The experimental site was located in Chahar Right Front Banner, Ulanqab City, Inner Mongolia Autonomous Region, China (40.92° N, 113.11° E). The region has a mean annual temperature of 9°C, an average frost-free period of approximately 120 days, and a mean annual precipitation of 360.8 mm. Precipitation, air temperature, and mean wind speed during the 2024 and 2025 growing seasons are shown in Figure 1. The soil is predominantly Calcisol, and the baseline soil fertility of the 0–20 cm layer is presented in Table 1. Field experiments were conducted in two consecutive years at different fields within the same area.

A split-plot design was employed, with oat cultivar assigned to the main plots and silicon input assigned to the sub-plots. Two cultivars with contrasting lodging resistance were selected: the lodging-resistant Mengyan 1 (MY1) and the lodging-susceptible Dingyan 2 (DY2). MY1 was developed by the Inner Mongolia Academy of Agricultural and Animal Husbandry Sciences. It has a plant height of 105.5 cm, a panicle length of 16.3 cm, and a spreading panicle type. The number of grains per panicle averages 54.6, with a thousand-grain weight of 34.0 g. Its growth period is about 90 days, classifying it as a mid-maturing cultivar.DY2 was bred by the Dingxi Academy of Agricultural Sciences. It has a plant height of 128 cm, a panicle length of 24.0 cm, and a compact lateral panicle type. The grains per panicle average 83.7, with a thousand-grain weight of 27.1 g. Its growth period is approximately 115 days, making it a mid-to late-maturing cultivar. Sodium metasilicate (Na₂SiO₃, analytical grade ≥ 99.0%) was used as the silicon source at five input rates: 0, 30, 60, 90, and 120 kg ha^-1, each with four replicates. Sodium metasilicate was uniformly mixed with the compound fertilizer and applied as a basal application to the respective treatment plots. The compound fertilizer was applied at a rate of 150 kg ha^-1 (N: P₂O₅: K₂O = 18: 18: 18). Additionally, urea was top-dressed at 75 kg ha^-1 (Total N ≥ 46.0%) both before and after the jointing stage. Sowing was carried out on 15 May 2024 and 11 May 2025. The seeding rate was 180 kg ha^-1. Each plot measured 20 m² with a row spacing of 25 cm. Furrows were opened mechanically, followed by manual seed broadcasting and soil covering. Throughout the growing season, weeding and irrigation were performed according to standard local agronomic practices.

During the grain-filling and milk stages, a 50 cm representative stem segment was collected from each plot, avoiding border and adjacent rows. From these segments, 15 uniformly growing, non-lodged plants were selected for agronomic trait measurements, which included:

Second basal internode length (cm): The length of the stem between the second node from the base and the third node was accurately measured using a ruler (counting from the soil surface).

Panicle fresh weight (g): The entire panicle was excised at the base, and fresh weight was determined using an electronic balance.

Stem breaking strength (N): Measured using a YYD-1A portable plant stem strength meter (Zhejiang Top Cloud Instrument Co., Ltd.). Leaf sheaths were first removed, and the second basal internode was placed in the instrument’s groove with a support span of 5 cm. The flat compression probe was positioned at the midpoint of the second basal internode, and force was applied downward at a uniform speed until the stem fractured. The recorded reading represents the stem breaking strength. The lodging resistance index was calculated using the following formula (Wang et al., 2019):

During the grain-filling and milk stages, a 50 cm representative stem segment was selected from each plot, avoiding border and adjacent rows. The second basal internode was excised, leaf sheaths were removed, and samples were placed in an oven at 105°C for 30 min to inactivate enzymes, then dried at 80°C to constant weight. The dried samples were ground for determination of mineral and lignin content.

