Synthesis, X-ray Structures, Electronic Properties, and O₂/NO Reactivities of Thiol Dioxygenase Active-Site Models

Abstract

Mononuclear nonheme iron complexes that serve as structural and functional mimics of the thiol dioxygenases (TDOs), cysteine dioxygenase (CDO) and cysteamine dioxygenase (ADO), have been prepared and characterized with crystallographic, spectroscopic, kinetic, and computational methods. The high-spin Fe(II) complexes feature the facially-coordinating tris(4,5-diphenyl-1-methylimidazol-2-yl)phosphine (^Ph2TIP) ligand that replicates the three histidine (3His) triad of the TDO active sites. Further coordination with bidentate L-cysteine ethyl ester (CysOEt) or cysteamine (CysAm) anions yielded five-coordinate (5C) complexes that resemble the substrate-bound forms of CDO and ADO, respectively. Detailed electronic-structure descriptions of the [Fe(^Ph2TIP)(L_S,N)]BPh₄ complexes, where L_S,N = CysOEt (1) or CysAm (2), were generated through a combination of spectroscopic techniques [electronic absorption, magnetic circular dichroism (MCD)] and density functional theory (DFT). Complexes 1 and 2 decompose in the presence of O₂ to yield the corresponding sulfinic acid (RSO₂H) products, thereby emulating the reactivity of the TDO enzymes and related complexes. Rate constants and activation parameters for the dioxygenation reactions were measured and interpreted with the aid of DFT calculations for O₂-bound intermediates. Treatment of the TDO models with nitric oxide (NO) – a well-established surrogate of O₂ – led to a mixture of high-spin and low-spin {FeNO}⁷ species at low temperature (−70 °C), as indicated by electron paramagnetic resonance (EPR) spectroscopy. At room temperature, these Fe/NO adducts convert to a common species with EPR and infrared (IR) features typical of cationic dinitrosyl iron complexes (DNICs). To complement these results, parallel spectroscopic, computational, and O₂/NO reactivity studies were carried out using previously-reported TDO models that feature an anionic hydrotris(3-phenyl-5-methyl-pyrazolyl)borate (^Ph,MeTp⁻) ligand. Though the O₂ reactivities of the ^Ph2TIP- and ^Ph,MeTp-based complexes are quite similar, the supporting ligand perturbs the energies of Fe 3d-based molecular orbitals and modulates Fe-S bond covalency, suggesting possible rationales for the presence of neutral 3His coordination in CDO and ADO.

Graphical Abstract

Synthetic models of the thiol dioxygenases (TDOs) have been prepared using a tris(imidazolyl)phosphine ligand that reproduces the atypical three histidine triad of the enzyme active sites. Geometric- and electronic- structure descriptions were obtained with crystallographic, spectroscopic, and computational methods. Dioxygen reactivity experiments confirmed that the complexes behave as functional TDO models, and insights into the O₂ activation mechanism were derived from kinetic studies and DFT. Exposure to NO yields multiple mono- and dinitrosyl iron complexes.

1. Introduction

The catabolism of a diverse array of cellular compounds depends upon mononuclear nonheme iron dioxygenases (MNIDs) that incorporate both atoms of O₂ into the product(s).^1–2 The best-known MNIDs are bacterial enzymes that catalyze the oxidative cleavage of aromatic and aliphatic C-C bonds and play a central role in the degradation of environmental pollutants (i.e., biodegradation).^3–5 MNIDs are also involved in human metabolism, as exemplified by cysteine dioxygenase (CDO) and cysteamine dioxygenase (ADO).^6–8 Both enzymes are thiol dioxygenases (TDOs) that convert an alkylthiol (RSH) to the corresponding sulfinic acid (RSO₂H) using O₂ (Scheme 1). CDO regulates the cellular concentration of L-cysteine (Cys) by performing the first step in the catabolism of this amino acid that leads to the formation of either taurine or sulfate as the end product.^9–10 Insufficient levels of CDO activity cause exogenous Cys to accumulate to deleterious levels, and this metabolic imbalance is associated with several neurological disorders, including Parkinson’s and Alzheimer’s diseases,^11–14 as well as autoimmune disorders.^15–16 ADO is required for the degradation of coenzyme A in mammals, and is the major contributor to the production of hypo(taurine) in the brain where levels of CDO are low.^6,17

Scheme 1.

Beyond its medical importance, CDO has attracted attention due to its atypical active-site structure. The overwhelming majority of MNIDs feature a high-spin Fe(II) center bound by one Asp (or Glu) and two His residues in a facial orientation (i.e., the 2-His-1-carboxylate (2H1C) facial triad).^18–19 In contrast, crystal structures of CDO have revealed a mononuclear Fe site with a neutral 3-histidine (3His) facial triad (Scheme 1).^20–21 The X-ray crystallographic data confirmed that Cys binds directly to Fe in a bidentate manner via its thiolate and amine donors, yielding a five-coordinate (5C) Fe(II) site capable of O₂ binding.²² While a crystal structure of ADO is currently lacking, sequence analysis suggests that it features a similar active-site structure.¹⁷ Since the publication of the first CDO structure in 2006, the 3His triad has also been observed in three microbial MNIDs: β-diketone dioxygenase (Dke1),^23–25 gentisate dioxygenase (GDO),²⁶ and salicylate dioxygenase (SDO).²⁷ The catalytic implications of this alteration in the first-sphere coordination sphere (3His vs. 2H1C triad) have been widely discussed, but no consensus explanation has emerged.²⁸

The first biomimetic models of CDO, prepared by Goldberg and coworkers, were based on 2,6-bis(imino)pyridine frameworks that provide three neutral N-donors in a meridional arrangement.^29–32 These complexes feature an exogenous or pendant aryl thiolate ligand that converts to a sulfonato (RSO₃⁻) donor upon exposure to O₂. The same group has utilized the pentadentate N4S chelate, N3PyS, to prepare TDO models in which the aryl thiolate is appended to a tripodal scaffold.³³ Reaction of the resulting Fe(II) complex, [Fe(N3PyS)(MeCN)]⁺, with O₂ leads to double-oxygenation of the thiolate donor, thereby mimicking the thiolate-to-sulfinate conversion catalyzed by CDO. Treatment of the same complex with nitric oxide (NO) generates a stable {FeNO}⁷ species that exhibits spectroscopic properties similar to those observed for the low-spin (S = 1/2) NO adduct of substrate-bound CDO.^34–35 Recently, Limberg and coworkers have prepared structural and functional CDO and ADO mimics supported by the hydrotris(3-phenyl-5-methylpyrazolyl)borate (^Ph,MeTp⁻) ligand and featuring bidentate L-cysteine ethyl ester (CysOEt) or cysteamine (CysAm) anions, respectively.^36–38 These 5C Fe(II) complexes react with O₂ to give the S-dioxygenated products, as confirmed by experiments involving isotopically-labeled O₂. De Visser and others have explored the O₂-activation mechanisms of CDO and select TDO models using computational methods.^39–42

While these previous biomimetic studies of CDO and ADO have yielded many exciting results, the models reported to date lack either the facial geometry or neutral charge of the 3His triad. As part of our efforts to prepare faithful Dke1 and GDO models, our group has found that easily-prepared tris(imidazolyl)phosphine (TIP) ligands nicely replicate the fac arrangement of three imidazole donors found in the active sites.^43–46 Here, we report the synthesis and structural characterization of two TDO models: [Fe(^Ph2TIP)(CysOEt)]BPh₄ (1) and [Fe(^Ph2TIP)(CysAm)]BPh₄ (2), where ^Ph2TIP is tris(4,5-diphenyl-1-methylimidazol-2-yl)phosphine (Scheme 2). These complexes are related to the TDO mimics of Limberg and coworkers (3 and 4 in Scheme 2); the critical difference is that 1 and 2 feature a neutral TIP scaffold, whereas 3 and 4 employ a monoanionic ^Ph,MeTp⁻ ligand. Our previous studies of Dke1 models indicated that TIP ligands accurately reproduce the coordination environment and donor strength of the 3His triad, whereas the properties of Tp⁻ ligands align better with the 2H1C triad.⁴³ Since the two sets of models are complementary, we have conducted parallel studies of 1/2 and 3/4 with the goal of illuminating the role of 3His coordination in promoting CDO/ADO catalysis. Therefore, this manuscript describes and compares the geometric structures, electronic properties, and O₂ reactivites of complexes 1–4, as revealed through a combination of crystallographic, spectroscopic, kinetic, and computational methods. The similitude of these models to the enzyme active site is evaluated by juxtaposing their structural and spectroscopic features to those reported in previous studies of CDO.

