A Carbodiimide-Fueled Reaction Cycle That Forms Transient 5(4H)-Oxazolones

Abstract

In life, molecular architectures, like the cytoskeletal proteins or the nucleolus, catalyze the conversion of chemical fuels to perform their functions. For example, tubulin catalyzes the hydrolysis of GTP to form a dynamic cytoskeletal network. In contrast, myosin uses the energy obtained by catalyzing the hydrolysis of ATP to exert forces. Artificial examples of such beautiful architectures are scarce partly because synthetic chemically fueled reaction cycles are relatively rare. Here, we introduce a new chemical reaction cycle driven by the hydration of a carbodiimide. Unlike other carbodiimide-fueled reaction cycles, the proposed cycle forms a transient 5(4H)-oxazolone. The reaction cycle is efficient in forming the transient product and is robust to operate under a wide range of fuel inputs, pH, and temperatures. The versatility of the precursors is vast, and we demonstrate several molecular designs that yield chemically fueled droplets, fibers, and crystals. We anticipate that the reaction cycle can offer a range of other assemblies and, due to its versatility, can also be incorporated into molecular motors and machines.

Introduction

In molecular self-assembly, molecules form supramolecular architectures held together through noncovalent interactions.¹⁻³ In recent years, such assemblies have led to various materials used in healthcare,^4,5 optoelectronics,^6,7 and catalysis,^8,9 among others.¹⁰ Most of these supramolecular materials exist in or close to equilibrium, contrasting biological assemblies sustained out of equilibrium by the constant transduction of energy harvested from chemical fuels through chemical reaction cycles. For example, the hydrolysis of ATP is catalyzed by actin, and the energy gained from this catalytic reaction cycle is used in the dynamic self-assembly of actin filaments.¹¹ Similarly, the energy released upon hydrolyzing GTP is used to fuel the dynamic assembly of tubulin into microtubules.¹²

In such chemically fueled self-assembly, a catalytic reaction cycle is responsible for extracting the chemical fuel’s energy by converting it into waste (Scheme 1A). A precursor molecule reacts with the chemical fuel, yielding the activated product and a waste molecule, i.e., the activation. In the deactivation, this activated product reverts to the precursor. Thus, the precursor catalytically converts fuel into waste while temporarily becoming activated. If the precursor is well-soluble but the product assembles, the energy harvested from the chemical fuel creates transient building blocks for self-assembly. The dynamic nature of the activation and deactivation process endows chemically fueled assemblies with inherently different material properties compared to their in-equilibrium counterparts, which include spatiotemporal control over the assemblies and their ability to self-heal.^13,14 Moreover, pattern formation and oscillatory behavior can be expected when feedback mechanisms are incorporated.¹⁵ Besides their unique material properties, synthetic counterparts of these materials serve as a model system for biological assemblies. Alternatively, the chemical fuel can be used to operate a molecular machine or motor.^16,17 The chemical energy of the fuel to waste conversion is in such designs used for directional rotation.

Scheme 1. Chemical Reaction Cycle That Regulates Molecular Assembly.

(A) Schematic representation of a chemically fueled reaction cycle coupled to self-assembly. Activation and deactivation regulate the self-assembly behavior of a building block. (B) The hydration of a carbodiimide fuel drives a chemically fueled reaction cycle to form an anhydride from two carboxylates or (C) an intramolecular anhydride or (D) an oxazolone from a peptide.

Thus, several chemically fueled reaction cycles have been designed to regulate molecules. The energy transduced by these cycles can come from the oxidation of reducing agents,¹⁸⁻²¹ hydrolysis of methylating agents,²² hydration of carbodiimides,²³⁻²⁹ and hydrolysis of ATP.^21,30 These cycles lead to a range of assemblies, including fibers,³¹⁻³³ droplets,^14,34 vesicles,³⁵ micelles,^18,36 colloids,^19,23 and particle clusters.^13,37 A well-studied cycle produces energy by hydrating carbodiimides.^23,38 In these cycles, carboxylates react with EDC (1-ethyl-3-(3-(dimethylamino)propyl)carbodiimide) to form an O-acylurea intermediate. A second carboxylate attacks this intermediate to form the anhydride product. Hydrolysis reverts the precursor, which constitutes the deactivation reaction (Scheme 1B and C).

