Exploration of the chemical reaction network.
The reaction cycle combines the phosphorylation of an amino acid (activation) and its subsequent dephosphorylation (deactivation). Phosphorylation requires a phosphorylating agent that we refer to as fuel. Deactivation through hydrolysis can be spontaneous and sped up by catalysis. Thus, a dynamically phosphorylated amino acid appears in response to fuel, which can only be sustained by the cycle’s continuous fuel consumption. We chose monoamidophosphate (MAP) as a phosphorylation agent because it is known to phosphorylate imidazole,28 yet, it is relatively inert towards hydrolysis. Indeed, we found MAP’s half-life of hydrolysis to be 16.5 ± 0.2 hours by 31P nuclear magnetic resonance (31P-NMR) in 500 mM 2-(N-morpholino)ethanesulfonic acid (MES) buffer at pH 6.5 and room temperature (Fig. 1A, Fig. S1, and Table S1). In contrast, in the presence of 75 mM of histidine, MAP’s half-life for observed hydrolysis decreased to 3.0 ± 1 hours (Table S2). Thus, histidine accelerated the conversion of 80 mM MAP, likely through its phosphorylation. We tested six further amino acids with phosphorylatable side groups—arginine, lysine, cysteine, serine, tyrosine, and aspartic acid (Fig. 1B).29, 30 For all amino acids, MAP’s half-life for observed hydrolysis remained similar to the half-life in the absence of amino acids (Table S2). Thus, none of the other amino acids observably catalyzed the fuel conversion. Indeed, no phosphorylated products were made in observable amounts by 31P-NMR (Fig. S2). We focused on histidine as a catalyst and tested two more prebioticically relevant fuels, i.e., diamidophosphate (DAP) and trimetaphosphate (TMP, Fig. 1C and Fig. S3). With 75 mM histidine, no significant amount of TMP was converted over 60 hours. In contrast, DAP was converted with an observable half-life of 12 ± 0.3 hours, which is significantly faster than without histidine at an observable half-life of 190 ± 50 hours.
We tested the pathway of the histidine-catalyzed fuel consumption. By 31P-NMR, we observed the conversion of MAP, the production of inorganic phosphate (Pi), and the emergence and decay 1-phospho-histidine (1-pH), 3-phospho-histidine (3-pH), and 1,3-bisphospho-histidine (1,3bpH, Fig. 1D, E and Fig. S4). Thus, all possible phosphorylated histidine species were observed. Even more exciting was that all species were only transiently present at the expense of MAP, pointing to the dynamic phosphorylation of histidine at the expense of MAP. The concentration of 3-pHis reached the highest maximum concentration of all species, which peaked at around 9 hours. The others, 1-pHis and 1,3-bispHis, peaked earlier and at lower concentrations. After 60 hours, all 75 mM histidine was deactivated to its dephosphorylated state, and all MAP was converted into inorganic phosphate. Besides the dynamic phosphorylation of histidine, no notable side reactions were observed. The same phosphorylated species were observed when we used DAP as a fuel. However, some side products were found after 60 hours. We conclude that histidine is a catalyst for the hydrolysis of MAP and DAP and is dynamically phosphorylated.
To quantitatively understand the reactions taking place in the cycle, we describe the following reactions in a kinetic model (See supporting notes Scheme S1-S20 and Table S1): the MAP hydrolysis to form inorganic phosphate (Fig. 1A), the reaction of MAP with histidine yielding 1-pHis or 3-pHis, the reaction of MAP with 1-pHis or 3-pHis to form 1,3-bisphospho-histidine (1,3-bispHis). Each of these activation reactions releases one molecule of ammonia. Additionally, we included in the model the dephosphorylation of each phosphorylated species (deactivation). A Levenberg-Marquardt fitting method fitted the model's rate constants to predict the experimental data (see supporting notes Scheme S1-S20 and Table S1). Intramolecular isomerization from 1-pHis to 3-pHis is not described in the model as it is known not to occur.31 In contrast, intermolecular phosphoryl transfer is known to occur.31, 32 But, we assumed that the hydrolysis of 1pHis is faster than the transfer to another histidine and thus do not describe the transfer. We determined the half-lives of the phosphorylated species from the fitted rate constants that could not be determined empirically. In line with the literature,33 it became clear that any phosphate group at the 1-pHis position was most labile, with a half-life of 1.79 hours and 1.88 hours for 1-pHis and 1,3-bpHis, respectively. In contrast, 3-pHis was less labile, with a half-life of 10.6 ± 0.7 hours and 2.88 hours for 1,3-bpHis. The activation rate constants for the different nitrogen atoms in the histidine were in the same range. Taken together, the MAP is catalytically converted by histidine, resulting in multiple transiently phosphorylated species. 3-pHis is the most long-lived, so its observed yield is the highest.
