Enhancing glucosinolate profiles in broccoli (Brassica oleracea) sprouts via electron beam irradiation: A functional food approach Authors: Juan Wang, Yaqin Wang, Zhaoyu Li, Xuezhi Zhao, Xirui Zhao, Hongju He, Guangmin Liu Published: 2026 ABSTRACT Developing non-thermal technologies to enhance health-promoting compounds is important for producing high-value functional foods. This study compared the effects of electron beam irradiation (EBI) applied to dry broccoli seeds before germination (FZZZ) with those of direct irradiation of sprouts (FZYM) on glucosinolate (GSL) enhancement. Low-dose FZZZ (2 kGy) increased total GSL content to 6556.85 μg/g, compared with 3174.58 μg/g in the control, whereas low-dose FZYM decreased total GSL content. Sprouts grown from irradiated seeds also showed increased vitamin C, soluble sugars, total free amino acids, and volatile sulfur-containing compounds. Metabolomic and amino acid analyses indicated that FZZZ increased the availability of sulfur amino acid precursors and remodeled methionine- and phenylpropanoid-related metabolism, whereas FZYM mainly induced stress-associated changes in lipid and nucleotide metabolism. qRT-PCR revealed selective regulation of genes involved in downstream GSL modification and hydrolysis rather than a coordinated activation of the entire GSL biosynthetic pathway. Overall, seed-stage EBI at 2 kGy is a promising strategy for enhancing GSL content in broccoli sprouts while maintaining growth and nutritional quality. Introduction Sprouts are young seedlings harvested before they develop true leaves (Liu et al., 2025). They are considered health-promoting foods because they are rich in nutrients and bioactive compounds that exhibit antioxidant activity and may help protect against cancer and cardiovascular diseases. Sprouting increases the levels of oligosaccharides and monosaccharides, free fatty acids, oligopeptides, amino acids, antioxidants, and vitamins (Liu et al., 2025). Among edible sprouts, broccoli (Brassica oleracea) sprouts are particularly valued for their potential anticancer and anti-inflammatory properties because of their high contents of antioxidants, vitamins, minerals, glucosinolates (GSLs), and GSL hydrolysis products (Matusheski et al., 2004). For instance, sprouting increases the GSL precursor of sulforaphane (SF) by up to 10 times compared with seeds and mature plants (Fahey et al., 1997). When broccoli cells are disrupted, GSLs are released from vacuoles and hydrolyzed by myrosinase; this process forms SF, a highly bioactive GSL hydrolysis product (Matusheski et al., 2004). Sulforaphane in broccoli sprouts has anti-cancer and anti-inflammatory properties. It enhances antioxidant enzyme activity and helps lower blood pressure, oxidative stress, and kidney inflammation (Fahey et al., 1997; Zhang et al., 2006). Several methods are currently used to regulate GSL content and increase the concentrations of bioactive compounds in broccoli sprouts. Selenium biofortification (using selenate or selenite) in broccoli sprouts significantly increases the content of Se-methylselenocysteine and preserves the GSL profile and content (Avila et al., 2013; Avila et al., 2014; Arscott and Goldman, 2019). UV irradiation (especially UV-B) promotes the accumulation of GSLs in broccoli sprouts. UV-B treatment significantly increases the levels of GSLs, particularly glucoraphanin (RAA), and enhances SF biosynthesis. This finding suggests that UV irradiation can stimulate secondary metabolic pathways, thereby leading to higher GSL production (Martínez-Zamora et al., 2021). Furthermore, cold plasma treatment (2 min, 30 kV) enhances the accumulation of GSLs, SF, and phenolic compounds in broccoli sprouts and is associated with 7460 differentially expressed genes and 6739 differential metabolites, indicating widespread transcriptional and metabolic responses (Lyu et al., 2023). The application of elevated CO2 (eCO2, 620 ppm) to three cultivars of broccoli sprouts (Southern Star, Prominence, and Monotop) for 9 days improved their nutritive and health-promoting values (Almuhayawi et al., 2020). Electron beam irradiation (EBI) is an emerging non-thermal technology characterized by high dose rates, short treatment times, and the absence of induced radioactivity in the treated food. It has been applied in food processing for microbial decontamination, enzyme inactivation, and shelf-life extension (Li et al., 2025). From a regulatory perspective, irradiation authorizations are commodity- and purpose-specific. In the United States, doses of up to 4.0 kGy are permitted for pathogen control and shelf-life extension in fresh iceberg lettuce and spinach, whereas doses of up to 8.0 kGy are permitted for microbial pathogen control in seeds intended for sprouting (U.S. Food and Drug Administration, 2012; U.S. Food and Drug Administration, 2026). These provisions provide relevant regulatory context, but do not automatically establish regulatory compliance for the use of EBI to enhance phytochemical accumulation in broccoli sprouts. Beyond its effects on microbial safety, EBI can induce physicochemical changes and metabolic reprogramming in plant-derived matrices, potentially influencing antioxidant systems and secondary metabolite profiles in a dose-dependent manner. Such effects may also result in dose- and matrix-dependent radiolytic products and changes in nutritional components, which should be considered when assessing the safety and suitability of EBI-treated sprouts. For example, electron beam and gamma irradiation can inactivate pathogens in commercial seed sprouts (Waje et al., 2009). In addition, EBI alters the functional properties and antioxidant activity of plant-derived substrates, indicating its broader capacity to modulate biofunctional traits (Wang et al., 2019). Recent work suggests that EBI intensity can reshape the profile of health-related constituents in broccoli sprout juice by modifying its sensory attributes and bioactive compounds (Yan et al., 2025). These advances underscore the dual function of EBI in green food processing: it can serve not only as a safety intervention but also as a tool for targeted matrix modification and nutrient enhancement. Despite these advances, research on the use of EBI to regulate broccoli sprout growth and enrich GSLs, as well as on the underlying metabolic pathways, remains limited. In particular, there is a critical knowledge gap regarding whether EBI is more effective when applied as a pre-germination seed priming technology or as a post-harvest direct processing treatment for sprouts. Therefore, this study aims to: (1) compare the effects of EBI applied to dry broccoli seeds before germination and directly to 5-day-old sprouts on sprout growth and GSL accumulation; (2) determine the effects of seed irradiation on selected nutritional components, including vitamin C, soluble sugars, and free amino acids, as well as sulfur-containing volatile compounds; (3) characterize stage-dependent metabolic reprogramming using widely targeted metabolomics and KEGG pathway enrichment analysis; and (4) examine the expression of genes involved in GSL and phenylpropanoid metabolism using quantitative real-time PCR. Collectively, these analyses were conducted to clarify the mechanisms underlying irradiation-induced changes in GSL accumulation and to evaluate the potential of seed-stage EBI for producing nutritionally enhanced broccoli sprouts. Materials and methods Chemicals and reagents High-purity GSL standards (>98%) were purchased from PhytoLab GmbH & Co. KG (Vestenbergsgreuth, Middle Franconia, Germany): sinigrin, gluconapin, glucobrassicanapin, progoitrin, glucoerucin (ERU), glucoraphanin, RAA, glucobrassicin (GBC), glucotropaeolin (TRO), and gluconasturtiin. 4-Methoxyglucobrassicin (4 ME) and neoglucobrassicin (NEO) were purchased from Extrasynthese (Z.I Lyon Nord, Genay Cedex France), while glucoalyssin (ALY) was obtained from Shanghai ZZBIO Co., Ltd. (Shanghai, China). High-performance liquid chromatography (HPLC)-grade methanol, liquid chromatography/mass spectrometry (LC/MS)-grade methanol, and LC/MS-grade formic acid were supplied by Fisher Scientific (Pittsburgh, PA, USA). Ultrapure water was obtained from Milli-Q purification systems (Millipore, MA, USA). Individual stock solutions of the 13 GSLs were prepared in 70% methanol. Materials Broccoli seeds (Brassica oleracea var. italica) were purchased from Jingyan Yinong (Beijing) Seed Industry Technology Co., Ltd. (Beijing, China). Plant materials were stored at 4 °C prior to use. Cultivation of broccoli sprouts Seeds were soaked in distilled water for 3 h and evenly spread on a germination plate. The plates were placed in germination boxes (20 cm × 14 cm × 10.5 cm; DYJ-S6108, Shandong Hengmei Electronic Technology Co., Ltd., Shandong, China), and distilled water was used as the cultivation medium. The germination boxes were maintained in darkness at 30 °C for 2 days and were then transferred to an artificial growth chamber, where the sprouts were grown hydroponically in distilled water. The growth chamber was maintained at 70% relative humidity under a 12 h light/12 h dark photoperiod. During the light period, the temperature was 25 °C and the light intensity was 90 μmol/m2/s; during the dark period, the temperature was 20 °C. The sprouts were cultivated for 7 days from sowing to harvest. One portion of each fresh sample was analyzed immediately for the relevant response variables, whereas the remaining portion was stored at −80 °C. Electron beam irradiation Electron beam irradiation was performed at China National Nuclear Corporation Jinhui Radiation Co., Ltd. using a Siemens system at six dose levels (2, 4, 6, 8, 10, and 12 kGy); non-irradiated samples served as the control (CK). Electron beam irradiation was performed using a 10 MeV electron accelerator at an electron flux of 3.125 × 1013 electrons cm−2 s−1. Each target dose was delivered in a single conveyor pass by adjusting the conveyor speed to 5.5, 4.0, 3.0, 2.1, 1.7, and 1.4 m/min for 2, 4, 6, 8, 10, and 12 kGy, respectively. Thus, the corresponding absorbed doses per pass were 2, 4, 6, 8, 10, and 12 kGy. In this continuous conveyor system, the exposure duration was determined by the residence time under the electron beam rather than by a separately preset irradiation time. Each treatment consisted of three biological replicates. In the seed-stage experiment, irradiation was applied to seeds prior to germination without further irradiation during sprout growth (FZZZ). In the sprout-stage experiment, irradiation was applied once to the sprouts on the fifth day of germination and seeds were not irradiated (FZYM). Determining the sprout germination rate and length Ten fresh broccoli sprouts with uniform growth and no visible mechanical damage were randomly selected from each treatment to determine the growth parameters. Sprout height was measured individually as the distance from the base to the apical growing point, and fresh weight was determined using an electronic balance (Sartorius Scientific Instruments (Beijing) Co., Ltd., SZJY-117). The number of germinated plants was recorded when the sprouts were collected, and the germination rate was calculated according to the following formula (1). All data were subjected to statistical analysis.