Protein-like proton exchange in a synthetic host cavity

Significance

Water drives molecular interactions central to the chemistry of life. Critical to life’s metabolic machinery, enzymes catalyze reactions with remarkable selectivities and efficiencies resulting from noncovalent interactions between water, substrate, and the enzyme active site. Here, amide hydrogen–deuterium exchange kinetics provide quantification of the electronic and steric factors that influence exchange between water and a highly charged host molecule whose protective outer shell and hydrophobic cavity allow it to function like an enzyme. Investigations suggest that the host’s interior microenvironment reacts with encapsulated water in a manner that is correlated to the host’s ability to mediate acid catalysis, thereby providing a tool for developing enzyme-like microenvironments to mediate acid- or base catalysis.

Abstract

The mechanism of proton exchange in a metal–ligand enzyme active site mimic (compound 1) is described through amide hydrogen–deuterium exchange kinetics. The type and ratio of cationic guest to host in solution affect the rate of isotope exchange, suggesting that the rate of exchange is driven by a host whose cavity is occupied by water. Rate constants for acid-, base-, and water-mediated proton exchange vary by orders of magnitude depending on the guest, and differ by up to 200 million-fold relative to an alanine polypeptide. These results suggest that the unusual microenvironment of the cavity of 1 can dramatically alter the reactivity of associated water by magnitudes comparable to that of enzymes.

Water-mediated proton transfer in molecular cavities plays an essential role in the function of nature’s remarkable metabolic machinery (1, 2). Of the broad diversity of enzyme-catalyzed reactions, those involving transfer of hydrogen are ubiquitous and of high fundamental importance (3). For instance, ATP synthase employs an internal chain of water molecules to mediate proton transfer across an electrochemical gradient (4, 5). This process drives ATP synthesis, which allows organisms to broadly meet their basic energy needs. Recent studies of tunneling in enzymatic hydron transfer reactions have other far-reaching and broad implications, namely the importance of the dynamic enzyme motions in driving catalysis (6, 7).

Of the methods used to study hydron transfer, kinetic studies of the mechanisms of amide hydrogen–deuterium exchange (HDX) have helped to elucidate the dynamic interactions between water and protein surfaces that are central to the unique capabilities of enzymes (8–13). Noncovalent interactions, such as electrostatic effects and hydrogen bonding, have also been shown to exert a significant influence on amide HDX rates. These properties are manifest in amide HDX rate constants that can vary by a factor of a billion for residues on the same protein (8). How these variations arise, and how such a property could be harnessed to drive enzyme-like reactivity, remain challenging questions.

In a growing effort to emulate the efficacy of biological receptors and enzymes, studies of the preparative, guest-binding, and catalytic properties of synthetic assemblies have been pursued (14–21). Whereas guest encapsulation has been reported in a variety of media, association processes of organic guests are generally more thermodynamically favorable in water (22, 23). Central to these studies is the notion that the displacement of a high-energy water cluster from a receptor drives guest association, and in some cases, subsequent catalysis. However, despite the abundance of structural data documenting diverse water clusters encapsulated in various host cavities, mechanistic descriptions of the dynamic processes between water and host are rare (24–32).

As part of the broader field of synthetic hosts previously mentioned, our group has developed a class of biologically inspired, anionic K₁₂Ga₄L₆ assemblies that exhibit catalytic properties similar to those of enzymes (1; Figs. 1 and 2A) (33, 34). Assembly 1 exhibits 12 intramolecular amide hydrogen bonds which, in analogy to structurally important peptide bonds found in polypeptides, preferentially stabilize the desired tetrahedral supramolecular structure over other conformers (35). Compound 1 and related hosts have been shown to catalyze several important chemical reactions with sizable rate accelerations (up to 10⁶) and unusual selectivity reminiscent of enzyme catalysis (34, 36–41). Unlike most enzymes, the reactions catalyzed by 1 are functionally and mechanistically very diverse. Despite these differences, those reactions proceeding with the largest rate enhancements in 1 all involve proton transfer steps. Spurred by this insight, we have investigated the fundamental mechanism of proton transfer in 1 through amide HDX kinetics. The purpose of these studies is twofold: first, to probe the fundamental mechanisms of acid-, base-, and water-mediated proton transfer in a host that is known to promote enzyme-like rate accelerations in acid-mediated processes; second, to explore the effect of guest binding on proton exchange, which provides insight into the interplay of solvent and guest exchange for a synthetic active site. The results of these studies are broadly relevant to the ligand–receptor binding events involving biomacromolecules and the development of synthetic enzyme mimics that emulate the utility of their biological counterparts.