Despite its success in fueling molecular assemblies, a significant downside of this reaction cycle is two unwanted side reactions. First, the O-acylurea can rearrange to an unwanted N-acylurea. This reaction is irreversible and scales inversely with the concentration of carboxylates. Second, the reactive O-acylurea can hydrolyze. This reaction results in the consumption of fuel without the formation of the wanted anhydride product. Both side reactions can be circumvented by precursors that carry a second carboxylate on the gamma position, thus forming an intramolecular anhydride (Scheme 1C).²³ However, these 1,3-dicarboxylate-based precursors drastically limit the scope of the reaction cycle to, for example, C-terminal aspartic acids and succinic acid derivatives.

Therefore, in this work, we developed a new reaction cycle based on the consumption of carbodiimides that yield a transient 5(4H)-oxazolone product (Scheme 1D). For brevity, oxazolones refer to 5(4H)-oxazolones in the following text. The reaction cycle is less prone to forming the unwanted N-acylurea due to a nearby nucleophile. Moreover, almost all carbodiimides activate the precursor into the transient product. Furthermore, the substrate scope of the reaction cycle is diverse, which we demonstrate by coupling it to several precursors, including peptides that assemble.

Results and Discussion

We used N-acetyl-l-phenylalanine (Ac-F-OH) as a model precursor (Figure 1A) to study the oxazolone-forming reaction cycle by high-performance liquid chromatography (HPLC). We compared the kinetics of the reaction cycle of N-acetyl-l-phenylalanine to phenylpropanoic acid because of its similar structure but lacking amide at the α position, which is required for oxazolone formation. Thus, we can compare the intermolecular anhydride-forming cycle (Scheme 1B) to the oxazolone-forming cycle (Scheme 1D). When 100 mM precursor reacted with 45 mM EDC in MES buffer (200 mM, pH 6.0), a transient peak was observed on the 254 nm channel of the analytical HPLC for both reaction cycles. For the cycle with phenylpropanoic acid, the transient peak was attributed to the intermolecular anhydride, which was confirmed by mass spectrometry and FT-IR-spectroscopy (Table S1 and Figures S2B, S3B).

Figure 1.

The robustness of the oxazolone reaction cycle with Ac-F-OH as a precursor. (A) Molecular structures of the species in the chemical reaction cycle. HPLC (markers) and kinetic model data (lines) as a function of time when fueled with EDC. All conditions are 100 mM precursor in 200 mM MES buffer. (B) With varying fuel concentrations at pH 6.0 at 21 °C, (C) at different pH values with 60 mM EDC at 21 °C, (D) at different temperatures with 60 mM EDC and pH 6.0. The error bars on the HPLC data represent the standard deviation of the mean value with n = 3.

In contrast, a transient peak emerged that corresponded to the oxazolone³⁹ for the Ac-F-OH as a precursor (Figure S2A). The oxazolone was confirmed by synthesizing it in an organic solvent⁴⁰ and characterizing it by HPLC, NMR, FT-IR, and mass spectrometry (Supporting Note 1). To follow the kinetics of the reactions, we used a quenching method,⁴¹ developed by our group, to measure the evolution of the concentration of the product, EDC, and the precursor as a function of time (Figure S1 and Figure S6). We used a kinetic model that predicts the concentrations of all species involved in the chemical reaction cycle every minute by solving a set of ordinary differential equations (Tables S5, S7, Supporting Note 2). The rate constants could be fit to accurately predict the reaction cycle’s evolution, and we could calculate the efficiency of the cycle, i.e., how much fuel is used to form the transient product of interest, i.e., the anhydride or oxazolone (Supporting Note 2). In the oxazolone-forming cycle, 91 ± 1% fuel was used to form the main product, while the value was 73 ± 1% in the intermolecular anhydride-forming cycle. Besides the product’s formation, the unwanted side product N-acylurea was formed in both cycles. Excitingly, only 2 ± 1% of the fuel is converted to the N-acylurea in the oxazolone-forming cycle, which starkly contrasts with the 26 ± 1% in the intermolecular anhydride-forming cycle (Figure S6A,B). Thus, the oxazolone formation avoids the formation of the unwanted side product N-acylurea to a large degree.