We embedded histidine in a peptide sequence and tested its reactivity to couple the reaction cycle to function in the future, for example, for producing self-sustaining protocells (vide infra) or for forming non-equilibrium self-assemblies. The most straightforward peptide design was placing the histidine between two glycines (G) and acetylating the NH2terminus (AcGHG-OH, Fig. 2C). We added MAP to 75 mM AcGHGOH under the same conditions as above and monitored the concentration profiles with 31P-NMR over 90 hours. AcG(3pH)G-OH appeared first, followed by AcG(1pH)G-OH, whereas in contrast to His, the bis-phosphorylated species was not observed (Fig. S6).
Because we did not observe the bis-phosphorylated species, we neglected it in the kinetic model. The half-life of Ac-G(3-pH)G-OH was experimentally determined to be 59.7 ± 6.4 h, and the one of Ac-G(1-pH)G-OH was predicted by the kinetic model to be 1.78 h (Table S1). Thus, for both histidine and Ac-GHG-OH, the 3-pH-isomer was the most stable. Yet, Ac-G(3-pH)G-OH’s deactivation was drastically slower than that of 3-phistidine (Fig. 1A), which aligns with our expectations.34, 35
The slow deactivation increased the overall reaction time to an impractical duration of hundreds of hours, thus, we tested whether the overall cycle could be accelerated by introducing catalysts for the dephosphorylation—we varied the pH and added pyridine to the reaction solutions. Specifically, we used pH 5.5, 6,5, and 7.5 with and without additional pyridine (5, 15, 25 mM). We monitored the concentration profiles in response to fuel (Fig. S5-S7) and calculated the half-life via the slope of the apparent first-order decay (without pyridine k− 3 Table S1, with pyridine k'−3, Table S3). At pH 5.5 without pyridine, the half-life of AcG(3pH)G-OH was 40.8 ± 9.5 hours, which increased to 59.7 ± 6.4 hours at pH 6.5 and to 143 ± 15 hours at pH 7.5 (Fig. 1B). These findings align with the known acid-lability of the P – N bond. 34, 36–40 28, 29, 32, 33, 37, 41 The observed half-life and, thus, overall cycle lifetime were reduced when the pyridine concentration increased. For example, from 59.7 ± 6.4 hours without pyridine at pH 6.5 to 12.1 ± 3.1 hours with 25 mM pyridine (Fig. 1B, purple). However, at these higher pyridine concentrations, the observed yield was low (Fig. 2D). Furthermore, the yield was too low at low pH (5.5) and high pyridine concentration to determine the apparent AcG(3pH)G-OH half-life.
We adjusted our kinetic model to predict the evolution of the reaction cycle with the peptide as the catalyst in the reaction cycle. For the kinetic model with pyridine present in the system, we neglected the intermolecular phosphate-group transfer between pyridine and pyridinio-N-phosphonate.42 We empirically determined the reaction rate constants of the direct fuel hydrolysis and the 3-phosphoisomer hydrolysis. The formation (kpPy) and hydrolysis (k− pPy) of pPy were determined by reacting the fuel with pyridine alone and fitting the data using COPASI (Table S1). We fixed those rate constants in the kinetic model and fitted the remaining others.
The kinetic model allows us to predict through which pathways the fuel was converted into inorganic phosphate, which we demonstrate in a Sankey diagram (Fig. 2E). At pH 6.5 and with 5 mM pyridine, 17 percent of the MAP was used to phosphorylate histidine directly. Most of the MAP was used to phosphorylate pyridine, and only a small fraction was transferred to the histidine. The majority of the MAP is hydrolyzed directly or hydrolyzed by phosphorylating pyridine. We used these calculations to produce the plots in Figs. 2F and G, which show the pathway through which MAP is hydrolyzed. Either it is hydrolyzed directly or through phosphorylating histidine (black and unwanted) or it is hydrolyzed through phosphorylating histidine (purple and wanted). At low pH values, the cycle was inefficient—only 1% of fuel was used to make the transient 1-pHis isomer and 5% to the 3-pHis isomer at pH 5.5 (Fig. 2F). The remainder of the fuel was lost through direct hydrolysis of MAP. The efficiency was much greater at a higher pH of 7.5, where 58% was used for the activation phosphorylation of histidine (1-pHis, 3pHis, and 1,3-pHis). Thus, decreasing the pH accelerates the total reaction cycle, but at the cost of efficiency. Adding pyridine has similar consequences for the reaction—the more pyridine added, the shorter the half-life of the phosphorylated histidine species, but at the cost of efficiency (Fig. 2G). For example, at pH 6.5 and without pyridine, 42% of our fuel was used to phosphorylate the histidine, whereas, with 25 mM pyridine, this number dropped to 13%.
