A natural solution from caterpillar cuticles for flexible impact-resistant materials Authors: Fenghou Yuan, Shuaifei Hu, Zihan Pang, Yuntian Cui, Xuliang Qian, Chaowei Shi, Tian Liu Published: 2026 ABSTRACT AbstractFlexible impact-resistant materials are essential for defense, transportation, and biomedical engineering due to their ability to combine strength, energy dissipation, and compliance. While nature offers robust solutions like beetle elytra or mantis shrimp hammers, designing materials that achieve these properties simultaneously remains a significant challenge. Here, we explore the cuticle of Ostrinia furnacalis (Asian corn borer) as a natural model. We identify two key structural proteins, FCSP-1 and FCSP-2, that form a β-sheet-rich matrix via liquid-liquid phase separation, adhering to chitin scaffolds. This matrix promotes lamellar structures that enhance energy dissipation through hydrogen bond disruption and secondary structure transformation under stress. NMR and molecular dynamics simulations further confirm that matrix formation relies on fine-tuned hydrogen bonding, electrostatic interactions, and π-π/cation-π interactions. Inspired by this mechanism, we developed a chitin-protein composite hydrogel with an energy dissipation capability of 8 MJ/m3, and its mechanical properties can be precisely tuned via in-vitro phosphorylation. This hydrogel demonstrates exceptional potential for protecting fragile objects.Subject terms: Gels and hydrogels, Entomology, Bioinspired materials, Proteins Introduction Impact-resistant materials play crucial roles in fields such as military, aviation, transportation, and biomedical engineering1–4. Nature offers a diverse portfolio of impact-resistant materials, from the rigid armor of beetle elytra to the dynamic hammer of mantis shrimp5–7. However, materials that are both flexible and impact-resistant for use in soft robotic actuators, personal protective equipment, or delicate biomedical implants remain a significant challenge2,8,9. These materials must simultaneously meet demanding requirements for strength, energy dissipation, and contact compliance2,10. Existing solutions, such as hydrogels constructed with hierarchical structures11–13 or incorporating shear-thickening fluids14–16, offer some promise but are still limited in their versatility and performance. The flexible cuticle of certain borer insects, such as the Asian corn borer (Ostrinia furnacalis), presents a natural template for flexible impact-resistant materials (Fig. 1a and Supplementary Fig. 1)17–21. The cuticle must endure mechanical impacts from the insect’s tunneling behavior while protecting its fragile internal tissues. Structurally, the caterpillar cuticle is composed primarily of chitin, over 100 flexible cuticle structural proteins (FCSPs)22,23, and small amounts of pigment. FCSPs comprise various families defined by sequence motifs, such as the chitin-binding Rebers-Riddiford (R&R) domain24,25, and are integral to cuticle integrity. A simple model suggests that the chitin forms a scaffold, with FCSPs filling the structure to create a lamellar arrangement, known for its efficient energy dissipation properties7,18,26–29. Molecular biology studies have shown that knocking down specific FCSPs via techniques like RNA interference can lead to disruptions in the lamellar structure or the formation of cavities, resulting in reduced mechanical properties30–32. However, the specific FCSPs that are key to energy dissipation, and the molecular mechanism by which they operate, are still not understood. Insect FCSPs modulate the function and mechanical properties of the cuticle. a Schematic showing the cuticle of Ostrinia furnacalis as powerful, flexible, impact-resistant material. b Composition of the integument cuticle via biochemical methods determination. c Uniaxial tensile test of native and deproteinized cuticles. Inset, comparison between toughness of cuticles. Data are presented as mean values ± SD (n = 10 individual samples). p = 0.0001. A two-tailed t-test was made for comparisons. d Scheme and cross-sectional scanning electron microscope (SEM) images of native and deproteinized cuticle. Scale bars are shown in the images. e The kinetics of the deformation recovery in the cuticle with/without treatment according to recovery stress-strain curves. Lines are fitted data, with rate constants k determined according to fitted curves. NC native cuticles recovered in PBS. DC deproteinized cuticles recovered in PBS. AIR, native cuticles recovered at 50% relative humidity (RH). Error bars correspond to s.d. (n = 10). The measure of center for the error bars indicated the mean value. f Scheme of the important role of hydrogen bonds (HBs) in the cuticle. In this study, we demonstrate that structural proteins are essential for the caterpillar cuticle’s energy dissipation capabilities, with their role closely tied to the breaking and reformation of hydrogen bonds (HBs). Through proteomic analysis, we identified FCSP-1 and FCSP-2 as two highly abundant FCSPs. FCSP-1 features a canonical RR1 domain, while FCSP-2 is a low-complexity protein, suggesting distinct but complementary roles. In reconstituted systems, we observed that they together form a β-sheet-rich matrix via liquid-liquid phase separation (LLPS) that adheres to chitin. Furthermore, this matrix was observed to promote the formation of the lamellar structure, and its energy dissipation is achieved through the reversible transformation between β-sheets and random coils. Incorporating these proteins into a chitin hydrogel significantly enhances its toughness, increasing it to 8 MJ/m3. Our findings thus propose a plausible molecular mechanism for energy dissipation in insect cuticles and offer new pathways for designing flexible, impact-resistant materials for a range of engineering applications. Results FCSPs are vital for the structure and mechanical properties of the cuticle The freshly dissected cuticle of O. furnacalis exhibits a hydrogel-like composition, consisting of approximately 84% water, 9% chitin, 5% protein, and 2% pigment (Fig. 1b). Notably, this cuticle demonstrates an impressive toughness of 0.62 MJ/m3 (Fig. 1c and Supplementary Figs. 2 and 3), surpassing that of most reported hydrogels33. Scanning electron microscopy (SEM) analysis revealed a distinct lamellar structure within the cuticle (Fig. 1d). Following the removal of proteins, the toughness of the cuticle decreased by 47%, and SEM images showed the disappearance of the lamellar structure. These findings indicate that FCSPs are crucial for preserving both the structural integrity and toughness of the cuticle. To further elucidate the role of FCSPs in energy dissipation, we conducted loading-recovery experiments under varying conditions. The native cuticle (NC), stretched and exposed to air with 50% relative humidity (AIR), demonstrated 40% recovery of maximum stress after 120 min, whereas the deproteinized cuticle (DC) exhibited significantly faster recovery (Fig. 1e and Supplementary Fig. 4). This suggests that the energy dissipation mechanism of FCSPs may involve sacrificial bonds, likely formed by main-chain HBs. The slow recovery of the cuticle in this context is hypothesized to result from the reformation of secondary structures required for HB reconstitution. To test this hypothesis, we monitored the performance recovery behavior of stretched NC under different conditions (Fig. 1e). The stretched NC submerged in water exhibited markedly slower recovery compared to its exposure to AIR. This may be due to excess water occupying the broken sites, forming water-mediated HBs, which impede the recovery of protein-protein and protein-chitin HBs. Additionally, the performance of stretched NC in 20% ethanol showed a faster recovery than in water. Ethanol is known to promote HBs formation in proteins34, which suggests that the HBs networks might be recovered in the recovery process. And this result further supports the idea that FCSP-mediated HBs networks are responsible for the efficient energy dissipation observed in the insect cuticle (Fig. 1f). Two interacting FCSPs form viscoelastic coacervates To identify the key FCSPs responsible for the energy dissipation performance of the cuticle in O. furnacalis, we employed LC-MS/MS for proteomic analysis. A total of 110 FCSPs were annotated and categorized into various protein families (Fig. 2a). Among these, FCSP-1 and FCSP-2 emerged