Scheme 2.

Formation of an iron-superoxo intermediate is the putative first step of O₂ activation in both TDOs and related complexes; however, the spin state and degree of S-radical character in the Fe/O₂ adduct remain matters of debate. In this report, the nature of O₂-bound 2 and 4 in various spin states has been probed via density functional theory (DFT) calculations, with a particular focus on the effect of ligand charge (TIP vs. Tp⁻) on energetics and Fe-S covalency. We have also explored the surprisingly complicated reactivity of 1 and 2 with NO – a well-established surrogate for O₂. Electron paramagnetic resonance (EPR) and IR experiments revealed that multiple Fe/NO species (both high- and low-spin) are formed, including a cationic dinitrosyl iron complex (DNIC). These results highlight the hemilability of scorpionate chelates, as well as the critical role of the active-site pocket in controlling the relative positioning of first-sphere ligands in CDO and ADO.

2. Experimental and Computational Methods

General Information

Unless otherwise noted, all reagents and solvents were purchased from commercial sources and used as received. Dichloromethane (CH₂Cl₂) was purified and dried using a Vacuum Atmospheres solvent purification system and stored under N₂. The synthesis and handling of air-sensitive materials were performed under inert atmosphere using a Vacuum Atmospheres Omni-Lab glovebox equipped with a freezer set to −30 °C. Nitric oxide (NO) from a cylinder (Airgas, Inc.) was purified by passage through an ascarite II column, followed by a cold trap at −78 °C to remove higher N_xO_y impurities. ¹⁵N-labeled NO was prepared via reaction of Na¹⁵NO₂ with ascorbic acid and aqueous Cu(II) chloride under an argon (Ar) atmosphere.⁴⁷ The L_N3 supporting ligands K(^Ph,MeTp) and ^Ph2TIP, as well as complexes 3 and 4, were prepared according to literature procedures.^{36–37,44,48}

Physical Methods

Elemental analyses were performed at Midwest Microlab, LLC in Indianapolis, IN. UV-vis absorption spectra were collected with an Agilent 8453 diode array spectrometer equipped with a cryostat from Unisoku Scientific Instruments (Osaka, Japan) for experiments at reduced temperatures. Infrared (IR) spectra were measured with a Nicolet Magna-IR 560 spectrometer. X-band EPR experiments were performed using a Bruker EleXsys E650 instrument equipped with an ER4415DM cavity, an Oxford Instruments ITC503 temperature controller, and an ESR-900 He flow cryostat. The program EasySpin (version 5) was used to simulate the experimental spectra.^{49 1}H NMR spectra were recorded in deuterated solvents using a Varian 400 MHz spectrometer. Mass spectrometric data were measured using either an Agilent 6850 gas chromatography-mass spectrometer (GC-MS) with a HP-5 column, or a Brucker matrix-assisted laser desorption/ionization (MALDI) instrument. Magnetic circular dichroism (MCD) data were collected using a Jasco Model J-715 spectropolarimeter, in conjunction with an Oxford Instrument SM-4000 8T magnetocryostat. The samples for these studies were prepared in CH₂Cl₂ and then diluted with butyronitrile to a final ratio of 3:7 (v/v), or 1:1 (v/v) for the NO-treated samples, thereby yielding glassy solutions upon freezing in liquid nitrogen. Potential artifacts due to glass strain were eliminated by taking the difference between spectra collected with the magnetic field aligned parallel and antiparallel to the direction of light propagation.

X-ray diffraction (XRD) data were collected at 100 K with an Oxford Diffraction SuperNova kappa-diffractometer (Rigaku Corp.) equipped with dual microfocus Cu/Mo X-ray sources, X-ray mirror optics, an Atlas CCD detector, and a low-temperature Cryojet device. The data were processed with the CrysAlis Pro program package, followed by numerical absorption correction based on Gaussian integration over a multifaceted crystal model and then empirical absorption correction using spherical harmonics, as implemented in SCALE3 ABSPACK scaling algorithm. Structures were solved using the SHELXS program and refined with the SHELXL program⁵⁰ within the Olex2 crystallographic package.⁵¹ X-ray crystallographic parameters are provided in Table S1 and further experimental details are available in the CIFs supplied in the Supporting Information.

[Fe(^Ph2TIP)(CysOEt)]BPh₄ (1)

L-cysteine ethyl ester hydrochloride (38 mg, 0.20 mmol) and triethylamine (62 μL, 0.44 mmol) were combined in CH₂Cl₂ and stirred until the solid completely dissolved. A solution of the [Fe(^Ph2TIP)(MeCN)₃](OTf)₂ precursor⁴⁴ (0.242 g, 0.20 mmol) in CH₂Cl₂ was added dropwise, giving rise to a bright yellow solution that was stirred for 30 minutes. The solvent was removed under vacuum and the resulting yellow residue was taken up in MeOH, and filtered. Addition of NaBPh₄ (68 mg, 0.20 mmol) caused immediate formation of a yellow precipitate. After solvent removal under vacuum, the resulting yellow powder was dissolved in a minimal amount of 1,2-dichloroethane and layered with MeOH. The yellow needles that formed after one day were collected, washed with MeOH, and dried under vacuum. Yield = 0.138 g (55 %). X-ray quality crystals were obtained by allowing the mother liquor to stand for one week at room temperature and under inert atmosphere. Crystallographic analysis revealed the presence of a MeOH solvate in the unit cell. Elemental analysis calculated (%) for C₇₇H₆₉BFeN₇O₂PS•CH₃OH: C, 72.84; H, 5.72; N, 7.62. Found: C, 72.56; H, 5.61; N, 7.70. IR (solution in CH₂Cl₂, cm⁻¹): ν = 3345 (w, ν(N-H)), 1728 (s, ν(C=O)). ¹H NMR (400 MHz, CDCl₃): δ = 15.54 (s, 1H, -NH), 15.14 (s, 1H, -NH), 14.86 (s, 9H, N-CH₃), 8.08 (s, 3H), 7.63 (s, 12H, BPh₄), 6.90 (s, 6H), 6.37 (s, 8H, BPh₄), 6.04 (s, 6H), 4.22 (s, 6H), 2.96 (q, 2H, -CO₂CH₂CH₃), 2.41 (broad s, 2H, -SCH₂C-), 2.01 (s, 1H, -H₂NCH-), 1.31 (t, 3H, -COOCH₂CH₃), −17.46 (broad s, 6H) ppm.

[Fe(^Ph2TIP)(CysAm)]BPh₄ (2)

Cysteamine hydrochloride (23 mg, 0.20 mmol) and triethylamine (62 μL, 0.44 mmol) were combined in CH₂Cl₂ and stirred until all the solid dissolved. A solution of [Fe(^Ph2TIP)(MeCN)₃](OTf)₂ (0.242 g, 0.20 mmol) in CH₂Cl₂ was added dropwise, giving an orange solution that was stirred for 30 minutes. The solvent was removed under vacuum and the resulting yellow residue taken up in MeOH. Addition of NaBPh₄ (68 mg, 0.20 mmol) caused immediate formation of a yellow precipitate. After evaporation of the solvent under vacuum, the resulting yellow powder was dissolved in a minimal amount of 1,2-dichloroethane and layered with MeOH. X-ray quality yellow crystals formed after one day at room temperature. The crystals were collected, washed with MeOH, and dried under vacuum. Yield = 0.015 g (18 %). Elemental analysis calculated (%) for C₇₄H₆₅BFeN₇PS: C, 75.19; H, 5.54; N, 8.29. Found: C, 76.02; H, 6.02; N, 7.89. IR (solution in CH₂Cl₂, cm⁻¹): ν = 3353 (w, ν(N-H)). ¹H NMR (400 MHz, CDCl₃): δ = 14.86 (s, 9H, N-1-CH₃), 8.19 (s, 8H, BPh₄), 8.05 (s, 4H, BPh₄), 7.34 (s, 8H, BPh₄), 6.85 (s, 6H), 6.03 (s, 3H), 5.93 (s, 3H), 4.08 (s, 6H), 3.89 (s, 2H, -NH₂), 3.21 (s, 6H), 2.63 (t, 1H, -CH₂CH₂-), 1.35 (t, 1H, -CH₂CH₂-), 1.23 (t, 1H, -CH₂CH₂-), 1.13 (t, 1H, -CH₂CH₂-), −19.26 (broad s, 6H, o-4-Ph) ppm.