(1)Germinationrate=NgNt×100% Ng refers to the number of seeds showing radicle emergence; Nt is the total number ofseeds sown. Determination of GSLs GSL analysis was performed according to the method described by Yu et al. (2021). The samples included day-7 sprouts obtained from two independent irradiation treatments (FZYM and FZZZ), as well as irradiated seeds. Briefly, 1.00 g of fresh frozen sample powder was placed in a 50 mL centrifuge tube and extracted with 10 mL of 80% (v/v) methanol containing 1000 ng/mL TRO as a process internal standard to monitor the extraction and analytical procedures. The mixture was incubated at 75 °C for 20 min and then sonicated at 40 kHz and 25 °C for 20 min using a KQ3200DA numerical-control ultrasonic cleaner (Kunshan Ultrasonic Instrument Co., Ltd., China). The extract was subsequently centrifuged at 3000 × g for 15 min at 4 °C. Subsequently, 1 mL of the supernatant was diluted to 10 mL with ultrapure water and filtered through a 0.22 μm nylon syringe filter prior to ultra high-performance liquid chromatography–tandem mass spectrometry (UHPLC-MS/MS) analysis. UHPLC-MS/MS analysis was performed using a Waters ACQUITY I-class UHPLC system coupled with a Xevo TQ-S micro triple quadrupole (QQQ) mass spectrometer (Waters, Milford, MA, USA). Separation of GSLs was achieved using a Waters ACQUITY UHPLC® BEH C18 column (100 × 2.1 mm, 1.7 μm) maintained at 30 °C. The injection volume was 2 μL. The mobile phases consisted of ultrapure water containing 0.1% (v/v) formic acid (A) and methanol (B) with a gradient elution program conducted at a flow rate of 0.2 mL/min as follows: 0–1.0 min, 90% A; 3.0 min, 75% A; 5.0 min, 40% A; 6.0 min, 0% A; 6.2–9.0 min, 90% A. MS/MS was conducted using electrospray ionization (ESI) in negative ion mode. The MS parameters were set as follows: desolvation temperature, 500 °C; capillary voltage, 2.50 kV; desolvation gas flow rate, 1000 L/h. The fragment voltage, collision energy (CE), and multiple reaction monitoring (MRM) transitions were individually optimized for each compound (Table S1). Individual glucosinolates were quantified using compound-specific external calibration curves prepared with the corresponding authentic standards. Calibration curves were constructed by plotting the analyte peak area against its concentration. Determination of vitamin C, soluble sugars, and free amino acids Fresh broccoli sprout samples subjected to CK (0 kGy), FZZZ-2 (2 kGy), and FZZZ-4 (4 kGy) treatments were homogenized and analyzed immediately. Vitamin C content was determined according to GB 5009.86-2025. Soluble total sugar content was determined using the Shaffer–Somogyi copper-reduction iodometric method according to the Chinese agricultural industry standard NY/T 1278-2007, and total free amino acid content was calculated as the sum of the 17 quantified amino acids. Three biological replicates were analyzed for each treatment. All results were expressed on a fresh-weight basis: vitamin C as mg/100 g FW and soluble sugars and free amino acids as g/100 g FW. Determination of sulfur-containing volatile compounds Fresh broccoli sprouts were frozen in liquid nitrogen and ground into a fine powder. A 0.20 g sample was transferred into a 20 mL headspace vial, and 5 μL of 2-methyl-3-heptanone solution (100 mg/L) was added as the internal standard. The vial was immediately sealed and equilibrated at 50 °C for 40 min with agitation at 450 rpm. An SPME fiber was then exposed to the vial headspace at 50 °C for 40 min under continuous agitation. Subsequently, the fiber was inserted into the GC inlet and thermally desorbed at 250 °C for 5 min. GC–MS analysis was performed using a Nexis GC-2030 gas chromatograph coupled to a GCMS-QP2020 NX mass spectrometer (Shimadzu, Kyoto, Japan). Volatile compounds were separated using an SH-PolarWax capillary column (60 m × 0.25 mm i.d., 0.25 μm film thickness; Cat. No. 227-36247-02, Shimadzu). Helium was used as the carrier gas at a constant flow rate of 1.0 mL/min. The oven temperature was maintained at 40 °C for 5 min, increased to 250 °C at 3 °C/min, and held at 250 °C for 15 min. The mass spectrometer was operated in electron ionization mode at 70 eV. The ion-source and interface temperatures were both 230 °C, and mass spectra were acquired in full-scan mode over an m/z range of 35–400. Compounds were identified by comparing their retention indices and mass spectra with those of authentic standards and entries in the NIST mass spectral library. They were semi-quantified using the internal-standard method and expressed as μg/kg fresh weight (FW). Metabolomic analysis Sample preparation and extraction Seven-day-old sprouts were vacuum freeze-dried for 63 h using a Scientz-100F freeze dryer (Ningbo Scientz Biotechnology Co., Ltd., Ningbo, China) and ground into a fine powder using an MM 400 mixer mill (Retsch GmbH, Haan, Germany) at 30 Hz for 1.5 min. Subsequently, 30 mg of the powdered sample was extracted with 1500 μL of 70% (v/v) aqueous methanol that contained an internal-standard mixture and had been pre-cooled to −20 °C. The samples were vortexed for 30 s at 30-min intervals for a total of six cycles. After centrifugation at 12,000 rpm for 3 min, the supernatants were filtered through a 0.22 μm membrane and transferred to autosampler vials for UHPLC–MS/MS analysis. UHPLC ESI-Q TRAP-MS/MS conditions Sample extracts were analyzed using an ExionLC™ AD UHPLC system coupled with a QTRAP tandem mass spectrometer (SCIEX). Separation was achieved with an Agilent SB-C18 column (1.8 μm, 2.1 × 100 mm) using mobile phase A (0.1% formic acid) and mobile phase B (acetonitrile containing 0.1% formic acid). The gradient program was as follows: 0–9.0 min, 5–95% B; 9.0–10.0 min, 95% B; 10.0–11.1 min, 95–5% B; and 11.1–14.0 min, 5% B. The flow rate, column temperature, and injection volume were 0.35 mL/min, 40 °C, and 2 μL, respectively. Detection was performed using ESI–MS/MS. The ESI source parameters were set as follows: source temperature, 500 °C; ion-spray voltage (IS), 5500 V in positive ion mode and −4500 V in negative ion mode; and ion source gas I, ion source gas II, and curtain-gas pressures, 50, 60, and 25 psi, respectively. Collision-activated dissociation was set to high. The QTRAP mass spectrometer was operated in QQQ mode for MRM acquisition, with nitrogen used as the collision gas at the medium setting. The DP and CE for each MRM transition were individually optimized. Specific MRM transitions were monitored within defined retention time windows according to the elution characteristics of the metabolites. Metabolites were identified by matching their MS/MS information with entries in the in-house MWDB database, followed by manual verification. Relative quantification was performed using the integrated peak areas of compound-specific Q1-to-Q3 MRM transitions, with Q1 and Q3 representing the precursor and characteristic product ions, respectively. Raw data were processed using Analyst 1.6.3 and MultiQuant software. The MRM transitions, ion forms, and metabolite-identification levels are provided in Table S2. The monitored Q1 and Q3 ions covered m/z 70.07-1241.32 and m/z 51.03-963.26, respectively. Because data acquisition was performed in MRM rather than conventional full-scan mode, a conventional full-scan mass range and mass-scan resolution were not applicable. KEGG annotation and enrichment analysis Identified metabolites were annotated using the KEGG COMPOUND database (http://www.kegg.jp/kegg/compound/). Then, annotated metabolites were mapped to the KEGG PATHWAY database (http://www.kegg.jp/kegg/pathway.html). RNA extraction and quantitative real-time PCR analysis Total RNA was extracted from fresh broccoli sprout tissues using TRIzol reagent (Thermo Fisher Scientific, USA), and its concentration and purity were determined using a NanoDrop 2000 spectrophotometer. First-strand cDNA was synthesized from 500 ng of total RNA using HiScript IV All-in-One Ultra RT SuperMix for qPCR (Vazyme, Nanjing, China). The cDNA was diluted tenfold and subjected to quantitative real-time PCR using PowerUp SYBR Green Master Mix (Thermo Fisher Scientific, USA) on a Roche real-time PCR system. The amplification program consisted of 95 °C for 2 min, followed by 40 cycles of 95 °C for 3 s and 60 °C for 30 s. The primer sequences are listed in Table S3. Statistical analysis All results were expressed as the arithmetic mean ± standard error. Data were processed using Microsoft Excel. Figures were generated with GraphPad Prism 10.0, and statistical analyses were performed using SPSS software (version 27, IBM, New York, NY, USA). Depending on the experimental design, significance was evaluated using one-way analysis of variance (ANOVA) or Student's t-test, and mean comparisons among treatments were conducted using the least significant difference method. Differences were considered statistically significant at p < 0.05. Unsupervised principal component analysis (PCA) was performed using the prcomp function in R (www.r-project.org). Hierarchical cluster analysis (HCA) results of samples and metabolites were visualized as heatmaps with dendrograms. Pearson's correlation coefficients (PCCs) between samples were calculated using the cor function in R and visualized as heatmaps. Both HCA and PCC were carried out using the R package ComplexHeatmap. For HCA, normalized signal intensities of metabolites (unit variance scaling) were visualized as a color spectrum. For two-group analysis, differential metabolites were determined by variable importance in projection (VIP >1) and absolute Log2FC (|Log2FC| ≥1.0). For multi-group analysis, differential metabolites were identified using the criteria VIP > 1 and ANOVA p < 0.05. VIP values were extracted from the orthogonal partial least squares discriminant analysis (OPLS-DA) model. Score plots and permutation plots were generated using the R package MetaboAnalystR. The data were log transformed (Log2) and mean centered before OPLS-DA. To avoid overfitting, a permutation test (200 permutations) was performed. Results and discussion Sprout-stage irradiation As shown in Fig. 1A and B, plant height and fresh weight were significantly reduced in the FZYM group when the irradiation dose exceeded 4 kGy. Direct sprout-stage irradiation produced a non-linear, dose-dependent response in GSL concentration and accumulation (Table 1). A total of six GSLs were identified (Fig. S1), namely ERU, RAA, ALY, NEO, 4 ME, and GBC. Total GSL concentration was lower than that of CK at 2–10 kGy, reaching 2058.39 and 543.69 μg/g FW at 6 and 8 kGy, respectively, compared with 3174.58 μg/g FW in CK. However, it rebounded to 3551.03 μg/g FW at 12 kGy, slightly exceeding the CK value. Total GSL accumulation also did not decrease uniformly: it increased from 111.11 μg in CK to 146.75 and 148.52 μg at 2 and 4 kGy, respectively, decreased markedly at 8 and 10 kGy, and then rebounded to 124.29 μg at 12 kGy. However, sprouts lost viability at 12 kGy, indicating that the apparent GSL rebound had no practical benefit and may represent a terminal defense response to lethal irradiation stress. Irradiation- and wounding-induced stresses can promote GSL biosynthesis through the activation of related genes and signaling pathways involving reactive oxygen species and ethylene (Banerjee et al., 2016; Villarreal-Garcia et al., 2016; Torres-Contreras et al., 2021).Fig. 1Effects of electron-beam irradiation on broccoli seed germination and sprout growth. (A, B) Plant height and fresh weight following irradiation at the sprout-growth stage; (C, D) germination rate and germination potential following seed irradiation; and (E, F) plant height and fresh weight following irradiation at the seed stage. Values are presented as means ± SD (A, B, E, and F: n = 10). Different lowercase letters indicate significant differences among irradiation doses (P < 0.05); bars sharing the same letter are not significantly different.Table 1Effects of irradiation on glucosinolate content and accumulation in broccoli sprouts.Table 1 Effects of electron-beam irradiation on broccoli seed germination and sprout growth. (A, B) Plant height and fresh weight following irradiation at the sprout-growth stage; (C, D) germination rate and germination potential following seed irradiation; and (E, F) plant height and fresh weight following irradiation at the seed stage. Values are presented as means ± SD (A, B, E, and F: n = 10). Different lowercase letters