Computational Insights into the Mechanism of Nitric Oxide Generation from S-Nitrosoglutathione Catalyzed by a Copper Metal-Organic Framework

Abstract

The controlled generation of nitric oxide (NO) from endogenous sources such as S-nitrosoglutathione (GSNO) has significant implications for biomedical implants due to the vasodilatory and other beneficial properties of NO. The water-stable metal-organic framework (MOF) Cu-1,3,5-tris[1H-1,2,3-triazol-5-yl]benzene has been shown to catalyze the production of NO and glutathione disulfide (GSSG) from GSNO in aqueous solution as well as in blood. Previous experimental work provided kinetic data for the catalysis of the 2 GSNO → 2 NO + GSSG reaction, leading to various proposed mechanisms. Herein, this catalytic process is examined using density functional theory. Minimal functional models of the Cu-MOF cluster and glutathione moieties are established, and three distinct catalytic mechanisms are explored. The most thermodynamically favorable mechanism studied is consistent with prior experimental findings. This mechanism involves coordination of GSNO to copper via sulfur rather than nitrogen and requires a reductive elimination that produces a Cu(I) intermediate, implicating a redox-active copper site. The experimentally observed inhibition of reactivity at high pH values is explained in terms of deprotonation of a triazole linker, which decreases the structural stability of the Cu(I) intermediate. These fundamental mechanistic insights may be generally applicable to other MOF catalysts for NO generation.

Graphical Abstract

INTRODUCTION

Water-stable catalytic generation of nitric oxide (NO) from endogenous sources has great potential for applications in biomedical implants, where NO release from the implant is known to have desirable vasodilatory and other beneficial health effects.^1–5 The Cu-1,3,5-tris[1H-1,2,3-triazol-5-yl]benzene (Cu-BTTri) metal-organic framework (MOF)^6–11 has been shown to catalyze homolytic sulfur-nitrogen cleavage of S-nitrosoglutathione (GSNO) in the presence of glutathione (GSH) to produce NO and glutathione disulfide (GSSG) in aqueous solutions. Given its stability and catalytic activity in both water and blood,^11–14 the solid-state Cu-BTTri catalyst is well-suited for biomedical applications.¹⁴

Recent experimental investigation^9–12 into the kinetics of Cu-BTTri-catalyzed NO production from GSNO revealed several key elements. First, a reaction stoichiometry of 2 GSNO → 2 NO + GSSG in aqueous solution (Figure 1) was established by direct NMR monitoring in aqueous solution.¹¹ Second, kinetic measurements as a function of the relative concentration of surface sites in combination with CN^– and 3,3′,3″-phosphanetriyltris benzene sulfonic acid catalyst poisoning experiments indicated that the low-coordinate copper sites on the surface of the ca. 0.6 ± 0.4 micron Cu-MOF particles are the kinetically dominant active sites.¹⁰ These surface active sites were proposed to be missing-linker defects that form when the Cu-BTTri particles stop growing during synthesis.¹⁰ Third, the rate of GSNO consumption at GSH saturation was measured experimentally to be 2.8 × 10^–4 M s⁻¹. Fourth, the rate was found to be first-order in [Cu-MOF] surface sites^{11, 15} and to be either first or second order in [GSNO].⁹ Finally, poisoning at pH values greater than pK_a ~5.6 was observed experimentally.⁹ This poisoning was proposed to be caused by deprotonation of a copper-bridging protonated triazole (free triazole pK_a = 9.3)^{16, 17} or deprotonation of the thiol proton in Cu-bound GSH (free GSH pK_a = 8.8).^{9, 17}

Figure 1.

In the presence of GSH, Cu-BTTri catalyzes the homolytic S–N cleavage of 2 GSNO, releasing 1 GSSG and 2 gaseous NO.