as the most abundant FCSPs in the procuticle, the flexible energy-dissipating layer of the cuticle. The two FCSPs account for 15.5% and 11.3% of the total, respectively (Fig. 2b, Supplementary Fig. 5, and Supplementary Table 1). FCSP-1 contains a low-complexity N-terminal domain (NTD) with alternating charged and uncharged regions, along with a C-terminal RR chitin-binding domain (CTD) (Fig. 2c). FCSP-2, in contrast, is a low-complexity protein, enriched in Ala, Val, Ser, Thr, Tyr, and Gln residues throughout its sequence. Such sequence characteristics suggested distinct but complementary roles of FCSP-1 and FCSP-2. Both FCSP-1 and FCSP-2 were predicted to undergo LLPS (Supplementary Figs. 6 and 7). Critical FCSPs of FCSP-1 and FCSP-2 co-coacervate to form a viscous matrix. a Proteomics of Ostrinia furnacalis’s integument cuticle (n = 3). b Composition of Ostrinia furnacalis’s integument FCSPs. c Diagram of sequences of FCSP-1 and FCSP-2 from identification in (a, b). d Chitin-binding activity of critical FCSPs, determined by two methods. Method 1, BCA protein assay. Method 2, gray scale scanning (p = 0.0008). Data are presented as mean values ± SD (n = 3 samples). e The confocal microscopic images showing FCSPs and mixture (MIX) conservation. Scale bars are 20 μm. f Turbidity of every single critical FCSP and their mixture at different mass ratios. Data are presented as mean values ± SD (n = 5 samples). g Glutaraldehyde crosslink assay between FCSP-1 and FCSP-2. h Laser scanning confocal microscope (LSCM) images and curves (i) of fluorescence recovery after photobleaching (FRAP) of FCSPs coacervates. Data are presented as mean values ± SD (n = 20 samples). Scale bars are 2 μm. j Fluorescence diffusion coefficients obtained from (i). A two-tailed t-test was made for comparisons (p = 0.0008 and 0.0409). Data are presented as mean values ± SD (n = 4 coacervates). Schemes showing the fluidity differences of FCSPs and MIX. k Fourier transform infrared spectroscopy (FTIR) spectrum showing the secondary structure changing during the FCSP-FCSP interaction. FCSP-1 was successfully expressed in a soluble form in Escherichia coli, whereas FCSP-2 predominantly formed inclusion bodies. Through bioinformatics analyses and phosphoproteomic analyses, we identified potential phosphorylation sites in FCSP-2, specifically within the S(phos)YSAAPA/V motifs (Supplementary Fig. 8 and Supplementary Table 2). Given that phosphorylation has been shown to enhance the solubility of structural proteins in other organisms35,36, we engineered a phospho-mimetic mutant of FCSP-2 by substituting serine with aspartic acid. This mutation significantly increased the protein’s solubility under physiological conditions. The chitin-binding capacity and LLPS behavior of FCSP-1 and FCSP-2 were subsequently characterized. FCSP-1 displayed strong chitin-binding activity, whereas FCSP-2 showed minimal binding, based on the chitin-binding assay (Fig. 2d and Supplementary Fig. 9a, b). Further analysis revealed that the chitin-binding function of FCSP-1 resides in its CTD, while its NTD does not contribute significantly (Supplementary Fig. 10). Interestingly, reducing temperature and the addition of ethanol enhanced FCSP-1’s chitin-binding activity (Supplementary Fig. 9c), suggesting a key role of HBs in the chitin-FCSP-1 interaction. Regarding LLPS behavior, FCSP-1 formed coacervates at a concentration of 0.4 mg/mL, while FCSP-2 did so at a much lower concentration of 0.1 mg/mL (Fig. 2e and Supplementary Fig. 11). When mixed in a physiological molar ratio (FCSP-1: FCSP-2 as 3: 2, total protein concentration 0.24 mg/mL, named MIX later), coacervation was significantly enhanced (Fig. 2e, f and Supplementary Fig. 11). This observation suggests a potential interaction between FCSP-1 and FCSP-2 that enhances LLPS. Glutaraldehyde cross-linking confirmed this interaction, as significant cross-linked complexes were observed in the MIX group, but not in the individual FCSP-1 or FCSP-2 samples (Fig. 2g and Supplementary Fig. 12). Fluorescence recovery after photobleaching (FRAP) experiments further demonstrated that the viscosity of the coacervates followed the order MIX > FCSP-2 > FCSP-1 (Fig. 2h–j), indicating that the coacervates formed by the mixed proteins exhibited tightly networked phase state. This distinctive internal structure may facilitate more efficient energy dissipation at the macroscopic scale. To gain deeper insights into the structural changes driving these interactions, we examined the secondary structures of the FCSP-1, FCSP-2, and MIX coacervates using FTIR (Fig. 2k and Supplementary Table 3). The amide I band of FCSP-1 peaked at 1645 cm⁻¹, while FCSP-2 peaked at 1640 cm⁻¹. The MIX coacervates, however, showed a peak at 1635 cm⁻¹ with a shoulder at 1625 cm⁻¹. Deconvolution of the amide I band of the MIX revealed an increased presence of β-sheets (β-crystals) (Supplementary Fig. 13), indicating a denser HBs network in the coacervates formed by the MIX compared to those formed by FCSP-1 or FCSP-2 alone. These results suggest that FCSPs coacervates may fuse into a toughening matrix, dissipating energy in the insect cuticle through their enhanced HBs networks37. Mechanisms of cuticle toughening via FCSPs coacervation To investigate the role of FCSPs coacervates in cuticle energy dissipation, we examined their interactions with chitin nanofibers (Fig. 3a). FCSP-1 successfully formed coacervates on the surface of chitin nanofibers, whereas FCSP-2 did not. Interestingly, the MIX (FCSP-1 and FCSP-2 in physiological ratio) not only formed coacervates that adhered to the chitin surface but also induced significant connections between coacervates, creating an interconnected matrix that linked separate chitin nanofibers. And the capability enables MIX to exhibit higher chitin binding capacity than either of the two individual FCSPs (Supplementary Fig. 14). Quartz crystal microbalance with dissipation monitoring (QCM-D) further confirmed that both FCSP-1 and FCSP-2 can form a viscoelastic matrix on chitin surfaces (Fig. 3b and Supplementary Fig. 15). Quasi-in-situ replenishment of viscous MIX coacervates facilitates hierarchical structure and improves the mechanical properties of DC. a SEM images of critical FCSPs binding to α-chitin fiber. Scale bars are 5 μm. b Simulation of modulus and viscosity from QCM-D showing the interaction and mechanical changes of FCSP-1 and FCSP-2. c Fluorescence intensity profile of dual-labeled MIX coacervates. Inset, LSCM images showing spatial distribution of FCSPs in MIX coacervates. d Predicted structure of MIX in AlphaFold338. e Fluorescence intensity profile showing spatial distribution of FCSPs and β-sheet in fused MIX assemblies. Inset, LCSM images. Scale bars, 500 nm. f Scheme of interaction and hierarchical assembly with chitin and hierarchical MIX coacervates. g LSCM images showing different binding activity of FCSPs to DC. Scale bars are 10 μm. h SEM image and schemes of the cuticle cross section for different reloaded durations. Scale bars, 1 μm. i Mechanical comparison between cuticles repaired with different FCSPs. Data are presented as mean values ± SD (n = 5 samples). j Scheme and k LSCM images of fluorescence resonance energy transfer (FRET) based on the tension sensor. Scale bars are 10 μm. l Quasi-in situ FRET profile of MIX coacervates. Data are presented as mean values ± SD (n = 5 samples). m FTIR spectrum showing the secondary structure changing during the shearing and recovery stages (n = 16). To better understand how FCSP-1 and FCSP-2 retain their chitin-binding ability while assembling into a matrix, we employed dual fluorescent labeling to track the hierarchical structural changes of coacervates over time (Fig. 3c and Supplementary Fig. 16). Early in the coacervation, FCSP-1 was predominantly localized at the exterior, while FCSP-2 concentrated within the core. This observation was corroborated by AlphaFold3 co-assembly modeling38. In such a model, the RR domain of FCSP-1 was positioned externally, while the NTD of FCSP-1 interacted with FCSP-2 to form the core (Fig. 3d), explaining the better chitin-binding ability of MIX coacervates. Over time, the coacervates assembled and fused into a uniform matrix, as indicated by the consistent β-sheet distribution observed through ThT staining, suggesting a uniform mechanical property (Fig. 3e). A schematic of the proposed mechanism is provided in Fig. 3f. To