Autocatalytic and oscillatory reaction networks that form guanidines and products of their cyclization

Autocatalytic and oscillatory networks of organic reactions are important for designing life-inspired materials and for better understanding the emergence of life on Earth; however, the diversity of the chemistries of these reactions is limited. In this work, we present the thiol-assisted formation of guanidines, which has a mechanism analogous to that of native chemical ligation. Using this reaction, we designed autocatalytic and oscillatory reaction networks that form substituted guanidines from thiouronium salts. The thiouronium salt-based oscillator show good stability of oscillations within a broad range of experimental conditions. By using nitrile-containing starting materials, we constructed an oscillator where the concentration of a bicyclic derivative of dihydropyrimidine oscillates. Moreover, the mixed thioester and thiouronium salt-based oscillator show unique responsiveness to chemical cues. The reactions developed in this work expand our toolbox for designing out-of-equilibrium chemical systems and link autocatalytic and oscillatory chemistry to the synthesis of guanidinium derivatives and the products of their transformations including analogs of nucleobases.


Materials and Methods
Thiourea, sodium 2-bromoethanesulfonate, methyl chloroformate, potassium thiocyanate, piperidine, bromoethane, 3-bromopropane-1amine hydrobromide, ethanethiol, sodium thiosulfate, iodine, potassium Chemical shifts for 1 H and 13 C are given in ppm relative to TMS. 1 H and 13 C spectra were calibrated using a residual solvent peak as an internal reference DMSO-d 6 ( 1 H NMR: δ = 2.50 ppm, 13 C NMR: δ = 39.52 ppm), D2O ( 1 H NMR: δ = 4.79 ppm). Data for the 1 H NMR spectra were reported as follows: chemical shift (ppm), The Chromatographic separation and mass analysis were performed on a Waters Acquity liquid chromatography system equipped with a PDA Detector (210 and 700 nm) and a Waters QDa mass detector with an electrospray ionization (ESI) and a mass range of 85-1250 m/z.

3-aminopropane-1-thiol (9)
Disulfide of 3-aminopropane-1-thiol from the previous experiment (500 mg, 2.7 mmole) in the form of a free base was dissolved in 15 ml of Et2O and mixed with 1 ml of EtSH (13 mmole). The vial with the reaction mixture was kept in a refrigerator for 24 h. Next, 3-aminopropane-1-thiol formed as flat hexagonal crystals on the walls of the vial. Next, the crystals were washed with diethyl ether and dried under vacuo.

N-(4-bromobutyl)phtalimide
This compound was synthesized using a modified literature procedure. 5 Potassium phtalimide (1.86 g, 10 mmole) was added to the solution of 1,4-dibromobutane (30 mmole, 3.9 mL) in DMF (10 mL) at room temperature. The reaction mixture was stirred at 80 o C for 24 hours. Then, it was filtered and washed with 2x20 mL of DMF. The filtrate was concentrated in vacuo and poured into an ice-water bath. The white solid, which precipitated from the water, was recrystallized from methanol to afford the desired product. Yield: 2.25 g, 80%.
The reaction mixture was stirred overnight and then was quenched with water (10 mL) and extracted with DCM. The organic phase was dried with Na2SO4 and evaporated in vacuo. The product was purified by gradient column chromatography (Hex/EtOAc, 0-25%). Yield: 1.5 g (43%

General protocol for the 1 H NMR kinetics experiments
The kinetics of the interaction between thiouronium salts and cysteamine (1) or its homologues (9 and 10) were monitored by NMR. Usually, we ran the experiment with a concentration of both compounds around 50 mM in phosphate buffer solution (1 M and pH 7.5 or 8). Thiouronium salt 8 exhibited very high activity; monitoring of the reaction progress in experiments where the starting concentrations of reactants were 50 mM was not possible by NMR. For measuring the kL value for 8, we used 1 mM solutions. All measurements were carried out at room temperature, and changes in concentrations of the reactants and products were monitored by integration of the characteristic signals in 1 H NMR spectra.

