Hierarchical structures in biology are driven by complex reaction pathways with corrective mechanisms to ensure their precision and function within the often-crowded living system. In cells, the number of reactive intermediates and partners engaged in the propagation of a single superstructure far exceeds that of synthetic supramolecular systems, often raising the question of whether such complexity is necessary from a structural perspective. The microtubule of the cytoskeleton, for example, is made up of simple tandem repeats of α- and β-tubulin proteins but involves chemical reactions by seven groups of post-translational modification1 and a proteome of >500 microtubule-associated proteins that collectively regulate its structure2,3. At first glance, the effort to build the microtubule seems to be excessive but recent findings in synthetic systems have found that supramolecular assembly processes are highly delicate4,5,6,7,8,9. Seemingly minimal (<0.01%) changes in liquid–liquid interfaces and composition and in fluid dynamics are as important in directing structure formation as well-established aspects such as concentration, temperature, molecular design and solvent polarity10,11,12. Given that the former set of factors are dynamic and heterogeneous within the cells, the magnitude and variety of control mechanisms necessary to guide structure formation in cells appear more reasonable.

Within the crowded cellular environment, the production of assembly precursors and their quantity are tightly regulated by interconnected reaction pathways. In this way, the reaction kinetics are coupled to the assembly landscape where the local chemical state and concentration influence the probability of metastable or kinetically trapped structures13. The importance of chemical reaction kinetics in controlling supramolecular assemblies has since been demonstrated in synthetic dissipative systems where advances in mimicking cellular homeostasis have showed recent success14,15,16,17,18,19,20,21,22,23. However, unlike synthetic analogues, biological precursors also modulate polymerization dynamics through structural plasticity, which reflects changes in the structural state without altering its chemical identity24. In microtubules, α-/β-tubulin at the polymerization terminal or the receding end are chemically identical yet they create opposite effects upon the global structure. Separate pathways then control the immediate environments surrounding the terminals where the cell can regulate the level of order/disorder within the superstructure. By creating additional nodes of complexity in reaction networks, biology gains access to another dimension of supramolecular features.

Inspired by this concept, we first establish a light-promoted molecular transformation cascade that organizes reversible and non-reversible covalent reactions to generate assembly precursors. The reaction kinetics will therefore dictate the availability of precursors over time while the reversible segment will modulate the lifetime of the precursors switching between active states. By controlling these parameters in the reaction cascade, we aim to demonstrate that assembly precursors can access different superstructures that are unavailable via conventional, direct supramolecular assembly.

As the first activator of the reaction cascade, we take the photoprotecting group (PPG) 2-nitroveratryloxycarbonyl (nvoc), for which the mechanism of photolysis and broad substrate capacity are well established25. The nvoc group absorbs ultraviolet light (365 nm) to form a zwitterionic excited state followed by an N=Onitro bond scission and the expulsion of CO2 (ref. 25). The resultant aromatic N=Onitroso group can oxidize thiols into disulfides via sequential addition and elimination steps26. By combining both reactions, we propose that thiol-containing molecules protected with the nvoc group would undergo a self-promoted transformation into disulfides in situ. As the thiol component, we design iso(Fmoc-I)nvocCA 1a, consisting of a masked cysteine connected via a thioester to form an isomerized backbone that temporarily blocks self-assembly. The nvoc group is installed at the N-terminus of Cys, thus regulating both the isomerization and self-promoted oxidation.

While the nvoc group undergoes a Norrish type II reaction to form the N=Onitroso group 5, the peptide performs an S,N-acyl rearrangement aligning the backbone and thus liberating the Cys-SH group, yielding Fmoc-ICA (3aλ, Fig. 1)25. The N=Onitroso by-product 5 promotes the main reaction sequence, oxidizing the free thiol of Fmoc-ICA 3aλ into its disulfide form DiFmoc-ICA 4aλ. We regulate the redox reversibility between 3aλ and 4aλ by introducing dithiothreitol (DTT) into the cascade, where the interconversion between both species is determined by the stoichiometry between N=Onitroso and DTT. The presence of both oxidation and reduction pathways therefore regulates the lifetime of 3aλ and 4aλ in solution, with the prevailing reaction defining the final species. To clarify the nomenclature, products formed within the cascade are denoted as 3aλ for Fmoc-ICA and 4aλ for DiFmoc-ICA, whereas the independently synthesized control peptides are termed 3a and 4a, respectively. Structure formation is compared by (1) direct assembly where precursors are diluted in a poor solvent, (2) cascade assembly where 1a is activated by light to eventually form the assemblies of 4aλ and (3) cycled cascade assembly where DTT is added to interconvert 3aλ and 4aλ (Fig. 1). Hence, the control peptides and their cascade counterparts possess identical chemical structures but differ in the way they are made available for assembly in solution.