elucidate the mechanical role of these FCSPs in the cuticle, we quasi-in situ reloaded FCSP-1, FCSP-2, and MIX into DCs. At a concentration of 0.1 mg/mL, coacervates were observed only in cuticles incubated with FCSP-1 or MIX, but not with FCSP-2 alone (Fig. 3g). This finding indicates that FCSP-FCSP and FCSP-chitin interactions are essential for coacervate formation within the cuticle. The loading capacities of FCSP-1, FCSP-2, and MIX were quantified as 2.6 mg, 0.4 mg, and 4 mg per cuticle, respectively (Supplementary Fig. 17), confirming that FCSP-1 enhances FCSP-2 adsorption onto the chitin scaffold. Following MIX reloading, the cuticle’s microstructure shifted from a reticular to lamellar pattern (Fig. 3h). Tensile tests revealed significant enhancements in strength and toughness, with MIX-reloaded cuticles showing an 18% increase in failure stress, a 99% increase in strain-to-fracture, and a 146% increase in fracture energy compared to DCs (Fig. 3i and Supplementary Fig. 18). These results underscore the critical role of the FCSPs matrix in modulating the cuticle’s mechanical properties and energy dissipation. To further confirm the involvement of the FCSPs matrix in energy dissipation, we developed a fluorescence resonance energy transfer (FRET)-based tension sensor capable of integrating into the coacervates to detect stress in protein aggregates during cuticle stretching (Fig. 3j). The sensor was designed using mECFP and mYPet linked by an elastic (GPGGA)₈ spider silk-derived linker. To allow the sensor to coacervate with MIX, two DP37 peptides from FCSP-1’s NTD were attached to its N- and C-termini. Tests confirmed that the sensor could be incorporated into MIX without impairing its LLPS capability (Supplementary Fig. 19). Notably, when the coacervates were stirred, the FRET effect decreased significantly, and fluorescence recovered by 87% after 20 min of rest (Supplementary Fig. 20). We then applied the FRET-incorporated MIX to DCs and observed a significant reduction in FRET signal upon tensile loading (Fig. 3k and Supplementary Fig. 21). Recovery of FRET effect was much slower in the cuticle-bound coacervates compared to free coacervates (Fig. 3l and Supplementary Fig. 21), likely due to the simultaneous reformation of HBs in the FCSPs matrix and structural recovery of the cuticle. These findings provide the first direct evidence of FCSPs' involvement in stress dissipation within composite biomaterials. The energy dissipation mechanism of the FCSPs matrix was further explored via FTIR analysis. Changes in the amide I bands indicated significant alterations in the secondary structure of MIX coacervates during loading and unloading cycles (Fig. 3m and Supplementary Table 4). Specifically, there was a pronounced decrease in β-crystal content post-tensile loading, accompanied by an increase in weaker β-sheet structures. After a 12 h recovery period, the β-crystal content rebounded to 90% of its pre-loading level (Supplementary Fig. 22). These shifts, typically associated with HBs formation and disruption, suggest that the HBs networks within the FCSPs matrix are essential for energy dissipation in the caterpillar cuticle. Interacting mechanism between two FCSPs To accurately delineate the interaction sites between FCSP-1 and FCSP-2, we initially investigated the specific regions within FCSP-1 responsible for this interaction. As depicted in Fig. 4a, the NTD of FCSP-1, unlike the C-terminal domain (CTD), interacts with FCSP-2 and facilitates coacervation. QCM-D analysis further corroborated that the NTD exhibits a significantly higher affinity for FCSP-2 compared to the CTD (Supplementary Fig. 23). Additionally, the conserved peptides DP37 and YF14, derived from the NTD of FCSP-1, demonstrated strong affinity for FCSP-2 (Fig. 4b and Supplementary Fig. 23). To further explore these interactions, we designed peptides DV36 and DV18 based on repetitive segments of FCSP-2 and study them together with DP37 and YF14 to facilitate subsequent nuclear magnetic resonance (NMR) and molecular dynamics (MD) simulations. The phase behavior of these four peptides was systematically examined under varying concentrations, temperatures, and ionic strengths. Our findings revealed that the conserved peptides exhibited phase behaviors consistent with their parent FCSPs (Fig. 4c–e and Supplementary Fig. 24). Notably, when mixed in physiological ratios (DP37:DV36 or YF14:DV18 at 1:5, hereafter referred to as MIXpep), the peptides demonstrated enhanced phase separation. Importantly, MIXpep coacervates formed by YF14 and DV18 underwent a liquid-to-solid phase transition, resulting in the formation of viscous gels (Fig. 4g). Interaction mechanism revelation via conserved peptides. a Turbidity of FCSP-2 and different parts of FCSP-1. Data are presented as mean values ± SD (n = 5 individual samples). b Conserved peptides of FCSPs. c Turbidity of single conserved peptides and their mixture at different peptide concentrations. Data are presented as mean values ± SD (n = 5 individual samples). d Turbidity of single conserved peptides and their mixture at different temperatures. Data are presented as mean values ± SD (n = 5 individual samples). e Turbidity of single conserved peptides and their mixture at different salt concentrations. Data are presented as mean values ± SD (n = 5 individual samples). f ζ potential of different peptides. Data are presented as mean values ± SD (n = 8 individual samples). g Confocal Microscopic images showing wild-type and L-mutant coacervates. Inset, fiber and gel formation in MIXpep coacervates. Scale bars, 10 μm. h 1H-13C-HSQC spectra of conserved peptides and mixtures. i Peptide-peptide interaction intensity diagram of different peptides. Red marks indicate residues involved in interactions. j Structural tendency diagram of different peptides. k Turbidity of different L-mutants at different peptide concentrations. Data are presented as mean values ± SD (n = 5 individual samples). (l) Secondary structure differences between wild-type and L-mutant peptides from the FTIR spectrum. Data are presented as mean values ± SD (n = 3 individual samples). The observed phase separation behaviors of these peptides under diverse conditions suggest critical molecular interactions governing coacervation. Enhanced coacervation at elevated temperatures and high ionic strengths indicates that hydrophobic interactions are principal driving forces, whereas electrostatic interactions serve as inhibitory factors (Fig. 4d, e). Furthermore, YF14 appears to promote phase separation by attenuating the surface charge of the coacervates through electrostatic interactions with DV18. This hypothesis is supported by ζ-potential measurements, which revealed that MIXpep coacervates possess significantly lower ζ-potentials compared to DV18 alone (Fig. 4f). To elucidate the intermolecular interactions between YF14 and DV18, solution-state NMR was employed. The introduction of both YF14 and DV18 into the system resulted in chemical shift perturbations predominantly in charged amino acids (Fig. 4h, i), along with noticeable changes in polar and hydrophobic amino acids (Fig. 4h, i). The 1H,13C-HSQC spectrum of aromatic residues also displayed chemical perturbations, indicating that aromatic residues are involved in the interactions (Supplementary Fig. 25). At initial stage of coacervation, there were no significant change in the two peptides secondary structure; YF14 remains in a disordered state, while DV18 exhibited as β-sheets (Fig. 4j). To further investigate the role of aromatic residues, we substituted tyrosine and phenylalanine with leucine in both peptides (YF14L, DV18L and L-mutant mixture MIXLpep) and assessed the phase separation behavior of the mutants. These mutations had minimal impact on the coacervation of individual peptides but significantly impaired YF14’s ability to promote DV18 coacervation (Fig. 4k) and prevented the formation of viscous gels (Fig. 4g). FTIR analysis revealed that wild-type peptide coacervates exhibited a higher proportion of β-sheet and β-crystalline structures compared to their mutant counterparts, suggesting that aromatic amino acids facilitate coacervate maturation and assembly through π-π interactions, thereby establishing a denser HBs network (Fig. 4l, Supplementary Fig. 26, and Supplementary Table 5). To gain deeper insights into the mechanisms of coacervation and the subsequent co-assembly of FCSP-1 