A variety of computational studies of chemical processes catalyzed by MOFs have provided structural, thermodynamic, and mechanistic insights.^18–33 Herein, we present theoretical calculations that address the following five fundamental questions relevant to the catalytic mechanism of NO release from GSNO at an active site of a Cu-BTTri MOF catalyst. First, what constitutes a minimal functional model for the MOF, the GSH, and the GSNO that will enable meaningful computational studies? This model must be able to not only retain the geometry of the putative active site of the MOF throughout the catalytic cycle, but also provide a physically reasonable thermodynamic landscape for catalysis. Second, is N- or S- coordination of GSNO to the Cu-MOF (or conceivably both) predicted in the thermodynamically favored catalytic cycle? The previously proposed proton-coupled electron transfer (PCET) mechanism assumed Cu-N coordination, but this previous proposal explicitly noted that Cu-S coordination of GSNO was possible and could not be ruled out.⁹ Third, what mechanism for the critical S–N cleavage would be consistent with the experimentally determined free energy span of 15.6 kcal/mol? Fourth, what is the turnover-limiting step in the thermodynamically favored mechanism from the computations, and is the mechanism consistent with the experimentally observed rate law? Fifth, what is the physical origin of the experimentally observed poisoning above pH 5.6?

In conjunction with experimental data from prior studies, these calculations establish a minimal functional model and provide fundamental mechanistic insights into this heterogeneously catalyzed reaction. The computationally determined mechanism that is consistent with the experimental findings listed above entails coordination of the GSNO to the Cu-MOF active site through its sulfur atom and involves a Cu(I) intermediate. Indeed, there is precedent in the literature for S–N homolysis via S-coordination of GSNO and other S-nitrosothiols to a variety of systems.^34–36 Furthermore, the computational studies suggest that the protonation site responsible for the experimentally observed inhibition of reactivity at high pH values in the Cu-BTTri catalyst⁹ is the triazole rather than the GSH. These types of insights have broader implications for Cu- and other metal-MOF catalysis.

METHODS

DFT Calculations

The DFT calculations were performed using both the B3P86-D3 functional^37–39 modified to include 15% exact exchange, as shown to provide an improved description of spin states in transition metal compounds,⁴⁰ and the TPSSh-D3 functional, a modified version of the TPSS-D3 functional that has been shown to predict accurate enthalpies of formation for inorganic complexes.^{39, 41, 42} As demonstrated below, these two functionals provide consistent results throughout this study. The def2-TZVP basis set⁴³ was used for the copper atoms, and the 6–311++G** basis set was used for all other atoms.^44–48 As the results were similar for the B3P86-D3 and TPSSh-D3 functionals, the results for the TPSSh-D3 functional are included only in the SI (Figs. S2–S4 and S7–S8). All calculations were performed in implicit water solvent using the conductor-like polarizable continuum model (C-PCM).⁴⁹ Spin contamination was monitored for all species and confirmed to remain negligible. All DFT calculations were performed using the Gaussian16 quantum chemistry program.⁵⁰ Charges and spin densities computed with the NBO7 package⁵¹ and the Atomic Polar Tensor (APT) charge model⁵² interfaced with Gaussian were used to verify the spin states. In addition to these population analyses, stability analyses were performed for species involving radical ligands, where various spin states are possible. Population analyses are provided in Tables S2–S4 of the SI, and spin contamination analyses are provided in Tables S5–S7 of the SI. Moreover, for benchmarking purposes, the free energies for this mechanism were also computed with a C-PCM approach that includes dispersion, repulsion, and cavitation effects (Figure S12). These effects were found to influence the quantitative, but not qualitative, characteristics of this mechanism.

The copper atoms in the Cu-BTTri MOF were assumed to be in the high-spin ferromagnetic state for this study, corresponding to a quartet for the isolated MOF cluster with three Cu atoms. Alternative spin configurations of the three copper centers are possible, but their investigation is susceptible to spin contamination in UDFT calculations and is beyond the scope of this study. When a radical is complexed to the active Cu, the local spin state for this Cu-ligand complex is denoted a closed-shell singlet, open-shell singlet, or triplet state, corresponding to an overall triplet, triplet, or quintet spin state, respectively. All three of these spin states were investigated for these species, and the lowest free energy species obtained was used in the free energy landscapes. For all mechanisms studied that involve a radical complexed to Cu, the local triplet state, corresponding to an overall quintet spin state, was found to have the lowest free energy.

