Structural Basis of Amyloid Fibril Assembly by Plant Seed Storage Proteins Authors: Yiling Zhang, Danni Li, Qinyue Zhao, Shiran Lv, Wencheng Xia, Yongyi Xu, Yingshan Li, Yunxia Li, Yaoyang Zhang, Dan Li, Bo Sun, Wenyan Kang, Cong Liu, Bin Dai Published: 2026 ABSTRACT Amyloid fibrils are highly ordered protein assemblies characterized by a cross-β architecture. A wide range of proteins can adopt amyloid states, contributing to both disease-related pathology and normal physiological function. While animal-derived amyloids have been extensively examined in atomic detail, amyloid fibrils formed by plant proteins remain relatively understudied. Here we systematically assess three major seed storage proteins—oat 12S globulin, soybean 7S globulin, and rice glutelin—under harsh cooking-like conditions (pH 2, 85 °C). Oat globulin and rice glutelin readily form fibrils in both purified preparations and whole-seed extracts, whereas soybean globulin forms fibrils only in purified preparations and remains largely amorphous in whole-seed extracts. Using cryo-electron microscopy, we determine the structure of oat globulin fibrils at 3.9 Å resolution. The fibril core adopts a compact triangular architecture with pseudo-threefold symmetry and is stabilized by extensive hydrophobic and aromatic packing. Our findings establish the molecular basis of amyloid formation in plant seeds and expand the structural landscape of amyloid fibrils beyond animal and microbial systems, providing a foundation for understanding amyloid formation in plant- and food-derived proteins. Introduction The structure of a protein fundamentally determines its biological function1–4. In addition to their native conformations, certain proteins can assemble into highly ordered amyloid fibrils, which are defined by a cross-β architecture stabilized by intermolecular hydrogen bonds5–8. Such fibrils exhibit remarkable stability, resistance to proteolytic degradation, and specific binding to diagnostic dyes such as Thioflavin T (ThT) and Congo Red (CR)9–16. Pathological amyloid fibrils are hallmarks of neurodegenerative diseases such as Alzheimer’s disease (AD) and Parkinson’s disease (PD), whereas functional amyloids fulfill diverse physiological roles17–20. Owing to their robust mechanical properties and intrinsic capacity for self-assembly, amyloid fibrils have attracted increasing interest as versatile building blocks in materials science, biomedicine, and even the food industry21–24. In recent years, fibrillar protein aggregates formed in food systems have emerged as an intriguing source of amyloid-like materials25,26. Food-derived amyloid fibrils can be induced by physicochemical treatments, including elevated temperature, acidic pH, or enzymatic hydrolysis27. Notably, these fibrils exhibit key structural features of classical amyloids, and have demonstrated functionality in food applications such as controlled nutrient release and texture modification, and functional food design. In addition, recent work has further supported the potential of oat protein nanofibrils in nutrition-related applications and has shown that food amyloid fibrils can be degraded under simulated gastrointestinal conditions28–30. Furthermore, their biocompatibility, low cost, and abundance make them attractive candidates for developing sustainable amyloid-based materials31–34. However, detailed structural and mechanistic insights into food-derived amyloid fibrils remain scarce. In particular, despite their abundance in food matrices, plant proteins have been little explored with respect to their molecular-level fibrillation behavior and structural features. This knowledge gap is striking considering that amyloids from animal and microbial sources have been extensively structurally characterized35,36. While some plant proteins are known to form amyloid-like aggregates, the structural basis of their fibrillation remains largely uncharacterized37. Plant proteins often possess distinct sequences and folds that could lead to distinct assembly pathways and fibril morphologies38. Given the inherent polymorphism of amyloid fibrils, key properties such as their stability, aggregation kinetics, and molecular interactions are tightly dictated by their underlying architecture35. Thus, resolving the atomic-level structure of a plant-derived fibril is critical for deciphering its assembly mechanism and for guiding the design of sustainable, functional amyloid materials in both biological and materials science contexts. In this study, we examine the amyloid-forming propensity of seed storage proteins from oat, soybean, and rice under acidic and high-temperature conditions that mimic harsh food-processing environments39. Recent reports have revealed the atomic-level fibril structures formed by the AI172 protein from Oryza sativa japonica and by subunits of soybean 11S globulin, further supporting that structurally ordered amyloid assemblies can arise from diverse plant proteins, while also underscoring the molecular diversity of plant-derived amyloid architectures40,41. Oat globulin readily forms β-sheet-rich amyloid fibrils in vitro, while rice glutelin and soybean globulin also assemble into fibrillar structures under the conditions tested, exhibiting structural and spectroscopic features consistent with amyloids. We then focus on the oat 12S globulin, determine the cryo-electron microscopy (cryo-EM) structure of its fibrils at high resolution. The resulting model reveals a compact triangular β-sheet core stabilized by interlocking hydrophobic interactions, uncovering a previously uncharacterized amyloid architecture in plant storage proteins. Our findings highlight seed storage