Integrated metabolomic and transcriptomic analysis reveal the flavor formation mechanism during oat grain development

Abstract

Oats are globally significant cereals that are appreciated for their nutritional composition and flavor profiles. However, the dynamic changes in volatile organic compounds (VOCs) and their associated biosynthetic genes during grain development remain poorly characterized. We hypothesized that the variation in oat flavor profiles was influenced by specific VOCs, which were synthesized via dedicated biosynthetic pathways and associated genes. To test this, GC–MS-based VOC profiling was performed during oat grain development in cultivar Neiyou 6, and the resulting semi-quantitative metabolomic data were integrated with a published transcriptomic dataset to identify candidate genes and pathways associated with aroma-related VOC accumulation. A total of 168 VOCs were identified, with terpenoids and esters as the dominant classes, accumulating predominantly during the Filling Stage. Based on database-derived odor descriptors, the identified VOCs were putatively associated with green, fruity, and sweet notes throughout grain development. KEGG pathway enrichment analysis identified monoterpenoid biosynthesis as a candidate pathway associated with aroma-related VOC accumulation; α-pinene oxide was identified as a representative differentially accumulated monoterpenoid. Integrated analysis of 113 differentially accumulated metabolites and 40,703 differentially expressed genes pinpointed four candidate cytochrome P450 genes (CYP1–4) that may relate to α-pinene oxide accumulation. WGCNA and correlation network analysis further indicated CYP2 and CYP3 as candidate genes associated with α-pinene oxide accumulation. These findings provide a descriptive and correlative foundation for understanding VOC dynamics during oat grain development and identify candidate targets for future functional studies aimed at improving oat flavor quality.

Highlights

• •168 VOCs were identified; terpenoids and esters dominated oat kernel development.
• •Most VOCs accumulated significantly at grain-filling stage; acetophenone at maturity.
• •113 differential VOCs were identified; green, fruity, and sweet odors predominated.
• •Monoterpene biosynthesis was a key pathway for flavor formation in oat grain.
• •CYP2/3 may associate with α-pinene oxide accumulation in oat grain.

1. Introduction

Oats are globally important cereal crops valued for their abundant nutritional composition, including β-glucan, antioxidants, minerals, and high-quality lipids (Campbell et al., 2021; Rasane et al., 2013), with industrial applications spanning oat rice, oatmeal, and oat milk (Alsado et al., 2023). Beyond nutrition, flavor critically determines consumer acceptance and product quality (Laaksonen et al., 2020). Volatile organic compounds (VOCs) play decisive roles in end-product quality by critically determining flavor formation and odor characteristics during both grain development and subsequent food processing (Xiang et al., 2025). Given the complexity and diversity of these VOCs, further research is needed to explore their trends during their crucial biosynthesis process, to better understand the formation of oat's distinctive flavor.

The biosynthesis of VOCs in cereal grains involves multiple interconnected metabolic pathways, including the lipoxygenase (LOX) pathway for fatty acid-derived volatiles, the mevalonic acid (MVA) and methylerythritol phosphate (MEP) pathways for terpenoids (Oni et al., 2025; Zhang, Pan, et al., 2024). Current research categorizes VOCs into three major groups: terpenoids, phenyl-propanoid/benzenoid derivatives, and fatty acid derivatives (Jesús et al., 2020). During rice grain development, total VOC content increased progressively as grains matured, with terpenoids representing a major constituent of the volatile profile (Zhang, Wu, et al., 2024). The monoterpenoid eucalyptol was identified as a key maturity indicator in the LL655 and SH-1 cultivars during mung bean seed development (Xiang et al., 2025). Alkanes and esters have been identified as the predominant classes of volatile metabolites during the grain development of waxy maize (Zhao et al., 2025, Zhao et al., 2025). In oats, terpenoids and esters have been identified as the dominant odor-active compound classes, contributing rose, honey, fatty, and cereal-nutty sensory attributes (McGorrin, 2019; Wang et al., 2023). While previous studies have primarily focused on the VOC components of cereal grains development, there has been a lack of research on the key aroma formation process, particularly the changes in key VOCs during grain development, particularly terpenoids. Addressing these gaps, the present study provides a correlative characterization of VOC dynamics during oat grain maturation, with a specific focus on terpenoid biosynthesis pathways.

