Atomic-precision π-driven peptide hydrogel nanofibers with ordered water channels

Abstract

Aromatic interactions organize molecules into ordered supramolecular architectures, while peptides form functional soft materials through hydrogen bonding and water-mediated assembly. In peptide-based systems, strong aromatic stacking is typically achieved by terminal capping, whereas terminally uncapped peptides organize water through polar end groups but rarely form highly ordered materials. Here we show that a π-extended aromatic unit can be integrated into a terminally uncapped peptide to create a class of supramolecular hydrogels with structural order. A pyrene-modified dipeptide hierarchically assembles into monodisperse helical nanofibers and self-healing hydrogels. Cryo-electron microscopy resolves the nanofibers at near-atomic precision (1.7 Å), revealing tightly packed protofilaments, continuous ordered water channels, and a unidirectional dipole extending along the fiber. These results demonstrate how reinforced aromatic stacking, polar interactions, and cooperative water organization can be orchestrated to generate emergent electrostatics and mechanical resilience, bridging conjugated materials and biomolecular matter, enabling functional soft materials inaccessible to either domain alone.

In peptide-based systems, aromatic stacking is usually achieved by terminal capping, but here the authors show that a π-extended aromatic unit can be integrated into an uncapped peptide using a pyrene-modified dipeptide that assembles into monodisperse helical nanofibers and forms hydrogels. Cryo-electron microscopy resolves the nanofibers at near-atomic resolution, revealing their tightly packed supramolecular architecture and continuous ordered water channels.

Introduction

Aromatic units play a central role as fundamental building blocks in the construction of supramolecular architectures^1–4. Their diverse π-interactions, ranging from π–π stacking^5–7 to CH–π interactions⁸ and cation–π interactions^9,10, not only guide self-assembly and molecular recognition, but also underpin the stability, orientation, and functional organization of supramolecular architectures^11–14. Expanding the π-surface area of molecular building blocks is a powerful strategy to amplify these π-interactions, enabling access to higher-order supramolecular structures with enhanced stability and emergent properties^15–17. Such enlarged π-planes can also endow the resulting supramolecular materials with mechanical and optoelectronic characteristics, positioning aromatic units among the most influential motifs in supramolecular design¹⁸.

In peptide-based supramolecular systems, backbone hydrogen bonding drives primary assembly, while aromatic residues such as phenylalanine contribute secondary π-interactions that reinforce and direct ordered nanostructures^19–22. Beyond natural residues, a widely used strategy has been to introduce noncanonical, π-expanded aromatic groups at peptide N-termini, leading to numerous mechanically robust peptide hydrogelators since the first examples in 1995 (Fig. 1a)^23–30. However, such N-terminal modification eliminates terminal polarity. By contrast, terminally uncapped (unprotected) peptides such as diphenylalanine (FF), the smallest known self-assembly unit (Fig. 1a)^31–33, use their ionic termini to organize water molecules into structured hydrogen-bond networks, giving rise to characteristic channel-like assemblies and functional properties^34,35. Importantly, for FF, the ordering of water has been demonstrated only through single-crystal X-ray structures and computational models. Despite the complementary advantages of π-extension and terminal polarity, their integration within a single peptide framework has remained largely unexplored³⁶.

Fig. 1. Strategy of creating hydrogelators.

a Supramolecular interactions directing the self-assembly of peptide hydrogels. b Molecular design concept and resulting properties demonstrated in this work.

This gap motivated us to design peptide building blocks that preserve terminal polarity while incorporating exceptionally strong π-interactions, aiming to create supramolecular hydrogelators with structural and mechanical characteristics. Here we report a terminally uncapped, non-natural dipeptide, FQ(Pyr), in which L-phenylalanyl-L-glutamine (FQ) is functionalized with a pyrene-based π-extended aromatic unit on the side chain (Fig. 1b). FQ(Pyr) undergoes highly ordered hierarchical self-assembly, forming nanoaggregates, uniform helical nanofibers (~100 Å diameter), and robust hydrogels through pH-triggered transitions. To resolve the architecture of the fibrils within the hydrogel, we employed cryo-electron microscopy (cryo-EM), which has recently emerged as a powerful tool for visualizing peptide helical assemblies in aqueous environments^37–44. Cryo-EM analysis at 1.7 Å resolution revealed an atomically precise supramolecular fiber featuring channel-like water-filled pores and a well-defined molecular arrangement stabilized by synergistic π and polar interactions. This structural detail clarifies the features of the helical fibers, the cooperative forces underlying their formation, and the principles governing peptide assembly in water. Overall, our findings demonstrate that combining terminal polarity with π-augmented aromatic groups fundamentally reshapes peptide self-assembly, enabling a previously inaccessible class of peptide-based hybrid supramolecular architectures and establishing a design framework for future molecular materials.