Lignin content: Acid-insoluble lignin (Klason lignin) was measured using the sulfuric acid method (Schwanninger and Hinterstoisser, 2002). Approximately 0.5 g of dried sample powder was accurately weighed into an acid-resistant container, and 15 mL of 72% sulfuric acid was added. The mixture was hydrolyzed in a 30°C water bath for 60 min. About 7.5 mL of distilled water was then added to dilute the acid to 4%, followed by autoclaving at 121°C for 60 min. After cooling to room temperature, the hydrolysate was filtered through a pre-weighed glass fiber filter. The insoluble residue was repeatedly washed with hot distilled water until the filtrate was neutral, then the filter and residue were dried at 105°C to constant weight. The dried residue was ashed in a muffle furnace at 600°C for 4h to remove ash, and the ash mass was recorded. Lignin content was calculated as:

Where m₀ is the sample weight (g), m₁ is the weight of acid-insoluble residue (g), and m₂ is the weight of the residue ash (g).

Lignin-biosynthetic enzymes: Sampling was initiated at 10 days after jointing in both cultivars and repeated at 10-day intervals for a total of five samplings. In each plot, a representative 50-cm stem segment was selected while avoiding border and near-border rows. The second basal internode was excised, leaf sheaths were removed, and samples were immediately stored in a dry-ice freezer. The activities of lignin-biosynthetic enzymes, including tyrosine ammonia-lyase (TAL), phenylalanine ammonia-lyase (PAL), 4-coumarate: CoA ligase (4CL), and cinnamyl alcohol dehydrogenase (CAD), were measured using ELISA kits.

Mineral content: Approximately 0.2 g of oat stem powder was weighed into a digestion tube, to which 5 mL of HNO₃ and 4 mL of H₂O₂ were added. Samples were digested for 30 min in a microwave digestion system (Mars6-Xpress, CEM Technology Co., Ltd., Matthews, NC, USA). The digestion tubes were then placed on a temperature-controlled apparatus (XMTG-7000, Gongsheng Instrument Co., Ltd., Yuyao, China) at 120°C until no yellow fumes were observed. The digested solution was transferred to a 50 mL volumetric flask and brought to volume with ultrapure water. Potassium, calcium, magnesium, and silicon concentrations were determined using an inductively coupled plasma optical emission spectrometer (ICP-OES, Agilent 5800, Agilent Technologies Inc., Santa Clara, CA, USA) (Barros et al., 2016b).

Data were compiled in Microsoft Excel 2019 prior to statistical analysis. Analysis of variance (ANOVA) together with Fisher’s Least Significant Difference (LSD) test (p < 0.05) was used to identify the effects of treatments on all the analyzed parameters. Prior to ANOVA, data were tested for normality using the Shapiro-Wilk test and for homogeneity of variance using Levene’s test (IBM SPSS Statistics, version 25.0). The random forest model was constructed using the “randomForest” package in R (v4.3.3) under the R Studio environment (v4.2.3). Variable importance and its statistical significance were assessed via permutation tests implemented in the rfPermute package, while the overall significance of the model was evaluated using the A3 package. Graphs were created using Origin 2024.

The lodging resistance index (LRI) of oat stems at the grain-filling stage was significantly affected by year, cultivar, and their interaction (Table 2, p < 0.001). In 2024, the LRI of Mengyan 1 (MY1) and Dingyan 2 (DY2) were 0.6 and 0.5, respectively, and decreased by 6.1% and 21.0% in 2025 (Figure 2A). Compared with MY1, the LRI of DY2 was 12.6% and 26.5% lower in 2024 and 2025, respectively (p < 0.01 and p < 0.001). Across both experimental years, the LRI at the grain-filling and milk stages was significantly influenced by the interaction between cultivar and silicon input (Table 2, p < 0.001). The LRI of MY1 reached its maximum at a silicon input of 60 kg ha^-1, whereas the optimal input for DY2 was 90 kg ha^-1. On a two-year average, relative to the no-silicon control (CK), the LRI of MY1 increased by 29.7% at the grain-filling stage and by 43.7% at the milk stage, whereas that of DY2 increased by 42.9% and 74.5%, respectively (Figure 2, p < 0.05).