Integrated metabolome and transcriptome analysis provides a systematic approach to study metabolites and their regulatory networks in crops, including Zanthoxylum (Fei et al., 2021), mango (Lei et al., 2022), citrus (Zhang et al., 2017). In oats, this approach has been successfully applied to dissect the molecular basis of nutritional quality traits. For example, temporal transcriptomic and metabolomic profiling across multiple oat varieties and developmental stages revealed that the dynamic expression of lipid and avenanthramide biosynthesis genes was consistent with the accumulation of corresponding metabolites during seed development (Hu et al., 2020). Comparative transcriptomic profiling of oat varieties with contrasting β-glucan contents further revealed stage-specific expression of CSL family genes that closely correlated with the dynamic accumulation of β-glucan (Qi et al., 2024). Moreover, transcriptome sequencing of developing oat grains identified 398 DEGs associated with plant hormone signal transduction and 107 DEGs associated with starch and sucrose metabolism, indicating these pathways as core regulators of grain development and quality formation (Wang et al., 2025). However, aroma-related VOCs, particularly terpenoids, have been rarely systematically investigated using integrated multi-omics approaches in developing oat grains.

Recently, the dynamic changes in VOCs during oat grain development and the candidate genes underlying their accumulation remain unclear. To address this knowledge gap, we hypothesized that dynamic changes of specific VOCs may influence the flavor formation characteristics during oat grain development, particularly for terpenoids, and conducted this study to characterize the underlying associations. Specifically, this study integrates GC–MS-based VOC profiling across four grain developmental stages with existing RNA-Seq data (Wang & Han, 2025), aiming to identify specific terpenoid-related biosynthetic pathways and candidate genes associated with terpenoid accumulation dynamics. Furthermore, the investigation proposes a potential association network in the biosynthesis of important aroma compounds in oats, providing new insights for future functional studies on oat flavor quality.

2. Materials and methods

2.1. Plant materials

Oat grains of cultivar Neiyou 6 ‘NY6’, selected for its high nutritional value and stable agronomic performance (Wang, 2020; Zhu, 2023), were obtained from the Inner Mongolia South Breeding Base located in Ledong County, Hainan Province, China (18°31′N, 108°55′E). Normal field management was conducted during the growing period, and pollination was recorded as 0 days post-pollination (dpp). The grains were collected at four developmental stages in February 2023: the Filling Stage (T1, 10–14 dpp), the Milk Stage (T2, 18–22 dpp), the Dough Stage (T3, 26–30 dpp), and the Maturity Stage (T4, 34–38 dpp) (Wang and Han, 2025). At each stage, lemmae were dissected and caryopses were collected using clean forceps. Three biological replicates (1.5 g per replicate) were prepared for each stage, with each replicate consisting of at least 150 caryopses from 15 independently sampled plants. The same biological material was used for both transcriptomic and metabolomic analyses to ensure direct comparability. All samples were immediately frozen in liquid nitrogen and stored at -80 °C until analysis.

2.2. Volatile extraction

Sample extraction was performed by Wuhan Metville Biotechnology Co., Ltd. (Wuhan, China). Oat grain samples were weighed and ground to a powder under liquid nitrogen and 500 mg aliquots were transferred into 20 mL headspace vials (Agilent Technologies, Palo Alto, CA, USA) containing saturated NaCl solution to inhibit enzymatic reactions. Vials were sealed with crimp-top caps fitted with TFE‑silicone septa (Agilent). Pooled quality control (QC) samples were prepared by combining equal aliquots from all experimental samples, with one QC sample was inserted after every 10 analytical samples to monitor instrumental stability and analytical reproducibility. Volatile extraction was performed by headspace solid-phase microextraction (HS-SPME); the DVB/CAR/PDMS fiber (120 μm; Agilent) was pre-conditioned at 250 °C for 5 min prior to use. Sample vials were equilibrated, after which the fiber was exposed to the headspace at 60 °C for 15 min for VOC absorption.