All free energy calculations included the zero-point energy and the vibrational and translational entropic contributions, as well as the solvation free energies. The rotational entropic contributions were omitted because the solvent is expected to inhibit rotations, as discussed elsewhere.⁵³ For comparison, the free energy landscape for the most thermodynamically favorable mechanism was also computed without omitting the rotational entropic contributions and was shown to be qualitatively similar (Figure S13). Evaluating the thermodynamics of the S-complexed Cu(I) mechanism required the free energy of a solvated proton. An isodesmic reaction could not be used to treat the solvated proton, as there was no experimental pK_a found for a suitable reference. Instead, the commonly used solvation free energy of –265.6 kcal/mol for a free proton was used.^{54, 55} This value does not influence the most relevant steps of the reaction pathway, but rather influences only one of the later steps.

Molecular Dynamics Simulation of the GSH Model

A molecular dynamics (MD) simulation of the full glutathione tripeptide was performed using the Amber 2018 package⁵⁶ with the ff99sb force field.⁵⁷ Parameters for the γ-glutamyl motif were taken from ref.⁵⁸. The tripeptide was solvated by TIP3P water⁵⁹ for at least 12 Å beyond the molecule in each direction to form a periodically replicated cubic box with side of length ~37 Å, and the total charge was neutralized with sodium counterions.⁶⁰ For equilibration, the solute was fixed by applying restraints with force constants 500 kcal/mol Å^–2 while the solvent and ions were minimized for 5000 steps using steepest descent (SD) and equilibrated with MD at 298.15 K (25°C) for 500 ps under NVT conditions, followed by 500 ps under NPT conditions at atmospheric pressure. After a series of minimizations in which the solute was tethered to its initial coordinates with the restraint force constant decreasing from 100 to 50 to 10 to 0 kcal/mol Å^–2, the system was heated by MD under NPT conditions to 298.15 K in 10 ps 50 K increments followed by 50 ps equilibration intervals. After heating to 298.15 K, the system was equilibrated by MD under NPT conditions for 5 ns, followed by equilibration by MD under NVT conditions for 10 ns. To sample the hydrogen-bonding motif, a 200 ns production MD trajectory was propagated under NVT conditions. Bonds involving hydrogen were constrained with the SHAKE algorithm⁶¹ for the MD. The NVT and NPT MD simulations were performed using a 1 fs timestep and Langevin dynamics^{62, 63} with a collision frequency of 2 ps⁻¹, and the Berendsen barostat⁶⁴ was used for the NPT simulations during equilibration. The electrostatics were treated using the particle mesh Ewald⁶⁵ method, and the nonbonded cutoff was 10 Å.

RESULTS AND DISCUSSION

Building the Cu-MOF Model

A minimal functional model of the Cu-BTTri MOF was determined using DFT to investigate the stability of the various models, as well as their ability to serve as a functioning catalyst via the mechanisms explored herein. The initial model systems studied consisted of a single copper site and model linkers. These systems, which are provided in the SI (Fig. S1), proved to be inadequate to retain the geometry of the copper sites present in the MOF.

A more extensive, three-copper cluster model was then constructed. Because no crystal structure is available for the Cu-BTTri MOF, the cluster was created by extracting the structure circled in Figure 2 from the isostructural crystal structure of the tetrazolate-based MOF H[Cu(DMF)₆][(Cu₄Cl)₃(BTT)₈–(H₂O)₁₂]·3.5HCl·12H₂O·16CH₃OH (Cu-BTT).⁷ The matching powder X-ray diffraction (P-XRD) patterns^{6, 7} of the Cu-BTT MOF shown in Figure 2 and the Cu-BTTri MOF used in the present study demonstrates that these two Cu-MOFs are isostructural despite the tetrazole linkers in the BTT MOF differing from the triazole linkers in the BTTri MOF. This allowed for the conversion of the Cu-BTT cluster to a Cu-BTTri cluster, the process of which is described below.

Figure 2.

The crystal structure of the tetrazolate-based Cu-BTT, from which the 3-Cu cluster (circled) was extracted to build a minimum surface-site model for the isostructural Cu-BTTri that is the basis for the present computational studies. Note that the surface nitrogen atoms were removed to create the cluster models. Color scheme: coral: copper; green: chloride; blue: nitrogen; red: oxygen; gray: carbon; white: hydrogen.

