A coordinated toolbox strategy integrating direct pathogen suppression, host-associated responses, and microbiome remodeling for crop protection

Abstract

Although integrated pest management promotes the coordinated use of multiple disease-control tactics, single sprayable formulations that consolidate complementary crop-protection functions remain underdeveloped. Here, we combined semi-rational screening with supramolecular co-assembly to construct a ternary nanoassembly comprising glycyrrhizic acid (GA, a plant-derived amphiphile), magnolol (MN, a plant-derived antifungal phenolic), and Zn²⁺, termed G-M@Zn NPs. Compared with free MN, G-M@Zn NPs enhanced MN photostability by 3.7-fold and increased rainfastness by 2.1-fold, indicating improved formulation persistence under simulated photolytic and wash-off stresses. The nanoformulation exhibited broad-spectrum antifungal activity in vitro against representative phytopathogenic fungi and, in a wheat-spike infection model of Fusarium graminearum, reduced the disease index from 52.5 ± 2.46 in the infected control to 22.3 ± 3.4. Under toxin-inducing conditions, G-M@Zn NPs also reduced deoxynivalenol (DON) accumulation and downregulated key TRI genes. Mechanistic assays linked the antifungal activity of G-M@Zn NPs to fungal membrane injury and mitochondria-associated oxidative stress in F. graminearum, with free MN eliciting qualitatively similar but weaker responses. G-M@Zn NP treatment was also accompanied by host defence-related physiological and metabolic responses involving redox regulation and tryptophan/indole-related pathways, together with treatment-associated shifts in the wheat spike microbiome. Preliminary biosafety assays indicated crop compatibility under the tested conditions and low acute toxicity toward earthworms. Serial-passage assays with F. graminearum showed only a modest increase in EC₅₀ after eight passages. Overall, this ternary co-assembly provides a strategy for integrating complementary crop-protection functions within a single multifunctional nanoformulation and supports IPM-compatible control of F. graminearum infection in wheat. It may also inform the design of bioinspired nanopesticide systems that combine improved formulation persistence with direct antifungal activity and crop-associated biological responses.

Graphical abstract

Supplementary Information

The online version contains supplementary material available at 10.1186/s12951-026-04644-3.

Introduction

Phytopathogenic fungi cause hundreds of crop diseases and therefore pose a persistent threat to global agriculture [1, 2]. The Food and Agriculture Organization of the United Nations (FAO) estimates that plant pests and diseases reduce global crop production by 20–40% each year, translating into economic losses exceeding USD 220 billion annually [3]. Currently, disease control relies predominantly on conventional chemical fungicides [4]. However, sustained and intensive fungicide use can accelerate resistance evolution, increases environmental burdens, and disturb agroecosystem functioning [4], thereby compromising long-term disease-management efficacy [5–7]. Integrated pest management (IPM) was developed to reduce dependence on any single tactic by coordinating agronomic, biological, physical, and limited chemical measures under monitoring- and threshold-guided decision-making [8, 9]. Although IPM provides an overarching framework for integrated disease management, crop-protection inputs that consolidate complementary functions within a single field-relevant formulation remain underdeveloped. In practice, direct pathogen suppression, formulation persistence, host defense-related responses, and microbiome-associated effects distributed across separate interventions, which can limit their coordination and robustness under field conditions [10, 11]. A representative example is Fusarium graminearum, a major causal agent of Fusarium head blight (FHB) in cereals. FHB not only reduces grain yield and quality but also contaminates grain with the mycotoxin deoxynivalenol (DON), thereby creating a dual challenge for crop productivity and food safety [12]. Current FHB management still relies heavily on fungicides, yet field control is often inconsistent because of unstable effective exposure, resistance risk, and limited functional breadth [13]. Most conventional inputs are designed primarily for direct pathogen suppression and are not intended to concurrently engage host defense-related responses or wheat-spike microbiome-associated effects. More broadly, this limitation is not unique to FHB: effective protection against phytopathogenic fungi often requires coordinated action across direct antifungal activity, environmental robustness, host-associated responses, and microbial ecological effects, but these functions are rarely embedded within a single practical input [14]. There is therefore a need for formulations that integrate these complementary functions in a field-relevant manner.

Nanoscale co-assembly offers a promising route to address this formulation-level gap [15]. By integrating multiple active components into a single supramolecular system, co-assembled nanoformulations have the potential to improve physicochemical robustness while combining complementary biological functions [16, 17]. Plant-derived natural products have been explored as antifungal leads owing to their broad bioactivity, biodegradability, and generally favorable environmental profiles [18]. For instance, magnolol (MN), a major biphenolic compound from Magnolia officinalis bark, exhibits broad-spectrum activity in vitro against various plant pathogens such as Colletotrichum gloeosporioides, Rhizoctonia solani, Fusarium culmorum, and Botrytis cinerea [19–21]. Glycyrrhizic acid (GA; throughout this manuscript, GA refers to glycyrrhizic acid rather than gibberellic acid), a plant-derived amphiphilic triterpenoid saponin, is well suited for supramolecular assembly and may influence biological interfacial organization [22, 23]. Certain phyllosphere-associated bacteria, including some Pantoea species, have been reported to suppress F. graminearum, suggesting that microbiome composition on the wheat spike may influence disease outcomes [24, 25]. Zinc (Zn), present predominantly as Zn²⁺, is an essential micronutrient for both plants and microbes. In plants, it contributes to enzyme function, redox regulation, and stress adaptation, and has also been linked to intracellular signaling and defense-related processes [26–28]. Despite their complementary properties, the agricultural deployment of these active agents, especially MN- and GA-containing combinations, remains constrained by several formulation challenges. These include the poor water dispersibility and photolability of MN, as well as the need to stabilize effective retention and persistence of the combined system at the plant-pathogen interface under field conditions [22, 29]. These considerations prompted us to ask whether the complementary chemical features and biological roles of MN, GA, and Zn²⁺ could be leveraged through ternary supramolecular co-assembly to improve interfacial availability while integrating multiple protective functions within a single field-relevant formulation [16].

Here, we combined semi-rational screening with supramolecular co-assembly to construct a ternary nanoformulation integrating GA, MN, and Zn²⁺ into a single nanoparticle system (G-M@Zn NPs). Using F. graminearum as a representative pathosystem while also evaluating antifungal activity in vitro against multiple phytopathogenic fungi, we investigated whether G-M@Zn NPs could provide direct antifungal efficacy and improved formulation persistence under field-relevant stresses while also being associated with host defense-related responses and microbiome shifts in the wheat-spike pathosystem. To define formulation effects on the pathogen, we examined membrane damage, mitochondria-linked oxidative stress, and related cellular injury in F. graminearum. To assess relevance within the FHB pathosystem, we performed wheat-spike infection assays and complemented them with coleoptile metabolomics under controlled conditions to capture treatment-associated host metabolic signatures. We further evaluated microbiome shifts on the wheat-spike surface. In parallel, we assessed biosafety across representative plant and non-target organism models and examined resistance-risk-related properties in F. graminearum to assess deployment-relevant characteristics of this ternary co-assembly.

Materials and methods

Preparation of G-M@Zn NPs

GA (10 mg) was dispersed in deionized water (10 mL) and heated at 80 °C under magnetic stirring for 20 min until a clear and homogeneous solution was obtained. Subsequently, ZnSO₄·7H₂O (1 mg, dissolved in deionized water) was added dropwise under continuous stirring, and the mixture was further stirred for 5 min to allow initial coordination. MN (2 mg) was dissolved in alkaline water (pH 9–10, adjusted with NaOH) to prepare a stock solution (2 mg/mL), and 1 mL of this solution was then added dropwise into the above mixture under constant stirring (600 rpm) to obtain the final GA: ZnSO₄·7H₂O: MN mass ratio of 5:0.5:1 (w/w/w). The resulting suspension was stirred for an additional 15 min to facilitate nanoparticle self-assembly. For pH screening, the pH of the mixture was adjusted to 3, 5, 7, or 9 prior to the self-assembly step, and pH 7 was selected for subsequent experiments. The mixture was then centrifuged at 12,000 × g for 15 min. The obtained precipitate was washed three times with deionized water (50 mL each time) under the same centrifugation conditions to remove unbound species, and the final pellet was freeze-dried to obtain G-M@Zn nanoparticles. The stepwise optimization workflow is summarized in Supplementary Tables S1-S4 and Fig. S1A-S1C.

