Treatment-induced metabolic shifts of soybean seeds through amino acid–zinc biofortification enhances oil quality and micronutrient density

Abstract

Background

Agronomic biofortification offers a practical approach to enhancing the nutritional and functional quality of soybeans (Glycine max L.); however, the combined metabolic effects of zinc (Zn) and amino acids (AAs) on seed composition remain poorly understood. This two-year field study assessed 16 AA × Zn foliar treatment combinations using an integrated analytical approach that includes mineral enrichment, oil content and yield, detailed fatty acid compositional data, lipid-quality indices, multivariate statistics, network modeling, and multi-criteria decision analysis.

Results

Combined application of amino acids and zinc significantly increased key nutrients—while maintaining stable levels of calcium and magnesium. Oil content increased by up to 15% and extracted oil mass increased by up to 13% with moderate-input treatments. Fatty acid profiles were consistently altered, with reductions in stearic and oleic acids and notable increases in linoleic (+ 5.9%) and eicosenoic acids (+ 9%), enhancing the unsaturated-to-saturated ratio without harming oxidative stability. Multivariate and compositional analyses revealed strong, reproducible proportional relationships among fatty acids across years, driven by coordinated Zn- and AA-dependent regulation of desaturase pathways. Lipid-health indices showed significant nutritional enhancements, with high-AA × moderate-Zn combinations producing the most cardio-protective lipid profiles. Network analysis identified oleic acid, linoleic acid, zinc, and nitrogen as central hubs linking nutrient status to lipid metabolism, while multi-criteria rankings showed that treatments using moderate-to-high doses of amino acids combined with moderate-to-high levels of zinc were generally among the best-performing options across ranking methods.

Conclusions

Overall, coordinated foliar application of amino acids and zinc offers a synergistic, environmentally friendly strategy to boost soybean nutrient density, improve oil yield, and promote a healthier lipid profile. The findings support amino acid–zinc biofortification as an efficient and eco-friendly approach for nutrition-focused soybean production.

Graphical Abstract

Supplementary Information

The online version contains supplementary material available at 10.1186/s12870-026-08751-z.

Introduction

Soybeans (Glycine max L.) play a key role in global food security, as one of the few crops that can provide high-quality protein, health-promoting oil, and valuable micronutrients for both human and livestock diets [1]. Its multifunctionality has become increasingly crucial as the world faces two significant nutritional challenges: persistent “hidden hunger,” primarily due to inadequate intake of micronutrients such as zinc (Zn) [2], and the growing demand for plant-based oils with enhanced cardiovascular and metabolic benefits [3]. Although breeding efforts have enhanced soybean yield potential, progress in increasing seed mineral density and improving oil quality has been relatively slow, leading to a renewed interest in agronomic biofortification as a flexible and cost-effective approach [4].

Zinc deficiency impacts more than three billion people and worsens in regions where cereal-based diets and high phytate intake limit Zn absorption [5]. Increasing the natural Zn content of staple foods—rather than relying only on supplements or food processing—offers a scalable and nutrition-sensitive solution [6]. Soybeans are considered a promising vehicle for Zn biofortification due to their high global consumption and favorable seed physiology; however, their ability to absorb and distribute Zn is particularly limited in calcareous, alkaline, or Zn-depleted soils [7]. Foliar Zn application bypasses many soil-related issues and improves phloem mobility, but its effectiveness heavily depends on leaf penetration, chelation, metabolic demand, and translocation capacity [8]. Meanwhile, amino acids (AAs) and protein hydrolysates have emerged as advanced biostimulants that can improve nutrient uptake and plant [9]. Amino acids serve not only as building blocks for proteins and signaling compounds but also as chelators that stabilize metal ions, influence membrane permeability, and activate cellular antioxidant defenses [10]. The foliar application of amino acids promotes nitrogen uptake and chlorophyll synthesis, while also enhancing plants’ resistance to abiotic stress [11]. These treatments maintain elevated stomatal conductance and photosystem II efficiency, hence enhancing photosynthetic resilience and biomass accumulation under water-scarce situations [12]. Additionally at the metabolic pathway level, Amino acid sprays markedly enhance the activities of nitrate reductase and glutamine synthetase at the metabolic level, thereby enhancing nitrogen usage efficiency [13]. However, the mechanistic interaction between AAs and micronutrients like Zn—especially whether their combined use results in synergistic effects on nutrient loading, lipid metabolism, and seed quality—remains inadequately studied [14].

Soybean oil quality depends on the proportions of saturated (SFA), monounsaturated (MUFA), and polyunsaturated fatty acids (PUFA) [15]. Changes in the oleic–linoleic–linolenic pathway, controlled by desaturase enzymes and the cellular redox state, significantly affect oxidative stability, nutritional value, and industrial uses [16]. Small metabolic shifts—such as increased zinc availability or enhanced AA-driven nitrogen assimilation—may alter carbon partitioning and desaturase activity, reshaping entire lipid biosynthesis pathways [17, 18]. However, few studies have investigated how foliar zinc and amino acid (AA) interactions impact the multivariate structure of soybean fatty acids or lipid health markers, such as iodine value, atherogenicity index, thrombogenicity index, or the health-promoting index [19].

Given the identified gaps, a comprehensive systems-level evaluation that integrates mineral composition, oil content, fatty-acid profiles, and multivariate statistical methods is urgently needed. There is a significant lack of agronomic studies that simultaneously consider compositional data analysis (CoDA), lipid quality indices, trait-correlation networks, and multi-criteria decision analysis (MCDM). Such an approach is crucial for identifying treatment combinations that not only increase seed Zn concentration but also improve lipid nutritional quality, boost oil yield, and maintain metabolic balance. In this study, we examined the effects of combining foliar applications of zinc and amino acids on nutrient levels, oil yield, fatty acid profiles, and lipid health indices in soybeans across two consecutive field seasons. By combining traditional nutrient measurements with compositional modeling, PCA, network analysis, and decision-making tools, the work provides a detailed understanding of how coordinated Zn × AA management alters soybean seed biochemistry. The central hypothesis is that moderate zinc application, when paired with optimized amino acid inputs, creates synergistic metabolic effects that: (1) increase macro- and micronutrient content, (2) shift fatty acid biosynthesis toward healthier unsaturated forms, and (3) preserve oil quality across seasons through improved nutrient–lipid interactions. By integrating agronomic biofortification with targeted biostimulation, this research presented a novel framework for producing nutritionally enriched, metabolically stable, and functionally superior soybean oil, thereby supporting both global nutritional objectives and sustainable crop management practices.

Materials and methods

Plant material and experimental site

Field experiments was performed during the 2022 and 2023 growing seasons at Akdeniz University, Antalya, Türkiye (36°53′54.4″N, 30°38′28.9″E). The soybean (Glycine max [L.] Merr.) cultivar ‘Sonya’ was chosen for its consistent performance and desirable seed quality traits. The seeds for Sonya cultivar were provided by Progen Seed Company (https://www.progenseed.com) in Türkiye. Pre-sowing soil samples (0–20 cm) were analyzed for physicochemical properties following standard protocols. The soil was identified as loamy, containing 1.23% organic matter, 20.9% CaCO₃, a pH of 7.7, and baseline micronutrient levels of 0.82 mg kg⁻¹ Zn, 5.51 mg kg⁻¹ Mn, and 0.88 mg kg⁻¹ Cu. Climatic data for both years (temperature, rainfall, humidity) were collected from the university meteorological station and summarized in Supplementary Figure S1.

Experimental design and treatments

A randomized complete block design (RCBD) with three replications was used, consisting of 16 treatment combinations of amino acid (AA) and zinc (Zn). The liquid amino acid products comprised total amino acids at 46.70%, total organic matter at 51.35%, organic nitrogen at 5.42%, organic carbon at 24.38%, and zinc in the form of zinc sulfate (ZnSO₄·7 H₂O). The chemical composition of the commercial amino acid individual amino acids per 100 g: 1.45 g glycine, 1.65 g proline, 0.78 g hydroxyproline, 0.72 g arginine, 0.56 g valine, 0.45 g methionine, 0.42 g glutamate, 0.37 g leucine, 0.34 g isoleucine, 0.28 g lysine, 0.25 g alanine, 0.23 g phenylalanine, 0.14 g serine, 0.13 g tyrosine, 0.12 g aspartate, 0.11 g histidine, 0.09 g asparagine, 0.06 g threonine, 0.04 g cystine, and 0.04 g tryptophan. Amino acids were applied at four concentrations (0, 50, 100, and 150 mL 100 L⁻¹; designated as AA₁–AA₄) and zinc at four rates (0, 2, 4, and 6 mg L⁻¹; Zn₁–Zn₄) using a 16 L mechanical sprayer. The treatments were arranged in four vertical lines, with each combination replicated in three blocks per year, resulting in 96 experimental units in two growing season. Sowing occurred in mid-May at 70 cm inter-row and 4–5 cm intra-row spacing, with each plot consisting of five plants.

Treatment application

All amino acids were diluted 100-fold with water and thereafter sprayed over the plant leaves each time. The control plants (AA1 and Zn1) were irrigated with tap water. Foliar treatments were performed on two occasions—during the three-leaf stage (BBCH 13) and the flowering stage (BBCH 60–62). Control plots (AA₁ × Zn₁) were irrigated alone with water.