2.3. GC–MS analysis

Chromatographic separation was performed on an Agilent 8890 GC system coupled with an Agilent 7000D triple quadrupole mass spectrometer (Agilent Technologies, Palo Alto, CA, USA), using a DB-5MS Capillary Column (30 m × 0.25 mm × 0.25 μm; Agilent J&W Scientific, Folsom, CA, USA). Helium was used as the carrier gas at a constant flow rate of 1.2 mL/min. The initial oven temperature was held at 40 °C for 3.5 min, ramped to 100 °C at 10 °C/min, then to 180 °C at 7 °C/min, and finally to 280 °C at 25 °C/min, for 5 min (Wang et al., 2022).

Mass spectrometry detection was performed using a quadrupole analyzer with electron impact (EI) ionization at 70 eV. The operating temperatures for the quadrupole mass detector, ion source, and transfer line were set to 150 °C, 230 °C, and 280 °C, respectively. Selected ion monitoring (SIM) mode was employed for the identification and quantification of analytes according to the National Standard of China (GB 23200.8–2016).

2.4. Metabolite identification and quantification

Compound identification was performed using a multi-criteria approach. Mass spectra were matched against the NIST 2020 Mass Spectral Library (match score ≥ 70), and retention indices (RI) were calculated using a C7–C40 n-alkane series (Sigma-Aldrich, St. Louis, MO, USA) under identical chromatographic conditions (RI tolerance ± 30 units). Based on multi-species data, literature references, authentic standards, and retention indices, an independent volatile compound database was established containing confirmed retention times and characteristic ions (Yuan et al., 2021). In SIM mode, one quantitative ion and 2–3 qualitative ions were assigned per compound for time-segmented detection according to elution sequence. A compound was positively identified when its retention time matched the database reference, all selected ions were present after background subtraction, and ion abundance ratios were consistent with reference values within defined tolerances (relative abundance >50%, ± 10%; 20–50%, ± 15%; 10–20%, ± 20%; ≤ 10%, ± 50%).

The internal standard (3-hexanone-2,2,4,4-d4) semi-quantitative method was used to calculate the relative metabolite contents. The formula used is as follows (Liang et al., 2024):

$${\mathrm{X}}_{\mathrm{i}}=\frac{{\mathrm{V}}_{\mathrm{s}}\times {\mathrm{C}}_{\mathrm{s}}}{\mathrm{M}}\times \frac{{\mathrm{I}}_{\mathrm{i}}}{{\mathrm{I}}_{\mathrm{s}}}\times {10}^{-3}$$

X_i represents the relative content (μg/g) of compound i in the test sample; V_s represents the volume (μL) of the internal standard added; C_s represents the concentration (μg/mL) of the internal standard; M represents the amount (g) of the test sample; I_s represents the peak area of the internal standard; I_i represents the peak area of compound i in the test sample.

2.5. Transcriptomic analysis

Transcriptomic data were obtained from our previously published RNA-Seq dataset generated from the same samples (Wang and Han, 2025). Total RNA was extracted using the RNAprep Pure Plant Kit (DP441; Tiangen Biotech, Beijing, China). RNA purity and integrity were assessed via 1% agarose gel electrophoresis and a NanoPhotometer spectrophotometer (Implen Inc., CA, USA). Sequencing libraries were constructed using the NEBNext Ultra RNA Library Prep Kit for Illumina (New England Biolabs, Ipswich, MA, USA) and sequenced on an Illumina platform (150 bp paired-end reads). Raw reads were processed using fastp to remove adapter-contaminated reads; paired-end reads were discarded if they contained > 10% undefined nucleotides (Ns) or if > 50% of bases had a quality score < 20. Clean reads were aligned to the oat OT3098 reference genome using HISAT2, and gene expression levels were quantified as FPKM using featureCounts. Differentially expressed genes (DEGs) were identified using DESeq2 with thresholds of |log₂FoldChange| ≥ 1 and FDR < 0.05.