proteins as a structurally diverse class of plant amyloid assemblies and support the potential use of plant-derived proteins in sustainable material development. Results Fibrillation of plant seeds and purified seed storage proteins We first tested whether whole seeds could form amyloid fibrils under harsh, cooking-like conditions. Oat, rice, and soybean seeds were incubated at 85 °C across a range of pH values (pH 2-8)42, then examined by negative-staining transmission electron microscopy (NS-TEM) (Fig. 1A). After 10 h of incubation at pH 2, abundant fibrillar structures were observed in the oat and rice seed samples, whereas no fibrils formed in the soybean samples (Fig. 1B, Fig. S1A). Consistent with these observations, SDS-PAGE analysis revealed time-dependent proteolysis of oat seed proteins at pH 2, with intact protein bands gradually diminishing and polypeptide fragments below ~14 kDa appearing over time (Fig. 1C, Fig. S1B). By contrast, samples incubated at pH values of approximately 4, 6, and 8 exhibited only minor degradation or remained largely intact. Mass spectrometry identified the dominant remaining band in the oat sample as 12S globulin, accounting for ~62% of total protein content, suggesting that peptides derived from this globulin are the main contributors to fibril assembly (Fig. 1D, E, Fig. S2). At pH 2, soybean seed proteins showed little evidence of proteolysis and failed to form fibrillar structures (Fig. S3A, B). By comparison, rice seeds underwent partial protein degradation and displayed limited fibril formation (Fig. S3C, D). The inability of soybean globulin to form fibrils may be attributed to its comparatively high lipid content, which could hinder amyloid aggregation, consistent with reported effects of lipid content on protein solubility and aggregation43–46.Fig. 1Oat, soybean, and rice seeds form fibrillar aggregates after 10 h of cooking.A Schematic diagram of the simulated cooking process of oat/soybean/rice seed. Oat, soybean, and rice seeds incubated under cooking-like conditions at pH ≈ 2, 4, 6, and 8 for 10 h, characterized by NS-TEM. Created in BioRender. Liu, C. (2026) https://BioRender.com/tdwatdq. B Effects of different pH conditions on fibrillation of oat/soybean/rice seeds. NS-TEM micrographs of samples. Scale bar, 500 nm. White arrows denote the fibrillar structures formed. Source data are provided as a Source Data file. Representative micrographs from three independent experiments with similar results are shown. C SDS-PAGE analysis of oat seed proteins during fibrillation at pH 2 over a certain timeframe. The experiment was independently repeated three times with similar results. D LC-MS/MS identification of protein band from oat seeds at 0 h in (C). E Schematic diagram of protein states during fibrillation. Protein monomers undergo partial proteolysis, yielding shorter peptides that subsequently nucleate and self-assemble into ordered β‑sheet-rich fibrils. Created in BioRender. Liu, C. (2026) https://BioRender.com/tdwatdq. Oat, soybean, and rice seeds form fibrillar aggregates after 10 h of cooking. A Schematic diagram of the simulated cooking process of oat/soybean/rice seed. Oat, soybean, and rice seeds incubated under cooking-like conditions at pH ≈ 2, 4, 6, and 8 for 10 h, characterized by NS-TEM. Created in BioRender. Liu, C. (2026) https://BioRender.com/tdwatdq. B Effects of different pH conditions on fibrillation of oat/soybean/rice seeds. NS-TEM micrographs of samples. Scale bar, 500 nm. White arrows denote the fibrillar structures formed. Source data are provided as a Source Data file. Representative micrographs from three independent experiments with similar results are shown. C SDS-PAGE analysis of oat seed proteins during fibrillation at pH 2 over a certain timeframe. The experiment was independently repeated three times with similar results. D LC-MS/MS identification of protein band from oat seeds at 0 h in (C). E Schematic diagram of protein states during fibrillation. Protein monomers undergo partial proteolysis, yielding shorter peptides that subsequently nucleate and self-assemble into ordered β‑sheet-rich fibrils. Created in BioRender. Liu, C. (2026) https://BioRender.com/tdwatdq. To confirm that specific proteins were forming the fibrils, we purified the major seed storage proteins from each species (oat 12S globulin, soybean globulin, and rice glutelin). Oat 12S globulin, which consists of an α-subunit (residues 25-317) and a β-subunit (residues 318-518) connected by disulfide bonds (Fig. S4A, B), was extracted from oat flour using an optimized alkaline extraction and acid precipitation protocol47. SDS-PAGE under reducing and non-reducing conditions confirmed the presence of the expected disulfide-linked subunits, and mass spectrometry verified the protein’s identity (Fig. S4C). We applied similar extraction procedures to soybean and rice, yielding purified soybean globulin and rice glutelin 48,49 (Fig. S5). We next incubated each purified protein under various pH values (≈ 2, 4, 6, 8) and temperature conditions (37 °C or 85 °C) to induce fibrillation (Fig. 2A–C, Fig. S6). Strikingly, all three proteins formed mature amyloid fibrils only under the harsh conditions of pH 2 and 85 °C, as visualized by NS-TEM (Fig. 2A–C, Fig. S6). Under these conditions, oat globulin also underwent extensive proteolysis predominantly at pH 2, yielding low-molecular-weight polypeptides, consistent with the notion that partial hydrolysis provides the building blocks for fibril assembly (Fig. 2A). Although SDS-PAGE revealed that limited proteolysis also occurred at pH 4 and pH 6, these conditions did not support fibril formation, likely because the