Reactivity with Dioxygen

Rate measurements were generally performed under conditions of O₂ saturation by bubbling O₂ (Airgas, Inc.) through CH₂Cl₂ solutions of 1 or 3 for several minutes. For both complexes, rates of reaction were measured four times at 20 °C, and changes in absorption intensity at two wavelengths were fit to exponential curves using the program IGOR. The reported k₁-values and uncertainties correspond to the average and standard deviation, respectively, of the resulting pseudo first-order rate constants. The concentration of O₂ in solution at various temperatures (T) was estimated using the formula: S = (LP_O2)/(TR), where L is the Ostwald coefficient (0.257 for CH₂Cl₂), P_O2 is the partial pressure of O₂, and R is the gas constant.^52–53 The determination of P_O2 accounted for the vapor pressure of CH₂Cl₂ (P_solv) as a function of T: P_O2 = 1 atm − P_solv. Lower concentrations of O₂ were obtained by injecting anaerobic solutions of the Fe(II) complex into O₂-saturated solutions of CH₂Cl₂ in varying amounts. Following established procedures,^36–37 the products of the oxygenation reactions were isolated by addition of 3M HCl to the reaction mixture and stirred for 3 hours. The aqueous layer was collected and the solvent removed under vacuum. The remaining residue was dissolved in MeOH and stirred with chelex for 12 hours. The solvent was removed by vacuum after filtration, and the resulting white solid was washed with toluene. ¹H NMR data of the reaction products were interpreted with the aid of published spectra and/or by comparison to spectra measured with commercially available material. Product of 1 + O₂ reaction: Yield = XX %; ¹H NMR (400 MHz, CD₃OD): δ = 1.28 (t, 3H, −OCH₂CH₃), 2.49 (dd, 1H, -CH₂S), 2.61 (dd, 1H, -CH₂S), 4.02 (m, 1 H, -CHNH₂), 4.20 (q, 2H, −OCH₂CH₃) ppm. MALDI-MS: m/z = 169.20, (calc. ). Product of 2 + O₂ reaction: Yield = XX %; ¹H NMR (400 MHz, CD₃OD): δ = 2.45 (t, 2H, -CH₂S), 3.15 (t, 2H, -CH₂NH₂) ppm. MALDI-MS: m/z = 169.20, (calc. ).

Density Functional Theory (DFT) Computations

Electronic-structure calculations were carried out using the ORCA 3.0 software package developed by Dr. F. Neese (MPI for Chemical Energy Conversion).⁵⁴ Computational models of 2 and 4 were obtained via unrestrained DFT geometry optimizations, using the X-ray crystal structures as starting points. For calculations involving 2, the ^Ph2TIP ligand was truncated by replacing the 4,5-diphenyl-1-methylimidazole donors with 4-methylimidazole rings; similarly, in calculations for 4 employed the ^Ph,MeTp⁻ ligand was modeled as ^Me,HTp⁻. Numerical frequency calculations verified that all structures corresponded to a local minimum with only real vibrational frequencies. The zero-point energies, thermal corrections, and entropy terms (vibrational, rotational, translational) were obtained from these frequency calculations. All calculations were carried out spin-unrestricted and utilized Ahlrichs’ valence triple-ζ basis set (TZV) and TZV/J auxiliary basis set, in conjunction with polarization functions on main-group and transition-metal elements (default-basis 3 in ORCA).^55–57 Solvent effects were accounted for using the conductor-like screening model (COSMO)⁵⁸ with a dielectric constant (ε) of 9.08 for CH₂Cl₂.

The DFT calculations employed different functionals depending on the nature of the species under examination and the property being computed. Geometric structures for the Fe(II) complexes and Fe/O₂ adducts were optimized using the Perdew-Burke-Ernzerhof (PBE) functional⁵⁹ with 10% Hartree-Fock exchange. The “spin-flip” feature of ORCA was employed to generate broken-symmetric wavefunctions for the S_tot = 2 and 1 states of Fe/O₂ adducts containing high-spin Fe centers. The transition state for the S-O_d bond forming reaction (where O_d is the distal oxygen atom of the iron-superoxo unit) was located by performing a relaxed surface scan along the S···O_d distance. The existence of the transition state was confirmed by the presence of one imaginary frequency, corresponding to the ν(S–O_d) mode.

Calculations of the Fe/NO adducts employed either the Becke-Perdew (BP86)^60–61 or TPSSh^62–63 functionals. In calculations of EPR parameters, the “core properties” with extended polarization [CP(PPP)] basis set⁶⁴ was used for the Fe atom and Kutzelnigg’s NMR/EPR (IGLO-III) basis set⁶⁵ for the NO ligand. The contribution of spin-orbit coupling to the g- and A-tensors was evaluated by solving the coupled-perturbed self-consistent field (CP-SCF) equations.^66–69 To compute the hyperfine coupling constants, a high resolution grid with an integration accuracy of 7.0 was used for the Fe and N atoms. Time-dependent DFT (TD-DFT) calculations employed the cam-B3LYP range-separated hybrid functional,⁷⁰ previously shown to yield good agreement between experimental and TD-DFT computed absorption spectra for CDO.⁷¹ Absorption energies and intensities were computed for 40 excited states with the Tamm-Dancoff approximation.^72–73 Isosurface plots of molecular orbitals were prepared using the ChemCraft program.

3. Results and Discussion

3.A. Synthesis and Solid State Structures

Complexes 1 and 2 were synthesized by first treating the previously-reported [Fe(^Ph2TIP)(MeCN)₃](OTf)₂ precursor⁴⁴ with either CysOEt•HCl or CysAm•HCl and two equivalents of triethylamine in CH₂Cl₂. After solvent removal, the resulting triflate salts were dissolved in MeOH and addition of NaBPh₄ caused immediate precipitation of 1 and 2 as bright yellow solids. X-ray quality crystals of 2 were grown by layering concentrated 1,2-dichloroethane (DCE) solutions with MeOH. In the case of 1, this procedure yielded analytically pure but poorly-diffracting needles; however, crystals suitable for X-ray analysis were eventually obtained by allowing the decanted mother liquor to sit for one week. The resulting crystal structure revealed that, during this time, the CysOEt ligand had undergone trans-esterification to give the complex [Fe(^Ph2TIP)(CysOMe)]BPh₄ (1^Me) instead. Despite this, ¹H NMR spectra of the first crop of crystals, which were employed in all spectroscopic and reactivity studies, indicated retention of the ethyl substituent (vide infra).

The X-ray structures of 1^Me and 2 reveal the presence of 5C monoiron(II) complexes with a bidentate S,N-cysteinato anion and κ³-^Ph2TIP chelate (Figure 1). In each case, the thiolate donor forms a hydrogen bond with a MeOH solvate molecule. The Fe(II) coordination geometries lie half-way between the square-pyramidal and trigonal bipyramidal limits, as quantified by the τ-values of 0.54 (1^Me) and 0.55 (2).⁷⁴ Significantly, the facial coordination mode of the ^Ph2TIP ligand ensures that a vacant site for O₂-binding is available cis to the thiolate ligand – a known requirement for dioxygenation of the sulfur atom.³¹ Selected metric parameters for the solid-state structures of 1^Me and 2 are provided in Table 1, in addition to previously-reported values for 3 and 4. The Fe-N_TIP bond distances are typical for high-spin Fe(II) complexes and fall within a narrow range of 2.15–2.20 Å. The Fe-N_Tp distances in 3 and 4 are slightly shorter on average, while also displaying greater variation with two short bonds of 2.12 Å (ave) and one long bond near 2.275 Å. The Fe-S1 and Fe-N_Cys bond distances change little across the 1–4 series with average values of 2.31 and 2.26 Å, respectively.