2.6. Multivariate statistical analysis

All measurements were performed in triplicate, and raw data were log₂-transformed and mean-centered prior to multivariate analysis. Principal component analysis (PCA) was analyzed using the R package (v4.1.0) with unit variance scaling for exploratory assessment of overall sample variability and inter-group separation. Orthogonal partial least squares discriminant analysis (OPLS-DA) was subsequently applied for supervised classification to discriminate volatile profiles across developmental stages (T1–T4), with model performance assessed by R²X, R²Y, and Q² statistics, with models considered robust when R²X > 0.5 and Q² > 0.5. Overfitting was excluded by 200-permutation testing. For multi-group comparisons, one-way ANOVA was first performed to assess overall differences among the four stages. Subsequently, Student's t-test was applied for pairwise comparisons of four stages (Liu et al., 2020), and false discovery rate (FDR) correction was applied to control for false positives, with adjusted P < 0.05 considered statistically significant. Hierarchical cluster analysis (HCA) and K-means clustering were performed to classify metabolites into groups with distinct temporal accumulation patterns, with results presented as heatmaps with dendrograms using the R package ComplexHeatmap. Differentially accumulated metabolites (DAMs) were identified using combined criteria: VIP > 1 in the OPLS-DA model and |log₂FC| ≥ 1 from a volcano plot of all detected VOCs (Zhang et al., 2024). The predicted sensory characteristics of DAMs were annotated using The Good Scents Company database (https://www.thegoodscentscompany.com/index.html) (Bhayani et al., 2023).

The identified metabolites were annotated using the KEGG Compound database (accessed 2022-07-20; Kanehisa & Goto, 2000) and mapped to the KEGG Pathway database (http://www.kegg.jp/kegg/pathway.html). Pathway enrichment significance was evaluated using the hypergeometric test, with all detected metabolites carrying valid KEGG annotations used as the background reference set (Xia and Wishart, 2016). Pathways with P ≤ 0.05 were considered significantly enriched. Weighted gene co-expression network analysis (WGCNA) was performed using the R package WGCNA based on Metware Cloud (cloud.metware.cn), aiming to select candidate genes associated with accumulation of aroma compounds. The soft power (β = 18) was determined based on the scale-free topology criterion, which required a scale-free topology fitting index R² > 0.9 (Langfelder and Horvath, 2008). The height for merging similar modules was 0.25, and the minimum module size was 50. A co-expression network was constructed using Cytoscape (version 3.8.2) based on the module with the highest correlation coefficient between module eigengenes and aroma compounds, with Pearson's correlation coefficient values > 0.6 and P < 0.05 selected for visualization (Cytoscape Consortium, San Diego, CA, USA).

3. Results

3.1. VOCs in oat grains

To investigate the synthesis and accumulation of aroma-related compounds, GC–MS profiling of VOCs was conducted during four developmental stages. The overlay analysis of the QC–TIC diagram (Fig. S1) and the Pearson correlation coefficient (Fig. S2) showed that the data recorded in this study had a good repeatability and reliability. In total, 168 VOCs were identified, including terpenoids, esters, heterocyclic compounds (Table S1), which terpenoids and esters each accounted for 19.05% of the total, followed by heterocyclic compounds at 18.45%, making them the primary components of VOCs in oat grains (Fig. 1A). These VOCs also appear in other crops (Liu et al., 2024).

Supplementary Fig. S1.

Overlap of the total ion flow (TIC) of mixed sample mass spectrometry analysis for quality control.

Supplementary Fig. S2.

Heat map of (Pearson's correlation coefficient) value among different samples including replicates.

Fig. 1.

VOCs at the different developmental stages of oat grain. (A) Ring chart of the proportion of 168 VOCs. (B) PCA analysis.(C) HCA of metabolite levels in different developmental stages of oat grains. T1: the Filling Stage, T2: the Milk Stage, T3: the Dough Stage, T4: the Maturity Stage.

The peak areas of the volatile compounds in the samples were subjected to PCA to visualize the overall distribution patterns of VOCs across the four developmental stages. PCA revealed clear separation among the four developmental stages, indicating distinct VOC composition profiles during grain development. Notably, samples from the same developmental stage clustered closely together, suggesting high consistency in VOC composition within each stage and demonstrating the reliability of the experimental replicates. PC1 and PC2 explaining 70.53% and 16.16% of the total variance, respectively (Fig. 1B). HCA was then conducted based on the identified VOCs, resulting in a heat map (Fig. 1C). The HCA results revealed distinct clustering patterns for the four grain samples of T1, T2, T3, and T4. Notably, the relative content of most VOCs significantly increased at T1 before decreasing during ripening. The total relative content of all detected VOC categories showed a decreasing trend across grain development, whereas terpenoids and esters showed the highest relative content at T1 among the predominant VOC categories (Table S2). However, a small amount of VOCs showed higher relative content at T4, such as acetophenone.

