Dynamic self-shrinking peptide hydrogels with shape memory and self-healing properties

Abstract

Three peptide amphiphiles based on aromatic amino acids have been successfully synthesized, purified, and thoroughly characterized. These peptides were found to form hydrogels at physiological pH 7.46, exhibiting a unique time-dependent self-shrinking behaviour. Notably, the extent and nature of this self-shrinking varied according to the specific amino acid sequence of each peptide. In addition to this, the hydrogels demonstrated remarkable self-healing abilities and 3D shape-memory properties. To investigate the role of sequence variation, particularly the position of the L-phenylalanine residue, coarse-grained molecular dynamics simulations were employed. These simulations aimed to elucidate how different sequences influence self-assembly into the nanoscale network structures characteristic of the hydrogels. Importantly, the computational findings showed strong agreement with the experimental results, confirming the formation of distinct hydrogel architectures driven by the peptide sequences.

---

Introduction

Low-molecular weight peptide gelators (LMWPGs) are small molecules that self-assemble into three-dimensional networks to form hydrogels with attractive mechanical and functional properties.^1–15 Composed of biocompatible, biodegradable amino acids,¹⁶ these peptide gels are typically reversible and are stabilized by noncovalent interactions, including π–π stacking, hydrogen bonding, van der Waals and ionic interactions.^17–20 Beyond static assemblies, LMWPGs can form dynamic self-shrinking hydrogels,²¹ in which ordered structures are formed and reorganized in response to environmental stimuli through reversible noncovalent bonding, where dynamic self-shrinking refers to the spontaneous organization of molecules or components into ordered structures.²² By harnessing the ability of materials to dynamically respond to external stimuli, scientists and engineers are continually pushing the boundaries of material design, leading to the creation of adaptive and functional materials with diverse practical applications.^2–6,23,24

Dynamic self-assembly occurs across scales, from supramolecular systems to biology (e.g., protein folding and membrane formation).^24–26 A good example of an in vivo instance of syneresis includes serum separating from the fibrin clot during blood coagulation. The concept underpins advances in materials science,^5,27 biotechnology,^3,6,28–30 nanotechnology,^23,31,32 drug delivery^6,33,34 and tissue engineering.^35–37

In most of the reported cases, sometimes supramolecular gels are stimuli-responsive, which induces sol-to-gel or gel-to-sol conversion using external stimuli like pH, temperature, light, or ions.^38–41 Sometimes, they show responsiveness towards external stimuli, but this time, they can also experience syneresis after using some external stimuli such as heat,⁴² pH,^42–44 pressure,⁴⁵ metal ions,^20,21,46 chemical,^47,48 and in the presence of UV light.⁴⁹ However, this is not always true; sometimes, the hydrogel can show shrinking behaviour without any external stimuli. There is only one report of such type of fascinating peptide material in the literature where a tripeptide-based gelator undergoes self-shrinking without the assistance of any external stimuli, which has been reported by our group.²¹ Such shrinking nature of supramolecular hydrogels makes them innovative materials for many naturalistic applications. In supramolecular chemistry, synthetic molecules also have the same kind of tendency to undergo syneresis. This process mainly happens when a change in the hydrophobic microenvironment takes place during the gel formation.⁵⁰ Sometimes, supramolecular gelators can experience shrinking due to the strong interaction between the hydrophobic gelator molecules. This shrinking or swelling nature of hydrogels is rare in the literature as inherent non-covalent interactions drive the self-assembly. There are reports that amphiphilic peptides can also form supramolecular aggregates through noncovalent interactions like supramolecular hydrogels. These interactions drive the self-assembly into hierarchical nanostructures such as fibers, vesicles, and sheets. Such assemblies exhibit tunable morphology, dynamic responsiveness, and biocompatibility, making them promising for applications in drug delivery, tissue engineering, and others.^51–53

In this work, we have developed three different amphiphilic tripeptides, G1, G2, and G3, by altering the position of the amino acid residues, one L-phenylalanine (F), and two L-tryptophan (W), and keeping the long chain fatty acyl derivative, myristyl (C14) at the N-terminal position. All three gelators can undergo spontaneous syneresis. However, the first one G1 shows a superfast shrinking ability due to strong π–π stacking and van der Waals interaction compared to the other two gelators, G2 and G3, as observed in our experimental and theoretical investigations. This is due to the formation of different types of microenvironments during the gelation time, which has been confirmed by morphological studies. To the best of our knowledge, this is the first demonstration that the order of positions of amino acid residues in the peptide chain alone can tune both the self-shrinking behaviour and hydrogel morphology in peptide gels. The self-shrunken gels are also self-healing, yielding a smart, reconfigurable soft material.

Experimental

Materials

Myristic acid (C₁₄-OH), L-tryptophan (L-trp), and L-phenylalanine (L-phe) were purchased from Sigma Aldrich. HOBt (1-hydroxybenzotriazole) and DCC (N,N′-dicyclohexylcarbodiimide) were purchased from SRL, India. The reagents hydrochloric acid (HCl), sodium hydroxide (NaOH) and sodium chloride (NaCl) were used during the separation of the crude product by using a separation funnel. Two NMR solvents CDCl₃ and DMSO-d₆ were purchased from Sigma Aldrich, USA and SRL India, respectively. Sodium dihydrogen phosphate and disodium hydrogen phosphate were purchased from Merck. Millipore Milli-Q grade water was used for all experiments.

Methods

Synthesis and characterisation. The peptides G1, G2 and G3 were synthesized using a conventional solution phase DCC/HOBt-mediated coupling method by a racemization-free fragment condensation strategy. All compounds were purified by column chromatography using silica gel (100–200 mesh size) as the stationary phase by using chloroform-ethyl acetate or chloroform-methanol as eluents. Finally, the purified compounds were characterized using ¹H NMR, ¹³C NMR and mass spectrometry. Detailed information about the synthesis and characterisation is given in the SI (Fig. S1–S48).

Preparation of the hydrogels. 2 mg gelator peptide (G1/G2/G3) has been weighed into a 5 mL screw-capped glass vial with the addition of 1 mL of 7.46 phosphate buffer solution. Then the glass vial was heated with a heat gun until the gelator molecules were dissolved completely. After that, the heated glass vials were cooled in a water bath for 1 min and immediate stable gel formation was observed, even before the solution reached the normal room temperature of 28 °C completely.

FTIR experimental details. Hydrogels and self-shrinking (syneresis) hydrogels were flash-frozen in liquid nitrogen and the water was removed by lyophilization to obtain the xerogels for ATR mode FTIR analysis. For the time-dependent FTIR study of G1, gelator solutions at 2 mg mL⁻¹ were prepared in four screw-cap vials. After gelation, one sample was frozen/lyophilized immediately (0 min), and the remaining vials were processed at 15 min, 30 min, and 90 min to capture syneresis-driven structural changes. The same protocol was applied for G2 and G3, enabling the comparison of the FT-IR spectra for both the initial hydrogels and their self-shrunk counterparts and monitoring of the interaction change during the syneresis process.