indicate significant differences among irradiation doses (P < 0.05); bars sharing the same letter are not significantly different. Table 1Effects of irradiation on glucosinolate content and accumulation in broccoli sprouts.Table 1 Effects of irradiation on glucosinolate content and accumulation in broccoli sprouts. Seed-stage irradiation In the seed-stage irradiation scheme (FZZZ), the germination performance showed a clear high-dose inhibition. Germination rates declined from 94.23% in the CK to 29.81% and 9.13% at 6 kGy and 8 kGy, respectively (Fig. 1C). Germination potential declined from 92.31% in the CK to 10.10%, 3.37%, 1.44%, and 0.48% at 6 kGy, 8 kGy, 10 kGy, and 12 kGy, respectively (Fig. 1D), which indicated that high-dose irradiation markedly impaired seed viability. Consequently, the germination of seeds irradiated at doses greater than 4 kGy was not practically feasible. Therefore, sprout growth and GSL accumulation were evaluated after seeds had been exposed to irradiation doses of 0, 2, or 4 kGy. Plant height decreased from 3.76 cm in the CK to 3.24 cm at 2 kGy and 2.66 cm at 4 kGy (Fig. 1E), whereas plant weight was not significantly affected by 2 kGy and 4 kGy irradiation (Fig. 1F). Despite these growth penalties, low-dose seed irradiation substantially enhanced the accumulation of GSLs in sprouts. At 2 and 4 kGy, the concentration of GSLs in sprouts increased to 6556.85 μg/g and 6216.78 μg/g, respectively, compared with 3174.58 μg/g in the CK. Similarly, the accumulation levels of GSL increased to 229.49 μg and 217.59 μg at 2 kGy and 4 kGy, respectively, compared with 111.11 μg in the CK (Table 1 and Table S4). The strongest effect was observed at 2 kGy, at which both metrics were 2.07 times those of the CK, whereas higher doses primarily reduced germination performance. Given the increased concentrations of GSLs in sprouts derived from irradiated seeds, it was examined whether seed irradiation increased the levels of GSL in seeds prior to germination. The total GSL content of seeds did not differ significantly among treatments, ranging from 18,287.96 to 19,819.02 μg/g (Table S5). These results suggest that the higher levels of GSL in sprouts after low-dose seed irradiation are more likely associated with metabolic regulation during germination than with greater seed reserves of GSL. This study found that EBI applied during the sprout growth stage was unfavorable for GSL accumulation in broccoli sprouts. In contrast, seed pre-irradiation followed by germination promoted GSL accumulation. Notably, GSL content did not differ significantly between seeds before and after irradiation. This result indicates that the enhancement of GSL accumulation in sprouts was not simply attributable to changes in the GSL reserves of the seeds. Instead, these contrasting outcomes likely reflect irradiation-induced metabolic reprogramming during germination and subsequent sprout development. Effects of seed-stage irradiation on selected nutritional components Seed irradiation significantly affected the selected nutritional components of broccoli sprouts (Fig. 2). Compared with the non-irradiated control, vitamin C content increased by 49.5% and 80.6% following the 2- and 4-kGy treatments, respectively (Fig. 2A). Soluble sugar content increased by 64.0% and 102.3% (Fig. 2B), while total amino acid content increased by 40.8% and 98.6%, respectively (Fig. 2C). Significant differences were observed among the control, 2-kGy, and 4-kGy treatments for all three nutritional indices (P < 0.05). These results indicate that irradiation of broccoli seeds at the tested doses did not cause depletion of vitamin C, soluble sugars, or total amino acids in the resulting sprouts.Fig. 2Effects of seed-stage electron-beam irradiation on selected nutritional components of broccoli sprouts. (A) Vitamin C content, (B) soluble sugar content, and (C) total amino acid content. CK represents the non-irradiated control, while 2 and 4 kGy indicate the absorbed irradiation doses applied to seeds before germination. Bars represent the means ± SD (n = 3 samples per treatment), and open circles represent individual sample values. Different lowercase letters within each panel indicate significant differences among treatments according to one-way ANOVA followed by Tukey's HSD test (P < 0.05). FW, fresh weight. Effects of seed-stage electron-beam irradiation on selected nutritional components of broccoli sprouts. (A) Vitamin C content, (B) soluble sugar content, and (C) total amino acid content. CK represents the non-irradiated control, while 2 and 4 kGy indicate the absorbed irradiation doses applied to seeds before germination. Bars represent the means ± SD (n = 3 samples per treatment), and open circles represent individual sample values. Different lowercase letters within each panel indicate significant differences among treatments according to one-way ANOVA followed by Tukey's HSD test (P < 0.05). FW, fresh weight. Effects of seed irradiation on GSL-related sulfur-containing compounds To complement the analysis of intact GSLs, sulfur-containing compounds were profiled to characterize potential downstream products associated with GSL hydrolysis in broccoli sprouts. A total of 21 sulfur-containing compounds were identified and classified into six chemical groups, including 12 isothiocyanates, three sulfur-containing nitriles, two sulfides, two sulfur-containing heterocycles, one thiocyanate, and one sulfur-containing alcohol (Table 2). Eighteen compounds were detected in CK, whereas all 21 compounds were detected in the 2- and 4-kGy treatments. The summed relative content of the 21 compounds was 6657.94 ± 549.34, 10,032.99 ± 469.12, and 4901.48 ± 710.19 μg/kg in CK, 2-kGy, and 4-kGy sprouts, respectively. Thus, the total content increased by 50.7% at 2 kGy but decreased by 26.4% at 4 kGy compared with CK. Among all detected compounds, 5-(methylthio)pentanenitrile was the most abundant, accounting for 88.8%, 90.8%, and 86.4% of the total sulfur-containing compounds in CK, 2-kGy, and 4-kGy sprouts, respectively.Table 2Sulfur-containing compounds in broccoli sprouts grown from electron beam-irradiated seeds.Table 2No.NameFormulac/lRIaRelative content (μg/kg)IentificationbCK2 kGy4 kGyIsothiocyanates1Isopropyl isothiocyanateC4H7NS1169/1177-3.06 ± 0.442.59 ± 0.03RI, S22-Butyl isothiocyanateC5H9NS1290/12873.83 ± 0.059.45 ± 1.5113.02 ± 0.04RI, S3Butyl isothiocyanateC5H9NS1322/1308-1.95 ± 0.211.79 ± 0.05RI, S4Isobutyl isothiocyanateC5H9NS1324/-2.90 ± 0.474.62 ± 1.497.20 ± 0.09RI, S53-Butenyl isothiocyanateC5H7NS1430/14536.15 ± 0.2015.02 ± 2.3613.04 ± 0.09RI, S6N-Pentyl isothiocyanateC6H11NS1465/14732.09 ± 0.035.95 ± 0.803.38 ± 0.02RI, S74-Methylpentyl isothiocyanateC7H13NS1533/152959.86 ± 3.80112.04 ± 13.577.46 ± 4.04RI, S8Hexyl isothiocyanateC7H13NS1575/158838.43 ± 5.88102.21 ± 17.2348.99 ± 2.85RI, S9Heptyl isothiocyanateC8H15NS1689/169626.74 ± 1.7534.33 ± 2.4123.21 ± 4.67RI, S103-(Methylthio)propyl isothiocyanateC5H9NS21964/197912.17 ± 1.7828.81 ± 1.7320.99 ± 0.53RI, S11Benzyl isothiocyanateC8H7NS2088/20718.98 ± 1.0811.30 ± 2.338.45 ± 0.30RI, S12Phenethyl isothiocyanateC9H9NS2230/2234344.59 ± 35.41363.40 ± 29.02286.02 ± 16.68RI, SSulfur-containing nitriles134-(Methylthio)butanenitrileC5H9NS1800/18063.89 ± 0.3010.55 ± 1.487.63 ± 0.86RI, S145-(Methylthio)pentanenitrileC6H11NS1855/18505913.87 ± 589.169109.00 ± 573.864237.07 ± 712.01RI, S156-(Methylthio)hexanenitrileC7H13NS2097/2089176.43 ± 11.42127.45 ± 19.3383.11 ± 0.27RI, SSulfides16Amyl methyl sulfideC6H14S1132/114032.68 ± 0.8345.22 ± 4.3338.67 ± 3.92RI, S17Diallyl sulfideC6H10S1138/1148-0.47 ± 0.070.33 ± 0.02RI, SThiocyanates18Thiocyanic acid methyl esterC2H3NS1288/127611.84 ± 0.3230.07 ± 5.0215.02 ± 0.41RI, SSulfur-containing heterocycles19TetrahydrothiopheneC4H8S1099/11094.17 ± 0.126.09 ± 0.834.37 ± 0.43RI, S20BenzothiazoleC7H5NS1932/19473.45 ± 0.015.93 ± 0.242.73 ± 0.03RI, SSulfur-containing alcohols214-(Methylthio)butanolC5H12OS1820/18125.86 ± 0.046.07 ± 1.046.38 ± 0.36RI, STotal6657.94 ± 549.3410032.99 ± 469.124901.48 ± 710.19acRI and lRI represent the calculated and literature retention indices, respectively, on a WAX capillary column.bIdentification methods: RI, identification by comparison of the calculated and literature retention indices; S, identification using authentic standards. Table 2Sulfur-containing compounds in broccoli sprouts grown from electron beam-irradiated seeds.Table 2No.NameFormulac/lRIaRelative content (μg/kg)IentificationbCK2 kGy4 kGyIsothiocyanates1Isopropyl isothiocyanateC4H7NS1169/1177-3.06 ± 0.442.59 ± 0.03RI, S22-Butyl isothiocyanateC5H9NS1290/12873.83 ± 0.059.45 ± 1.5113.02 ± 0.04RI, S3Butyl isothiocyanateC5H9NS1322/1308-1.95 ± 0.211.79 ± 0.05RI, S4Isobutyl isothiocyanateC5H9NS1324/-2.90 ± 0.474.62 ± 1.497.20 ± 0.09RI, S53-Butenyl isothiocyanateC5H7NS1430/14536.15 ± 0.2015.02 ± 2.3613.04 ± 0.09RI, S6N-Pentyl isothiocyanateC6H11NS1465/14732.09 ± 0.035.95 ± 0.803.38 ± 0.02RI, S74-Methylpentyl isothiocyanateC7H13NS1533/152959.86 ± 3.80112.04 ± 13.577.46 ± 4.04RI, S8Hexyl isothiocyanateC7H13NS1575/158838.43 ± 5.88102.21 ± 17.2348.99 ± 2.85RI, S9Heptyl isothiocyanateC8H15NS1689/169626.74 ± 1.7534.33 ± 2.4123.21 ± 4.67RI, S103-(Methylthio)propyl isothiocyanateC5H9NS21964/197912.17 ± 1.7828.81 ± 1.7320.99 ± 0.53RI, S11Benzyl isothiocyanateC8H7NS2088/20718.98 ± 1.0811.30 ± 2.338.45 ± 0.30RI, S12Phenethyl isothiocyanateC9H9NS2230/2234344.59 ± 35.41363.40 ± 29.02286.02 ± 16.68RI, SSulfur-containing nitriles134-(Methylthio)butanenitrileC5H9NS1800/18063.89 ± 0.3010.55 ± 1.487.63 ± 0.86RI, S145-(Methylthio)pentanenitrileC6H11NS1855/18505913.87 ± 589.169109.00 ± 573.864237.07 ± 712.01RI, S156-(Methylthio)hexanenitrileC7H13NS2097/2089176.43 ± 11.42127.45 ± 19.3383.11 ± 0.27RI, SSulfides16Amyl methyl sulfideC6H14S1132/114032.68 ± 0.8345.22 ± 4.3338.67 ± 3.92RI, S17Diallyl sulfideC6H10S1138/1148-0.47 ± 0.070.33 ± 0.02RI, SThiocyanates18Thiocyanic acid methyl esterC2H3NS1288/127611.84 ± 0.3230.07 ± 5.0215.02 ± 0.41RI, SSulfur-containing heterocycles19TetrahydrothiopheneC4H8S1099/11094.17 ± 0.126.09 ± 0.834.37 ± 0.43RI, S20BenzothiazoleC7H5NS1932/19473.45 ± 0.015.93 ± 0.242.73 ± 0.03RI, SSulfur-containing alcohols214-(Methylthio)butanolC5H12OS1820/18125.86 ± 0.046.07 ± 1.046.38 ± 0.36RI, STotal6657.94 ± 549.3410032.99 ± 469.124901.48 ± 710.19acRI and lRI represent the calculated and literature retention indices, respectively, on a WAX capillary column.bIdentification methods: RI, identification by comparison of the calculated and literature retention indices; S, identification using authentic standards. Sulfur-containing compounds in broccoli sprouts grown from electron beam-irradiated seeds. cRI and lRI represent the calculated and literature retention indices, respectively, on a WAX capillary column. Identification methods: RI, identification by comparison of the calculated and literature retention indices; S, identification using authentic standards. Isothiocyanates constituted the largest chemical class in terms of the number of compounds detected. Following tissue disruption, myrosinase removes glucose from the parent GSL and produces an unstable thiohydroximate-O-sulfate aglucone. In the absence of active specifier proteins, this intermediate generally rearranges spontaneously to form the corresponding isothiocyanate. In the present study, the summed volatile isothiocyanate contents were 505.74, 692.14, and 506.14 μg/kg in CK, 2-kGy, and 4-kGy sprouts, respectively. Phenethyl isothiocyanate was the predominant isothiocyanate in all three treatments, with contents of 344.59, 363.40, and 286.02 μg/kg, respectively. Sulfur-containing nitriles were less diverse but quantitatively dominated the detected compounds. Their summed contents were 6094.19, 9247.00, and 4327.81 μg/kg, accounting for 91.5%, 92.2%, and 88.3% of the total sulfur-containing compounds in CK, 2-kGy, and 4-kGy