Across both experimental years, the stem breaking strength (SBS) of the second basal internode was significantly influenced by cultivar, silicon input rate, and their interaction (Table 2, p < 0.001). The SBS of DY2 was 19.4% and 13.4% lower than that of MY1 at the grain-filling and milk stages, respectively (Figure 3, p < 0.001). Maximum SBS for MY1 was achieved at a silicon input rate of 60 kg ha^-1, while for DY2 it occurred at 90 kg ha^-1. On a two-year average, relative to the no-silicon control (CK), SBS of MY1 increased by 40.7% at the grain-filling stage and by 52.7% at the milk stage, whereas that of DY2 increased by 51.3% and 42.7%, respectively (p < 0.05).

The lignin content of the second basal internode in oat stems was significantly affected by year, cultivar, and their interaction (Table 2, p < 0.001). In 2024, the lignin contents of MY1 and DY2 were 16.1% and 14.0%, respectively, and decreased by 12.4% and 15.1% in 2025 (Figure 4A). Compared with MY1, the lignin content of DY2 was 12.7% and 15.3% lower in 2024 and 2025, respectively (p < 0.001). Across both experimental years, stem lignin content was also significantly influenced by cultivar, silicon input, and their interaction (Table 2, p < 0.001). The lignin content of MY1 reached its maximum at a silicon input of 60 kg ha^-1, whereas the optimal input for DY2 was 90 kg ha^-1. On a two-year average, relative to the no-silicon control (CK), lignin content of MY1 increased by 12.5% at the grain-filling stage and by 14.6% at the milk stage, whereas that of DY2 increased by 12.4% and 17.0%, respectively (Figure 4, p < 0.05).

The activities of lignin-biosynthetic enzymes in the second basal internode of oat stems were significantly affected by silicon input (Table 2, p < 0.001). In 2024, enzyme activities in both MY1 and DY2 followed a unimodal pattern, peaking at 30 days after jointing (DAJ). In 2025, MY1 exhibited a bimodal pattern with the highest peak at 40 DAJ, whereas DY2 remained unimodal, peaking at 20 DAJ (Figure 5). The activities of TAL, PAL, 4CL, and CAD in the second basal internode peaked at a silicon input of 60 kg ha^-1 for MY1, whereas for DY2, the optimal silicon input was 90 kg ha^-1. On a two-year average, relative to the no-silicon, the activities of TAL, PAL, 4CL, and CAD in the second basal internode increased by 15.6%, 30.6%, 20.4%, and 17.6% for MY1, and by 21.2%, 25.9%, 22.0%, and 25.9% for DY2, respectively.

At the grain-filling stage, the K, Ca, and Si contents in the second basal internode of oat stems were significantly affected by cultivar, silicon input, and their interaction (Table 2, p < 0.001). Compared with MY1, the K, Ca, and Si contents of DY2 stems were 13.0%, 11.6%, and 6.0% lower in 2024 and 8.0%, 13.4%, and 5.6% lower in 2025, respectively (Table 3). The K, Ca, and Si contents in MY1 stems were highest at a silicon input of 60 kg ha^-1, whereas the optimal rate for DY2 was 90 kg ha^-1. On a two-year average, relative to the no-silicon, these treatments increased K, Ca, and Si contents by 25.5%, 26.1%, and 25.7% in MY1, and by 22.4%, 35.1%, and 40.3% in DY2, respectively (p < 0.05).

At the milk stage, the K and Si contents in the second basal internode were significantly influenced by cultivar, silicon input, and their interaction (Table 2, p < 0.001). Compared with MY1, the K and Si contents of DY2 stems were 38.6% and 7.7% lower in 2024 and 4.0% and 6.5% lower in 2025, respectively (Table 3). The K and Si contents in MY1 stems were highest at a silicon input of 60 kg ha^-1, whereas the optimal rate for DY2 was 90 kg ha^-1. On a two-year average, relative to the no-silicon, these treatments increased K and Si contents by 34.5% and 16.1% in MY1, and by 32.1% and 19.7% in DY2, respectively (p < 0.05).

