A Chemically Fuelled Molecular Automaton Displaying Programmed Migration of Zn2+ Between Alternative Binding Sites

Abstract A molecular system comprising a cationic zinc complex and an amino acid‐derived ambident ligand having phosphate and carboxylate binding sites undergoes a series of rearrangements in which the metal cation migrates autonomously from one site to another. The location of the metal is identified by the circular dichroism spectrum of a ligated bis(2‐quinolylmethyl)‐(2‐pyridylmethyl)amine (BQPA) chromophore, which takes a characteristic shape at each binding site. Migration is fuelled by the decomposition of trichloroacetic acid to CO2 and CHCl3, which progressively neutralises the acidity of the system as a function of time, revealing in sequence binding sites of increasing basicity. The migration rate responds to control by variation of the temperature, water content and triethylamine concentration, while an excess of fuel controls the duration of an induction period before the migration event.

We have previously shown that the selective binding of a Zn 2 + cation to a set of monoprotic ligands of differing pK a may be controlled by addition of acid or base, modulating the protonation states of the various ligands in such a way that the Zn 2 + cation associates chemoselectively with the most basic of the available free anions. [21] Exchange between two equilibrium states, as one ligand was displaced by another, was detected by changes in the circular dichroism (CD) spectrum of a bis(2-quinolylmethyl)-(2-pyridylmethyl)amine chromophore (BQPA, Figure 1a(i)) [22,23] also bound at the Zn 2 + centre. We now show that regioselective metal-ligand coordination may be controlled in an out-of-equilibrium system in which pH evolves autonomously as a result of a chemically fuelled deacidification process [ Figure 1a(iii)]. When a single ligand with multiple binding sites [ Figure 1a(ii)] is used instead of a set of monodentate ligands, we find that the Zn 2 + complex migrates autonomously from one binding site to another as the deacidification unmasks in sequence one site after another in a stepwise manner (Figure 1b). The location of the complex is revealed by the BQPA chromophore whose CD signal is distinguishably different at the alternative binding sites of the ligand.

Results and Discussion
The metabolite phosphothreonine 1 offers an archetype of the type of polyprotic, ambident ligand that is required, [24] combining potential binding sites of differing acidity with chirality capable of inducing a detectable CD signal. With 1 as a model, we set about constructing suitable synthetic mimics that could later be combined into a polyprotic target ligand. The ability of the phosphate binding site to induce a CD signal was optimised by first varying the steric demand of the phosphate Osubstituent R 1 in L -threonine analogues 2 a-c, then varying the N-acyl substituent R 2 (Table 1), whilst ensuring that all groups tested were acid-stable. Ligands 2 d-f were derivatised as nbutyl amides to mimic the local environment in a peptide ligand. The best performing analogue 2 f features bulky tertbutyl groups at the phosphate oxygen and α-nitrogen atoms, which were both required for strong CD signal induction.
Carboxylic acid 3, derived from L-hydroxyproline (Hyp), displays (i) a mesitylsulfonyl group as a sterically demanding, acid-stable nitrogen protecting group, (ii) a Boc-γ-aminobutyryl (GABA) side chain as a spacer unit and handle for further derivatisation, and (iii) a trans relationship between the carboxylate and γ-oxygen to minimise steric interaction between the binding site and the side chain. Equimolar 0.25 mM solutions of Zn(BQPA) · 2ClO 4 and either 2 f or 3 in MeCN were titrated with triethylamine ( Figure 2). Upon deprotonation, both 2 f and 3 induced strong, positive CD signals at 239.5 nm from the BQPA ligand. Phosphate 2 f reached a maximum of + 15,200 deg dm 2 mol À 1 , while carboxylate 3 reached a maximum of + 20,000 deg dm 2 mol À 1 at 1 equivalent of base. Further addition of base led to no change in signal for either ligand, indicating complete deprotonation with 1 equiv. NEt 3 in both cases.
The two optimised ligand structures 2 f and 3 were merged to give the Thr-GABA-Hyp derived peptide 4, which was synthesised from L-Hyp in nine steps with an overall yield of 25 % (see Supporting Information for synthetic details). In order to track the movement of the Zn 2 + cation between the binding sites by CD, the two binding sites must induce Cotton effects of opposite signs in the BQPA location reporter. We therefore used D-Thr to invert the sign of the CD signal generated by binding at the phosphate station. The GABA spacer unit gives a total of 16 bonds between binding sites, reducing the risk that both sites will bind simultaneously to the Zn 2 + .