sprouts, respectively. Nitrile formation represents an alternative branch of GSL hydrolysis. Epithiospecifier proteins and nitrile-specifier proteins can redirect the unstable GSL aglucone away from spontaneous isothiocyanate formation and toward nitrile production. Previous research has shown that 4-(methylthio)butyl GSL, i.e., glucoerucin, can generate either 4-(methylthio)butyl isothiocyanate or 5-(methylthio)pentanenitrile, depending on the hydrolysis conditions and specifier-protein activity (Hanschen and Schreiner, 2017). The predominance of 5-(methylthio)pentanenitrile in the present study is therefore consistent with substantial metabolic partitioning of glucoerucin-derived intermediates toward the nitrile-forming branch. Relative to CK, the nitrile content increased by 51.7% at 2 kGy but decreased by 29.0% at 4 kGy. These changes largely accounted for the corresponding variation in the total content of sulfur-containing compounds. The remaining compound classes occurred at considerably lower levels. The total sulfide contents were 32.68, 45.69, and 39.00 μg/kg in CK, 2-kGy, and 4-kGy sprouts, respectively, with amyl methyl sulfide as the major sulfide. Diallyl sulfide was detected only after irradiation. The only detected thiocyanate, thiocyanic acid methyl ester, increased from 11.84 μg/kg in CK to 30.07 μg/kg at 2 kGy and subsequently decreased to 15.02 μg/kg at 4 kGy. Organic thiocyanates can be formed when thiocyanate-forming proteins redirect GSL aglucones with suitable side-chain structures, although the precise precursor of thiocyanic acid methyl ester in the present samples could not be established (Kuchernig et al., 2011). The two sulfur-containing heterocycles, tetrahydrothiophene and benzothiazole, had combined contents of 7.62, 12.02, and 7.10 μg/kg in CK, 2-kGy, and 4-kGy sprouts, respectively. Both compounds reached their highest levels at 2 kGy. Unlike isothiocyanates and nitriles, sulfides, sulfur-containing alcohols, and sulfur-containing heterocycles are not necessarily the immediate products of myrosinase-mediated GSL hydrolysis. They may also originate from secondary transformation of unstable sulfur-containing intermediates or from other sulfur metabolic pathways. Therefore, their direct precursor-product relationships cannot be assigned without isotope-tracing or enzyme-activity experiments. The dose-dependent variation in sulfur-containing compounds was generally consistent with that of the total GSL content, which increased from 3174.58 μg/g FW in CK to 6556.85 μg/g FW at 2 kGy and then decreased slightly to 6216.78 μg/g FW at 4 kGy. Both total GSLs and sulfur-containing degradation products therefore reached their highest levels following the 2 kGy treatment, suggesting that the enhanced accumulation of GSL precursors contributed, at least partly, to the increased formation of sulfur-containing products. However, the total content of sulfur-containing compounds decreased markedly at 4 kGy, whereas the total GSL content showed only a relatively small reduction compared with that at 2 kGy. This disproportionate response indicates that the abundance of degradation products was not determined solely by the GSL substrate pool. Irradiation-induced changes in myrosinase activity, specifier protein activity, cellular compartmentalization, and the subsequent stability or transformation of hydrolysis products may also have affected the conversion efficiency and product distribution of GSLs. Metabolites in broccoli sprouts To characterize treatment-dependent metabolic shifts and to identify the key pathways and regulatory nodes underlying differential GSL accumulation, metabolomic analysis was conducted on non-irradiated sprouts (CK), sprouts germinated from seeds irradiated at 2 kGy (FZZZ_2) and 4 kGy (FZZZ_4), and sprouts directly irradiated at 2 kGy (FZYM_2) and 4 kGy (FZYM_4). Principal component analysis To initially assess the overall metabolic differences among samples and the variability within each group, PCA was performed on nine samples. The results showed good discrimination among the sample groups (Fig. 3A). Along principal component 1 (PC1; 42.13% of the variance explained), the FZYM and FZZZ groups were clearly separated from one aother, and both were distinctly separated from CK. Along principal component 2 (PC2; 21.76% of the variance explained), FZYM_2 and FZYM_4 were clearly separated, while the FZYM and FZZZ groups also remained distinct from CK.Fig. 3(A) Principal component analysis (PCA) of the samples; (B) Pearson correlation analysis; (C) overview of the metabolites detected in each sample; and (D) overview of differential metabolites among samples. (A) Principal component analysis (PCA) of the samples; (B) Pearson correlation analysis; (C) overview of the metabolites detected in each sample; and (D) overview of differential metabolites among samples. Within-group correlation analysis was conducted to evaluate the reproducibility of biological replicates. Pearson's correlation coefficient (r) calculated using the built-in cor function in R and used to assess reproducibility. As shown in Fig. 3B, the |r| values within each group were close to 1, which indicated reliable biological reproducibility within the groups in this study. Overview of metabolites in each sample group As shown in Fig. 3C, untargeted metabolomics detected more than 3100 metabolites in each group. The metabolite profiles had similar class compositions and were dominated by amino acids and their derivatives, terpenoids, lipids, alkaloids, flavonoids, and phenolic acids. The CK group contained 3150 metabolites, with major classes including amino acids and derivatives (544), terpenoids (397), lipids (379), and alkaloids (370). Irradiation generally increased the number of detected metabolites. In the sprout-irradiated treatments, 3247 and 3232 metabolites were detected in FZYM_2 and FZYM_4, respectively. The corresponding counts in selected classes were 560 and 554 for amino acids and their derivatives, 405 and 420 for terpenoids, and 321 and 314 for flavonoids. For seed irradiation followed by germination, FZZZ_2 (3162 metabolites) showed slight shifts, mainly in lipids (383) and alkaloids (381). Notably, FZZZ_4 had the highest number of detected metabolites (3,273), including 334 flavonoids and 265 phenolic acids.). These results suggest that higher-dose seed irradiation may enhance secondary metabolism during germination and contribute to greater metabolic diversity and GSL enrichment. Identification of differential metabolites Orthogonal partial least squares discriminant analysis Metabolic differences among the sample groups were determined using OPLS-DA. In the model, R2X and R2Y represent the explained variance of X and Y, respectively, while Q2 reflects the predictive ability of the model (Q2 > 0.9 indicates an excellent model, while Q2 >0.5 represents an acceptable model). The OPLS-DA models showed high predictive performance (Q2) and a good model fit (R2X and R2Y). Clear separation was observed among CK, FZYM_2, FZYM_4, FZZZ_2, and FZZZ_4, which indicated distinct metabolic profiles across treatments. The OPLS-DA score plots for the pairwise comparisons are shown in Fig. S2. Differences in metabolites of broccoli sprouts under different irradiation methods As shown in Fig. 3D and Fig. S3, direct irradiation at the sprout stage (FZYM) induced significant metabolic changes. The FZYM_2-versus-CK comparison yielded 892 differential metabolites, of which 543 were upregulated and 349 were downregulated. The upregulated metabolites were mainly lipids (106) and terpenoids (91), whereas the downregulated metabolites were mainly phenolic acids (43), alkaloids (39), and flavonoids (31). At 4 kGy, the FZYM_4-versus-CK comparison identified 1374 differential metabolites, including 825 upregulated and 549 downregulated metabolites. The upregulated metabolites were mainly lipids (147), amino acids and their derivatives (142), and terpenoids (135), whereas the downregulated metabolites were mainly flavonoids (72) and phenolic acids (64). Consistently, FZYM_4 vs. FZYM_2 identified 773 differential metabolites (481 up and 292 down), with upregulated metabolites dominated by lipids (97), terpenoids (84), and amino acids and derivatives (71), and downregulated metabolites mainly distributed in flavonoids (39), others (38), and alkaloids (32). Collectively, these patterns suggest that sprout-stage irradiation preferentially stimulates lipid- and terpenoid-associated metabolism while suppressing some phenolic- and flavonoid-related processes. Because these pathways contribute to membrane remodeling and stress signaling, the observed changes are consistent with coordinated reprogramming of stress-responsive metabolism. This interpretation is consistent with the central role of membrane lipid remodeling and signaling lipids in abiotic stress adaptation (Liu et al., 2019) and with radiation-driven reactive oxygen species (ROS)/oxidative processes that can promote membrane perturbation and lipid peroxidation responses (Tan et al., 2023). Terpenoids are widely recognized as stress-responsive metabolites involved in defense and signaling. These functions provide a plausible basis for their preferential induction under sprout-stage irradiation (Boncan et al., 2020). In contrast, seed irradiation followed by germination (FZZZ) produced extensive metabolic changes, with a stronger bias toward downregulation at 2 kGy (Fig. 3D and S3)A total of 1301 differential metabolites were identified in the FZZZ_2-versus-CK comparison, including 412 upregulated and 889 downregulated metabolites. The upregulated metabolites were mainly phenolic acids (53) and flavonoids (38), whereas the downregulated metabolites were mainly amino acids and their derivatives (276), terpenoids (134), and lipids (80). This pattern indicates substantial perturbation of primary and secondary metabolism during early germination. At 4 kGy, the FZZZ_4-versus-CK comparison identified 1272 differential metabolites, including 633 upregulated and 639 downregulated metabolites. The upregulated metabolites were mainly lipids (133), metabolites classified as “others” (130), and flavonoids (105), whereas the downregulated metabolites were mainly amino acids and their derivatives (217), alkaloids (73), and terpenoids (65). Notably, the dose comparison (FZZZ_4 vs. FZZZ_2) revealed 906 differential metabolites (738 upregulated and 168 downregulated), with significantly increased levels of lipids (159), amino acids and derivatives (150), and terpenoids (111). This pattern may reflect a stronger compensatory metabolic response at higher doses during germination. The strong bias toward downregulation in the FZZZ_2-versus-CK comparison—particularly among amino acids and their derivatives, terpenoids, and lipids—suggests that irradiation before germination substantially perturbs the reactivation of primary metabolism required for early seedling establishment. This process includes nitrogen mobilization and the use of lipid and protein reserves (Rosental et al., 2014). The shift at 4 kGy (FZZZ_4 vs. FZZZ_2) toward broad upregulation of lipids, amino acids/derivatives, and terpenoids is consistent with stress-adaptive (compensatory) metabolic activation as germination proceeds under higher stress load. This pattern is compatible with radiation-triggered stress priming/hormetic frameworks described for plants under ionizing radiation (Kim, 2025). Stage-dependent differences were further supported by comparisons under the same dose (Fig. 3D and Fig. S3). FZZZ_2 vs. FZYM_2 produced 1708 differential metabolites (523 up and 1185 down), with upregulated features mainly in flavonoids (83) and alkaloids (51), while downregulated features were dominated by amino acids and derivatives (312), terpenoids (189), and lipids (160). Similarly, FZZZ_4 vs. FZYM_4 yielded 1660 differential