DLS experimental details. Dynamic light scattering (DLS) was performed on dispersions prepared by diluting the 2 mg mL⁻¹ hydrogel (or self-shrinking hydrogel). 10 µL of gel was added to 990 µL of Milli-Q water (final concentration was 0.02 mg mL⁻¹), gently inverted, briefly bath-sonicated for 1 min, and shaken for 10 min to disperse the self-assembled structures. The dispersion was transferred to a clean cuvette, equilibrated at 25 °C for 5 min, and the data were measured; only Milli-Q water served as a control without the gel.

Atomistic (AA) simulation details

To bridge the gap between experimental observations and computational insights, we have used large-scale coarse-grained (CG) MD simulations to reduce the number of degrees of freedom and enable the simulation of larger assemblies over microsecond timescales. The CG model was developed using fully atomistic simulation using protocols as described below. The atomistic structures of the three self-shrinking tripeptide gelators G1 (C14WFWOH), G2 (C14WWFOH), and G3 (C14FWWOH) were built using Avogadro software.⁵⁴ The initial box size for the all-atom (AA) MD simulation was 5 × 5 × 5 nm³ containing a single tripeptide, solvated in TIP3P water. The solvated system was energy minimized using the steepest descent technique to remove any bad contacts. This was followed by a 20 ns equilibration run in the NVT ensemble and 15 ns NPT run at 300 K and 1 atm pressure. Finally, a 50 ns production run using the NPT ensemble was performed. An integration time step of 2 fs, with the CHARMM36m^55,56 force field, was used for all the simulations using GROMACS 2023 software.⁵⁷ PME was used to compute long-range electrostatics with a real space cutoff radius of 1.2 nm. The same cutoff was also used for the Lennard-Jones interactions, and the bonds involving hydrogen atoms were constrained with the LINCS algorithm. The all-atom MD trajectories of the three tripeptide gelators were coarse-grained using the PyCGTOOL⁵⁸ package in combination with MARTINI v2.2 topology mapping.^59–61 For visualization and snapshot generation of both atomistic and coarse-grained MD (molecular dynamics) simulation trajectories, VMD tools were utilized.⁶²

Coarse-grained (CG) MD simulation details

As mentioned in the last paragraph, CG coordinates and initial bond topology parameters for the MARTINI v2.2^60,61 model were generated for the three tripeptide gelators (G1, G2, and G3) using atomistic MD trajectories, where the mapping followed the MARTINI v2.2 AA-to-CG scheme implemented through the PyCGTOOL package, as shown in Fig. 1a and b.

Fig. 1 (a) Atomistic (AA) representation of the G1 (C14WFWOH) tripeptide; the atoms are coloured as N: blue, O: red, C: cyan, and H: white. (b) Coarse-grained (CG) representation from the atomistic bead mapping of the G1 tripeptide, using PyCGTOOL and MARTINI v2.2 FFs.

This Martini v2.2 CG force field uses four-to-one mapping where four atoms and associated hydrogens are represented by one CG bead and three-to-one mapping for aromatic substituents. Accordingly, the aromatic Phe (F) molecule is represented by three side chain beads for the side ring structure and one bead for the backbone main chain, while the Trp (W) side chain and main chain are represented by the five CG beads (Fig. S49). To investigate the self-shrinking behaviour, we have prepared three systems composed of 1200 tripeptide gelators (G1, G2, and G3) and coarse-grained water molecules. In the CG model, four atomistic water molecules are represented by a single charge-neutral CG water bead. The initial configurations consisted of 1200 randomly placed tripeptides within a 22 × 22 × 22 nm³ cubic box, corresponding to a peptide concentration of 0.2 mg mL⁻¹. We have also added CG chloride and CG sodium ions in the requisite amount to electrically neutralize the system and attain the salt concentration of 0.4 mg ml⁻¹ along with the CG sodium (Na⁺) and negatively charged CG dihydrogen phosphate (H₂PO₄⁻) and hydrogen phosphate (HPO₄²⁻) anions to mimic experimental conditions. All the CG-MD simulations have been carried out with GROMACS 2023 software. The systems were coupled to external temperature and pressure baths of 300 K and 1 atm, respectively, using the Berendsen coupling methods. Both electrostatic and van der Waals interactions were handled using CG shifted potentials with a cutoff distance of 1.2 nm. All the initial systems were energy minimized using the steepest descent algorithm; then the systems were heated to 300 K. The system was then equilibrated at 300 K for 15 ns and finally production simulations for 4 µs at 300 K with an integration time step of 10 fs.

Various structural, energetic, and dynamic properties were carried out using in-house codes and the gmx analysis tool of GROMACS. 4 µs long CG-MD trajectories were used to calculate the following quantities: solvent accessible surface area (SASA), number of clusters, radius of gyration, radial distribution function (RDF), stacking, and non-bonded interaction energy of the tri-peptide main and side chains to monitor the self-assembly process at different time intervals.

Results and discussion

Gelation study and dynamic self-shrinking

In this study, three peptides with the same molecular formula and identical amino acid composition, but differing in amino acid sequence, keeping the myristic acid-like amphiphilic molecule in the N-terminal side (Fig. 2a), were used. All these synthesized compounds are C14WFWOH (G1), C14WWFOH (G2), and C14FWWOH (G3). These tripeptide gelators form hydrogels in the 7.46 phosphate buffer solution. The minimum gelator concentration (MGC) was 0.4–0.5 mg mL⁻¹ (0.04–0.05 wt%), indicating that our synthesized gelators are super gelators in nature at 7.46 PBS medium. However, we have studied the pH dependent gelation and it was found that all of these gelators are capable of forming self-shrinking hydrogels within a pH range of 6.5 to 13.8, indicating their robust self-assembly behaviour under moderately acidic to basic environments. These gelators are capable of forming shrinking hydrogels in the phosphate buffer having pH 6.5–8.5 and in the presence of a sodium hydroxide base within the pH range 8.0–13.8. Among the three gelators, G1 shows very fast self-shrinking, and this happens within 3 h to complete the shrinking process by releasing 90% of water. On the other hand, G2 and G3 take 7 days to complete self-syneresis with 80–90% release of expelled water (Fig. 2b). As the hydrogel starts self-shrinking slowly, its transparency decreases due to the increase of the opaque nature (Fig. 2b). The initial observation suggests that this difference in shrinking rate is due to the difference in amino acid position in the peptide sequence. Moreover, the water release kinetics for G1 to G3 gelators follows first-order release kinetics (Fig. 2c and Fig. S50). This fast water release kinetics indicates that there may be some different morphological transformation mechanism for which the G1 tripeptide undergoes very fast self-shrinking over the other two G2 and G3 gelators. So, to get more insight into it, some experimental and theoretical studies were performed.