Sampling and nutrient analysis

At the harvest, three plants randomly selected from each plot, and their seeds were analyzed for macronutrient (N, P, K, Ca, Mg) and micronutrient (Zn, Mn, Cu) concentrations. The Kjeldahl technique was used to ascertain nitrogen concentration. Phosphorus analysis included digesting samples in concentrated H₂SO₄-H₂O₂, with amounts quantified colorimetrically by the molybdate-blue technique [20]. Concentrations of K, Ca, Mg, Zn, Mn, and Cu were quantified using inductively coupled plasma optical emission spectroscopy (ICP-OES; Perkin Elmer Optima DV7000).

Oil extraction and fatty acid analysis

For each biological replicate, oil was extracted from a composite sample consisting of 20 mature soybean seeds collected from each plot. The seeds were dried, ground, and subjected to Soxhlet extraction using petroleum ether (40–60 °C). The extracted oil was evaporated at 70 °C to remove solvent residues. Fatty acid methyl esters (FAMEs) were prepared following Garcés and Mancha [21] and analyzed by GC–MS/FID (Shimadzu GC-MS-QP2020NX, SH-Stabilwax column, 60 m × 0.25 mm × 0.25 μm; injector 250 °C, split 40:1). Fatty acids were identified by the Wiley W9N11 library and quantified as a percentage of total oil.

Fatty acid dataset and compositional analysis

Nine major fatty acids—palmitic (16:0), stearic (18:0), oleic (18:1), linoleic (18:2 n-6), linolenic (18:3 n-3), arachidic (20:0), eicosenoic (20:1), behenic (22:0), and lignoceric (24:0)—were quantified for each treatment. Because fatty acid data are compositional, all values were expressed as parts of a unit sum and analyzed using log-ratio transformations [22]. Zeros, when present, were replaced by half the minimum non-zero value, and closure normalization ensured Σx_i = 1. Centered log-ratio (clr) and isometric log-ratio (ilr) transformations were used for multivariate analysis.

Compositional statistics

clr-transformed data were subjected to principal component analysis (PCA) to visualize treatment effects, and the Aitchison variation matrix was used to assess pairwise proportionality between fatty acids. A multivariate analysis of variance (MANOVA) tested the impact of Year, AA, Zn, and their interactions on overall fatty acid composition using Pillai’s trace. All analyses were performed in Python (pandas, numpy, scipy, statsmodels), and visualizations generated with matplotlib and seaborn.

Lipid quality indices

The nutritional and functional quality of soybean oil was evaluated using a comprehensive set of lipid health indices derived from treatment-specific fatty acid profiles for each year of the study. Total saturated (SFA), monounsaturated (MUFA), and polyunsaturated (PUFA) fatty acids were first measured, followed by the calculation of physiologically relevant ratios, including UFA/SFA, PUFA/SFA, and the essential fatty acid balance n-6/n-3, which serve as indicators of oil fluidity, metabolic behavior, and dietary value. Desaturation efficiency was evaluated using the stearoyl-CoA desaturase index (SCD18 = C18:1 / (C18:0 + C18:1)) [23], indicating the enzymatic transformation of stearic acid into oleic acid. To assess the oxidative and nutritional characteristics of the oil, composite metrics were calculated: the iodine value (IV) [24] and unsaturation index (UI) [25] to quantify total double-bond content and lipid fluidity; the atherogenicity index (AI) and thrombogenicity index (TI) [26] to evaluate cardiovascular risk based on the ratio of pro- and anti-atherogenic fatty acids; the health-promoting index (HPI) [27] as a comprehensive indicator of lipid quality; and the peroxidizability index (PI) [28], which assesses the oil’s susceptibility to oxidative degradation based on PUFA structure. All indices were determined for each treatment combination (AA × Zn) in both years using standard equations applied to the percentage of individual fatty acids. The stability of these metrics over time was assessed through cross-year comparisons to determine whether AA–Zn supplementation consistently improved the nutritional quality of soybean oil across different environmental conditions.

Biofortification and nutrient indices

Nutrient enrichment was evaluated using the enrichment factor (EF = treatment/control) and the response ratio (RR = nutrient yield/control yield). A nutrient density score (NDS = [(HPI × UFA/SFA)/SFA] normalized 0–100) was calculated. Percent recommended dietary allowance (%RDA) of linoleic (18:2) and linolenic (18:3) acids per 30 g seed serving was based on FAO/WHO adequate intake values. Nutrient yield per area (NYA) was derived as concentration × yield (kg ha⁻¹).

Multi-criteria decision analysis (MCDM)

To rank treatments, six methods—TOPSIS, VIKOR, PROMETHEE II, ELECTRE III, WASPAS, and MOORA—were applied independently for each year and the pooled dataset. Ranks were aggregated using Borda count, and concordance among methods was assessed by Kendall’s W and Spearman’s ρ. Sensitivity analyses were conducted to assess the stability of rankings under varying weighting schemes.

Statistical analysis

The dataset comprised 96 observations (16 treatments × 3 replicates × 2 years). Factorial ANOVA (Year × AA × Zn) was performed using SAS 9.4 and verified in R 4.3. Tukey’s HSD (p < 0.05) was applied for mean separation. Results are reported as mean ± SD (n = 3) with significant groups indicated by letters. Percentage changes were calculated relative to the control within each year. Figures were prepared in R and Python for publication.

Results

Macronutrient composition

Across both experimental years, the combined foliar application of amino acids (AA) and zinc (Zn) significantly influenced the macronutrient composition of soybean seeds. Across both years, AA, Zn, and the interaction of AA×Zn substantially influenced seed macronutrient content (Table S1, Table S2). The annual means are shown in Fig. 1A (Year 1) and Fig. 1B (Year 2). Amino acids were applied at four concentrations—AA₁ = 0 mL 100 L⁻¹ (control), AA₂ = 50 mL 100 L⁻¹, AA₃ = 100 mL 100 L⁻¹, and AA₄ = 150 mL 100 L⁻¹—while zinc was supplied as ZnSO₄·7 H₂O at four levels: Zn₁ = 0 mg L⁻¹ (control), Zn₂ = 2 mg L⁻¹, Zn₃ = 4 mg L⁻¹, and Zn₄ = 6 mg L⁻¹. Nitrogen concentration was primarily enhanced by AA supplementation, with Zn alone showing minor influence. In Year 1, the control (AA₁×Zn₁) contained 5.98% N; Zn applied without AAs (AA₁×Zn₂–Zn₄) caused mixed responses (− 6.2% to + 3.0%), whereas AA addition markedly increased N.

Fig. 1.

Integrated heatmap of nutrient, oil, and fatty-acid traits in soybean seeds under different amino acid × zinc (AA × Zn) treatments during 2 years. Each column represents a single biochemical or nutritional trait, and each row corresponds to a specific treatment combination (AA1–AA4 × Zn1–Zn4). Individual color gradients are scaled independently for each trait (min–max normalization) to highlight within-trait variability, while the numeric values inside cells denote actual measured means. Greener or warmer tones indicate relatively higher intensities within that trait. The heatmap illustrates that AA₄×Zn₄ and AA₃×Zn₃ produced the highest overall concentrations of macronutrients and micronutrients, with concurrent improvements in oil yield and quality parameters

Moderate AA levels (AA₂ and AA₃) raised N to 6.40–6.96% (+ 7–16%), and the highest rate (AA₄) achieved 7.53% under AA₄×Zn₄ (+ 25.9% vs. control). In Year 2, similar trends were observed, with AA₄ combinations producing the most significant gains ( ~ + 21%). Phosphorus (P) showed year-dependent variation: in Year 1, control seeds (AA₁×Zn₁) had 1.40% P, which increased to 2.02% (+ 44.6%) under AA₁×Zn₃ (0 mL AA, 4 mg L⁻¹ Zn). Moderate AA + Zn treatments (AA₂–AA₄ with Zn₁–Zn₂) maintained high P levels (≈ 1.97–2.01%, + 41–44%), but excessive combinations, such as AA₄ × Zn₄, reduced P slightly (− 10.6%). In Year 2 (control = 1.45%), Zn alone produced small gains (+ 9.4%, AA₁×Zn₃), and AA₂×Zn₁ (50 mL AA, 0 mg L⁻¹ Zn) gave the highest P (1.71%, + 18.2%). Potassium (K) exhibited the most significant variability: in Year 1 (control = 0.50%), values ranged from − 19.9% (AA₂×Zn₁) to + 57% (AA₄×Zn₄—the most significant improvement across all nutrients and years), whereas in Year 2 (control = 0.39%), most treatments were within ± 6%, except for reductions at AA₄×Zn₃ and AA₄×Zn₄ (− 17.1% and − 23.1%). Calcium (Ca) also responded strongly in Year 1: the control (AA₁×Zn₁ = 0.39%) increased by + 34.2% with AA₁×Zn₂ (0 mL AA, 2 mg L⁻¹ Zn), while AA₄×Zn₂ (150 mL AA, 2 mg L⁻¹ Zn) gave the maximum value (0.56%, + 44.4%). In Year 2 (control = 0.45%), effects were more minor (+ 9% at AA₄×Zn₂, − 14.9% at AA₄×Zn₄). Magnesium (Mg) showed moderate shifts: in Year 1 (control = 0.22%), increases of 7–21% were recorded at AA₂×Zn₁ (50 mL AA, 0 mg L⁻¹ Zn) and AA₄×Zn₂ (150 mL AA, 2 mg L⁻¹ Zn), both reaching 0.27%, whereas high Zn slightly reduced Mg. In Year 2 (control = 0.20%), minor increases (+ 6–7%) occurred under AA₁×Zn₃ (0 mL AA, 4 mg L⁻¹ Zn) and AA₃×Zn₃ (100 mL AA, 4 mg L⁻¹ Zn), while substantial decreases (− 26.7% and − 28.3%) appeared at AA₄×Zn₃ and AA₄×Zn₄. Overall, amino acid supplementation was the main driver of nutrient enhancement, with moderate doses (AA₂ = 50 mL 100 L⁻¹ and AA₃ = 100 mL 100 L⁻¹) providing the most consistent improvements. In contrast, excessive combinations (AA₄×Zn₃–Zn₄ = 150 mL AA with 4–6 mg L⁻¹ Zn) occasionally induced antagonistic or inhibitory effects, possibly due to ionic competition or metabolic saturation. Although the magnitude of responses was slightly lower in Year 2, the overall pattern of treatment influence remained consistent across both seasons.