and FCSP-2, we employed computational simulations. FINCHES39 predictions indicated that the regions spanning residues 64–101 of FCSP-1 and residues 21–325 of FCSP-2 exhibited significantly enhanced contact probabilities (Fig. 5a). The overlap between these contact regions and the DP37 (YF14) and DV18 peptides underscores the critical role these peptides play in the co-assembly of FCSP-1 and FCSP-2. Pathway of conserved peptides assembly. a Heatmap of interaction tendency driven by disordered regions, via FINCHES tool39. b Phase separation structures of MIXpep and DV in all-atom MD, with surface charge distribution shown (n = 3). c Interfacial interactions between peptides in the context of MIXpep. Purple, YF14, black, DV18. The Greek letters represent different chains. d Solvent accessible area (SASA) of MIXpep and DV systems at critical interacting residues. Data are presented as mean values ± SD (n = 18 dots for MIXpep, containing 3 individual simulations and 6 YF14 peptides or DV18 peptides in each simulation. n = 36 dots for DV18, containing 3 individual simulations and 12 DV18 peptides in each simulation). e Structures of MIXpep and L-mutant (MIXLpep) before and after disassembly, with surface hydrophilicity distribution shown (n = 3). f Number of HBs in each frame during the disassembly process (n = 3). g Secondary structures percentages of MIXpep and MIXLpep during the disassembly. Data are presented as mean values ± SD (n = 3 individual simulations). h Number of π-mediated contacts in each frame during the disassembly. Solid lines indicate the curve fitted according to the spline function. And the Pearson correlation coefficients were shown. i Scheme of key peptide assembly pathways. In MIXpep system, molecular proximity and multivalent interactions between peptides triggered by electrostatic interactions begin to form (e.g., π-mediated and hydrophobic interactions). The changed local chemical environment induces further reinforcement of molecular assembly and interactions, resulting in the formation of solid-like gels. In contrast, the absence of electrostatic shielding in systems lacking YF14, along with the lack of additional reinforcement in MIXLpep, results in the inability to form stable assemblies. We conducted all-atom MD simulations to investigate the phase separation of DV18 alone and in combination with YF14 (Fig. 5b, Supplementary Fig. 27, and Supplementary Table 6). In contrast to the loosely assembled DV18 system, the MIXpep system displayed stable assembly starting at 235 ns. Despite the MIXpep system containing a higher proportion of charged residues, it exhibited a lower surface potential and stronger electrostatic interactions, as −766.5 kJ/mol at the peptide interface compared to the DV18 system (Fig. 5c, Supplementary Fig. 27, and Supplementary Tables S7 and 8), consistent with ζ-potential and NMR results (Fig. 4f, i). Solvent-accessible surface area analysis revealed that the MIXpep system had a significantly lower surface area for charged residues compared to the DV18 system (Fig. 5d), suggesting that coacervation facilitates the encapsulation of more charges within the assembled structures (Fig. 5i). To investigate the role of aromatic residues in the co-assembly of YF14 and DV18, we performed all-atom MD simulations of the MIXpep and MIXLpep disassembly process, based on AlphaFold3-predicted structures. While MIXpep assemblies remained stable for up to 500 ns with a ΔGtotal of −133.2 kJ/mol (Fig. 5e and Supplementary Fig. 27), the MIXLpep assemblies disassembled more readily (with a ΔGtotal of −84.0 kJ/mol). The MIXpep system exhibited more hydrogen bonding and a higher content of regular secondary structures (30.4% of β-sheet) compared to MIXLpep (only 18.1% of β-sheet, Fig. 5f, g, Supplementary Fig. 28, Supplementary Table 5, and Supplementary Tables 7–9). That tendency is in good agreement with the secondary structure in FTIR (Fig. 4l), indicating the strong tendency of assembly and stability. Analysis also revealed that π-mediated interactions contributed a lot to the stability of the assemblies, remaining intact during the all-atom MD process, with a Pearson’s coefficient of -0.082 over time (Fig. 5h), indicating stable π-mediated interactions with time. In summary, we propose a molecular mechanism for the assembly process of YF14 and DV18 (Fig. 5i). Electrostatic interactions between YF14 and DV18 promote the formation of binary coacervates, which in turn reduce the surface charge and facilitate further co-assembly of the two peptides. Subsequently, π-mediated interactions play a critical role in stabilizing the coacervates, further enhancing gelation and promoting the formation of β-sheet-rich structures. These findings provide valuable insight into how FCSP-1 and FCSP-2 co-assemble to form a stable FCSP matrix that endows the cuticle with extraordinary mechanical properties. Design and fabrication of biomimetic impact-resistant materials We demonstrate that the formation of a matrix by FCSPs, including FCSP-1 and FCSP-2, via LLPS, plays a critical role in the strength and toughness of the Ostrinia furnacalis caterpillar cuticle. Phosphorylation has been shown to influence the fluidity of coacervates in other structural proteins, such as caddisfly silk40, cell cycle-specific protein Ki-6741, β-amyloid42, and artificial materials43. Notably, in vitro phosphorylation and dephosphorylation of FCSP-2 regulate its coacervation behavior and the fluidity of the resulting coacervates (Fig. 6a). The FRAP experiments showed that the fluorescence recovery was adjustable in the mixed-protein (MIX) coacervates with various phosphorylated-state FCSP-2, indicating the lowest internal molecular mobility and the formation of a more constrained network. Based on this, we hypothesize that biomimetic hydrogels derived from these materials could exhibit superior impact resistance, with their properties modifiable via phosphorylation of the FCSPs matrix (Fig. 6b). This material will provide a unique platform for designing programmable, high-performance materials for demanding applications. Biomimetic damper based on chitin and key FCSPs. a Curves and fluorescence diffusion coefficients of FCSPs’ coacervates FRAP under different phosphorylation states. Data are presented as mean values ± SD (n = 4 coacervates). b Scheme of in-situ reinforcement for insect cuticle and FCSP-chitin dampers via reversible phosphorylation modification strategy. c Radar map showing tensile mechanical performance of FCSP-chitin hydrogels under different phosphorylation states (n = 10). d Diagram of NCs and hydrogels reinforced by in situ phosphorylated FCSPs. Data are presented as mean values ± SD (n = 10 individual samples). e Ashby plot based on the hydrogel’s fracture energy versus tensile strength. The data in the Ashby plot include Ostrinia furnacalis’s cuticle and bioinspired damper in this work, chitosan (hydrated), cellulose hydrogels, chitin hydrogels (physically crosslinked), chitin hydrogels (double networked), alginate and BSA/alginate hydrogels, gelatin and gelatin/tannic hydrogels, hydrogels based on ChNF, ChNF/Ca2+, ChNF (surface deacetylated), ChNF/gelatin hydrogels, silk, and skin. f Dissipated energy in falling-ball experiments according to rebound height on different commercial materials. Data are presented as mean values ± SD (n = 5 samples). The dissipated energy during the first bounce cycle is shown. PAAM, polyacrylamide. HDPE high-density polyethylene, PP polypropylene, PVC polyvinyl chloride, PU polyurethane, EPE expanded polyethylene, CFRP cis-1,4 polybutadiene. The red star represents the composite hydrogel in this work. Insets, Digital photos of falling-ball experiments on a bio-inspired damper. Inset, dissipative mechanism proposed in this work. g Scheme and bar chart of flexible protective tests on BSISA-chitin and FCSP-chitin hydrogels. Digital photograph showing the protective effect of FCSP-chitin hydrogels on fragile hydrogel balls. The protective effect is attributed to a combination of compliance, stress redistribution, and energy dissipation of the FCSP-chitin matrix. Data are presented as mean values ± SD (n = 3 individual tests). To test this hypothesis, we prepared a series of hydrogels containing both phosphorylated and dephosphorylated FCSP-2 and characterized their mechanical properties, with a chitin hydrogel scaffold according to the binary-solvent-induced self-assembly (BSISA) strategy that we previously developed. Our results show that hydrogels containing dephosphorylated FCSP-2 exhibited the weakest mechanical properties, followed by those with phosphorylated FCSP-2. Interestingly, a hydrogel prepared first with phosphorylated FCSP-2, then dephosphorylated in situ, displayed the best mechanical properties (Fig. 6c and Supplementary Figs. 29 and 30). This hydrogel achieved a strength of 15.64 MPa and a work of fracture of 8.24 MJ/m3, corresponding to 3.65- and 14.31-fold improvements, respectively, compared to the natural caterpillar cuticle (Fig. 6d). The Ashby plot further highlights that the strength and work of fracture of our hydrogel surpass those of nearly all physically crosslinked hydrogels derived from natural materials (Fig. 6e). These exceptional properties suggest promising potential for applications requiring both flexibility and impact resistance, such as energy-dissipative components in soft robotics, wearable protective gear, or cushioning layers for fragile devices. To evaluate the flexible protective performance under dynamic loading, we evaluated the impact resistance of the FCSP-chitin hydrogels with two experiments. First, falling ball tests assessed the energy dissipation of the hydrogels (Fig. 6f and Supplementary Fig. 31). The FCSP-chitin hydrogel stabilized the falling ball rapidly, showing a 100% reduction in rebound height compared to a chitin-only hydrogel during the initial rebound, indicating superior energy absorption under the test conditions. Compared to other commercial impact-resistant materials, the FCSP-chitin hydrogels exhibited superior buffering performance under the tested conditions, as evidenced by lower rebound height and greater energy absorption during the first impact cycle. Boundary effects may influence the impact response of thin (0.5 mm) layers, yet the observed protective advantage at this biologically relevant thickness is statistically significant. Building on these observations, we next evaluated the hydrogels’ ability to protect fragile objects in a realistic protective scenario. When a soft hydrogel ball was protected in FCSP-chitin hydrogel, it remained intact after a 50 g weight was dropped from a height of 1 m (Fig. 6g and Supplementary Movie 1). In contrast, commonly used packaging materials and chitin hydrogels failed to protect the objects and exhibited a violent rebound under identical conditions, underscoring the potential of FCSP-chitin hydrogels in protective packaging applications and functional energy-absorbing materials (Fig. 6g, Supplementary Fig. 32, and Supplementary Movie 2). FCSP-chitin hydrogels exhibited better protective performance under low-intensity impact, which could be attributed to a combination of compliance, stress redistribution, and energy dissipation (Supplementary Figs. 33 and 34). More broadly, this work demonstrates a viable strategy for creating tunable, high-toughness hydrogels with direct implications for fields ranging from soft robotics to biomedical device coatings. Discussion Previous studies on the mechanisms of natural biomaterials have primarily focused on rigid materials such as bone44, stomatopod dactyl club45, beetle elytra26, shell hinges46, and squid beaks5. Notably, seminal work on systems like the squid beak and mantis shrimp hammer has elucidated design principles for achieving extreme hardness or gradient stiffness, often employing strategies such as covalent cross-linking to reinforce protein coacervates within chitin scaffolds. In contrast, we propose a plausible impact-resistant mechanism of the soft cuticle in boring insects. In our reconstituted systems, the matrix formed by two structural proteins through LLPS can fill the chitin scaffold, and the entire system is stabilized by non-covalent interactions, including hydrogen bonds, electrostatic interactions, and π-mediated interactions. We suggest that the reversible, dynamic network facilitates energy dissipation at the molecular level, manifesting as macroscopic viscoelasticity. This viscoelasticity enables compliance, stress redistribution, and impedance matching, allowing the material to withstand repeated impacts without permanent deformation or catastrophic failure while retaining softness. This work thus offers a plausible molecular mechanism for the observed mechanical behavior and provides a valid bioinspired design route for synthetic impact-resistant materials. Further analysis revealed that in boring insects, the homology between these two proteins is higher, suggesting that this co-assembly mechanism may represent a molecular pattern by insects to meet their physiological needs of tunneling (Supplementary Fig. 35 and Supplementary Table 10). High sequence homology likely enhances molecular cooperativity, drives more efficient LLPS, and the formation of a seamlessly integrated dynamic network. It is therefore plausible that this system has evolved to meet specific mechanical demands, such as generating the sustained driving force required for boring. Consistent with this hypothesis, the energy dissipation mechanism we describe operates optimally within a defined regime of moderate impact frequency and intensity, which is precisely the condition encountered in the insect’s natural boring behavior. While this inherent specialization may define the core advantage of this bioinspired hydrogel, i.e., flexible impact resistance within its operational window. Thus, our work bridges fundamental understanding and application by demonstrating the design of a bioinspired material inspired by this plausible mechanism. We have designed a biomimetic hydrogel with a work of fracture as high as 8.6 MJ/m3, significantly surpassing other high-performance hydrogels previously reported for flexible impact resistance47,48, such as salt-welding strategy49 and double network hydrogels50. Therefore, this hydrogel holds great potential for energy absorption and the protection of fragile objects. Even more remarkably, the hydrogel’s mechanical properties can be further tuned through biomimetic phosphorylation post-fabrication, allowing the same hydrogel to protect fragile objects of varying properties. More broadly, this study establishes a biomimetic design, leveraging LLPS-driven protein assembly and post-assembly phosphorylation for engineering advanced materials for soft robotics, adaptive protective wearables, and bio-integrated devices requiring programmable mechanical response. In summary, this work proposes a distinct biological strategy for impact resistance, where LLPS of two structural proteins forms a dynamic network through reversible non-covalent interactions, achieving both efficient energy dissipation and softness. The bioinspired hydrogel designed based on this principle not only exhibits outstanding impact resistance but also demonstrates, for the first time, the capability for mechanical programming via phosphorylation. Altogether, this LLPS-driven design extends the performance range of such materials and integrates additional stimuli-responsive elements. These programmable hydrogels thus establish a versatile platform for advanced applications in soft robotics, adaptive wearables, and bio-integrated devices. Methods Ethics statement The insect used in this study is the Asian corn borer (Ostrinia furnacalis), an invertebrate species. According to institutional and national regulations, ethical approval is not required for research involving invertebrates. All experimental procedures complied with international animal welfare standards, and every effort was made to minimize animal suffering. Sample collection was carried out under relevant permits and in accordance with local legislation. Materials Raw α-chitin powder extracted from crab shells and 1,1,1,3,3,3-hexafluoroisopropanol (HFIP) were purchased from TCI Chemical Co., Ltd (Shanghai, China) and Macklin Biochemical Co., Ltd (Shanghai, China), respectively. DP37 peptide was purchased from GL Biochem Co., Ltd (Shanghai, China). Rectangular polydimethylsiloxane (PDMS, Sylgard 184, Dow Corning Co., Ltd, USA) was used to prepare hydrogels as molds. Unless otherwise stated, all the other reagents were purchased from Aladdin Inc., and all materials were used without further purification. Insect cuticle collection After being fed until the fifth day of the fifth instar, the Asian corn borer Ostrinia furnacalis was dissected to scrape away all the tissue. Then, cuticles