Results

Hydrogel formation and mechanical properties of uncapped, π-augmented dipeptide FQ(Pyr)

The non-natural dipeptide FQ(Pyr) was synthesized by introducing a pyrenyl group as a π-augmented aromatic unit on the side chain of L-phenylalanyl-L-glutamine (FQ) [synthesis section of Supplementary]. As a control, we also prepared and analyzed FQ(Ph), a smaller analogue with a phenyl substituent. Both compounds were subjected to a pH-triggered gelation assay using glucono-δ-lactone (GdL) as a slow acidifier (Fig. 2a, Supplementary Movie S1)^45,46. Upon acidification from pH 12 to ~4, FQ(Ph) formed a white opaque gel, whereas FQ(Pyr) yielded a transparent self-supporting hydrogel (Fig. 2b). Notably, the critical gelation concentration (CGC) of FQ(Pyr) was substantially lower than that of FQ(Ph), with gelation observed at around 0.1 wt% versus 0.75 wt%, respectively (Fig. 2b and Supplementary Fig. S1). Here, CGC is defined as the minimum concentration at which the sample retains water at the bottom of the vial upon inversion. Below the CGC, although self-assembly into fibrous structures occurs, an interconnected network capable of immobilizing water is not fully formed. The critical gelation pH (CGpH) was determined by varying the amount of GdL (Fig. 2b and Supplementary Fig. S2). Both peptides gelled across a broad pH window [FQ(Pyr): ~ pH 9–10; FQ(Ph): ~ pH 8], but FQ(Pyr) initiated gelation at higher pH.

Fig. 2. Hydrogelators FQ(Pyr) and FQ(Ph).

a The method of hydrogelation and structure of FQ(Pyr) and FQ(Ph). b Properties of FQ(Ph) and FQ(Pyr) hydrogels. Critical gelation concentration (CGC) and critical gelation pH (CGpH) were measured by using the vial inversion test. CGpH were measured by using 0.75 wt% solution. Storage modulus G’ and loss modulus G” were estimated by frequency sweep of 0.75 wt% hydrogels by adding 10 mg/mL GdL at 0.9 Hz, 0.1% shear strain and 20 °C. c Amplitude sweep measurement of 0.75 wt% hydrogels obtained by adding 10 mg/mL GdL. Amplitude sweeps were performed at 1 Hz and 20 °C in the shear strain range: 0.1–100%. Data are shown as mean ± s.d. (n = 3 independent experiments). d Thixotropy measurement of 0.75 wt% FQ(Pyr) hydrogel triggered by adding 10 mg/mL GdL. The measurements were performed at 1 Hz and 20 °C with shear strain 0.1% (blue) and 100% (red).

Rheological measurements revealed marked differences in hydrogel mechanics. Frequency sweeps showed characteristic gel behavior with G’ (storage modulus) exceeding G” (loss modulus) for both systems (Fig. 2b and Supplementary Fig. S3). However, in amplitude sweeps, FQ(Pyr) hydrogels maintain elasticity up to 2% strain, whereas FQ(Ph) hydrogels failed at much lower deformations (Fig. 2c). Meanwhile, an FQ(Pyr) hydrogel prepared near the CGpH condition (pH 9) was subjected to rheological measurements, which confirmed gel-like behavior (Supplementary Fig. S4). However, the gel exhibited lower mechanical strength than that observed at pH 4, indicating that the gel strength is strongly dependent on pH. Furthermore, FQ(Pyr) gels displayed strong self-healing ability (Fig. 2b). Upon mechanical shaking, both gels became heterogeneous with liquid separation from the solid, but only FQ(Pyr) spontaneously reformed a homogeneous gel after resting for 24 hours (Supplementary Fig. S5). Thixotropy measurements under cyclic shear confirmed rapid gel–sol–gel transitions with immediate recovery of G’ upon release of high strain (Fig. 2d and Supplementary Fig. S6). Together, these results indicate that π-extension enhances not only the assembly propensity of the dipeptide but also the mechanical resilience of the resulting hydrogel network.

Nanostructure morphology and assembly behavior of FQ(Pyr)

Transmission electron microscopy (TEM) and scanning electron microscopy (SEM) imaging revealed pronounced differences in fibril morphology between FQ(Pyr) and FQ(Ph) in both gels and sols. Although both dipeptides formed fibrillar networks in their hydrogels, the morphologies were completely different (Fig. 3a, b). FQ(Ph) formed rigid, tape-like microscale fibrils with broad size heterogeneity (200–1200 nm), while FQ(Pyr) assembled into flexible nanoscale fibrils with near-uniform diameters of ~15 nm (n = 100) (Fig. 3c). These structural differences are consistent with the opaque, weak hydrogel of FQ(Ph) and the transparent hydrogel of FQ(Pyr). Despite the added bulk of the pyrene unit, FQ(Pyr) adopted a more dynamic and entangled fibrillar morphology, suggesting that π-extension favors flexibility alongside robust assembly.

Fig. 3. The dynamic self-assembly process of FQ(Ph) and FQ(Pyr).

a TEM images of hydrogel and solution in the dry state. The sol samples were directly dried on the grid, whereas the hydrogel samples were first diluted to form a dispersed solution, which was then applied and dried on the grid. In the top TEM images, samples were prepared from FQ(Ph) in pH 12 solution at 0.1 wt% (left), pH 4 solution at 0.1 wt% (middle), dilution sample of partial hydrogel at 0.5 wt% which is under CGC (0.75 wt%) (right), while in the bottom images they were prepared from FQ(Pyr) in pH 12 solution at 0.0005 wt% (left), pH 12 solution at 0.03 wt% (middle), dilution sample of hydrogel at 0.5 wt% (right). b SEM images of hydrogel of FQ(Ph) (top) and FQ(Pyr) (bottom). c Fibril diameter distribution. The diameters were measured from the TEM image of the hydrogel at 0.5 wt% (n = 100). d Schematic overview of the transformation of FQ(Pyr) morphology. e The fluorescence spectra depending on concentration of FQ(Pyr) and ionic state of dipeptide at pH 12 (left) and pH 4 (right). FQ(Pyr) were prepared over a concentration range (6.3–3000 μM (pH 12)/ (0.63–200 μM (pH 4) by using 0.05 M NaOH aq. (pH 12) or 0.5 M acetate-buffer (pH 4). Steady-state fluorescence emission spectra (λ_ex = 340 nm, λ_em = 375–550 nm (pH 12)/ λ_ex = 330 nm, λ_em = 375–550 nm (pH 4)) were recorded.