Next, we compared the kinetics of the oxazolone-forming reaction cycle to the intramolecular anhydride-forming cycle (Scheme 1C, Figure S6C,D). We used 100 mM Ac-D-OH as a precursor at pH 6.0 fueled with 30 mM EDC, and we measured the concentration of anhydride over time by using the same quenching method. In this reaction cycle, we found no N-acylurea and we used the kinetic model to calculate the efficiency of the cycle to be 99 ± 1%. Excitingly, the half-life of the transient intramolecular anhydride is 50 s, which is 85 times shorter than the oxazolone. In other words, we now have two reaction cycles that form transient products efficiently and have vastly different half-lives (Tables S6, S7).

A challenge in developing a reaction cycle is its robustness toward fuel concentration, pH, and temperature, which we tested for the oxazolone-forming cycle next. We measured the effect of the concentration of fuel added while the pH was fixed at 6.0 and the temperature at 21 °C. We fueled 100 mM Ac-F-OH in 200 mM MES with 30, 50, or 60 mM EDC (Figure 1B and Figure S7A). The oxazolone concentration increased at the expense of EDC in the first hour. It then decreased in the following 6 h, highlighting that the deactivation reaction of the cycle is relatively slow. Indeed, using the kinetic model, we established the half-life of the oxazolone to be 71 min under the applied conditions (Table S7). After the reaction cycle with 30, 50, or 60 mM EDC had been completed, there was roughly 1 mM of the N-acylurea (Figure S7B).

We tested the robustness of the reaction cycle toward pH (100 mM precursor with 60 mM EDC) by measuring the cycle at pH 5.0, 6.0, and 7.0 at 21 °C (Figure 1C and Figure S7C). At pH 6.0, the activation rate constant was 7.56 × 10^–4 mM^–1 min^–1, which increased by 3-fold when we worked at pH 5.0 due to the higher reactivity of EDC (Table S7). The increased activation rate made the overall reaction cycle much faster. In contrast, at pH 7.0, the activation reaction was only 5.95 × 10^–5 mM^–1 min^–1. Interestingly, the half-life of the oxazolone product stayed more or less unchanged in this pH range and ranged between 30 min at pH 7.0 and 71 min at pH 6.0. Moreover, under the conditions of pH 5.0 and 6.0, the N-acylurea formation is only of 1 ± 0.5% and 2 ± 1%, respectively, compared to 6 ± 1% at pH 7.0 (Figure S7D). Finally, the effect of temperature was determined at pH 6.0 (100 mM precursor with 60 mM EDC) by varying it from 21, 30, and 35 °C, respectively. The reaction cycle can reach a comparable efficiency of oxazolone for different temperatures (Figure 1D and Figure S7E,F). However, the half-life of the oxazolone decreased from 71 min at 21 °C to 28 min at 35 °C.

Next, we tested the versatility of the new reaction cycle by testing a range of carboxylate-based derivatives. In these experiments, we were interested in how much of the chemical fuel is used to form the wanted product oxazolone, i.e., the efficiency of the reaction cycle. We also determined the percentage of the added EDC used to form unwanted side product N-acylurea. The remainder is the EDC lost due to the direct hydration of EDC or the hydrolysis of the O-acylurea. The concentration of N-acylurea could be determined by HPLC directly. Due to the transient nature of the oxazolone, we used the kinetic model to calculate the percentage of the EDC used to form oxazolone (Supporting Note 2). The first precursors we tested were N-acetyl-protected amino acids (Ac-). We used 100 mM precursor with 30 mM EDC in MES buffer (200 mM, pH 6.0) at 21 °C. For Ac-F-OH, 2 ± 1% of the EDC was converted into the unwanted N-acylurea (Table 1), while for Ac-G-OH, Ac-A-OH, Ac-H-OH, Ac-M-OH, Ac-V-OH, Ac-S-OH, and Ac-R-OH, the formation of N-acylurea was too little to integrate on HPLC, and we assume the values are less than 1% (Table 1, Figure S8A–G). More than 90 ± 1% of the EDC was used to drive the activation reaction for Ac-G-OH, Ac-A-OH, Ac-H-OH, Ac-M-OH, and Ac-F-OH. In other words, the chemical fuel is used highly efficiently to form the main product. However, when Ac-R-OH was used as a precursor, only 53% of the fuel was used to form oxazolone. We hypothesize that the hydrophilic and cationic side chain enhances the hydrolysis of the O-acylurea intermediate (Table S8 and Supporting Notes 2). This reaction is reflected in the K value, which our kinetic model uses to describe the reactions of the O-acylurea; that is, K is the ratio between the hydrolysis to urea vs conversion to oxazolone. If K is far less than 1, oxazolone formation is more favored than the hydrolysis to urea. Among all amino acids in Table S8, only Ac-R-OH has a K value of 0.867, close to 1, which means the hydrolysis of the O-acylurea becomes competitive. Moreover, the hydrolysis of its oxazolone was also very fast. Therefore, the formation of oxazolone for Ac-R-OH was decreased.