Fig. 1.

(Left) Space-filling model of K₁₂Ga₄L₆ tetrahedron 1. (Right) Rotated image of 1, where a ligand has been removed to reveal the tetrahedron’s interior.

Fig. 2.

(A) Structurally important hydrogen bonding in tetrahedron 1 (lines represent ligands and spheres represent Ga³⁺ centers), and polypeptides. (B) General scheme of solvent displacement from 1 by a guest molecule.

Results and Discussion

Initial studies revealed a puzzling relationship between the observed rate of amide HDX and the identity of the encapsulated guest. In the absence of guest, ¹H NMR analysis revealed that 1 undergoes first-order amide HDX relatively rapidly, as evidenced by the loss of a single protio amide resonance at 13.1 parts per million (ppm) (100 mM phosphate buffer, pD 8.40, 25 °C). In the presence of a cationic guest, this rate decreases by a magnitude that is commensurate with the binding affinity of the host–guest pair. This relationship is represented in Fig. 3, where with excess guest, the log(k_obs) of amide HDX correlates linearly with log(K_a) (42) of the encapsulated guest. Thus, although it was clear that the presence of a strongly bound cationic guest inhibits host amide HDX, further investigations were necessary to explain this relationship.

Fig. 3.

Linear free-energy relationship between log(K_a) and log(k_obs) of amide HDX. All measurements were obtained using two equivalents of guest.

To probe the influence of guest encapsulation on the mechanism of amide HDX, 1 was treated with increasing concentrations of guest, and proton exchange rates were measured (pD 8.40). In all cases, the rate of amide HDX decreased with increasing guest concentration until one equivalent of guest had been added (Fig. 4). Increasing the concentration of added guest past one equivalent did not cause as large a decrease in k_obs, suggesting that a saturation limit in guest had been obtained, which binds the interior of 1 and strongly inhibits proton exchange. In this manner, rate constants for HDX could be fit to k_obs = k₃f_A + k₂f_B, where f_A and f_B are the mole fractions of species A and B, which may be characterized by the absence or presence of encapsulated water (Fig. 5). This fitting enabled k₃ to be evaluated for G = NMe₄⁺, NPr₄⁺, and PEt₄⁺ (see the Supporting Information). Based on this analysis, HDX proceeds up to 30× faster in B than A. We propose that encapsulated solvent provides a contribution to this difference (8–10 solvent molecules are expected to be encapsulated, based on thermogravimetric analysis) (42). It is also possible that HDX in the presence of excess G occurs simultaneously with guest exchange, as relevant guest self-exchange processes likely involve ligand distortion and occur 10–1,000× faster than the observed rates of HDX (43). Eyring analysis of HDX (G = PEt₄⁺) revealed a large negative entropy of activation that is similar to the negative standard entropy of guest exchange and consistent with either hypothesis. The mechanism in Fig. 5 also provides a conceptual link between host–guest-influenced HDX and the Linderstrøm–Lang model of structurally hindered proton exchange by proteins (12, 13). Taken collectively, these data support a proton exchange mechanism involving encapsulated water, which can be modified by the presence of a competing guest.

Fig. 4.

Effect of increasing guest equivalents on k_obs of amide HDX with G = PEt₄⁺, NMe₄⁺ (pD 8.40; lines represent fits of experimental points to Eq. 1. Error bars are omitted for clarity; see the Supporting Information).

Fig. 5.

(A) General guest-dependent mechanism for amide H/D exchange at constant pH and (B) rate law describing this process, where f_A and f_B are the fractions of host existing as A and B, respectively, and guest self-exchange is faster than k₂, k₃ (42).