Fig. 1: Photolytic cascade generation of molecular components and its effect on the formation of different supramolecular nanostructures.
figure 1

The molecular section describes a series of covalent reactions, the combined kinetics of which determine how the assembly precursor is produced in solution. By exposing the precursors to different combinations of reactions, the supramolecular pathway is altered, producing several morphologies despite having the same chemical identity. The reaction pathway of 1a after ultraviolet irradiation leads to the decaged isopeptide 2aλ, and rearrangement/linearization and oxidation of the thiol 3aλ to the disulfide 4aλ. The nvoc side product 5 enters the cycle as the oxidant to fuel the final step. The addition of DTT reverses the oxidation, causing 3aλ and 4aλ to rapidly interconvert, thereby delaying the production of 4aλ. Depending on the assembly pathway (direct, cascade, cycled cascade) of 4a/4aλ, different supramolecular nanostructures and chirality are produced.

Results and discussion

Peptide synthesis was carried out using Fmoc solid-phase peptide synthesis on Wang resin using N,N′-diisopropylcarbodiimide (DIC) and ethyl (2Z)-2-cyano-2-(hydroxyimino)acetate (OxymaPure) for the iterative coupling steps (Fig. 2). To construct an isomerized Cys backbone, Fmoc-cysteine-monomethoxytrityl (Fmoc-Cys(Mmt)-OH) was used to enable orthogonal installation of the nvoc group via 6-nitroveratryl chloroformate on the N-terminus and Steglich esterification at the S-terminus. The so-called isopeptide was cleaved off from the resin with trifluoroacetic acid (TFA):triisopropyl silane (TIPS):H2O (95:2.5:2.5) and purified by high-performance liquid chromatography (HPLC) to afford 1a. As reference compounds, Fmoc-ICA 3a and DiFmoc-ICA 4a were synthesized using modified protocols (Supplementary Information). The serine analogue, where the cysteine of 1a was substituted by serine to form 1b, was also synthesized to elucidate the role of the thiol group in the oxidation reaction.

Fig. 2: Solid-phase synthesis of the isopeptide derivatives.
figure 2

Synthesis of iso(Fmoc-I)nvocXA by solid-phase peptide synthesis: (1) piperidine; (2) Fmoc-X(protecting group)-OH, DIC, ethyl cyanohydroxyiminoacetate (oxyma); (3) 6-nitroveratryl chloroformate (nvoc-Cl), triethylamine; (4) dichloromethane, TIPS, TFA; (5) Fmoc-Ile-OH, DIC, 4-dimethylaminopyridine; (6) TFA, TIPS, H2O.

The molecular transformation of each step in the cascade was characterized by HPLC in methanol:NH4HCO3 buffer (1:1) at pH 7.4. This condition provides pH control and ensures that reaction intermediates and products remain in their molecular state and can therefore be quantified, using tryptophan as an internal standard. The optimum ultraviolet irradiation (365 nm) time of 180 s gave a deprotection conversion of 94% (Supplementary Fig. 20). Upon irradiation, the formation of each intermediate—the deprotected isopeptide 2aλ and the S,N-rearranged peptide 3aλ—and oxidation to the disulfide 4aλ was monitored over 24 h (Fig. 3b). Separately synthesized 3a and 4a were used as references for the retention times. Upon successful photodeprotection, the S,N-acyl shift to 3aλ began immediately and proceeded with a t1/2 (time required to attain 50% conversion) of 30 min. In comparison with the O,N-acyl shift27 from 2bλ to 3bλ with t1/2 > 6 h (Fig. 3c), the thiolate was demonstrated to be a better leaving group with an accelerated rearrangement kinetics. Although no further chemical changes were detected with the serine analogue 3bλ, the cysteine derivative 3aλ was further oxidized by the N=Onitroso into 4aλ with a t1/2 = 5 h (Fig. 3d). The control reference 3a remains stable in its reduced form for 24 h under ambient conditions, confirming that its oxidation into 4aλ was the result of the cascading step of the N=Onitroso by-product (Supplementary Fig. 25). We also studied the cascade reaction in an aprotic solvent, acetonitrile, in which the oxidation step of 3aλ was delayed by more than a factor of 4 (t1/2 > 24 h) (Supplementary Fig. 26).