Table 1. N-Acetylated Amino Acids as a Precursor in the Oxazolone-Forming Reaction Cycle^*.

^*Various precursors compared for their efficiency in forming oxazolone and N-acylurea. All yields are expressed as a fraction of the EDC added. We used our kinetic model for the oxazolone efficiency and assumed a calculation error of ±1%. For the N-acylurea efficiency, we used HPLC data. All conditions are in 200 mM MES at pH 6.0, 21 °C.

^aThe formation of N-acylurea was negligible, and we assume the values are less than 1%.

^b100 mM precursor with 30 mM EDC.

^c0.6 mM precursor with 0.6 mM EDC.

^d25 mM precursor with 12.5 mM EDC.

^e25 mM precursor with 12.5 mM EDC with 20 mM pyridine.

^fThis cycle produces the corresponding intramolecular anhydride.

In contrast, for Ac-I-OH and Ac-L-OH, i.e., amino acids with bulkier side groups, much more N-acylurea was generated (33 ± 1% and 9 ± 1%, respectively, Table 1 and Figure S8H,I). We hypothesize the difference is a result of the increased steric hindrance of the bulky groups on the α position. Nevertheless, the remainder of the EDC was used to create oxazolone, pointing to an energy-efficient reaction cycle for these precursors.

When the amino acid was protected with a 9-fluorenylmethoxycarbonyl (Fmoc) or N-(tert-butoxycarbonyl) (Boc), no oxazolone was found, and the reaction cycle converts more than 50 ± 1% of the EDC into N-acylurea (Table 2 and Figure S9). It is noteworthy that carbamates have lower pK_a’s than analogous amides.^42,43 Therefore, the alkoxycarbonyl protecting group decreases its nucleophilicity. Moreover, we suspect that these bulky protecting groups’ steric hindrance will affect its reactivity too. When we used di- or tripeptides with a glycine at the C-terminal position, the amount of N-acylurea generated increased with increasing peptide length from Ac-G-OH (<1%) to Ac-FG-OH (21 ± 1%) and Ac-F₂G-OH (Ac-FFG-OH, 27 ± 1%) (Table 2 and Figures S16A, S18A). This drawback can be overcome by using pyridine, which is known to suppress the formation of N-acylurea,⁴⁴ which we screened next. Indeed, adding 20 mM pyridine decreased the N-acylurea formation to 4 ± 1% of the EDC while keeping the reaction cycle efficient (Figure S16B). Meanwhile, pyridine speeds up the deactivation, resulting in a shorter half-life of the oxazolone (Figure S16C, Table S9).

Table 2. Precursors in the Oxazolone-Forming Reaction Cycle^a.

^aSee footnote of Table 1.

Based on these data, we describe the requirements for designing suitable precursors in the oxazolone-forming reaction cycle (Scheme 2). It has to possess a C-terminal carboxylic acid and no nearby nucleophiles other than an amide group on the α position for intramolecular heterocyclization. Racemization of the α position by oxazolone formation is unavoidable.^45,46 Therefore, an achiral amino acid (e.g., glycine) is preferred on the C terminus. Finally, bulky R¹ and R² groups and alkoxycarbonyl protecting groups will increase the amount of N-acylurea formed. If bulky protecting groups or larger peptides are required, pyridine can be used to maintain the high efficiency of this reaction cycle.

Scheme 2. Detailed Schematic Representation of the Oxazolone Formation and the Reaction Cycle and Proposed Guidelines for the Design of Precursors.