Fig. 2 (a) Structure of the peptide amphiphiles (G1/G2/G3) that form the hydrogels at PBS of pH 7.46, which undergoes self-shrinking with time. Peptide gelator, G1 induced 2 mL PBS hydrogel's (b) time-dependent water-releasing pictures and (c) water-releasing kinetics.

FT-IR analysis

The Fourier transform infrared (FT-IR) analysis revealed strong peaks in the range of 1630–1680 cm⁻¹, indicating the presence of amide carbonyl groups, and broad peaks between 3400 and 3500 cm⁻¹, which correspond to the acid hydroxyl groups in both the hydrogel and self-shrinking states of all the gelators. Additionally, peaks were observed between 1500 and 1600 cm⁻¹ and 3250 and 3500 cm⁻¹, corresponding to the bending and stretching frequencies of the N–H bond, respectively (Fig. S51a and b). Time-dependent FTIR studies were conducted on the hydrogel after 0 min, 15 min, 30 min, and 90 min of self-shrinking for the dried hydrogel obtained from gelator G1 to comprehend the shrinking mechanism. It was noted that the bending frequency at 1545 cm⁻¹ shifted to 1547 cm⁻¹, and the weak N–H stretching frequency at 3295 cm⁻¹ merged with the broad acid –OH group and almost disappeared after the hydrogel shrinking. However, the position of the amide carbonyl group remained unchanged even after the hydrogel shrinking (Fig. S51a). For gelators G2 and G3, the N–H bending peak at 1539 cm⁻¹ and the amide carbonyl group at 1632 cm⁻¹ were observed to shift to 1543 cm⁻¹ and 1638 cm⁻¹, respectively (Fig. S51b). Lastly, the peaks at 2921 cm⁻¹ and 2852 cm⁻¹ were attributed to the asymmetric and symmetric stretching vibrations of –CH₂, confirming that the alkyl spacers of all gelators are highly ordered and packed in a zigzag parallel and antiparallel manner in the hydrogel matrix (Fig. S51a and b).

XRD analysis

An X-ray diffraction analysis was conducted to verify the stacking interaction or any multi-layered structure formation during the self-shrinking of the hydrogels. The peak observed in the range of 3.30–3.90 Å (2θ = 26.27–22.17°) is attributed to the π–π stacking interaction between the gelator molecules (Fig. S52a–f). Additionally, distinct peaks were identified at 2θ values of 14.42°, 20.77°, 29.34°, 36.06°, 41.12°, and 50.51°, which are approximately 7° apart, indicating the formation of a highly closely-packed multi-layered structure in the case of the G1 based dried hydrogel due to hydrogel syneresis (Fig. S52d). In contrast, G2 and G3 exhibited ordered peaks at 13.32–13.38°, 20.24–21.44°, 27.01–27.04°, 36.06–36.39°, and 42.39–42.45°, which are spaced approximately 6° or its multiples apart (Fig. S52e and f). The higher peak at 50.51° for G1 was absent for the G2 and G3 gelators. Before the shrinking of the G1 and G2-based hydrogels, no such ordered peaks were detected (Fig. S52a and b). In contrast, the XRD pattern of the G3 gelator-based hydrogel remained unchanged before and after shrinking (Fig. S52c and f).

CD study

The circular dichroism (CD) spectra of G1, G2, and G3 exhibit features characteristic of aromatic residues. Bands centered at 303 nm (G1), 300 nm (G2), and 299 nm (G3) arise from electronic transitions of the tryptophan indole moiety, while peaks at 254 nm (G1), 264 nm (G2), and 266 nm (G3) reflect combined contributions from phenylalanine and tryptophan. Additional features at 244 nm (G2) and 248 nm (G3) are consistent with π–π stacking among the phenylalanine rings (Fig. 3a).

Fig. 3 (a) Circular dichroism spectra of G1, G2, and G3 at 1.5 mM concentration with 10 mm path length. (b)–(d) Time dependent CD spectra of G1, G2 and G3, respectively, at 1.5 mM concentration with 10 mm path length.

Time-dependent CD measurements show progressive growth of the indole band in each system—303 nm for G1 (Fig. 3b), 300 nm for G2 (Fig. 3c), and 299 nm for G3 (Fig. 3d)—indicating that the tryptophan side chains experience increasingly ordered local environments. This trend supports a time-dependent enhancement in molecular organization consistent with self-assembly. Moreover, the phenylalanine-associated π–π stacking bands are intensified for G2 (244 nm) and G3 (248 nm), suggesting the strengthening of the aromatic interactions during assembly.

Rheological study

We performed rheological experiments on all three peptide-based hydrogels to understand well their mechanical strength and flow characteristics. We employed frequency sweep analysis to determine the viscoelastic properties, specifically focusing on the storage modulus (G′) and loss modulus (G″), which respectively represent the elastic (solid-like) and viscous (liquid-like) components of the materials. The frequency sweep analysis revealed that the storage modulus (G′) consistently exceeded the loss modulus (G″), confirming the solid-like nature of our materials. Before the shrinking, the storage modulus ranged from 10² to 10³ for all three hydrogels formed by the three hydrogelators. Interestingly, among the three systems, the hydrogel derived from gelator G1 displayed a noticeably different rheological profile compared to those based on G2 and G3. The G1-based hydrogel showed a gradual and continuous increase in both G′ and G″ values over time during the measurement. This trend can be attributed to the rapid self-shrinking property of the G1 hydrogel. The release of water results in a more densely packed molecular network, wherein the gelator molecules come into closer contact, promoting enhanced non-covalent interactions such as π–π stacking and hydrogen bonding. This molecular rearrangement leads to a significant increase in gel rigidity and mechanical strength, as reflected in the rising moduli. In contrast, G2 and G3-based hydrogels showed relatively stable moduli over time, suggesting much slower dynamic reorganization. This observation underscores the unique mechanical adaptability of the G1 hydrogel and highlights its potential utility in applications in the bio-materials field, particularly in sustained and targeted drug release applications where responsive stiffness and structural transformation are desired (Fig. S53).

Amplitude-sweep rheology defined the linear viscoelastic region (LVE) for hydrogels of G1–G3 and it remained linear up to ∼7.49% (G1), ∼1.82% (G2), and ∼1.95% (G3) shear strain. Beyond the LVE, the moduli were decreased and the G′/G″ crossover (yield point) occurred at ∼25.3% (G1) and ∼17.9% (G2, G3). To assess thixotropy, a time-dependent step-strain experiment was carried out at alternating 0.01% and 60% strain. At 60% the network yielded gel to sol transition; upon returning to 0.01% strain, the moduli were recovered, demonstrating the reversible, reproducible self-healing property of the hydrogel. These rheological features demonstrate the self-shrinking hydrogels’ suitability for foreseeable applications in wound healing and drug delivery (Fig. S54).