Micronutrient composition

Across both experimental years, the micronutrient composition of soybean seeds responded markedly to zinc (Zn) and amino acid (AA) supplementation. Across both years, AA, Zn, and the interaction of AA×Zn substantially influenced seed micronutrient content (Table S1, Table S2). The annual means are shown in Fig. 1A (Year 1) and Fig. 1B (Year 2). Amino acids were applied at four rates—AA₁ = 0 mL 100 L⁻¹ (control), AA₂ = 50 mL 100 L⁻¹, AA₃ = 100 mL 100 L⁻¹, and AA₄ = 150 mL 100 L⁻¹—while Zn was supplied as ZnSO₄·7 H₂O at Zn₁ = 0 mg L⁻¹ (control), Zn₂ = 2 mg L⁻¹, Zn₃ = 4 mg L⁻¹, and Zn₄ = 6 mg L⁻¹. Among micronutrients, seed Zn concentration exhibited the most pronounced and consistent increases, following a clear dose-dependent trend. In Year 1, control plants (AA₁×Zn₁; 0 mL AA, 0 mg L⁻¹ Zn) contained 64.27 mg kg⁻¹ Zn. Zinc application alone (AA₁×Zn₂–Zn₄; 0 mL AA, 2–6 mg L⁻¹ Zn) nearly doubled this value, reaching 126.09 mg kg⁻¹ under AA₁×Zn₄ (+ 96.2%). The inclusion of AAs further enhanced Zn accumulation: AA₂ (50 mL AA) increased Zn to 133.18 mg kg⁻¹ (+ 107.2%), AA₃ (100 mL AA) to 143.09 mg kg⁻¹ (+ 122.7%), and AA₄ (150 mL AA) peaked at 148.00 mg kg⁻¹ (+ 130.3%). In Year 2, the pattern persisted—control seeds (AA₁×Zn₁; 67.33 mg kg⁻¹) increased to 129.91 mg kg⁻¹ (+ 92.9%) under Zn alone and reached 151.53 mg kg⁻¹ (+ 125%) under AA₄×Zn₄ (150 mL AA, 6 mg L⁻¹ Zn), confirming the robust synergistic effect between AAs and Zn in enhancing Zn bioaccumulation. Manganese (Mn) also responded positively, particularly to AA supplementation. In Year 1, control seeds (AA₁×Zn₁) contained 56.13 mg kg⁻¹ Mn. Zinc alone (AA₁×Zn₂–Zn₄) produced modest increases (+ 12–16%), while AAs caused substantial improvements: AA₂ (50 mL AA) elevated Mn to 68.33–74.53 mg kg⁻¹ (+ 21.7–32.8%), AA₃ (100 mL AA) peaked at 85.73 mg kg⁻¹ (+ 52.7%), and AA₄ (150 mL AA) exceeded 80 mg kg⁻¹ across Zn treatments, with AA₄×Zn₃ (6 mg L⁻¹ Zn) reaching 86.13 mg kg⁻¹ (+ 53.4%). In Year 2, the control (AA₁×Zn₁; 61.73 mg kg⁻¹) showed more variable responses: Zn alone (AA₁×Zn₂–Zn₄) ranged from − 3.8% to + 19.0%, while AA₂ (50 mL AA) and AA₃ (100 mL AA) generally improved Mn content (+ 30.6% and + 13.7%, respectively). AA₄ (150 mL AA) maintained strong enhancements (up to 80.78 mg kg⁻¹, + 30.9%), although high-input treatments such as AA₄×Zn₄ (150 mL AA, 6 mg L⁻¹ Zn) slightly reduced Mn (− 5.1%), suggesting antagonistic effects at excessive foliar concentrations. Copper (Cu) displayed an inverse trend: Zn alone typically reduced or only slightly altered Cu, whereas AAs consistently increased it. In Year 1, the control (AA₁×Zn₁) averaged 24.87 mg kg⁻¹ Cu, while Zn alone (AA₁×Zn₂–Zn₄) reduced Cu by 2–11% (lowest at AA₁×Zn₂, − 10.5%). In contrast, AA supplementation significantly improved Cu levels—AA₂ (50 mL AA) raised it to 29.80–37.20 mg kg⁻¹ (+ 19.8–49.6%), AA₃ (100 mL AA) to 29.67–33.63 mg kg⁻¹ (+ 19–35%), and AA₄ (150 mL AA) peaked at 31.20 mg kg⁻¹ (+ 25.5%). In Year 2, control seeds (AA₁×Zn₁) contained 20.07 mg kg⁻¹ Cu, and Zn alone (AA₁×Zn₂–Zn₄) caused minimal changes (− 1.3% to + 14.6%). AAs again enhanced Cu accumulation, though less dramatically than in Year 1: AA₂ reached 24.40 mg kg⁻¹ (+ 21.6%), AA₃ 25.40 mg kg⁻¹ (+ 26.6%), and AA₄ 23.73 mg kg⁻¹ (+ 18.3%). However, reductions reappeared at high combined doses—AA₄×Zn₃ (150 mL AA, 4 mg L⁻¹ Zn; −3.3%) and AA₄×Zn₄ (150 mL AA, 6 mg L⁻¹ Zn; −11.0%)—mirroring the antagonistic response observed for Mn. Collectively, these findings confirm that moderate amino acid supplementation (AA₂–AA₃ = 50–100 mL 100 L⁻¹) in combination with zinc fertilization (Zn₂–Zn₃ = 2–4 mg L⁻¹ Zn) optimally enhanced micronutrient enrichment in soybean seeds, while excessive inputs (AA₄×Zn₄ = 150 mL AA, 6 mg L⁻¹ Zn) tended to induce nutrient imbalance through competitive or inhibitory interactions.

Oil content and yield

Across both experimental years, soybean seed oil concentration and oil yield exhibited variable responses to foliar amino acid (AA) and zinc (Zn) applications, with moderate improvements under several treatment combinations but also reductions, particularly at higher AA × Zn levels (Fig. 1A and B & Table S1). In Year 1, the baseline oil content in control seeds was 19.79%. Zinc alone slightly enhanced oil concentration, peaking at 21.62% (+ 9.2%) under AA₁ × Zn₃. AA supplementation produced mixed outcomes: AA₂ resulted in a modest decline at Zn₂ (− 3.3%) but improved oil content to 22.75% (+ 14.9%) under Zn₄; AA₃ showed steady increases, with the maximum at 21.94% (+ 10.9%) under Zn₃; AA₄ enhanced oil content under Zn₂ (+ 11.7%) but reduced it slightly under Zn₄ (− 2.4%). In Year 2, with a control baseline of 19.80%, Zn application again raised oil content by ~ 10% under Zn₂ and Zn₃. AA effects were treatment-dependent, with AA₂ lowering oil under Zn₂ (− 3.1%) but producing the highest increase across both years at 22.89% (+ 15.6%) under Zn₄. AA₃ produced consistent increases (6.5–13.1%), peaking under Zn₃, while AA₄ improved oil under Zn₂ (+ 12.1%) but reduced it under Zn₄ (− 3.6%).

In terms of extracted oil mass, Year 1 control plants produced 0.29 g of seed oil. Zn alone had inconsistent effects: AA₁×Zn₃ improved yield to 0.31 g (+ 6.9%), while AA₁×Zn₂ and AA₁×Zn₄ reduced it by − 8.0% and − 10.3%, respectively. AA₂ improved yield at Zn₁ (+ 5.7%) and peaked at 0.33 g (+ 12.6%) under Zn₄, but other Zn levels caused reductions (− 6.9% to − 16.1%). AA₃ outcomes ranged from − 20.7% at Zn₂ to a slight gain at Zn₃ (+ 1.1%), while AA₄ produced limited benefits (up to + 3.4% under Zn₂) but a sharp reduction under Zn₄ (0.23 g, − 19.5%). In Year 2, the control yield was 0.28 g of seed oil. Zn supplementation under AA₁×Zn₃ gave the highest increase of the study at 0.34 g (+ 20%), although this effect was marginally non-significant, while most other Zn-only treatments reduced yields (− 9.4% to − 10.6%). AA₂ modestly improved yields at Zn₄ (0.30 g, + 7.1%), AA₃ peaked at 0.31 g (+ 8.2%) under Zn₃ but declined at other Zn levels, and AA₄ provided small gains at Zn₁ (+ 7.1%) and Zn₂ (+ 9.4%) but again reduced yield sharply under Zn₄ to 0.23 g (− 20%).

