A W/Cu synthetic model for the Mo/Cu cofactor of aerobic CODH indicates that biochemical CO oxidation requires a frustrated Lewis acid/base pair

Abstract

Constructing synthetic models of the Mo/Cu active site of aerobic carbon monoxide dehydrogenase (CODH) has been a long-standing synthetic challenge thought to be crucial for understanding how atmospheric concentrations of CO and CO₂ are regulated in the global carbon cycle by chemolithoautotrophic bacteria and archaea. Here we report a W/Cu complex that is among the closest synthetic mimics constructed to date, enabled by a silyl protection/deprotection strategy that provided access to a kinetically stabilized complex with mixed O²⁻/S²⁻ ligation between (bdt)(O)W^VI and Cu^I(NHC) (bdt = benzene dithiolate, NHC = N-heterocyclic carbene) sites. Differences between the inorganic core’s structural and electronic features outside the protein environment relative to the native CODH cofactor point to a biochemical CO oxidation mechanism that requires a strained active site geometry, with Lewis acid/base frustration enforced by the protein secondary structure. This new mechanistic insight has the potential to inform synthetic design strategies for multimetallic energy storage catalysts.

Graphical Abstract

INTRODUCTION

Chemolithoautotrophic bacteria and archaea metabolize carbon monoxide (CO) as their source of both carbon and energy, accounting for consumption of 10⁸ tons/year of CO that helps maintain homeostasis in the global carbon cycle.¹ Although CO poisons most organisms through irreversible inhibition of metalloproteins, these particular microorganisms feature carbon monoxide dehydrogenase (CODH) metalloenzymes that are able to use CO productively by catalyzing energy release in the form of its 2-electron oxidation to carbon dioxide (CO₂). Several anaerobic bacteria feature CODH enzymes with a Ni/Fe cofactor capable of catalyzing the reversible CO + H₂O « CO₂ + 2e⁻ + 2H⁺ redox reaction in both directions,² while aerobic bacteria and archaea instead employ a Mo/Cu cofactor that is catalytically active only in the oxidation direction.^3,4 Fundamental understanding of the chemical mechanisms that allow CODH enzymes to store energy in inorganic chemical bonds and release it on demand will aid in the design of synthetic catalysts for energy storage and conversion.⁵

Aerobic CODH belongs to the xanthine oxidase family of oxotransferase enzymes that typically have square pyramidal (MCD)(X_eq)Mo^VI(O_ax)(=S) active sites (MCD = molybdopterin cytosine dinucleotide; X = O or OH; ax and eq denote axial and equatorial, respectively) which catalyze oxygen atom transfer by Mo^VI=O_eq/Mo^IV redox cycling using a reactive site in the equatorial plane of the metal coordination sphere.⁶ The unusual feature of the CODH homologue is the presence of a copper ion bound to the S-atom, resulting in a heterobinuclear active site with an unsupported Mo^VI–(μ₂-S)–Cu^I bridge that has no precedent in synthesis or biology (Figure 1a).³ In fact, all other known molybdopterin and tungstopterin enzymes have mononuclear, rather than dinuclear, Mo or W active sites.^6–9 The presence of a heterobinuclear active site in CODH is suggestive of a cooperative mechanism, but the detailed pathway for CO oxidation is still under debate.^10–15 Although synthetic models could provide valuable insights regarding structural and functional aspects of this unusual cofactor, constructing such models has eluded most synthetic efforts for about two decades.^16,17 Here we report synthesis and characterization of (bdt)(O)(X)W^VI–(μ₂-S)–Cu^I(NHC) complexes (bdt = benzenedithiolate, X = O or OSiⁱPr₃, NHC = a N-heterocyclic carbene) that mimic the oxidized (MCD)(O_ax)(X_eq)Mo^VI–(μ₂-S)–Cu^I(S_Cys) cofactor in CODH (X = O or OH, S_Cys = cysteine) in aspects unmatched by other models (Figure 1b). Experimental and computational data from this model system provide new mechanistic insight into the mechanism of CO oxidation by the native cofactor, which we propose uses a frustrated Lewis acid/base pair¹⁸ enforced by the protein secondary structure.