Fig. 3: Kinetic profile of iso(Fmoc-I)nvocXA derivatives showing accelerated conversion for 1a and recycling pathway.
figure 3

a, Reaction pathway of the two isopeptides after irradiation. b, HPLC kinetics of iso(Fmoc-I)nvocCA 1a in methanol:NH4HCO3 buffer (1:1) over the 24 h after irradiation. After deprotection, iso(Fmoc-I)CA 2aλ rearranges into the linear peptide Fmoc-ICA 3aλ and oxidizes into the dimer DiFmoc-ICA 4aλ. c, HPLC kinetics of iso(Fmoc-I)nvocSA 1b in methanol:NH4HCO3 buffer (1:1) over the 72 h after irradiation. After deprotection, iso(Fmoc-I)SA 2bλ rearranges into the linear peptide Fmoc-ISA 3bλ. d, Concentrations of the intermediates and the product over time (1aλ) calculated from the calibration curves of 3a and 4a. The intermediate 2aλ was calibrated with 3a. e, HPLC kinetics of iso(Fmoc-I)nvocCA 1a 100 µM irradiated with DTT 10 mM in methanol:NH4HCO3 buffer (1:1). f, Molar proportion (χ) of 3aλ versus 4aλ after 6 h and 24 h for 1aλ, 1aλ + 10 equiv. of DTT and 1aλ + 10 equiv. of benzyl mercaptan. Data are presented as ±s.d., n = 3.

The mechanism was investigated in parts by decoupling the cascade, separating the N=Onitroso production step and the thiol oxidation. Therefore, nvoc-glycine 6 was synthesized and, via irradiation, found to produce the N=Onitroso intermediate 5 (retention time, RT = 12.27 min) with quantitative conversion (Supplementary Fig. 33). In contrast to most findings that observed the nitrosobenzaldehyde to be the cleavage product, we found intermediate 5 to be the carboxylic acid derivative that was reported to form in oxygenated aqueous conditions (Supplementary Fig. 34)28. The addition of separately synthesized 3a into the reaction successfully consumed the intermediate 5 to form 4a (Supplementary Fig. 35). This reaction pathway was further confirmed by using benzyl mercaptan (BzSH, RT = 17.01 min) as a substitute for 3a, which resulted in the formation of the oxidized BzS–SBz (RT = 21.70 min) (Supplementary Fig. 36). Each molecule of N=Onitroso 5, apart from side reactions, is capable of oxidizing up to 4 equiv. of thiols26.

From these findings, we postulate that the cascade from 1aλ to 4aλ can be modulated by interfering with the redox-sensitive species. As expected, complete inhibition of step 3aλ to 4aλ was observed by introduction of excess BzSH (10 equiv.), which reacted competitively with N=Onitroso 5 (Fig. 3f). Next, we further increase the complexity of the system by adding a selective reductant (DTT) into the cascade which reduces disulfide 4aλ back to thiol 3aλ. The coexistence of DTT and N=Onitroso 5 therefore forms a redox cycle, causing molecules of 3aλ and 4aλ to rapidly interconvert, with the prevailing stoichiometry and their kinetics determining the final redox state of the peptide at equilibrium. Hence, as long as DTT is present in solution, full oxidation by N=Onitroso 5 could not be achieved and would be observed as a net delay in the formation of 4aλ (Fig. 3e). Although DTT also possesses thiol groups, its intramolecular cyclization favoured an oxygen-promoted radical elimination from the N=Onitroso 5 core in the presence of water and oxygen29,30. The radical mechanism proposed would cycle the arylnitroso compound, which preserves the oxidation capacity of 5. This was confirmed by treating N=Onitroso 5 with excess DTT (10 equiv.) for 24 h before adding 1 equiv. of thiol 3a for a further 24 h incubation. The first 24 h of pretreatment, complete cyclization of DTT occurred, and in the following 24 h, complete oxidation of 3a to 4a showed the activity of N=Onitroso 5 (Fig. 4). In contrast, pretreatment with excess BzSH (10 equiv.) completely consumed N=Onitroso 5 in the first phase, and hence no subsequent oxidation of 3a was observed when added in the second phase (Fig. 4).