Morphological study

Initially, xerogel state scanning electron microscopic (SEM) studies were performed for the hydrogels to understand the morphological features of these tripeptide gelators. The gelator G1 shows an interconnected vesicle-like morphological structure formed by one vesicle with another. However, G2 and G3-based hydrogels show nano-fiber-like morphological structures in their respective hydrogel states. The SEM study confirms that the morphological features of G1 (Fig. 4a) are different from G2 and G3 tripeptide gelators (Fig. 4b and c).

Fig. 4 The SEM images of G1 (a), G2 (b) and G3 (c) peptide-based hydrogels at 7.46 PBS in their xerogel state. HR-TEM images of (d) before and (g) after the shrinking of the hydrogel formed by the gelators G1, (e) before and (h) after the shrinking of the hydrogel formed by the gelator G2, and (f) before and (i) after the shrinking of the G3 based hydrogel at 7.46 PBS.

Transmission electron microscopic (TEM) studies of the xerogels derived from the G1, G2, and G3-based shrinking hydrogels were also performed to elucidate the hydrogel-shrinking mechanism. The G1 shows a vesicle like morphological structure to close packed network structure (Fig. 4d and g). Whereas, both the G2 and G3 hydrogels show a nanofiber-like morphological structure before shrinking (Fig. 4e and f), and after their shrinking the nano fiber-like morphological structure transformed into a fiber-to-fiber fiber-interacted network structure for the G2 gelator peptide or fiber-to-fiber-intercepted nano rod-like structure for the G3 gelator peptide (Fig. 4h and i).

As both the SEM and TEM studies required the drying of the sample, there is a chance of morphological transformation. So, for real-time sample analysis, we also performed optical microscopic studies, which confirmed that the G2 and G3 gelators show nanofiber-like morphological structures. However, the G1 gelator forms a more complex network structure using the vesicle-like nanostructure formed by the gelator molecules (Fig. S55). These developments of morphological structure were further examined in the next part of the morphological experiment through the Cryo-TEM study.

To understand clearly, a Cryo-TEM study was performed for the time-dependent shrinking hydrogels. The hydrogel exhibited a vesicle-like morphological structure before shrinking, which was also confirmed by the SEM study. Then, the time-dependent shrinking of the hydrogel for 0 min, 15 min, 30 min, and 90 min shows that slowly, the vesicle-like unit comes closer to forming a fiber chain. Then, these fiber chains come closer to each other to create a network structure (Fig. 5a). The other two gelators, G2 and G3, show nano fiber-like network structures in their hydrogel states before the shrinking. However, the shrinking hydrogels show that these nanofibers are coming closer to forming nano rod-like structures due to fiber-to-fiber interaction (Fig. 5b and c). So, in the case of the G1 gelator, the inherent hydrophobic microenvironment is responsible for shrinking, whereas in the case of G2 and G3 hydrogelators, the fiber-to-fiber interaction and the intrinsic hydrophobicity both are responsible for the self-shrinking of these hydrogels.

Fig. 5 Cryo-TEM images for time-dependent morphological changes during the self-shrinking of (a) G1, (b) G2 and (c) G3 hydrogelators in 7.46 phosphate buffer solutions at room temperature, 28 °C.

DLS study

The DLS study is a powerful tool for studying the diffusion behavior of macromolecules in solution. Hence, the hydrodynamic radii can be calculated from the diffusion coefficient, depending upon the size and shape of the macromolecule. This study also demonstrates the homogeneity of our peptide in its hydrogel states. A DLS study was performed to support the morphological data of the G1-based peptide hydrogel's vesicle-like structure, which shows that the d (nm) of the vesicle was 58.77 to 68.06 nm and more than 55% of the vesicle size is under 58.77 nm (Fig. S56a and b). After shrinking, the nano fiber-like structure is formed, so the latter study confirms that the size after the formation of the nanofiber is large and the d (nm) size comes within the range 58.77 to 54.68 nm with a very high range confirming the hydrogelators morphological transformation from a vesicle to close-packed nano fiber network structure. In this case, only 5–6% of the gelator exists in its initial 58.77 to 68.06 nm range (Fig. S56c and d).

Morphology of the self-shrinking tripeptides from CG molecular dynamics

To understand the dynamic self-assembly and morphological evolution of the various tripeptide (G1, G2, and G3) assemblies, we calculated several structural and thermodynamic properties as shown in Fig. 6–8. Fig. 6 presents the time-dependent morphological evolution of the G1 tripeptides. Initially, all the 1200 G1 tripeptides were placed randomly in a cubic simulation box. As the simulation time progressed (t = 15 ns), randomly placed tripeptides begin to aggregate into smaller clusters. Between t = 500 ns and t = 4 µs, these smaller aggregates merged into a single large cluster, giving rise to interconnected network-like morphologies as shown in Fig. 6a–e.

Fig. 6 Instantaneous snapshots from the self-assembly pathway of 1200 G1 tripeptides starting from an initial random state in a cubic box of edge length 22 nm. (a) Energy minimized randomly placed G1 tripeptides at t = 0 ns, (b) conformation of their self-assembly after t = 15 ns of CG-MD equilibration, (c)–(e) morphologies of G1 at different time intervals from 500 ns to 4 µs. The blue frame in the snapshots shows the simulation boxes.

Fig. 7 Instantaneous snapshots of 1200 G2 tripeptide gelators' self-assembly dynamics in a 22 × 22 × 22 nm³ cubic box at different time intervals. (a) Energy minimized conformations of the randomly placed G2 tripeptides, (b) conformation after t = 15 ns of CG-MD simulation time, and (c)–(e) final morphologies of the G2 tripeptide at different time frames from t = 500 ns to t = 4 µs. At t = 2 µs, the G2 tripeptides shrink to nanorods, then near t = 4 µs, the final morphology becomes like an oblong shape.

Fig. 8 Representative snapshots of 1200 G3 tripeptide self-assembly in a 22 × 22 × 22 nm³ box at different time intervals, leading to the formation of nanorods. (a) Initial energy minimized structure (t = 0 ns), (b) conformation after t = 15 ns of CG-MD simulations, (c)–(e) morphologies at different time intervals from t = 500 ns to t = 4 µs time points. The final structures gradually evolve into nanorods. The simulation box is shown in the blue frame.

While investigating the molecular origin of the distinct morphologies, we observed a rapid and spontaneous self-aggregation of the G1 tripeptides transitioning from smaller clusters to an extended network-like morphology within the first few nanoseconds. This morphological transformation of G1 is primarily driven by strong π–π stacking interactions between the aromatic side chains (SC–SC) and main chains (MC–MC) of the tripeptides. The CG beads representing aromatic rings in the tripeptides are designated as side chains (SC), while the backbone beads are labelled as main chains (MC) and based on their sequence positions from the N-terminus, the side chains are labelled as SC1, SC2, and SC3 as shown in Fig. S49a, with a detailed schematic provided in the SI. The rapid self-shrinking and enhanced stacking in G1 are attributed to the central placement of the Phe (F) residue, which facilitates stronger π–π interactions compared to Trp (W) at the center.