Both dipeptides also displayed different morphology transformations in solution (Fig. 3a). Spherical nanoaggregate intermediates were observed in basic conditions for both peptides. At concentration below CGC, FQ(Ph) formed nanoaggregates. In contrast, FQ(Pyr) displayed a concentration-dependent transition from nanoaggregates to fibrils (Fig. 3a, d). Under decreasing pH, FQ(Pyr) fibrils interwove into three-dimensional networks, culminating in hydrogelation (Fig. 2d). At sufficiently high concentrations, FQ(Pyr) even forms a highly viscous solution even under basic conditions (Supplementary Fig. S7), consistent with a concentration-driven increase in fibrous assemblies that eventually entangle^47,48.

Since FQ(Pyr) clearly exhibits a strong assembly propensity, we determined its critical assembly concentration (CAC) using concentration-dependent pyrene fluorescence (Fig. 3e and Supplementary Figs. S8,9)^49,50. Under basic conditions (pH 12), the polarity-sensitive I₁/I₃ intensity ratio indicated a CAC of ~320 μM. Under acidic conditions (pH 4), the emergence of a prominent excimer peak at lower concentrations revealed enhanced π–π interactions and tighter pyrene stacking. Based on the I₁/I₃ ratio, the CAC was estimated to be 9.5 μM. The pH dependence likely reflects changes in ionization state (Fig. 3d and Supplementary Fig. S10). At pH 12, FQ(Pyr) exists predominantly in an anionic state and is well-solubilized in water, whereas at lower pH it becomes zwitterionic, enabling electrostatic complementarity that facilitates pyrene π–π stacking. Therefore, the formation of the zwitterionic species appears to be a key driving force for self-assembly and gelation.

High-resolution cryo-EM of FQ(Pyr) fibril in the hydrogel

To determine the three-dimensional supramolecular architecture of the FQ(Pyr) fibrils, we employed cryo-EM, achieving an overall resolution of 1.7 Å, ranking among the highest reported for helical assemblies⁵¹ (Fig. 4a, b, c, Supplementary Fig. S11 and Supplementary Movie S2). By contrast, cryo-EM studies of peptide hydrogels have so far been constrained to resolutions approximately 3 Å,^37–44 reflecting the intrinsic heterogeneity and dynamic disorder of these systems. The near-atomic resolution attained here overcomes a long-standing limitation and establishes a benchmark for the structural analysis of supramolecular hydrogels. This exceptional resolution enabled construction of highly detailed molecular models, revealing precise positions of carbon, nitrogen, as well as distinct indications of certain hydrogen atom positions. (Supplementary Fig. S12). The structure elucidates the conformation of FQ(Pyr), the planarity of its aromatic rings, and the arrangement of coordinated water molecules. All FQ(Pyr) molecules adopt a distinctive U-shaped conformation (Fig. 4d). Quantum chemical calculations at the B3LYP/def2SVP level (SCRF/SMD, water) predict that the most stable isolated conformer is also bent due to intramolecular hydrogen bonding and CH–π interactions, but its geometry differs from that observed in the fibril, indicating that intermolecular interactions substantially reshape the conformational energy landscape (Supplementary Fig. S13).

Fig. 4. Architecture of the filament.

a Lateral view of the density map of the filament. b Longitudinal slice through the density map of the filament, showing the layered arrangement of the structure. Notice the hydrated pores. Molecules are colored by location. c Model of a single layer of the filament. Dipeptides are colored by conformation (blue for conformation 1, pink for conformation 2, and green for conformation 2’. Notice molecule 12, showing alternative locations. d FQ(Pyr), as shown in the density map, as modelled in the structure, and its chemical formula.

The cryo-EM structure reveals a twisted helical filament with a diameter of 100 Å, exhibiting a repeating pattern every 350 Å (Supplementary Fig. S14a, b). The high degree of order is evident from power spectra computed from particles of different filament regions, showing reflections visible out to 1.5 Å (Supplementary Fig. S14c). The filament’s structure is characterized by a specific helical arrangement, with a rise of 2.358 Å and a twist of 90.6°, and it follows a C2 symmetry. The filament can be thought as a 4-start helix, where two strands are C2-symmetric and offset from the other two strands by the helical rise (2.358 Å). Each of the stacked strands in the 4-start helix consists of 12 FQ(Pyr) molecules, with a local rise of 4.72 Å and twist of 1.22 degrees (Fig. 4a, b). The combination of the ~1400 Å helical pitch, C2 symmetry, and 90.6° twist between neighboring strands gives rise to the observed 350 Å repeating pattern.