3.2. Differential VOCs at different developmental stages

The OPLS-DA model showed good separation between groups,with all R²X and Q² values exceeding 0.9. Model validity was assessed by 200-permutation testing, which demonstrated no signs of overfitting (Fig. S3). A total of 113 DAMs were identified based on this criterion (Table S3). The classification and relative content of detected DAMs indicated that the relative total content of aroma compounds exhibited a progressive decline from T1 to T4, mainly heterocyclic compounds, aldehyde, and terpenoids (Fig. 2A, Fig. S4). The proportion of terpenoids increased progressively during grain development, whereas aldehydes showed a decreasing trend from T1 to T4. These results indicate that oat grain maturation is accompanied by dynamic shifts in VOCs composition, with the identified DAMs potentially contributing to shifts in aroma-related VOC composition throughout grain development.

Supplementary Fig. S3.

Permutation test plots (200 permutations) of OPLS-DA models across pairwise developmental stage comparisons.

Fig. 2.

Characterization of VOCs at different developmental stages. (A) Classification and proportions of VOCs in different stages. (B) K-Means analysis grouped the DAMs into ten clusters at different developmental stages. (C) Flavor wheel of DAMs. Inner circle, number of common differential metabolites. Middle circle, related odor description. and Outer circle, characteristic component KEGG annotation. T1: the Filling Stage, T2: the Milk Stage, T3: the Dough Stage, T4: the Maturity Stage.

Supplementary Fig. S4.

Volcano plots showing the differential metabolites.

K-Means clustering of 113 DAMs resulted in 10 distinct groups (Fig. 2B, Table S4). Subclasses 2, 5, 6, 8, 9, and 10 exhibited elevated accumulation at T1 relative to later stages, indicating that representative terpenoids such as linalyl acetate, 1-methyl-4-(1-methylethenyl)-1,2-cyclohexanediol, and cis-dihydrocarvone showed elevated accumulation at T1 relative to later developmental stages. This is consistent with previous studies (He et al., 2024; Wang et al., 2024).

3.3. Odor activity value analysis

To provide a preliminary characterization of the potential sensory attributes associated with identified DAMs, predicted sensory annotations were retrieved from The Good Scents Company database (Bhayani et al., 2023) and visualized as a flavor wheel representing the top 10 predicted sensory note categories (Fig. 2C, Table S3). The DAMs were distributed across 10 predicted odor categories, predominantly green (26), fruity (18), and sweet (16), consistent with previous reports on cereal volatile profiles (Liang et al., 2024). Representative contributors to the green note D197 (α-pinene oxide) and included XMW1401 (n-valeric acid cis-3-hexenyl ester); the fruity note was associated with D202 (3-hexen-1-ol, propanoate, (Z)-) and KMW0288 (heptanoic acid, ethyl ester); and the sweet category featured terpenoids such as KMW0432 (2,6-octadienal, 3,7-dimethyl-, (Z)-) and D192 (fenchol), consistent with the established role of terpenoids and esters as key aroma VOCs in cereals (Wang et al., 2023).

3.4. Pathway enrichment analysis of DAMs

KEGG pathway enrichment analysis of DAMs identified five significantly enriched pathways across pairwise comparisons: metabolic pathways, biosynthesis of secondary metabolites, monoterpenoid biosynthesis, limonene and pinene degradation, and tropane, piperidine and pyridine alkaloid biosynthesis (Fig. 3). Among these, monoterpenoid biosynthesis was consistently enriched in T4 vs. T1 and T2 vs. T1 comparisons, suggesting that terpenoid-related metabolic changes were pronounced during early grain development. No enriched pathways were detected in the T3 vs. T2 and T3 vs. T1 comparisons, likely due to metabolic stabilization during this period, resulting in fewer DAMs and insufficient pathway annotation coverage. Based on these results, monoterpenoid biosynthesis is selected as the focus for subsequent integrated transcriptomic and metabolomic analysis.