Next, we studied whether we could induce self-assembly with the new chemical reaction cycle. We designed a small library of peptides to form fibers based on Ac-F_nG-OH. We increased the product’s propensity to assemble by increasing the number of phenylalanines from n = 0 to 3 (Figure 2A). For Ac-G-OH and Ac-F₁G-OH, no evidence of self-assembly was found (Figure 2B,C, Figure S17A,B). In contrast, when we fueled a solution of 30 mM Ac-F₂G-OH and 15 mM pyridine with 80 mM EDC at pH 5.0, 21 °C, the turbidity increased rapidly, forming a hydrogel within 30 min (Figure 2B, Figure S17A). Confocal microscopy confirmed the presence of a fibrillar network (Figure 2C), and cryo-TEM showed the nanometer details of this fiber network (Figure 2D). The hydrogel lasted just over an hour, which was confirmed by oscillatory rheology (Figure 2E). That showed a rapid increase in the sample’s storage and loss modulus after applying 80 mM chemical fuel, and the storage modulus was always higher than the loss modulus, indicating that the material was in a gel state for 1.5 h.^27,32,47,48 The loss of the hydrogel coincides with the oxazolone concentration vanishing (Figure S18B). Less than 1% as a fraction of fuel was used to form the unwanted side product N-acylurea, suggesting that the transient oxazolone was responsible for the transient self-assembly. When only 50 mM EDC was added, no gel was formed, and no change in the loss and storage moduli was observed (Figure S17C).

Figure 2.

Using the oxazolone-forming reaction cycle to induce transient hydrogels. (A) Molecular design of fiber-forming chemically fueled peptides. All conditions are in 200 mM MES at pH 5.0, 21 °C. (B) Time-lapse photography of solutions of the various precursors in response to EDC. Conditions are 30 mM Ac-FG-OH, Ac-F₂G-OH with 80 mM EDC and 15 mM pyridine, and 4 mM Ac-F₃G-OH without EDC and pyridine. (C) Confocal microscopy images of Ac-FG-OH and Ac-F₂G-OH in response to EDC after 0.5 h and 4 mM Ac-F₃G-OH without EDC. All the scale bars correspond to 10 μm. (D) Cryo-TEM images of the solutions of Ac-F₂G-OH at 65 min after fueling with EDC and 4 mM Ac-F₃G-OH without fuel. All the scale bars correspond to 100 nm. (E) Storage and loss modulus as a function of time, measured by plate–plate rheology of 30 mM Ac-F₂G-OH with 80 mM EDC and 15 mM pyridine. Solid circles represent the storage modulus (G′); empty circles represent the loss modulus (G″). The error bars represent the standard deviation of the average (n = 3).

Finally, we found that a solution of 4 mM Ac-F₃G-OH formed a hydrogel before the application of fuel at pH 5.0 at 21 °C (Figure 2B), further confirmed by rheology (Figure S17C). Confocal microscopy confirmed the presence of a dense fibrillar network (Figure 2C). Cryo-TEM showed the nanometer details of this fiber network (Figure 2D). Taken together, our chemical reaction cycle can convert solutions into hydrogels at the expense of chemical energy.

Next, we explored whether our new reaction cycle could create chemically fueled transient emulsions of oil droplets in water, which have been powerful in delivering hydrophobic drugs with tunable kinetics.^14,49 Thus, we replaced the fiber-forming domain in the previous peptides (Ac-F_n) with an aliphatic domain. Above, we described that N-acetyl-G-OH (Ac-G-OH) does not assemble (Figure S17A,B). Thus, we started with N-butyl-G-OH (C₄-G-OH) and N-hexyl-G-OH (C₆-G-OH). In this series, the hydrophobicity of the precursor is increased with a simple aliphatic tail aiding the phase separation (Figure 3A). For C₄-G-OH, no evidence of phase separation was found (Figure 3B, C, and D). In contrast, when we fueled 200 mM C₆-G-OH with 100 mM EDC in the presence of 10 mM pyridine in 200 mM MES at pH 6.0, 21 °C, the turbidity increased rapidly and peaked at around 20 min while it completely vanished after 85 min (Figure 3B,D). Confocal microscopy confirmed the presence of oil droplets (Figure 3C). By combining the turbidity data with the kinetics of the kinetic model, we obtained the critical concentration for droplet formation of 25 mM of the C₆-G-OH oxazolone (Figure S19B). We found that the initial fuel concentration could tune the lifetime of the emulsion. With 50 mM EDC, no turbidity was observed, while the lifetime increased linearly against the fuel added (Figure 3E).

Figure 3.