Verification of sulfate removal

To assess the removal of residual free sulfate ions derived from ZnSO₄·7H₂O during purification, the supernatant from the final washing step was collected and analyzed using a BaCl₂ precipitation test. Briefly, a 5 mL aliquot of the washing supernatant was acidified with dilute HCl to pH 1–2, followed by addition of 0.5 mL of 0.1 M BaCl₂ solution. The mixture was allowed to stand at room temperature for 5–10 min. No visible white precipitate or turbidity was observed, indicating that free sulfate ions in the final washing supernatant were below the visual detection limit under the applied conditions. As a negative control, deionized water was treated identically with dilute HCl and 0.1 M BaCl₂. As a positive control, a dilute ZnSO₄ solution was treated in the same way to confirm the formation of visible white BaSO₄ precipitate.

Nanoparticle characterization and elemental analysis

Nanoparticle formation and physicochemical characteristics were verified by dynamic light scattering (particle size and polydispersity index), zeta-potential analysis, and electron microscopy (TEM/SEM), as described in the characterization section. The elemental composition of the purified G-M@Zn nanoparticles was further examined by energy-dispersive X-ray spectroscopy (EDS). No detectable sulfur signal was observed under the applied conditions.

Statistical analysis

Unless otherwise stated, technical replicates refer to repeated measurements of the same biological sample, biological replicates refer to independently prepared samples, cultures, or experimental units (for example, independent dishes or field plots), and independent experiments refer to complete repetitions of the same experiment performed on separate occasions. For antifungal plate assays, each treatment included three independent plates per experiment, and the entire experiment was independently repeated three times. DLS measurements were performed on three independently prepared samples, with three measurements per sample. RT-qPCR analyses included three biological replicates per treatment, each with three technical replicates. For microscopy-based analyses, three independent samples were examined per treatment, and five microscopic fields were quantified for each sample. Metabolomic analyses were performed using four biological replicates per treatment. For microbiome profiling, four spike samples were pooled to generate one composite biological replicate, and three composite biological replicates were analyzed per treatment [30, 31]. For field experiments, each treatment consisted of three replicate plots arranged in a randomized complete block design. For disease assessment, 50 spikes were scored per plot, one plot-level disease index was calculated for each plot, and the resulting three plot-level values were used for statistical analysis. Unless otherwise stated, data are presented as mean ± standard deviation (SD). For univariate data, statistical significance among multiple groups was assessed using one-way ANOVA in SPSS, followed by Tukey’s multiple-comparisons test. Differences were considered significant at P < 0.05.

For additional methods, see Supporting Information. Figures S1-S19 and Tables S1-S9 are also provided therein.

Results and discussion

Design rationale

The practical deployment of fungicides in the field has long been constrained by limited stability under agronomic conditions and insufficient delivery to pathogen-relevant interfaces [32, 33]. To address this challenge, we developed a semi-rationally optimized nanoformulation by integrating molecular complementarity, supramolecular co-assembly, and deployment-relevant design criteria within a single formulation. MN is a plant-derived lignan with broad antifungal activity in vitro [34], yet its practical utility is constrained by poor water dispersibility and by the difficulty of maintaining effective exposure on heterogeneous plant surfaces under light-exposed field conditions [19, 29, 35]. To address these limitations, we implemented a multiscale computational workflow to identify a co-assembly partner capable of supporting MN co-assembly and formulation stabilization. Screening a library of fifty natural small molecules prioritized GA as a co-assembly partner for further study. Although another candidate showed a slightly lower calculated interaction energy, candidate selection was not based on docking score alone. Instead, compounds were first screened for their potential to support co-assembly with MN and were then further prioritized by considering raw-material cost, source accessibility, preparation feasibility, and potentially beneficial functional attributes relevant to formulation design. Within this framework, GA was selected because it combined a highly favourable predicted interaction energy (-24.28 kcal/mol), practical accessibility, and an amphiphilic triterpenoid saponin scaffold well suited for supramolecular assembly and interfacial behavior (Fig. 1A, Table S5) [36]. Consistent with this selection logic, experimental evaluation of eight representative molecules showed that nanoparticle formation and interaction strength generally followed the computational ranking (Table S6). Based on this binary design, we next introduced Zn²⁺ as a coordination-active node to reinforce the assembly and potentially contribute additional biological functionality. Given its ability to form reversible coordination interactions with both MN and GA, Zn²⁺ is more reasonably interpreted as a coordination-active component associated with GA-MN domains rather than as a discrete outer layer. Accordingly, its incorporation was expected to enhance structural robustness and interfacial persistence [37]. In addition, Zn is an essential micronutrient for both plants and microbes and, in plants, has been linked to redox regulation, intracellular signaling, and defense-related processes [38, 39]. Together, this multi-scale and coordination-assisted design strategy provides a practical route for converting a deployment-limited antifungal natural product and a complementary amphiphilic co-assembly partner into a field-relevant nanofungicide with improved structural robustness, interfacial performance, and multifunctional potential.

Fig. 1.

Characterization of the co-assembled G-M@Zn NPs. (A) Interactions of MN with Compound 1 (1), Compound 10 (2), and GA (3) predicted by molecular docking. π-π interactions (pink) and hydrogen bonds (green) are indicated. (B) Particle size distributions of G-M@Zn NPs; (C) ζ-potentials of GA, MN, ZnSO₄ and G-M@Zn NPs. (D) TEM images of G-M@Zn NPs; (E) UV-vis spectra of GA, MN, and G-M@Zn NPs. (F) FT-IR spectra of GA, MN, GA-Zn and G-M@Zn NPs. (G) ¹H NMR spectra of GA, MN, and G-M@Zn NPs. (H) Molecular dynamics simulation of G-M@Zn NPs at different simulation time. GA, glycyrrhizic acid; MN, magnolol; G-M@Zn NPs, glycyrrhizic acid-magnolol@Zn nanoparticles

Synthesis of nanoparticles

Guided by this semi-rational supramolecular design strategy, we next synthesized glycyrrhizic acid- magnolol nanoparticles (G-M@Zn NPs) via a one-pot reaction of MN, GA and Zn²⁺ at 80 °C for 15 min. G-M@Zn NPs exhibited an average hydrodynamic diameter of 154.6 ± 3.27 nm, with a PDI of 0.215 ± 0.03 and a zeta potential of -30.27 mV (Fig. 1B and C), indicating the formation of a uniformly dispersed, colloidally stable nanosystem with a balanced hydrophobic-hydrophilic character. Transmission electron microscope (TEM) images further revealed that G-M@Zn NPs were spherical with an average diameter of ~ 139 nm, while EDS mapping confirmed that Zn was distributed throughout the nanoparticles (Fig. 1D and Fig. S2A-S2G). Such a ~ 140 nm spherical architecture may facilitate interfacial contact and more uniform surface deposition on plant tissues, thereby potentially contributing to the biological performance observed in subsequent antifungal assays [40].