Remarkably, the filament contains multiple hydrated pores: a central pore of ~15 Å diameter (Fig. 4b, c, orange) and four lateral pores of ~8 Å (Fig. 4b, c, green). While the lateral pores are fully filled with complexed water molecules, the central pore contains at least two ordered hydration layers (Fig. 4b, c, Supplementary Fig. S12), forming a highly organized water channel. The overall geometry of the filament dictates that the central pore runs straight along the filament’s axis, while the lateral pores follow the helical twist.

Each layer contains 4 rotationally symmetric repeating units, each comprising 12 FQ(Pyr) molecules (Fig. 4c). The zwitterionic C- and N-termini are either exposed on the outer surface of the filament, in contact with the solvent, or are positioned lining the internal hydrated channels, while the hydrophobic pyrene and phenyl moieties pack inwards, generating a well-defined amphiphilic architecture. The hydrophobic interior and solvent-exposed termini together confer strong water retention, which is critical for hydrogel formation. This amphiphilic arrangement likely underlies the exceptional robustness of FQ(Pyr) hydrogelator.

Peptide conformation in the protofilament assembly

Within the helical filament, each FQ(Pyr) molecule belongs to one of two distinct conformation families, as defined by the relative orientation of the pyrene and benzene rings. In each asymmetric unit, eight molecules adopt conformation 1 (Figs. 4c, 5a, blue), where the pyrene and benzene rings are nearly parallel, with an average interplanar angle of 28.4° (Fig. 5a, blue, and Supplementary Movie S3). In this arrangement, weak intramolecular CH–π interactions are observed between a benzene CH group and the pyrene π-surface, with an average distance of 3.2 Å. The remaining four molecules adopt conformation 2, in which the aromatic rings are nearly perpendicular, with an interplanar angle of ~85° (Fig. 4c, Fig. 5a, pink, Supplementary Movie S4). Molecule ⑪ adopts a related conformation, termed 2’, which differs by a 180° rotation of the pyrene ring (Fig. 4c, green, Supplementary Movie S5), but behaves functionally similarly to conformation 2.

Fig. 5. Longitudinal arrangement of filament elements.

a Arrangement and longitudinal interaction of molecules with conformation 1 (left, blue) and 2 (pink, right) in protofilament assembly. The stacks are stabilized by extensive π–π interactions, as well as backbone and water-mediated hydrogen bonds, here represented as dashed lines. The distances between stacked aromatics refers to the intraplanar distance. b Pairwise interactions between protofilaments, with the respective angles annotated. Angles are defined between the segments defined by the carboxy α carbon, and the glutamate N, projected on the layer plane.

All molecules stack vertically to form protofilaments, yielding two distinct protofilament types, corresponding to the two conformational classes. Each helical layer contains four asymmetric units, each with 12 molecules; the full helical assembly is thus constructed from 48 protofilaments (4 × 12), each generated by the vertical stacking of molecules of the same conformations while preserving the overall helical twist.

Intermolecular interactions in protofilament assembly

Three major classes of intermolecular interactions stabilize vertical stacking within protofilaments (Fig. 5a). First, strong π–π stacking interactions occur between both pyrene-pyrene and benzene-benzene pairs. These interactions adopt parallel-displaced geometries with an average interplanar distance of ~3.5 Å (Fig. 5a). Remarkably, these π–π contacts persist across all elementary fibrils, suggesting a reinforcement by exceptionally strong and continuous π–π stacking interactions.

All conformers exhibit a pronounced segregation of the polar functionalities. Both amide bonds rotate such that all amide nitrogen atoms lie on one face of the molecule, and all carbonyl oxygen atoms on the opposite face (Supplementary Fig. S15a,b). This consistent orientation creates very distinct electrostatic surfaces and enables two amide–oxygen hydrogen bonds between molecules in adjacent layers, one pair on the peptide bond and the other one between glutamine’s amide nitrogen and oxygen atoms (Fig. 5a). Importantly, this alignment generates a unidirectional dipole across every protofilament, which is preserved along the entire fibril, implying the presence of a macroscopically ordered electrostatic field (Supplementary Fig. S15c,d).

Lastly, the exposed C- and N-termini participate in an extended hydrogen bonding network mediated by highly ordered water molecules, positioned between, and interacting with neighboring layers. The water molecules exhibited interactions with the N-terminal and the C-terminal polar functional group of the closest dipeptide, while also engaging with the N-terminal region of a more distal dipeptide. These water molecules bridge multiple layers simultaneously, increasing intralayer and interlayer cohesion. It appears that the zwitterionic state of the termini is vital in strengthening the water-mediated intralayer interactions, thus enabling the formation of the full filament, in a pH-dependent manner. Ordered water molecules likely play an important role by mediating reversible hydrogen-bonding networks which enables rapid reorganization of supramolecular interactions after shear-induced disruption, thereby contributing to efficient self-healing.

Interactions between protofilaments

Protofilaments engage in defined lateral interactions that produce paired configurations (Fig. 5b). Neighboring layers are intercalated, where each layer is longitudinally displaced by the helical rise (half the helical rise of each strand in the 4-start helix; 2.358 Å), presumably to optimize stacking (Fig. 5b). The geometry of each pair depends on both conformational type and the relative orientation (parallel or antiparallel) of adjacent molecules. With two conformational families, four pairing modes arise: parallel conformer 1—conformer 1 (e.g. ⑦-⑧) parallel conformer 1– conformer 2 (e.g. ①-②), antiparallel conformer 1—conformer 1 (④-⑨), and antiparallel conformer 1 – conformer 2 (⑥-⑦) (Fig. 5b).