Fitting the rate constants
The model of direct ligation was used to determine the rate constants ( ) of the reactions. Best-fitted parameters were obtained by global fitting the data from the kinetic measurements to the model. The model was input to the software COPASI and the fittings were performed using the build-in function of "Parameter Estimation". The applied algorithm for fitting was Evolutionary-Programming (number of generations, 2000; population size, 30). 9 The fitting results are shown in Fig. 1 to 6. The imperfection of the fitting might be partially associated with the catalytic effect of 2-mercaptoethylguanidine (11), which is probably a better leaving group than MESNA thiol or ethanethiol. The catalytic effect of thiols that are good leaving groups is a well-known phenomenon in native chemical ligation, 10 and it should be expected in the thiol-assisted formation of guanidines.
Model: Direct ligation between thiouronium salts and the homologues of cysteamine with a rate of .
The dynamics of the model is described by the following differential equations (1) to (4): Supplementary Figure 1. Global fitting of the reaction between 3 (50 mM) and 1 (50 mM) in phosphate buffer pH 8. The best value of (4.77×10 -3 min -1 mM -1 ) was obtained after parameter fitting. pH 8. The best value of (1.47×10 -3 min -1 mM -1 ) was obtained after parameter fitting.
Supplementary Figure 3. Kinetic data of the reaction between 3 (50 mM) and 10 (50 mM) in phosphate buffer pH 8. The reaction did not proceed during the 500-minute period.
Supplementary Figure 5. Global fitting of the reaction between 5 (50 mM) and 1 (45 mM) in phosphate buffer pH 8. The best value of (0.111×10 -3 min -1 mM -1 ) was obtained after parameter fitting. pH 8. The best value of (0.246 min -1 mM -1 ) was obtained after parameter fitting.   Hence, we ensured that peak m/z=157 belongs to the bicyclic amide 24. To get an additional evidence that the compound with m/z=156 is bicyclic amidine 22, we synthesized 3-(2-iminothiazolidin-3-yl)propanenitrile. This nitrile after being dissolved in potassium phosphate buffer pH=8.00 (reaction 2). immediately decomposed, producing some amount of a compound whose spectrum resembles the one from bicylic amide 24. The results of these studies can be summarized in three conclusions: 1) One of the products of the reaction 1 is bicyclic amide 24. It can be separated from the reaction mixture using silica column chromatography. Its spectrum showed complete correspondence with 24 that we synthetized by the alternative procedure.

Supplementary
2) Product with m/z=156 is bicyclic amidine 22. Its 1 H NMR spectrum matched with the amidine 22 obtained from the reaction 3 as reference. This product, however, is unstable during workup and converts to the bicyclic amide 24.
3) In buffer, thiol 20 undergoes spontaneous cyclization into amidine 22, which then leads to the production of amide 24.
Those conclusions can be summarized into the following mechanism: According to our prior knowledge, the ligation of thiouronium salts with (H2N-(CH2)3-S)2 is much slower than one with cystamine (see Figure 2, main text). Here this observation is also supported, since guanidine formation is slower than in the reaction 1. No cyclization product was observed in the kinetic experiment as expected from the slower intramolecular cyclization of 3-mercaptopropylguanidines than of 2-mercaptoethylguanidines.
However, increasing concentrations of reactants and the reaction time indeed has led to the production of bicyclic products. Based on these data, we conclude that 21 first cyclizes into bicyclic amidine 23 which then hydrolyses to bicyclic amide. This finding let us to assume that the reaction 4 proceeds via the analogous mechanism to the reaction 1.

Flow set-up
The flow set-up consisted of four main components: (i) syringe pumps, (ii) a micro continuously stirred tank reactor (CSTR), (iii) a microfluidic mixer (or two mixers), and (iv) a flow cell. We used NEMESYS Low Pressure Syringe pumps from CETONI in all our experiments. The construction of the micro-CSTR, microfluidic mixers and the flow cell is described in detail below.

Micro-CSTR.
The reactor consisted of the main glass body, a glass bottom, rubber connectors to tubing, and a stirring bar. The glass body was manufactured by Witeg Co. with a custom design. The bottom consists of ~0.2 mm cover glass that was glued to the main body with silicon glue after the stirring bar (4/1 mm) had been inserted into the main body. The tubing connectors were made from rubber plungers from 1 ml plastic syringes by punching holes for tubing in the middle of them. To control the temperature inside the reactor, it was installed on a copper plate that had a water circulation system connected to a thermostat (Fig. 28). The copper plate was equipped with a clamping system and thermal paste (CPU grease) was used for better heat transfer.
Importantly, the CSTR outlet tubing was made of polyethylene with an inner diameter of 0.4 mm to reduce the transfer time from CSTR to the mixer.
Supplementary Figure 28. Continuous stirred-tank reactor installed on the thermostated copper plate.