Mechanism of G-M@Zn NPs formation

To elucidate the assembly mechanism of G-M@Zn NPs, we integrated spectroscopic analyses with molecular dynamics (MD) simulations to examine how noncovalent interactions and coordination chemistry cooperatively govern nanoparticle formation. In the MD simulations, the GA-MN-Zn system evolved from an initially dispersed configuration into a compact aggregate. The RMSD of the aggregate non-hydrogen atoms increased during the early stage of the simulation and then approached a plateau at ~ 3.0 nm (Fig. S3A). However, RMSD plateauing alone was not used as the sole criterion for convergence. Instead, the combined evolution of the radius of gyration (Rg), solvent-accessible surface area (SASA), nonbonded interaction energies, and periodic-boundary-corrected trajectory snapshots indicated progressive compaction of the assembly. In particular, Rg and SASA both decreased markedly during the early stage and showed no major systematic drift after ~ 20–40 ns (Fig. S3B and S3C). To further visualize this process, snapshots extracted every 20 ns from 0 to 100 ns were examined (Fig. 1H). The system underwent substantial compaction by ~ 40 ns, although local rearrangement toward a more continuous and densely packed aggregate remained evident at later time points. Together with the stabilization of nonbonded interaction energies, these observations indicate that the system approached a reasonably equilibrated assembled state after approximately 40 ns. Zn²⁺ ions were enriched around GA-MN-associated domains rather than remaining freely dispersed, consistent with a role for Zn²⁺ in reinforcing the supramolecular assembly [37]. In parallel, hydrogen-bond analysis showed that GA-water interactions dominate the hydrogen-bond network, whereas MN-water hydrogen bonds progressively decline over the simulation (Fig. S3D-S3I). This preferential hydration pattern suggests a heterogeneous supramolecular organization with spatially differentiated microenvironments, in which GA is enriched in the more hydrated interfacial region whereas MN is preferentially associated with more hydrophobic domains [41]. To quantify the driving forces for GA-MN association, we decomposed the non-bonded interaction energy into short-range Coulombic (Coul-SR, -561.5 ± 106.8 kJ/mol) and Lennard-Jones (LJ-SR, -1465.9 ± 84.2 kJ/mol) components (Fig. S3J), showing that dispersion-dominated hydrophobic interactions are the primary stabilizing contributions. Both terms became stable after ~ 40 ns, consistent with the aggregation timescale inferred from the trajectory analyses. A two-dimensional Gibbs free-energy surface constructed as a function of RMSD and Rg identified a global minimum conformation (Fig. S3K), in which the aromatic rings of MN packed against hydrophobic regions of GA through predominantly hydrophobic and π-associated contacts, providing a molecular-level picture of GA-MN co-assembly.

To experimentally assess these interaction motifs, we used a combination of UV-vis, FT-IR and NMR spectroscopy to examine possible aromatic interactions and GA-MN-Zn association in the assembled state. The principal UV absorption band of GA at ~ 258 nm originates from the conjugated chromophore of the triterpenoid saponin nucleus. Co-assembly with MN and Zn²⁺ induces a hypsochromic shift of this band to 249 nm, consistent with an altered electronic microenvironment upon co-assembly and Zn²⁺ incorporation (Fig. 1E). Fourier-transform infrared (FTIR) spectroscopy revealed characteristic bands of GA, including a broad O-H stretching band at 3435.22 cm^− 1, a C = O stretching band at 1743.65 cm^− 1, and carboxylate-associated bands at 1643.64 and 1469.97 cm^− 1 (Fig. 1F). In addition, the band at 1091.71 cm^− 1 was assigned to C-O vibrations of the glucuronic acid moieties in GA. Upon introducing Zn²⁺, the O-H stretching band shifted to 3406.29 cm^− 1, and the carboxylate (COO⁻) stretching bands exhibited pronounced changes in band shape and intensity, consistent with the carboxylate groups serving as important contributors to Zn²⁺ coordination. In the GA-MN-Zn co-assembled network, the O-H region became markedly broadened and shifted to lower wavenumbers, supporting the formation of an extended hydrogen-bonding network within the nanoparticles. Meanwhile, the aromatic C = C skeletal stretching band of MN shifted from 1496.76 cm^− 1 to 1485.19 cm^− 1 in G-M@Zn NPs, indicating an altered aromatic microenvironment consistent with possible π-π interactions upon co-assembly. Collectively, these spectral changes support that coordination, hydrogen bonding, and aromatic/hydrophobic associations cooperate to stabilize the supramolecular architecture of G-M@Zn nanoparticles. XPS further corroborated Zn incorporation and the presence of a Zn-O coordination environment in G-M@Zn NPs. The Zn 2p_3/2 /Zn 2p_1/2 doublet is clearly observed, and O 1s deconvolution reveals a Zn-O component in addition to organic oxygen species, supporting the presence of Zn in a Zn-oxygen coordination environment within the co-assembled nanoparticles (Fig. S4A-4 C). In addition, ¹H NMR spectroscopy was employed to interrogate the molecular-level association between GA and MN (Fig. 1G). Upon addition of GA, the phenolic -OH resonance of free MN (δ = 9.00 ppm) underwent pronounced exchange broadening and became undetectable, indicative of substantially accelerated proton exchange in DMSO-d₆ and consistent with hydrogen-bond-associated perturbation of the phenolic hydroxyl environment through interaction with GA’s oxygenated motifs. In parallel, GA resonances from the glucuronic acid region displayed reproducible perturbations (e.g., shifted from 3.54 to 3.52 ppm), and the anomeric H-1″ signal shifted from 4.40 to 4.35 ppm (Δδ = 0.05 ppm), supporting a GA-localized electronic-environment change upon complexation. Moreover, a subtle displacement in the MN aromatic region (shifted from 6.93 to 6.92 ppm, Δδ = 0.01 ppm) suggests an altered aromatic microenvironment, potentially reflecting weak aromatic/hydrophobic contacts in the associated state.

Overall, the combined simulations and spectroscopy support a cooperative assembly mechanism for G-M@Zn NPs, in which GA-MN association is driven primarily by dispersion-dominated hydrophobic/π-rich contacts, while Zn²⁺ reinforces the assembly through coordination with GA oxygenated groups. Together with the extended hydrogen-bonding network, these interactions provide a molecular-level basis for the enhanced structural robustness of the co-assembled nanoparticles.

Physicochemical properties and interfacial behavior of G-M@Zn nanoparticles

Although semi-rationally designed nanomaterials can be optimized at the molecular and supramolecular levels, their suitability for field crop protection ultimately depends on agriculturally relevant physicochemical properties such as responsive release, photostability, surface retention, and adhesion on plant tissues [42]. In our system, free MN exhibited rapid photodegradation under controlled near-UVA irradiation, which may limit its practical utility under light exposure [43]. Accordingly, we compared the photostability of free MN and G-M@Zn NPs under 365 nm irradiation. Free MN degraded rapidly, with only 15.5% remaining after 32 h. In contrast, G-M@Zn NPs retained 58.1% of MN over the same period. Pseudo-first-order kinetic fitting yielded half-lives of 12.6 h for free MN and 46.5 h for MN in G-M@Zn NPs, corresponding to an approximately 3.7-fold enhancement in photostability (Fig. 2A and B). These results indicate that supramolecular assembly markedly improved the resistance of MN to near-UVA-induced degradation, likely because the co-assembled matrix partially attenuated incident UV and reduced direct photoexposure of MN [44]. It should be noted that this assay was designed as a comparative near-UVA stress test rather than a full-spectrum simulation of natural solar irradiation.

Fig. 2.

Performance characteristics of G-M@Zn NPs. (A) Time-dependent residual MN content in free MN and G-M@Zn NPs during irradiation. (B) Pseudo-first-order kinetic fitting of MN photodegradation in free MN and G-M@Zn NPs. (C) Cumulative release of MN from G-M@Zn NPs in different media (pH 3, 5, 7). (D) Contact angle of GA, MN and G-M@Zn NPs. (E) Spreading performance of GA, MN and G-M@Zn NPs on wheat leaves. (F) Snapshots of droplet impact process on PTFE surface for water, MN and G-M@Zn NPs; scale bar, 2 mm. All droplets were released from a vertical distance of 15 cm. (G) SEM images of glumes after spray application. GA, glycyrrhizic acid; MN, magnolol; G-M@Zn NPs, glycyrrhizic acid-magnolol@Zn nanoparticles

Given that F. graminearum colonization of wheat spike tissues is often associated with local acidification (approximately pH 4–5), we next asked whether G-M@Zn NPs would exhibit relatively restrained release under near-neutral conditions while becoming more release-permissive under acidic infection-relevant conditions [45]. Using a dialysis-based release assay, G-M@Zn NPs showed higher MN release under acidic conditions, reaching 93.0% and 89.5% cumulative release at pH 3 and 5 within 48 h, compared with 70.2% at pH 7 (Fig. 2C). The release profiles were well described by a first-order model, consistent with diffusion-dominated transport under a concentration gradient (Table S7). This pH sensitivity may arise from protonation of GA, which weakens electrostatic interactions and/or Zn²⁺-mediated coordination that otherwise reinforce the co-assembled framework, thereby accelerating MN dissociation and outward transport [44].