Fig. 3.

KEGG pathway enrichment of the DAMs. (A) T4 vs. T3, (B) T4 vs. T2, (C) T4 vs. T1, (D) T2 vs. T1. T1: the Filling Stage, T2: the Milk Stage, T3: the Dough Stage, T4: the Maturity Stage. Each bubble in the plot represents a metabolic pathway whose abscissa and bubble size jointly indicate the magnitude of the impact factors of the pathway, a larger bubble indicates a higher impact factor. The bubble color represents the p-values of the enrichment analysis, with darker colors indicating a higher degree of enrichment.

3.5. Correlation analysis of transcriptomics and metabolomics

To explore the transcriptional basis of terpenoid accumulation during oat grain development, metabolomic data were integrated with published transcriptomic data (Wang and Han, 2025), enabling construction of a monoterpenoid biosynthesis pathway map (Fig. 4, Table S5). The MEP and MVA pathways are fundamental to terpenoid synthesis (Gui et al., 2022). In the MVA pathway, 37 DEGs were identified across 7 catalytic steps, with most genes showing peak expression at T4, suggesting upregulation of terpenoid precursor supply during late grain development. In the MEP pathway, 11 DEGs were identified across 7 catalytic steps; DXR1 (1-deoxy-D-xylulose-5-phosphate reductoisomerase), CMK1 (4-(cytidine 5-diphospho)-2-C-methyl-D-erythritol kinase), HDS1/2 ((E)-4-hydroxy-3-methylbut-2-enyl diphosphate synthase), HDR3 ((E)-4-hydroxy-3-methylbut-2-enyl diphosphate reductase) showed peak expression at T4, while MDS1/2 (2-C-methyl-D-erythritol 24-cyclodiphosphate synthase) and HDR1/2 showed peak expression at T1. These results suggest putatively that most terpenoid precursors accumulated early in oat grain development.

Fig. 4.

Pathway analysis of DAMs and DEGs involved in “ monoterpenoid biosynthesis”. MVA, the mevalonate pathway. MEP, the 2-C-methyl-d-erythritol 4-phosphate pathway. AACT, acetyl-CoA acetyltransferase. HMGS, 3-hydroxy-3-methylglutaryl-CoA synthase. HMGR, 3-hydroxy-3-methylglutaryl-CoA reductase. MVK, mevalonate kinase. PMK, phosphomevalonate kinase. MPDC, diphosphomevalonate decarboxylase. DXS, 1-deoxy-D-xylulose-5-phosphate Synthase. DXR, 1-deoxy-D-xylulose-5-phosphate reductoisomerase. MCT, 2-C-methyl-derythritol 4-phosphate cytidylyltransferase. CMK, 4-(cytidine 5-diphospho)-2-C-methyl-D-erythritol kinase. MDS, 2-C-methyl-D-erythritol 24-cyclodiphosphate synthase. HDS, (E)-4-hydroxy-3-methylbut-2-enyl diphosphate synthase. HDR, (E)-4-hydroxy-3-methylbut-2-enyl diphosphate reductase. IDI, isopentenyl-diphosphate isomerase. GPPS, geranyl diphosphate synthase. TPS, terpene synthase and CYP, cytochrome P450.

In the downstream monoterpenoid biosynthesis, geranyl diphosphate synthase (GPPS) condensed isopentenyl diphosphate and dimethylallyl diphosphate to form geranyl diphosphate, which was subsequently cyclized by terpene synthase (TPS) to produce (-)-α-pinene. Four candidate cytochrome P450 genes (CYP1–4) were identified as potentially catalyzing the oxidation of (-)-α-pinene to α-pinene oxide. Among these, CYP2/3 showed high expression at T1 but declined at later stages, while CYP1/4 exhibited markedly elevated expression at T4. These expression patterns suggest stage-specific transcriptional regulation of monoterpenoid biosynthesis. The contrasting expression patterns between CYP2/3 and CYP1/4 across developmental stages may be associated with the observed accumulation pattern of α-pinene oxide.