Major fatty acid composition

Across both experimental years, soybean seed fatty acid profiles responded in distinct and often contrasting ways to foliar amino acid (AA) and zinc (Zn) applications (Fig. 1A-B & Table S3). Palmitic acid was largely unaffected, while stearic and oleic acids were consistently reduced, and linoleic and linolenic acids were enhanced under most treatments. Palmitic acid levels remained stable, with Year 1 control seeds containing 10.16% and treatments ranging narrowly between 10.14 and 10.46%, showing only marginal shifts (+ 2.7% at most under AA₄×Zn₄ and minor declines under AA₃×Zn₁ and AA₄×Zn₂). In Year 2, the control was slightly higher (10.50%), and most treatments caused small decreases (− 0.5% to − 3.3%), except for AA₄×Zn₄, which produced a modest increase (+ 0.5%). This confirms that palmitic acid was insensitive primarily to AA × Zn management. Stearic acid, in contrast, showed consistent and pronounced declines across both years. In Year 1, the control averaged 4.88%, and Zn alone reduced levels by 9.8% to 13.9%, while AA supplementation sustained reductions of 7.3% to 14.3%, with the lowest value at AA₃×Zn₃ (4.19%). Year 2 followed the same trend, with a control of 4.89% and decreases ranging from − 6% to over − 30%. The sharpest decrease was at AA₄×Zn₂ (3.11%, − 36.4%), highlighting a substantial negative impact of both Zn and AAs on stearic acid. Oleic acid was also consistently reduced, though high AA treatments partially mitigated losses. In Year 1, the control contained 25.02%, which declined steadily under Zn alone to 21.94% (− 12.3%) at AA₁×Zn₄, while most AA combinations resulted in further decreases (− 8 to − 14%), except AA₄×Zn₂ (23.91%, − 4.4%) and AA₄×Zn₃ (24.47%, − 2.2%), which minimized reductions. Year 2 results mirrored these patterns: the control (25.05%) decreased to 21.50% (− 14.1%) under AA₃×Zn₁, but AA₄×Zn₂ and AA₄×Zn₃ again moderated declines to − 4.7% and − 2.2%, respectively, showing some alleviation under high-AA management. Conversely, linoleic acid consistently increased under AA and Zn across both years. In Year 1, the control averaged 52.87%, and Zn alone raised it progressively up to 55.96% (+ 5.8%) at Zn₄. AA supplementation produced similar improvements, with AA₂×Zn₁ and AA₂×Zn₂ reaching 2.00% and 1.97% ( ≈ + 5.6–5.7%), while AA₄×Zn₃ recorded the lowest increase (+ 1.6%) but remained above control. Year 2 (control = 52.85%) displayed almost identical trends, with Zn alone elevating levels to 56.03% (+ 6.0%) and AA₂×Zn₂ peaking at 55.89% (+ 5.7%). Linolenic acid also improved consistently, though to a moderate extent. In Year 1, the control averaged 5.79%, and Zn alone boosted it to 6.51% (+ 12.5%) at Zn₂, while AA₂ and AA₃ combinations further increased it, with AA₃×Zn₁ reaching 6.66% (+ 15.1%), the highest across both years. AA₄ treatments generally supported gains (up to + 11.5%), except AA₄×Zn₂, which showed only a negligible effect (+ 0.5%). In Year 2, the control again averaged 5.79%, and treatment patterns closely paralleled those of Year 1. Zn and AAs raised linolenic acid by 6–15%, peaking at 6.68% (+ 15.2%) under AA₃×Zn₁, while AA₄×Zn₂ remained the weakest response (+ 1.0%).

Minor fatty acid composition

Across both experimental years, minor long-chain saturated and monounsaturated fatty acids in soybean seeds were generally reduced by foliar amino acid (AA) and zinc (Zn) applications, with only modest increases observed for eicosenoic acid under specific conditions (Fig. 1A-B & Table S4). Arachidic acid (C20:0) consistently declined across all treatments. In Year 1, the control averaged 0.40%. Zn alone reduced levels by − 6.6% to − 14.0%, and AA supplementation further suppressed them, with AA₃×Zn₃ reaching the minimum at 0.34% (− 15.7%). Year 2 results mirrored this trend, with a control of 0.41% and reductions ranging from − 8.2% to − 14.8%, with the lowest value under AA₃×Zn₃ (0.35%). Behenic acid (C22:0) also decreased under most treatments. In Year 1, the control was 0.39%. While AA₁×Zn₂ slightly increased content (+ 1.7%), higher Zn and AA combinations reduced it by − 4% to − 8.5%, with AA₃×Zn₃ recording the lowest value (0.36%). Year 2 followed the same pattern, with control at 0.39% and most treatments reducing levels by 1% to 7.7%, with minima again observed at AA₃×Zn₃ and AA₄×Zn₂ (0.36%, − 7.7%). Lignoceric acid (C24:0) was similarly suppressed. In Year 1, the control averaged 0.19%, and Zn alone lowered it by − 6.9% to − 15.5%, while AA treatments accentuated this decline, with AA₄×Zn₂ reaching 0.16% (− 17.2%). In Year 2, the control was slightly higher at 0.20%, but reductions remained consistent, ranging from − 9.8% to − 19.7% under Zn-only and AA combinations, with the lowest values at AA₂×Zn₄ (0.16%) and AA₃×Zn₃ (− 16.4%). In contrast, eicosenoic acid (C20:1) showed modest but consistent improvements under some treatments. In Year 1, the control averaged 0.22%, and Zn supplementation increased it by + 9.1% under Zn₂ and Zn₃, while AA additions produced small gains of + 4–7% at AA₂×Zn₄, AA₃×Zn₂, and AA₃×Zn₃ (up to 0.24%, + 7.6%). Year 2 displayed nearly identical patterns, with Zn₂ and Zn₃ increasing content by + 7.5–9.0%, and AA₂×Zn₄ and AA₃×Zn₃ providing the most substantial AA-driven enhancements (+ 4.5%). However, some AA combinations slightly reduced eicosenoic acid (− 1.5% to − 3.0%).

Trait networks and structural equation modeling

Correlation- and model-based network analyses provided new insights into the interconnectedness of soybean traits, highlighting relationships among fatty acid composition, oil-related parameters, and nutrient levels. The Pearson correlation network (|r| ≥ 0.40) showed dense connections among unsaturated fatty acids, with linoleic and linolenic acids forming a closely linked group, indicating proportional variation across treatments (Fig. 2a). Oleic acid played a central role in connecting the saturated and polyunsaturated fractions. Pearson network shows a strong FA module and a separate nutrient module; Zn and N did not meet the |r|≥0.40 threshold. Micronutrients, especially Zn and Mn, also exhibited positive associations with unsaturated fatty acids, supporting their influence on oil quality.

Fig. 2.

Integrated correlation and model-based trait network of soybean fatty acid composition and nutrient interactions. The figure summarizes multi-layered relationships among soybean traits derived from (a) Pearson correlation (|r| ≥ 0.40), b partial correlation (|ρ| ≥ 0.15), and c Graphical Lasso networks, alongside (d) a structural equation model (SEM). Modules of unsaturated fatty acids (oleic, linoleic, and linolenic), macronutrients (N, P, K, Ca, and Mg), and micronutrients (Zn, Mn, and Cu) are highlighted in distinct colors. Node size reflects degree centrality, and edge thickness indicates the strength of connections. The SEM (panel d) displays standardized path coefficients linking nutrient availability → fatty acid groups (PUFA, MUFA, SFA) → oil content → oil yield, illustrating nutrient-driven enhancement of unsaturation and yield

The partial correlation network (Ledoit–Wolf precision matrix, |ρ| ≥ 0.15) pruned indirect associations and highlighted direct dependencies (Fig. 2b). Here, linoleic–linolenic remained strongly connected, confirming their robust proportionality. Partial correlation retained only a few direct dependencies, notably a Zn–oil_content–N link (plus Zn–Mg), while most FA–nutrient links were removed. Notably, palmitic acid was more isolated, reflecting its relative independence from nutrient enrichment and highlighting treatment-specific modulation of saturated fatty acids.

The sparse Graphical Lasso network further distilled the system into a compact set of conditional dependencies (Fig. 2c). Communities were clearly partitioned: one dominated by unsaturated fatty acids (oleic, linoleic, linolenic), another by micronutrients (Zn, Mn, Cu), and a third by macronutrients (N, P, K, Ca, Mg). Oil content and yield occupied inter-modular positions, acting as connectors that linked nutrient status with lipid fractions. Centrality analyses across networks consistently identified oleic acid, linoleic acid, nitrogen, and zinc as hub traits, characterized by high degree and eigenvector centralities, which highlights their systemic influence.

To test directional hypotheses, a structural equation model (SEM) was constructed (Fig. 2d). Standardized path coefficients indicated that Zn, N, and P were significant positive predictors of PUFA_grp (β = 0.10–0.16), while SFA_grp was weakly promoted by Ca but negatively related to oil content. Oil content was strongly influenced by PUFA and MUFA groups (β = 0.22–0.49), whereas saturated fatty acids had a suppressive effect (β ≈ − 0.22). Oil yield mainly depended on oil content (β = 0.48) but was also directly affected by N, P, and Zn. The model explained up to 60% of the variance in oil content and yield, suggesting that nutrient enrichment alters fatty acid balances to favor unsaturation and enhance oil productivity.