extent of hydrolysis and conformational destabilization was insufficient to induce amyloid assembly. These findings demonstrate that both an acidic environment (pH≈ 2) and high temperature (≈ 85 °C) are essential for the fibrillation of these seed storage proteins, and that proteolytic fragmentation is likely a prerequisite for their assembly into fibrils. Circular dichroism (CD) spectroscopy revealed secondary-structure changes in all three seed storage proteins upon fibrillation (Fig. 2D). After fibrillation, both oat and soybean globulins exhibited a clear spectral shift, with the negative band converging to ~218 nm, while oat globulin additionally showed a pronounced positive band in the 185-200 nm region, together indicating the formation of β-sheet-dominated structures. By contrast, rice glutelin showed a distinct structural transition from an α-helix-rich conformation, characterized by negative bands at ~208 and 222 nm, to a β-sheet-enriched structure with a sharp minimum around ~218 nm.Fig. 2Characterization of oat globulin, soybean globulin, rice glutelin fibrillation.A Effects of different pH conditions on oat globulin fibrillation (pH≈ 2, 4, 6, 8) within 12 h, as visualized by NS-TEM (left). White arrows denote the fibrillar structures formed. Fibrils of oat globulin gradually formed only at pH 2 conditions as the incubation time increased, reaching a mature state at 12 h. SDS-PAGE analysis of different samples is shown (right). Oat globulin samples underwent proteolysis exclusively under pH 2 conditions, producing peptide fragments over the course of incubation. Source data are provided as a Source Data file. Representative micrographs from three independent experiments with similar results are shown. The experiment was independently repeated three times with similar results. B Effects of different pH conditions on soybean globulin fibrillation (pH≈ 2, 4, 6, 8) within 12 h, as visualized by NS-TEM. White arrows denote the fibrillar structures formed. Fibrils of soybean globulin gradually formed only at pH 2 conditions as the incubation time increased, reaching a mature state at 12 h. Representative micrographs from three independent experiments with similar results are shown. C Effects of different pH conditions on rice glutelin fibrillation (pH≈ 2, 4, 6, 8) within 12 h, as visualized by NS-TEM. White arrows denote the fibrillar structures formed. Fibrils of rice glutelin gradually formed only at pH 2 conditions as the incubation time increased, reaching a mature state at 12 h. Representative micrographs from three independent experiments with similar results are shown. D Transition of secondary structures from monomer (blue) to fibril (red) of oat globulin, soybean globulin and rice glutelin was monitored using CD. Source data are provided as a Source Data file. Characterization of oat globulin, soybean globulin, rice glutelin fibrillation. A Effects of different pH conditions on oat globulin fibrillation (pH≈ 2, 4, 6, 8) within 12 h, as visualized by NS-TEM (left). White arrows denote the fibrillar structures formed. Fibrils of oat globulin gradually formed only at pH 2 conditions as the incubation time increased, reaching a mature state at 12 h. SDS-PAGE analysis of different samples is shown (right). Oat globulin samples underwent proteolysis exclusively under pH 2 conditions, producing peptide fragments over the course of incubation. Source data are provided as a Source Data file. Representative micrographs from three independent experiments with similar results are shown. The experiment was independently repeated three times with similar results. B Effects of different pH conditions on soybean globulin fibrillation (pH≈ 2, 4, 6, 8) within 12 h, as visualized by NS-TEM. White arrows denote the fibrillar structures formed. Fibrils of soybean globulin gradually formed only at pH 2 conditions as the incubation time increased, reaching a mature state at 12 h. Representative micrographs from three independent experiments with similar results are shown. C Effects of different pH conditions on rice glutelin fibrillation (pH≈ 2, 4, 6, 8) within 12 h, as visualized by NS-TEM. White arrows denote the fibrillar structures formed. Fibrils of rice glutelin gradually formed only at pH 2 conditions as the incubation time increased, reaching a mature state at 12 h. Representative micrographs from three independent experiments with similar results are shown. D Transition of secondary structures from monomer (blue) to fibril (red) of oat globulin, soybean globulin and rice glutelin was monitored using CD. Source data are provided as a Source Data file. Structural characterization of oat globulin fibrils We characterized the morphology of oat globulin fibrils using NS-TEM and atomic force microscopy (AFM). NS-TEM imaging revealed that the fibrils were long, unbranched, and morphologically uniform (Fig. 3A). AFM imaging further showed a periodic surface topology along the fibril length and a distinct left-handed helical twist, consistent with what AFM studies describe as “twisted fibrils” 50,51 (Fig. 3B).Fig. 3Structural characterization of oat globulin fibrils.A NS-TEM micrograph of oat globulin fibrils. Scale bar, 500 nm. Representative micrographs from three independent experiments with similar results are shown. B AFM micrograph of oat globulin fibrils (left). Scale bar, 50 nm. Fibrils exhibit distinct helical periodicity and display a clear left-handed chirality (right). Representative micrographs from three independent experiments with similar results are shown. C Structural analysis of polymorph 1. Cryo-EM micrograph of oat globulin fibrils. Scale bar, 100 nm. 2D classification of oat globulin fibrils. The half-pitch was determined from 686-pixel box-size class averages in 2D classification. 