The crystal structure of the tetrazolate-based Cu-BTT cluster was converted to Cu-BTTri by substitution of the pore-facing nitrogen atoms with carbon atoms. The P-XRD patterns indicate that the triazole moieties bound to a given copper must be equivalent. Given that the P-XRD pattern of Cu-BTTri must match that of the tetrazolate-based MOF, equivalent nitrogen atoms must be complexed to a given copper center. Therefore, only two orientations are possible: either the nitrogen atoms facing the center copper must be replaced by carbon (N-surface), or the outer nitrogen atoms facing solvent must be replaced by carbon (N-pore). Geometry optimizations using both DFT functionals predicted contortions of the cluster in the N-surface orientation. This prediction is illustrated in Figure 3A, which shows the sterically clashing HCCH motifs. However, the N-surface orientation is predicted to be 2–3 kcal/mol lower in free energy than the N-pore orientation (Table S1). Despite this thermodynamic comparison, the structural contortions deviating away from the crystal structure in the N-surface orientation, the resulting high strain to the surrounding framework, and the stabilizing hydrogen bond present in the N-pore orientation upon protonation of a triazole led us to choose the N-pore orientation for our subsequent calculations. Additionally, the discussion in section 3.3 of the SI strongly supports the N-pore orientation for the S-complexed Cu(I) mechanism, which was found to be the only energetically viable mechanism that we explored (see below).

Figure 3.

Two Cu-BTTri MOF models examined computationally. Note that the view in this figure is from the back of the 3-Cu cluster, which has been rotated sideways with respect to the view in Fig. 2. Color scheme: coral: copper; green: chloride; blue: nitrogen; gray: carbon; white: hydrogen. (A) The model with the non-ligating nitrogen atoms facing the surface (N-surface orientation) exhibits steric clashing of HCCH motifs and lacks the stabilizing N–H---N hydrogen bond, leading to significant distortions of the MOF from the isostructural reference. (B) The model with the non-ligating nitrogen atoms facing each other (N-pore orientation) exhibits a N–H---N hydrogen bond (dashed line) that contributes to the cluster stability. The model shown in (B) was chosen for the mechanistic studies. In these models, two surface copper centers are bridged to a pore copper center by four triazole ligands, and all three copper centers are coordinated to a chloride ion.

Based on this analysis, the MOF cluster model used for this study contains triazoles with inner nitrogen atoms. Elemental analysis of the tetrazolate-based MOF has revealed a ratio of one proton to every eight triazole linkers.^6–8 Therefore, the 3-Cu cluster, which contains four triazoles, should contain either one or no protonated triazoles. We therefore examined all proposed mechanisms for the 3-Cu cluster with no protonated triazoles and with a single protonated triazole. Upon protonation of one of the triazoles, this model is stabilized by the N–H---N hydrogen bond shown in Figure 3B. The choice of the nitrogen atom to protonate has minimal effect (< 0.5 kcal/mol) on the free energy of the cluster, both when the cluster has GSNO and GSH complexed to it, as indicated by the experimental rate law, and when only water molecules are coordinated. The results for the singly-protonated cluster model are shown in the main text, and the results for the deprotonated cluster are reported in the SI (Figs. S5–S8). The protonated model is necessary for the S-complexed Cu(I) mechanism because the Cu(I) intermediates do not retain the structure without the proton on the triazole linker, as discussed below.

GSH and GSNO Models

The considerable conformational flexibility of the tripeptides GSH and GSNO necessitated their truncation to facilitate the calculations. This simplification is not expected to influence the electronic structure at the Cu-MOF surface active site. GSH and GSNO were shortened in a manner that retained the relevant nitrogen and carboxylate groups, resulting in cysteine and S-nitrosocysteine models for GSH and GSNO (Fig. 4A). The thiol proton of GSH could potentially be involved in a PCET reaction during the catalytic mechanism, as indicated by experiments observing inhibition of reactivity at high pH values.⁹ The computed proton affinities of the model GSH moiety and the protonated triazole were found to be virtually identical. Hence, both are predicted to deprotonate within the same pH range, consistent with their respective experimental pK_a values of 8.8 and 9.3 when not ligated to copper.^{16, 17} Specifically, the computed free energies of deprotonation for model GSH and protonated triazole when ligated to copper differ by only 0.26 or 1.1 kcal/mol when computed with the B3P86 or TPSSh functional, respectively. Moreover, the carboxylic acid pK_a of GSH (3.8) is lower than the pH of the experimental buffer (4.5), so it is possible the thiol proton diffuses to buffer in the case of proton transfer from sulfur to the carboxylate in GSH. This possibility is explored in one of the mechanisms discussed below.