Figure 1.

(a) Active site comparisons of the xanthine oxidase family of metalloenzymes and the aerobic CODH homologue; (b) the synthetic model complexes reported here; previous model systems reported by (c) Young and (d) Holm; and (e) our working hypothesis for kinetic stabilization of an active site model.

RESULTS AND DISCUSSION

Synthetic heterobinuclear Mo/Cu complexes and multimetallic Mo/Cu clusters with multiple S-atom bridges have been known for decades,¹⁶ but the singular example of a synthetic system with unsupported Mo–(μ₂-S)–Cu bridges was reported originally by Kirk and Young in 2006 (Figure 1c).¹⁹ No reactivity with CO has been observed for any derivative of this system,²⁰ consistent with the fact that the equatorial oxygen sites were irreversibly deactivated by arylation. Furthermore, these complexes were isolated in the Mo^V/Cu^I redox state rather than the Mo^VI/Cu^I state needed for CO oxidation in the native enzyme, lacked dithiolate ligation at Mo to mimic the molybdopterin group, and featured coordinatively saturated Cu^I sites that were therefore deactivated towards cooperative CO activation. Synthetic attempts by the groups of Tatsumi,^21,22 Holm,^23,24 and others^25–28 at overcoming some or all of those shortcomings failed to produce unsupported M–(μ₂-S)–Cu bridges (M = Mo or W). The prevailing view from these studies is that M(O_eq)–(μ₂-S)–Cu systems are unstable towards intermolecular O/S disproportionation that produces thermodynamically stable M(μ₂-S)₂Cu cores. As an illustrative example, Holm found that a [(bdt)(O_ax)W(μ₂-S)₂Cu(NHC)]⁻ complex was produced regardless of whether a [Cu(NHC)]⁺ synthon was added to [(bdt)W(O)₂(S)]²⁻ (1) or to [(bdt)W(O)(S)₂]²⁻ (Figure 1d).²³ After several of our own attempts that yielded M(μ₂-S)₂Cu complexes and upon reflecting on these synthetic precedents, our plan was to resolve Holm’s disproportionation problem by deactivating the equatorial oxygen towards disproportionation as demonstrated by Young, only here with a removable protecting group that could be cleaved to reveal the reactive oxygen site only after the M–(μ₂-S)–Cu bridge had been assembled (Figure 1e).

To test our hypothesis, we prepared [(bdt)W(O_ax)(OSiⁱPr₃)(S)]⁻ (1-[Si]) bearing a silyl-protected oxygen that we planned to cleave with fluoride at the opportune time.²⁵ Addition of (IPr)CuOTf to 1-[Si] (Figure 2a) formed a major product and two minor products as indicated by ¹H NMR spectroscopy (IPr = N,N’-bis(2,6-di-iso-propylphenyl)imidazol-2-ylidene). Crystallization allowed us to characterize the major product as orange-colored (O_ax)(ⁱPr₃SiO)W(μ₂-S)(μ₂:k²,k¹-S-bdt)Cu(IPr) (2) in 55% yield, and the composition of 2 was confirmed by observing the parent [M]⁻ ion (m/z = 996.1 and 998.1) by MALDI-TOF mass spectrometry (Figure S9). Although our protecting group strategy had succeeded in preventing O/S disproportionation in the major product, to our surprise the solid-state structure of 2 featured one of the thiolate sulfurs from the bdt group occupying a bridging position between W^VI and Cu^I (Figure 2b). Further crystallization from the mother liquor provided green crystals of a minor product, (bdt)(ⁱPr₃SiO)W(μ₂-S)₂Cu(IPr) (3), indicative of the unwanted disproportionation process (Figure S1). We assume the mass balance from this reaction to be (bdt)(ⁱPr₃SiO)W(μ₂-O)₂Cu(IPr) or fragments thereof, though we have been unable to grow X-ray quality crystals of any other minor products to date. Complex 2 exhibited charge-transfer transitions in the UV-Vis spectrum at 341, 387, and 460 nm, which are shifted slightly from the values for 1-[Si] of 300, 350, and 490 nm.²⁵ Similarly, complex 3 exhibited transitions at 332, 418, 480, and 602 nm that are shifted only slightly from the values for [(bdt)(^tBuPh₂SiO)W(μ₂-S)₂]⁻ of 332, 426, 488, and 588 nm. These modest perturbations to the W^VI ligand-to-metal charge transfer transitions indicate that the anionic [(bdt)W^VI]⁻ fragments are not significantly perturbed upon coordination to cationic [Cu(IPr)]⁺.