We designed histidine-bearing peptides to form self-sustaining complex coacervate droplets upon dynamic phosphorylation. We used the peptide R30 (NH2-Arg30-OH) as a polycation and peptide 1 (Ac-Tyr(OMe)-Asp-His-Asp-Asp-NH2) as a polyanion (Fig. 3B). At pH 7.5, peptide 1 carries three negative charges (-3). In this state, it has a weak affinity for R30 with a binding constant of 1.19 ± 0.02 mM as determined by isothermal calorimetry (ITC, Fig. S9). It can form complex coacervate-based droplets, but only at very high concentrations. Inspired by the design rules of polyanion and polycation-based complex coacervation systems described by Keating,43 we anticipated that the dynamic phosphorylation of the histidine in peptide 1 would increase its affinity for the polycation by increasing its overall charge from − 3 to -5.
We tested for the formation of self-sustaining complex coacervate droplets by mixing 20 mM peptide 1, 12.5 mM MAP, and 50 mM R30 in 500 mM MES buffer at pH 7.5. We produced microreactors in which the droplets could be analyzed by mixing the aqueous solution with a perfluorinated oil. We used a microfluidic droplet generator to produce microreactors of reaction solution surrounded by perfluorinated oil that we analyzed by confocal microscopy (Fig. 3C). As the concentration of phosphorylated peptide 1 increased, we witnessed the first coacervate-based droplets after 9 hours. The droplets were round and rapidly fused, indicating their liquid nature (Movie S1). Besides, a Fluorescence Recovery After Photobleaching (FRAP)-assay on Cy5-labeled R30 further corroborated their liquid nature (Fig. 3D and E, Fig. S10). The droplet rapidly recovered after bleaching with a half-life time (t1/2) of 2.64 ± 0.39 s from which we could derive a diffusivity constant (D) of 0.295 ± 0.04 µm2/s.
After all droplet material had fused to one large droplet, it sunk to the bottom of the reactor. We tracked the average volume of the droplet material in every microreactor with time. From these measurements, we derived the droplet volume relative to its reactor. This fraction increased in the first 47 hours to just under 1% as the reaction cycle produced increasing amounts of phosphorylated peptide 1 (Fig. 3F). Over the next hundred hours, it decayed until no droplets were found after 243 hours.
We analyzed the kinetics of the reaction cycle by HPLC and 31P-NMR and refitted our kinetic model to the obtained data (Fig. 3F and Fig. S11). From the kinetic model, we can derive that the droplets dissolved when the total concentration of phosphorylated species fell below 262 µM. We can also conclude that the droplets, after roughly 30 hours, can only comprise 3-pHis as a phosphorylated species because 1-pHis has already been hydrolyzed. Next, we tested how the chemical fuel can affect the macroscopic behavior of the active droplets using optical density at 500 nm as a readout. To produce sufficient droplet material for a reproducible increase in optical density, we added 2.5% PEG8000 as a crowding agent.44 We added various amounts of MAP as fuel. Above 10 mM of MAP, we found evidence of the first droplets. We then determined the time needed for the optical density to pass a threshold which we define as the lag time (Fig. 3G and S13). With increasing amounts of fuel, the lag time decreased from 8.6 ± 1.17 hours with 10 mM MAP to 2.55 ± 0.13 hours with 50 mM of fuel. Notably, the decrease in the lag time tended to level off beyond 40 mM of fuel. With increasing fuel, the maximum turbidity reached, as a measure for the maximum amount of droplet material, increased too (Fig. 3H and S13).
Our experiments show that the new reaction cycle we introduced can be used to create active droplets. In future works, we envision functionalizing the droplets with catalysts and reagents so that they can control their fate. For example, a catalyst could inhibit or accelerate dephosphorylation. To demonstrate that such an approach is viable, we tested whether other components, besides the two peptides, could partition into the droplets. We monitored the partitioning by confocal microscopy of Cy3-labeled single and double-stranded DNA, single-stranded RNA, and polystyrene sulfonate. Both the single-stranded DNA and RNA partitioned well and homogeneously throughout the droplets. Excitingly, the double-stranded DNA and polystyrene sulfonate partitioned well and seemed to form microdomains of high concentration within the droplets.