Figure 4.

Models of the GSH and GSNO ligands. (A) The truncated GSH (left) and GSNO (right) models used in the electronic structure calculations, where the S–H---O motif is circled in GSH. (B) The full GSH tripeptide readily samples the S–H---O hydrogen-bond conformation present in the truncated GSH model, according to molecular dynamics simulations. Color scheme: blue: nitrogen; red: oxygen; yellow: sulfur; gray: carbon; white: hydrogen. Both species are neutral.

Compared to the full tripeptide, the GSH model is biased to exhibit an S–H---O hydrogen bond motif as shown in Figure 4. To establish whether the full tripeptide could access this motif, the full GSH tripeptide was simulated with molecular dynamics using a molecular mechanical force field. This hydrogen bond motif was defined in terms of an S—O distance less than 3.4 Å and an SHO angle greater than 135° (Fig. 4B). Using this definition, the S–H---O hydrogen bond motif was sampled by the full GSH peptide for 0.5% of a 200 ns trajectory, corresponding to sampling this motif approximately every nanosecond. This average timescale is much faster than the experimentally determined timescale of catalysis, which is on the order of milliseconds.⁹ Thus, this hydrogen bond would be accessible if it were required for a PCET mechanism.

Investigating Possible Catalytic Mechanisms

Next, we investigated three possible catalytic mechanisms motivated by the experimental data and the six possible mechanisms considered previously based on the experiments.⁹ In the first computationally studied mechanism, denoted the N-complexed mechanism, GSNO coordinates to the copper center via its nitrogen, and Cu remains in the +2 oxidation state throughout the cycle. In the second mechanism, denoted the S-complexed mechanism, GSNO coordinates to the Cu center via its sulfur, and again Cu remains in the +2 oxidation state. In the third mechanism, denoted the S-complexed Cu(I) mechanism, GSNO coordinates to the Cu center via its sulfur, and Cu is converted to the +1 oxidation state during the catalytic cycle. Both DFT functionals predicted stable complexation of GSH, GSNO, and NO• for the N-complexed mechanism, as well as stable complexation of GSH, GSNO, and GS• for the S-complexed mechanisms. Moreover, for all three mechanisms, the proton is found to move from the sulfur to the carboxylate oxygen of copper-bound GSH for the entire catalytic cycle. Note that the carboxylate group retains the proton in these calculations, although under experimental conditions the proton would be transferred to the pH = 4.5 buffer. Investigation of the S-complexed Cu(I) mechanism with deprotonation of GSH upon coordination to copper indicated that this protonation state does not qualitatively alter the free energy landscape (Figures S10 and S11).

To provide a benchmark for the mechanisms investigated, we employed the energetic span model of Kozuch and Shaik^{66, 67} to obtain the energetic span from the turnover frequency (TOF) extracted from the experimental data. For the case of a first-order kinetic rate law with respect to the concentration of GSNO, denoted [GSNO], the standard TOF can be expressed as⁶⁷

$$\text{TOF}°= \frac{1}{2}\frac{1}{\left[\text{Cu}\right]}\frac{c°}{\left[\text{GSNO}\right]}\left|\frac{\text{d}\left[\text{GSNO}\right]}{\text{d}t}\right|$$ (1)

where c° denotes the standard reactant concentration of 1 M. After using Eq. (1) to extract the value of the TOF° from the experimental data, the energetic span of the cycle was determined from the expression

$$\text{TOF}°\approx \frac{k_{\text{B}}T}{h}\text{exp}\left[-\frac{\delta \epsilon }{k_{\text{B}}T}\right]$$ (2)

where k_B is the Boltzmann constant, T is the temperature, h is Planck’s constant, and δε is the energetic span. Following this procedure, the experimentally measured rate of GSNO consumption at GSH saturation was found to correspond to an energetic span of 15.6 kcal/mol. For the cycles considered, the energetic span obtained from the DFT calculations is the difference in free energy of the highest free energy intermediate and the lowest free energy intermediate that occurs prior to the highest intermediate. Note that this span is a lower limit because the transition state will be higher in free energy than the intermediate with the highest free energy.