Figure 2.

(a) Synthetic methodology; solid-state structures (C and N as wireframes, H not shown, all other atoms at 50% probability ellipsoids) of (b) silyl-protected 2 (only one of four molecules from the asymmetric unit shown) and (c) deprotected anion 4 (only the major component of disordered S–W–O unit shown).

Low-temperature addition of [Et₄N][F] to 2 provided orange-colored [(bdt)(O_ax)W(μ₂-O_eq)(μ₂-S)Cu(IPr)][NEt₄] (4), which was isolated in 45% recrystallized yield. The orange color of 4 arises from absorptions at 340 and 390 nm, which if present in the enzyme system would be masked by intense absorptions in the 350–550 nm range from the [2Fe-2S] electron transfer sites and the FAD cofactor contained within the aerobic CODH scaffold.³ The optical transitions for 4 are shifted only slightly compared to those for [(bdt)W(O)₂S]²⁻ at 320 and 405 nm,²⁵ again indicative of the relatively minor perturbation in W^VIO₂S electronic structure upon interaction with [Cu(IPr)]⁺ in 4. The solid-state structure of 4 showed that, upon oxygen deprotection, an equatorial rearrangement had occurred by which one of the W^VI=O units displaced the bridging bdt thiolate from 2 to occupy the bridging position between W^VI and Cu^I (Figure 2c). Complex 4 is among the closest synthetic models for the Mo/Cu cofactor in aerobic CODH reported to date, differing only by the substitution of Mo for W (due to lack of synthons for the Mo analogue of 1-[Si]) and the presence of a short M^VI=O_eq→Cu^I dative interaction that is absent in the native system (vide infra). Although several binuclear complexes with mixed μ-S²⁻/μ-O²⁻ ligation have been structurally characterized before, including dimolybdenum examples,^29–31 to our knowledge complex 4 is the first heterobinuclear M(μ-O)(μ-S)M’ complex to be structurally characterized. It should be noted that molybdenum and tungsten complexes with identical ligand sets tend to be isostructural/isometric^32,33 and often exhibit similar chemical behavior, as evident from the related Mo and W members of the oxotransferase class of enzymes to which aerobic CODH belongs.^6,7 The W^VI(μ₂-S)(μ₂-O)Cu^I core of 4 exhibited significant positional disorder crystallographically. However, the possibility of 4 being a W^VI(μ₂-S)₂Cu^I complex was ruled out by crystallographic refinement (Table S5) and successful observation of the parent [M]⁻ ion (m/z = 840.6 and 842.6) by MALDI-TOF mass spectrometry (Figure S10). Additionally, 4 lacks the notably intense, low-energy optical transition at ~480 nm that is characteristic of W^VI(O)(S)₂ complexes.²⁵

