Towards our goal to access a minimal chemical-based system that can show oscillation, a self-assembly process was designed to install both negative and positive feedback loops with suitable time delays under non-equilibrium conditions. Firstly, if the assembly is catalytic and can promote its own degradation, then the negative feedback could be installed (Fig. 1a). Further, if one of the products of the catalytic degradation or waste can facilitate the regeneration of the assembly, the time delayed positive feedback loop will be installed (Fig. 1b). This coupling of negative and positive feedback loops might yield an oscillatory system (Fig. 1c). With these in mind, the building block ABC was rationally designed (Fig. 1d). The building block featured a kinetically stable but thermodynamically activated hydrophobic ester (BC, Fig. 1d). This aromatic ester had free aldehyde group that could couple via a dynamic imine bond with the amine group present in the polar residue (A) to access the amphiphilic building block ABC which may have the propensity to self-assemble (Fig. 1d). For the polar residue A, molecular histidine was used (Fig. 1d). Free histidine usually has poor hydrolytic roles yet they show an enhancement of their catalytic proficiencies when they are featured in supramolecular assemblies via multivalency, co-operative effects and so forth3,23. Finally, the dynamic imine bond was selected as this linkage is known for its acute responsiveness towards pH switches which was used to generate the second feedback (positive feedback). Hence, an ester was synthesized via the condensation of organic acid 4-formylbenzoic acid (B, Fig. 1d) and 4-nitrophenol (C), which was used as the hydrophobic part and upon catalytic hydrolysis would create a pH gradient to provide positive feedback towards the imine bond formation.
Temporal gelation at pH 7.5 buffer (one cycle)
We started with a simplified system where only one feedback (hydrolytic activity) would be operational while the other would be inoperative. Hence, the feedback that could generate from pH changes was deactivated by using a buffer system. BC was mixed with A at different molar concentrations with fixed molar ratio of 1:1 (DMSO/buffer solution, 60% v/v, pH 7.5). The samples with molar concentrations of 50 mM each (A and BC) gradually became viscous, followed by the transformation into self-supporting gel at around 2h (Fig. 2b). The gel showed weakening with time and manifested a gel-to-sol transition at 5h (Fig. 2b).
The transient gels were monitored by visible light scattering spectroscopy (λ = 700 nm) to measure the turbidity of the system. The rapid increase in turbidity was observed till 2.5h which exactly matched with the gelation time, followed by the gradual decrease that finally plateaued around ca. 11h (Fig. 2e). To gain more insight about the temporal gelation, transmission electron microscopy (TEM) was performed at different time intervals. TEM micrographs demonstrated the temporal generation of self-assembled fibrillar structures was responsible for gelation. At time ~0h, when the sample was in sol state, sparsely populated fibrillar structures could be seen (Fig. 2c). However, at ca. 2h (gel state), the network like morphology was seen (Fig. 2c). After ca. 5h, when the sample turned to sol, very few fibrillar structures could be seen (Fig. 2c). Further, similar morphological transition was witnessed in scanning electron microscopy (SEM, Supplementary Fig. S1). Time-dependent rheology was performed to investigate the system in more detail on the macroscopic level. The gradual increase in storage modulus followed by the decrease (after 2h) in storage modulus, confirmed the temporal change in material property of the system (Supplementary Fig. S2 and S3).
At this point, to investigate the self-assembly and autonomous disassembly process, both the imine and ester bonds were monitored by different analytical techniques. To probe the formation of the imine bond, time dependent nuclear magnetic resonance (NMR) spectroscopy was performed. 1H NMR of the sample containing A and BC (see Supplementary Information for more details) manifested the generation of imine with the proton peak at δ 8.41 ppm, which intensified till ca. 2h. The imine condensation even in a largely aqueous medium (~ 40% water) suggested the role of the self-assembly process, as observed in previous reports24. Notably, the intensity of the peak started to decline after 2-3h and suggested hydrolysis of the imine bond (Supplementary Fig. S4). The time gap in the increase and decrease of the imine peak intensity broadly corroborated with the lifetime of the gel. This suggested that the presence of the imine building block ABC is responsible for gelation and subsequent hydrolysis of the Schiff base leads to disassembly. Controls done without aldehyde group did not lead to gelation to suggest the importance of the imine building block (Supplementary Fig. S5). Further quantification of the extent of imine formation was not possible due to peak broadening that resulted from low mobility of the molecules at the gel state.