were washed thoroughly with phosphate-buffered saline. To exclude the influence of the tissues, specimens for post-translational modifications (PTMs) were collected in lysis solution with inhibitors of phosphatase and phosphokinase at 4 °C, including 100 mM phenylmethylsulfonyl fluoride (PMSF), 10 mM sodium fluoride (NaF), and 20 mM sodium pyrophosphate (Na4P2O7). For cuticle deproteinizing, NCs were soaked at 50 °C in 92.5% H2O, 5% 30% H2O2, and 2.5% 10 N NaOH until de-pigmentation and deproteinization were achieved, then the deproteinized samples were applied to experiments. After being freeze-dried, the pristine cuticle was turned into dry specimens51. FCSP's collection of Asian corn borer To extract FCSPs from the cuticle, cuticle samples were ground evenly in liquid nitrogen. Then, the SDT buffer (4% SDS, 100 mM Tris-HCl, and 1 mM DTT) was used for lysis and protein extraction at 37 °C for 2 h. After that, the sample was centrifuged at 12,000×g at 4 °C for 40 min. Protein was digested by trypsin, and the peptides were desalted on C18 cartridges (Empore SPE Cartridges C18, volume 3 mL, Sigma-Aldrich Co., Ltd, China). Peptide samples were concentrated by vacuum centrifugation and reconstituted in 0.1% (v/v) formic acid. Phosphorylation identification of FCSPs in NC Samples were reconstituted in 1.4 mL of precooled immunoaffinity purification buffer (IAP buffer, 50 mM MOPS, 10 mM Na2HPO4, 50 mM NaCl, pH 7.2) for phosphor-peptide enrichment. Then TiO2 beads were added into the mixture, shaken for 40 min, and centrifuged for phosphor-peptides enrichment. The beads were washed 3 times. Phosphor-peptides were eluted and concentrated, then dissolved in 0.1% (v/v) formic acid for MS analysis. And the ground cuticle was separated with SDS-PAGE and stained with the fluorescent phosphoprotein stain Pro-Q Diamond (Molecular Probes, Eugene, OR, USA)36. LC-MS/MS analysis LC-MS/MS spectra were performed by a Q-Exactive mass spectrometer (Thermo Scientific) that was coupled to Easy nLC (Thermo Scientific) for 120 min. The sample solution was separated using a C18-reversed-phase analytical column (3 μm resin, Thermo Scientific), with a linear gradient of elution buffer (84% acetonitrile and 0.1% formic acid). The mass spectra were acquired in positive ion mode over a range of 300–1800 m/z. To identify and estimate protein abundance, fragmentation spectra were searched against the Ostrinia furnacalis full-protein database by MaxQuant 1.5.3.17 software. The MaxQuant identification indexes were set as follows: Trypsin, max missed cleavages was 2, fixed modifications were carbamidomethyl, variable modification was Phospho (STY)/Phospho (Y). The database was selected as NCBI_moths_Ostrinia furnacalis. Peptide FDR ≤ 0.01, site FDR ≤ 0.01, protein ≤ 0.01. Each sample was analyzed three times. The mass spectrometry proteomics data have been deposited in the Proteomexchange Consortium database [http://proteomecentral.proteomexchange.org] with the dataset identifier PXD027277 as our group’s previous work52. And data of phosphoproteomics are available at iProX53 [https://www.iprox.cn/page/home.html] with the dataset identifier IPX0015633000. Composition analysis The composition analysis was conducted according to generalized biochemical analytical methods51. Briefly, samples for content determination were prepared by alkaline peroxidation or acid hydrolysis. First, the water on the surface of the freshly dissected cuticle was gently removed, and the sample was immediately weighed (wet weight, Ww), freeze-dried, and reweighed (to ascertain water content, Wd). Then, the water content (ωwater) was calculated. For chitin content, samples were soaked at 50 °C in a solution of 92.5% H2O, 5% 30% H2O2, and 2.5% 10 N NaOH, all by volume. The peroxidation treatment was repeated until full hydrolysis was achieved. The remaining solid was then freeze-dried and weighed again to determine the chitin content (ωchitin). Acid hydrolysis was performed to determine pigment content (ωpig). Samples were soaked in vacuum at 110 °C for 24 h in a solution of 5% phenol (an antioxidant) in 6 M HCl. The soluble degradation products were centrifuged at 10,000 rpm for 30 min. The sediment was collected, freeze-dried, and weighed for pigment content. At last, the protein content (ωprotein) was acquired by difference. Recombinant cuticular protein expression and purification The recombinant plasmids containing FCSPs were transformed into BL21 Star (DE3) Escherichia coli competent cells (Novagen) individually. The strains were fermented in LB medium containing 50 mg/L kanamycin to an OD600 of 0.8 at 37 °C. Protein expression was induced with 0.5 mM Isopropyl-β-D-thiogalactopyranoside (IPTG) at 37 °C for 5 h. The expressed bacteria were collected by centrifugation at 10,000 rpm at 4 °C for 10 min, and the pellets were suspended in Tris-HCl buffer (200 mM NaCl, 50 mM Tris-HCl, pH 7.4) and lysed under ultrasonication for 5 min with an ultrasonic cell disruptor (JY 92-IIDN, Yiheng Co., Ltd, China). For FCSP-1, the soluble protein was centrifuged at 17,000×g at 4 °C for 40 min. The supernatant was bound to a Ni-NTA column (Cytiva Co., Ltd, Sweden) and eluted with buffer (200 mM NaCl, 50 mM Tris-HCl, 500 mM imidazole, pH 7.4). Then, the protein solution was dialyzed in a cellulose dialysis membrane soaked in ice-cold dialysis buffer (20 mM NaCl, 20 mM Tris-HCl, pH 7.4), to remove imidazole at 4 °C overnight. For FCSP-2, the insoluble inclusion body was collected by centrifugation at 10,000 rpm for 20 min, purified five times with 50 mL PBS, and then purified five times with 50 mL 1 M urea buffer (1 M urea, 100 mM NaCl, 20 mM Tris-HCl, 1 mM EDTA, 1% (v/v) TritonX-100, pH 8.0). The purified inclusion body was then dissolved in 8 M urea buffer and dialyzed in dialysate buffer (20 mM NaCl, 20 mM Tris-HCl, pH 7.4) for protein refolding. Micro-structure imaging Micro-structure imaging was collected using a field emission-scanning electron microscope (SEM, Hitachi SU8200, Japan), with 2 kV accelerating voltage. For cross-sectional images, specimens were frozen-quenched and prepared. In detail, the dissected cuticle samples were rapidly frozen by plunging into slushy nitrogen. They were then broken off with pre-cooled tweezers under liquid nitrogen to expose a clean cross-section. The fractured samples were transferred under vacuum to a freeze-dryer (lyophilizer) and dried for 48 h to sublime the ice without introducing capillary forces that could collapse the hydrated structure. Finally, the dried samples were sputter-coated with a 3-nm platinum layer prior to SEM imaging before the imaging of the surfaces and the cross sections of the fibers. Chitin-binding activity analysis A α-chitinous substrates binding assay was performed to verify the chitin-binding capabilities of FCSPs. To ensure that the α-chitin was insoluble, the α-chitinous substrates (from crab shell) were prewashed with a buffer of 20 mM Tris-HCl, 100 mM NaCl (pH 7.4). Then the reaction mixtures (100 μL) consisting of 5 mg/mL α-chitin and 0.5 mg/mL purified recombinant proteins in binding buffer (20 mM Tris-HCl, 100 mM NaCl, pH 7.4) were incubated with end-over-end rotation at 60 rpm for 12 h at 4 °C. The mixtures were then centrifuged at 17,000×g at 4 °C for 10 min. The chitin was washed three times with 1 mL of washing buffer (20 mM Tris-HCl, 500 mM NaCl, pH 7.4). Finally, the chitin-binding activity of proteins was determined by two methods, the BCA Protein Assay Kit and SDS-PAGE followed by CBB staining. In detail, the chitin-binding activity was analyzed indirectly by BCA determination and directly by gray-scale scanning. To determine the co-assembly properties of α-chitin and recombinant cuticular proteins, α-chitin nanofiber from Phaeocystis globosa (NIES-1396) was cultured and purified as chitinous substrates. The chitin nanofiber and proteins were co-incubated as above. Then the co-assembly was freeze-dried and detected with FE-SEM. Quasi-in situ co-assembly was performed with purified protein and deproteinized insect cuticle. Like the chitin-binding assay, cuticle and protein were co-incubated for 12 h at 4 °C. Then the replenished effect was determined with free protein. Laser scanning confocal microscope (LSCM) and FE-SEM were used for further determination. In vitro LLPS assay In vitro LLPS assays were conducted in a total of 100 μL. Purified proteins in 20 mM Tris-HCl, 20 mM NaCl, pH 7.4 were mixed