Examining the metrical parameters (Table 1) for the inorganic cores of complexes 2, 3, and 4 revealed clear trends as the bridging ligand opposite the μ₂-S²⁻ bridgehead were varied from a bdt thiolate in 2 to a μ₂-S²⁻ in 3 to a μ₂-O²⁻ in 4: the W···Cu interatomic distance decreased steadily from 2.757(2) Å to 2.581(2) Å, and the W–(μ₂-S)–Cu angle decreased steadily from 77.0(1)° to 69.0(2)°. Due to the acute sulfur angle in 4, a M^VI=O_eq→Cu^I dative interaction is evident from the 2.377(8)-Å Cu···O_eq distance. (It should be noted that crystallographic disorder makes bond metrics at the light μ₂-O-atom within the inorganic core of 4 somewhat unreliable. See Supplementary Information for further details.) The ultimate result is that the inorganic core of 4 takes on a “closed” form that is distinct from the Mo/Cu cofactor in the CODH enzyme characterized by Dobbek from oligotropha carboxidovorans,³ which has an “open” form with an obtuse Mo–(μ₂-S)–Cu angle of 113°, a long Mo···Cu interatomic distance of 3.74 Å, and no significant Mo^VI=O_eq→Cu^I interaction evident from the 3.36-Å Cu···O_eq distance. The structure of 4 can be viewed as the relaxed, ground-state structure of this inorganic core in the absence of protein matrix effects. The structure of the native cofactor in CODH, by contrast, can be viewed as a strained geometry whose strain energy is compensated for by non-covalent effects derived from the protein secondary structure (vide infra).

Table 1.

Comparative structural metrics

Complex	∠[M–(μ2-S)–Cu] (°)	d[M···Cu] (Å)a	d[Cu–(μ2-Xeq)] (Å)b
2 (M = W)	77.0(1)	2.758(2)	2.242(2)
3 (M = W)	72.449(7)	2.6194(9)	2.206(2)
4 (M = W)	69.0(2)	2.581(2)	2.377(8)
CODH cofactor (M = Mo)c	113	3.74	3.36

^aNon-bonding interatomic distance.

^bX_eq is the bridging ligand across from the conserved μ₂-S²⁻ ligand: S_bdt for 2, S²⁻ for 3, O²⁻ for 4.

^cData taken from PDB 1N5W.

To further probe the strain energy associated with distorting the active site from the closed form seen in 4 to the open form seen in CODH, we initiated a DFT study at the BVP86/LANL2DZ ECP (Cu/W)/6–311+G(d) level of theory³⁴ with implicit THF solvation (IEFPCM model). A truncated model for 4 in which the 2,6-di-iso-propylphenyl groups were replaced with methyl groups was used. The computationally optimized structure was found to have a W–(μ₂-S)–Cu angle of 72°, closely matching the experimentally determined value of 69.0(2)° for 4. The optimized coordinates also reinforced the W^VI=O_eq→Cu^I formulation (vs. W^VI–O–Cu^I) of the bonding, with the W=O–Cu unit calculated to be unsymmetrical with short W-O and long Cu-O distances (d_WO = 1.81 Å, d_CuO = 2.10 Å). Despite removal of the bulky NHC substituents, the DFT model overestimated the W···Cu interatomic distance at 2.71 Å, still well below the observed distance of 3.36 Å in the native cofactor. Next, we performed a scan by varying the W–S–Cu angle and allowing the other geometric parameters to optimize freely at each point. As expected, opening this angle to more obtuse values correlated with increased W···Cu and Cu···O distances (Figure 3a). The W···Cu interatomic distance was found to increase monotonically over a 65–125° range of angles at sulfur. On the other hand, a discontinuity in Cu···O distances was observed between sulfur angles of 100° and 105°, corresponding to a conformational change in which the equatorial oxygen atom rotated away from the Cu^I and the bdt ligand rotated towards the Cu^I once the Cu···O_eq interaction was sufficiently weak (>3 Å). The conformation at acute sulfur angles (≤100°) resembles the solid-state structure of 2 indicating onset of S(bdt)→Cu^I interaction, while the conformation at obtuse angles (≥105°) resembles the solid-state structure of 2. In terms of energetics, a global minimum was located at a sulfur angle of ~72°, and a higher-energy local minimum was found at ~105° (Figure 3b). The energy surface is relatively shallow, consistent with the large error in calculated W···Cu interatomic distance compared to the experimentally determined value. Based on this angular dependence, the difference in energy between the global minimum and the 113° angle observed in the Mo/Cu cofactor of CODH is 4–5 kcal/mol. This value estimates a lower limit for the strain energy present in the native cofactor, which is likely to be higher due to the inability of the enzymatic system to undergo such dramatic conformational changes. For this reason, all further calculations compared the “closed” form at the global minimum to the “open” form at a W–(μ₂-S)–Cu angle of 100°, just prior to the conformational change in the model case. The optimized structure of hypothetical [(bdt)(O_ax)W(O_eq)(μ-S)Cu(SMe)]²⁻ (Table S4) was found to be very similar to the monoanionic NHC analogue with a Cu···O_eq distance of 2.08 Å, indicating that the slight π-accepting character of the NHC ligand does not play a significant role in stabilizing conformer with the “closed” core.