Using the oxazolone-forming reaction cycle to induce transient droplet formation. (A) Molecular design of oil droplet-forming chemically fueled N-butyl amino acids. (B) Time-lapse photography of solutions of the various precursors in response to EDC. The condition is 200 mM C₄-G-OH, C₆-G-OH with 100 mM EDC, and 10 mM pyridine in a 200 mM MES buffer at pH 6.0 at 21 °C. (C) Confocal microscopy images of the samples described in B after 0.5 h. All the scale bars correspond to 2 μm. (D) Corresponding absorbance at 600 nm as a measure of turbidity. (E) The lifetime of the droplets in the cycle of 200 mM C₆-G-OH with 10 mM pyridine against initial fuel concentration. Error bars represent the standard deviation of the average (n = 3).

Finally, we explored the new reaction cycle to form amphiphile-based assemblies. We designed and synthesized precursors based on SO₃-C_n-G-OH (Supporting Note 3) inspired by the structure of dodecyl sulfate, a common anionic amphiphile. We assumed that the corresponding oxazolone of SO₃-C_n-G-OH should behave the same as dodecyl sulfate and self-assemble to form micelles. Surprisingly, no micelles formed in the cycle. Instead, we found needle-like crystals in response to EDC. For this class of building blocks, we also investigated a molecular library and increased the hydrophobicity by using a longer carbon chain (Figure 4A). For SO₃-C₁₂-G-OH, no crystals formed in the cycle fueled with EDC (Figure 4B). But we could use this derivative to determine the rate constants for our kinetic model in the absence of assemblies. Crystals emerged in the cycle using 30 mM SO₃-C₁₄-G-OH fueled with 30 mM EDC in 200 mM MES at pH 6.0, 21 °C, that remained for around 16 to 18 h, as evidenced by turbidity data (Figure 4B,D) combined with microscopy (Figure 4C). When crystals were formed, the oxazolone (Figure S20) was present much longer than the nonassembling SO₃-C₁₂-G-OH (Figure S21), which we explain by a self-protection mechanism.³⁸ Briefly, the crystals protect their oxazolone from hydrolysis because they exclude water. The hydrolysis thus only occurs on the fraction that remains in the solution. Using the kinetic model, we could determine that this concentration was roughly 5 mM of the oxazolone of SO₃-C₁₄-G-OH. Sharp peaks in a powder X-ray diffraction spectrum (p-XRD) confirmed that the needle-like precipitates in the cycle are crystalline (Figure 4E). Finally, for 15 mM SO₃-C₁₆-G-OH, needle-like crystals form without fuel, likely because the 16-carbon chain provides enough high hydrophobicity to induce self-assembly (Figure 4B,C and Figure S22).

Figure 4.

Using the oxazolone-forming reaction cycle to synthesize transient crystals. (A) Molecular library of crystal-forming chemically fueled precursors based on SO₃-C_n-G-OH. All conditions are in 200 mM MES at pH 6.0, 21 °C. (B) Time-lapse photography of solutions of the various precursors in response to EDC. Conditions are 30 mM SO₃-C₁₂-G-OH and SO₃-C₁₄-G-OH with 30 mM EDC and 15 mM SO₃-C₁₆-G-OH without EDC. (C) Microscopy images of 30 mM SO₃-C₁₄-G-OH in response to 30 mM EDC after 0.5 h and 15 mM SO₃-C₁₆-G-OH without EDC. All the scale bars correspond to 100 μm. (D) Turbidity as a function of time of 30 mM SO₃-C₁₄-G-OH in response to 30 mM EDC in 3 independent experiments. (E) Powder X-ray diffraction spectrum of crystals of oxazolone from SO₃-C₁₄-G-OH.

Conclusion

We introduced a new chemical reaction cycle fueled by a carbodiimide to drive 5(4H)-oxazolones formation. Unlike other reaction cycles, this one is robust and operates under different temperatures, pH, and fuel input and applies to a wide range of precursors. It uses carbodiimide-based fuel extremely efficiently and does not suffer from significant side reactions. Based on the requirements for the design of precursors, various oxazolones and self-assemblies can be obtained. Future work should explore its potential in forming different self-assemblies and developing new supramolecular materials.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/jacs.3c00273.

• Materials and methods description, turbidity data, rheology data, NMR data, FT-IR data, HPLC data, confocal data, mass spectrometry data, and p-XRD data (PDF)

German Federal Ministry of Education and Research (BMBF), Deutsche Forschungsgemeinschaft (411722921, EXC-2094-390783311, and the International Research Training Group ATUMS (IRTG 2022)), China scholarship council and European Research Council (ERC starting grant 852187).

The authors declare no competing financial interest.