Figure 1.

(a) Thermal ellipsoid plot (50% probability) obtained from the X-ray crystal structure of 2. The BPh₄⁻ counteranion, non-coordinating solvent molecules, and all hydrogen atoms have been omitted for clarity. (b) Overlays of the crystallographically-derived structures of 1^Me (light gray) and the Cys-bound active-site of CDO (pink) derived from PDB 4TJO. The phenyl substituents of the TIP^Ph2 ligand are not shown.

Table 1.

Selected Bond Distances (Å) and Bond Angles (°) for Complexes 1–4 and the Cys-Bound CDO Active Site as Determined by X-ray Crystallography.

Bond Distances	1	2	3 a	4 b	CDO c
Fe1-N1	2.170(2)	2.169(2)	2.112(2)	2.153(2)	1.893
Fe1-N3	2.153(2)	2.178(2)	2.106(2)	2.110(2)	2.199
Fe1-N5	2.195(2)	2.183(2)	2.275(2)	2.249(2)	2.110
Fe1-S1	2.3107(6)	2.3051(5)	2.3122(9)	2.3175(6)	2.291
Fe1-N7	2.245(2)	2.248(2)	2.290(3)	2.252(2)	2.262
Bond Angles
N1-Fe1-N3	96.73(6)	98.67(6)	95.36(9)	95.97(6)	99.9
N1-Fe1-N5	87.90(6)	86.32(6)	82.29(9)	86.08(6)	98.6
N1-Fe1-S1	125.08(5)	120.73(4)	134.21(7)	129.17(5)	112.9
N1-Fe1-N7	89.01(6)	88.43(6)	89.30(9)	90.71(6)	85.7
N3-Fe1-N5	83.02(6)	82.88(6)	85.29(9)	82.16(6)	86.0
N3-Fe1-S1	137.22(5)	139.15(4)	130.07(7)	134.18(5)	146.8
N3-Fe1-N7	87.80(6)	86.20(6)	92.04(9)	87.38(7)	96.9
N5-Fe1-S1	104.99(4)	108.23(4)	104.07(6)	105.47(5)	94.1
N5-Fe1-N7	169.90(6)	167.03(6)	170.88(9)	168.67(6)	174.3
S1-Fe1-N7	84.63(5)	84.64(4)	84.32(7)	84.96(5)	80.8

^aReference 36.

^bReference 37.

^cObtained from an X-ray structure of Cys-bound CDO (PDB 4JTO; 2.0 Å resolution).

The superposition of complex 1^Me upon the active-site structure of Cys-bound Fe(II)-CDO⁷⁵ in Figure 1(b) highlights the structural similarities between the synthetic and enzymatic coordination environments. The most noticeable discrepancy concerns the relative orientations of the imidazole rings. The His donors in the enzyme active site possess greater rotational freedom than their synthetic counterparts, as the latter are tethered to a central P-atom at the 2-position. The CDO crystal structure exhibits Fe-N_His distances of 2.20, 2.11, and 1.89 Å (Table 1); the smallest value (corresponding to His86) is much shorter than expected based on synthetic CDO models and typical Fe-N_His bond distances in proteins.⁷⁶ The Fe-N_Cys and Fe-S_Cys bond distances of CDO nicely match those measured for 1^Me (Table 1) and the orientations of the cysteinato ligands with respect to the facial N3 triad are quite similar, although CDO lies somewhat closer to the square-pyramidal limit (τ-value of 0.46). The high degree of structural resemblance between CDO and 1^Me suggests that these five-coordinate sites are similarly primed for O₂ activation.

3.B. Solution-State Spectroscopic Properties

Examination of CDCl₃ solutions of 1 and 2 using the Evans NMR method provided effective magnetic moments of 5.16 and 5.02 μ_B, respectively, consistent with mononuclear, high-spin (S = 2) ferrous complexes. The ¹H NMR spectra of 1 and 2 exhibit paramagnetically-shifted peaks in the 15 to −20 ppm range(Figures S1 and S2). The three imidazole donors are spectroscopically equivalent due to dynamic averaging of the CysOEt/CysAm positions on the NMR time scale. Based on our prior analysis of ^Ph2TIP-based Fe(II) complexes,⁴⁴ the intense downfield peak at 15 ppm is assigned to the N-1-CH₃ groups, while the broad upfield peak near −18 ppm is attributed to ortho protons of the 4-Ph substituents. Resonances arising from the CysOEt and CysAm ligands experience only modest shifts and broadening upon ligand binding to Fe. The splitting pattern due to the ethyl ester moiety is clearly discernable in the spectrum of 1, indicating that the CysOEt ligand remains intact. In short, the ¹H NMR data confirm that the solution- and solid-states structures are essentially the same.

Solutions of 1 and 2 in CH₂Cl₂ are pale yellow due to the presence of overlapping absorption bands (ε ~ 1000 M⁻¹cm⁻¹) in the near-UV region that are attributed to S→Fe(II) charge transfer (CT) transitions (vide infra). To obtain further information regarding the electronic excited states of these complexes, magnetic circular dichroism (MCD) spectra of 1 and 2 were measured in frozen 3:7 solutions of CH₂Cl₂:butyronitrile. The absorption and MCD spectra of the two complexes are nearly identical, suggesting that the cysteinato substituent (–CO₂Et or –H) has little effect on overall electronic structure. As shown in Figures 2 and S3, the MCD spectra exhibit two intense, positive features near 23,800 and 26,800 cm⁻¹. The broadness of the lower-energy band suggests that it is composed of at least two electronic transitions. In all cases, the intensities of the MCD bands increase with decreasing temperature, displaying the C-term behavior expected for paramagnetic S = 2 systems. Significantly, the MCD spectra of 1 and 2 also exhibit a temperature-dependent band at 11,270 cm⁻¹ (887 nm) that is barely perceptible in the absorption spectra. Based on its large C₀/D₀ ratio of ~2.0 (where C₀ and D₀ are related to the MCD and absorption intensities, respectively⁷⁷), this peak likely arises from an Fe(II) d-d transition. This assignment is supported by previous spectroscopic studies, which revealed that the highest-energy d-d transition of 5C Fe(II) complexes lies between 8,400 and 13,500 cm^−1.78

Figure 2.

Electronic absorption and MCD spectra collected for complexes 2 (top) and 4 (bottom). The absorption spectra were measured at room temperature in CH₂Cl₂. The MCD spectra were obtained for frozen, glassy solutions in a 3:7 mixture (v/v) of CH₂Cl₂:butyronitrile at 4 K with a magnetic field of 7 T.

For the sake of comparison, Figure 2 also displays the absorption and MCD spectra of complex 4 measured under identical conditions. Only two S→Fe(II) CT bands are apparent, and these are blue-shifted by 2700 cm⁻¹ relative to those of complex 2. The d-d band experiences a smaller energy increase of 300 cm⁻¹. Similar shifts in band energies are observed when spectra of CysOEt-containing 1 and 3 are compared (Figure S3).

3.C. Computational Analysis of Fe(II) Complexes

The spectroscopic results were interpreted with the aid of time-dependent DFT (TD-DFT) studies that employed the cam-B3LYP range-separated hybrid functional.⁷⁰ These calculations were performed with geometry-optimized structures of 2 and 4 that are in good agreement with the crystallographic structures (Table S2). Consistent with the experimental data, the TD-DFT computed absorption spectrum for 2 is dominated by three S→Fe(II) CT transitions in the region below 30,000 cm⁻¹ (Figure S4). The energy-level diagram provided in Figure 3 depicts the ligand- and Fe(II)-based molecular orbitals (MOs) involved in the relevant transitions. The donor MOs are localized on the CysAm ligand and contain primarily S(3p) character: the S_σ MO lies along the Fe-S bond, while the S_π MO is perpendicular to the bond. The two lowest-energy CT bands in the TD-DFT spectrum arise from S_π→Fe(d_xz)/Fe(d_yz) transitions centered around 25,000 cm⁻¹, which nicely matches the experimental band at 23,800 cm⁻¹. To higher energy, TD-DFT predicts a S_σ→Fe(II) (σ→σ*) transition at 29,100 cm⁻¹ that likely corresponds to the experimental feature at 26800 cm⁻¹. The highest-energy Fe d-d transition, attributed to the Fe(d_xy)→Fe(d_x2–y2) excitation, is predicted to appear as a weak feature at 13,750 cm⁻¹. Thus, the calculated energies agree resonably well with the experimental values, with a root-mean-square deviation of only 2100 cm⁻¹ for the bands observed by MCD.