Microfluidic mixers.
We made two types of microfluidic mixers: (i) with two inlets (required for all the experiments); (ii) with three inlets (required only for experiments studying the response to chemical stimuli).
The fabrication protocol for the device with three inlets is the same and it will be described for the device with two inlets. First, three hypodermic needles were cut to ~1 cm length. The needles were placed in a 5 cm plastic petri dish in such a way that the ends of two of them touch each other and the end of the third needle is ~1 cm from the junction. Next, the petri dish was filled with a mixture of Sylgard® 184 silicone elastomer with 9 wt% of the curing agent. The mixture was degassed under reduced pressure and cured for 2 hours at 65 °C to form a polydimethylsiloxane (PDMS) elastomer. After curing, the needles were removed and PDMS was removed from the petri dish. The channel between the junction and the third inlet was manually scratched, and the device was plasma bonded to a microscope glass slide (Fig. 29).

Supplementary Figure 29. Static mixer made of PDMS.
Flow cell. Initially we used a commercially available Helma flow cell with a 0.2 mm light pass; however, it produced unsatisfactory results, most likely because of the non-homogeneous flow in the channel with a high aspect ratio (0.2/4 mm) and at very low flow rates, which characterizes our experiments. In addition, rectangular cells (similar to the ones we used previously) 11 are prone to the accumulation of gas bubbles.
Therefore, we decided to build a flow cell with a round cross-section, which would provide most of the homogeneous flow, despite its ability to refract light. Since we analyzed the changes in the concentration of the same compound (2-nitro-5-thiobenzoate) in the same solution (phosphate buffer), the problems associated with calculating the absorbance can be partially resolved by using a nonlinear calibration curve. To build the cell, we used a glass capillary with a 1 mm internal diameter. Using a gas burner, we made two necks 1 cm apart in this capillary. The capillary was cut at these necks, and two needles (d = 0.8 mm) were inserted into the necks. The connections of needles to the capillary were sealed with a minimal amount of silicone hermetic. The capillary, together with needles inserted into it, was placed into a petri dish on top of two metal cylinders (1.2 mm in diameter, made from thick needles). The PDMS elastomer was poured on top of the whole assembly and cured.
Once PDMS was cured, the rectangular piece of PDMS was cut and the needles were removed, leaving a connection for the tubing (Fig. 30). Finally, the cell was attached to a glass slide and placed in the focus of the beam of the UV-Vis spectrometer.
We noted that a more convenient way of embedding a glass capillary into PDMS would be to use a twostage process that eliminates the use of metal cylinders as support for the capillary. Here, one needs to make a cured layer of PDMS in a petri dish and to place the capillary directly on top of it.
The embedding of the capillary into PDMS plays a critical role for this cell; it eliminates sharp changes in the refractive index at the boundary of the round capillary because the refractive index of glass (~1.52) is much closer to the refractive index of PDMS (1.42) than to the refractive index of air (1). Therefore, it allows minimizing the refraction of the beam to an acceptable level.

The standard protocol for oscillations in flow (except for experiments that study the response to chemical stimuli)
Four syringes were filled with the required solutions as described below:  163 mg of thiouronium salt 8 were dissolved in 2.5 ml of HPLC-grade water, and this solution was