On the other hand, foliar “locking” on healthy tissues must be balanced against efficient release at infection sites [46]. To quantify the basis for this “locked-on” behavior, we measured the wetting and interfacial properties of the formulations on healthy leaf surfaces. G-M@Zn NPs markedly improved leaf wetting, lowering the contact angle from 83.4–98.8° (water/MN/GA) to 22.0°, and concomitantly reducing surface tension from 44.9 to 52.2 to 39.8 mN/m (Fig. 2D and S5A). Together with the accelerated MN release observed under acidic conditions in vitro, these measurements support a putative “lock-and-release” functional profile, featuring improved surface retention on healthy leaf tissues together with greater release permissiveness under infection-relevant acidic conditions. Moreover, interfacial “locking” is governed not only by static wetting but also by transient droplet-surface interactions during spraying [47]. Many conventional fungicide formulations exhibit insufficient interfacial adhesion and energy dissipation on leaf surfaces, leading to substantial off-target losses via rebound, splashing, and run-off, which compromises on-leaf deposition and increases environmental burden [48]. Therefore, assessing spray-relevant adhesiveness and anti-splash performance under agriculturally meaningful conditions is essential [46]. To provide a simplified low-surface-energy model surface, PTFE was used as a hydrophobic reference interface. PTFE was intended to approximate the non-wetting character of waxy plant surfaces, rather than to reproduce the full microstructure or transport channels of the cuticular layer [49, 50]. High-speed imaging captured droplet impact behavior, and impact-dynamics descriptors including time to maximum spreading, rebound behavior, and splashing outcome were quantified to evaluate whether G-M@Zn NPs suppressed rebound and splashing on hydrophobic surfaces, behaviors expected to favor stronger interfacial retention under simplified spray-relevant hydrophobic conditions [46]. As shown in Fig. 2F and Videos S1, water and G-M@Zn NPs droplets, the latter applied at an MN-equivalent concentration of 100 µg/mL, reached their maximum spreading within a representative time of 2.2 ms, whereas MN at 100 µg/mL required a slightly longer time to reach maximum spreading (representative value: 2.6 ms). Notably, MN exhibited a pronounced irregular/serrated contact line (fingering-type contact-line instability) near the maximum-spreading regime, consistent with impact-induced interfacial instabilities on low-surface-energy substrates at high impact speeds. Subsequently, water and MN droplets underwent splashing followed by rebound and detachment from the surface within representative times of 17.0 ms and 18.2 ms, respectively. In sharp contrast, the G-M@Zn NPs droplet showed neither splashing nor rebound; instead, it maintained continuous contact with the substrate and displayed a progressively expanding deposition footprint, consistent with reduced off-surface liquid loss and sustained interfacial contact on the hydrophobic model interface.

To quantitatively resolve droplet-impact dynamics, we summarized the temporal evolution of the normalized spreading diameter Dt/D0 and the normalized rebound height Ht/D0, where D0, Dt and Ht denote the initial droplet diameter, the spreading diameter at time t and the vertical distance from the substrate surface to the droplet apex, respectively. For all liquids, Dt/D0 increased rapidly after impact to reach the maximum spreading diameter (Dmax), followed by a retraction stage; however, the retraction kinetics and final outcomes differed markedly. After reaching Dmax at 2.2 ms (water) and 2.6 ms (MN), the droplets continued to retract and their Dt/D0 values decayed to zero within 17.0 ms and 18.2 ms, respectively, indicating rebound and complete detachment from the PTFE surface. In sharp contrast, the G-M@Zn NPs droplet exhibited only minor recoil and its Dt/D0 did not decrease to zero, consistent with sustained contact and stable deposition on the substrate (Fig. S5B). Concordantly, the peak Ht/D0 values for water (0.82) and MN (1.13) were substantially higher than that for G-M@Zn NPs (0.22), indicating more efficient dissipation of retraction kinetic energy by G-M@Zn NPs (Fig. S5C). Collectively, these results demonstrate that G-M@Zn NPs effectively suppress droplet rebound and splashing on hydrophobic surfaces, thereby supporting stronger interfacial retention and suggesting a reduced propensity for off-surface liquid loss under this hydrophobic model condition.

In practical crop protection settings, such improved retention may be relevant to future efforts aimed at reducing re-application frequency or application rates, although this possibility requires direct field validation [51]. To account for variation in surface orientation under practical application conditions, spray deposition was evaluated across different tilt angles [52]. To isolate the role of surface inclination under controlled conditions, we used PTFE as a hydrophobic model interface and assessed droplet retention on substrates tilted at 30° and 60°. Increasing the tilt angle amplifies the downslope gravitational driving force (mg × sin θ); once this force exceeds the interfacial pinning/adhesion associated with contact-angle hysteresis, droplets become prone to sliding, coalescence and run-off, leading to liquid loss and reduced deposition efficiency [53]. Despite the enhanced run-off tendency at high inclination, G-M@Zn NPs still exhibited substantially higher retention on the tilted substrates than MN, consistent with improved apparent interfacial adhesion and greater resistance to gravity-driven losses (Fig. S6A and S6B, Videos S2 and S3).

Having established this trend on the PTFE model, we next assessed deposition on freshly excised wheat flag leaves collected at anthesis at a shallow inclination (15°). Under identical spray distance and 15° leaf inclination, free MN rapidly beaded into discrete droplets on the wheat leaf surface, whereas G-M@Zn NPs spread into a more continuous wetted film across the leaf surface, yielding a more continuous wetted footprint (Fig. 2E). This retention advantage remained evident at steeper inclinations: on wheat leaves tilted at 30° and 60°, G-M@Zn NPs showed significantly stronger surface retention than MN, consistent with enhanced resistance to gravity-driven run-off and improved on-leaf residence (Fig. S7 and S8, Videos S4 and S5).

To test whether interfacial locking translates into resistance against rain-driven wash-off, a field-relevant stress, we performed a simulated rain-wash assay to evaluate rainfastness [54]. As shown in Fig. S9, the non-washed controls showed no significant differences between MN (100 µg/mL) and G-M@Zn NPs at an MN-equivalent concentration of 100 µg/mL, indicating comparable initial deposition and good measurement reproducibility. After rain washing, only 21.6% of MN remained on the leaf surface, whereas 45.3% was retained when delivered as G-M@Zn NPs under identical conditions. Because the two formulations showed similar pre-wash recovery, the higher post-wash retention of G-M@Zn NPs is more likely attributable to stronger interfacial locking rather than differences in initial loading. This improved rainfastness may arise from the nanoscale formulation’s greater ability to associate with the leaf surface micro/nanotopography, together with the multiple polar groups of GA that could contribute to stronger interfacial interactions, thereby potentially reducing wash-off under flowing water [55].

In field settings, FHB is initiated at anthesis when airborne ascospores or conidia of F. graminearum land on susceptible floral tissues, germinate, and develop infection hyphae that colonize surfaces such as the glume and lemma [56]. To interrogate this infection-front interface, we used SEM to visualize the adhesion and surface distribution of G-M@Zn NPs on wheat glumes. The glume surface is decorated with regularly patterned papillae (~ 10 μm in diameter), and free MN beads on the waxy cuticle, failing to form a continuous deposit. In contrast, incorporation of GA enabled G-M@Zn NPs to form a dense, film-like coverage on the glume surface (Fig. 2G). This enhanced deposition is consistent with the interfacial properties of the GA-containing co-assembled formulation, as supported by the lower contact angle and reduced surface tension measured for G-M@Zn NPs relative to MN (Fig. 2D and Fig. S5A). Collectively, these observations highlight the potential of G-M@Zn NPs as a candidate formulation for FHB management.

Together, Together, improved near-UVA stability, stronger interfacial retention, enhanced rainfastness, and acid-enhanced release behavior provide a mechanistic basis for the improved antifungal performance of G-M@Zn NPs under agriculturally relevant stress conditions.