Fig. 4: Decoupling the redox process by pretreating N=Onitroso with different thiols.
figure 4

HPLC chromatograms of 6-nitroveratryl glycine (nvoc-Gly) 6 irradiated with 10 equiv. DTT (left) or BzSH (right) after 24 h. The intermediate N=Onitroso 5 is consumed by BzSH to produce BzS–SBz, but is not consumed by DTT/DTTox. Fmoc-ICA 3a (1 equiv.) was then added and after a further 24 h samples were analysed by HPLC again. Oxidation into 4a was observed for the DTT-pretreated reaction due to the presence of N=Onitroso 5. LC traces at 214 nm (black) and 350 nm (red) are shown for each step.

To understand the interplay of DTT and N=Onitroso in the redox cycle, we ran the full cascade over 24 h with varying amounts of DTT (5–100 equiv., 0.5 – 10 mM) and elucidated the reaction kinetics (Fig. 5a). By plotting the molar proportion of 3aλ/4aλ, we observed a time-dependent suppression of the formation of disulfide 4aλ because the reduction by DTT initially dominated the reaction pathway and rapidly reduced 4aλ as soon as this was formed (Fig. 5b). Eventually, at lower DTT ratios, oxidation to 4aλ dominated as DTT supply decreased over time. Consequently, by increasing the amounts of DTT and thus the reduction rate, the delay in the formation of 4aλ was prolonged as long as sufficient N=Onitroso 5 was present (Fig. 5b,c). For example, at 1 mM of DTT, synthesized 4a was reduced to 3a with a t1/2 = 2 h, whereas the DTT-free cascade oxidation into 4aλ by N=Onitroso 5 occured with t1/2 = 5 h (Fig. 5d). Comparing these reaction rates at 1 mM DTT, the reduction by DTT controls the initial phase of the redox cycle, where 4aλ was only transiently present due to its rapid reconversion to 3aλ. Therefore, steady production of 4aλ could only occur when the rate of oxidation by N=Onitroso 5 began to dominate after 6 h (Fig. 5b).

Fig. 5: Modulation of the cascade by DTT.
figure 5

a, Reaction pathway of 1aλ with DTT. N=Onitroso 5 and DTT form a redox cycle between 3aλ and 4aλ where they undergo multiple redox reactions. Denoted by DTT*, DTT also undergoes a separate, radical oxidation pathway with 5. b, Molar proportion of 3aλ versus 4aλ with different concentrations of DTT over 24 h. c, HPLC chromatograms of 1aλ with different concentrations of DTT. The peak of 5 is shown in the 350 nm trace. d, Molar proportion of 3aλ versus 4aλ for the reduction of 50 µM 4a with 1 mM DTT and the cascade of 100 µM 1aλ with 1 mM DTT over time. Note that 50 µM disulfide 4a is equivalent to 100 µM of thiol 1aλ.