Comparative cluster heatmap analysis across years

The side-by-side cluster heatmaps of treatments and traits offered valuable insights into the consistency and variability of treatment responses across the two experimental years (Fig. 3). In both years, treatments were grouped into distinct clusters mainly defined by their fatty acid composition, extracted oil mass characteristics, and nutrient enrichment patterns. However, the strength of clustering and the positions of high-performing treatments shifted depending on seasonal conditions. In Year 1, clustering was primarily driven by compositional traits, with treatments clearly separating along PUFA- versus MUFA-dominated profiles. Treatments with higher amino acid levels (AA₃–AA₄) combined with moderate zinc (Zn₂–Zn₃) consistently clustered together, reflecting their shared ability to increase the proportion of unsaturated fatty acids and improve lipid quality indices. Minor fatty acids (arachidic, behenic, lignoceric) contributed little to the separation, clustering showed comparatively lower dispersion/contrast than major fatty acids, consistent with their stability and low variability across treatments. Micronutrients such as Zn and Mn showed moderate clustering strength in Year 1, with zinc-enriched treatments forming a small sub-cluster. However, their differentiation was less pronounced than for fatty acid traits. Oil content and oil yield traits occupied central positions in the trait dendrogram, indicating intermediate responsiveness across treatments during the Year 1. In contrast, Year 2 displayed a markedly different clustering structure, reflecting stronger treatment differentiation driven by nutrient enrichment and yield-related traits. This year, zinc treatments had a greater influence, with Zn₂ and Zn₃ forming distinct sub-clusters across multiple amino acid levels, suggesting that Year 2’s environmental conditions enhanced zinc responsiveness. Oil yield became a more decisive factor in treatment separation, with AA₃–AA₄×Zn₂–Zn₃ treatments clustering together at the high-yield end of the heatmap, while controls or high-zinc (Zn₄) treatments clustered separately with lower performance profiles. Macronutrient traits such as N and P also showed increased dispersion in Year 2, with high-AA treatments forming clusters characterized by higher nitrogen and phosphorus content, indicating more substantial year-specific effects on macronutrient uptake. Comparing the two heatmaps highlights the dynamic nature of treatment responses: while the top-performing treatments (AA₃–AA₄ with Zn₂–Zn₃) remained consistent across years, their clustering patterns revealed different trait drivers of superiority. In Year 1, superiority mainly was related to compositional traits, especially improved unsaturation profiles. In Year 2, it was driven both by composition and yield, reflecting enhanced nutrient density and oil productivity. Treatments involving Zn₄ remained consistently isolated in both years, indicating that excessive zinc disrupted the desirable balance of traits and was often associated with reduced oil quality.

Fig. 3.

Comparative hierarchical cluster heatmaps of soybean treatments and traits across two experimental years. a and b are the first and second years of study, respectively

Principal component analysis of trait and treatment relationships across years

The PCA biplots for Year 1 and Year 2 show how fatty acid composition, oil traits, and nutrient levels together influenced the differentiation of soybean treatments (Fig. 4). The first two principal components captured most of the total trait variation, with PC1 explaining nearly half of the variance and PC2 accounting for 15–20%. This suggests these two axes effectively summarize the main patterns of treatment differences. In Year 1, the PCA biplot revealed a clear separation of treatments along PC1, mainly driven by fatty acid profiles. Traits like linoleic and linolenic acids (PUFAs) loaded positively on PC1, while palmitic and stearic acids (SFAs) loaded negatively, positioning treatments with higher unsaturation — especially AA₃–AA₄ combined with Zn₂–Zn₃ — toward the right side of the plot. These treatments were clearly different from control and low-input treatments (AA₁×Zn₁ or AA₁×Zn₂), which clustered nearer to the origin or toward the SFA side, indicating lower nutritional enhancement. PC2, though explaining less variance, further separated treatments based on oil yield, micronutrient enrichment (notably Zn and Mn), and yield-related traits. Treatments with higher oil content and better micronutrient profiles tended to cluster higher along PC2, reflecting multidimensional improvements in nutritional quality. The trait vectors also showed that oil yield was positively associated with PUFA levels, suggesting that increasing unsaturation also improved productivity in Year 1. In Year 2, the overall PCA structure was similar but also had notable differences from Year (1) Linoleic and linolenic acids again dominated PC1, confirming that PUFA enrichment remained the main axis of treatment differentiation across years. However, in Year 2, vectors for macronutrients like nitrogen and phosphorus aligned strongly with PUFA traits along PC1, indicating that environmental conditions that year favored simultaneous accumulation of unsaturated fatty acids and seed macronutrients. This pattern underscores that the combined AA and Zn treatments not only improved oil quality but also increased nutrient density more effectively in Year (2) Treatments with moderate zinc levels (Zn₂, Zn₃) and high amino acid supplementation (AA₃, AA₄) again positioned toward the PUFA-rich, nutrient-dense area of the biplot. Still, their separation from control treatments was more pronounced than in Year 1, reflecting more substantial effects under Year 2 conditions. In Year 2, PC2 was mainly influenced by oil content and Zn concentration, which contributed to treatment separation along the vertical axis. Interestingly, while Zn₄ treatments remained distinct, they were located away from PUFA and oil quality traits, reinforcing that excessive zinc application suppressed desirable nutritional outcomes. Comparing the two years highlights both stability and adaptability in treatment responses. The stability is seen in the consistent placement of AA₃–AA₄×Zn₂–Zn₃ treatments as superior, PUFA-rich, and nutrient-enhancing in both years. The adaptability is evident in the shifting relationships of secondary traits — where oil content and yield were more independent drivers in Year 1, but in Year 2, these traits aligned more closely with PUFAs and macronutrients, strengthening the nutritional benefits of high-input treatments. This suggests that environmental conditions in Year 2 not only enhanced the oil’s nutritional quality but also integrated productivity and nutrient enrichment more closely within the same treatment group.

Fig. 4.

Principal Component Analysis (PCA) biplots illustrating trait–treatment relationships in soybean across two experimental years. a and b are the first and second years of study, respectively

Comparative stability analysis across traits using AMMI1 and GGE biplots

The AMMI1 and GGE biplots, created for major fatty acids, minor fatty acids, oil-related traits, and nutrient compositions, offered an integrated perspective on treatment stability and responsiveness over the two years of the experiment (Figure S2 A-G). Overall, the patterns showed that although treatment effects were highly significant for most traits, the level of stability and genotype × environment (G × E) interaction varied greatly depending on the biochemical class studied. For major fatty acids, treatments consistently exhibited broad separation along the AMMI1 IPC1 axis, indicating considerable G×E interaction driven by differences between Years 1 and 2. Linoleic and linolenic acids displayed extensive treatment dispersion, suggesting that PUFA proportions were highly sensitive to environmental variation and nutrient supply. In contrast, oleic acid showed relatively lower IPC1 dispersion, reflecting more stable expression across the years.

The GGE biplots reinforced these observations, with PC1 capturing most of the variance and placing Years 1 and 2 in opposite quadrants relative to PUFA-dominant treatments, highlighting the nutritional plasticity of fatty acid composition under foliar biofortification. Minor fatty acids (arachidic, behenic, eicosenoic, and lignoceric) showed more compact distributions in both AMMI1 and GGE plots, indicating overall stability and limited responsiveness to treatments. The low variance explained by IPC1 for these traits aligns with their minor role in total oil content and limited modulation through foliar amino acid and zinc supplementation. These results suggest that minor fatty acids are mostly environmentally buffered, with minimal fluctuations across years, reducing their effectiveness in distinguishing treatment stability. In contrast, oil content and oil yield exhibited strong G×E effects. The AMMI1 biplots for these traits revealed clear treatment differentiation along IPC1, indicating that the interaction between foliar nutrition and seasonal conditions had a significant impact on oil productivity. GGE biplots consistently placed treatments associated with higher oil content and yield on the same side of PC1 across both years, with Year 2 vectors demonstrating a greater magnitude. This indicates that favorable conditions in Year 2 amplified treatment benefits, especially for combinations of AA₃–AA₄ with Zn₂–Zn₃, which not only produced higher oil yields but also clustered closer to the year vectors, reflecting strong responsiveness. Conversely, low-input treatments and controls clustered near the origin, indicating limited responsiveness and minimal contribution to variation. For micronutrient content (Zn, Mn, Cu), GGE biplots revealed distinct patterns, with Zn levels exerting strong influence along PC1, confirming treatment-driven enrichment. Zn and Mn were more sensitive to environmental and treatment interactions, while Cu remained relatively stable. AMMI1 further showed that Zn enrichment was highly year-dependent, with IPC1 values sharply separating treatments across the two years. This demonstrates that foliar Zn supply strongly interacted with seasonal conditions to influence seed micronutrient levels. Macronutrients (N, P, K, Ca, Mg) showed diverse stability profiles. Nitrogen and phosphorus experienced strong G×E interactions, with treatments achieving higher mean values positioned further along IPC1, especially in Year 2, indicating greater responsiveness under favorable conditions. Potassium and magnesium were more stable, clustering tightly around the origin in both AMMI1 and GGE plots. Calcium showed moderate responsiveness to treatments and some environmental buffering, displaying intermediate behavior.