To probe the coupling of the transient gelation with the temporal formation of imine, the hydrolysis of the ester (BC) was monitored via time resolved HPLC. The rate of formation of C was monitored at λ= 318 nm (Fig. 2f). Notably, the rate profile showed an initial burst hydrolysis of the ester in the first 2.5h with catalytic rate 1.5 mM/h. This rise was followed by a gradual decline of the catalytic rates and rate plateaued after 5h. The initial burst suggested the gel state was catalytic while the sol state was unable to hydrolyse the remaining ester regardless of the presence of large amount of histidine. This suggested that the self-assembly process was coupled with the catalytic hydrolysis and subsequently was responsible for its own degradation, thus installing a negative feedback loop. It is important to note here that the product B also features the aldehyde functionality which can couple with the histidine via the imine bond. However, from NMR, no imine bond could be seen to be formed after 4h (Supplementary Fig. S4). Control experiments done with mixture of C with A and B did not show the generation of any imine proton peak from NMR. The mixture of A and B also did not access any gel or network like morphology. These results suggested that the self-assembling capability of the product (AB or ABC) is vital for the imine bond formation.
Temporal gelation at pH 6.5 buffer (one cycle)
As noted, the above experiments were done at pH 7.5. We were interested to investigate the system at a pH where the imine formation will be facilitated in the medium. For this, the pH of the medium was chosen to be 6.5 (DMSO/buffer, 0.1 M, 60% v/v) as low pH facilitates Schiff base formation via catalytic elimination of water25. At this pH, mixture of A and BC accessed a self-supporting gel within 0.5h, which was significantly faster than the gelation time required at pH 7.5 (Figs. 2d, e). The gel at pH 6.5 was stable for a longer time (lifetime of 4h compared to 2.5h in case of pH 7.5, Fig. 2g). Further, rapid rise in the turbidity in case of pH 6.5 compared to that observed for pH 7.5 supported the faster gelation in acidic pH (Fig. 2e). The prolonged lifetime was also confirmed from the turbidity measurement at pH 6.5. These observations suggested the rapid imine formation at acidic pH (6.5) facilitated greater extent of building block (ABC) generation that in turn led to longer lifetime of the self-assembled state. The greater extent of self-assembled networks was also supported by rheology measurements which showed improved mechanical strength of the gel (Fig. 2g, Supplementary Fig. S6 and S7).
The rate of hydrolysis at pH 6.5 was monitored by HPLC. The formation of C was seen to be accelerated significantly as it showed a burst release (for the first 0-3h, 4.4mM/h, Fig. 2e). Subsequently, the rate showed decline (for 3-5h, 1.7mM/h) before plateauing after 5h. The burst phase once again overlapped with the lifetime of the gel and reinforced the importance of the self-assembled catalytic microenvironment22. The rate of hydrolysis at pH 6.5 was almost three-fold higher than that was observed at pH 7.5 (Fig. 2f, inset). Higher hydrolytic rates at lower pH again supported the importance of the self-assemblies. The overall consumption of the ester was also significantly higher at the acidic pH.