Figure 3.

DFT models: (a) dependence of Cu···O and W···Cu distances and (b) energy on W–S–Cu angle; (c) CO binding models for the closed (unrestricted S angle) and open (100° angle at S) forms.

Two mechanisms have been proposed to account for CO oxidation catalyzed by the Mo/Cu cofactor of CODH.⁴ Based on crystallographic studies with the substrate analogue CNⁿBu, Dobbek proposed that CO activation occurs by insertion into the Cu–S bond to generate a Mo(μ-CO₂S)Cu intermediate that subsequently undergoes hydrolysis to expel CO₂ or bicarbonate.³ That hypothesis has gained further support from recent QM/MM computational modeling.^14,15 By contrast, purely quantum mechanical calculations have indicated that such a thiocarbonate species likely represents an off-cycle thermodynamic well in the reaction energy landscape.^10,11,13 Instead, those studies proposed a mechanism in which the Cu^I site binds CO, and then the proximal Mo^VI=O_eq unit adds to the bound Cu^I–CO unit at carbon to produce a Mo(μ₂-S)(μ₂:O,C-O₂C)Cu intermediate that expels CO₂. Recently, spectroscopic studies have provided evidence for a partially reduced Mo^V(OH)–(μ₂-S)–Cu^I–CO intermediate consistent with the latter proposal.¹² Our collected observations provide further evidence for this pathway from the perspective of synthetic modeling. Exposure of complex 4 to CO at a range of temperatures (up to 40°C) and pressures (up to 6 atm) failed to produce any detectable reaction according to ¹H NMR spectroscopy, presumably because the electrophilic Cu^I site and the nucleophilic W^VI=O unit have quenched each other in the “closed” configuration. In the native cofactor, this Lewis acid/base pair is kept separated in the “open” form, enabling cooperative behavior wherein the Cu^I site acts as a Lewis acid to coordinate CO and the Mo^VI=O_eq unit acts as a Lewis base towards the bound CO. Consistent with this proposal, coordination of CO to the Cu^I site in the “open” model was associated with a calculated binding free energy of ΔG = 3.9 kcal/mol (ΔH = −7.2 kcal/mol), in good agreement with the computed value of 7.9 kcal/mol for CO binding to Cu^I for the native enzyme system.¹¹ A key structural feature is that the CO-bound structure in the closed form does not position the CO ligand near the W=O_eq group, whereas in the open form the CO and W=O_eq groups are well positioned to form a new C-O bond for ultimate CO₂ liberation without requiring much transition-state reorganization (Figure 3c). The Mo/Cu cofactor of CODH can thus be viewed as a Mo^VI=O/Cu^I frustrated Lewis pair (FLP),¹⁸ where the frustration is enforced by the protein secondary structure preventing acid/base quenching. Ison has studied systems in which metal-oxo units serve as Lewis bases in FLPs,³⁵ we have shown that Cu^I can act as the Lewis acid in FLPs,³⁶ and Erker has studied FLP-mediated CO activation processes.^37–39 Thus, it is reasonable to propose cooperative CO activation by a FLP system comprised of a Mo^VI=O/Cu^I frustrated pair.