Overview of treatment effects on fatty acid composition

Over both years, applying amino acids (AA) and zinc (Zn) as a foliar spray resulted in clear and repeatable changes in the fatty acid profile of soybean seed oil. The data showed that oleic (C18:1), linoleic (C18:2), and linolenic (C18:3) acids accounted for over 85% of all fatty acids, making them the primary factors influencing oil quality and unsaturation (Fig. 1). Saturated fats—palmitic (C16:0) and stearic (C18:0)—represented a smaller yet significant portion (< 15%), affecting melting point and oxidative stability. The combination of foliar Zn and amino acids consistently improved the unsaturation pattern. Throughout both years, high amino acid levels (AA₃–AA₄) paired with moderate Zn levels (Zn₂–Zn₃) yielded the best lipid profiles: higher oleic acid, lower linoleic acid, and a balanced reduction in linolenic acid. These changes shifted the ω-6/ω-3 ratio toward the ideal physiological range (≈ 8–10), enhancing the oil’s nutritional quality. Conversely, lower amino acid levels (AA₁–AA₂) or no Zn added resulted in higher amounts of polyunsaturated acids, which, while raising iodine value, can weaken oxidative stability and reduce storage life. Over both years, environmental factors had only a slight effect on total lipid levels, not on the order of fatty acids. Year 2 had marginally higher oleic acid and lower linoleic acid, indicating more stable desaturation processes under the nutritional treatment. This consistency suggests that foliar feeding influenced lipid biosynthesis pathways at a physiological level rather than causing year-specific environmental effects.

Because fatty-acid data are inherently compositional (components sum to one), Aitchison and centered log-ratio (CLR) transformations were used to reveal proportional structures (Table S5, Table S6, Table S7, Table S8). The Aitchison variation matrices showed a close relationship between oleic and linoleic acids (variation < 0.02), suggesting a strong substitutive relationship in biosynthesis via the Δ12-desaturase pathway. Similarly, linoleic and linolenic acids exhibited moderate variance (≈ 0.03), indicating additional desaturation steps through Δ15-desaturase. Minor components (arachidic, behenic, lignoceric acids) displayed higher variance (0.06–0.08), reflecting looser control and greater environmental sensitivity. CLR transformations yielded a more interpretable multivariate structure. Positive CLR coordinates for oleic acid and eicosenoic acid, along with negative loadings for linoleic and linolenic acids, defined the main compositional gradient that distinguished the treatments. The groups AA₃×Zn₂, AA₄×Zn₂, and AA₄×Zn₃ consistently appeared at the positive end of this gradient, while controls and low-AA × low-Zn treatments clustered at the opposite end. This geometric separation was highly consistent across years, underscoring the biological stability of treatment effects.

To interpret compositional changes physiologically, several derived lipid indices were calculated: Iodine Value (IV), Unsaturation Index (UI), Peroxidizability Index (PI), Atherogenicity Index (AI), Thrombogenicity Index (TI), and Health-Promoting Index (HPI) (Table S9, Table S10). Across all treatments and years, the mean IV was 131.3 ± 1.1, and the UI averaged 151.7 ± 1.3, indicating moderate unsaturation with good oxidative resistance. The peroxidizability index (PI ≈ 135 ± 2.3) exhibited a balanced sensitivity to oxidation, sufficient to maintain fluidity without compromising shelf life. Cardiovascular indices showed consistently low AI (≈ approximately 1.96) and TI (≈ approximately 5.78), along with a relatively high HPI (≈ approximately 0.51), indicating a healthy lipid profile typical of oleic-rich soybean oil. No significant year effect was seen for any lipid index (p > 0.05), and coefficients of variation stayed below 1%. This consistency confirms the robustness of the AA×Zn modulation and the stability of fatty acid biosynthesis under various field conditions. The combined indices indicate that foliar nutrient management enhances oil quality by reducing the amount of atherogenic, saturated fatty acids and increasing monounsaturated fats, which benefit cardiovascular health.

The stearoyl-CoA desaturase index (SCD18), which reflects the enzymatic conversion of stearic acid (C18:0) to oleic acid (C18:1), remained notably stable over the years—0.839 (Y₁) and 0.840 (Y₂)—with minimal variation (SD ≈ 0.01) (Table S11). This consistency indicates that the treatments influenced the substrate pools rather than changing enzyme activity itself. The stable SCD18 values imply that desaturase enzymes are regulated in a homeostatic manner even under different zinc and amino acid regimes. Since Zn is a known cofactor for many dehydrogenase and oxidase systems, its foliar application likely affected membrane lipid desaturation indirectly by modulating carbon and nitrogen metabolism, which helped maintain adequate NADPH levels for desaturase reactions. Similarly, amino acid supplementation may have increased the availability of acetyl-CoA and malonyl-CoA through improved nitrogen assimilation and a more balanced transamination process. Collectively, these factors enhanced oleic acid synthesis without inducing oxidative stress, as indicated by unchanged PI and stable HPI.

Principal component analysis (PCA) of CLR-transformed fatty acids and derived indices revealed the overall structure of variation (Table S12, Table S13). For Year 1, the first two components explained 73.4% of the total variance (PC₁ = 47.3%, PC₂ = 26.2%), while in Year 2, the explained variance increased to 80.5% (PC₁ = 53.7%, PC₂ = 26.8%). PC₁ represented the unsaturation axis, positively associated with oleic acid, iodine value, and unsaturation index, and negatively correlated with linoleic and linolenic acids. PC₂ captured secondary variation among minor long-chain fatty acids (behenic, lignoceric, eicosenoic) and was sensitive to subtle differences in desaturation pathways. The PCA biplots showed clear clustering of treatments: High AA × moderate Zn (AA₃–₄×Zn₂–₃) formed a compact group in the upper-right quadrant, characterized by high oleic acid, elevated IV/UI, and low AI/TI. Low AA × low Zn (AA₁×Zn₁) and control treatments clustered in the lower-left quadrant, dominated by polyunsaturated linoleic and linolenic acids. Intermediate groups occupied transitional regions, reflecting partial compositional changes. Comparing PCA loadings across years showed nearly identical orientation of variable vectors, indicating strong structural reproducibility. The cumulative variance explained by PC₁ + PC₂ (> 70% in both years) confirms that a small number of latent factors—primarily the oleic–linoleic balance—governed most of the observed compositional variability.

Fatty-acid balance analysis (isometric log-ratios) and nutritional desaturation structure (NDS) metrics further supported PCA findings (Table S14). The ratios C18:1/C18:0 and C18:2/C18:1 served as sensitive indicators of desaturase activity. Under increasing amino acid inputs, the C18:1/C18:0 ratio rose steadily, indicating improved monounsaturated production, whereas C18:2/C18:1 declined, consistent with partial suppression of Δ12-desaturase. The integrated NDS plots revealed a hierarchical desaturation pattern: saturated → Monounsaturated → Polyunsaturated, which was enhanced at the first step but moderated at the second, resulting in an overall shift toward oleic accumulation. This metabolic balance supports higher oxidative stability while maintaining essential linoleic content for human nutrition. Zinc played a synergistic role in maintaining enzyme cofactor integrity, and amino acids provided nitrogen scaffolds for enhanced lipid biosynthesis, together fine-tuning the desaturation cascade.

Despite minor climatic fluctuations from year to year, all analytical layers (compositional, biochemical, and multivariate) converged on a consistent interpretation: nutrient supplementation with amino acids and Zn significantly altered oil composition in a reproducible manner. The mean differences between years for key indices (IV, UI, AI, TI, HPI) were all within ± 0.5%, indicating strong temporal stability. The PCA structures overlapped considerably, and SCD18 values remained nearly identical, indicating that treatment effects are intrinsic and physiologically rooted rather than environmentally dependent. The combined evidence therefore supports the hypothesis that co-application of Zn and amino acids improves lipid quality through metabolic reprogramming of desaturation pathways without disrupting overall homeostasis.

The integrated results outline a clear mechanistic story. Foliar Zn functions as a structural cofactor for enzymes involved in oxidative metabolism, supporting NADPH regeneration and maintaining membrane integrity. Meanwhile, amino acids act as metabolic regulators, supplying building blocks for protein synthesis and balancing carbon and nitrogen to feed lipid biosynthesis pathways. The synergy between zinc and amino acids ensures efficient carbon flow toward oleic acid accumulation, while moderating the synthesis of linoleic and linolenic acids, ultimately improving the nutritional quality of lipids. These biochemical effects are reflected in consistent statistical and multivariate patterns, including increased oleic acid content, a higher unsaturation index with stable peroxidizability, lowered atherogenic and thrombogenic potentials, and better clustering around ideal compositional ratios. This integrated performance indicates that amino acid–zinc combinations with moderate to high input levels often provided the most favourable results for enhancing oil quality across various conditions, although the exact treatment ranking varied among MCDM methods. Over two separate growing seasons, foliar applications of amino acids and zinc reliably shifted soybean seed oil composition toward a monounsaturated, cardioprotective profile. The consistent fatty acid levels, lipid indices, and multivariate patterns across years indicate strong physiological control and reproducibility. Overall, the combined CoDA–PCA–index approach confirms that balanced amino acid and zinc inputs can sustainably enhance the nutritional and industrial value of soybean oil without disrupting desaturation pathways or compromising oxidative stability. This method offers a practical farming solution for producing high-value, healthful oils suitable for various production settings.