Titration of an equimolar solution of Zn(BQPA) · 2ClO 4 (0.25 mM, MeCN) and 4 with NEt 3 showed the behaviour we were seeking. Initially, the neutral ligand does not bind appreciably to zinc, and only a weak CD response is measured ( Figure 3b). As the first equivalent of base is added, the more acidic phosphate site ('station 1' of Figure 1) is revealed, and a negative CD signal develops at 239.5 nm as the [Zn(BQPA)] 2 + complex migrates to that site. A second equivalent of base further deprotonates the ligand to form a dianion, revealing a carboxylate site ('station 2' of Figure 1) capable of binding Zn 2 + more tightly than the phosphate does. As a result, Zn 2 + migrates (either intra-or intermolecularly) from the phosphate to the carboxylate, and the BQPA reports the change in environment in the form of a switch to a positive CD signal at 239.5 nm, a response that is complete on addition of 3 equiv. of base. Figure 1. (a) Components of the system: (i) A Zn 2 + cation tightly bound to a BQPA ligand whose induced CD spectrum reports the local chiral environment of the Zn 2 + ; (ii) an ambident ligand with a phosphate and a carboxylate binding site, each in a local chiral environment; (iii) trichloroacetic acid as a fuel whose decomposition removes acidic species from the system. (b) Operation of the system. In the presence of acid, the [Zn(BQPA)] 2 + complex remains in solution. As the acidity decreases according to a time regime determined by the consumption of the fuel, the metal moves first to the less basic (station 1) and then to the more basic site (station 2).   Evidence that this binding selectivity arises from the difference in pK a between the two stations was provided by calculations using the COSMO-RS [25][26][27] method (see Supporting Information). Using this method, we estimate pK a values in pure acetonitrile of 15.0 � 2.0 for station 1 and 21.8 � 1.3 for station 2. The difference between the pK a values of the stations is most likely over 3 pK a units, enough to provide almost complete pHdependent binding selectivity. We furthermore estimate, using weighted average positive sigma (WAPS) values as computational descriptors of charge delocalisation, [28] that addition of 1-2 % water will not substantially alter this pK a difference.
The binding constant of the carboxylate station was estimated to be K = 1.19 � 0.15 × 10 6 M À 1 at 25°C by titration of [Zn(BQPA)] 2 + with an equimolar mixture of 3 and the phosphazene base tBuP 1 (NMe 2 ) 3 (see Supporting Information). 3 displayed 1 : 1 host:guest binding behaviour. The binding isotherm of 2 f was more complex (see Supporting Information).
Autonomous migration of Zn 2 + must be powered by a fuel, for which we chose TCAH, which decomposes under basic conditions by decarboxylation of its conjugate base TCA À . The kinetics of this reaction have been extensively studied in water, organic solvents, and solvent mixtures. [29][30][31][32] Since the decomposition of TCAH produces non-acidic products CO 2 and CHCl 3 , it is accompanied by an overall loss of acidity. The consequent time-dependent increase in pH has been exploited to fuel molecular motion in supramolecular assemblies including rotaxane shuttling, [33] rotary motors, [34] catalyst activation, [35] crystallisation [36] and gelation. [37] α-Cyano carboxylic acids have similarly been used as a chemical fuel to control the pHdependent conformations of catenanes [38] and calixarenes, [39,40] and also to temporarily release Zn 2 + from an aza-crown ether host in a self-sorting multicomponent system. [41] We applied the TCAH-fuelled pH change to the multi-site binding system using ligand 4. The operation of the autonomous system is illustrated in Figure 4. The resting state is one in which both binding sites of 4 are deprotonated (Figure 4a), and the Zn 2 + occupies the more basic carboxylate station (station 2).  Addition of two equivalents of TCAH (pK a (MeCN) = 10.93) [42] protonates both sites, priming the automaton by releasing [Zn(BQPA)] 2 + into solution (station 0) (b). The deprotonated fuel decomposes slowly and continuously, transferring protons from the binding sites to chloroform irreversibly. Decomposition of the first equivalent of TCA À deprotonates the more acidic phosphate site (station 1), which captures from solution the [Zn(BQPA)] 2 + ions (c). Further decomposition of the remaining TCA À deprotonates the carboxylic acid, causing the zinc to migrate to the carboxylate station and returning the system to the resting state (station 2) (a). In this way, a dynamic system emerges in which the binding selectivity of the metal complex is autonomously shifted from one site, or station, of the ligand to another in a manner that is programmed by the time-course of the deacidification of the solution.