with the corresponding buffer and incubated for 30 min at room temperature (25 °C). The mixtures were placed into 96-well cell culture plates. Then the turbidity was determined at 405 nm. Five repeats were studied. Preparation of FCSP-chitin complex hydrogels The chitin hydrogel was prepared according to the BSISA strategy that we previously developed54 (BSISA-chitin hydrogel). In detail, after being purified and dried, chitin (100 mg, adjustable, from crab shell) was added to 5 mL of HFIP solution and vortexed at room temperature for two days to aid in the dissolution of the chitin. Then the mixture was ultrasonicated for 1 h for deaeration. The mixture was then plated (Jet Bio-Filtration Co., Ltd, Guangzhou, China) and stored exposed to air at room temperature, until the solution was reduced to the pre-set concentration (2% w/v). A mixture of IPA/HFIP with a specific concentration of IPA was then added to the chitin/HFIP solution. The mixture was stored in PDMS molds hermetically at room temperature for 1 day for gelation. The resultant gel was an organic gel. And then the organic gel was washed at least six times with deionized water for solvent exchange and annealing the self-assembly of chitin. Thus, the BSISA-chitin hydrogel was prepared. To prepare FCSP-chitin hydrogels, FCSP-1 and FCSP-2 (in their native or phosphorylated forms, as indicated) were incorporated into the network via incubation in protein solution, allowing chitin-binding and coacervation-mediated protein matrix formation within the hydrogel pores. Then the FCSPs were replenished into the chitin hydrogel through chitin binding. The replenishment effect was determined by detecting the protein concentration in solution. Tensile testing All the mechanical tests were conducted on a uniaxial MTS machine (10 N load cell, MTS-E40, MTS system Co., Ltd, China). The mechanical sensor was 50 N, and 1 mm/min was set as the stretching velocity. Insect cuticles were carefully loaded with clamps with an initial length of 5 mm. And the two sides of the spline were wrapped by elastic foam to avoid sample damage during clamping. Mechanical performance testing of the hydrogel was measured with a uniaxial MTS machine as above. For each measurement, ten specimens were measured in this manner. Self-recovery tests For self-recovery of cuticle in different solution conditions, cuticles were tested after recovery time, for different treatment times. For NCs recovered in PBS (NC), the cuticle was stretched first and then recovered in PBS for different times. For DCs recovered in PBS (DC), the cuticle was deproteinized and stretched. Then the DC was recovered in PBS for different times. For NCs recovered at 50% relative humidity (NC in AIR), the NC was stretched first and then recovered in air (RH = 50%) for different times. For NC in ethanol, the NC was stretched first and then recovered in 20% ethanol for different times. The load rate was 1 mm/min for all tests. And for each measurement, 10 specimens were measured at room temperature (25 °C). FRAP test FRAP experiments were performed on an Olympus FV1000 (Japan). FCSP-1 and FCSP-2 were labeled with either rhodamine B (RhoB) or fluorescein isothiocyanate (FITC). Imaging was performed with a 488 nm laser. Circular bleach regions were drawn in the protein droplets. Recovery curves were fit to an infinite boundary model in three dimensions to calculate the recovery timescale τ55. The diffusion coefficient was calculated as D = R2/τ. And the same Olympus microscope was used for co-localization of two key proteins and quasi-in situ co-assembly of protein on DC. Glutaraldehyde crosslinks of proteins To verify the close distance of key proteins in condensates, glutaraldehyde cross-links were performed. Stock glutaraldehyde solutions (25%, w/v) were added to make 10 μM concentration, and the mixture was incubated for 4 h at room temperature. The suspension was dialyzed in dialysate buffer (20 mM NaCl, 20 mM Tris-HCl, pH 7.4) overnight to remove free glutaraldehyde and terminate the crosslinks. Then the crosslinks were identified by SDS-PAGE. QCM-D The interaction and adsorption of guest proteins on the host-substance surfaces was investigated using QCM-D (QCM-D E4, Biolin Co., Ltd, Sweden). For interaction between chitin and protein, the SiO2-coated quartz sensor chip (QSX 303, Biolin Co., Ltd, Sweden) was used. For all protein-protein interaction experiments, commercially available QSX 301 (Biolin) sensors coated with gold under a chromium adhesion layer were used56. The sensor chips were cleaned by UV-ozone treatment for 10 min, rinsed twice with deionized (DI) water, and dried with N2. The changes in resonance frequency (ΔF) and dissipation (ΔD) were recorded at the fundamental frequency (4.95 MHz) and its 3rd, 5th, 7th, 9th, and 11th overtones. All experiments were performed at a constant flow rate of 100 μL·min-1 and a constant temperature of 25 °C. The baseline drift was subtracted from the data. The surface density (Δm) was calculated from the negative shift in ΔF and ΔD using the Sauerbrey equation. The layer parameters to be fitted were kept within boundaries: viscosity (0.0001–0.1 kg/(m·s)), shear (100 – 1) × 105 Pa). In this work, the 7th overtone signal was adopted for this calculation57,58. Stress visualization based on FRET In order to evaluate the stress of coacervates upon loading, a FRET sensor was designed based on a conserved peptide segment of FCSP-1, i.e., DP37. We used the sensor module of mECFP/mYPet (donor/acceptor). And a (GPGGA)8 molecular linker from spidroin was inserted into the fluorescence donor and acceptor. And two DP37 were distributed on either flank of the sequence (shown in supporting information). Purification of the DP37-FRET stress sensor was performed on an Ni-NTA column as above. Buffer with 100 mM imidazole was used to elute DP37-FRET. After desalination via dialysis, the recombinant proteins FCSP-1, FCSP-2, and DP37-FRET were mixed in a ratio of 2:2:1. Fluorescent images were analyzed with Fiji, a distribution of ImageJ. Fourier transform infrared spectroscopy (FT-IR) In order to study the secondary structure of protein, FCSPs were deuterated first. In detail, Protein samples for FTIR analysis were prepared in deuterated buffer (20 mM Tris-DCl in D2O, pD 7.4) and were subjected to multiple rounds of thermal cycling (lyophilization and re-dissolution in D2O). And then, an attenuated total reflectance-Fourier transform infrared (ATR-FTIR) spectrometer (Nicolet iS5 FT-IR Spectrometer, Thermo Scientific) was used, equipped with a Diamond/ZnSe crystal. Measurements started after stabilizing the sample state for 1 min. For each measurement, 64 scans were co-added with a nominal resolution of 4 cm–1. In detail, the FTIR spectra were collected with the software OMNIC, and the background (D2O) was subtracted in the software Peakfit (version 4.12). The semiqualitative content of three primary conformations/secondary structures of FCSPs, including β-sheet, random coils, α-helix, and β-turns, was analyzed by the spectral deconvolution of the amide I band (1580–1720 cm−1). The range of Gaussian peaks for each molecular conformation was assigned59. NMR spectroscopy and data collection All NMR experiments were conducted on a Bruker Avance III 600 MHz spectrometer equipped with a cryoprobe. The typical concentration of the samples was 2 mM. The data were processed using Topspin 4.1.0 (Bruker) and analyzed with NMRFAM-SPARKY60. Two-dimensional 1H–1H COSY (correlation spectroscopy), TOCSY (total correlation spectroscopy), and NOESY (nuclear Overhauser spectroscopy) spectra were acquired for the peptides DV18, YF14, and their mixtures at a concentration of 2 mM in a 50 mM phosphate buffer (comprising Na2HPO4 and NaH2PO4) at pH 7.0. Two-dimensional 1H–13C HSQC (heteronuclear single quantum coherence) spectra were acquired for the peptides DV18, YF14, and their mixtures (MIXpep) at a concentration of 2 mM in a 50 mM phosphate buffer (comprising Na2HPO4 and NaH2PO4) with all the protons in the buffer replaced by deuterium. The experiments involving individual peptides and peptide mixtures were conducted under identical experimental conditions and utilized the same experimental parameters. All-atom MD simulations Atomistic MD simulations were conducted using GROMACS 2022.461, applying the Amber99SB-ILDN force field62. Because this force field was specifically optimized to address known limitations of