The parallel 1-1 pair exhibits a slight outward tilt (~10°) bringing pyrene rings into proximity, whereas the parallel 1-2 pair displays a modest inward tilt (~20°), directing aromatic units towards the polar termini. Optimal packing occurs in the antiparallel 1-1 pair, which achieves almost precise alignment, in contrast to the more kinked antiparallel 1-2 pair (Fig. 5b). These subtle variations fine-tune the amphiphilic arrangement of each layer and, consequently, the global architecture of the helix. Notably, parallel 1-1 pairs occupy the fibril periphery, while parallel 1-2 pairs line the pore surfaces.

Hierarchical assembly pathway

Inspection of the single-layer assembly revealed four sequential geometric rules that collectively generate the hierarchical supramolecular structure (Fig. 6a). Starting from a single FQ(Pyr) molecule, lateral placement of an antiparallel partner followed by a parallel partner (rules (i) and (ii)) produces the first-order hierarchical assembly, a down-up-up trimer (Fig. 6b). Applying rule (iii), a perpendicular molecule associates with the parallel pair, initiating a second round of trimer formation. Four consecutive applications of rules (i)–(iii) assemble the second-order structure: a cluster of molecules forming a small pore (Fig. 6b).

Fig. 6. Hierarchical assembly of FQ(Pyr) filaments.

a Each layer of the filament comprises repetitive elements that conform to four consecutive geometric rules. The empty boxes indicate the space where the next molecule will be added. b Applying rules (i-iv) (a) in sequence, enables assembly of the layer. c Schematic representation of the FQ(Pyr) filament.

The pyrene groups that are arranged in parallel following the assembly of the down-up-up trimer act as a long apolar interaction surface, long enough to accommodate the next FQ(Pyr) molecule, which is placed perpendicularly to the rest (Fig. 6a, rule (iii)). Further strengthening this interaction, the polar end of the down molecule of the trimer interacts with the polar part of the incoming perpendicular FQ(Pyr). This molecule then acts as the initiation site for the assembly of another trimer or was already part of one. This process repeats 4 times, causing the assembly to circularize around an ~8 Å diameter pore (Fig. 6b).

In a final assembly step, the units interact laterally through an exposed hydrophobic surface. This surface consists of the two pyrene rings near the corners of each tetramer of trimers (Fig. 6a, rule (iv)). Again, four of these units are required to circularize the structure, this time around the larger, ~15 Å pore (Fig. 6b). The filament is expected to maximize packing efficiency while minimizing lateral exposure of hydrophobic surfaces to the polar solvent. This is likely achieved through the distinct packing angles adopted by each conformer, resulting in the final structure observed by cryo-EM (Fig. 6b).

Iteration of this process yields the complete layer architecture, characterized by four small pores surrounding a larger central pore (Fig. 6b). We infer that layer formation proceeds simultaneously with longitudinal protofilament growth, driven by extensive π-stacking along the filament axis (Fig. 6c). However, lateral interactions introduce angular constraints that would otherwise generate structural strain. Molecules in conformer 2 locally relieve this strain, facilitating encircling of the pore or enabling optimized packing around the pore (Fig. 6b). Extending the structure indefinitely beyond four asymmetric units would introduce steric clashes and unfavorable interactions, favoring formation of finite helical filament rather than crystalline lattices (Fig. 6c).

Discussion

Our findings demonstrate that integrating a π-augmented aromatic unit into an uncapped dipeptide enables a level of supramolecular order that bridges molecular precision and macroscopic material function. The FQ(Pyr) system assembles into uniform ~100 Å helical nanofibers whose near-atomic cryo-EM structure reveals tightly packed protofilaments stabilized by reinforced π-stacking, polar segregation, and highly organized water channels. These interactions give rise not only to an unusual combination of mechanical rigidity and self-healing behavior in the bulk hydrogel, but also to a unidirectional dipole that is preserved across the entire fiber; an emergent property rarely observed in peptide-based materials. Furthermore, the presence of a unidirectional dipole may give rise to functionalities, such as piezoelectric behavior, field-responsive alignment behavior and directed ion transport through the water channels. The hierarchical assembly rules extracted from the structure provide a blueprint for how uncapped peptides and π-augmented aromatics can cooperate to create hybrid architectures inaccessible to either motif alone.

More broadly, this work defines a class of supramolecular architectures at the materials–biology interface, in which synthetic π-systems are integrated with peptide-based building blocks at near-crystalline precision. By uniting the organizational principles and rigidity of conjugated aromatic materials with the aqueous affinity, polarity, and adaptability of biomolecular scaffolds, the resulting nanofibers bridge molecular-scale order and macroscopic function in a manner unattainable by either domain independently. Beyond the specific system reported here, these findings establish a general design strategy for creating programmable supramolecular materials that couple electronic, structural, and biological functionalities.

Beyond the materials advances, this study also represents a major conceptual and technical breakthrough. Although cryo-EM has transformed protein structural biology, its application to synthetic supramolecular materials has been constrained by intrinsic heterogeneity and dynamic disorder. Our results show that appropriately designed peptide–aromatic hybrids can overcome these limitations, enabling cryo-EM to function as a general, high-resolution platform for resolving synthetic supramolecular architectures. By closing the gap between molecular design and structural determination, this work positions cryo-EM not merely as an analytical tool, but as a driver of discovery in supramolecular materials chemistry, opening routes toward rationally engineered soft matter with structural and functional fidelity.