Figure 3.

Energy-level diagram for the spin-down (β) molecular orbitals (MOs) obtained from a spin-unrestricted DFT calculation for complex 2. MOs are labeled according to their principal contributor. DFT-generated isosurface plots of the MOs are also provided.

Comparison of the DFT results obtained for 2 and 4 indicate that substitution of neutral ^Ph2TIP with anionic ^Ph,MeTp⁻ causes an nearly uniform 4450 cm⁻¹ increase in the energies of the Fe 3d set of orbitals. Hence, the S→Fe(II) CT bands experience a substantial blue-shift upon conversion of 2→4, while the highest-energy d-d transition is largely unaffected – consistent with the spectroscopic data described above. The S_σ→Fe(II) (σ→σ*) transition of complex 4, predicted to occur at 33,270 cm⁻¹, is likely obscured by the onset of π–π* transitions in the near-UV region, which explains the presence of only two CT bands in the absorption/MCD spectra of 3 and 4. Moreover, an analysis of the relative compositions of the half-occupied Fe 3d-based MOs indicates that the Fe–S bond in 2 is more covalent than its counterpart in complex 4, which is also reflected by minor differences in the Mulliken spin populations of the S atoms (0.19 spins in 2 and 0.15 spins in 4). This finding is consistent with the fact that the TIP-based complexes exhibit more intense S→Fe(II) CT transitions than the Tp-based complexes (Figures 2 and S3), given that CT intensity is proportional to the amount of ligand character in the metal-based MOs.

Substrate-bound Fe(II)-CDO likewise exhibits two S→Fe(II) CT bands in its MCD spectrum; however, both of these features appear above 30,000 cm⁻¹, much higher in energy than the corresponding bands of complexes 1–4.⁷⁹ This large discrepancy in CT energies is rather surprising given the high-degree of structural similarity between the first coordination spheres of 1 and Cys-bound Fe(II)-CDO, as discussed above (Figure 1). Noncovalent interactions within the active site of CDO presumably modulate the relative energies of the thiolate- and Fe 3d-based MOs, thereby increasing the CT energies.

3.D. O₂ Reactivity: Experimental Results

Not surprisingly, the O₂ reactivity of the ^Ph2TIP-based complexes closely mirrors that reported previously by Limberg and coworkers for 3 and 4. As shown in Figure 4, treatment of 1 with O₂ results in the slow decay of the S→Fe(II) CT bands to eventually yield a featureless spectrum, indicating loss of the iron-thiolate bond. Similar spectral changes were observed upon O₂ exposure of complexes 3 and 4,^36–37 as well as Cys-bound Fe(II)-CDO.⁷⁹ The reaction products were isolated after stirring solutions of 1 and 2 in O₂-saturated CH₂Cl₂ for six hours, followed by acidic work-up. ¹H NMR spectra revealed that the 1 + O₂ reaction yields the ethyl ester of L-cysteine sulfinic acid as the only observable CysOEt-derived product (Figure S5). Similarly, exposure of 2 to O₂ yields hypotaurine (HTau) as the dominant product, although an unidentified CysAm-derived compound is also present in smaller amounts (HTau:unknown ratio of 5:1; Figure S6). The identities of the sulfinic acid products were confirmed by comparison to literature spectra (cysteine sulfinic acid ethyl ester) or commercial samples (HTau), in addition to mass spectrometric analysis. Thus, each complex in the 1–4 series behaves as a structural and functional TDO mimic.

Figure 4.

Time-resolved absorption spectra for the reaction of complex 1 (conc. = 0.53 mM) with O₂; the spectra shown here were collected at intervals of 30 min. The reaction was performed at 20 °C in O₂-saturated CH₂Cl₂. The path length of the cuvette was 1.0 cm. Inset: Plots of absorption intensity as a function of time for reaction of 1 and 3 with O₂ at 20 °C in O₂-saturated CH₂Cl₂. The wavelengths monitored were 360 and 355 nm for 1 and 3, respectively. Absorption intensities were normalized using the equation provided in the plot.

Rate measurements for the reaction of 1 and 3 with O₂ were generally performed in O₂-saturated CH₂Cl₂ solutions ([O₂] = 5.8 mM at 20 °C) with Fe(II) concentrations of less than 0.60 mM. Under these conditions, the decay in absorption intensity versus time followed first-order behavior past three half-lives (Figures 4 and S7) for both complexes, and initial rates increased linearly with O₂ concentration (Figure S8). Thus, the reactions are first-order with respect to both Fe and O₂, permitting a detailed kinetic analysis. At 20 °C, the 1 + O₂ reaction proceeds with a pseudo-first-order rate constant (k₁) of 1.9(4) × 10⁻⁴ s⁻¹. The 3 + O₂ reaction exhibits a smaller k₁-value of 1.2(3) × 10⁻⁴ s^−1.80 Thus, the identity of the L_N3 supporting ligand has only a modest impact on the O₂ reaction rate, with the neutral ^Ph2TIP ligand offering a slight rate advantage over the anionic ^Ph,MeTp⁻ ligand.

Activation parameters for the 1 + O₂ reaction were determined by monitoring rates at temperatures between 35 and −30 °C. Analysis of the linear Eyring plot (Figure 5) provided an activation enthalpy (ΔH^‡) of 12(1) kcal/mol and activation entropy (ΔS^‡) of −25(4) cal/(mol·K), resulting in an activation free energy (ΔG^‡) of 19(2) kcal/mol at 293 K. These values are similar to previously-reported parameters for O₂ activation by mononuclear Fe complexes,⁵² including closely-related 5C complexes with ^Ph2TIP and Tp⁻ ligands.^81–82 The large negative value for ΔS^‡ indicates that the rate-determining step has an associative nature, which is common for reactions involving O₂. However, as discussed in the next section, the activation parameters likely reflect multiple elementary reactions, and thus the values must be interpreted with caution.

Figure 5.

Eyring plot for the reaction of 1 with O₂ in O₂-saturated CH₂Cl₂ over a temperature range of 35 to −30 °C. Second-order rate constants (k₂) were obtained by dividing the pseudo-first-order rate constant by [O₂] at the specified temperature.

3.E. O₂ Reactivity: Computational Results

The first step in the proposed O₂ activation mechanisms of CDO and related models complexes is the generation of an iron-superoxo species, followed by nucleophilic attack of the distal O atom on the thiolate ligand to give a four-membered Fe-O-O-S ring. Computational studies suggested that this second step is rate-limiting in both the enzyme and complex 3.^39,83 To better understand the kinetic data reported above, we have examined the thermodynamics of O₂ binding and O–S bond formation via DFT calculations for complexes 2 and 4.⁸⁴ These calculations employed the PBE functional with 10% HF exchange, which has been shown provide reliable geometries and relative energies for Fe/O₂ species.⁸⁵

The adduct that forms upon O₂ binding to a high-spin Fe(II) complex has three possible spin states: triplet (S = 1), quintet (S = 2), or septet (S = 3). The septet and quintet states are best described as high-spin Fe(III) centers coupled to a superoxide radical (O₂^•−) in either a ferromagnetic or antiferromagnetic manner, respectively.⁸⁶ The triplet state arises from antiferro-magnetic coupling between a high-spin Fe(II) center and a neutral O₂ ligand. However, we also considered the possibility that the S = 1 Fe/O₂ adduct consists of a low-spin Fe(III) ion ferromagnetically coupled to O₂^•−, as this configuration yielded the lowest-energy Fe/O₂ model in de Visser’s study of complex 3.³⁹ The resulting computational models are labeled ^spin[2-O₂]^(HS,LS) and ^spin[4-O₂]^(HS,LS), where HS and LS indicate whether the Fe center is high-spin or low-spin, respectively. Metric parameters of the geometry-optimized structures are provided in the Supporting Information (Table S3). Whereas de Visser found that O₂ binding to 3 triggers dissociation of a pyrazole donor regardless of spin state,³⁹ nearly allo of our optimized Fe/O₂ models are 6C with Fe-N_TIP/Tp bonds lengths ≤ 2.31 Å; the only exceptions being the ^S=3[2-O₂]^HS and ^S=2[2-O₂]^HS models, in which one of the Fe-N_TIP bonds elongates to ~2.4 Å during geometry optimization to yield quasi-6C structures (Table S3).