Radar chart visualization of the six lipid quality indices—iodine value (IV), unsaturation index (UI), atherogenicity index (AI), thrombogenicity index (TI), health-promoting index (HPI), and peroxidizability index (PI)—clearly illustrated the multivariate response of soybean oil composition to amino acid (AA) and zinc (Zn) supplementation across both experimental years (Figure 5). In Year 1, polygons corresponding to high-AA and moderate-Zn treatments (AA₃×Zn₂ and AA₄×Zn₃) expanded markedly outward along the IV, UI, and HPI axes while contracting along AI and TI, indicating increased unsaturation, improved cardiovascular quality, and stable oxidative resistance. Year 2 exhibited the same geometric hierarchy, with slightly higher IV and UI values, but preserved a proportional structure, confirming the reproducibility of treatment effects across environments. The combined-mean radar chart integrated both years and revealed a smooth performance gradient, with low-input treatments (AA₁×Zn₁–Zn₂) forming compact inner polygons and optimized treatments (AA₃–AA₄×Zn₂–Zn₃) occupying the outermost regions, reflecting a synergistic enhancement in all oil-quality dimensions. The opposing directions of the IV/UI versus AI/TI axes visually captured the metabolic trade-off between unsaturation and atherogenicity. At the same time, the moderate expansion of PI confirmed that these gains were achieved without compromising oxidative stability. Overall, the balanced and nearly circular polygons of AA₃×Zn₂ and AA₄×Zn₃ treatments represent the most desirable lipid profiles—highly unsaturated, low in atherogenic and thrombogenic potential, and metabolically stable—demonstrating that coordinated amino acid and zinc nutrition can reproducibly reprogram fatty-acid biosynthesis toward healthier and more stable soybean oil.

Fig. 5.

Radar charts of lipid quality indices (IV, UI, AI, TI, HPI, and PI) in soybean oil under AA×Zn treatments for (a) Year 1, b Year 2, and c the combined-mean radar chart

Multi-Criteria Decision-Making (MCDM) integration of treatment performance across years

The comprehensive MCDM analysis, conducted using six decision-making algorithms—TOPSIS, VIKOR, PROMETHEE II, ELECTRE III, WASPAS, and MOORA— revealed clear but method-dependent differences in treatment performance (Table S15). Among individual methods, TOPSIS consistently ranked AA₄×Zn₄ first (score 1.0), supported by AA₂×Zn₃ and AA₂×Zn₄ as high-performing alternatives. Meanwhile, VIKOR, which emphasizes compromise between the ideal and worst solutions, selected AA₁×Zn₂, AA₄×Zn₃, and AA₃×Zn₂ as its leading options. The outranking-based methods, PROMETHEE II and ELECTRE III showed the same ranking order, both placing AA₂×Zn₄ first, followed by AA₂×Zn₁ and AA₃×Zn₄, while AA₄×Zn₄ and AA₄×Zn₃ also remained within the upper-ranking group. A closely related pattern was obtained with WASPAS, where AA₂×Zn₄, AA₂×Zn₁, and AA₃×Zn₄ occupied the first three positions, again supporting the competitive performance of these combinations. Under MOORA, the ranking was led by AA₄×Zn₄, followed by AA₂×Zn₄ and AA₃×Zn₄, with AA₄×Zn₂ and AA₄×Zn₃ also appearing among the strongest alternatives. Spearman correlation among methods showed high agreement (ρ > 0.8) between TOPSIS, PROMETHEE II, ELECTRE III, and MOORA, while WASPAS and VIKOR displayed moderate differences due to their unique aggregation approaches. Taken together, these results indicate that although the precise order differed among decision models, treatments combining moderate-to-high doses of amino acids with moderate-to-high levels of zinc were consistently favored. Specifically, low-input treatments typically held lower positions, while AA₂×Zn₄, AA₃×Zn₄, AA₄×Zn₄, and AA₄×Zn₃ consistently appeared among the best-performing options across multiple ranking frameworks. The conclusion that stronger nutritional intervention produced the most balanced improvement in nutrient accumulation, oil-related traits, and overall treatment desirability is supported by this overall pattern, particularly when combined with amino acid supplementation and zinc enrichment.

Discussion

This study demonstrated that combining amino acid (AA) biostimulants with foliar zinc (Zn) fertilization can synergistically reprogram soybean seed metabolism, yielding results that neither input alone has achieved previously. The combined application significantly increased seed Zn concentration (up to + 128%) along with other minerals (e.g., N, + 23%; P, + 31%; K, + 39%; Cu, + 33%) without disrupting the balance of Ca and Mg. These findings confirm that AAs can improve micronutrient uptake and transport, likely by acting as natural chelators and carriers [29]. Foliar-applied AAs may form complexes with Zn that enhance leaf penetration and phloem mobility, while also boosting root and leaf functions (e.g., increased N assimilation and chlorophyll production) [30]. Few previous studies have examined such AA–Zn interactions in field-grown soybeans, so our results fill a significant knowledge gap. The novelty is in demonstrating that a tailored nutrient and biostimulant strategy can simultaneously raise micronutrient levels and modify primary metabolism (oil synthesis) in seeds [31]. This integrated biofortification approach offers a fresh perspective, shifting from single-factor solutions toward coordinating multiple physiological pathways for better crop quality. Biofortification implies the enhancement of the nutritional quality of crops using methods such as selective breeding, genetic manipulation, fermentation, and agronomic techniques, including the use of fertilizers or other nutrients [32]. The main objective is to rectify nutritional deficiencies and enhance overall dietary quality, especially in low-income and areas with limited resources where access to varied food items is restricted and treatments like supplementation or industrial fortification may be unfeasible [33]. In this context, our AA–Zn foliar approach serves as an agronomic biofortification method, as it concurrently increased seed Zn enrichment (up to + 128%) and enhanced the overall mineral profile and oil quality, therefore supporting both nutritional and functional quality objectives.

Encouragingly, the AA–Zn treatments enhanced agronomic performance across two seasons. Oil yield increased by up to approximately 13% at moderate input levels, without reducing seed protein content or overall yield. This is notable because soybeans often show a trade-off between protein and oil [34]. However, by supplying additional N through AAs and critical Zn cofactors, the plants seem to overcome resource limitations and fill seeds more completely [35]. In fact, certain mid-level combinations (e.g., AA₂–AA₃ × Zn₂–Zn₃) produced more pods and larger seeds, resulting in higher oil output—a result consistent across both years. These gains correspond with reports that Zn fertilization can increase seed yield and oil content up to an optimal dose [36]. For example, Jat et al., [37] observed an increase in soybean oil percentage with Zn application, and our work reinforces that sufficient Zn availability is a limiting factor for oil accumulation. At the same time, we found that excessive inputs yield diminishing returns. The highest treatment (AA₄×Zn₄) did not further boost yield or Zn enrichment and was somewhat less effective than moderate doses. This indicates a saturation point or slight antagonism at very high concentrations – a phenomenon also observed in the study where supra-optimal Zn can reduce seed set or interfere with other processes nutrients [36, 38]. Thus, more is not always better; our results highlighted the importance of optimized dosing to maximize AA–Zn benefits without causing toxicity or nutrient imbalances. The robust yield and biofortification gains attained here (even in calcareous soil conditions) highlight the agronomic viability of AA–Zn foliar feeding as a tool to produce nutrient-enriched soybeans [39].

A key finding is that AA–Zn treatments altered the seed fatty acid profile, showing a coordinated treatment-induced shifts in seed composition and fatty acid profiles. We observed a consistent decrease in stearic acid (C18:0, a saturated fatty acid) along with a decline in oleic acid (C18:1, a monounsaturated fatty acid), combined with notable increases in linoleic acid (C18:2, + 5.9%) and eicosenoic acid (C20:1, + 9%). The increase in total polyunsaturates (PUFA) compared to saturates suggests that fatty acid desaturase enzymes were upregulated or activated by the treatments [40]. This is biologically plausible: Zn is known as a cofactor for fatty acid desaturases and elongases [41], so enhanced Zn uptake could directly stimulate the conversion of saturated to unsaturated fatty acids. In particular, the decrease in oleic acid, accompanied by a corresponding increase in linoleic acid, suggests that Δ¹²-desaturase (FAD2) activity was increased, resulting in a greater conversion of oleate (18:1) to linoleate (18:2) [42]. At the same time, the increase in eicosenoic acid (20:1) indicates higher activity in fatty acid elongation pathways [43]; zinc-dependent elongase enzymes may have become more active, lengthening 18-carbon chains to 20-carbon monounsaturated fats [41]. Amino acid supplementation likely enhanced this process by improving the plant’s overall redox status and energy metabolism—conditions in which desaturases function effectively [44]. AAs such as cysteine and glutamate can strengthen cellular antioxidant defenses (e.g., glutathione production), while Zn acts as a cofactor for antioxidant enzymes like superoxide dismutase [45]. Linolenic acid (C18:3) did not exhibit a significant increase, indicating that the transition to elevated C18:2 occurred without a corresponding rise in C18:3. This pattern aligns with variations in the relative flow via FAD2- compared to FAD3-associated stages; however, enzyme activity were not quantified, rendering this conclusion speculative [46]. Such coordinated shifts indicate a treatment-induced shifts in the oleic–linoleic–linolenic pathway due to improved Zn nutrition and N assimilation [47]. Importantly, these compositional changes were highly consistent across both years, as shown by our CoDA and PCA analysis, which revealed nearly identical proportional fatty acid relationships between 2022 and 2023.