In contrast, the G2 and G3 tripeptides, where the Phe (F) residue is positioned at the termini, exhibit slower aggregation and transition from nanofibers to nanorod-like structures as shown in Fig. 7a–e and 8a–e, respectively. The terminal placement of Phe (F) disrupts efficient stacking, causing zigzag arrangements of the side chains and backbones and leading to weaker SC–SC and MC–MC interactions.

Moreover, we observed a pronounced local ordering of side chain–side chain (SC–SC) in the network/fibrous-like morphology formed by G1. This is mostly driven by central Phe (F)-mediated stacking and the presence of both parallel and antiparallel side chain alignment in G1 as shown in Fig. 9a. This leads to dominant interactions between similarly stacked intermolecular side chains (e.g., SC1–SC1, SC2–SC2 and SC3–SC3 parallel alignment) and between oppositely positioned terminal side chains (e.g., SC1–SC3, antiparallel alignment). Interestingly, such kind of local ordering was absent in the morphology of G2 and G3 as shown in Fig. 9b and c. Thus, our CG-MD results show strong alignment with experimental observations (final conformations), particularly in capturing the distinct morphologies and stabilities in terms of faster self-shrinking of G1 compared to G2, and G3 assemblies, as shown in Fig. 9 and 10b.

Fig. 9 Representation of the final morphologies of tripeptide gelators from the 4 µs long CG-MD simulations – (a) the network-like morphology of the G1 tripeptides, showing the local ordering of SC–SC (smaller domains as highlighted in green box) due to stronger pi–pi stacking interactions, (b) and (c) the oblong and nanorod-like morphology for the G2 and G3 tripeptides, due to weaker pi–pi stacking interactions and zigzag reorganization of the tripeptides due to them engaging in a different kind of stacking interaction between different types of tripeptide residues.

Fig. 10 (a) The time evolution of the solvent-accessible-surface-area (SASA) for the three tripeptides for 4 µs simulation time. (b) Distribution of the number of clusters formed for the 1200 tripeptides during 4 µs CG-MD simulation time.

To get deeper insights into the structural stability of the morphologies of three tri-peptide systems (for G1, G2, and G3), we investigated the relationship between tripeptide aggregation and hydrophobicity by calculating the solvent-accessible-surface-area (SASA). Initially, the random dispersion of tripeptides in the solvent results in a higher SASA. As the system evolves with simulation time, the peptides self-assemble through non-bonded van der Waals interactions and π–π stacking, leading to a progressive decrease in SASA over time, as shown in Fig. 10a. Among the three systems, the G3 tripeptides exhibit the lowest SASA values, which can be attributed to their tightly packed nanorod-like morphology. This compact structure limits water penetration, indicating a higher degree of hydrophobic aggregation compared to the slightly more open network-like morphology of G1 and the intermediate oblong-like structure of the G2 tripeptides.

To study the kinetics of the self-assembly process of the three tripeptides, we have investigated the cluster size distributions over 4 µs CG-MD simulation time. We have calculated the number of clusters as a function of simulation time, as shown in Fig. 10b. Among the three systems, the G1 tripeptide exhibits the fastest dynamic self-shrinking, with the number of clusters decreasing to a minimum within 160 ns. This is in qualitative agreement with the experimental observation. In contrast, G2 and G3 reach similar levels of aggregation at around 300 ns and 500 ns, respectively, as indicated by the vertical dashed lines in the inset of Fig. 10b. The rapid aggregation in G1 is driven by close contacts and higher stacking energy between identical intermolecular side chains (e.g., SC1–SC1, SC2–SC2, SC3–SC3 for parallel alignments, and SC1–SC3 for antiparallel alignments) across all 1200 tripeptides. The number of contacts and minimum distances between SC and MC also qualitatively suggest strong π–π stacking for G1, which would result in a different morphology than G2 and G3 tripeptides, as illustrated in Fig. S57–S59 in the SI.

Below we provide a brief qualitative description of the stacking interactions. The aromatic stacking between the side and main chains can be present in several modes, such as face-to-face, edge-to-face, and offset configurations as described in Fig. S49b in the SI. Considering that the Phe–Phe (F–F) amino acid residue exhibits stronger stacking and stable interactions in peptide assemblies, and due to the central positioning of F, the G1 tripeptide facilitates the network-like morphology due to strong stacking between the G1 tripeptide pairs, with identical side chains (e.g. SC1–SC1, SC2–SC2, and SC3–SC3). In the absence of reorganization, the G1 tripeptides are also aligned anti-parallel; therefore, SC1–SC3 interaction also becomes dominant as shown in Fig. 11a–c. Thus, the stronger stacking and central placement of Phe (F) in G1 lead to a network-like structure, distinct from the oblong/nanorod morphologies obtained for G2 and G3.

Fig. 11 Representations of π–π stacking interactions for G1 and G3 tripeptides. (a)–(c) Stacking interactions between identical intermolecular side chains for parallel (e.g., SC1–SC1, SC2–SC2, SC3–SC3) and antiparallel (SC1–SC3) arrangements in the G1 tripeptide final cluster. (d)–(f) Representations of stacking interactions between intermolecular side chains of different types (e.g., SC2–SC3, SC1–SC2) due to weaker stacking and more reorganization for G3. The tripeptide G2 also exhibits stacking patterns similar to G3 due to the identical terminal positioning of Phe (F), which is not shown here.

In the case of the G2 and G3 tripeptides, the terminal Phe (F) position leads to dominant stacking between non-identical side chains (e.g., SC1–SC2, SC2–SC3), with fewer identical side chain interactions. Therefore, the combined effect of disrupted central stacking symmetry and increased reorganization of Trp (W) results in oblong or nanorod-like structures for G2 and G3, as illustrated in (Fig. 11d–f).

To elucidate the average structural organization of the dynamic tripeptide gelators and the spatial distribution of amino acid residues within the assembled morphologies, we computed the intermolecular radial distribution function, g(r), as shown in Fig. 12a–f. In G1 tripeptides, strong π–π interactions between identical side chains drive close packing through parallel and antiparallel alignments. This is evident from the RDF, where the SC2–SC2 (F–F) pair shows a significantly higher first peak (red curve, Fig. 12a) than SC1–SC1 and SC3–SC3 (W–W), highlighting the dominant role of central F–F stacking in G1 self-assembly (Fig. 12a–c). While the Phe (F) exhibits stronger intermolecular stacking than Trp (W), its terminal placement in G2 and G3 leads to more zigzag reorganization rather than the parallel and antiparallel stacking. The zigzag arrangement in G2 and G3 is evident from the pair distribution functions of different side chain combinations. Overall, G2 and G3 exhibit higher peak intensities than G1, reflecting side chain reorganization and the lack of parallel or antiparallel alignment (Fig. 12d–f). The position of Phe (F) plays a key role in determining tripeptide morphology. To further probe internal structure, RDFs from the assembly centre were computed, revealing spatial distributions of hydrophilic and hydrophobic residues as shown in Fig. S60 in the SI.