Analyzing the closed and open forms according to the quantum theory of atoms in molecules (QTAIM)⁴⁰ provided further insight into the FLP hypothesis, in particular by identifying bond critical points (BCPs) and ring critical points (RCPs) in the electron density topology that are associated with covalent bonds and delocalized ring currents, respectively. At the relaxed sulfur angle (Figure 4a), QTAIM analysis confirmed the presence of a BCP between Cu^I and O_eq in the electron density map, consistent with the proposed W^VI=O_eq→Cu^I formulation. Despite the short metal-metal distance in this closed form, no BCP was located along the W···Cu path, indicating an absence of direct metal-metal bonding. Instead, a RCP was located near the centroid of the W^VI(μ-O)(μ-S)Cu^I four-membered core. The large and negative curvature of electron density at this RCP,⁴¹ as well as a relatively low Shannon aromaticity index,⁴² are consistent with significant aromaticity of this inorganic ring that is comparable to the aromaticity of a benzene unit (Figure S11). Upon binding of CO to the closed form of the complex, the QTAIM metrics indicate a loss of aromatic character (e.g. a two-fold decrease in negative curvature at the RCP). In other words, binding of CO to the closed form is accompanied by the penalty of partially breaking aromaticity in the core of the active site. At the sulfur angle of 100°, no BCP was located along the Cu···O_eq path, consistent with rupture of the W^VI=O_eq→Cu^I interaction in the strained geometry (Figure 4b). Furthermore, no RCP was located near the W···Cu path. In other words, binding of CO to the “frustrated” open form does not require any disruption of aromaticity in the active site. In the protein environment, the strain energy associated with Lewis acid/base frustration and with disruption of electron delocalization is likely compensated by secondary structure factors as well as crucial non-covalent interactions in the secondary coordination sphere near the active site. Indeed, in the computational models described above, BCPs in the electron density map were located between the oxygen centers and nearby C-H bonds from the NHC ligand’s N-CH₃ groups, indicating that even weak hydrogen bond donors are able to interact with the W=O groups.

Figure 4.

Electron density maps for the (a) closed and (b) open models. Contour lines are drawn in the S–Cu–O_eq plane in each case. Bond critical points and ring critical points are shown in green and red, respectively, with accompanying electron density values r(r) shown in black.

Several structural aspects of the native CODH structure are in accordance with the hypothesis presented herein (Figure 5a). As mentioned above, the protein binding pocket holds the Mo^VI and Cu^I sites far apart, creating a strained μ₂-S²⁻ geometry. This open geometry prevents meaningful Mo^VI=O_eq→Cu^I interaction, with the resulting charge build-up on the equatorial oxygen being stabilized by hydrogen bond donation from Glu763, a universally conserved residue in all xanthine oxidases, and from Gly569 (Figure 5b). Coordinative unsaturation at the Cu^I site is stabilized by a 3.15-Å contact with the N-terminus of its terminal Cys388 ligand. Additionally, a solvent H₂O ligand in the crystal structure of the oxidized enzyme is in position both to stabilize the Cu^I center at a distance of 2.42 Å and/or donate a hydrogen bond to O_eq,⁴ even inducing trigonal Cu^I coordination in some structural variants.^43,44 These weak and non-covalent interactions are necessary for the CODH secondary structure to enforce a frustrated Mo^VI=O/Cu^I Lewis pair. We anticipate that this insight about the special features that activate the Mo/Cu cofactor of CODH towards CO oxidation will inform design of synthetic catalysts for small-molecule activation and energy storage.

Figure 5.

(a) Conceptual schematic of the FLP concept applied to heterobimetallic CO activation; (b) Depiction of secondary coordination-sphere stabilization of the strained geometry in aerobic CODH.