Next, we studied the impact of the cascade pathway and redox cycling on structure formation through self-assembly at a standardized condition of methanol:NH4HCO3 buffer (1: 9) at pH 7.4. Reducing the organic solvent content to 1:9 (from 1:1 used in the reaction cycle in Fig. 3) was necessary to promote self-assembly. To confirm that the cascade proceeded efficiently under reduced amounts of methanol, we performed a reverse analysis (by redissolving the assemblies in 50% methanol) and we showed that the conversion of 1aλ to 4aλ was complete at 24 h (Supplementary Fig 24). The critical assembly concentration (CAC) was determined by Proteostat assay, a general indicative fluorescence probe for peptide nanostructures (Supplementary Figs. 4446)31. The control peptides 3a and 4a showed a CAC of 140 µM and 50 µM, respectively, whereas 4aλ showed a CAC of 30 µM. The higher fluorescence intensity produced by 4a suggested that its superstructures contain higher molecular order than those of 3a. By considering the findings of the CAC, we henceforth standardized the monomer concentrations to 200 µM, 24 h incubation time at room temperature for all subsequent structural studies to ensure that the formed superstructures are at equilibrium. Temperature-dependent 1H NMR spectroscopy studies revealed that the assemblies of 3a and 4a (Supplementary Figs. 47 and 48) remained stable and did not depolymerize within 20–60 °C, taking into consideration the boiling point of methanol.

Circular dichroism (CD) spectroscopy analysis of 3a and 4a revealed major differences in the secondary structures of both assembled nanostructures. The control peptide 3a showed a positive ellipticity at 187 nm and a negative Cotton effect at 205 nm attributed to the n → π* transitions of the peptide backbone typical of a distorted α-helical structure (Fig. 6b)32. Assemblies of 4a exhibit different secondary structural features that are indicative of higher order. Positive ellipticity at 191 nm and negative ellipticities at 208 nm and 222 nm indicated an atypical twisted β-sheet structure, corroborating the increased fluorescence detected by the Proteostat assay (Fig. 6b)32. Additionally, the characteristic π → π* transition of the Fmoc group at 255, 274 and 294 nm could only be found in the corresponding spectra of 4a, indicating that the Fmoc group played a critical role in the emergence of supramolecular chirality during the assembly process. Having elucidated the supramolecular assembly characteristics by the direct assembly of 3a and 4a, we then studied how the reaction cascade would guide the assembly landscape. The reaction cascade to form monomer 4aλ affords self-assembled superstructures with combined secondary signatures of 3a in the region <240 nm, and Fmoc interactions of 4a in the region >240 nm. The possibility that the cascade did not drive to completion was excluded because post-assembly analyses by HPLC confirmed the total conversion of 4aλ and the absence of 3aλ (Supplementary Fig. 24). It is intriguing that 4a and 4aλ, although molecularly identical, adopted different secondary structures within the same solvent. When the cascade was performed with redox cycling through DTT, the final superstructures revealed the emergence of a new positive ellipticity at 222 nm and an increase in the intensity of the Fmoc contribution at >240 nm (Fig. 7c). These observations increased in magnitude with increasing amounts of DTT.

Fig. 6: Influence of assembly pathways (direct, cascade, cycled cascade) on structure formation, chirality and long-range order.
figure 6

a, Direct assembly of 3a and 4a. b, CD spectra of 3a and 4a indicating distorted α-helical and atypical twisted β-sheet structures, respectively. c, Schematic of the cascade and cycled cascade assembly.

Fig. 7: Influence of assembly pathways (direct, cascade, cycled cascade) on structure formation, chirality and long-range order.
figure 7

a, TEM image of 3a revealing a network consisting of thin (9 ± 2 nm) fibres. b, TEM image of 4a revealing twisted fibres with defined lengths (480 ± 170 nm) and periodicity (69 ± 9 nm). c, CD spectra of 4aλ with different equivalents of DTT indicating the emergence of secondary chirality at 222 nm and overall order as a function of redox cycling. DTT in its reduced and oxidized form (2 mM) as controls showed no contribution to the CD signature. d, TEM image of 4aλ showing fragmented fibres with average lengths of 260 ± 150 nm. e, TEM image of 4aλ with 10 equiv. DTT revealing the emergence of longer fibres (920 ± 420 nm). Scale bars in a,b,d,e, 250 nm. f, Thickness and length of the individual peptide fibres of monomers 3a, 3aλ, 4a and 4aλ obtained from TEM. Data are presented as mean ± s.d., n = 14. Statistical significance was calculated by analysis of variance with a Tukey post-hoc test: *P < 0.05, **P < 0.01, ***P < 0.001.