3D classification of polymorph 1. Resolution of 3D reconstruction is indicated. D Structural analysis of polymorph 2. Cryo-EM micrograph of oat globulin fibrils. Scale bar, 100 nm. 2D and 3D classification of polymorph 2. The half-pitch was determined from 686-pixel box-size class averages in 2D classification. Structural characterization of oat globulin fibrils. A NS-TEM micrograph of oat globulin fibrils. Scale bar, 500 nm. Representative micrographs from three independent experiments with similar results are shown. B AFM micrograph of oat globulin fibrils (left). Scale bar, 50 nm. Fibrils exhibit distinct helical periodicity and display a clear left-handed chirality (right). Representative micrographs from three independent experiments with similar results are shown. C Structural analysis of polymorph 1. Cryo-EM micrograph of oat globulin fibrils. Scale bar, 100 nm. 2D classification of oat globulin fibrils. The half-pitch was determined from 686-pixel box-size class averages in 2D classification. 3D classification of polymorph 1. Resolution of 3D reconstruction is indicated. D Structural analysis of polymorph 2. Cryo-EM micrograph of oat globulin fibrils. Scale bar, 100 nm. 2D and 3D classification of polymorph 2. The half-pitch was determined from 686-pixel box-size class averages in 2D classification. We next determined the fibril structure at atomic resolution using cryo-EM. 5000 cryo-EM micrographs were collected on a 300 kV electron microscope, and 69,486 fibril segments were manually picked for image processing (Supplementary Table 1). Two predominant fibril polymorphs were evident in the two-dimensional (2D) class averages: a straight polymorph (~65% of segments) and a twisted polymorph (~35% of segments). Given the challenges of reconstructing straight, non-twisted filaments with current helical algorithms, we focused on the twisted fibrils, which exhibited clear helical periodicity (Fig. 3C, D). Within the twisted polymorph population, 2D classification further identified two fibril polymorphs with different helical half-pitches, comprising approximately 61% and 39% of the segments, respectively (Fig. 3C, D). Subsequent three-dimensional (3D) classification converged on two closely related 3D structures corresponding to these polymorphs. Using helical reconstruction, we obtained a 3D density map of the major fibril polymorph (Polymorph 1, ~61%) at an overall resolution of ~3.88 Å (Fig. 3C, Fig. S7). This fibril has a half helical pitch of ~54.7 nm, a helical rise of 4.80 Å per repeating unit, and a helical twist of −1.58° per subunit (Fig. 4A). Using the ~3.9 Å cryo-EM density map, we built an atomic model of the predominant fibril structure (Polymorph 1)52. The fibril core architecture and residue registry were assigned by integrating features resolved in the density map with the peptide segments identified by tandem mass spectrometry mapping to oat globulin (Fig. 4B, Fig. S8). This approach ensured that the atomic model was consistent with both the structural features observed in cryo-EM and the sequences detected in the fibril-enriched fraction. Together, these complementary datasets support the reliability of the amino acid sequence assignment within the fibril core. To better characterize the oat globulin fibril sample, we performed low-molecular-weight SDS-PAGE, revealing prominent bands around ~12 kDa and ~4.2-6.4 kDa (Fig. S8C). We interpret this as reflecting fragmented or heterogeneous fibril-associated species rather than a direct readout of the ordered core size. Structural analysis indicates that the fibril core comprises residues Val369 to Asn403 of oat globulin, which fold into four β-strands arranged in a conserved triangular fold (Fig. 4C). A topology diagram maps these β1-β4 strands onto the full primary sequence of the oat globulin monomer, highlighting the segment that forms the fibril core (Fig. 4C-D).Fig. 4Cryo-EM structures of oat globulin fibrils.A Cryo-EM density map and structural models of oat globulin fibril. Fibril width, length of half pitch, helical rise and twist angle are indicated. The twist angle is graphically illustrated. Graphing was performed with UCSF Chimera v1.13. B LC-MS/MS of gel band from SDS-PAGE of the sample. Oat globulin contains three domains, including signal peptide (pink), α-subunit (blue), and β-subunit (green). The primary sequence of the β-subunit is presented, with the sequences of distinct peptides detected by LC-MS/MS highlighted in red. The SDS-PAGE experiment was independently repeated three times with similar results. Source data are provided as a Source Data file. C Overlay of the oat globulin fibril structure with the density map via rigid-body fitting. β-strands are indicated. The topology diagram of oat globulin is shown at the bottom. D Three-layer structural model of oat globulin fibrils is shown in cartoon and displayed in different views. Cryo-EM structures of oat globulin fibrils. A Cryo-EM density map and structural models of oat globulin fibril. Fibril width, length of half pitch, helical rise and twist angle are indicated. The twist angle is graphically illustrated. Graphing was performed with UCSF Chimera v1.13. B LC-MS/MS of gel band from SDS-PAGE of the sample. Oat globulin contains three domains, including signal peptide (pink), α-subunit (blue), and β-subunit (green). The primary sequence of the β-subunit is presented, with the sequences of distinct peptides detected by LC-MS/MS highlighted in red. The SDS-PAGE experiment was independently repeated three times with similar results. Source data are provided as