3.6. Correlation between candidate gene expression and α-pinene oxide accumulation

To identify candidate genes potentially associated with the accumulation of specific VOCs, WGCNA was performed on 40,703 DEGs (Table S6) identified across four developmental stages, with all differentially accumulated monoterpenoids used as traits. Using hierarchical clustering and dynamic tree cut, we generated 11 modules (Fig. 5A, Table S7). The results reveal that several modules show correlation with α-pinene oxide (D197) accumulation. Among them, the MEturquoise module showed the strongest positive correlation (r = 0.83, P = 8.3 × 10^-4), while the MEblue module exhibited the strongest negative correlation (r = -0.87, P = 2.3 × 10^-4). Furthermore, all candidate genes (CYP1–4) showed statistically significant correlations with both the MEturquoise and MEblue modules (P < 0.05).

Fig. 5.

Correlation analysis between gene modules and α-pinene oxide accumulation. (A) Heatmap representing the correlation between gene modules and the relative content of all monoterpenes. The color gradient represents the correlation values, with red indicating positive correlations and blue indicating negative correlations between gene expression modules and α-pinene oxide levels. (B) The correlation network for CYP1–4 and α-pinene oxide. The red line represents positive correlation with |r| ≥ 0.6 and P < 0.05. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

The MEturquoise and MEblue modules, exhibiting the strongest correlations with α-pinene oxide accumulation, were selected for further analysis. To explore the associations between candidate genes (CYP1–4) and α-pinene oxide accumulation, a correlation network analysis was performed (Fig. 5B, Table S8). Among the four candidate genes, CYP2 (AVESA.00001b.r3.2Ag0001534) (r = 0.62, P = 0.03) and CYP3 (AVESA.00001b.r3.2Cg0002249) (r = 0.86, P = 0.00034) showed strong positive correlations (|r| ≥ 0.6, P < 0.05) with α-pinene oxide accumulation (Langfelder and Horvath, 2008), whereas CYP1 and CYP4 did not meet the threshold and were therefore excluded from the network. These correlation-based associations suggest that CYP2 and CYP3 may be co-expressed with α-pinene oxide accumulation during oat grain development.

4. Discussion

Oats are a nutrient-rich cereal cultivated and consumed globally (Ren et al., 2022). The cultivar NY6 used in this study is an independently bred naked oat variety that has passed DUS testing and possesses the highest β-glucan content (7.05%) among 23 domestic naked oat varieties (Wang, 2020). Flavor is a fundamental quality trait of oats, yet the dynamic changes in VOCs during grain development and the candidate biosynthetic genes underlying their accumulation remain poorly characterized (Wang et al., 2023). This study provides a descriptive and correlative characterization of VOC accumulation dynamics across four developmental stages, integrating GC–MS-based metabolomics with published transcriptomic data to identify candidate genes associated with aroma-related VOC accumulation.

Consistent with previous reports (McGorrin, 2019; Wang et al., 2023), terpenoids and esters were the dominant VOC classes in oat grains. The present study extends beyond prior work by revealing that these compounds accumulated predominantly at T1 and declined progressively toward maturity, a pattern that warrants further investigation. Sensory annotations retrieved from The Good Scents Company database suggested that the identified DAMs were putatively associated with green, fruity, and sweet notes, broadly consistent with previous studies on oat aroma (Wang et al., 2023). These annotations are database-derived and represent predicted rather than experimentally validated sensory descriptors; experimental sensory validation remains a priority for future work. Notably, α-pinene oxide was putatively associated with a green note and identified as a representative differentially accumulated monoterpenoid, consistent with the role of volatile terpenoids in flavor formation in other crops (Xiuyin et al., 2023)

KEGG enrichment analysis identified monoterpenoid biosynthesis as a candidate pathway enriched in comparisons involving T1, suggesting that terpenoid-related metabolic changes may be pronounced during early grain development. This observation is consistent with the reported roles of monoterpenoid pathways in flavor transition during dill and papaya development (Wang et al., 2025; Zhang et al., 2024). Compared with previous studies focused primarily on processed oat products (Bai et al., 2022; Parker et al., 2000), the present study provides a systematic characterization of VOC dynamics during grain development, representing a previously unaddressed research gap. In the MEP pathway, the early-stage peak expression of MDS1/2 and HDR1/2 is consistent with findings in other species (Feng et al., 2024; Zhao et al., 2023), suggesting that expression patterns are consistent with potential precursor accumulation at T1, though functional validation in oats is required to confirm. Similarly, the coordinated expression patterns observed in the MVA pathway are consistent with synchronized precursor accumulation during early grain development, a pattern reported across multiple plant species (Chen et al., 2025; Wang et al., 2019).