Methods

General synthesis of dipeptides

FQ(Ph) and FQ(Pyr) were synthesized by sequential amide coupling of Fmoc-L-Glu(O^tBu)-OH with the corresponding aromatic amine, Fmoc deprotection, coupling with Boc-L-Phe-OH, and final acid-mediated deprotection to afford the target compounds as TFA salts. The products were purified by flash chromatography and characterized by NMR spectroscopy and high-resolution mass spectrometry. Detailed synthetic procedures and analytical data are provided in the Supplementary Information.

General preparation of hydrogel

The dipeptide [FQ(Ph) or FQ(Pyr)] was dissolved in 0.05 M NaOH aq. (pH 12.3) with strong vortexing and sonication at 40 °C for 20 min. The final pH of the dipeptide solution ranged from 11 to 12, depending on the concentration and molecular structure of the compound. If the pH were to decrease below this level, solubility would become limited; therefore, 1 M aqueous NaOH was added as needed to adjust the final pH to 11–12. Regardless of the pH at this stage, glucono-δ-lactone (GdL, 10 mg/mL) was added to the basic peptide solution and mixed gently by hand. After this, the sample was allowed to stand, with gelation taking place overnight at room temperature.

For example, to prepare a 0.75 wt% hydrogel, 100 μL of 0.05 M NaOH aq. was used per 0.75 mg of dipeptide. After complete dissolution, GdL was added at a ratio of 1 mg per 100 μL of peptide solution (from a 10 mg/mL GdL stock solution), followed by gentle mixing and incubation at room temperature overnight.

Measurement of hydrogel

pH measurements

pH measurements were performed on a Mettler Toledo™ SevenCompact pH meter S220 (Mettler Toledo, Columbus, USA) equipped with a Mettler Toledo™ InLab® Ultra-Micro-ISM® electrode (Mettler Toledo, Columbus, USA). The pH meter was calibrated using pH 2.00, 4.01, 7.00, 9.21 and 11.00 buffer solutions and then rinsed with deionized water. The probe was inserted into the solution and pH readings were recorded.

Critical gelation concentration (CGC) measurements

The CGC was determined by a vial inversion test. The dipeptide solutions were prepared at 1.0 wt%, 0.75 wt%, 0.5 wt%, 0.25 wt%, 0.1 wt%. 0.05 wt%. The hydrogelation occurred by using GdL10 mg/mL, which was assessed by inverting the vial and assessing whether the material regained its self-supporting structure.

Critical gelation pH (CGpH) measurements

The CGpH was determined by a vial inversion test on hydrogels prepared using varying amounts of GdL. The dipeptide solutions were prepared at 0.75 wt%. The CGpH was defined as the pH at the boundary between a hydrogel and a solution or viscous solution. After resting overnight, the prepared samples were performed the inversion test and measuring the pH.

Critical aggregation concentration (CAC) determination using pyrene fluorescence measurement

A The CAC of dipeptide FQ(Pyr) was determined via the pyrene fluorescence assay. Samples were measured with an Edinburgh FLS920 (Edinburgh Instruments Ltd., UK) fluorescence spectrometer equipped with a 450 W Xe arc lamp was used as the excitation source, a single grating monochromator in both the excitation and the emission sides and coupled to a PMT-980 HAMAMATSU photomultiplier as detector at room temperature. The emission spectra were corrected for detection and optical spectral response of the fluorescence spectrometer through a calibration curve supplied by the manufacturer.

The samples were analyzed using high precision 1.4 mL SUPRASIL quartz glass cuvettes with 10 mm path length. Data processing was realized using OriginPro 2025 (64-bit) SR1 10.2.0.196 (ORIGINLAB Corperation, Northampton, USA).

FQ(Pyr) were prepared over a concentration range (6.3–3000 μM (pH 12)/ (0.63–200 μM (pH 4)) by using 0.05 M NaOH aq. (pH 12.32) or 0.5 M acetate-buffer (pH 4.45). Steady-state fluorescence emission spectra (λ_ex = 340 nm, λ_em = 375–550 nm (pH 12)/ λ_ex = 330 nm, λ_em = 375–550 nm (pH 4)) were recorded. The ratio of the first and third vibronic peaks (I₁/I₃ at ~383 and 402 nm) was plotted against log[FQ(Pyr)], and the CAC under pH 12 was defined at the inflection point of the resulting curve. On the other hand, the CAC under pH 4 was also defined by plotting the ratio of the first vibronic peak and excimer pear (I₁/I₃ at ~383 and 461 nm).

Self-healing test

A Self-healing ability was qualitatively evaluated by a vial inversion test. After 0.75 wt% hydrogel formation by using GdL10 mg/mL, the sample was subjected to vortex mixing to fully disrupt the hydrogel. The resulting sol-like dispersion was then allowed to stand undisturbed at room temperature for 1 day. The recovery of the gel state was assessed by inverting the vial to check whether the material regained its self-supporting structure.