a Source Data file. C Overlay of the oat globulin fibril structure with the density map via rigid-body fitting. β-strands are indicated. The topology diagram of oat globulin is shown at the bottom. D Three-layer structural model of oat globulin fibrils is shown in cartoon and displayed in different views. Surface electrostatic and hydrophobic analyses indicated that hydrophobic interactions predominantly stabilize the fibril core (Fig. 5A, B). The four β-strands wrap around a tightly packed hydrophobic core composed largely of aliphatic residues (Ala, Val, Ile) plus one Trp. Key residues within the hydrophobic core include Trp381, Ile383, Ala385, Val388, His390, Ile392, Gly394, Ala396, Val398, and Val400. Continuous hydrophobic stacking along the fibril axis likely underpins the protofilament’s stability. Notably, the fibril core has a distinctive triangular cross-section. Specific inter-sheet hydrophobic contacts lock the three β-sheet faces together (Fig. 5C). For example, Ile383 and Ala385 in β2 pack against Val388 in β3 to create one corner of the triangle. Similarly, Ile392 on β3 packs against Ala396 and Val398 in β4 to form the second corner. Finally, Trp381 and Val400 form a cross-strand “clasp” that secures the assembly, reinforcing the triangular core (Fig. 5C).Fig. 5Structural analysis of oat globulin fibrils.A Surface electrostatic potential of the fibril core. Red denotes regions of negative electrostatic potential; blue denotes regions of positive electrostatic potential. B Hydrophilic and hydrophobic distribution of the fibril core. Hydrophilicity (yellow) to hydrophobicity (green) is indicated by the color bar. C Oat globulin fibril structure highlighting hydrophobic interactions (zoomed-in views) within a single protofilament. Residues involved in intraprotofilament contacts are shown as spheres. Structural analysis of oat globulin fibrils. A Surface electrostatic potential of the fibril core. Red denotes regions of negative electrostatic potential; blue denotes regions of positive electrostatic potential. B Hydrophilic and hydrophobic distribution of the fibril core. Hydrophilicity (yellow) to hydrophobicity (green) is indicated by the color bar. C Oat globulin fibril structure highlighting hydrophobic interactions (zoomed-in views) within a single protofilament. Residues involved in intraprotofilament contacts are shown as spheres. The minor fibril polymorph (Polymorph 2) appeared to have a similar overall fold to Polymorph 1 based on 3D classification (Fig. 3C, D). However, we were unable to obtain a high-resolution reconstruction for Polymorph 2, likely due to its greater structural heterogeneity. Even though we estimated helical symmetry parameters for Polymorph 2, its density map remained poorly resolved, with diffuse fibril outlines and a lack of distinct structural features, which precluded atomic model building (Fig. 3D). In parallel, we collected cryo-EM data for soybean and rice seed protein fibrils. In both cases, 2D class averages showed no clear periodic features (Fig. S9). This lack of well-defined classes suggests these fibrils did not adopt a uniform helical symmetry, precluding reliable 3D reconstruction and atomic-level structural analysis. Discussion Previous studies have shown that several plant seed storage proteins, including soybean 7S and 11S globulins, rice glutelins and barley storage proteins, as well as certain food-derived proteins, can assemble into amyloid-like fibrils under defined physicochemical conditions, indicating that amyloid formation is not restricted to animal or disease-associated proteins39,53–57. Building on this foundation, our results demonstrate that seed storage proteins from multiple plant species possess an intrinsic capacity to undergo conformational rearrangement and form β-sheet-rich fibrils when exposed to specific environmental triggers, supporting the concept that amyloid-like folding may represent a broadly conserved and sequence-encoded property of proteins rather than a phenomenon confined to pathological contexts. By systematically comparing the fibril-forming behavior of oat, rice, and soybean proteins and resolving the cryo-EM structure of oat fibrils, our study extends previous observations and provides additional structural and mechanistic insight into the polymorphism of plant-derived amyloids, thereby offering a useful framework for investigating protein stability and environmentally induced folding transitions. We further demonstrated that exposure of either oat globulin or rice glutelin, whether in seeds or in purified form, to simulated cooking conditions (elevated temperature and acidic pH) is sufficient to induce amyloid fibril formation. This finding highlights the strong inherent amyloid-forming propensity of these seed storage proteins and points to an underexplored potential for such endogenous aggregation in plants. Moreover, the ability to induce fibrillation in minimally processed, natural proteins underscores their promise as cost-effective, abundant, and sustainable building blocks for the development of bio-inspired materials58–60. The high-resolution cryo-EM structure of the oat globulin fibril revealed a triangular, self-locking β-sheet architecture stabilized predominantly by hydrophobic interactions. Although such a fibrillar structure almost certainly does not form in seeds under normal physiological conditions, its spontaneous formation under specific in vitro conditions illustrates a powerful sequence-encoded propensity for ordered aggregation. The resulting fibril is highly compact, pseudo-threefold symmetry, and thermodynamically stable, offering insight into the design principles of plant-based