Antifungal activity of G-M@Zn NPs

Current control of F. graminearum remains largely reliant on chemical fungicides; however, their intensive use has accelerated resistance development, underscoring the need for more sustainable alternatives [13]. To assess whether G-M@Zn NPs assembled from plant-derived actives could serve as an alternative or complement to conventional fungicides, we quantified the in vitro antifungal activities using a colony-diameter assay. As a component control, GA@Zn was prepared at the same GA: Zn mass ratio as in G-M@Zn NPs but without MN. As shown in Fig. S10A, GA@Zn exhibited only limited antifungal activity even at a relatively high concentration, indicating that the GA@Zn scaffold alone contributed little direct fungistatic activity. By contrast, G-M@Zn NPs showed stronger inhibition than free MN in the colony assay (Fig. 3A; Table 1), with the EC₅₀ decreasing from 9.26 to 7.02 mg/L, which may be related to the sustained-release behavior of the nanoassembly [57]. Benchmarking against two registered fungicide formulations further placed the in vitro activity of G-M@Zn NPs below tebuconazole SC (EC₅₀ = 2.35 mg/L) but within a similar range to carbendazim SC (EC₅₀ = 5.91 mg/L) (Table 1). More notably, G-M@Zn NPs retained substantially higher antifungal activity than free MN after UV irradiation, likely owing to GA-mediated photoprotection (Fig. S10B). Furthermore, G-M@Zn NPs exhibited broad-spectrum antifungal efficacy across multiple genera, with strong inhibition against Magnaporthe oryzae, Verticillium dahliae, Botrytis cinerea, Fusarium sambucinum, and Alternaria alternata (Fig. 3A and Table S8). Collectively, these results suggest the broad-spectrum efficacy and UV-resilient activity of G-M@Zn NPs. Given the agricultural importance of F. graminearum as the causal agent of FHB and a major producer of deoxynivalenol, we next used it as a model pathogen to elucidate the antifungal mechanism of G-M@Zn NPs.

Fig. 3.

Inhibitory effects of MN and G-M@Zn NPs against Fusarium graminearum and other plant pathogenic fungi. (A) Inhibitory effects of MN and G-M@Zn NPs against F. graminearum, Magnaporthe oryzae, Verticillium dahliae, Botrytis cinerea, Fusarium sambucinum, and Alternaria alternata. (B) Scanning electron microscope (SEM) and propidium iodide (PI) staining images of F. graminearum hyphae exposed to sterile water (control), MN, and G-M@Zn NPs. GA, glycyrrhizic acid; MN, magnolol; G-M@Zn NPs, glycyrrhizic acid-magnolol@Zn nanoparticles

Table 1.

In vitro antifungal activities of magnolol and glycyrrhizic acid-magnolol@Zn co-assembled nanoparticles against F. graminearum

Pathogen	Treatments	Toxicity equation	r 2	EC50 (mg/L)	95% CIa (mg/L)
Fusarium graminearum	MN	y = 1.305x + 3.7389	0.9049	9.26	6.33–13.48
	G-M@Zn NPs	y = 0.9955x + 4.1573	0.9657	7.02	5.51–8.96
	Tebuconazole SCb	y = 0.4577x + 4.8302	0.9633	2.35	1.65–3.35
	Carbendazim SCc	y = 0.8087x + 4.3763	0.9493	5.91	4.00-8.72

^aCI, confidence interval. ^bTebuconazole SC, commercial tebuconazole suspension concentrate (43%). ^cCarbendazim SC, commercial carbendazim suspension concentrate (50%)

Antifungal mechanism of G-M@Zn NPs

SEM imaging of hyphae collected from liquid YEPD cultures revealed intact hyphae with smooth surfaces in the control group. Treatment with MN (25 µg/mL) caused surface roughening accompanied by localized swelling and irregular collapsed or constricted regions, whereas G-M@Zn NPs applied at an MN-equivalent concentration of 25 µg/mL induced more pronounced shrinkage and collapse (Fig. 3B). To determine whether these morphological alterations were associated with membrane damage, we performed propidium iodide (PI) staining, for which fluorescence serves as a marker of membrane permeabilization. In agreement with the morphological observations, the PI-positive area fraction increased from 0.09 ± 0.02% in the control group to 0.93 ± 0.25% after MN treatment and further to 1.98 ± 0.16% after G-M@Zn NPs treatment, indicating a graded increase in membrane permeability in F. graminearum (Fig. S11A). The control group showed negligible PI fluorescence, consistent with preserved membrane integrity, whereas MN treatment produced a detectable PI-positive signal, indicating membrane damage. By contrast, G-M@Zn NPs elicited significantly stronger PI uptake than MN, consistent with greater membrane permeabilization (Fig. 3B). These membrane-disruption phenotypes align with previous reports that MN perturbs fungal membrane homeostasis via reprogramming of lipid metabolism [19], and that GA has been reported to interact with membrane sterols and alter membrane permeability [55]. Together, these observations suggest that GA-containing co-assembly may enhance the membrane-disruptive effects associated with MN delivery, a response that was further accompanied by stronger mitochondria-associated oxidative-stress-related signals, as examined below.

We next used TEM to examine whether this surface and membrane damage was accompanied by intracellular ultrastructural disruption. Compared with the control, MN induced ultrastructural abnormalities including cytoplasmic condensation and increased structural heterogeneity, whereas G-M@Zn NPs caused more pronounced intracellular vacuolation and loss of ultrastructural organization (Fig. 4A). These ultrastructural changes were consistent with the stronger PI staining observed in the G-M@Zn NPs group, further supporting compromised cellular integrity. To determine whether such damage was associated with oxidative stress, we performed DCFH-DA staining, for which fluorescence is commonly used as a broad indicator of intracellular ROS-associated oxidative signal accumulation. Because GA is a structural component of the co-assembled nanoparticles, free GA was included as a component control in the oxidative-stress assays. Quantitatively, background-corrected mean fluorescence intensity increased from 1.57 ± 1.60 a.u. in the control group to 29.16 ± 0.78 a.u. after GA treatment (78.43 µg/mL), and rose further to 87.80 ± 2.31 and 92.94 ± 4.32 a.u. after MN (25 µg/mL) and G-M@Zn NPs treatment at an MN-equivalent concentration of 25 µg/mL, respectively (Fig. 4B and S11B). We further assessed mitochondria-associated oxidative perturbation using MitoSOX staining, a commonly used indicator of mitochondrial superoxide-associated fluorescence. MitoSOX fluorescence showed a similar pattern, increasing from 2.67 ± 2.52 a.u. in the control group to 28.39 ± 4.80, 64.67 ± 5.51 and 101.33 ± 4.16 a.u. after GA, MN and G-M@Zn NPs treatment, respectively, with the G-M@Zn NPs group showing the strongest signal (Fig. 4B and S11C). Together, these data suggest that G-M@Zn NPs induce a cellular injury profile characterized by enhanced membrane permeabilization, intracellular ultrastructural disruption, and stronger mitochondria-associated oxidative-stress-related signals in F. graminearum, which is consistent with their enhanced antifungal activity [58]. Having established a pathogen-directed cellular injury profile, we next examined host-associated responses in wheat spikes under field-relevant conditions by measuring malondialdehyde (MDA), a marker of lipid peroxidation and oxidative membrane injury, together with defense/antioxidant enzyme activities and stress-responsive physiological indices, including proline and soluble protein.

Fig. 4.

Mechanistic study of G-M@Zn NPs against F. graminearum. (A) Transmission electron microscopy (TEM) images of F. graminearum hyphal cells after different treatments. Representative ultrastructural features are labeled in the images, including the cell wall (CW), plasma membrane (PM), nucleus (N), vacuole (V), and mitochondrion (M). (B) CLSM images of F. graminearum showing intracellular ROS (DCFH-DA) and mitochondrial superoxide (MitoSOX). GA, glycyrrhizic acid; MN, magnolol; G-M@Zn NPs, glycyrrhizic acid-magnolol@Zn nanoparticles

G-M@Zn NPs confer field-relevant protection associated with defense-related responses in wheat spikes

Zn is an essential micronutrient involved in plant defense and redox homeostasis [27]. In plants, Zn status can influence ROS balance partly through Zn-dependent antioxidant systems, particularly Cu/Zn-superoxide dismutases, and appropriate Zn supplementation has often been associated with reduced oxidative damage under stress conditions [27, 59]. However, these effects are context- and dose-dependent, since both Zn deficiency and excess can perturb redox balance. Yet conventional Zn supplementation does not necessarily translate into robust disease resistance, underscoring the need for approaches that integrate spatially controlled micronutrient delivery with defence-related physiological responses to combat F. graminearum [38]. This pathogen causes FHB and primarily infects wheat spikes around anthesis [56], often severely reducing grain set and leading to blighted or empty spikelets [56]. In a randomized complete block field trial using spikelet spot inoculation, G-M@Zn NPs reduced the disease index to 22.3 ± 3.4 (mean ± SD, n = 3 plots), compared with 52.5 ± 2.5 in the infected control; the corresponding values for the GA, MN, and ZnSO₄ treatments were 43.2 ± 4.0, 30.2 ± 3.9, and 41.0 ± 6.1, respectively (Fig. 5A and F and S12A). The greater reduction in disease severity achieved by G-M@Zn NPs than by free MN suggests that their field performance is unlikely to reflect the intrinsic antifungal activity of MN alone. We therefore examined whether this protection was associated with reduced oxidative injury and coordinated changes in defense-related readouts in wheat spikes.