Nine indicators were identified associated with lignin-biosynthetic at the grain-filling stage based on a random forest model (R² = 0.85), ranked by importance as follows: CAD activity at 30 DAJ, 4CL activity at 30 DAJ, TAL activity at 10 DAJ, Mg content at grain-filling stage, TAL activity at 30 DAJ, Ca content at grain-filling stage, K content at grain-filling stage, PAL activity at 20 DAJ, and PAL activity at 30 DAJ (Figure 6A). In contrast, eleven key indicators were identified at the milk stage (R² = 0.91), ranked by importance as: CAD activity at 30 DAJ, TAL activity at 30 DAJ, 4CL activity at 40 DAJ, 4CL activity at 30 DAJ, TAL activity at 50 DAJ, PAL activity at 40 DAJ, 4CL activity at 50 DAJ, PAL activity at 30 DAJ, Mg content at grain-filling stage, 4CL activity at 10 DAJ, and PAL activity at 50 DAJ (Figure 6B). These results indicated that lignin-biosynthetic enzyme activity at 30 days after jointing represents a critical regulatory period for lignin synthesis, with CAD activity exerting the dominant influence on stem lignin accumulation at both the grain-filling and milk stages, lignin-biosynthetic enzyme activities contributed more to lignin biosynthesis than mineral contents. The relative contributions of the indicators indicate that lignin-biosynthetic enzyme activities have a greater influence on lignin biosynthesis than mineral contents.

The basal stem is the primary region that supports plant weight and resists external mechanical forces, and the mechanical strength of the second basal internode is considered a key determinant of lodging resistance in oat (Yu et al., 2025). Previous studies have shown that appropriate silicon input can improve the morphological structure of the basal stem and enhance stem mechanical strength, thereby increasing lodging resistance (Jiang et al., 2025; Wu et al., 2025; Yang et al., 2024). This study found that silicon input significantly affected the lodging resistance index and stem breaking strength of the second basal internode in oat (Table 2, p < 0.001), and the lodging resistance index was closely correlated with stem breaking strength during both the grain-filling stage (R² = 0.76) and the milk stage (R² = 0.87), consistent with previous reports (Figure 7A). Moreover, the optimal silicon input differed between cultivars: the lodging-resistant cultivar performed best at 60 kg·ha^-1, whereas the lodging-susceptible cultivar performed better at 90 kg·ha^-1 (Figures 2, 3, p < 0.05). These differences likely reflect cultivar-specific variations in silicon uptake, transport, and utilization, which are primarily associated with the activity of silicon transporters and the expression levels of their corresponding genes (Manivannan and Ahn, 2017; Mandlik et al., 2020). Notably, regardless of silicon input level, the lodging resistance index and stem breaking strength of the lodging-resistant cultivar Mengyan 1 were consistently higher than those of the lodging-susceptible cultivar Dingyan 2, indicating that oat lodging resistance is strongly influenced by genetic characteristics (Table 2, p < 0.001). In addition, environmental factors are key determinants of crop lodging (Mohammadi et al., 2020). Year had a significant effect on the stem lodging resistance index (Table 2, p < 0.001). compared with 2024, the lodging resistance index of both cultivars decreased overall in 2025, likely due to extreme weather events, such as persistent heavy rainfall, occurring during the critical grain-filling period from July to August, which increased the incidence of lodging (Figure 1).

Lodging in crops principally stems from insufficient stem mechanical strength, and lignin localized in the cell walls of mechanical tissues substantially increases cell-wall rigidity, thereby enhancing stem mechanical strength and lodging resistance (Liu et al., 2018b). Lignin content in the second basal internode is positively correlated with lodging resistance index and mechanical strength (Liu et al., 2024b; Luo et al., 2019). This study found that silicon input significantly increased lignin content in the second basal internode of oat stems, and lignin content was closely correlated with the lodging resistance index (Figure 7B, R² = 0.80 at the grain-filling stage, R² = 0.64 at the milk stage) and stem breaking strength (Figure 7C, R² = 0.52 at the grain-filling stage, R² = 0.63 at the milk stage).Studies in maize (Zea mays L.) and rice (Oryza sativa L.) have also shown that silicon input can increase stem lignin content and enhance stem mechanical strength, thereby reducing the risk of lodging (Shah et al., 2023; Jiang et al., 2025). Appropriate silicon input promotes integration of lignin oligomers, whereas excessive silicon may enhance oligomeric repulsion and inhibit further lignin polymerization (Djikanović et al., 2024), which aligns with the unimodal response of lignin content to increasing silicon observed in both cultivars. Moreover, lignin content in the stems of the lodging-resistant cultivar was higher than that in the lodging-susceptible cultivar, indicating that its greater lodging resistance may be attributed to lignin accumulation (Zhang et al., 2022).