Through integrated WGCNA and correlation network analysis, CYP2 (AVESA.00001b.r3.2Ag0001534) and CYP3 (AVESA.00001b.r3.2Cg0002249) were identified as candidate genes positively co-expressed with α-pinene oxide accumulation, representing the most probable targets for future functional validation. α-Pinene oxide is defined as an epoxide of α-pinene (ChEBI database), and plant cytochrome P450 enzymes are well-established catalysts for monoterpene oxidation; the initially produced monoterpenes are typically oxidized or conjugated by CYP enzymes during specialized metabolism (Amiripour et al., 2019; Howyzeh et al., 2018). Based on these observations, we presume that CYP450-mediated epoxidation of α-pinene may contribute to α-pinene oxide accumulation in oat grains, and that CYP2 and CYP3 may be associated with this process, though these remain correlative inferences pending functional validation. The scope of the current study is limited to a single cultivar and reliance on bioinformatic correlation analyses without functional validation. To gain a more comprehensive understanding of oat aroma accumulation, future investigations incorporating multi-cultivar comparisons, metabolomic flux analysis, and gene function verification will be essential to fully elucidate the mechanisms underlying oat flavor formation.

5. Conclusion

This research elucidates widely-targeted metabolomics and transcriptomics to systematically characterize VOC accumulation dynamics and their potential associations with candidate genes across four developmental stages of oat grains. Terpenoids and esters were identified as the dominant VOC classes, accumulating predominantly at the Filling Stage (T1) and declining progressively toward maturity. The putative association of identified DAMs with green, fruity, and sweet sensory notes highlighted the flavor relevance of stage-specific VOC dynamics, though experimental sensory validation remains necessary. The consistent enrichment of monoterpenoid biosynthesis in early developmental comparisons, combined with elevated upstream gene expression at T1, suggested that the T1 putatively represented a critical stage for terpenoid-driven aroma accumulation. Among the identified DAMs, α-pinene oxide emerged as a representative monoterpenoid of interest, and CYP2 (AVESA.00001b.r3.2Ag0001534) and CYP3 (AVESA.00001b.r3.2Cg0002249) were identified as candidate genes positively co-expressed with its accumulation through integrated WGCNA and correlation network analysis. These findings are correlative in nature and provide a foundation for future functional studies to elucidate the molecular mechanisms underlying VOC accumulation in oat grains.

The following are the supplementary data related to this article.

Supplementary Table S1

168 VOCs were identified by GC–MS.

Supplementary Table S2

Dynamic changes in the contents of various volatile compound classes and total VOCs during grain development.

Supplementary Table S3

The basic information of detected 113 DAMs.

Supplementary Table S4

113 DAMs based on K-means analysis.

Supplementary Table S5

DEGs involved in the monoterpenoid biosynthesis.

Supplementary Table S6

The total number of DEGs detected in the four RNA-seq samples.

Supplementary Table S7

Correlation between gene expression and module eigengenes.

Supplementary Table S8

Correlation of α-pinene oxide with candidate genes.

Supplementary material 1

Field cultivation and VOCs analysis methods of Oat.

CRediT authorship contribution statement

Yao Qin: Writing – original draft, Software, Methodology, Investigation, Formal analysis, Data curation. Ting Wang: Methodology, Conceptualization. Huiyan Liu: Formal analysis. Yongda Guo: Software. Xiaohong Yang: Investigation. Yan Yang: Data curation. Dorjpagma Shagdarsuren: Validation. Bing Han: Writing – review & editing, Supervision, Resources, Project administration, Conceptualization.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability

Data will be made available on request. The RNA-Seq data have been submitted to the Genome Sequence Archive at the National Genomics Data Center (GSA: CRA019946), Link address: https://ngdc.cncb.ac.cn/gsa/search?searchTerm=CRA019946.