metabolites (639 up and 1021 down), with upregulated metabolites enriched in others (132), flavonoids (129), and phenolic acids (72), and downregulated metabolites dominated by amino acids and derivatives (310), terpenoids (168), and lipids (122). At the same dose, the FZZZ-versus-FZYM comparisons yielded extensive metabolic differences. These findings suggest that irradiation before germination may alter the trajectory of metabolic reactivation during germination, whereas irradiation at the sprout stage may preferentially engage membrane-centered and signaling-associated metabolism. This aligns with the concept that germination requires coordinated activation of core metabolic programs (energy, carbon–nitrogen balance, and reserve mobilization) that are highly sensitive to stress perturbation (Rosental et al., 2014), while established tissues often exhibit rapid lipid remodeling/signaling responses to abiotic challenges (Liu et al., 2019). In Brassicales vegetables such as broccoli (Brassica oleracea var. italica), GSLs are sulfur- and nitrogen-containing secondary metabolites that contribute to plant defense and stress interactions (Galanakis, 2021). In the present study, GSLs accumulated to higher levels in the seed-irradiation–germination treatments (FZZZ_2 and FZZZ_4, with FZZZ_2 being the highest), whereas sprout-stage irradiation (FZYM_2 and FZYM_4) reduced GSLs relative to CK. Notably, these GSL shifts occurred alongside stage-specific remodeling of phenylpropanoid/flavonoid-related features. A mechanistic framework for interpreting such co-variation is provided by genetic evidence that GSL–phenylpropanoid crosstalk can be driven by GSL-pathway intermediates (aldoximes or their derivatives), which negatively impact phenylpropanoid output (Kim et al., 2020). This regulation is attributed to oxime accumulation rather than the absence of end-product GSLs, and may extend beyond Arabidopsis because aldoximes occur in diverse plant species (Kim et al., 2020). Mechanistically, aldoxime-associated crosstalk involves transcriptional regulation of a set of Kelch-domain F-box (KFB) genes. The encoded KFB proteins promote phenylalanine ammonia-lyase (PAL) ubiquitination, thereby enhancing proteasome-mediated PAL turnover and constraining the entry step of phenylpropanoid biosynthesis (Kim et al., 2020). Within this context, the lower abundances of several flavonoid- and phenolic-related metabolites in FZYM are consistent with altered phenylpropanoid-associated metabolism; however, the present data do not establish changes in PAL regulation or metabolic flux. By contrast, germination after seed irradiation (FZZZ) was associated with greater GSL accumulation and a larger proportion of upregulated phenolic and flavonoid metabolites. This implies that the balance between GSL-associated defense chemistry and phenylpropanoid allocation is strongly constrained by the developmental stage under electron beam stress. KEGG annotation and enrichment analysis of differential metabolites KEGG annotation showed that 142–245 differential metabolites were annotated across the comparison groups and mapped to 88–122 metabolic pathways. Overall, the annotated metabolites were predominantly assigned to the KEGG categories “metabolic pathways” and “biosynthesis of secondary metabolites,” which accounted for approximately 47.31–54.93% and 23.77–31.69% of the annotated metabolites, respectively. These categories were followed by cofactor biosynthesis and several pathways related to amino acid and fatty acid metabolism. KEGG pathway enrichment analysis KEGG pathway enrichment analysis based on differential metabolites revealed clear divergence in the types of significantly enriched pathways between FZYM and FZZZ. In FZYM, the enriched pathways were dominated by unsaturated fatty acid, aromatic amino acid, phenylpropanoid, and nucleotide metabolism. This pattern suggests stress-responsive metabolic reorganization associated with membrane remodeling, signaling, and repair processes. In FZYM_2 vs. CK, differential metabolites were significantly enriched in linoleic acid metabolism, biosynthesis of phenylacetic acid derivatives, and phenylalanine metabolism (Fig. 4A, p < 0.05). In FZYM_4 vs. CK, differential metabolites were primarily enriched in the biosynthesis of cinnamic acid derivatives, and the metabolism of linoleic acid and purines (Fig. 4B, p < 0.05). When comparing doses directly (FZYM_4 vs. FZYM_2), significant enrichment mainly involved the metabolism of tryptophan and purines (Fig. 4C, p < 0.05). Overall, enrichment of pathways related to unsaturated fatty acids, aromatic amino acids, phenylpropanoids, and nucleotides after sprout-stage irradiation suggested activation of metabolic modules involved in membrane lipid regulation, stress signaling, repair, and energy metabolism.When considered together with the GSL phenotype—in which FZYM-2 and FZYM-4 had lower GSL contents than CK—this enrichment profile is consistent with acute stress-related metabolic reallocation. Specifically, metabolism may have shifted toward immediate defense and repair processes rather than the accumulation of biosynthetic defense metabolites such as GSLs.Fig. 4KEGG annotation and enrichment analysis of differentially abundant metabolites in each pairwise comparison of broccoli sprout samples. (A): FZYM_2 vs CK, (B): FZYM_4 vs CK, (C): FZYM_4 vs FZYM_2 (D): FZZZ_2 vs CK, (E): FZZZ_4 vs CK; (F): FZZZ_4 vs FZZZ_2; (G): FZZZ_2 vs FZYM_2; (H): FZZZ_4 vs FZYM_4. BPAD, Biosynthesis of phenylacetic acid derivatives; ALAM, alpha−Linolenic acid metabolism; FFB, Biosynthesis of kaempferol aglycones I; BKAI, Biosynthesis of kaempferol aglycones; BPCS, Biosynthesis of p−coumaroyl sugars; SDGB, Stilbenoid, diarylheptanoid and gingerol biosynthesis; BUTT, Biosynthesis of ursane−type triterpenoids; BBD, Biosynthesis of benzaldehyde derivatives; BCAD, Biosynthesis of cinnamic acid derivatives; BSAD, Biosynthesis of salicylic acid derivatives; LAB, Luteolin aglycones biosynthesis; GDM, Glyoxylate and dicarboxylate metabolism; BCD, Biosynthesis of catechol derivatives; BOPCAD, Biosynthesis of other p−coumaric acid derivatives; BVAD, Biosynthesis of vanillic acid derivatives; BCIII, Biosynthesis of coumarins III; GSTM, Glycine, serine and threonine metabolism; UOTQB, Ubiquinone and other terpenoid−quinone biosynthesis; PHST, Plant hormone signal transduction; BBAD, Biosynthesis of benzoic acid derivatives; BVPSM, Biosynthesis of various plant secondary metabolites; VLID, Valine, leucine and isoleucine degradation; GPI-AB, Glycosylphosphatidylinositol (GPI)−anchor biosynthesis; 2-OAM, 2−Oxocarboxylic acid metabolism; PGI, Pentose and glucuronate interconversions; AAM, Ascorbate and aldarate metabolism; ASNSM, Amino sugar and nucleotide sugar metabolism; BNS, Biosynthesis of nucleotide sugars; BTAD, Biosynthesis of tartaric acid derivatives; IPM, Inositol phosphate metabolism; PTTB, Phenylalanine, tyrosine and tryptophan biosynthesis; BLAII, Biosynthesis of kaempferol aglycones II; CMM, Cysteine and methionine metabolism; BSM, Biosynthesis of secondary metabolites; PPM, Phosphonate and phosphinate metabolism; BFII, Biosynthesis of flavanols II; NKGB, Neomycin, kanamycin and gentamicin biosynthesis; BAI, Biosynthesis of anthocyanins I; APM, Arginine and proline metabolism; BFAD, Biosynthesis of ferulic acid derivatives; BVA, Biosynthesis of various alkaloids; CSWB, Cutin, suberine and wax biosynthesis; BCII, Biosynthesis of coumarins II; AAM, Arachidonic acid metabolism; CTM, Cobalamin transport and metabolism. KEGG annotation and enrichment analysis of differentially abundant metabolites in each pairwise comparison of broccoli sprout samples. (A): FZYM_2 vs CK, (B): FZYM_4 vs CK, (C): FZYM_4 vs FZYM_2 (D): FZZZ_2 vs CK, (E): FZZZ_4 vs CK; (F): FZZZ_4 vs FZZZ_2; (G): FZZZ_2 vs FZYM_2; (H): FZZZ_4 vs FZYM_4. BPAD, Biosynthesis of phenylacetic acid derivatives; ALAM, alpha−Linolenic acid metabolism; FFB, Biosynthesis of kaempferol aglycones I; BKAI, Biosynthesis of kaempferol aglycones; BPCS, Biosynthesis of p−coumaroyl sugars; SDGB, Stilbenoid, diarylheptanoid and gingerol biosynthesis; BUTT, Biosynthesis of ursane−type triterpenoids; BBD, Biosynthesis of benzaldehyde derivatives; BCAD, Biosynthesis of cinnamic acid derivatives; BSAD, Biosynthesis of salicylic acid derivatives; LAB, Luteolin aglycones biosynthesis; GDM, Glyoxylate and dicarboxylate metabolism; BCD, Biosynthesis of catechol derivatives; BOPCAD, Biosynthesis of other p−coumaric acid derivatives; BVAD, Biosynthesis of vanillic acid derivatives; BCIII, Biosynthesis of coumarins III; GSTM, Glycine, serine and threonine metabolism; UOTQB, Ubiquinone and other terpenoid−quinone biosynthesis; PHST, Plant hormone signal transduction; BBAD, Biosynthesis of benzoic acid derivatives; BVPSM, Biosynthesis of various plant secondary metabolites; VLID, Valine, leucine and isoleucine degradation; GPI-AB, Glycosylphosphatidylinositol (GPI)−anchor biosynthesis; 2-OAM, 2−Oxocarboxylic acid metabolism; PGI, Pentose and glucuronate interconversions; AAM, Ascorbate and aldarate metabolism; ASNSM, Amino sugar and nucleotide sugar metabolism; BNS, Biosynthesis of nucleotide sugars; BTAD, Biosynthesis of tartaric acid derivatives; IPM, Inositol phosphate metabolism; PTTB, Phenylalanine, tyrosine and tryptophan biosynthesis; BLAII, Biosynthesis of kaempferol aglycones II; CMM, Cysteine and methionine metabolism; BSM, Biosynthesis of secondary metabolites; PPM, Phosphonate and phosphinate metabolism; BFII, Biosynthesis of flavanols II; NKGB, Neomycin, kanamycin and gentamicin biosynthesis; BAI, Biosynthesis of anthocyanins I; APM, Arginine and proline metabolism; BFAD, Biosynthesis of ferulic acid derivatives; BVA, Biosynthesis of various alkaloids; CSWB, Cutin, suberine and wax biosynthesis; BCII, Biosynthesis of coumarins II; AAM, Arachidonic acid metabolism; CTM, Cobalamin transport and metabolism. In contrast, FZZZ showed a more pronounced enrichment pattern involving nucleotide metabolism, membrane lipid and carbohydrate metabolism, and hormone-related pathways, consistent with developmental reconstruction during germination and potentially providing a metabolic context more conducive to GSL accumulation. In FZZZ_2 vs. CK, differential metabolites were significantly enriched in the metabolism of purines, nucleotides, and glycerolipids, the biosynthesis of zeatin and p-coumaroyl sugars, and pentose and glucuronate interconversions (Fig. 4D, p < 0.05). In FZZZ_4 vs. CK, the significantly enriched pathways included nucleotide metabolism; cysteine, methionine, and glycerolipid metabolism; and multiple flavonoid-related pathways, including the biosynthesis of flavones, flavonols, and kaempferol-related compounds (Fig. 4E, p < 0.05). For the dose comparison (FZZZ_4 vs. FZZZ_2), differential metabolites were significantly enriched in the metabolism of linoleic acid, alpha-linolenic acid, and purines, and the biosynthesis of flavones, flavonols, and cinnamic acid derivatives (Fig. 4F, p < 0.05). Importantly, the enrichment of cysteine and methionine metabolism in FZZZ_4 vs. CK was highly relevant to the observed increase in GSLs in FZZZ treatments. GSLs are sulfur-containing defense metabolites, and their accumulation is often supported by sulfur assimilation and sulfur amino acid metabolism (Chan et al., 2019). Therefore, the involvement of cysteine- and methionine-related metabolism during germination after seed irradiation may increase precursor availability and provide a metabolic context for subsequent GSL accumulation. This interpretation is consistent with the higher GSL contents in FZZZ_2 and FZZZ_4 than in CK. Comparisons between developmental stages at the same irradiation dose further indicate that irradiation timing is a key determinant of pathway prioritization. In FZZZ_2 vs. FZYM_2, differential metabolites were significantly