In the N-complexed mechanism (Fig. 5), GSH coordinates to the copper center, followed by coordination of GSNO to Cu by its nitrogen atom and subsequent S–N bond homolysis, resulting in a complexed NO radical and, importantly, a dissociated GS• radical. In this mechanism the GS• radical then rapidly combines with another GS• radical to form the observed GSSG disulfide, and NO• leaves as a second product. The Cu-BTTri cluster model was able to form all putative intermediates and predicts favorable coordination of GSH and GSNO, as implicated by the experimentally determined rate law. Population analyses indicate that all copper centers retain a +2 oxidation state and a ferromagnetic spin state for the three copper atoms throughout the cycle (Table S2). These properties are demonstrated by the significant NBO spin density on each copper site, the >1 APT charge for each copper center (excluding the open-shell singlet species, which was not found to participate in the mechanism based on its higher free energy), and the 9 d-electrons on each copper site.

Figure 5.

N-complexed mechanism for the Cu-BTTri model with one protonated triazole as shown in Figure 3B. In free GSH, the thiol proton is located on sulfur (top middle). In computations, the proton is found to move from the sulfur and remain on the carboxylate oxygen of GSH for the entire catalytic cycle in this proposed mechanism, neglecting proton transfer to the buffer. Subsequent complexation of GSNO via the nitrogen (right intermediate) leads to homolytic S–N bond cleavage (transition state, bottom), resulting in dissociation of GS• (bottom left) in this mechanism. Dissociation of NO• (upper left) restores the activated catalyst.

In this mechanism, however, the free energy difference (ΔG) between the NO• coordinated complex and the thermodynamically stable GSH and GSNO coordinated complex is too high (ca. 40 kcal/mol, Fig. 6) to be consistent with the estimated energetic span of the cycle of 15.6 kcal/mol based on the experimental data.⁹ As such, this step is most likely thermodynamically and kinetically inaccessible, suggesting that this mechanism is unviable. Note that for all mechanisms studied, the catalyst is activated after the initial coordination of GSH to copper, and thus the catalytic cycle does not include the starting Cu(II)(2H₂O) species.

Figure 6.

Calculated free energy landscape of the N-complexed catalytic cycle. The free energies are calculated relative to the sum of the free energies of the initial species. Cu(II) represents the active copper site in oxidation state +2 in the 3-Cu model cluster. Species in parentheses are complexed to the active copper site. Symbols within the square brackets denote the local spin state formed when the radical species is ligated to the active copper site. The Cu-NO local triplet state (Cu-NO[T]) is shown because it was found to have the lowest free energy. The landscape predicts favorable coordination of GSH and GSNO but includes a significantly uphill step of ca. 40 kcal/mol leading to S–N bond cleavage. Hence, this mechanism, which was previously favored⁹ on the basis of the available evidence at the time, is predicted to be unlikely based on the computations.

Alternatively, mechanisms analogous to Figure 5 can be drawn where GSNO coordinates to Cu(II) through sulfur.⁹ The S-complexed mechanism, in which GSNO coordinates to Cu by sulfur (Fig. 7), is consistent with previous work⁶⁸ reporting that complexation of S-nitrosothiols to copper by sulfur is less thermodynamically favorable, but weakens (i.e., elongates) the S–N bond. This alternative mechanism involves the dissociation of NO• rather than GS• first upon S–N cleavage, resulting in a lower ΔG associated with the S–N homolysis step. Population analyses indicate that all copper centers retain a +2 oxidation state and a ferromagnetic spin state throughout the cycle (Table S3) using the same criteria as described above for the N-complexed mechanism. Ultimately, the intermediate following S–N homolysis was too high in energy for this mechanism to be viable (Fig. 8). A GS• + GS• radical recombination to form GS-SG over adjacent copper sites was also considered but deemed unlikely given that the surface copper sites are separated by over 5 Å, and thus the barrier for the GS• recombination would likely be high. Other mechanisms, such as those involving N- and S-complexed intermediates complexed to neighboring copper sites, are conceivable but are beyond the scope of the present study.

Figure 7.