Rheology

Rheological measurements were carried out on an Anton Paar Modular Compact Rheometer MCR 102 (Anton Paar GmbH, Graz, Austria) with Anton Paar RhepCompass V1.20.40.496 (Anton Paar GmbH, Graz, Austria) analysis software. For all rheological measurements the rheometer was equipped with a P PTD200 + H PTD200 photo cell and a PP25 spindle (24.94 mm cone plate) with the measurement gap set to 0.300 mm. Tests were performed at 20 °C. Data processing was conducted using OriginPro 2025 (64-bit) SR1 10.2.0.196 (ORIGINLAB Corporation, Northampton, USA).

Amplitude sweeps were carried out at fixed frequency f = 1 Hz while changing the sheer strain from 0.1% to 100%. And frequency sweep measurements were carried out at strain 0.1% while changing the frequency from 0.1 Hz to 100 Hz. Thixotropic behavior was assessed using a time-dependent shear protocol. Rheological measurements were performed at a fixed frequency of 1 Hz. Prior to initiating the shear cycles, a low shear strain of 0.1% was applied for 150 s to equilibrate the sample. Each cycle consisted of applying a high shear strain of 100% for 50 s, followed by a low shear rate of 0.1% for 300 s. This sequence was repeated for three consecutive cycles without interruption. Amplitude sweep, frequency sweep and thixotropic measurement were conducted independently 3 times each on three samples.

All samples were prepared as 0.75 wt% hydrogels triggered by GdL, following the procedure described in the “Preparation of hydrogel” section. However, for measurements of the pH 9 gel, HCl was used as a pH-adjusting agent to prepare the hydrogel, and a concentration of 0.5 wt% was employed.

Viscosity measurements were carried out changing the sheer rate from 1 Hz to 7 Hz.

Room-temperature transmission electron microscopy (TEM)

TEM measurements were performed using a Thermo Fisher Scientific FEI Titan Themis G3 60-300 transmission electron microscope (Thermo Fisher, Waltham, Massachusetts, USA) equipped with a high-brightness field emission gun (X-FEG), a monochromator, a quadrupole energy dispersive X-ray system, a high-angle annular dark field detector (Fishione Model 3000), an image Cs corrector, and a fast CMOS camera (Ceta 2 speed upgrade 4k x 4k). The microscope was operated at an acceleration voltage of 300 kV. For sample preparation, aqueous solutions were drop-casted (1 μL) onto formvar/carbon coated copper grids (Plano EM, S162 or Micro to Nano, 22-1MFC20-50) after cleaning in an oxygen-plasma oven (Diener electronics ZEPTO Plasma-Surface-Technology Plasma Oven) (only using Plano EM, S162) and drying overnight under ambient conditions. The solutions of dipeptide (pH 12) were directly used as sample, whereas the hydrogel/partial hydrogel samples were diluted with water (water 100 μL + A small amount of gel scooped using a spatula) to form a dispersed solution, which was then applied on the grid. Images were analyzed using TIA version 5.2 (FEI) and ImageJ version 1.54 g (National Institutes of Health, USA, Java 1.8.0_345).

Scanning electron microscopy (SEM)

SEM measurements were performed on a Zeiss CrossBeam 340 (Zeiss, Oberkochen, Germany) equipped with an In-lens SE detector. The acceleration voltage was set to 1 or 2 kV, an aperture size of 30 µm was chosen and the secondary electrons were detected at a working distance of 2 mm. SEM samples were obtained by spreading the hydrogel or diluted solution of the hydrogel with water on silicon wafers and drying overnight under ambient conditions. Images were analyzed using ImageJ version 1.54 g (National Institutes of Health, USA, Java 1.8.0_345).

Cryogenic transmission electron microscopy (cryo-EM) sample preparation and data collection

4 μL of 5 mg/mL FQ(Pyr), prepared as described, were plunged on glow discharged C-flat CF-2/2-4Au (400 mesh) grids in a Vitrobot Mark IV (force 10, 5.5 s), in 14 °C and 95% humidity, in liquid ethane. The grids were then clipped and inserted into a Krios G4 electron cryo-microscope operating at 300 kV, equipped with an e-CFEG, and a SelectrisX energy filter set at a 10 eV slit. Images were collected using EPU 3.11.0 at 270,000x magnification (pixel size 0.454 Å) with a total dose of 50 e^-/Ų in faster aquisition mode (using image-beam shift with AFIS (aberration free image shift)) with a defocus range between −0.5 µm and 1.6 µm on a Falcon 4i detector, in EER format.