enriched in the metabolism of purines, nucleotides, linoleic acid, and alpha-linolenic acid, and phenylalanine/tyrosine/tryptophan biosynthesis (Fig. 4G, p < 0.05). In FZZZ_4 vs. FZYM_4, differential metabolites were mainly enriched in the metabolism of nucleotides, purines, and pyrimidines, as well as pathways related to anthocyanins and p-coumaric acid derivatives (Fig. 4H, p < 0.05). Collectively, these results demonstrate that irradiation timing (seed vs. sprout) can reshape enrichment patterns across nucleotide metabolism, fatty acid metabolism, and polyphenol-related pathways, and potentially shift the final output of GSL-related defense metabolism. Differential abundance analysis and its correspondence with GSL changes The differential abundance (DA) score is a pathway-level metric used to summarize the direction of metabolic changes across a pathway: DA > 0 indicates an overall tendency toward upregulation of pathway-associated metabolites, whereas DA < 0 represents an overall tendency toward downregulation. For FZYM, DA analysis consistently suggested that irradiation prioritized lipid-associated stress responses and selected secondary metabolic processes while suppressing specific branches related to phenylacetic acid derivatives and p-coumaroyl sugars.In FZYM_2 vs. CK, positive DA scores were observed for linoleic acid and alpha-linolenic acid metabolism and for several secondary metabolic pathways, including the biosynthesis of stilbenoids, diarylheptanoids, gingerols, ursane-type triterpenoids, and coumarins V, indicating greater pathway activity in FZYM_2 (Fig. 5A). In contrast, the biosynthesis of phenylacetic acid derivatives, p-coumaroyl sugars, and salicylic acid derivatives, as well as lipoic acid metabolism, showed DA < 0, which indicated relatively higher activity in the CK. Similarly, in FZYM_4 vs. CK, there was clear upregulation (DA > 0) of phenylalanine metabolism, oxidative phosphorylation, and the biosynthesis of ursane-type triterpenoids and fatty acid pathways, whereas downregulation (DA < 0) was observed in the biosynthesis of catechol derivatives, p-coumaroyl sugars, phenylacetic acid derivatives, and certain coumarin-related branches (Fig. 5B). Taken together, both doses shared a consistent pattern at the sprout stage. Irradiation promoted unsaturated fatty acid metabolism and parts of aromatic amino acid and phenylpropanoid derivative metabolism, and suppressed the biosynthesis of p-coumaroyl sugars and phenylacetic acid derivatives. The increase in unsaturated fatty acid metabolism is consistent with the well-documented roles of lipid unsaturation in maintaining membrane function during abiotic stress and in lipid-associated signaling (He and Ding, 2020). Because ionizing radiation can trigger ROS production and oxidative stress signaling in plants, strengthened lipid remodeling can be interpreted as a stress-responsive adjustment that supports membrane remodeling and related energy-demanding processes (Kim, 2025). Meanwhile, aromatic amino acids serve as precursors for the phenylpropanoid network, and phenylpropanoid-derived metabolites are widely implicated in plant stress tolerance and defense, supporting the view that activation of this module is associated with coordinated metabolic reprogramming toward selected defense-related metabolites (Ninkuu et al., 2025). In contrast, the reduced p-coumaroyl sugar signals are compatible with a downshift in hydroxycinnamate glycosylation pools since hydroxycinnamate glucose esters (including 4 coumaroyl glucose) are common plant conjugates and can be stress-responsive through UDP glycosyltransferase-mediated regulation (Babst et al., 2014). Finally, phenylacetic acid is a natural auxin, and its derivatives and conjugates are integral to phenylacetic acid homeostasis; therefore, suppressed biosynthesis of phenylacetic acid derivatives may indicate irradiation-associated perturbation of hormone-related conjugation metabolism (Perez et al., 2023).Fig. 5Differential abundance (DA) score of differentially abundant metabolites of each pairwise comparison of sprout. (A): FZYM_2 vs CK, (B): FZYM_4 vs CK, (C): FZYM_4 vs FZYM_2 (D): FZZZ_2 vs CK, (E): FZZZ_4 vs CK; (F): FZZZ_4 vs FZZZ_2; (G): FZZZ_2 vs FZYM_2; (H): FZZZ_4 vs FZYM_4. BPAD, Biosynthesis of phenylacetic acid derivatives; ALAM, alpha−Linolenic acid metabolism; FFB, Biosynthesis of kaempferol aglycones I; BKAI, Biosynthesis of kaempferol aglycones; BPCS, Biosynthesis of p−coumaroyl sugars; SDGB, Stilbenoid, diarylheptanoid and gingerol biosynthesis; BUTT, Biosynthesis of ursane−type triterpenoids; BBD, Biosynthesis of benzaldehyde derivatives; BCAD, Biosynthesis of cinnamic acid derivatives; BSAD, Biosynthesis of salicylic acid derivatives; LAB, Luteolin aglycones biosynthesis; GDM, Glyoxylate and dicarboxylate metabolism; BCD, Biosynthesis of catechol derivatives; BOPCAD, Biosynthesis of other p−coumaric acid derivatives; BVAD, Biosynthesis of vanillic acid derivatives; BCIII, Biosynthesis of coumarins III; GSTM, Glycine, serine and threonine metabolism; UOTQB, Ubiquinone and other terpenoid−quinone biosynthesis; PHST, Plant hormone signal transduction; BBAD, Biosynthesis of benzoic acid derivatives; BVPSM, Biosynthesis of various plant secondary metabolites; VLID, Valine, leucine and isoleucine degradation; GPI-AB, Glycosylphosphatidylinositol (GPI)−anchor biosynthesis; 2-OAM, 2−Oxocarboxylic acid metabolism; PGI, Pentose and glucuronate interconversions; AAM, Ascorbate and aldarate metabolism; ASNSM, Amino sugar and nucleotide sugar metabolism; BCAD, Biosynthesis of caffeic acid derivatives; BNS, Biosynthesis of nucleotide sugars; BTAD, Biosynthesis of tartaric acid derivatives; IPM, Inositol phosphate metabolism; PTTB, Phenylalanine, tyrosine and tryptophan biosynthesis; BLAII, Biosynthesis of kaempferol aglycones II; BKAI, Biosynthesis of kaempferol aglycones I; BKAII, Biosynthesis of kaempferol aglycones II; CMM, Cysteine and methionine metabolism; BSM, Biosynthesis of secondary metabolites; PPM, Phosphonate and phosphinate metabolism; BFII, Biosynthesis of flavanols II; NKGB, Neomycin, kanamycin and gentamicin biosynthesis; APM, Arginine and proline metabolism; BSAD, Biosynthesis of sinapic acid derivatives; FFB, Flavone and flavonol biosynthesis; BFAD, Biosynthesis of ferulic acid derivatives; BVAD, Biosynthesis of various alkaloids; CSWB, Cutin, suberine and wax biosynthesis; ACM, Arachidonic acid metabolism; CTM, Cobalamin transport and metabolism. Differential abundance (DA) score of differentially abundant metabolites of each pairwise comparison of sprout. (A): FZYM_2 vs CK, (B): FZYM_4 vs CK, (C): FZYM_4 vs FZYM_2 (D): FZZZ_2 vs CK, (E): FZZZ_4 vs CK; (F): FZZZ_4 vs FZZZ_2; (G): FZZZ_2 vs FZYM_2; (H): FZZZ_4 vs FZYM_4. BPAD, Biosynthesis of phenylacetic acid derivatives; ALAM, alpha−Linolenic acid metabolism; FFB, Biosynthesis of kaempferol aglycones I; BKAI, Biosynthesis of kaempferol aglycones; BPCS, Biosynthesis of p−coumaroyl sugars; SDGB, Stilbenoid, diarylheptanoid and gingerol biosynthesis; BUTT, Biosynthesis of ursane−type triterpenoids; BBD, Biosynthesis of benzaldehyde derivatives; BCAD, Biosynthesis of cinnamic acid derivatives; BSAD, Biosynthesis of salicylic acid derivatives; LAB, Luteolin aglycones biosynthesis; GDM, Glyoxylate and dicarboxylate metabolism; BCD, Biosynthesis of catechol derivatives; BOPCAD, Biosynthesis of other p−coumaric acid derivatives; BVAD, Biosynthesis of vanillic acid derivatives; BCIII, Biosynthesis of coumarins III; GSTM, Glycine, serine and threonine metabolism; UOTQB, Ubiquinone and other terpenoid−quinone biosynthesis; PHST, Plant hormone signal transduction; BBAD, Biosynthesis of benzoic acid derivatives; BVPSM, Biosynthesis of various plant secondary metabolites; VLID, Valine, leucine and isoleucine degradation; GPI-AB, Glycosylphosphatidylinositol (GPI)−anchor biosynthesis; 2-OAM, 2−Oxocarboxylic acid metabolism; PGI, Pentose and glucuronate interconversions; AAM, Ascorbate and aldarate metabolism; ASNSM, Amino sugar and nucleotide sugar metabolism; BCAD, Biosynthesis of caffeic acid derivatives; BNS, Biosynthesis of nucleotide sugars; BTAD, Biosynthesis of tartaric acid derivatives; IPM, Inositol phosphate metabolism; PTTB, Phenylalanine, tyrosine and tryptophan biosynthesis; BLAII, Biosynthesis of kaempferol aglycones II; BKAI, Biosynthesis of kaempferol aglycones I; BKAII, Biosynthesis of kaempferol aglycones II; CMM, Cysteine and methionine metabolism; BSM, Biosynthesis of secondary metabolites; PPM, Phosphonate and phosphinate metabolism; BFII, Biosynthesis of flavanols II; NKGB, Neomycin, kanamycin and gentamicin biosynthesis; APM, Arginine and proline metabolism; BSAD, Biosynthesis of sinapic acid derivatives; FFB, Flavone and flavonol biosynthesis; BFAD, Biosynthesis of ferulic acid derivatives; BVAD, Biosynthesis of various alkaloids; CSWB, Cutin, suberine and wax biosynthesis; ACM, Arachidonic acid metabolism; CTM, Cobalamin transport and metabolism. The lower GSL contents observed in FZYM_2 and FZYM_4 than in CK may reflect metabolic prioritization. GSL biosynthesis is tightly connected to primary sulfur assimilation and sulfur amino acid metabolism, which supplies reduced sulfur and related precursors, and the production of GSLs is expected to impose measurable allocation costs within the broader metabolic network (Borpatragohain et al., 2016). Sprout-stage irradiation may alter the allocation of carbon, nitrogen, and sulfur resources and increase investment in repair and membrane-associated responses. This interpretation is consistent with the central roles of membrane lipid remodeling and lipid-linked signaling during abiotic stress and with the increased repair demands caused by ionizing-radiation-induced ROS signaling (Fig. 5C) (Yu et al., 2021). Under such conditions, the observed decrease in GSL accumulation may result from reduced investment in net GSL biosynthesis and/or increased GSL turnover or downstream conversion. GSL pools can be coordinately regulated according to nutrient availability and can undergo internal turnover under stress-like conditions to remobilize resources (Jeschke et al., 2019). For FZZZ, DA analysis revealed a distinct pathway pattern that reflected developmental reconstruction during germination and a different balance of metabolic allocation. In FZZZ_2 vs. CK, DA > 0 was observed in glycerolipid and inositol phosphate metabolism and the biosynthesis of caffeic acid derivatives, neomycin, kanamycin, and gentamicin, which indicated higher activity in FZZZ_2 (Fig. 5D). Conversely, DA < 0 was found in the metabolism of nucleotides, pyrimidines, ascorbate, and aldarate, as well as plant hormone signal transduction, and the biosynthesis of cinnamic acid derivatives, which indicated higher pathway activity in the CK. In FZZZ_4 vs. CK (Fig. 5E), positive DA scores were observed for flavonoid-related pathways, including the biosynthesis of kaempferol aglycones I and II and other flavonoids, as well as for neomycin, kanamycin, and gentamicin biosynthesis and glycerolipid metabolism. In contrast, negative DA scores were observed for nucleotide, amino sugar, nucleotide sugar, ascorbate, and aldarate metabolism and for the biosynthesis of p-coumaroyl sugars and nucleotide sugars, indicating greater activity in CK. Consistent tendencies were observed across both seed irradiation doses. Irradiation promoted glycerolipid and inositol phosphate metabolism, which are widely linked to membrane lipid remodeling and lipid-associated signaling during stress adaptation (Yu et al., 2021). In contrast, irradiation suppressed the metabolism of nucleotides, ascorbate, and aldarate, and p-coumaroyl sugar biosynthesis, which suggested that there was reduced investment in nucleotide supply and antioxidant-related processes during early metabolic reconstruction (Niehaus et al., 2022). In the dose