Table 2 summarizes the computed thermodynamic parameters for the eight Fe(II) + O₂ → Fe/O₂ reactions examined here. For both [2-O₂] and [4-O₂], the four spin states lie with 3 kcal/mol of each other. This result is typical of ferric-superoxo species, which commonly possess a large number of close-lying electronic states.⁸⁶ The enthalpic (ΔH_rxn) contributions to O₂ binding are generally insignificant or slightly favorable, yet the overall process is endergonic (ΔG_rxn = 8–13 kcal/mol) due to the unfavorable entropic effects characteristic of bimolecular reactions. Although the differences are within the uncertainty of the methodology, the computed ΔG_rxn-values are uniformly smaller for the [4-O₂] models relative to their [2-O₂] counterparts, suggesting that the Tp-based complexes have a higher affinity for O₂ than the TIP-based complexes. This conclusion is supported by the fact that the computed ν(O-O) frequencies for the [4-O₂] models are 15–25 cm⁻¹ lower in energy compared to their counterparts in the analogous [2-O₂] models (Table 2), implying that the ^Ph,MeTp⁻ ligand promotes greater charge transfer from Fe(II) to O₂. Analysis of the Mulliken populations revealed that the S-donors in the Fe/O₂ adducts possess a significant amount of spin-density, with alpha-spin-values of 0.33 and 0.27 in the S = 2 structures of [2-O₂] amd [4-O₂], respectively. This partial radical character, which is consistently larger in the TIP-containing models, likely promotes formation of the S-O bond in the next step of the mechanism.

Table 2.

Energetic Parameters Computed for O₂ Binding to Complexes 2 and 4, and Comparison of Superoxo Stretching Frequencies in the Resulting [FeO₂] Adducts.

Stot (Fe spin) a	reactants	ΔHrxn	ΔSrxn b	ΔGrxn b	ν(O-O) (cm−1)
S = 3 (HS)	2 + O2	0.5	10.3	10.8	1222
	4 + O2	−1.3	10.6	9.3	1197
S = 2 (HS)	2 + O2	2.3	10.8	13.1	1213
	4 + O2	0.7	11.2	11.9	1198
S = 1 (HS)	2 + O2	1.7	11.2	12.9	1269
	4 + O2	−0.3	12.3	12.0	1243
S = 1 (LS)	2 + O2	−3.3	14.6	11.3	1183
	4 + O2	−7.1	15.1	8.0	1160

^aS_tot is the overall spin of the [FeO₂] adduct, whereas HS and LS indicate the spin-state of the Fe center itself.

^bThermodyanmic parameters were computed assuming a temperature of 298.15 K.

The majority of computational studies of CDO and synthetic mimics suggest that formation of the critical Fe(II)-O-O-S cyclic intermediate occurs along the S = 2 potential energy surface (PES). ^39–40,83 Our DFT calculations revealed that conversion of ^S=2[2-O₂]^HS and ^S=2[4-O₂]^HS to the corresponding peroxo species occurs with overall transition-state barriers of 25 and 23 kcal/mol, respectively, relative to the Fe(II) and O₂ starting materials (Scheme 3). The barrier of ~12 kcal/mol for O_d-S bond formation is in line with previous DFT studies of WT CDO and related models, which reported values around 10 kcal/mol for the same mechanistic step.^40,87 The difference in transition-state energies between ^S=2[2-O₂]^HS and ^S=2[4-O₂]^HS lies within the estimated error of the computations (± 2 kcal/mol), consistent with our experimental finding that the identity of L_N3 does not have a major impact on dioxygenation rates. Significantly, the computed activation parameters for ^S=2[2-O₂]^HS (ΔH^‡ = 10.5 kcal/mol, ΔS^‡ = −48 cal/(mol•K)) are in reasonably good agreement with the experimentally-determined values of ΔH^‡ = 12(1) kcal/mol, ΔS^‡ = −25(4) cal/(mol•K). In the DFT-optimized transition-state geometry derived from ^S=2[2-O₂]^HS (Scheme 3), the O-O bond is positioned directly over the Fe-S bond, giving rise to a S···O_d distance of 2.18 Å. The O–O and Fe–S bonds are elongated by 0.040 and 0.212 Å, respectively, compared to the ^S=2[2-O₂]^HS adduct. The degree of S-radical character in the transition state, as indicated by Mulliken spin populations, is significantly greater for TIP-based 2 (0.18 spins) than Tp-based 4 (0.11 spins). The distribution of unpaired spin density in the transition-state structure (Figure S9) indicates that the S–O_d bond arises from overlap between the S-based p_z-orbital and the O₂-based π* orbital pointed along the S–O_d axis.

Scheme 3.

Relative energies (in kcal/mol) for the initial steps in the reaction of 2 (red lines) and 4 (blue lines) with O₂ at 298.15 K. Geometry-optimized structures of ^S=2[2-O₂]^HS (left) and ^S=2[2-O₂]^TS (right) are also provided and select bond lengths are indicated in angstroms (Å).

For the sake of completeness, we also examined the cyclization reaction along the S = 1 PES originating from the ^S=1[2-O₂]^LS and ^S=1[4-O₂]^LS species, as this process yielded the lowest-energy transition state in de Visser’s study of complex 3.³⁹ However, the barrier for the S =1 route is nearly 10 kcal/mol higher in energy than the S = 2 pathway, suggesting that the former is not a viable option for the dioxygenation mechanism if the Fe center remains six coordinate.

As shown in Figure S10, the energies of the resulting Fe(II)-O-O-S cyclic intermediates on the S = 2 PES are roughly equal to those of the preceding Fe/O₂ adducts. The next step in the mechanism requires O–O bond cleavage to yield an high-spin oxoiron(IV) unit bound to a sulfoxide ligand. Our DFT calculations indicate that the transition-state barrier for O-O bond cleavage is quite small (~2 kcal/mol; Figure S10), providing further evidence that formation of the Fe-O-O-S species is the rate-limiting step in the dioxygenation mechanism.

3.F. Nitric Oxide Reactivity: Experimental Results

Using nitric oxide (NO) as a surrogate for O₂ is a common strategy for gaining insights into the electronic structures of ferrous centers in MNIDs and related complexes, since the reaction gives rise to an EPR-active {FeNO}⁷ species (using the Enemark-Feltham notation, where the superscript indicates the sum of Fe 3d and NO π* electrons). A notable feature of CDO is that the enzyme-substrate complex binds NO to yield a low-spin (S = 1/2) Fe/NO adduct, as opposed to a high-spin (S = 3/2) adduct as is generally observed for MNID enzymes with the 2H1C triad.⁸⁸

Our studies found that exposure of 1 and 2 to NO yields different species depending on the temperature of the reaction. For both complexes, the Fe/NO adducts generated at −70 °C (1^NO and 2^NO) feature two absorption bands near 440 and 660 nm. As shown in Figure 6, these spectra closely resemble those reported previously for high-spin, six-coordinate (6C) {FeNO}⁷ complexes with similar ligand sets, such as [Fe(NO)(4-TIP^Ph)(acac)]⁺ and [Fe(NO)(^Ph2Tp)(acac^PhF3)] (where 4-TIP^Ph = tris(2-phenylimidazol-4-yl)phosphine and acac^PhF3 = anion of 4,4,4-trifluoro-1-phenyl-1,3-butanedione; the structure of the latter complex was verified crystallographically).⁴⁵ The corresponding X-band EPR spectra (Figure 6; inset) exhibit an axial S = 3/2 signal (g ~ 4.0, 2.0), also characteristic of 6C {FeNO}⁷ species.⁸⁹ The intensity of the S = 3/2 signal was measured between 10 and 50 K for 1^NO, and simulation of the data provided a D-value of 6.7(3) cm⁻¹ and E/D-ratio of 0.005. We also generated Fe/NO adducts of 3 and 4 (3^NO and 4^NO, respectively), as Limberg and coworkers did not report NO-binding studies for these complexes. The electronic absorption and EPR data of 3^NO and 4^NO (not shown) are nearly identical to those obtained for 1^NO and 2^NO, indicating that the L_N3 ligand has little effect on NO reactivity or the spectroscopic features of the resulting species. Therefore, only the iron-nitrosyl adducts derived from 1 and 2 will be discussed in detail.