Cryo-EM image processing

All data processing was completed in CryoSPARC ver. 4.6.2, unless otherwise stated. All parameters were set to their default values, unless otherwise stated. We collected 13,139 multi-frame EER movies, which were each fractionated into 40 frames and subjected to Motion Correction and CTF estimation using CryoSPARC’s Patch Motion Correction and Patch CTF jobs, using default settings. Of those, 7204 were selected for further analysis, based on CTF fit resolution (1.5–12 Å), estimated mean defocus (3000–16,000 Å), estimated astigmatism (0–1000 Å), and total cumulative motion (0–50 pixels). Using filament tracing (filament_diameter = 120, min_distance = 0.118, min_filament_length = 1.5, lowpass_res = 12, branch_radius_remove = 0.5, and max_prune_dist = 0.2), we picked (−169 <Local power <2552, Curvature <0.0046, and Sinuosity <1.47895) and extracted 813,485 unbinned particles (box size: 416 pixels), which were subjected to two rounds of 2D classification (100 classes, and then 50 classes and class2D_max_res = 3). The classes giving well-aligning 2D class averages were selected, and the resulting 358,791 particles were used for ab-initio reconstruction, and an initial, symmetry-less initial reconstruction. The helical parameters were initially estimated using the Symmetry search utility and trial and error, and later optimized by iterative refinements. Finally, the resulting alignments and corresponding 2.12 Å reconstruction were used for Global and Local CTF refinement, and subsequently for Reference Based Motion Correction, resulting in a 1.75 Å reconstruction. A round of 3D classification, followed by re-optimization of the per-particle defocus resulted in our final 1.65 Å reconstruction, from 263,876 particles (Settings for final refinement: refine_init_twist = 90.62, refine_init_shift = 2.36, refine_sym_enforce_r = 3, refine_pg_symmetry = C2, refine_sym_search = 3, refine_sym_twist_min = 90.5, refine_sym_twist_max = 90.7, refine_sym_shift_min = 2.3, refine_sym_shift_max = 2.4, refine_locproc_start_res = 2, refine_res_init = 6, refine_num_final_iterations = 4, refine_res_gsfsc_split = 3.5, refine_gs_resplit = true, refine_gs_split_by_mic = true, refine_scale_min = true, refine_dynamic_mask_start_res = 4). The helical parameters finally converged to the following values: Symmetry: C2, helical twist: 90.609 degrees, helical rise: 2.358 Angstroms, helical symmetry order: 6 (Supplementary Fig. S10). Averaged power spectra were calculated in CryoSPARC ver. 5.0.4 using the particles of 2 representative 2D classes (Supplementary Fig. S14).

Computational methods

Initial molecular geometries of the PAH-dipeptide were generated in Avogadro using the MMFF94 force field, followed by brief MMFF94-based pre-relaxation. Conformational sampling was performed with the CREST program employing the GFN2-xTB Hamiltonian, which provides broad exploration of rotamers for flexible systems. The resulting conformer ensemble was filtered to structures within 10 kcal•mol^–1 of the CREST global minimum. From this filtered ensemble, 47 representative conformers were selected for DFT geometry optimization. Among them, 44 structures converged normally and were used for the free-energy ranking and subsequent discussion. DFT geometry optimization and frequency calculations were carried out using B3LYP-D3(BJ)/def2-SVP with the SMD solvation model for water. The optimized Cartesian coordinates and corresponding relative Gibbs free energies of the 44 normally terminated structures are provided as Supplementary Data 1 and 2, respectively. The CREST-generated rotamer ensemble within 10 kcal·mol^–1 of the CREST global minimum is provided as Supplementary Data 3.

Author contributions

A.U., B.J.R. and K.I. conceived the concept of π-augmented peptide assembly. K.I. and B.J.R. directed the project. C.G managed the cryo-EM analysis. A.U. performed all the synthesis and the measurement of hydrogel properties. L.S. and D.M. conducted TEM and SEM experiments with A.U. A.U., D.E., B.J.R., A.Y., K.A. and K.I. performed and analyzed synthesis, structure determination, and photophysical property experiments. H.K. conducted calculation of monomer structure. A.N. performed and analyzed cryo-EM measurement. G.B. conducted modeling based on cryo-EM data. G.B. and C.G. analyzed the structure of assembly based on the modeling. All authors analyzed and checked the data. A.U., G.B., B.J.R., C.G. and K.I. wrote the manuscript with feedback from other authors. All authors have given approval to the final version of the manuscript.

Peer review

Peer review information

Nature Communications thanks Edward Egelman and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Funding

This work was supported by JSPS KAKENHI (grant no 25H00429 to K.I.), JST-ASPIRE (grant no JPMJAP2508 to K.I., B.J.R. and C.G.), RIKEN Pioneering Project (to K.I.) and JSPS Promotion of Joint International Research (grant no 22K21346 to A.U. and A.Y.), B.J.R. acknowledges funding by the Deutsche Forschungsgemeinschaft (IRTG 2678 Project ID 437785492). The cryo-EM data were collected at Cryo-EM SoN, the cryo-EM infrastructure of the University of Münster, funded by the DFG (project ID 496113311).

Data availability

Materials and methods, hydrogel property measurements, TEM/SEM measurements, cryo-EM measurements are in Main text, and synthetical procedures and NMR spectra are available in the Supplementary Information. The cryo-EM map has been deposited in the Electron Microscopy Data Bank (EMDB) under the accession code EMD-58514. The optimized Cartesian coordinates of the 44 normally terminated DFT-optimized PAH–dipeptide conformers are provided as Supplementary Data 1. The corresponding Gibbs free energies, relative Gibbs free energies, and Boltzmann populations are provided as Supplementary Data 2. The CREST-generated rotamer ensemble within 10 kcal·mol^–1 of the CREST global minimum is provided as Supplementary Data 3. Additional raw data are available from the corresponding authors upon request.

Competing interests

The synthesis and bio-applications of π-extended hydrogel partly described in this manuscript have been filed by RIKEN and University of Münster (Ayaka Ueda, Kazuma Amaike, Hideya Kono, Daisuke Shimizu, Bart Jan Ravoo, and Kenichiro Itami: Japanese Patent Application No. 2026-086868). These authors declare no other competing interests. The remaining authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-75984-9.