Fig. 12 (a)–(c) Radial distribution function between pairs of identical intermolecular side chains (e.g., SC1–SC1, SC2–SC2, SC3–SC3), for G1 and similarly for G2 and G3. (d)–(f) Representation of RDF between different types of intermolecular side chains (e.g., SC1–SC2, SC1–SC3, etc.) for G1, G2, and G3 aggregates.

To probe aromatic stacking strength between the pair of tripeptides, we have simulated three pairs of tripeptides in water in separate simulations. We have calculated the binding free energy between two vertically stacked G1, G2, and G3 tripeptides using metadynamics simulations. In the metadynamics simulations, the COM distance between the two tripeptides was used as the reaction coordinate. As shown in Fig. 13a–c, the G1 tripeptides exhibit the strongest binding free energy of −17.55 kcal mol⁻¹, compared to −15.67 kcal mol⁻¹ for G2 and −15.34 kcal mol⁻¹ for G3. This result clearly supports our hypothesis that central Phe (F) in G1 enables stronger π–π stacking than the terminal Phe (F) in G2 and G3.

Fig. 13 Binding free energies between the centre-of-mass (COM) of two vertically stacked tripeptides using metadynamics simulations for – (a) G1 (−17.55 kcal mol⁻¹), (b) G2 (−15.67 kcal mol⁻¹) and (c) G3 (−15.34 kcal mol⁻¹). Binding free energy values clearly indicate that the order of stacking energy follows as G1 > G2 > G3.

To identify the key interactions driving self-assembly, we analysed non-bonded energy contributions (Fig. 14a and b). The fibrous network-like morphology of the G1 tripeptide shows significantly lower total potential energy than the oblong/nanorod-like structures of G2 and G3 (Fig. 14c and d), indicating higher stability for G1 due to central Phe (F) mediated π–π stacking. The dominant SC–SC van der Waals interactions highlight the role of aromatic stacking, which is consistent with the number of contacts and RDF analysis. Overall, the morphological differences arise from sequence-dependent variations in SC–SC and MC–MC interactions. In G1, strong SC–SC hydrophobic stacking gives rise to a well-connected network-like final structure. In contrast, keener competition between SC–SC and MC–MC interactions in the G2 and G3 hydrogels yields less ordered zigzag nanorods and oblong morphologies. Overall, our theoretical CG models provide a reliable qualitative understanding and mechanistic insights consistent with experimental findings.

Fig. 14 Time-averaged non-bonded interaction energies of the G1, G2, and G3 tripeptide hydrogels – (a) between side chain–side chain (SC–SC), (b) between main chain–main chain (MC–MC), (c) total van der Waals (vdW) energy for all 1200 tripeptides, and (d) total potential energy (vdW + electrostatics) for all 1200 tripeptides in the simulation box.

3D-shape memory and self-healing study

3D shape memory materials are intelligent materials that can change shape in response to external stimuli, such as temperature or light, or sometimes without any external stimuli. In a recent report the shape memory behaviour of a hydrogel was demonstrated by Shanmugam and co-workers showing the gel's ability to retain and recall the shape of the container in which it was initially formed, even after shrinking. The hydrogel, formed from fluorenylmethoxycarbonyl-β-L-phenylalanine (Fmoc-β-Phe), a low-molecular-weight (LMW) gelator, self-assembles into a 3D network at physiological pH and room temperature. After gelation, it spontaneously undergoes syneresis – a process where water is expelled from the hydrogel matrix-resulting in volume shrinkage over time. When the hydrogel is formed in containers of different shapes (e.g., circular, square, rectangular), the hydrogel maintained the 3D shape even after shrinking, confirming its shape memory capability.²²

In this report, our prepared hydrogels undergo syneresis three-dimensionally and can regain their original state by simply heating. This kind of 3D-shrinking nature can be used to make different types of shapes (cylindrical and rectangular) using glass vials and glass cuvettes (Fig. 15a–c). The shape memory behaviour of the hydrogel is demonstrated by its ability to retain and recall the shape of the container in which it was initially formed, even after shrinking. Our reported hydrogels not only can be used for their 3D shape memory performance but also can be used as self-healing materials. The self-healing properties of these hydrogels are due to the reversible nature of the peptide interactions, allowing the material to reform and heal after damage. This is achieved through various mechanisms such as hydrogen bonding, ionic interactions, and hydrophobic interactions. To test the self-healing nature of our prepared hydrogel, methylene blue and Congo red charge self-shrinking are kept in contact for 60 min (Fig. 15). It was observed that the two different dye-induced shrinking gels are joined together by using H-bonding, ionic, and hydrophobic interactions, indicating the intrinsic self-healing nature of the shrinking hydrogel.

Fig. 15 3D nature of the shrinking hydrogel in a (a) cylindrical, (b) rectangular and (c) square shape glass vial. (d)–(f) Self-healing nature of the shrinking hydrogel with time.

Conclusions

In this report, three distinct tripeptides were designed by altering the position of the L-phenylalanine (F) amino acid residue and each was capable of forming hydrogels in a phosphate buffer at pH 7.46. These hydrogels display a striking time-dependent self-shrinking phenomenon by expelling water molecules and spontaneously assembling into densely packed nanorod structures. This positional modification significantly influences the self-shrinking behaviour of the resulting hydrogelators. Among the three, one exhibited a remarkably rapid self-shrinking rate, while the other two demonstrated comparatively slower kinetics. This difference in the shrinking mechanism stems from the stronger π–π stacking and van der Waals interactions. Interestingly, these tripeptide-based hydrogels exhibit three-dimensional shape-memory properties and function as second-generation self-healing materials, capable of restoring their original structure through reversible noncovalent interactions. Their smart behaviour highlights their potential for development into advanced and intelligent biomaterials in the future. Self-shrinking hydrogels offer autonomous volume reduction that expels solvent and densifies gel networks, enabling drug depots with tunable release, wound dressings that manage exudate and promote closure, self-tightening tissue sealants, and microfluidic valves/pumps requiring no external power. Shrinkage-driven preconcentration can improve analytical sensitivity, while programmable contraction provides chemo-mechanical actuation for soft robotics and 4D materials. In tissue engineering, controlled compaction can offer maturation cues and customizable porosity. Reversible shrink–swell cycles further provide separations and filtration. To explore the impact of sequence variation, specifically the positioning of the L-phenylalanine residue, coarse-grained molecular dynamics simulations were conducted. These simulations provided insights into how different peptide sequences affect the self-assembly process, leading to the formation of nanoscale network structures typical of the hydrogels. Notably, the computational results closely mirrored the experimental observations, validating that the distinct hydrogel architectures are governed by the sequence-dependent assembly behaviour of the amphiphilic peptides.