fixed-charge force fields, including those related to side-chain conformational sampling and the description of key non-covalent interactions such as π–π and cation–π contacts. The initial conformations of the phase separation structures of MIXpep and DV18 were generated using Packmol63, where 12 peptides were randomly placed within each cubic box (with equivalent concentrations ranging from 9.1 to 49.2 mM according to different experimental purposes). The initial conformations of the peptide mixture were predicted by AlphaFold338, maintaining a 1:1 molar ratio of YF14 and DV18. The resulting structures were solvated using the TIP3P water model, and a cubic simulation box was constructed to facilitate the computation of molecular interactions, with 100 mM of NaCl (i.e., 120 Na+ and Cl− in a 2048.4 nm3 cubic box). The dimensions of the simulation box were carefully chosen to minimize the effects of periodic boundary conditions. Non-bonded interactions were truncated at a cutoff distance of 10 Å. The system was first equilibrated by applying the steepest descent method for 5000 steps, with peptide atoms restrained at 10 kcal/mol. Subsequently, the systems were gradually heated to 298 K over 100 ps, with a 20 kcal/mol restraint on the peptide complex. This was followed by a 1 ns isothermal-isobaric (NPT) simulation and a 1 ns canonical (NVT) simulation, each performed with a 5 kcal/mol restraint. Finally, a 500 ns MD run was executed to further equilibrate the system and enable extensive sampling. And the quantitative analyses were performed only over equilibrated frames (250–500 ns). All MD simulations were carried out using 2 fs time steps, with temperature control managed by the V-rescale thermostat. And the Berendsen thermostat was used for isobaric coupling during the equilibration phase. In order to obtain a more comprehensive understanding of the energetic contributions, the molecular mechanics/Poisson-Boltzmann surface area (MM/PBSA) method64 was employed to decompose the binding free energy. This methodology allows for the quantification of individual contributions from electrostatic interactions, van der Waals forces, polar solvation energy, and nonpolar solvation energy to the overall binding free energy. During the calculations, 20 ns intervals were extracted from key snapshots along the 500 ns MD simulation trajectory, with one representative frame sampled every 10 ps. This sampling strategy ensures comprehensive coverage of the system and allows for accurate characterization of the interactions. Each simulation was performed for three repeats. And the dataset was deposited in Figshare (10.6084/m9.figshare.30193804). In-situ phosphorylation modification of FCSP-2 based on the reaction of sodium tripolyphosphate (STP) and λ alkaline protein phosphatase FCSP-2 was dissolved in buffer (20 mM Tris-HCl, pH 7.4) at a concentration of 0.1–0.5 mg/mL. STP was added at a final concentration of 0.5 wt.% and stirred vigorously at room temperature for 3 h. Then, the solution was dialyzed into the buffer to remove impurities, and the phosphorylation-modified proteins were obtained. In situ dephosphorylation based on. Protein samples were dialyzed against BTP buffer (20 mM BTP, 100 mM NaCl, 1 mM MnCl2) at pH 7.4. And then FCSP-2 was treated with λ alkaline protein phosphatase (0.1 U/μL) at 37 °C. Falling-ball tests In order to study the dissipative contribution of the FCSPs matrix to the chitin base, classical falling-ball tests were performed. In brief, a 60-mm diameter hydrogel (or other commercially obtained materials, with a thickness of 0.5 ± 0.1 mm) pad was used. An iron plate was used as a control. A glass ball with 9 mm in diameter was dropped onto the different base materials from a height of 15 cm and the bounce height of every bounce was measured using video. All tests used n = 5 independent replicates. Protective tests on fragile hydrogel balls To evaluate the protective capability of the FCSP-chitin hydrogel, a soft hydrogel sphere (1 cm diameter, modulus < 10 kPa) was used as a model fragile object. In the encapsulation test, a sphere was embedded in a 5-mm-thick FCSP-chitin hydrogel block and impacted by 25-g and 50-g weights dropped from 50 and 100 cm. The samples were fixed by the pipelines. In the coating test, a sphere was covered with a 3-mm-thick hydrogel layer and impacted by dropped weights. BSISA-chitin hydrogel was tested under identical conditions as a control. After impact, sphere integrity was visually inspected for rupture or irreversible deformation. Each condition was tested in triplicate. Quantitative flexible protective performance test A fragile hydrogel ball (0.23 g, diameter 1 cm, modulus < 10 kPa) was launched upward from 2 m to impact a 3 mm protective hydrogel (FCSP‑chitin or chitin‑only) fixed beneath a 10 N load cell (MTS E40, ±0.5% accuracy, ≤ 1 ms response). Force‑time signals were recorded at 100 Hz. Dynamic calibration against high‑speed video (1000fps) gave a measured impulse within ±5% of the theoretical value. Peak force was read from the trace, and impulse was integrated over the contact time. All conditions used n = 5 independent replicates. Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Supplementary information Supplementary Information Description of Additional Supplementary Files Supplementary Movie 1 Supplementary Movie 2 Reporting Summary Transparent Peer Review file Source data Source Data Acknowledgements The authors would also like to acknowledge Prof. Huajian Gao and Prof. Jing Yu from the School of Aerospace Engineering of Tsinghua University and Prof. Qing Yang and Prof. Xiangyu Liang from the Agricultural Genomics Institute at Shenzhen, Chinese Academy of Agricultural Sciences for their valuable discussions on this work. We thank Hefei National Research Center for Interdisciplinary Sciences at the Microscale for the assistance with NMR measurements. Author contributions T.L. designed and supervised the study. F.Y., S.H., Z.P. and Y.C. carried out the experiments and analyzed the data. Z.P. and X.Q. carried out the all-atom MD and analyzed the data. S.H. and C.S. carried out the NMR measurements. F.Y. and T.L. wrote the manuscript. Peer review Peer review information Nature Communications thanks Valentina Perricone and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available. Funding The work was supported by the National Natural Science Foundation of China (31871959, 32170502 and 32572810). Data availability The data that support the findings of this study are available from the corresponding author on reasonable request. Source data are provided with this paper. Proteomics data are available at the Proteomexchange Consortium database [http://proteomecentral.proteomexchange.org] with the dataset identifier PXD027277. Data of phosphoproteomics are available at iProX [https://www.iprox.cn/page/home.html] with the dataset identifier IPX0015633000. And the dataset of all-atom MD and NMR is available at Figshare with the DOI [10.6084/m9.figshare.30193804]. Unless otherwise stated, all data supporting the results of this study can be found in the article, supplementary, and source data files. Source data are provided with this paper. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. These authors contributed equally: Fenghou Yuan, Shuaifei Hu, Zihan Pang, Yuntian Cui. Supplementary information The online version contains supplementary material available at 10.1038/s41467-026-75849-1. References Associated Data Supplementary Materials Supplementary Information Description of Additional Supplementary Files Supplementary Movie 1 Supplementary Movie 2 Reporting Summary Transparent Peer Review file Source Data Data Availability Statement The data that support the findings of this study are available from the corresponding author on reasonable request. Source data are provided with this paper. Proteomics data are available at the Proteomexchange Consortium database [http://proteomecentral.proteomexchange.org] with the dataset identifier PXD027277. Data of phosphoproteomics are available at iProX [https://www.iprox.cn/page/home.html] with the dataset identifier IPX0015633000. And the dataset of all-atom MD and NMR is available at Figshare with the DOI [10.6084/m9.figshare.30193804]. Unless otherwise stated, all data supporting the results of this study can be found in the article, supplementary, and source data files. Source data are provided with this paper.