S-complexed mechanism for the Cu-MOF model with one protonated triazole as shown in Figure 3B. In this mechanism, GSNO coordinates to copper via sulfur rather than nitrogen (right intermediate). This intermediate leads to NO• dissociation upon S–N homolysis (transition state, bottom), followed by GS• dissociation (upper left). As observed for the N-complexed mechanism, the proton remains on the carboxylate oxygen of GSH for the entire catalytic cycle, neglecting proton transfer to the buffer.

Figure 8.

Calculated free energy landscape of the S-complexed catalytic cycle. The free energies are calculated relative to the sum of the free energies of the initial species. Cu(II) represents the active copper site in oxidation state +2 in the 3-Cu model cluster. Species in parentheses are complexed to the active copper site. Symbols within the square brackets denote the local spin state formed when the radical species is ligated to the active copper site. The Cu-GS local triplet state (Cu-GS[T]) is shown, as the corresponding open-shell singlet was 0.8 kcal/mol higher in energy. Complexation of GSNO via sulfur reduces the negative change in free energy upon coordination of GSNO relative to the N-complexed mechanism (Figs. 5 and 6). However, the free energy landscape for this mechanism again includes a prohibitively large ΔG upon GS• dissociation, and therefore this mechanism is predicted to be unlikely.

The final, thermodynamically favored and hence proposed mechanism is the S-complexed Cu(I) mechanism. In this case, a GSSGH⁺ reductive elimination at Cu circumvents the thermodynamic cost of dissociating GS• by the direct formation of a disulfide bond (Fig. 9). Population analysis of this mechanism strongly indicates a reduction of Cu(II) to Cu(I) upon GSSGH⁺ elimination, as demonstrated by the annihilation of spin density on the active copper, and the significant reduction in APT charge (Table S4). Unlike the other mechanisms, the free energy landscape of this mechanism (Fig. 10) avoids the thermodynamically unfavorable GS• dissociation step — a key finding from the present studies. Instead, a ligand-ligand reductive elimination step produces a Cu(I) intermediate that is able to complex a second equivalent of GSNO, accomplish a second S–N cleavage (releasing a second NO), and regenerate, post protonation, the resting state. The significantly smaller free energy span within this catalytic cycle provides strong support for its likelihood over the other two proposed mechanisms.⁶⁹ Although the rate law cannot be determined from the thermodynamic landscape, the highest reaction free energy in Figure 10 corresponds to the first S–N cleavage step, which therefore most likely has the highest activation free energy. If this assumption were correct, this mechanism would be consistent with a first-order dependence on [GSNO] in the rate law, even though it involves a second GSNO.

Figure 9.

S-complexed Cu(I) mechanism for the model with one protonated triazole as shown in Figure 3B. This mechanism involves coordination of GSNO to Cu(II) via sulfur (top right intermediate), followed by dissociation of NO• (bottom right intermediate). A subsequent copper reduction-coupled GSH-GS• reductive elimination reaction results in a formally Cu(I) intermediate (bottom intermediate). To complete the catalytic cycle, complexation of a second GSNO molecule (bottom left intermediate) followed by Cu(I) to Cu(II) oxidation-assisted S–N homolysis, results in production of a second NO• (top left intermediate). Protonation of the resulting anionic ⁻GS ligand restores Cu-GSH. The proton remains on the carboxylate oxygen of GSH for the entire catalytic cycle, neglecting proton transfer to the buffer. See Figure S10 for an alternative related mechanism in which the proton transfers to the buffer in the first GSH coordination step at the top.

Figure 10.

Calculated free energy landscape of the S-complexed Cu(I) catalytic cycle. The free energies are calculated relative to the sum of the free energies of the initial species. Cu(X) represents the active copper site in oxidation state X in the 3-Cu model cluster. Species in parentheses are complexed to the active copper site. Symbols within the square brackets denote the local spin state formed when the radical species is ligated to the active copper site. The Cu-GS local triplet state (Cu-GS[T]) is shown, as the corresponding open-shell singlet was 0.8 kcal/mol higher in energy. Noteworthy in this mechanism is that a step producing free GS• is avoided by an elimination reaction at Cu forming GSSGH⁺. Hydration of Cu(I) by one water molecule, rather than two water molecules, following GSSGH⁺ elimination, avoids a 20-electron species, as discussed in Section 3.3 of the SI.