Lignin biosynthesis is chiefly regulated by the activities of lignin-biosynthetic enzymes. Tyrosine ammonia-lyase, phenylalanine ammonia-lyase, 4-coumarate: CoA ligase, and cinnamyl alcohol dehydrogenase are central enzymes in the lignin-biosynthetic pathway in oat stems, and increased activities of these enzymes promote lignin accumulation, thereby enhancing stem mechanical strength and lodging resistance (Zheng et al., 2017; Wang et al., 2022). Lignin biosynthetic enzymes activities were significantly affected by silicon input (Table 2, p < 0.001), and under the optimal silicon input, both lignin content and lignin biosynthetic enzymes activities reached their highest levels in the two oat cultivars (Figure 4 and 5). Previous studies indicate that silicon enhances stem lodging resistance by upregulating the expression of lignin-biosynthetic genes, increasing the activities of corresponding enzymes, and thereby promoting lignin deposition to reinforce the cell-wall structure (Xue et al., 2023; Jiang et al., 2025; Shah et al., 2023). Moreover, analysis using a random forest model indicated that lignin biosynthetic enzymes activities at 30 days after jointing represent a critical regulatory period for lignin biosynthesis, and cinnamyl alcohol dehydrogenase activity plays a dominant role in stem lignin accumulation during both the grain-filling and milk stages (Figure 6). Previous studies indicate that the regulation of lignin biosynthesis by lignin biosynthetic enzymes is stage-dependent (Wang et al., 2022). In oat stems, internodes shift after jointing from longitudinal elongation to secondary cell wall thickening and mechanical tissue lignification, requiring increased lignin biosynthetic enzymes activity to promote lignin accumulation.

Mineral elements play a key regulatory role in crop growth and development as well as in the formation of stem mechanical strength (Liu et al., 2024a). Silicon input significantly promotes the uptake and utilization of K, Ca, and Si in crops (Kupdhoni et al., 2025), which is consistent with the results of the present study. Minerals in stems contribute to lodging resistance through two main pathways. First, they activate genes involved in lignin biosynthesis and enhance the activities of lignin-biosynthetic enzymes, thereby promoting lignin accumulation. Second, they are deposited in cell walls or intercellular spaces through processes such as calcification and silicification, which directly strengthen stem mechanical properties (Huang et al., 2023; Ma et al., 2025; Wu et al., 2023). This study showed that stem Mg content responded weakly to silicon input, with a significant effect only in Dingyan 2 in 2025 (Table 2, p< 0.05), possibly because silicon promotes Mg uptake only under Mg deficiency (Buchelt et al., 2020). However, the random forest model identified Mg as a key indicator, suggesting that its role in lignin synthesis may not depend on accumulation, but may indirectly promote lignin synthesis in a catalytic manner. Across the two-year trial, stem K content in both cultivars was significantly higher in 2025 than in 2024, which may be attributable to intensified lodging in 2025, leading to the redistribution and accumulation of K due to its high mobility (Mostofa et al., 2022). Excessive K can inhibit the activity of lignin-biosynthetic enzymes through ion competition or signal regulation, thereby reducing lignin accumulation (Wu et al., 2023), which is consistent with the significant negative correlation observed in this study between stem K content and lignin-biosynthetic enzyme activity at 30 days after jointing (Figure 8). Overall, silicon input promoted mineral accumulation in oat stems, and Ca, Mg, and Si were significantly and positively correlated with the activities of lignin-biosynthetic enzymes, indicating that mineral elements contribute, at least in part, to the activation of these enzymes and thereby enhance lignin biosynthesis. However, the specific physiological mechanisms by which individual minerals regulate lignin biosynthesis and lodging resistance remain to be further systematically elucidated.