Feedback from pH variation allows regelation (two cycles)
We noted that the remarkable differences in the lifetimes of gels, the extent of self-assembly and the subsequent negative feedback from hydrolysis was purely decided by the environmental pH (Figs.2e-g). We envisaged that if the system can itself register a gradual pH drop which is synchronized with the time delays in catalysis, it might be possible to realize an oscillatory gel; an unprecedented observation in literature thus far (Figs. 1c and 3a). The slowly forming waste (B) generated via the hydrolytic cleavage of BC can be the proton source to independently decrease the environmental pH that can consequently facilitate the regeneration of ABC utilizing the remaining esters in the medium to restart the cycle (Fig. 3a). Hence, unbuffered medium (Milli Q water) was used to achieve the pH switch within the medium. The mixture of A and BC (50: 50mM, Sol1) in 60% v/v DMSO/water showed an initial pH of 7.35 due to the presence of A. To monitor whether the proposed production of B was able to increase the acidity of the system, the pH was measured with time. Indeed, the pH of the system gradually declined from 7.35 to 6.1 in a span of ca. 48h (Fig. 3d). The visible light scattering measurement of the unbuffered system showed increase in turbidity till 2h that was followed by an expected decline. Remarkably, a distinct increase of turbidity was observed after 15h, which again peaked at 24h followed by the substantial decline in intensity (Fig. 3e). This observation suggested that the phenomenon of the creation of catalytic microenvironment was registered twice as a function of time with a single batch of reactant addition (closed system). Importantly, we could observe the sol-gel-sol-gel-sol transition (Fig. 3b). Starting from the initial sol state (Sol1), a self-supporting gel (Gel1) was accessed at around 1.5-2h which converted into sol (Sol2) approximately within 5h (Fig. 3b). The regelation was observed at ca. 23-24h (Gel2, Fig. 3b). This was followed by the formation of viscous solution (Sol3,) in 60% of the samples while the rest 40% remained as weak gel after ca. 48h (done in 10 different sets of vials under identical experimental conditions, Fig. 3b).
Moreover, two cycles of assembly and disassembly were also suggested from time-dependent TEM investigations. Distinct network like morphologies was seen to be accessed at ca. 2h and 24h respectively, while in other time periods, significantly lesser extent of networks was observed (Fig. 3c). Time-resolved rheology could also register the temporal generation and regeneration of higher mechanical strength of the Gel1 and Gel2 in the 1st and 2nd cycle, respectively. The higher storage modulus (G') value of the Gel1 (G' = 5567 Pa) and Gel2 (G' = 4092 Pa) states compared to Sol1 (G' = 709 Pa), Sol2 (G' = 1572 Pa) and Sol3 (G' = 295 Pa) underpinned the autonomous development of self-assembled structures at Gel1 and Gel2 states that resulted in gelation (Fig. 4a). To gain more details about the self-assembly process, time dependent fluorescence spectroscopy was performed. DPH was used as a dye, a well-known fluorescent reporter which shows augmented intensity when bound to hydrophobic microenvironment26. Expectedly, in the 1st cycle (Sol1-Gel1-Sol2), the gradual increase in fluorescence intensity was observed which peaked at Gel1 state around 2h, then started to decrease at 3h (Fig. 4b). Around 24h re-intensification of fluorescence intensity was observed followed by decrease with ageing (Fig. 4b). The real time pH change in the macroscopic gel was monitored by using Methyl Red indicator that was added to the system from t~0h. The temporal change in colour of the medium from yellow to orange demonstrated the autonomous reduction of the pH within the system (Fig. 4c).
Further, time dependent NMR was performed with the unbuffered system. As expected, the generation of imine proton peak was observed at δ 8.41 ppm which gradually intensified at ca. 1h (Gel1) followed by the decline in peak intensity (Fig. 4d). The peak registered a resurgence in intensity at around 20-24h (Gel2, Fig. 4d). Subsequently, the imine proton peak started to diminish after 24h and plateaued around 48h. This result supported that indeed the waste B acted as positive feedback to regenerate the imine within the unbuffered system and consequently regelation was observed.
Extensive time-dependent HPLC was performed to monitor the release of C with respect to time and the hydrolytic activity was probed. As hypothesized, the positive feedback from pH which resulted in the re-formation of imine and subsequent regelation was expected to play a critical role in reviving the hydrolytic capability. Indeed, after the plateauing of rates (first cycle, 9-15h), an increase in the production rate of C could be detected again, at the similar timeline where regelation and re-intensification of imine were observed (Fig. 4e, orange band, 19-24h).
After confirming the temporal gelation (Gel1) followed by regelation (Gel2), the concentration distribution of the substrate as well as the products that were presented and produced within the medium were quantified from HPLC (Fig. 4e). Initially, at high pH the higher concentration of the BC drove the gelation thus overcoming the low imine conversion rate. Whereas in the case of Gel2, the low pH of the system helps in the rapid imine formation, thus the gel was formed even at low concentrations of BC. With further decrease in pH no gel was formed because of insufficient gelation concentration at pH 6.1. Efforts are underway to increase the number of cycles of assembly and disassembly by varying the concentrations and molecular components.