Author contributions

B. M. carried out the synthesis, analysis, all experiments, data interpretation of the experimental work and initially drafted the manuscript. S. M. carried out the theoretical studies and initially drafted the manuscript for this theoretical study. T. M. helped in the experiments regarding the DLS, FT-IR, Cryo-TEM, designed the Graphical abstract and improved the figure design of the initial draft of the manuscript. A. B. and P. K. M. conceived, designed and coordinated the study, and composed the manuscript to its final form. All the authors gave their final approval for publication.

Conflicts of interest

There are no conflicts to declare.

Data availability

The data supporting this article have been included as part of the supplementary information (SI). Supplementary information: instrumentation; synthesis, characterisation and spectra (¹H, ¹³C-NMR and HR-MS), theoretical data for stacking interactions, FT-IR data, PXRD data, rheological data, DLS data, and theoretical data. See DOI: https://doi.org/10.1039/d5sm00859j.

Acknowledgements

B. M. would like to acknowledge IACS, Kolkata, India; S. M. and T. M. would like to thank CSIR, New Delhi, India, for their financial support. P. K. M. acknowledges the Department of Science & Technology (DST), India for financial support and the Science and Engineering Research Board (SERB) for financial and computational support (No. CRG/2021/003659).

Notes and references

1. F. Sheehan, D. Sementa, A. Jain, M. Kumar, M. Tayarani-Najjaran, D. Kroiss and R. V. Ulijn, Chem. Rev., 2021, 121, 13869–13914,  DOI:10.1021/acs.chemrev.1c00089.
2. F. Rodríguez-Llansola, B. Escuder and J. F. Miravet, J. Am. Chem. Soc., 2009, 131, 11478–11484,  DOI:10.1021/ja902589f.
3. W. Tan, Q. Zhang, M. C. Quiñones-Frías, A. Y. Hsu, Y. Zhang, A. Rodal, P. Hong, H. R. Luo and B. Xu, J. Am. Chem. Soc., 2022, 144, 6709–6713,  DOI:10.1021/jacs.2c02238.
4. A. R. Hirst, B. Escuder, J. F. Miravet and D. K. Smith, Angew. Chem., Int. Ed., 2008, 47, 8002–8018,  DOI:10.1002/anie.200800022.
5. B. O. Okesola and D. K. Smith, Chem. Soc. Rev., 2016, 45, 4226–4251,  10.1039/c6cs00124f.
6. K. Basu, A. Baral, S. Basak, A. Dehsorkhi, J. Nanda, D. Bhunia, S. Ghosh, V. Castelletto, I. W. Hamley and A. Banerjee, Chem. Commun., 2016, 52, 5045–5048,  10.1039/c6cc01744d.
7. B. Mondal, B. Hansda, T. Mondal, P. Pal, K. Basu and A. Banerjee, Langmuir, 2024, 40, 21876–21883,  DOI:10.1021/acs.langmuir.4c03210.
8. T. Mondal, S. Roy, A. Das, S. Banerjee, B. Mondal, N. Chatterjee and A. Banerjee, Chem. Commun., 2025, 61, 8204–8207,  10.1039/d5cc00660k.
9. K. Basu, B. Mondal, A. Das Mahapatra, N. Nandi, D. Basak and A. Banerjee, J. Phys. Chem. C, 2019, 123, 20558–20566,  DOI:10.1021/acs.jpcc.9b04414.
10. T. Mondal, S. Patra, B. Mondal, P. Ghosh, I. W. Hamley and A. Banerjee, ACS Appl. Polym. Mater., 2024, 6, 11383–11391,  DOI:10.1021/acsapm.4c01965.
11. A. Baral, S. Roy, A. Dehsorkhi, I. W. Hamley, S. Mohapatra, S. Ghosh and A. Banerjee, Langmuir, 2014, 30, 929–936,  DOI:10.1021/la4043638.
12. P. Chakraborty, T. Guterman, N. Adadi, M. Yadid, T. Brosh, L. Adler-Abramovich, T. Dvir and E. Gazit, ACS Nano, 2019, 13, 163–175,  DOI:10.1021/acsnano.8b05067.
13. B. Hansda, B. Mondal, S. Hazra, K. S. Das, V. Castelletto, I. W. Hamley and A. Banerjee, Soft Matter, 2023, 19, 8264–8273,  10.1039/D3SM00883E.
14. B. Mondal, V. K. Gupta, B. Hansda, A. Bhoumik, T. Mondal, H. K. Majumder, C. J. C. Edwards-Gayle, I. W. Hamley, P. Jaisankar and A. Banerjee, Soft Matter, 2022, 18, 7201–7216,  10.1039/D2SM00562J.
15. T. Mondal, A. Chatterjee, B. Hansda, B. Mondal, P. Sen and A. Banerjee, Soft Matter, 2024, 20, 1236–1244,  10.1039/D3SM01291C.
16. A. M. Garcia, M. Melchionna, O. Bellotto, S. Kralj, S. Semeraro, E. Parisi, D. Iglesias, P. D’Andrea, R. De Zorzi and A. V. Vargiu, et al. , ACS Nano, 2021, 15, 3015–3025,  DOI:10.1021/acsnano.0c09386.
17. S. Bera, S. Mondal, B. Xue, L. J. W. Shimon, Y. Cao and E. Gazit, Nat. Mater., 2019, 18, 503–509,  DOI:10.1038/s41563-019-0343-2.
18. D. Niu, Y. Jiang, L. Ji, G. Ouyang and M. Liu, Angew. Chem., Int. Ed., 2019, 58, 5946–5950,  DOI:10.1002/anie.201900607.
19. J. A. Foster, K. K. Damodaran, A. Maurin, G. M. Day, H. P. G. Thompson, G. J. Cameron, J. C. Bernal and J. W. Steed, Chem. Sci., 2016, 8, 78–84,  10.1039/C6SC04126D.
20. L. Adler-Abramovich and E. Gazit, Chem. Soc. Rev., 2014, 43, 6881–6893,  10.1039/c4cs00164h.
21. S. Basak, N. Nandi, S. Paul, I. W. Hamley and A. Banerjee, Chem. Commun., 2017, 53, 5910–5913,  10.1039/c7cc01774j.
22. D. K. Duraisamy, P. D. Sureshbhai, P. Saveri, A. P. Deshpande and G. Shanmugam, Chem. Commun., 2022, 58, 13377–13380,  10.1039/d2cc05507d.
23. J. Chen, T. Wang and M. Liu, Inorg. Chem. Front., 2016, 3, 1559–1565,  10.1039/c6qi00238b.