Figure 6.

Absorption spectrum of 1^NO (black solid line) generated via reaction of 1 with NO at −70 °C in CH₂Cl₂. Previously-reported⁴⁵ UV-vis spectra of [Fe(NO)(4-TIP^Ph)(acac)]⁺ (blue dashed line) and [Fe(NO)(^Ph2Tp)(acac^Ph,F3)] in MeCN and CH₂Cl₂, respectively, are shown for comparison (these spectra have been shifted upwards for clarity). Inset: X-band EPR spectrum of 1^NO in frozen CH₂Cl₂. Parameters: frequency = 9.380 GHz; power = 2.0 mW; modulation amplitude = 2.0 G; T = 10 K.

In addition to the dominant S = 3/2 resonances arising from 1^NO and 2^NO, EPR spectra of samples exposed to NO at low temperatures contained variable amounts of a minor S = 1/2 signal (Figures 6 and S11). The concentrations of the low-spin species, labeled 1^NO-LS and 2^NO-LS, were generally less than 10% of total Fe as determined by double integration of the signal at 77 K.⁹⁰ Notably, these rhombic spectra exhibit 1:1:1 triplet superhyperfine splitting in each of the three g-values, corresponding to a ¹⁴N nucleus with A_x,y,z = 36, 36, and 45 MHz in the case of 2^NO-LS (the complete set of spin-Hamiltonian parameters is provided in Table 3). This pattern coverts to a 1:1 doublet when the samples are prepared with ¹⁵NO (Figures 6 and S11), confirming that the splitting is due to the N-atom of a single NO ligand. Therefore, we can rule out the possibility that the 1^NO-LS and 2^NO-LS signals are due to formation of a DNIC. Possible structures for these S = 1/2 species derived from DFT calculations will be discussed in the next section.

Table 3.

Experimental and Computed EPR Parameters for Low-Spin {FeNO}⁷ Species

					14N A-values (MHz) a
Species	Method	gx	gy	gz	Ax	Ay	Az
1NO-LS	exper	2.009	2.037	2.072	46	41	44
2NO-LS	exper	2.007	2.035	2.070	45	37	36
S=1/2[2-NO]6C	DFT/BP86	1.985	2.014	2.042	43	68	39
	DFT/TPSSh	1.986	2.004	2.022	31	81	36
S=1/2[2-NO]5C	DFT/BP86	2.005	2.021	2.047	50	46	41
	DFT/TPSSh	2.007	2.031	2.054	53	34	28
[Fe(NO)(N3PyS)]+	exper b	1.962	2.007	2.047	40	59	40
6C complex	DFT/BP86	1.976	2.006	2.025	29	62	32
	DFT/TPSSh	1.975	2.001	2.019	19	82	23
Fe(OEP)(NO)	exper c	2.015	2.057	2.106	41	50	43
Fe(II)CDO-Cys/NO	exper d	1.979	2.028	2.071	<30	100	<30

^aIn most cases, the orientations of the computed g- and A-tensors did not align. Thus, spectral simulations with the computed Euler angles were used to determine the A-value corresponding to each g-value.

^bReference 34.

^cOEP = octaethylporphyrin(2−); references 97 and 99.

^dReference 88.

The reaction of 1 (or 2) with NO at room temperature (RT) initially generates a chromophore with absorption features matching those observed at −70 °C. However, 1^NO and 2^NO are not stable at elevated temperatures, quickly converting to the same new species (RT^NO) characterized by a weak absorption band at 910 nm (Figure 8). Time-dependent absorption spectra for this reaction display an isosbestic point at 820 nm, indicating clean conversion of 1^NO or 2^NO to RT^NO without the build-up of intermediates (Figure S12). EPR spectra collected for frozen solutions of RT^NO exhibit a sharp, rhombic S = 1/2 signal that lacks hyperfine splitting (Figure 8; inset), along with a very weak peak at g = 4.0 arising from residual 1^NO or 2^NO. Unfortunately, despite repeated attempts, we were unable to grow crystals of RT^NO suitable for X-ray crystallography.

Figure 8.

Electronic absorption (top, black) and MCD (bottom, red) spectra of RT^NO prepared via reaction of NO with complex 1 at ambient temperature. The absorption spectrum was measured in CH₂Cl₂. The MCD spectrum was obtained for a frozen, glassy solution in 1:1 (v/v) CH₂Cl₂:butyronitrile at a temperature of 4 K and a magnetic field of 7 T. Inset: X-band EPR spectrum (blue line) of RT^NO in frozen CH₂Cl₂. Parameters: frequency = 9.497 GHz; power = 0.5 mW; modulation amplitude = 1.0 G; T = 77 K.

Notably, the absorption and EPR features of RT^NO are quite similar to those reported for cationic, dinitrosyl iron complexes (DNICs) with the general formula [Fe(NO)₂(L_N2)]⁺ (where L_N2 is a neutral N,N-donor).^91–94 Like RT^NO, these S = 1/2 {Fe(NO)₂}⁹ species commonly exhibit a rhombic g-tensor centered at 2.03 and a broad absorption band in the near-IR region with an extinction coefficient of ~100 M⁻¹ cm⁻¹. To determine whether RT^NO corresponds to a DNIC, IR spectra of NO-treated 1 and 2 were measured at RT. The resulting spectra display two intense peaks at 1745 and 1817 cm⁻¹, both of which downshift by ~34 cm⁻¹ upon ¹⁵NO substitution (Figure S13). These frequencies and isotope shifts are consistent with prior studies of [Fe(NO)₂(L_N2)]⁺ complexes, which revealed that the out-of-phase and in-phase combinations of ν(NO) modes occur within energy ranges of 1720–1770 and 1790–1840 cm⁻¹, respectively. Furthermore, the separation between NO stretching frequencies (Δν_NO) is diagnostic of coordination number and nuclearity,⁹⁵ and the Δν_NO-value of 72 cm⁻¹ measured for RT^NO is typical of four-coordinate (4C) {Fe(NO)₂}⁹ species.^91–94

The formation of {Fe(NO)₂}⁹ DNICs from Fe(II)-thiolate precursors requires formation of 0.5 equivalent of disulfide to provide the additional electron. Analysis with GC-MS found that L-cystine ethyl ester was the only CysOEt-derived product generated during the 1 + NO reaction at RT. Thus, our collective experimental results suggest that DNIC formation occurs via the reaction:

$${[\text{Fe}(\text{NO})({}^{\text{Ph}2}\mathrm{T}\text{IP})(\text{CysOEt})]}^{+}({\mathbf{1}}^{\mathbf{NO}})+\text{NO}\to {[\text{Fe}{(\text{NO})}_2({}^{\text{Ph}2}\mathrm{T}\text{IP})]}^{+}({\mathbf{RT}}^{\mathbf{NO}})+1/2\text{RS}-\text{SR},$$

where R = -CH₂CH(NH₂)CO₂Et

This conclusion is supported by the fact that the spectroscopic parameters of RT^NO are identical regardless of whether 1 or 2 serves as the precursor, indicating that the thiolate ligand is no longer attached to the iron-containing product.