Transmission electron microscopy (TEM) studies were conducted under the same conditions, that is, methanol:NH4HCO3 buffer (1: 9) at pH 7.4 (Fig. 6), to investigate the morphologies of the formed supramolecular structures. TEM samples from the cascade were prepared, in situ, from the reaction mixture after 24 h, similar to the CD experiments. Each variation (for example, by DTT) was cross-checked via post-assembly HPLC analyses to ensure complete conversion to 4aλ. For the control peptides, direct assembly of 3a resulted in the formation of a network consisting of thin (9 ± 2 nm) peptide fibres (Fig. 7a), whereas 4a formed short and unusually defined fibres (Fig. 7b). Statistical analyses revealed that 4a appears to have a critical length (480 ± 170 nm) with a twist periodicity of 69 ± 9 nm (Fig. 7f). Compared to 4aλ produced by the cascade, the resulting morphologies were largely fragmented fibres with a strong tendency to form localized clusters among few mature fibres (Fig. 7d). Without the irradiation step to initiate the reaction cycle, no assemblies were observed (Supplementary Fig 49). With redox cycling via DTT (10 equiv.), the delayed production of 4aλ showed a notable growth (>3-fold) in the length of the fibres (920 ± 420 nm) (Fig. 7e). In addition to the bulk analyses from CD, these TEM studies demonstrated that reaction pathways for the in situ generation of supramolecular assemblies can be used to modulate long-range structural order.


We have developed a reaction cascade that highlights the trademark structural plasticity in nature that enables different hierarchical states to be formed by one type of chemical identity. By merging reaction dynamics with a modulated global redox environment, a disulfide assembly precursor (4aλ) has been found to exhibit different nanostructural orders, lengths and chirality. The disulfide bond is important in coordinating the aromatic contribution of the monomer towards structure propagation and the overall secondary structure of fibrils. Based on this principle, redox cycling between the thiol (3aλ) and disulfide (4aλ) results in the emergence of unique chiral signatures, ultimately forming different assembly states of 4aλ. The presented approach encompasses a number of strategies adopted in nature to regulate hierarchical structures through controlled and reversible production of assembly intermediates. We believe that this integrative chemical concept provides a fresh perspective on the bottom-up synthesis of nanostructures and expands the repertoire of dynamic supramolecular architectures that can be formed under mild conditions.


Light-induced S,N-acyl shift/O,N-acyl shift (HPLC)

The isopeptides were dissolved in 300 µl methanol (200 µM) and combined with 300 µl of an aqueous NH4HCO3 solution (5 mM, pH 7.4) to yield a 100 µM peptide solution. The peptide solutions were irradiated with ultraviolet light (365 nm) for 3 min while stirring at 300 r.p.m. and tryptophan (10 µM) was added as an internal standard. Aliquots of 10 µl were analysed at certain time points in an HPLC set-up. The kinetics were monitored for 24 h (1a) and 72 h (1b).

CD spectroscopy

For the measurement of the irradiated iso(Fmoc-I)nvocCA, the peptide was dissolved in 60 µl methanol (2 mM) and combined with 540 µl of an aqueous NH4HCO3 solution (5 mM, pH 7.4) to yield a 200 µM peptide solution. The solution was then irradiated with ultraviolet light (365 nm) for 10 min and incubated for 24 h. CD spectra were recorded at wavelengths from 300 to 185 nm with a bandwidth of 1 nm, data pitch of 0.2 nm and a scanning speed at 20 nm min−1 at 20 °C.


The isopeptide was dissolved in 60 µl methanol at a concentration of 2 mM and diluted with 540 µl NH4HCO3 buffer (5 mM, pH 7.4) to yield 100 µl solutions (200 µM). Then, the peptide was irradiated for 10 min (365 nm) and 100 µl of this solution incubated for 24 h. TEM grids were prepared by pipetting 3 µl solution onto a Formvar-coated copper grid and incubating these for 5 min. After the incubation, the solutions were removed with filter paper, and the grids were stained with 7 µl 4% uranyl acetate solution for 5 min. The grids were washed three times with MilliQ water and dried before being measured.