amyloid biomaterials. Notably, the triangular β-sheet framework provides a robust scaffold for the rational engineering of nanostructures from plant proteins, advancing the development of biodegradable, environmentally friendly materials. Amyloid fibril morphology is highly sensitive to assembly conditions. Parameters such as temperature, ionic strength, incubation time, protein concentration, and buffer composition can influence fibril periodicity, curvature, and polymorph distribution, and even small variations may shift the system toward different conformations61,62. Consistent with this, both the Spc‑SH3 system and food-derived proteins show condition-dependent differences in fibril morphology under changes in incubation regime, pH, heating mode, or salt concentration63–66. Therefore, the differences in AFM-observed periodicity and morphology between our oat globulin fibrils and those reported likely reflect slight variations in experimental conditions rather than technical inconsistencies47. These observations highlight the inherent polymorphism of amyloid fibrils and the importance of carefully controlled assembly conditions. Consistently, the cryo-EM structure we resolved corresponds to the dominant polymorph under our conditions, representing the stable, conformationally homogeneous fibril core (V369-N403). Other minor fibril species or flexible/disordered regions outside this core were present in the sample but could not be reconstructed at high resolution due to low abundance and structural heterogeneity. This highlights that while cryo-EM provides a detailed view of the ordered core, fibril polymorphism and condition-dependent structural variations are inherent features of amyloid assemblies. In exploratory in vitro assays, given the broad interest in heterotypic seeding mechanisms, we examined whether these plant-derived fibrils could influence the aggregation of disease-associated substrates (α-syn, FUS LC, and TDP-43 LC)67–69. Among the tested assemblies, soybean- and rice-derived fibrils showed negligible or only minor effects, whereas oat globulin fibrils markedly promoted α-syn aggregation, with only a modest acceleration observed for FUS LC and no detectable effect on TDP-43 LC under the same conditions (Fig. S10A). Notably, the cross-seeding efficiency of oat globulin fibrils remained lower than that of homologous α-syn PFFs, indicating a comparatively weaker heterotypic seeding capacity under the tested conditions (Fig. S10B). Consistently, oat protein fibrils were reported to modulate amyloid-β (Aβ) aggregation in vitro, further supporting system-dependent interactions between plant-derived amyloids and disease-associated proteins70. These observations were obtained under controlled in vitro conditions and should not be overinterpreted in a pathological context. Importantly, a recent study on food-derived amyloid fibrils has shown that such assemblies can be degraded under simulated gastrointestinal conditions, suggesting limited persistence of intact assemblies in the digestive tract71. As the present in vitro experiments do not address physiological exposure, tissue distribution, or passage across biological barriers, no direct pathological inference can be drawn from these data. Further in vivo studies will be required to determine whether such biochemical interactions occur under physiological conditions. In conclusion, our study demonstrates that plant seed storage proteins can serve as amyloid-forming systems under specific conditions. By characterizing their fibrillation behavior and determining the cryo-EM structure of oat globulin fibrils, we provide insight into the molecular features underlying plant amyloid assembly. These findings expand the current structural understanding of amyloids and support the potential use of plant-derived amyloid materials in sustainable material applications. Methods Experiments of whole plant seeds under cooking-like conditions Oat, soybean, and rice seeds were coarsely ground into small fragments using a mortar and pestle. The ground materials were suspended in deionized water adjusted to pH 2.0, 4.0, 6.0, or 8.0 (10 mg/mL). The suspensions were incubated at 85 °C for 10 h to simulate prolonged cooking conditions. After heating, samples were cooled to room temperature and subsequently characterized by negative-staining transmission electron microscopy (NS-TEM) to examine the morphology of the resulting protein aggregates. Protein purification Oat globulin was extracted using an optimized alkaline solubilization–acid precipitation protocol47. Briefly, commercial oat flour was defatted twice with petroleum ether (1:6, w/v), followed by air-drying in a fume hood. The defatted powder was suspended in deionized water (1:6, w/v), adjusted to pH 9.2 with NaOH, and stirred for 25 min at room temperature. The mixture was filtered through a 200-mesh nylon cloth and centrifuged to remove insoluble debris. The supernatant was adjusted to pH 4.8 using 1 M HCl to induce protein precipitation. After stirring for 1 h, the precipitate was collected by centrifugation, washed with deionized water, and freeze-dried to obtain purified oat globulin. Soybean 7S globulin and rice glutelin were purified using an isoelectric precipitation method48,49. Preparation of seed storage protein fibrils and pre-formed fibrils (PFFs) Oat globulin, soybean globulin, and rice glutelin were incubated in a pH 2 buffer at 85 °C in a ThermoMixer (Eppendorf) for 12 h at a concentration of 5 mg/mL. The resulting fibril samples were subsequently used for NS-TEM, atomic force microscopy, and cryo-EM sample preparation. For incubated fibrils of oat globulin, soybean globulin, and