To assess the extent to which acid- and base-mediated proton exchange processes are accelerated or retarded in 1, the effects of bulk solution pD on HDX rates were investigated. Mechanistic studies have revealed that amide HDX proceeds by acid-, base-, and to a lesser extent, water-mediated mechanisms, of which the dominant mechanism depends on bulk solution pH or pD. Experimental rate constants obtained at various pD values were fit to the logarithmic form of Eq. 2 (in Fig. 7D), from which rate constants for exchange catalyzed by acid (k_D), base (k_OD), and water (k_W) were obtained. These rate constants for host containing NMe₄⁺, PEt₄⁺, and solvent are tabulated in Table 1. The influences of the host are seen in the 10⁵-fold acceleration of k_D in solvent-filled host, and in the deceleration of k_OD by 10⁸-fold in PEt₄⁺-filled host. For comparison, rate constants for a representative alanine polypeptide are included (44). The contributions of the various mechanisms are evident in the parabola-like relationship observed between log(k_obs) of amide HDX and pD (Fig. 6).

Fig. 7.

Acid- (A), base- (B), and water- (C) mediated amide H/D exchange mechanisms, which can be described by (D) Eq. 2.

Table 1.

Rates of amide H/D exchange in 1

Rate constant	PEt4+⊂1	NMe4+⊂1	(D2O)n⊂1	PDLA*
logkD M−1⋅s−1	2.80 ± 0.03	3.81 ± 0.14	5.41 ± 0.04	0.0171
logkOD M−1⋅s−1	0.16 ± 0.02	1.15 ± 0.09	3.16 ± 0.04	8.48
logkW s−1	−3.62 ± 0.01	−2.84 ± 0.03	−2.25 ± 0.02	−3.04

^*Rate constants (298 K) for PDLA were calculated using Arrhenius parameters from ref. 44.

Fig. 6.

Influence of bulk solution pD on k_obs of amide HDX. Varying the presence and type of guest affords order-of-magnitude changes in k_obs, whose minimum pD values are shifted up to 6 pD units relative to poly-DL-alanine (PDLA). Lines represent fits of experimental k_obs values to Eq. 2 (Fig. 7D; see Table 1 for rate constants).

It is apparent that the rate minimum for HDX in 1 occurs at pD ∼9, considerably higher than the minimum at pH ∼3 for polypeptides in aqueous solution. This minimum is where the dominant mechanism of amide HDX switches from acid to base catalysis. The shift of the minimum for 1 corresponds to a 200,000-fold increase in k_D and decreases in k_OD [compare (D₂O)_n⊂1; Table 1] relative to a representative polypeptide. These dramatic differences can be understood in terms of the mechanisms for acid and base catalysis, shown in Fig. 7 (45). Regardless of whether the acid-catalyzed exchange proceeds via N protonation (more likely in a polyanionic host) or via the imidic acid, the transition state bears a positive charge. In contrast, the transition state for the base-catalyzed exchange bears a negative charge. The positive charge is stabilized and the negative charge is destabilized by electrostatics, resulting from the polyanionic nature of 1. Similar behavior was seen in amide HDX in anionic (and cationic) micelles (9).

In the presence of a guest (NMe₄⁺ or PEt₄⁺), the pD-dependent curves in Fig. 5 are lowered and flattened. The data in Table 1 show that this is because both k_D and k_OD are decreased 40- to 100-fold for NMe₄⁺ and 400- to 1,000-fold for PEt₄⁺. This effect is not simply electrostatic, as it operates on both k_D and k_OD. Instead, when the interior of 1 is blocked by G, both acid- and base-mediated exchanges are retarded. We attribute this difference to a steric effect, whereby the guest interferes with the HDX, but we cannot be more specific as to how this operates.

It is surprising that k_W is hardly affected by the presence of G. The range of variation is only about 10-fold, and k_W is nearly the same as for a representative polypeptide. It is unusual for k_W to be detectable above the errors in measuring the contributions of k_D and k_OD. What makes detection possible in the presence of G is that both of those contributions are retarded relative to the water-mediated pathway, leading to curves that are flattened near the pD minimum, as can be seen in Fig. 6.