Fig. 5.

G-M@Zn NPs trigger plant defense responses against F. graminearum. (A–F) Representative phenotypes of wheat spikes under different treatments, including healthy control, infected control, free GA (313.7 µg/mL), free ZnSO₄·7H₂O (46.6 µg/mL, corresponding to 10.6 µg/mL Zn), free MN (100 µg/mL), and G-M@Zn NPs applied at an MN-equivalent concentration of 100 µg/mL, corresponding to approximately 313.7 µg/mL GA and 10.6 µg/mL Zn. Representative spike images were recorded at 21 dpi. (G) Malondialdehyde (MDA) content. (H) Superoxide dismutase (SOD) activity. (I) Phenylalanine ammonia-lyase (PAL) activity. (J) Polyphenol oxidase (PPO) activity. (K) Peroxidase (POD) activity. (L) Catalase (CAT) activity in wheat spikes from the indicated groups at 3dpi. Data are presented as mean ± s.d. (n = 10 spikes were sampled per treatment from the field experiment). Different letters indicate significant differences (one-way ANOVA followed by Tukey’s test, P < 0.05). GA, glycyrrhizic acid; MN, magnolol; G-M@Zn NPs, glycyrrhizic acid-magnolol-Zn nanoparticles

Protection was associated with reduced lipid peroxidation and coordinated changes in complementary defense-related readouts

To assess whether field protection was associated with reduced oxidative injury in wheat spikes, we first measured MDA levels across the inoculated treatments. Among the inoculated groups, MDA levels were lowest in wheat spikes treated with G-M@Zn NPs (11.4 ± 1.2 nmol/g FW), significantly lower than those in the infected control and the GA, MN, and ZnSO₄ treatments, consistent with attenuated infection-associated lipid peroxidation and oxidative membrane injury (Fig. 5G). We next quantified the activities of five defense-related enzymes at 3 days post inoculation (dpi). These enzymes represent complementary defense readouts, including ROS detoxification (SOD and CAT), phenylpropanoid-associated responses (PAL), peroxide-driven oxidative defenses (POD), and phenolic oxidation-related defense responses (PPO) [60]. As shown in Fig. 5H-L, GA and ZnSO₄ showed limited direct growth inhibition in plate assays but still increased defense-related enzyme activities in planta, indicating that host-associated physiological responses can be detected even in treatments with limited direct fungistatic activity. This comparison provides an in planta reference for interpreting the stronger and more coordinated changes in defense-related readouts associated with G-M@Zn NPs treatment. G-M@Zn NPs were associated with the broadest enhancement across both antioxidant and phenylpropanoid-linked defense readouts, while simultaneously lowering MDA levels—a combination consistent with reduced oxidative injury and a more balanced redox-related response under infection. This response pattern was further accompanied by increased soluble protein accumulation and elevated proline levels in wheat spikes, consistent with altered stress adaptation and osmoprotective responses (Fig. S12B and S12C). Collectively, these readouts indicate that the nanoformulation was associated with stronger coordination of defense-related metabolic and redox responses than free GA or Zn salts, potentially linked to improved local Zn availability and retention at infection-relevant interfaces [61]. Since FHB initiates and progresses on the spike surface, we next profiled spike phyllosphere communities to examine whether this response pattern coincides with a remodeling of the surface microenvironment [62].

G-M@Zn NPs are associated with early spike-surface microbiome features related to disease and yield outcomes

At 3 dpi, G-M@Zn NPs treatment was already associated with early shifts in spike-surface microbiome configuration. Because this time point was chosen to capture early treatment-associated responses during F. graminearum infection, the dataset is best interpreted as an early-phase community snapshot rather than a stabilized microbiome [31]. Profiling of spike-surface bacterial (16 S rRNA) and fungal (ITS) communities revealed kingdom-specific patterns: bacterial alpha-diversity metrics showed a tendency toward higher Shannon diversity and lower richness (lower Chao1), whereas fungal alpha-diversity metrics showed a tendency toward lower Shannon diversity with no clear change in richness (Fig. 6A and D). Bray-Curtis PCoA indicated treatment-associated shifts in both bacteriome and mycobiome composition, with a more consistent displacement observed for bacteria than for fungi; however, PERMANOVA did not detect a statistically significant overall between-group difference under the current sample size. PERMDISP further showed a significant reduction in bacterial within-group dispersion under G-M@Zn NPs treatment, whereas fungal dispersion did not differ significantly between groups (Fig. 6K and N). Taxonomically, infection was accompanied by a Fusarium-dominated mycobiome, whereas G-M@Zn NPs were associated with a lower relative representation of Fusarium and coincided with increased representation of genera such as Cladosporium, which was interpreted conservatively at the genus level because this taxon includes species with diverse ecological roles (Fig. 6E, G and H). In the bacterial community, treated spikes showed enrichment of Pantoea—identified by LEfSe (including Pantoea vagans) as a discriminant feature of the treatment group—while infected spikes were associated with taxa such as Buchnera/Buchnera aphidicola, Candidatus Regiella, and Exiguobacterium (Fig. 6F, I and J). Notably, the enrichment of Pantoea is consistent with prior reports that some Pantoea spp. antagonize F. graminearum via secreted bioactives (e.g., herbicolin A) targeting ergosterol-containing membrane microdomains [25]. Importantly, several taxa highlighted by LEfSe also showed exploratory cross-time associations. When their relative abundances at 3 dpi were correlated against disease index and 100-grain weight measured at 21 dpi, disease-enriched lineages tended to correlate positively with later disease severity and negatively with later yield-related performance, whereas multiple treatment-enriched lineages showed the opposite pattern (Spearman; ^*P < 0.05, ^**P < 0.01) (Fig. 6O and P, S13A and S13B). Collectively, these data suggest that G-M@Zn NPs treatment was associated with early shifts in the spike-surface microbiome, including reduced bacterial community dispersion and reduced Fusarium representation, which may be linked to reduced disease pressure and improved yield-related outcomes. One possible interpretation is that improved interfacial persistence of the formulation may influence the local infection-associated surface environment, but direct evidence linking formulation retention, tissue status, and microbiome restructuring is still lacking.

Fig. 6.

Field microbiome responses associated with G-M@Zn NPs treatment during F. graminearum infection. Spike-surface microbiomes were profiled at 3 dpi, representing an early acute-phase community state rather than a fully stabilized endpoint community. (A-D) Fungal and bacterial alpha-diversity indices (Chao1 and Shannon, OTU level); (E, F) Genus-level community composition; (G-J) Relative abundances of selected genera; (K, L) PCoA of fungal and bacterial communities based on Bray-Curtis dissimilarity; group differences were evaluated by PERMANOVA, and the corresponding R² and P values are shown in the plots. (M, N) Box plots showing within-group community dissimilarity based on Bray-Curtis distances for fungal (M) and bacterial (N) communities at the OTU level; group distributions were compared using a two-sided Wilcoxon rank-sum test; (O, P) Spearman correlation heatmaps showing the associations between the relative abundances of selected bacterial taxa (O) or fungal taxa (P) at 3 dpi and disease index or 100-grain weight measured at 21 dpi. Red and blue indicate positive and negative correlations, respectively. Asterisks indicate significance levels of the correlations (^*P < 0.05, ^**P < 0.01). Data points represent three composite biological replicates per group, each generated by pooling four spikes

A synchronized coleoptile infection model reveals treatment-associated metabolic shifts during G-M@Zn NPs-mediated protection