rice glutelin, the fibril-containing pellet was separated from the monomer-containing supernatant by centrifugation. Then, the fibril-containing pellet was resuspended, and its concentration was determined as the fibril concentration. Oat globulin, soybean globulin and rice glutelin PFFs were prepared by fibrils sonication at 20% power on ice for 25 times (1 s on, 1 s off). NS-TEM Five microliters of fibril solution were incubated on a 200-mesh glow-discharged copper grid (Zhongjingkeyi Technology Co., Ltd., Beijing) for 45 s. Then, the grid was washed with double-distilled water followed by 2% w/v uranyl acetate for another 45 s and dried in air. The samples were imaged by a Tecnai T12 transmission electron microscope (FEI) operated at 120 kV. Atomic force microscopy (AFM) A drop of five microliters of fibril solution was loaded on a clean mica surface for 5 min at room temperature and washed by double-distilled water to remove unbound fibrils. Then, images were captured by Nanoscope V Multimode 8 (Bruker) with SNL-10 probes (a constant of 0.35 N m−1) on ScanAsyst air mode in 1 Hz scan rate. The following data processing were carried out in the supplied software Nanoscope Analysis (version 1.5, Bruker). Circular dichroism (CD) spectroscopy CD spectroscopy was performed to characterize the secondary structure of both monomeric and fibrillar protein samples. Fibrils were obtained by incubating protein under self-assembly conditions and subsequently collected by centrifugation at 12,000 × g for 10 min to remove residual soluble monomers. The resulting fibril pellet was resuspended in acidified ddH₂O (pH 2.0, adjusted with HCl) and centrifuged again at 12,000 × g for 10 min to ensure complete removal of residual monomers. The final pellet was then resuspended in the same buffer, gently mixed, and briefly centrifuged at low speed (1-2 min) to remove bubbles prior to CD measurements. Monomeric samples were freshly prepared in the same acidified buffer (pH 2.0) at a final concentration of 2 mg/mL. To ensure sample clarity and remove potential insoluble impurities or pre-existing aggregates, the freshly prepared solutions were briefly centrifuged (12,000×g, 5min), and the supernatant was used directly for CD measurements. All samples (200 μL) were transferred into a 1 mm pathlength quartz cuvette. CD spectra were recorded on a Chirascan spectropolarimeter (Applied Photophysics, UK) at 25 °C. Baseline correction was carried out using the corresponding buffer. Spectra were acquired over the wavelength range 260-190 nm with a scanning speed of 0.5 s/nm and a bandwidth of 1 nm. Each measurement was scanned three times and averaged to improve the signal-to-noise ratio. Raw data were processed according to the manufacturer’s protocol, including blank subtraction and spectral smoothing, and plotted using GraphPad Prism 8. Mass spectrometry For monomeric proteins, samples were mixed with SDS loading buffer (containing 2% SDS, 5% β-mercaptoethanol, 10% glycerol, 62.5 mM Tris-HCl, pH 6.8, and 0.01% bromophenol blue), heated at 100 °C for 5 min, and loaded onto 4–20% SDS-PAGE gels. The gels were stained with Coomassie Brilliant Blue R-250, and protein bands of interest were excised for in-gel digestion and liquid chromatography-tandem mass spectrometry (LC-MS/MS) analysis. For fibrillar samples, fibril suspensions were collected by centrifugation (12,000×g, 40 min) and washed twice to remove soluble components. The resulting pellets were resuspended in SDS loading buffer and heated at 100 °C for 30 min to disrupt the fibril structure and dissociate the assembled peptides. These samples were analyzed by SDS-PAGE to confirm the presence of fibrillar assemblies, while the corresponding soluble fractions were separately subjected to LC-MS/MS for protein identification. The pellets were resuspended in SDS loading buffer and heated at 100 °C for 30 min, thereby denaturing and hydrolyzing the highly aggregated fibrillar proteins and yielding peptide fragments. These samples were analyzed by SDS-PAGE to confirm the presence of fibrillar assemblies, while the corresponding soluble fractions were separately subjected to LC-MS/MS for protein identification. Protein bands of interest were excised from SDS-PAGE gels, carefully avoiding keratin contamination, and cut into ~1 mm³ cubes. Gel pieces were destained by sequential incubation in 100 mM ammonium bicarbonate/acetonitrile (1:1, v/v) and neat acetonitrile until clear, followed by dehydration. Proteins were reduced with DTT at 56 °C for 30 min and alkylated with iodoacetamide at room temperature in the dark. After another dehydration step, the gel pieces were rehydrated in trypsin solution and incubated overnight at 37 °C. Peptides were extracted with formic acid/acetonitrile, combined with extracts from a final acetonitrile shrinking step, dried at 60 °C, and redissolved in 0.1% formic acid or stored at −20 °C. The peptide mixture was analyzed using an online EASY-nL-LC 1000 coupled with an Orbitrap Q-Exactive HF mass spectrometer. The sample was loaded directly onto a 15 cm home-made capillary column (C18-AQ, 1.9 μm resin, 100 μm I.D.). A gradient was developed using mobile phase A (0.1% FA, 2% ACN, and 98% H2O) and B (0.1% FA, 2% H2O, and 98% ACN), with a 60-min gradient set as 8% B at 4 min, 22% B at 45 min, 30% B at 53 min, 92% B at 57 min and 95% B at 60 min. A static flow rate of 300 nL/min was used. Mass spectra were acquired in a data-dependent mode with one full scan in the Orbitrap (m/z: 350-1500; resolution: 60,000; AGC target value: 3,000,000; maximal injection time: 20 ms; exclusion duration: 30 s), followed by MS2 scans in the linear trap (resolution: 