Evidence for the water-mediated pathway has historically been tenuous, which has limited the establishment of a firm mechanism. Water-mediated HDX is highly unlikely to proceed simply through NH removal by D₂O, as this process would have a maximum rate constant of 10⁻⁵ s⁻¹, based on an amide pK_a of ∼15 and an encounter-controlled reverse rate constant of 10¹⁰ M⁻¹⋅s⁻¹. We instead propose the mechanism shown in Fig. 7C. Ordinarily this path too is unlikely, with a rate constant of 10⁻⁶ s⁻¹, based on a pK_a of about −1 for the O-protonated amide, a pK_W of ∼15 in D₂O, and an encounter-controlled rate constant of 10¹⁰ M⁻¹⋅s⁻¹ for deprotonation by OD⁻. However, here the negative charge on the host can raise the pK_a of the O-protonated amide up to 4 units (46), and retard encounter only slightly, leading to an estimate consistent with the observed k_W. However, we cannot explain why the steric effect that retards k_D and k_OD is not also operative on k_W.

In summary, the observed rates of amide HDX in biologically inspired, anionic K₁₂Ga₄L₆ supramolecular assemblies 1 were correlated to the affinities of alkylammonium and phosphonium cations for the host interior. Guest titration experiments and modeling suggest that amide HDX proceeds through two different pathways contingent on the presence or absence of a guest, justifying this correlation. The rates of HDX catalyzed by acid and base are shifted by a factor of up to 200 million relative to those of an alanine polypeptide, an effect that we attribute to the polyanionic nature of 1. Collectively, these experiments suggest that amide HDX occurs through encapsulation of water and that the unique electrostatics of the host dramatically alters reactivity of this transient hydrate. These results improve both our understanding of water in molecular recognition, and contributions of proton transfer to multistep acid catalysis in supramolecular cavities. Knowledge of these properties may enable the more judicious development of both synthetic and biomolecular microenvironments. More broadly, these insights are envisioned to link the fundamental processes of hydron transfer and enzyme-mimic structure, thereby providing a significant step toward the development of catalysts that rival the utility of biological enzymes.

Experimental

General Procedures, Instrumentation, and Materials.

Unless otherwise noted, reactions and manipulations were performed using standard Schlenk techniques or in an oxygen-free wet box. NMR spectra were obtained on Bruker AV-500 or AV-600 spectrometers. Chemical shifts are reported as δ in ppm relative to residual protonated solvent resonances. Unless otherwise noted, chemicals were obtained from commercial suppliers and used without further purification. The synthesis of 1 and complexes with NMe₄⁺, NPr₄⁺, NEt₄⁺, and PEt₄⁺ have been described (33, 43). All solvents were degassed under N₂ for 20 min before use. Buffered solutions were made by titrating a K₃PO₄ solution with DCl until the desired pD was reached. Reported pD and [OD^–] values are corrected for the glass electrode artifact (i.e., pD = pH_read + 0.40) (47) and for solvent dissociation constants (48). Least-squares fits were obtained using Origin (OriginLab) or Microsoft Excel. Rate constants were fit to log(k_obs) = log(k_D[D⁺] + k_OD[OD^–] + k_W). Association constants (K_a) for the encapsulation of NMe₄⁺ [log(K_a/M⁻¹) = 2.51(1)] and NPr₄⁺ [log(K_a/M⁻¹) = 3.12(5)] were measured by NMR, UV-vis, and isothermal titration calorimetry (42). The association constants for PEt₄⁺ and NEt₄⁺ have been reported previously (42, 43).

General Procedure for Amide HDX Kinetic Experiments.

In a typical experiment, host 1 (5 mg, 1.4 μmol) was treated with 0–5 equivalents of NMe₄Cl, NPr₄Br, NEt₄Cl, or PEt₄I in H₂O. After 10 min, water was removed in vacuo, affording a yellow solid. This solid was then dissolved in 500 µL 100 mM deuterated phosphate buffer, after which the resulting solution was transferred to a standard NMR tube. The tube was then inserted in a preheated NMR probe [25.0(1) °C] and the reaction was followed within 3 min of its preparation with single-scan ¹H spectra acquired every 20–60 s. Decay of the integral of an amide NH resonance was assessed against either the aromatic resonances of 1 or an internal standard of sodium 3-trimethylsilylpropanesulfonate. First-order rate constants were obtained using the equation I = (I_o − baseline)exp(−k_obs) + baseline, where I_o is the initial NH integral (44). In the case of solvent-filled host, only the final ∼20% conversion was observed and k_obs was obtained using an adjustable baseline parameter. Initial rate kinetics were used to obtain k_obs where the reaction was slower, such as in the presence of PEt₄⁺. To reliably fit k_obs in the cases of slow HDX, the baseline parameter was constrained to zero. Large differences in KCl concentration, i.e., 20 and 570 mM, afforded no measurable change in HDX rate [k_rel = 1.0(1)], demonstrating that this salt does not influence the HDX rates. Guest exchange of PEt₄⁺ for NEt₄⁺ was likewise unaffected by moderate changes in bulk solution acidity [pD_corr 6.47 and 10.45; k_rel = 1.1(1)]. Triplicate measurements of HDX [solvent-filled host, 25.0(1) °C, pD 8.40] had an SE in k_obs of 7%, which is reflected in the fits of the next section.