24. L. Qin, P. Duan, F. Xie, L. Zhang and M. Liu, Chem. Commun., 2013, 49, 10823–10825,  10.1039/c3cc47004k.
25. K. A. Dill and J. L. Maccallum, Science, 2012, 338, 1042–1046.
26. Y. Liu, Y. Wu, Z. Luo and M. Li, iScience, 2023, 26, 106279,  DOI:10.1016/j.isci.2023.106279.
27. B. Mondal, D. Bairagi, N. Nandi, B. Hansda, K. S. Das, C. J. C. Edwards-Gayle, V. Castelletto, I. W. Hamley and A. Banerjee, Langmuir, 2020, 36, 12942–12953,  DOI:10.1021/acs.langmuir.0c02205.
28. J. Li, Y. Kuang, Y. Gao, X. Du, J. Shi and B. Xu, J. Am. Chem. Soc., 2013, 135, 542–545,  DOI:10.1021/ja310019x.
29. S. Zhang, Nat. Biotechnol., 2003, 21, 1171–1178,  DOI:10.1038/nbt874.
30. J. B. Matson, R. H. Zha and S. I. Stupp, Curr. Opin. Solid State Mater. Sci., 2011, 15, 225–235,  DOI:10.1016/j.cossms.2011.08.001.
31. S. Fleming and R. V. Ulijn, Chem. Soc. Rev., 2014, 43, 8150–8177,  10.1039/C4CS00247D.
32. T. Schnitzer, M. Schnurr, A. F. Zahrt, N. Sakhaee, S. E. Denmark and H. Wennemers, ACS Cent. Sci., 2024, 10, 367–373,  DOI:10.1021/acscentsci.3c01284.
33. J. Mayr, C. Saldías and D. Díaz Díaz, Chem. Soc. Rev., 2018, 47, 1484–1515,  10.1039/C7CS00515F.
34. J. M. Wolfe, C. M. Fadzen, Z.-N. Choo, R. L. Holden, M. Yao, G. J. Hanson and B. L. Pentelute, ACS Cent. Sci., 2018, 4, 512–520,  DOI:10.1021/acscentsci.8b00098.
35. M. Miotto, R. M. Gouveia, A. M. Ionescu, F. Figueiredo, I. W. Hamley and C. J. Connon, Adv. Funct. Mater., 2019, 29, 1807334,  DOI:10.1002/adfm.201807334.
36. S. Pande, F. Pati and P. Chakraborty, ACS Appl. Bio Mater., 2024, 7, 5885–5905,  DOI:10.1021/acsabm.4c00879.
37. S. Deb, S. Gupta, S. Bose, T. Mondal, B. Mondal and A. Banerjee, ACS Appl. Bio Mater., 2025, 8, 3061–3075,  DOI:10.1021/acsabm.4c01891.
38. S. Panja and D. J. Adams, Chem. Soc. Rev., 2021, 50, 5165–5200,  10.1039/d0cs01166e.
39. M. P. Conte, N. Singh, I. R. Sasselli, B. Escuder and R. V. Ulijn, Chem. Commun., 2016, 52, 13889–13892,  10.1039/c6cc05821c.
40. K. Komatsu, et al. , Chem. Commun., 2009, 9205–9207.
41. B. L. Abraham, P. Agredo, S. G. Mensah and B. L. Nilsson, Langmuir, 2022, 38, 15494–15505,  DOI:10.1021/acs.langmuir.2c01394.
42. S. Kiyonaka, S.-L. Zhou and I. Hamachi, Supramol. Chem., 2003, 15, 521–528.
43. L. Qin, F. Xie, X. Jin and M. Liu, Chem. – Eur. J., 2015, 21, 11300–11305,  DOI:10.1002/chem.201500929.
44. S. L. Zhou, S. Matsumoto, H. D. Tian, H. Yamane, A. Ojida, S. Kiyonaka and I. Hamachi, Chem. – Eur. J., 2005, 11, 1130–1136,  DOI:10.1002/chem.200400677.
45. M. P. Conte, N. Singh, I. R. Sasselli, B. Escuder and R. V. Ulijn, Chem. Commun., 2016, 52, 13889–13892,  10.1039/C6CC05821C.
46. L. Qin, F. Xie, P. Duan and M. Liu, Chem. – Eur. J., 2014, 20, 15419–15425,  DOI:10.1002/chem.201404035.
47. J. Chen, T. Wang and M. Liu, Chem. Commun., 2016, 52, 11277–11280,  10.1039/c6cc05968f.
48. T. Sugiura, T. Kanada, D. Mori, H. Sakai, A. Shibata, Y. Kitamura and M. Ikeda, Soft Matter, 2020, 16, 899–906,  10.1039/c9sm01969c.
49. F. Xie, L. Qin and M. Liu, Chem. Commun., 2016, 52, 930–933,  10.1039/c5cc08076b.
50. S. Panja, B. Dietrich and D. J. Adams, Angew. Chem., Int. Ed., 2022, 61, e202115021,  DOI:10.1002/anie.202115021.
51. S. Y. Lee, C. L. Chen and S. Zhang, Biomacromolecules, 2024, 25, 1250–1263,  DOI:10.1021/acs.biomac.4c00050.
52. P. K. Dutta, R. Banerjee and M. H. Stenzel, Soft Matter, 2023, 19, 5467–5480,  10.1039/D3SM00363A.
53. F. Yang, T. Wu and M. Liu, Chem. – Eur. J., 2024, 30, e202400134,  DOI:10.1002/chem.202400134.
54. M. D. Hanwell, et al. , J. Cheminf., 2012, 4, 17,  DOI:10.1186/1758-2946-4-17.
55. J. Huang, S. Rauscher, G. Nawrocki, T. Ran, M. Feig, B. L. de Groot, H. Grubmüller and A. D. MacKerell, Nat. Methods, 2017, 14, 71–73,  DOI:10.1038/nmeth.4067.
56. R. B. Best, X. Zhu, J. Shim, P. E. M. Lopes, J. Mittal, M. Feig and A. D. MacKerell, J. Chem. Theory Comput., 2012, 8, 3257–3273,  DOI:10.1021/ct300400x.
57. M. J. Abraham, T. Murtola, R. Schulz, S. Páll, J. C. Smith, B. Hess and E. Lindahl, SoftwareX, 2015, 1–2, 19–25,  DOI:10.1016/j.softx.2015.06.001.
58. J. A. Graham, J. W. Essex and S. Khalid, J. Chem. Inf. Model., 2017, 57, 650–656,  DOI:10.1021/acs.jcim.7b00096.
59. D. H. de Jong, G. Singh, W. F. D. Bennett, C. Arnarez, T. A. Wassenaar, L. V. Schäfer, X. Periole, D. P. Tieleman and S. J. Marrink, J. Chem. Theory Comput., 2013, 9, 687–697,  DOI:10.1021/ct300646g.
60. S. J. Marrink, H. J. Risselada, S. Yefimov, D. P. Tieleman and A. H. de Vries, J. Phys. Chem. B, 2007, 111, 7812–7824,  DOI:10.1021/jp071097f.
61. L. Monticelli, S. K. Kandasamy, X. Periole, R. G. Larson, D. P. Tieleman and S.-J. Marrink, J. Chem. Theory Comput., 2008, 4, 819–834,  DOI:10.1021/ct700324x.
62. W. Humphrey, A. Dalke and K. Schulten, J. Mol. Graphics, 1996, 14, 33–38,  DOI:10.1016/0263-7855(96)00018-5.

---

Footnote

† Contributed equally.

---