Based on the S-complexed Cu(I) mechanism shown in Figure 10, the calculated energetic span of the cycle was determined to be 14.8 kcal/mol, which is the difference in free energy of the Cu(I)(H₂O) intermediate and the Cu(II)(GSH)(GSNO) intermediate within the catalytic cycle. Thus, this computed value is consistent with the energetic span of 15.6 kcal/mol extracted from experiment. Moreover, as our estimate of the energetic span from the experimental data assumed a first order dependence on [GSNO], this agreement indirectly corroborates the proposed first-order dependence of the rate law on [GSNO].⁹ Additional calculations for an alternative related mechanism in which GSH deprotonates upon coordination to copper are qualitatively consistent with the S-complexed Cu(I) mechanism shown here (Figures S10 and S11). Thus, the S-complexed Cu(I) mechanism is consistent with the key experimental findings: a stable yet active Cu-BTTri MOF surface catalytic site, the observed rate of GSNO consumption, the catalytic cycle, inhibition of reactivity at high pH values that could be due to triazole deprotonation, and the experimentally observed rate law.

CONCLUSIONS

The calculations presented herein provide new insights into the mechanism by which the Cu-BTTri MOF generates NO• from endogenous sources. The computational studies answered critical questions that remained from the experimental work and provided evidence inconsistent with the previously favored mechanism involving only Cu(II) and a single GSNO in the catalytic cycle. The currently proposed S-complexed Cu(I) mechanism is consistent with all of the available experimental evidence.^9–12 This mechanism invokes GSH coordination to copper through its sulfur atom as well as a Cu(I) intermediate. The computational studies also elucidated the critical role of protons and their probable sites of action in the mechanism. The GSNO S–N cleavage process is enabled by proton transfer, but not in the context of a concerted PCET mechanism. Protonation of one of the triazole linkers is found to be essential for catalysis because it stabilizes the Cu-MOF model cluster throughout the catalytic cycle, especially the Cu(I) intermediate. The calculations therefore suggest that the inhibition of reactivity observed experimentally above a pH of ca. 5.6 may arise from structural destabilization of the active site via deprotonation of the triazole linker. Note that larger MOF cluster models or periodic MOF models may impart structural stability to the deprotonated Cu(I) intermediate. Understanding the precise site of protonation and the molecular effects of removing that proton is difficult to explore experimentally. Additionally, the critical finding that a step resulting in dissociation of free GS• is thermodynamically prohibitive in the present model system is noteworthy and may be of broader significance and applicability.

The cluster model used in this work has been shown to be sufficient in that geometry optimizations do not deviate significantly from the isostructural tetrazolate-based MOF crystal structure, and the energetic span of the most thermodynamically favorable mechanism is consistent with the experimental data. However, this model does not account for other possible mechanistic scenarios. For example, neighboring MOF components could participate in catalysis, although the distance between copper centers may prevent direct participation. Moreover, various types of MOF defects could play a critical role in catalysis.^70–72 Our findings do not preclude these possibilities, but rather provide a relatively simple mechanistic hypothesis for this complex catalytic system.

The combination of these calculations and the previous experimental work^9–11 has led to an enhanced mechanistic understanding of the heterogeneous catalytic process in which GSNO is converted to NO• and GSSG by the Cu-BTTri MOF. This understanding is expected to assist in the rational design of related MOF catalysts to generate nitric oxide for a wide range of chemical and biomedical applications and to help guide research and applications of other MOF systems involving Cu redox components or the BTTri ligand system. Moreover, these insights may be more generally applicable to other catalytic processes involving GS• and related thiyl radical intermediates.

ASSOCIATED CONTENT

The Supporting Information is available at https://pubs.acs.org/ and contains: discussion of the initial single copper models; free energy analysis of triazole orientations within the 3-Cu cluster; free energy landscapes of the TPSSh functional and the deprotonated 3-Cu cluster; population analyses of the Cu(II) and Cu(I) intermediates for the singly protonated 3-Cu model computed with the B3P86 functional; spin contamination analyses of the intermediates; calculation of the energetic span based on the experimental rate of GSNO consumption; additional calculations for the S-complexed Cu(I) mechanism; structures of the intermediates.