Protocol for studies of responses of oscillators to chemical stimuli
To study the response of oscillators to chemical stimuli, we developed a system that allowed us to add an extra reagent to the oscillatory mixture while keeping the concentrations of all other components constant. In this system, the content of the syringes that supply cystamine to the phosphate buffer, and Ellman's reagent remained the same as described in the previous section. K3[Fe(CN)6] (5 mM) was added to the syringe containing maleimide (5 mM) and acrylamide (321 mM) for the experiment with a "mixed" oscillator. These three syringes were connected to the CSTR and the mixer identically to the original set-up. However, the inlet of the CSTR through which 8 was supplied in the original set-up was connected to the outlet of a mixer with three inlets (see Figure 6 in the paper). These three inlets were connected to syringes with (i) 8 (336 mM) or a mixture of 8 (150 mM) with acetylthiocholine (150 mM) in water, (ii) pure water, (iii) a solution of the compound; thus, it was the effect we intended to study (i.e., furfuryl alcohol (600 mM), thiosemicarbazide (120 mM), glyoxal (300 mM)). The flow from the syringe (i) was always ½ from the flow from syringes connected directly to CSTR. The combined flow from syringes (ii) and (iii) was also ½ from the flow from syringes connected directly to CSTR. By changing the ratio of flows from syringes (ii) and (iii), we controlled the dilution and consequently, the final concentrations of the CSTR in the furfuryl alcohol, thiosemicarbazide, and glyoxal.

Data processing for the oscillatory experiments
The results of the oscillatory experiments were then plotted using MATLAB 2019a; they are presented in the following Supplementary Figures 32 to 35. The script "plotall.m" has the following operations: (1) it loads the data files and removes noise from the original data, (2) it checks the baselines and uses the corresponding methods to convert the absorbance to the concentration, and (3) it saves the converted files as new data files with the names "nonoiseCon" plus the original name. This script also displays the data in time-concentration plots just for debugging purposes. The other MATLAB file, "plotOscillator.m", plots each of the oscillation experimental results, which corresponded to the data points in the phase plot in Figure 6d in the paper.

Model of the autocatalytic reactions
Supplementary Figure 36. Details of the numerical model for the autocatalytic reaction between thiouronium salts and cystamine (17). Three parameters are involved: is the rate of the thiouronium salt hydrolysis; is the rate of the ligation; is the rate of the disulfide exchange. All the reactions in the model are assumed to obey the mass-action law.
The numerical model of the reaction between thiouronium salts and 17 is shown in Fig. 36 where is the rate of thiouronium salt hydrolysis, is the rate of the disulfide exchange, and is the overall rate of the ligation. The model was built in COPASI and parameters were fitted using the build-in function of the parameter estimation. Here, was fixed to 0.444 (M -1 S -1 ) according to our previous research. 11 The other two parameters ( and ) were estimated from least squares curve fitting and the Levenberg-Marquardt algorithm was used. The COPASI model was attached as a separate file.

Complete model of the oscillator
Supplementary Figure 40. The model used for the numerical simulation of the oscillation in CSTR. Three modules were added to the autocatalytic model, the in and out-flow, delay by maleimide, and negative feedback using acrylamide. The parameters involved here are as follows: kH is the rate of the thiouronium salt hydrolysis; kL is the rate of the ligation; kSS is the rate of the disulfide exchange; kAAm is the rate of the reaction between thiol and acrylamide; kMal is the rate of the reaction between thiol and maleimide, and FvV is the space velocity.
All reactions except for flow are assumed to follow the mass-action law. The rate law for flow supply and removal is described as FvV × concentration.
The numerical model of the oscillation in CSTR (Fig. 40) was built according to the previous literature, 11 which was simulated using MATLAB (Fig. 41-43). The behavior of the system was described by a set of ODEs The dynamics of the model is described by the following system of ordinary differential equations (ODEs)

Three-variable model and linear stability analysis of the oscillator
To analyze the behavior of the oscillator, we used the three-variable model, which is identical to the one used in the analysis of the thioester-based oscillator. 11 The model of the thiouronium salt oscillator is reduced to three variables by applying three approximations: first, the positive feedback loop is described by simple quadratic autocatalysis with the rate constant k1; second, the negative feedback loop (the reaction with acrylamide) is described by a first-order reaction with rate constant k3, and third, the end products can be neglected.
With these assumptions, the system of reactions can be described by: where The steady states are defined by: To analyze the stability of the steady states, we used the Jacobian matrix (J), and determined the eigenvalues (λ) such that the following is not invertible: We used a MatLab script (available on request) modified from our previous work to compute and plot a 2D map of the oscillatory and steady-state regions (Fig. 6d in the paper). 12 We noted that the script uses the input parameters "param.K_AAm" and [Acrylamide], which are related to k3 by k3 = param.K_AAm*[Acrylamide].