To disentangle treatment-associated metabolic responses from the combined effects of field heterogeneity, spike-surface microbiome variation, and asynchronous infection progression, we next performed untargeted metabolomics in a synchronized wheat coleoptile infection system. Although coleoptiles do not phenocopy spike-based head blight symptoms, they provide a reproducible and experimentally tractable wheat-F. graminearum interaction model for resolving early plant metabolic responses under controlled conditions [63]. We therefore used this assay as a complementary mechanistic system rather than a direct substitute for spike-based head blight phenotyping [64]. PCA showed a pronounced infection-driven displacement from healthy controls, whereas nanoformulation-treated samples shifted partially back toward the healthy metabolic space (PC1 = 47.1%, PC2 = 13.0%), consistent with a partial shift toward, but not full restoration of, the healthy metabolic profile (Fig. 7A). A complementary supervised PLS-DA analysis showed a consistent group-separation pattern (Fig. S14). Overall between-group differences in metabolite profiles were further supported by adonis (PERMANOVA; R² = 0.5074, P = 0.001). Differential metabolites were defined using VIP > 1, FDR (q) < 0.05 and |log₂FC| > log₂ (1.2), with exogenous/artefactual annotations removed prior to set and pathway analyses. Infection induced broad metabolic remodeling, dominated by upregulation (149 up versus 20 down) (Fig. 7B). In contrast, treatment on the infected background produced a more selective counter-shift (12 up versus 22 down) (Fig. 7C). Notably, treated tissues remained distinct from uninfected controls (99 up versus 14 down) (Fig. 7D), consistent with the establishment of a treatment-associated metabolic state rather than a simple return to baseline.

Fig. 7.

Coleoptile metabolomic responses to F. graminearum infection and G-M@Zn NPs treatment. (A) PCA score plot. (B-D) Volcano plots of differential metabolites identified using VIP > 1, false-discovery-rate-adjusted q < 0.05, and |log₂FC| > log₂(1.2). (E) KEGG enrichment of differential metabolites, with pathways ranked by adjusted P values. (F) Heatmap of selected metabolites (Infected vs. Healthy). (G-I) Box plots of representative metabolites. n = 4 biologically independent samples per group. Statistical significance in (G-I) was assessed by one-way ANOVA followed by Tukey’s multiple-comparison test

In the Infected vs. Healthy contrast, KEGG enrichment indicated broad remodeling of defense-related metabolism, with prominent signals in tryptophan metabolism and the biosynthesis of various plant secondary metabolites, alongside pathways such as flavonoid biosynthesis, plant hormone signal transduction, ascorbate and aldarate metabolism, and phenylalanine metabolism (ranked by adjusted P values) (Fig. 7E). To focus pathway analysis on metabolites that were both infection-associated and treatment-responsive, we defined a treatment-responsive, infection-associated subset by combining the overlap between Infected vs. Healthy and G-M@Zn NPs vs. Infected (14 metabolites) with the three-way intersection shared across all comparisons (11 metabolites), yielding a 25-metabolite panel (Fig. S15A). KEGG enrichment of this panel highlighted two prominent pathways—tryptophan metabolism and biosynthesis of various plant secondary metabolites (Fig. S15B)—suggesting that G-M@Zn NPs treatment preferentially affected a focused subset of infection-perturbed metabolic pathways rather than globally resetting the metabolome. Consistent with this interpretation, HeatmapTree visualization of representative differential metabolites revealed a coherent infection-associated metabolic signature across replicates, including elevated indole/tryptophan-associated features (e.g., N-methyltryptamine and an indole conjugate) and phenylpropanoid-related metabolites (e.g., coumarin and isoscopoletin), some of which may be involved in signaling during host–pathogen interactions (Fig. 7F). This infection-associated pattern was partially attenuated in the treated samples. Notably, several lipid and conjugated features followed the same directional trend in the heatmap, highlighting classes that are incompletely captured by KEGG mapping but may nonetheless track infection-associated membrane and oxidative-stress states.

Within this 25-metabolite panel (Table S9), indole/tryptophan-associated features—including N-methyltryptamine and an IAA-related indole conjugate (indole-3-acetyl-L-aspartic acid)—together with phenylpropanoid-linked metabolites (coumarin, isoscopoletin, methyl 4-hydroxycinnamate, and coumaramido ferulamidobutane) showed infection-associated increases that were consistently attenuated by treatment (Fig. 7G-I; Fig. S16A-S16C). In parallel, several membrane-associated lipid signatures—including a ceramide-like feature (Pe-Cer(16:3_2O/26:6)), phosphatidic acids (PA(20:4/22:4) and PA(6-keto-PGF1α/P-16:0)), and a fatty acyl lysine (N-eicosapentaenoyl lysine)—followed the same directional shift (Fig. S16D-S16G), consistent with membrane-associated and redox-stress-related metabolic perturbations.

Considered together with the reduced MDA levels and altered defense-related enzyme activities observed in field spikes, these metabolomic patterns are consistent with a treatment-associated attenuation of infection-linked physiological disruption. The wheat-spike microbiome shifts observed in parallel may reflect changes in the local infection-associated surface environment, although the underlying ecological drivers were not directly resolved in this study [65]. More broadly, these findings support an association between formulation persistence, attenuated infection-associated metabolic disruption, and host- or surface-associated response patterns, while the causal relationships among these layers require further investigation [66, 67].

Effects of G-M@Zn NPs on DON accumulation and TRI-related gene expression

To determine whether this protective pattern was also accompanied by suppression of fungal toxigenic potential, we next quantified DON accumulation and the expression of key TRI biosynthetic genes under toxin-inducing conditions. DON is a major virulence-associated mycotoxin produced by F. graminearum and a key contributor to the food-safety risk associated with FHB [13]. Under toxin-inducing culture conditions, the DON content in the control group reached 429.4 ± 18.98 µg/g. Compared with the control, treatment with MN at 25 µg/mL reduced DON accumulation to 264.1 ± 29.62 µg/g, whereas G-M@Zn NPs applied at the equivalent MN concentration further decreased DON content to 183.2 ± 11.73 µg/g, indicating a stronger inhibitory effect of G-M@Zn NPs on DON accumulation than free MN (Fig. S17A). To further examine whether this reduction in DON accumulation was accompanied by transcriptional changes in the fungal toxigenic program, we next quantified the expression of key TRI biosynthetic genes by RT-qPCR. GA@Zn alone had little effect on the expression of TRI5 and TRI6 relative to the control, whereas both MN and G-M@Zn NPs significantly downregulated these genes. For TRI10, G-M@Zn NPs produced the lowest transcript level among the tested treatments, while MN showed a partial reduction (Fig. S17B). Given that TRI5 encodes trichodiene synthase, which catalyses the committed step in DON biosynthesis, and that TRI6 and TRI10 function as key transcriptional regulators of the TRI gene cluster, these transcriptional patterns are consistent with the reduced DON accumulation observed under G-M@Zn NP treatment and support attenuation of the TRI-associated transcriptional response under toxin-inducing conditions [56].

Biosafety and resistance liability

As a plant-derived, water-based formulation intended for agricultural use, we next evaluated its crop compatibility. Using broad bean (Vicia faba L.) as a dicot model, we found that G-M@Zn NPs did not reduce germination, with rates comparable to the control (Fig. 8B). In wheat, a monocot model, G-M@Zn NPs did not suppress early seedling growth, with seedling height remaining comparable to the control after 7 d treatment (Fig. 8A, C and E). The contents of chlorophyll a, chlorophyll b, and total chlorophyll in wheat leaves were likewise maintained rather than depleted (Fig. 8F). Representative seedling phenotypes further showed no visible growth retardation or developmental abnormality in the G-M@Zn NPs group over the early observation period. Together, these data indicate good plant compatibility under the tested conditions. We next extended the safety assessment to a non-target soil organism. In an earthworm filter-paper contact screening assay, no mortality was observed after 48 h exposure at 1.2 mg/cm², indicating that the LC₅₀ exceeded the highest tested dose under the screening conditions used here (Fig. S18) [46].

Fig. 8.