comparison FZZZ_4 vs. FZZZ_2 shown in Fig. 5F, the higher dose showed DA > 0 for the metabolism of linoleic acid, alpha-linolenic acid, and several defense-related pathways such as isoflavonoid biosynthesis, whereas cobalamin transport and metabolism exhibited DA < 0, which indicated relatively higher activity in FZZZ_2. Linoleic acid and alpha-linolenic acid pathways are closely connected to oxylipin and jasmonate-related lipid signaling, which are widely implicated in stress responses and defense regulation; therefore, their stronger activation at 4 kGy is consistent with a more intensive stress-responsive remodeling state (Wasternack and Song, 2017). Isoflavonoids are also defense-associated flavonoid branch metabolites, which supports the interpretation that higher-dose germination engages broader defense chemistry (Wang et al., 2024). When integrated with the GSL phenotype, FZZZ-2 showed the highest GSL content, followed by FZZZ-4, and both exceeded CK. These DA patterns support the idea that 2 kGy may represent a more favorable balance between effective induction and metabolic cost. Low-dose irradiation is commonly discussed within radiation hormesis and stress priming frameworks, where moderate stress can establish a primed metabolic state without eliciting excessive damage responses (Kim, 2025). This balance is also consistent with the broader concept that defensive metabolism can impose resource and energy costs. Therefore, stronger and more widespread stress activation at higher doses may divert substrates and reducing power toward repair and membrane remodeling, thereby limiting further increases in specific defense-metabolite pools such as GSLs (Neilson et al., 2013). Finally, DA analysis of the stage comparisons shown in Fig. 5G (FZZZ_2 vs. FZYM_2) and 5H indicated that seed irradiation followed by germination more strongly upregulated pathways associated with carbohydrate supply, glycosylation, and membrane construction than did sprout irradiation. These pathways included amino sugar, nucleotide sugar, glycerolipid, and inositol phosphate metabolism and nucleotide sugar biosynthesis. This is consistent with the central roles of nucleotide sugars as sugar donors for glycosylation and complex carbohydrate production during growth processes, and with glycerolipid and inositol phosphate-related metabolism supporting membrane biogenesis and lipid-associated signaling under stress and development (Figueroa et al., 2021). Conversely, sprout irradiation maintained or enhanced alpha-linolenic acid metabolism and several nucleotide-, folate-, and ascorbate-related modules. These changes were compatible with stronger engagement of oxylipin- and jasmonate-linked lipid signaling and with the established antioxidant roles of ascorbate in ROS detoxification and redox signaling (Deleu et al., 2019). These stage-dependent DA patterns reinforce a coherent explanation for the opposite GSL responses. Under the low-dose conditions included in the metabolomic analysis, seed-stage irradiation was associated with a metabolic background that supported GSL accumulation during germination, whereas direct sprout-stage irradiation was associated with acute stress- and membrane-related responses and lower GSL concentrations (Yu et al., 2021). EBI-induced modulation of the methionine pathway and GSL production To elucidate the mechanisms by which EBI modulates GSL accumulation in broccoli sprouts, we integrated targeted metabolomics with pathway analysis and focused on cysteine and methionine metabolism (Fig. 6A) and the GSL biosynthetic pathway (Fig. 6B). The metabolic responses exhibited distinct patterns depending on the irradiation stage (FZZZ or FZYM).Fig. 6Heatmap of metabolites in the cysteine and methionine metabolism pathway (A), heatmap of metabolites in the glucosinolate biosynthesis pathway (B), and potential pathway of electron beam irradiation regulating glucosinolate biosynthesis and metabolism (C). Heatmap of metabolites in the cysteine and methionine metabolism pathway (A), heatmap of metabolites in the glucosinolate biosynthesis pathway (B), and potential pathway of electron beam irradiation regulating glucosinolate biosynthesis and metabolism (C). The above results suggest that FZZZ may function as a metabolic priming treatment (specifically at 2–4 kGy), which significantly enhances the concentration and total accumulation of GSLs. Heatmap analysis of the cysteine and methionine pathway (Fig. 6A) revealed a significant upregulation of upstream precursors in FZZZ-treated sprouts compared with CK, including L-serine, L-homoserine, and L-methionine. These amino acids are upstream substrates for sulfur assimilation and methionine-derived secondary metabolism. Their increased availability may expand the precursor pool for GSL biosynthesis and promote downstream conversion toward GSL formation (Deleu et al., 2019; Sønderby et al., 2010). Consistent with this pathway-level metabolomic pattern, independent free-amino-acid analysis showed significant dose-dependent increases in serine, cysteine, and methionine in sprouts derived from irradiated seeds. Relative to CK, methionine increased by 49.4% and 114.3% following the 2- and 4-kGy treatments, respectively, while serine and cysteine showed similar significant increasing trends (p < 0.05; Fig. 2C). These results independently corroborate the expansion of the sulfur-amino-acid and methionine-related precursor pools following seed irradiation. However, because total GSL accumulation was slightly greater at 2 kGy than at 4 kGy, precursor availability alone did not determine GSL output and should not be interpreted as direct evidence of increased biosynthetic flux. Within the GSL biosynthetic pathway (Fig. 6B), a marked enrichment was observed in key aliphatic intermediates, particularly dihomomethionine. Dihomomethionine is a central intermediate generated through successive methionine side-chain elongation reactions and directly serves as a precursor for major aliphatic GSLs (Sønderby et al., 2010). This enrichment is consistent with increased availability of methionine-derived intermediates and a potential adjustment of the side-chain elongation process; however, it does not by itself demonstrate increased metabolic flux.Major aliphatic GSLs, such as RAA and ERU, reached their highest measured levels at 4 kGy. This finding suggests that FZZZ may provide a favorable metabolic context for phytochemical synthesis during germination. In contrast, the low-dose FZYM treatments included in the metabolomic analysis (2 and 4 kGy) showed lower total GSL concentrations than CK and were associated with an abiotic stress response. Metabolomic profiling showed a substantial accumulation of S-adenosyl-L-homocysteine (SAH) and L-homocysteine in FZYM groups (Fig. 6A). SAH is a feedback inhibitor of S-adenosyl-L-methionine-dependent methyltransferases, and its accumulation is commonly associated with reduced methylation within the methionine cycle. In this context, the observed increase in SAH in FZYM-treated sprouts is consistent with a potential constraint on methylation-dependent steps involved in aliphatic GSL biosynthesis. In the low-dose FZYM treatments analyzed metabolomically, the concentrations of aliphatic GSLs such as RAA and ERU were lower than those in CK, potentially owing to altered biosynthesis, turnover, or stress-induced degradation. In contrast, indole GSLs such as 4 ME exhibited a distinct response. Under low-dose FZYM (2–4 kGy), the levels of 4 ME showed transient stability or even stress-induced increases relative to aliphatic counterparts. Indole GSLs are closely associated with plant defense signaling and are preferentially induced under abiotic and biotic stress conditions (Sønderby et al., 2010; Clay et al., 2009; Martínez-Ballesta et al., 2013). This shift toward indole GSLs suggests a preferential accumulation of defense-related metabolites under stress conditions associated with direct electron beam exposure (Fig. 6C). In summary, EBI influences GSL accumulation in a stage-dependent manner. FZZZ is associated with increased precursor availability and altered methionine-derived GSL metabolism, which may contribute to increased GSL accumulation. In contrast, FZYM is associated with changes in methionine cycle-related metabolism and a shift toward indole GSLs under stress conditions. These findings provide a mechanistic basis for the potential use of EBI in the biofortification of broccoli sprouts. Effects of irradiation on glucosinolate metabolism-related gene expression To provide transcriptional evidence for the metabolic responses to irradiation, the expression of 13 genes associated with glucosinolate metabolism and stress-related phenylpropanoid metabolism was determined by qRT-PCR (Fig. 7). These genes covered multiple stages of the glucosinolate pathway, including methionine side-chain elongation, represented by methylthioalkylmalate synthase 1 (MAM1); aldoxime formation and core-structure biosynthesis, represented by cytochrome P450 monooxygenase 79F1 (CYP79F1), cytochrome P450 monooxygenase 83A1 (CYP83A1), glutathione S-transferase phi 11 (GSTF11), glutathione S-transferase tau 20 (GSTU20), UDP-dependent glycosyltransferase 74C1 (UGT74C1), and desulfoglucosinolate sulfotransferase 18 (SOT18); secondary side-chain modification, represented by flavin monooxygenase glucosinolate S-oxygenase 2 (FMO-GS-OX2), flavin monooxygenase glucosinolate S-oxygenase 5 (FMO-GS-OX5), and 2-oxoglutarate-dependent dioxygenase AOP3 (AOP3); and glucosinolate hydrolysis, represented by myrosinase (MYR). Phenylalanine ammonia-lyase 1 (PAL1) and phenylalanine ammonia-lyase 2 (PAL2) were included as marker genes of phenylpropanoid metabolism and the general stress response. The functions of MAM1, CYP79F1, CYP83A1, and UGT74C1 in methionine chain elongation, aldoxime formation, oxime metabolism, and thiohydroximate glucosylation, respectively, have been experimentally demonstrated in Arabidopsis (Grubb et al., 2014; Kroymann et al., 2001; Naur et al., 2003; Olson-Manning et al., 2015). The corresponding genes involved in glucoraphanin biosynthesis, side-chain modification, and hydrolysis have also been characterized during broccoli germination (Liu et al., 2024).Fig. 7Effects of irradiation on the expression of glucosinolate-metabolism- and phenylpropanoid-related genes in CK, FZZZ_2, and FZZZ_4 broccoli sprouts. Panels (A-M) show the relative expression of (A) methylthioalkylmalate synthase 1 (MAM1), (B) cytochrome P450 monooxygenase 79F1 (CYP79F1), (C) cytochrome P450 monooxygenase 83A1 (CYP83A1), (D) glutathione S-transferase phi 11 (GSTF11), (E) glutathione S-transferase tau 20 (GSTU20), (F) UDP-dependent glycosyltransferase 74C1 (UGT74C1), (G) desulfoglucosinolate sulfotransferase 18 (SOT18), (H) flavin monooxygenase glucosinolate S-oxygenase 2 (FMO-GS-OX2), (I) flavin monooxygenase glucosinolate S-oxygenase 5 (FMO-GS-OX5), (J) 2-oxoglutarate-dependent dioxygenase AOP3 (AOP3), (K) myrosinase (MYR), (L) phenylalanine ammonia-lyase 1 (PAL1), and (M) phenylalanine ammonia-lyase 2 (PAL2) in the CK, 2 kGy, and 4 kGy groups. Bars represent means ± SD, and open circles indicate individual replicate measurements (n = 6). Different lowercase letters indicate significant differences among treatments, as determined by one-way ANOVA of log2-transformed relative expression followed by Tukey's multiple-comparison test (adjusted P < 0.05). Panel (N) presents a heatmap of the mean log2 fold changes in FZZZ-2 and FZZZ-4 relative to CK. Blue and orange indicate decreased and increased expression, respectively. Asterisks indicate significant differences relative to CK after Tukey adjustment: *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. Effects of irradiation on the expression of glucosinolate-metabolism- and phenylpropanoid-related genes in CK, FZZZ_2, and FZZZ_4 broccoli