15,000; AGC target value: 50,000; maximal injection time: 25 ms; isolation window: 1.6 m/z). Low-Molecular-Weight SDS-PAGE Protein samples were prepared by mixing with Tricine sample buffer (containing SDS and reducing agent) and heated at 95 °C for 30 min, followed by brief centrifugation. Samples and molecular weight markers were loaded onto 16% Tricine-SDS polyacrylamide gels. Electrophoresis was performed at 120 V for 3 h at 4 °C. Following electrophoresis, gels were fixed in 50% isopropanol, 10% acetic acid, and 40% water for 30 min, and stained with Coomassie Brilliant Blue G-250. Excess dye was removed by washing with water, and gels were imaged for visualization of protein bands. Cryo-electron microscopy (Cryo-EM) sample preparation and data collection Oat globulin, soybean globulin, and rice glutelin fibrils were applied to glow-discharged holey carbon copper grids (R2/1, 300 mesh, Quantifoil), and then plunge frozen in liquid ethane after blotting with filter paper by using Vitrobot Mark IV (Thermo Fisher). Cryo-EM images of fibrils were obtained using a Tecnai T12 transmission electron microscope (FEI). Cryo-EM data were collected of oat globulin fibrils in a 300 kV Titan Krios G4 transmission electron microscope (Thermo Fisher) with a BioContinuum K3 direct detector (Gatan), using a GIF Quantum energy filter (Gatan) with a slit width of 20 eV to remove inelastically scattered electrons. 40 movie frames per micrograph were recorded in ×105,000 magnification with a pixel size 0.83 Å pixel−1 at super-resolution mode. Cryo-EM data collection was performed by software EPU (Thermo Fisher) with 2 s exposure time and −1.0 to −2.0 μm defocus value in a total dose of 55 e-Å2. Image processing MotionCorr2 for motion correction was implemented to correct the beam-induced motion of movie frames with dose-weighting72. Then, CTFFIND-4.1.8 was used to estimate the contrast transfer function of motion-corrected images73. Next, fibrils were manually picked by using the manual picking method of RELION version 3.174. Helical reconstruction Helical reconstruction was performed in RELION version 3.1, including particle extraction, two-dimensional (2D) classification, three-dimensional (3D) classification, 3D auto refinement and post-processing72. For the oat globulin dataset, 69,486 manually picked fibrils from 5000 micrographs were extracted into segments in a box size of 360 pixels with an inter-box distance of 30.0 Å. Then, several iterations of 2D classification were applied to extracted segments with a decreasing in-plane angular sampling rate from 8° to 0.5°, and the T = 2 regularization parameter. Several iterations of 2D classification were performed to select particles which belong to the same polymorph. An initial 3D model was built by selected 2D classes. Cryo-EM micrographs, and 2D classification averages were used together to calculate an initial twist angle. A value of 4.8 Å was used as an initial value of helical rise. When the local search of symmetry for helical twist and rise was carried out after the separation of β strands, the clearest class was selected for following 3D auto refinement. Finally, the overall resolution of twisted filament oat globulin fibrils was reported at 3.88 Å, according to the gold-standard FSC = 0.143 criteria. For soybean globulin dataset, 154,319 manually picked fibrils from 3000 micrographs were extracted into segments in a box-size of 1024 pixel with an inter-box distance of 30.0 Å. Then, several iterations of 2D classification were applied to the extracted segments, with a decreasing in-plane angular sampling rate from 8° to 1° and the T = 2 regularization parameter. Several iterations of 2D classification were performed to select particles which belong to the same polymorph, but we could not classify these particles through 2D classification. For the rice glutelin dataset, 35,619 manually picked fibrils from 1000 micrographs were extracted into segments in two box sizes of 1024 pixels and 960 pixels with an inter-box distance of 30.0 Å. Then, several iterations of 2D classification were applied to extracted segments with a decreasing in-plane angular sampling rate from 8° to 1°, and the T = 2 regularization parameter. Several iterations of 2D classification were performed to select particles which belong to the same polymorph, but we could not classify these particles through 2D classification. Atomic model building and refinement Based on the density maps after post-processing program, the atomic models of twisted filament oat globulin fibrils were built de novo in COOT, respectively74. Then, three-layer models were generated in software Chimera and refined by real-space refinement program of PHENIX75,76. Additional details about three models were shown in Supplementary Table 1. Thioflavin T (ThT) kinetic assay FUS LC, α-syn and TDP-43 LC monomers were incubated in specific buffers with or without oat globulin, soybean globulin, and rice glutelin PFFs in a black 384-well plate (Thermo Scientific). ThT was added to the reaction mixture at a final concentration of 10 μM. The fluorescence intensity was recorded using a Varioskan Flash Spectral Scanning Multimode Reader (Thermo Scientific), measuring at 440 nm (excitation) and 485 nm (emission) wavelengths with shaking at 900 rpm at 37 °C. At least three replicates were performed for each sample. GraphPad Prism 9 was applied for graphing, with mean ± SD. Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Supplementary information Supplementary Information Reporting Summary Transparent Peer Review file Supplementary Information Reporting Summary Transparent Peer Review file Source data Source Data Source Data