Fits of Eq. 1 to k_obs with Variable [G].

The plots below represent experimental k_obs of amide HDX at variable [G], presented with fitted k_obs values using Eq. 1 (see the main text). Fitted k_obs were obtained using least-squares fitting of log(k_obs) to the experimental log(k_obs) data. Parameters K, [H]_T, k₂, and k₃ were adjustable and [H] was calculated by solving the quadratic (Eq. S4) derived below, thereby enabling calculation of f_A and f_B from K, [H]_T, and [G]_T. The quadratic (Eq. S4) is obtained from Eqs. S1 and S2, where [H]_T is the sum of [HG] and [H], and [G]_T is the sum of [HG] and [G]. Notably, fitted K values in the fittings below are larger than independently determined encapsulation constants (see General Procedures, Instrumentation, and Materials). We attribute this difference to (i) steric effects resulting from external association and guest exchange processes, and (ii) the possible existence of a partially solvent-filled host that exhibits a slow HDX. Nonetheless, k₂ and [H]_T are in agreement with experimentally determined values. Close agreement between k_W and k₃ was also observed.

$${\left[H\right]}_T=K\left[H\right]\left[G\right]+\left[H\right]=\left[H\right]\left(1+K\left[G\right]\right),$$ [S1]

$${\left[G\right]}_T=K\left[H\right]\left[G\right]+\left[G\right]=\left[G\right]\left(1+K\left[H\right]\right),$$ [S2]

$${\left[H\right]}_T=\left[H\right]+\left[H\right]K{\left[G\right]}_T/\left(1+K\left[H\right]\right),$$ [S3]

$$K{\left[H\right]}^2+\left(1+K{\left[G\right]}_T-K{\left[H\right]}_T\right)\left[H\right]-{\left[H\right]}_T=0.$$ [S4]

Fig. S1.

Representative time vs. concentration plot. Typical concentration of amide NH vs. time plot used to obtain k_obs for HDX (guest = NMe₄⁺), where the baseline parameter is constrained to zero. Similar results are obtained when the baseline parameter is adjustable (I_o = 6.35 mM, baseline = 0.31 mM, k = 3.61 × 10⁻³ s⁻¹).

Fig. S2.

Eyring plot representing the dependence of amide HDX k_obs on temperature [guest = PEt₄⁺; pD 8.40, ∆H^‡: 71(1) kJ·mol⁻¹; ∆S^‡: −77(2) J·mol⁻¹⋅K⁻¹].

Fig. S3.

Fit of amide HDX k_obs to Eq. 1 (model) with variable concentrations of PEt₄⁺ (k₂ = 5.8 × 10⁻³ s⁻¹, k₃ = 2.2 × 10⁻⁴ s⁻¹, logK = 5.7, pD 8.40; [H]_T = 2.2 mM).

Fig. S4.

Fit of amide HDX k_obs to Eq. 1 (model) with variable concentrations of NPr₄⁺ (k₂ = 5.6 × 10⁻³ s⁻¹, k₃ = 9.3 × 10⁻⁴ s⁻¹, logK = 4.2, pD 8.40; [H]_T = 1.8 mM).

Fig. S5.

Fit of amide HDX k_obs to Eq. 1 (model) with variable concentrations of NMe₄⁺ (k₂ = 5.7 × 10⁻³ s⁻¹, k₃ = 1.5 × 10⁻³ s⁻¹, logK = 4.9, pD 8.40; [H]_T = 2.0 mM).