≥2-fold concentration differences between control and experimental groups; (2) For experiments with biological replicates, metabolites must additionally meet VIP ≥ 1 in the OPLS-DA model, as this threshold serves as a validated indicator of statistically significant differences in inter-group discrimination. A Venn diagram was used to visualize the relationships among differential metabolites across groups, while a volcano plot and cluster heatmap illustrated their expression levels and patterns. GO annotation, KEGG pathway annotation, and enrichment analysis were employed to screen and analyze differential metabolic pathways. The DESeq2 package ([59]http://www.bioconductor.org/packages/release/bioc/html/DESeq2.html ) was used to identify differentially expressed genes (DEGs). Differential metabolites and genes were normalized to log2 fold change using the R core program (Pearson correlation coefficient > 0.8) and mapped to common KEGG pathways for integrated pathway analysis. 2.6. qRT-PCR Differentially expressed genes (DEGs) were selected to validate the transcriptome data. Gene-specific primers were designed using Primer-BLAST (NCBI), and β-actin from foxtail millet served as an internal reference for normalizing gene expression levels (Supplimentary Table 1). 3. Results 3.1. Metabolite profiling and accumulation during grain-filling stages in foxtail millet Using widely targeted metabolite technology, we identified a total of 552 metabolites in the grains of foxtail millet cultivar JG21 during the grain-filling stages, including NMB. These metabolites were categorized into 32 classes, encompassing 4 primary metabolism categories (210 metabolites) and 17 secondary metabolism categories (342 metabolites). Notably, the distribution ratios of these metabolites varied significantly between JG21 and NMB ([60]Fig. 1A, B), and they were further predicted to represent the principal components in foxtail millet grains. The study revealed that flavonoids and their derivatives were the most abundant secondary metabolites in the foxtail millet, included eight classes of substances: flavonols, flavones, flavone carbosides, flavone-lignin, flavanones, isoflavones, catechins, and anthocyanins. Furthermore, a comparative analysis between JG21 and NMB identified significant differences in a total of 144 metabolites across 27 categories. Of these, 79 metabolites were significantly abundant in JG21 than in NMB, whereas 65 metabolites were markedly higher in NMB than in JG21. Notably, the differential metabolites were predominantly enriched in flavonoids and their derivatives, hydroxycinnamoyl derivatives, and lipids and their derivatives ([61]Fig. 1C). These variations in metabolite profiles are likely to play a pivotal role in shaping the quality differences between the two cultivars. Fig. 1. [62]Fig. 1 [63]Open in a new tab Metabolites identified in grains of foxtail millet. A. Metabolic profiles and relative abundance distribution in grains of JG 21. B. Metabolic profiles and relative abundance distribution in grains of NMB. C. Comparative analysis of differentially accumulated metabolites between the two cultivars. We conducted a cluster analysis of the metabolites present in the grains of JG21 and NMB at the early S2 and late S4 grain-filling stages. For each sample type, all biological replicates clustered together, while clear differences were observed between sample types, reflecting the distinct characteristics of the cultivars and developmental stages. At the late S4 grain-filling stage, JG21 and NMB were distinctly separated based on their metabolite profiles ([64]Fig. 2A). When comparing the grain-filling stages S2 and S4 of JG21 and NMB, 186 and 180 metabolites showed significant differences, respectively. In the early grain-filling stage S2, 153 differential metabolites were identified, while 144 differential metabolites were identified in the late grain-filling stage S4 for the two cultivars. Among these, 108 metabolites exhibited similar changes in both cultivars ([65]Fig. 2B). Based on the previously identified metabolite pathways, 78 metabolites specifically altered in JG21 were associated with 76 metabolic pathways, whereas 72 metabolites uniquely changed in NMB were linked to 52 pathways during the grain-filling stages, as detailed in Supplementary Tables 2 and 3. Therefore, differences between the two cultivars were also evident at the early stage, likely contributing to or influencing the observed differences in mature grains. Fig. 2. [66]Fig. 2 [67]Open in a new tab Comparative metabolic profiling of foxtail millet cultivar. A. Heatmap of differential metabolites between two grain-filling stages (S2 vs. S4) in JG21 and NMB. The heatmap color scale (ranging from −3 to 3) represents Z-score-normalized metabolite abundance across triplicate biological samples of the JG21 and NMB cultivars. Positive (red) and negative (green) values indicate higher or lower metabolite levels relative to the mean, respectively. B.Venn diagram of differential metabolites from grain filling stages S2-S4 of JG21 and NMB. (For interpretation of the references to color in this figure legend, the