sprouts. Panels (A-M) show the relative expression of (A) methylthioalkylmalate synthase 1 (MAM1), (B) cytochrome P450 monooxygenase 79F1 (CYP79F1), (C) cytochrome P450 monooxygenase 83A1 (CYP83A1), (D) glutathione S-transferase phi 11 (GSTF11), (E) glutathione S-transferase tau 20 (GSTU20), (F) UDP-dependent glycosyltransferase 74C1 (UGT74C1), (G) desulfoglucosinolate sulfotransferase 18 (SOT18), (H) flavin monooxygenase glucosinolate S-oxygenase 2 (FMO-GS-OX2), (I) flavin monooxygenase glucosinolate S-oxygenase 5 (FMO-GS-OX5), (J) 2-oxoglutarate-dependent dioxygenase AOP3 (AOP3), (K) myrosinase (MYR), (L) phenylalanine ammonia-lyase 1 (PAL1), and (M) phenylalanine ammonia-lyase 2 (PAL2) in the CK, 2 kGy, and 4 kGy groups. Bars represent means ± SD, and open circles indicate individual replicate measurements (n = 6). Different lowercase letters indicate significant differences among treatments, as determined by one-way ANOVA of log2-transformed relative expression followed by Tukey's multiple-comparison test (adjusted P < 0.05). Panel (N) presents a heatmap of the mean log2 fold changes in FZZZ-2 and FZZZ-4 relative to CK. Blue and orange indicate decreased and increased expression, respectively. Asterisks indicate significant differences relative to CK after Tukey adjustment: *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. Compared with CK, the 2 kGy treatment significantly increased the expression of MYR, PAL1, and PAL2, whereas no significant changes were detected in the upstream glucosinolate biosynthetic genes MAM1, CYP79F1, or CYP83A1. In contrast, the 4 kGy treatment induced a broader transcriptional response, with significant increases in UGT74C1, GSTF11, FMO-GS-OX2, FMO-GS-OX5, MYR, PAL1, and PAL2. Moreover, the expression levels of GSTF11, FMO-GS-OX2, MYR, PAL1, and PAL2 were significantly higher under 4 kGy than under 2 kGy. No significant treatment effects were observed for MAM1, GSTU20, CYP83A1, CYP79F1, or AOP3. Although the overall treatment effect for SOT18 was significant, none of the pairwise comparisons reached significance in the Tukey test. When integrated with the metabolomic results, the qRT-PCR data provide transcriptional support for selective remodeling of GSL metabolism, but do not indicate coordinated activation of the entire biosynthetic pathway. MAM1 catalyzes the initial condensation reaction in methionine side-chain elongation (Kroymann et al., 2001), whereas CYP79F1 converts chain-elongated methionine derivatives into aldoximes and has been identified as a major flux-controlling step in aliphatic glucosinolate biosynthesis (Olson-Manning et al., 2015). CYP83A1 subsequently metabolizes the aliphatic aldoximes during formation of the glucosinolate core structure (Naur et al., 2003). The absence of significant induction of these three genes, particularly under 2 kGy, therefore does not support a direct conclusion that irradiation promoted methionine side-chain elongation or activated the entire de novo glucosinolate biosynthetic pathway. Nevertheless, the total glucosinolate concentration increased significantly from 3174.58 μg g−1 FW in CK to 6556.85 and 6216.78 μg g−1 FW under 2 and 4 kGy, respectively. Importantly, although the mean glucosinolate concentration was numerically higher under 2 kGy than under 4 kGy, the two irradiation treatments were not significantly different. The high glucosinolate content under 2 kGy, in the absence of significant induction of MAM1, CYP79F1, and CYP83A1, suggests that glucosinolate accumulation cannot be explained solely by increased transcription of the measured biosynthetic genes. It may additionally involve transient expression responses occurring before the sampling point, post-transcriptional regulation, changes in enzyme activity, substrate availability, transport, storage, or altered rates of glucosinolate turnover. Temporal uncoupling between glucosinolate-related gene expression and glucosinolate accumulation has also been observed during broccoli germination (Liu et al., 2024). The broader transcriptional response under 4 kGy was mainly associated with downstream core-structure formation and side-chain modification. In particular, induction of UGT74C1 is consistent with enhanced glucosylation of thiohydroximate intermediates (Grubb et al., 2014), whereas the increased expression of FMO-GS-OX2 and FMO-GS-OX5 suggests enhanced S-oxygenation and remodeling of aliphatic glucosinolate side chains (Liu et al., 2024). However, this stronger transcriptional response did not produce a higher total glucosinolate concentration than 2 kGy. This apparent discrepancy indicates that the expression of downstream modification genes may influence glucosinolate composition rather than necessarily increasing the total glucosinolate pool. In addition, MYR, which is associated with glucosinolate hydrolysis, was significantly induced by both irradiation doses and showed its highest expression under 4 kGy. This response may indicate an increased capacity for glucosinolate turnover or formation of hydrolysis products, although enzyme-activity and hydrolysis-product measurements would be required to confirm this possibility. The marked induction of PAL1 and PAL2, particularly under 4 kGy, is consistent with activation of a general irradiation-induced stress response because PAL catalyzes the entry step of phenylpropanoid biosynthesis and PAL genes are responsive to abiotic environmental stimuli (Olsen et al., 2008). Collectively, these results indicate that irradiation modulated multiple components of glucosinolate metabolism and activated phenylpropanoid-associated stress responses. The 2 kGy treatment achieved the numerically highest glucosinolate level without coordinated induction of the upstream marker genes, whereas 4 kGy produced a stronger downstream modification, hydrolysis-related, and stress-responsive transcriptional signature without further increasing total glucosinolate accumulation. Therefore, the qRT-PCR data support irradiation-induced regulation and remodeling of glucosinolate metabolism but do not, by themselves, demonstrate enhanced metabolic flux, activation of sulfur assimilation, or inhibition of methyltransferases by SAH. Confirmation of these mechanisms would require enzyme-activity assays, sulfur-assimilation marker genes, hydrolysis-product analysis, or isotope-based metabolic-flux measurements. Implications for practical applications and risk–benefit considerations Beyond mechanistic insights, the present findings have implications for the practical application of EBI in broccoli sprout production. In particular, seed priming emerges as a technically attractive intervention point, as it requires only a single treatment prior to germination, minimizes interference with subsequent cultivation processes, and is inherently compatible with standardized, factory-based sprout production systems. Furthermore, the consistently greater GSL accumulation observed at low irradiation doses suggests that seed priming may improve the batch-to-batch uniformity of functional quality, a key consideration in industrial sprout production. At the same time, potential risks associated with irradiation-based interventions must be carefully considered. In this study, high irradiation doses markedly impaired seed germination performance, which underscored a narrow margin between effective stimulation and detrimental outcomes at the seed stage. Although low-dose treatment promoted the accumulation of GSL, EBI induced oxidative processes in biological matrices (Diehl, 2002), which may influence other quality attributes, such as lipid oxidation or redox-sensitive metabolites (Brewer, 2009), particularly under higher doses or suboptimal processing conditions. Nevertheless, the nutritional responses observed in the present study did not indicate depletion of the selected nutrients at the tested seed-irradiation doses. Compared with the non-irradiated control, vitamin C, soluble sugar, and total amino acid contents were significantly increased in sprouts derived from seeds treated at 2 and 4 kGy. These increases may be associated with irradiation-induced metabolic responses during seed germination, although the underlying mechanisms require further investigation. Because only selected nutritional components were determined and radiolytic products were not directly analyzed, these results should not be interpreted as a comprehensive assessment of the chemical or toxicological safety of the irradiated sprouts. In addition, regulatory frameworks governing the use of ionizing irradiation in food systems and consumer acceptance of irradiated products vary across regions and may represent non-technical constraints on broader implementation. Based on the combined evaluation of GSL enrichment, germination performance, and metabolic responses, an irradiation dose of 2 kGy appears to represent a favorable operational window for broccoli seed priming, providing substantial functional enhancement with limited adverse effects. A dose of 4 kGy may serve as a secondary option; however, its application requires careful assessment of germination robustness and overall product quality. Together, these considerations highlight the importance of balancing functional benefits against biological and regulatory risks when translating irradiation-based strategies from experimental systems to industrial practice. Conclusions In conclusion, this study demonstrates that low-dose electron beam seed priming (2 kGy) serves as a powerful non-thermal processing strategy to significantly enhance glucosinolate (GSL) accumulation in broccoli sprouts. This pre-germination pretreatment triggers targeted metabolic reprogramming without altering initial seed GSL reserves. In contrast, direct irradiation of 5-day-old sprouts decreased GSL content at low dose, indicating that the timing of irradiation is critical for the targeted regulation of glucosinolate profiles. In addition, compared with non-irradiated control sprouts, sprouts grown from seeds irradiated at 2 and 4 kGy contained higher levels of vitamin C, soluble sugars, total free amino acids, and sulfur-containing volatile compounds.Metabolomic analysis showed that seed-stage irradiation was associated with changes in sulfur amino acid precursor availability and methionine- and phenylpropanoid-related metabolism, whereas direct sprout irradiation mainly induced stress-associated changes in lipid and nucleotide metabolism. The qRT-PCR results further indicated selective regulation of genes involved in downstream GSL modification, hydrolysis, and phenylpropanoid-associated stress responses rather than coordinated activation of the entire de novo GSL biosynthetic pathway. Overall, seed-stage EBI at 2 kGy represents a promising strategy for enhancing GSL accumulation and selected nutritional attributes in broccoli sprouts. Nevertheless, further studies on enzyme activities, metabolic flux, radiolytic products, toxicological safety, and process scalability are required before industrial application. CRediT authorship contribution statement Juan Wang: Investigation, Methodology, Data curation, Software, Writing–original draft, Writing–review & editing. Yaqin Wang: Investigation, Methodology, Writing–review & editing. Zhaoyu Li: Data curation, Software, Writing–original draft. Xuezhi Zhao: Methodology. Xirui Zhao: Methodology, Data curation. Hongju He: Methodology, Project administration, Resources. Guangmin Liu: Investigation, Methodology, Writing–review & editing, Funding acquisition. Funding This research was funded by the 10.13039/501100007934Innovation and Capacity-Building Project by the Beijing Academy of Agriculture and Forestry Sciences (KJCX20240402, KJCX20240331) and Beijing Innovation Consortium of Agriculture Research System (BAIC01-2026). Declaration of competing interests 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.