Phytotoxicity assessment of G-M@Zn NPs in early plant growth. (A) Schematic of the germination and seedling-growth assays; (B) Germination percentage of Vicia faba. (C) Germination percentage of wheat; (D) Representative wheat seedling phenotypes under the indicated treatments; (E) Wheat seedling length; (F) Chlorophyll a, chlorophyll b and total chlorophyll contents in wheat. Data are mean ± s.d.; different letters indicate significant differences (one-way ANOVA followed by Tukey’s test, P < 0.05). GA, glycyrrhizic acid; MN, magnolol; G-M@Zn NPs: glycyrrhizic acid-magnolol@Zn nanoparticles

To assess the stability of antifungal sensitivity under sustained selection pressure, F. graminearum was serially passaged for eight generations in the presence of the formulation (25 µg/mL). Over this period, only a modest shift in sensitivity was observed, with the EC₅₀ increasing from 7.02 to 9.42 µg/mL (approximately 1.34-fold; Fig. S19). These results indicate no evidence of rapid resistance escalation over the tested period. However, because the number of passages was limited, they do not support a strong conclusion regarding long-term resistance risk under agricultural conditions. Even so, short-term serial-selection assays can still provide an informative first-pass readout of phenotypic stability under defined selection pressure, while durability under field conditions requires longer-term selection experiments and field monitoring. Viewed together with the multi-component composition of G-M@Zn NPs and the multiple cellular injury phenotypes described above, the limited sensitivity shift observed during short-term passaging is compatible with a formulation whose antifungal effects may not depend on a single canonical molecular target [68–71]. Whether such multi-component inhibitory features translate into slower resistance emergence than conventional single-target fungicides will require direct comparative evolution experiments. In this sense, the present results provide a preliminary rationale for further evaluating multi-component cooperative formulations as a strategy for limiting rapid resistance emergence.

Projected field Zn loading and deployment considerations

Although Zn is an essential micronutrient and foliar Zn application has been widely explored and applied in wheat production, incorporation of a Zn-containing component into a nanoformulation still warrants environmental consideration [72]. Repeated field application could, in principle, contribute to cumulative soil Zn loading. Based on the recorded spray consumption in our plot trial, the applied spray volume was approximately 111.1 L/ha, corresponding to a projected Zn input of ~ 1.18 g Zn/ha per application. This spray volume is comparable to those commonly used in field studies of Fusarium head blight management in wheat, which are often reported around 112–150 L/ha [73]. Moreover, the projected Zn loading in our system is substantially lower than the kg/ha-scale Zn inputs commonly used in agronomic Zn biofortification studies in wheat. At the same time, previous work has shown that the long-term effects of Zn-based nanomaterials in soil-wheat systems, as well as their transformation, speciation and bioavailability in soils, require explicit field-based evaluation rather than assumption from composition alone [74]. Thus, although the present data do not support a strong conclusion regarding long-term environmental safety, they do indicate that per-application Zn loading was low relative to commonly reported agronomic Zn inputs under our plot conditions. Likewise, although the in vitro comparison with commercial fungicide formulations cannot by itself define realistic field competitiveness, it provides an important first benchmark for the intrinsic antifungal potency of G-M@Zn NPs relative to registered products. Considered together with the disease suppression observed in our plot trial, these results support further consideration of G-M@Zn NPs as a candidate formulation for FHB management. Direct head-to-head comparison with registered fungicides on plant surfaces and under field conditions will nevertheless be important for defining its deployment value under realistic disease-management scenarios. More broadly, this combined efficacy-loading-fate perspective may provide a useful framework for evaluating other metal-containing agro-nanoformulations, for which in vitro potency, field-scale performance and long-term environmental fate should be assessed in an integrated manner [75].

Conclusion

Extant crop protection is constrained less by a shortage of potent antifungal molecules than by a persistent mismatch between molecular activity and field performance [5]. Many active ingredients that perform well in vitro fail to deliver consistent field efficacy because sustained exposure at plant-pathogen interfaces is difficult to establish and maintain: it is limited by uneven retention across heterogeneous crop tissues and further eroded by photodegradation and wash-off [2, 35]. The hydrophobic, wax-structured surfaces of wheat spikes, including glumes, impose an additional barrier to effective deposition by limiting droplet wetting and adhesion; impact-driven rebound, splashing, and run-off compound these losses. Together, these processes reduce on-target deposition and shorten residence time [76]. Compensatory practices—such as repeated applications and higher doses—raise production costs, increase environmental load, and intensify selection pressure for resistance, thereby undermining durable control and the goals of sustainable IPM [77]. In the FHB pathosystem, this mismatch is further compounded by the need to mitigate not only disease development but also mycotoxin accumulation.

To address these constraints, we developed a coordination-assisted supramolecular co-assembly strategy to integrate three functionally complementary components—MN, GA, and Zn²⁺—into a single, water-processable nanoformulation (G-M@Zn NPs). Guided by semi-rational computational screening, GA was selected as an effective co-assembly partner for MN. Non-covalent interactions—notably dispersion/π-π contacts and an extended hydrogen-bonding network—collectively stabilize the co-assembled architecture, offering a plausible explanation for the enhanced stability and aqueous processability of MN [78]. The introduction of Zn²⁺ as a coordination-guided “locking node” was designed to reinforce assembly integrity while also introducing a micronutrient component with potential physiological relevance [55]. Reversible Zn-O coordination, together with the extended hydrogen-bonding network, yields a compact nanostructure that remains stable in water yet responds to infection-associated acidity, aligning with a “lock-and-release” design logic.

This supramolecular architecture was subsequently evaluated for field-relevant physicochemical and interfacial behaviors at plant-pathogen interfaces. Through measurements of UV durability, rainfastness, wetting, interfacial adhesion, and spray impact (rebound and splashing) outcomes, we aimed to quantify the formulation’s ability to reduce interfacial losses on hydrophobic tissues. Importantly, its lock-and-release behavior is consistent with a profile that combines improved retention on healthy, hydrophobic surfaces with accelerated release under acidic, infection-like conditions. This increases the likelihood that active ingredients remain present where and when they are most needed, particularly on the spike surface around anthesis where FHB initiates and progresses.

Beyond direct fungistasis, the data suggest that the Zn-containing nanoformulation was associated with reduced oxidative damage and coordinated changes in defense-related physiological and metabolic readouts at the infection interface [66]. In wheat spike assays, treatment reduced lipid peroxidation and was associated with coordinated changes across antioxidant- and phenylpropanoid-/oxidative-defense-related readouts. Untargeted metabolomics in a synchronized coleoptile infection system further revealed selective attenuation of infection-driven metabolic reprogramming, with pathway mapping highlighting defense-linked axes centered on tryptophan/indole metabolism and secondary metabolite biosynthesis. Concurrently, the spike surface microbiome showed treatment-associated features that were exploratorily associated with later disease severity and yield-related traits. Taken together, these results are consistent with a model in which interfacial persistence of G-M@Zn NPs may alleviate infection-associated oxidative and metabolic burden while supporting a less disrupted physiological profile in infected tissues; whether the accompanying microbiome shifts reflect a downstream ecological consequence of improved surface condition, contribute to protection, or both, remains unresolved. Separately, under toxin-inducing conditions, G-M@Zn NPs reduced DON accumulation and repressed key TRI biosynthetic genes, indicating that the formulation may also constrain fungal toxigenic output in addition to limiting disease development [79]. Collectively, improved on-target deposition and durability under rain and UV, together with the observed disease- and toxin-suppression effects, warrant future field optimization studies to determine whether dose or application frequency can be reduced without compromising disease and toxin control. If confirmed, such improvements could help lower input demand and off-target losses while supporting food-safety-relevant FHB management and broader IPM objectives [8].

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Chenguo Li: Writing – original draft, Methodology, Investigation, Conceptualization. Chaojian Wang: Methodology, Investigation. Shimin Wu: Investigation. Kaishun Di: Methodology, Investigation. Qiuyu Zhu: Methodology, Investigation. Xiaoge Yang: Methodology. Mengna Wang: Methodology, Investigation. Zhicheng Wang: Methodology. Guangyue Li: Methodology. Pingfang Tian: Writing – review & editing, Supervision, Resources, Methodology, Funding acquisition, Conceptualization.

Funding

National Key Research and Development Program of China (2023YFA0914700).

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

We confirm that all methods involving animals, plants, and related materials were performed in accordance with relevant institutional, national, and international guidelines and regulations.

Competing interests

The authors declare no competing interests.