The ability to manipulate soybean oil composition through foliar nutrition is an emerging concept, and our findings both agree with and expand upon earlier research work. Zinc’s role in fatty acid metabolism has been suggested in previous studies. For example, Sharifi [48] reported that adding Zn (with biofertilizers) increased the unsaturated fatty acid fraction in soybean oil, thereby enhancing oil quality indices. This supports our observation of a higher unsaturated-to-saturated ratio under Zn enrichment. On the other hand, some studies have observed different effects depending on context. Shahsavari et al. [49] found that Zn fertilization of oilseeds increased certain saturated fats (palmitic, stearic, behenic acids) along with linoleic acid, and suggested that Zn in seeds might prevent PUFA oxidation, helping to preserve more saturated lipids [49]. In our case, saturates actually declined, indicating that with sufficient AA co-treatment, the treatment-induced shift favored desaturation over the buildup of saturates. It is possible that the amino acids reduced oxidative stress enough that preserving saturates was unnecessary [50]– instead, the additional Zn could be used for desaturase function, producing more PUFA [51]. Additionally, differences in Zn dosage or plant genotype might explain the discrepancy: we used moderate Zn levels alongside AAs, while very high Zn alone (as in some reports) could overwhelm the system and result in less desirable fatty acids profiles. Notably, we also observed a rise in seed oil content (up to a 15% increase in oil) with optimal AA–Zn treatments. This result supports other studies where Zn fertilization increased oil content in soybean [36], and related crops (e.g. rapeseed and sunflower), likely by promoting photosynthate flow to seeds and activating fatty acid synthesis enzymes [52, 53]. However, previous foliar Zn trials on soybean without biostimulants often showed minimal or no improvement in seed oil or protein content [36]. The clear improvements observed in our study highlight the value of AA co-application: by enhancing nitrogen nutrition and potentially hormone-like signaling, AAs provided the missing stimulus that enabled Zn’s benefits to appear in seed composition [54].

From a nutritional perspective, the oil produced with AA–Zn treatment was healthier and more stable than the control oil, as demonstrated by various lipid indices. The increase in linoleic acid (an essential ω-6 fatty acid) and the maintenance of α-linolenic acid (ω-3) ensure a high level of polyunsaturates, which are well-known to improve plasma cholesterol profiles when they replace saturated fats in the diet [55]. Indeed, the atherogenic index (AI) and thrombogenic index (TI) in our treated samples were significantly lower than in untreated soybeans (data not shown here), indicating a reduced amount of saturated fats that can promote arterial plaque and blood clot formation. A lower AI is associated with lower LDL-cholesterol levels [56], and a lower TI similarly indicates a fatty acid profile that is less likely to contribute to thrombosis [57]. Conversely, the health-promoting index (HPI) was significantly increased by the treatments, indicating a higher ratio of hypocholesterolemic to hypercholesterolemic fatty acids. A high HPI (or h/H ratio) points to fats that may protect against cardiovascular disease [58] – our top AA–Zn combinations yielded HPI values higher than those usually seen in conventional soybean oil. It’s essential to note that these nutritional benefits were achieved without compromising oxidative stability. Although the iodine value (IV) and unsaturation index (UI) of the oil increased slightly due to the presence of more double bonds from linoleic acid, the simultaneous changes—such as the decrease in ultra-unsaturated linolenic acid and the rise in monounsaturated 20:1—helped maintain a balanced peroxidation index. In practical terms, the enhanced oil would have shelf stability similar to that of regular soybean oil, especially since the enrichment of Zn and Cu in seeds may increase the natural antioxidant enzymes or tocopherol content that protect against rancidity [36]. The improved fatty acid profile also offers industrial advantages: oils richer in linoleic acid (and moderate in oleic acid) are valuable for specific food and chemical uses. The consistent performance of the high-AA + moderate-Zn treatments in yielding a cardio-protective lipid profile across two years is promising.

One of the strengths of this study was the use of multivariate and modeling approaches (CoDA, network analysis, SEM) to uncover the complex interactions involved. Network modeling revealed that oleic and linoleic acids form a key antagonistic pair within the metabolic network, and intriguingly, Zn and N (total nitrogen) emerged as important hubs closely linked to these fatty acids. This provides a mechanistic insight: adequate Zn and N nutrition are essential for shifting the oleic/linoleic balance, likely through co-regulation of desaturase gene expression or activity. Our structural equation modeling further supported this idea, showing that the positive influence of Zn on linoleic content was partially mediated by improved N status (as a proxy for protein/enzyme synthesis) and by direct effects on enzymatic pathways (consistent with Zn’s role as a cofactor). In essence, AAs and Zn work together: AAs enhance the plant’s nitrogen-dependent biosynthetic capacity (more enzymes and antioxidants), while Zn activates specific enzymes (desaturases/elongases) that modify fatty acid structures [59]. Another key point is that no strong antagonism between Zn and other micronutrients was observed at optimal levels: seed Mn and Cu actually increased alongside Zn, indicating that the AA-mediated improvement in overall nutrient uptake offset any competitive issues. Only at very high Zn levels might negative feedback occur (e.g., internal Zn excess downregulating its own transport or inducing mild stress). These nuanced insights, supported by our multi-criteria decision analysis (which combined yield, nutrition, and quality measures), led us to conclude that the best results come from a balanced treatment—typically AA at 75–100% of the maximum rate combined with Zn at 50–75% of the maximum. These combinations (e.g., AA₃×Zn₂ or AA₃×Zn₃) consistently optimized desirable traits across both years, whereas adding cost and complexity to AA₄×Zn₄ provided no additional benefits. This type of optimization aligns perfectly with sustainable agronomy goals: finding the optimal balance where biological gains and resource efficiency are both maximized. Our findings clearly demonstrate that targeted, moderate biostimulant–micronutrient interventions can produce significant improvements in crop quality without the downsides of over-fertilization.

Conclusions

In conclusion, this two-year field study provided strong evidence that foliar co-application of amino acids and zinc can reprogram soybean seed metabolism in ways that simultaneously increase micronutrient density and enhance oil quality. Key findings include: (1) a substantial rise in seed Zn content (doubling in some cases) along with higher levels of N, P, K, Cu, and Mn, confirming the effectiveness of the AA–Zn strategy for broad nutrient enrichment; (2) a measurable increase in oil yield (up to 15% higher oil content) without reducing protein yield, likely due to improved pod development and seed filling under moderate AA–Zn applications; and (3) a positive shift in fatty acid composition—characterized by lower saturated fat and a higher proportion of unsaturated fats—which led to improved lipid health indices (lower AI and TI, higher HPI, etc.) and a more heart-healthy oil profile. Importantly, these benefits were consistently observed across two growing seasons, highlighting the robustness of the approach in real-world field conditions. The biochemical effects of AA–Zn interaction were reflected in the coordinated regulation of desaturase pathways, producing oils with increased linoleic acid and maintained linolenic acid—crucial for essential fatty acid nutrition—while preserving oxidative stability. This rebalancing of seed carbon and fatty acid metabolism, driven by targeted foliar inputs, demonstrates a form of metabolic biofortification that complements traditional breeding and genetic engineering efforts.

From an agronomic and nutritional standpoint, the significance of this work was clear. Agronomically, combining amino acid biostimulants with micronutrient foliar feeds can serve as a practical tool to improve crop quality. Farmers and crop advisors might apply AA–Zn foliar sprays during key growth stages to produce soybeans that not only yield well but are also pre-enriched with Zn and have better oil characteristics. This supports sustainable agriculture goals: the approach uses naturally derived compounds and targeted application, reducing the need for excessive soil fertilization and increasing nutrient use efficiency. Nutritionally, the enriched soybeans can help address Zn deficiency (known as “hidden hunger”) in populations relying on plant-based diets, while the altered oil profile offers additional benefits for cardiovascular health and food industry use. Biochemically, our results confirm that plant nutrient status can be manipulated to change primary metabolism results that can lead to new research on crop biofortification strategies that go beyond yield and basic nutrition to include functional food quality.

The success of the AA–Zn approach may influence policies for biofortification initiatives. It provides an accessible, low-cost agronomic method to improve soybean nutrition, which can be promoted by agricultural extension services and governments in areas with zinc-poor soils or malnutrition. Demonstrating that yield, nutrient content, and oil quality can all be enhanced at once supports a more integrated crop improvement strategy, potentially guiding policies and incentives for farmers to cultivate nutrient-dense, healthful crops. Additionally, enhancing the natural quality of soybean oil, one of the most widely used edible oils, strengthens food system resilience: healthier, more stable, nutrient-rich oil lessens dependence on post-harvest fortification, extends shelf life, and reduces waste. Over time, adopting these biofortified, high-quality soybeans broadly could diversify dietary nutrient sources, help mitigate nutritional deficiencies in at-risk populations, and offer farmers a premium product.

Abbreviations

AA

amino acids

Zn

zinc

CoDA

detailed fatty acid compositional data

MCDM

multi-criteria decision analysis

RCBD

randomized complete block design

FAMEs

Fatty acid methyl esters

PCA

principal component analysis

MANOVA

multivariate analysis of variance

SFA

saturated fatty acids

MUFA

monounsaturated fatty acids

PUFA

polyunsaturated fatty acids

NDS

nutrient density score

Authors’ contributions

EY, MK: conceptualization, investigation, formal analysis, writing-original draft preparation, and software. IS: Methodology, formal analysis, and visualization. BG, SH, AS, OK: Software, Data curation. MQ, MG: Validation. MK: Writing-review. EY, MK, IS: Review and editing, supervision, and funding acquisition. EY: Methodology, Writing-review and editing, Funding acquisition. All the authors have read and approved the final manuscript.

Funding

There was no funding.

Data availability

All data generated or analyzed during this study are included in this published article and its supplementary information files.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.