The molecular automaton was constructed by mixing Zn(BQPA) · 2ClO 4 (0.25 mM), 4 (1 equiv), NEt 3 (3 equiv.) in MeCN containing 2 % H 2 O, [43] and primed by adding TCAH (3 equiv.) which initiated operation of the automaton starting at station 0. Progress of the system through the programmed series of steps was monitored by molar ellipticity at 239.5 nm over 20 min (Figure 5a). As we had hoped, the BQPA 'location reporter' signal decreased over the first few minutes, reaching a negative maximum of À 3,000 deg dm 2 mol À 1 at around 4 minutes, as the zinc migrated to the phosphate station 1. The location reporter signal then swung to positive and increased in magnitude over the next 16 minutes, corresponding to carboxylate deprotonation and migration of Zn 2 + to the carboxylate station 2.
Increased amounts of both NEt 3 and TCAH up to 5 equivalents were accompanied by an increase in the rate of operation; the rapid growth of the phosphate signal showed that the zinc arrived earlier at the phosphate station 1, with the journey time decreasing from 420 seconds (at 2 equiv.) to 110 seconds (at 5 equiv.). Similarly, it migrated faster to the carboxylate station 2, as the carboxylate signal plateaued earlier following a steeper gradient. [44] Higher temperatures likewise led to faster TCA À decomposition and faster Zn 2 + migration. Samples containing Zn-(BQPA) · 2ClO 4 (0.25 mM, MeCN with 2 % H 2 O), 4 (1 equiv.), NEt 3 (2 equiv.) and TCAH (2 equiv.) were monitored over 30 minutes at temperatures between 20-40°C (Figure 5b). Steeper gradients were observed as the temperature was increased; the time taken for Zn 2 + to arrive at the phosphate station approximately halved with every 5°C increase. The time taken for the system to return to the carboxylate station also decreased at higher temperatures; the output signal plateaued after 6 minutes at 40°C, while no plateau was observed within 30 minutes below 25°C. [45] Since the TCAH stock solution in MeCN required a water content of 20 vol % to prevent spontaneous decomposition, [43] it followed that the rate of Zn 2 + migration could also be controlled by varying the water content of the sample. Samples
With excess TCAH the site-hopping event may be delayed by an induction period, as migration to the phosphate site can only begin once excess TCAH is depleted. [39] Samples containing Zn(BQPA) · 2ClO 4 (0.25 mM, MeCN with 2 % H 2 O), 4 (1 equiv.), NEt 3 (2 equiv.) and TCAH (2-6 equiv.) were monitored over 20 minutes at 35°C (Figure 5d). The excess TCAH leads to a travel delay which increases linearly by an average of 139 seconds for each additional equiv. of TCAH (0.25 mM), giving a zero-order decomposition rate constant of k = 1.8 × 10 À 6 mol dm À 3 s À 1 during the induction period. TCA À concentration remains constant until TCAH is fully consumed, but once excess TCAH is depleted, the Zn 2 + starts its journey first to the phosphate and then the carboxylate station, but at the same rate, regardless of the initial TCAH concentration.
Finally, multiple additions of fuel were used enforce restarts of the migration program. A sample of Zn(BQPA) · 2ClO 4 (c o = 0.025 mM, [46] MeCN with 2 vol % H 2 O), 4 (1 equiv.) and NEt 3 (2 equiv.) was consecutively treated with 1.0, 2.0, 3.0, 0.5 and 1.5 equiv. of TCAH, with a 10 minute observation period between fuel pulses (Figure 5e). By varying the amount of TCAH added, the migration process could be reprogrammed to start at any point between stations 0 and 2. After each fuel pulse, the system returned to the same resting state output of + 9,100 deg dm 2 mol À 1 within 10 minutes. A further experiment in which the automaton was repeatedly restarted with ten successive additions of 2.0 equiv. TCAH (see Supporting Information) indicated that the operating rate decreased marginally with each cycle, owing to accumulation of the water required to stabilize the fuel.

Conclusion
We have demonstrated migration of a cationic metal complex [Zn(BQPA)] 2 + between different sites on an ambidentate peptide ligand 4, in which the location of the metal is controlled by the pH of the medium. Coupling the migration of the metal between sites with TCA À decomposition creates a chemically fuelled molecular automaton, in which autonomous reorganisation of its components followed a time course that could be programmed using five independent variables. The absolute configuration of the ligand binding sites controlled the sign and magnitude of the output signals, while initial TCA À concentration, temperature and water content all influenced the rate of movement between sites. An excess of TCAH fuel allowed the start time of the motion to be delayed. The total operating time of the automaton was typically on the order of hundreds of seconds, a convenient timescale for continuous monitoring by optical techniques. The migration of the metal between stations and the factors controlling fuel decomposition are general and need not necessarily be limited to systems with a CD output; their highly programmable behaviour shows their potential to be used as controllers for more complex supramolecular assemblies.
Synthetic procedures, characterisation of novel compounds (including 1 H and 13 C NMR spectra), miscellaneous CD data (base titrations and binding isotherms) and pK a calculation methodology may be found in the Supporting Information.