To further probe the electronic properties of RT^NO, we collected MCD spectra for frozen solutions in 1:1 (v/v) CH₂Cl₂:butyronitrile. Whereas the absorption spectrum of RT^NO exhibits few well-defined features, the MCD spectrum measured at 4 K and 7 T shows four distinct bands between 10,000 and 25,000 cm⁻¹ (Figure 8). Variable-temperature variable-field (VTVH) MCD data collected at 907 nm (Figure S14) confirm that the MCD bands arise from a S = 1/2 species, consistent with the EPR results. To the best of our knowledge, these data represent the first MCD analysis of a DNIC. The origins of the observed absorption/MCD features were elucidated with the help of TD-DFT calculations, as discussed in the following section.

3.G. Nitric Oxide Reactivity: Computational Results

DFT calculations for Fe/NO species derived from complexes 1 and 2 utilized either the nonhybrid BP86 or hybrid meta-GGA TPSSh functionals, which are known to provide accurate geometries and spectroscopic parameters for both mono- and dinitrosyl iron complexes.⁹⁴ Geometry optimizations of the high-spin {FeNO}⁷ adduct obtained upon NO binding to 2 yielded quasi-6C structures, labeled ^S=3/2[2-NO]^6C, with an elongated distance of 2.5 Å for the Fe–N_TIP bond trans to the nitrosyl (Table S4) – indicative of the strong trans influence of NO ligands.⁹⁶ Interestingly, during the BP86 geometry optimizations of the low-spin {FeNO}⁷ adduct, the trans imidazole fully dissociated and formed an intramolecular H-bond with the amine group of CysAm, yielding the 5C structure ^S=1/2[2-NO]^5C. In order to generate a low-spin 6C model (i.e., ^S=1/2[2-NO]^6C), it was necessary to constrain the trans Fe–N_TIP bond distance to 2.3 Å. The structures and relative energies of the three BP86-optimized models are provided in Figure 9. The ^S=1/2[2-NO]^5C structure is more stable than both ^S=3/2[2-NO]^6C and ^S=1/2[2-NO]^6C by several kcal/mol. When the TPSSh functional is employed instead of BP86, a low-spin {FeNO}⁷ species with a quasi-6C structure (trans Fe-N_TIP distance of 2.41 Å) was found to exist in a local minimum. While the three TPSSh-generated structures are fairly close in energy, the ^S=1/2[2-NO]^5C structure remains the most stable (Figure 9).

Figure 9.

DFT optimized structures and relative energies (in kcal/mol) of the ^S=3/2[2-NO]^6C, ^S=1/2[2-NO]^6C, and ^S=1/2[2-NO]^5C models computed using the BP86 and TPSSh functionals. The structures were derived from the BP86 calculations.

To determine whether the ^S=1/2[2-NO]^5C structures corresponds to the mononitrosyl S = 1/2 species observed via EPR spectroscopy (Figure 7), we calculated g-values and ¹⁴N A-tensors using both functionals. We also performed parallel calculations for [Fe(NO)(N3PyS)]⁺ – a structurally characterized 6C, low-spin {FeNO}⁷ complex prepared by Goldberg and coworkers.³⁴ As shown in Table 3, our DFT methodology accurately reproduces the EPR features of this complex, namely, its rhombic g-tensor centered near g_e = 2.00 and large A_y-value. Importantly, the calculations are consistent with prior studies of low-spin {FeNO}⁷ complexes, both heme and nonheme, which have revealed that certain spin-Hamiltonian parameters are diagnostic of coordination number.^97–100 Specifically, the g_x-values of 6C species are always less than the free-electron value (g_e = 2.002), whereas g_x > g_e for 5C {FeNO}⁷ species. Additionally, the A-tensors of 6C complexes exhibit relatively large anisotropy and a dominant A_y-value, whereas 5C complexes possess relatively isotropic ¹⁴N A-tensors in which the A_x-value is the largest component (Table 3). The EPR parameters of 1^NO-LS and 2^NO-LS closely align with those computed for ^S=1/2[2-NO]^5C, as well as those observed for 5C NO adducts of ferrous-porphyrin complexes.^97–100 Thus, we conclude that the observed low-spin signal is due to dissociation of the trans imidazole donor in a fraction of the {FeNO}⁷ adducts prepared at low temperature.

Figure 7.

EPR spectra of 2^NO-LS (black solid line) generated via reaction of 1 with ¹⁴NO (top) or ¹⁵NO (bottom) at −70 °C in CH₂Cl₂. Parameters: frequency = 9.492 GHz; power = 2.0 mW; modulation amplitude = 1.0 G; T = 77 K. Simulations of the experimental data (gray lines) were generated using the spin-Hamiltonian parameters listed in Table 3.

We also generated computational models with the formula [Fe(NO)₂(^MeTIP)]⁺ to further confirm that RT^NO corresponds to a cationic DNIC. These calculations employed the broken-symmetry approach to generate the proper S = 1/2 wavefunction, best described as a high-spin Fe(III) center coupled antiferromagnetically to two NO⁻ (S = 1) ligands.¹⁰¹ Regardless of the functional used, geometry optimizations of the DNIC^DFT models converged to four-coordinate (4C) structures with a κ²-TIP ligand, further demonstrating the hemilability of the TIP framework. The computed spectroscopic properties of the DNIC^DFT models are summarized in Table 4. DFT is known to overestimate ν(NO) frequencies, with hybrid functionals generally performing worse than nonhybrid functionals.^94,101 While our results follow this pattern, the BP86 functional proved quite effective in estimating the energy splitting between the in-phase and out-of-phase combinations of ν(NO) modes of RT^NO. Both functionals yielded rhombic g-tensors centered near g = 2.03, consistent with our EPR data and prior studies of DNICs.^91–94 The computed A-tensors are highly anisotropic with two small A-values (< 10 MHz) and one moderate value of 25 or 39 MHz, in accordance with the absence of any detectable ¹⁴N hyperfine splitting in the experimental RT^NO EPR spectrum.

Table 4.

Comparison of Computed Spectroscopic Parameters for DNIC^DFT Models with Experimental Values Measured for RT^NO.

	DNICDFT (BP86)	DNICDFT (TPSSh)	RTNO (exper)
ν(NO)sym, ν(NO)asym	1843, 1776 cm−1	1897, 1816 cm−1	1817, 1745 cm−1
gx, gy, gz	2.013, 2.028, 2.058	2.027, 2.033, 2.061	2.01, 2.04, 2.07
Ax, Ay, Az (14N7)	3, 22, 3 MHz	7, 36, 13 MHz	not observed
Ax, Ay, Az (14N8)	4, 5, 25 MHz	1, 8, 39 MHz

Consistent with the experimental absorption spectrum of RT^NO, TD-DFT calculations of DNIC^DFT predict a series of weak bands (ε-values < 400 M⁻¹cm⁻¹) in the visible region (Figure S15). The donor MOs for these transitions contain approximately equal amounts of Fe 3d and NO π* character due to the highly-covalent nature of the iron-nitrosyl bonds. In contrast, the acceptor orbital is the Fe(d_x2_-y2)-based MO that lies in the N_TIP-Fe-N_TIP plane and lacks contributions from the NO ligands (Figure S15). Thus, the transitions have both ligand-field and NO⁻→Fe(III) CT character, but their intensities are limited by the poor spatial overlap between donor and acceptor MOs.

4. Conclusions

The syntheses, X-ray crystallographic structures, electronic properties, and O₂/NO reactivities of two novel TDO active-site models (1 and 2) have been reported. These 5C Fe(II) complexes consist of a bidentate “substrate” ligand (CysOEt or CysAm) and the facially-coordinating ^Ph2TIP supporting ligand (Scheme 2). In our estimation, 1 and 2 are the most structurally accurate models of the CDO and ADO active sites prepared to date, as the ^Ph2TIP supporting ligand reproduces the neutral charge and all-imidazole coordination of the 3His triad. The electronic structures of 1 and 2, along with those of the previously-reported ^Ph,MeTp-based TDO models (3 and 4), were elucidated with spectroscopic (absorption, MCD) and computational (DFT) techniques. These studies yielded specific assignments for the three overlapping S→Fe(II) CT transitions in the near-UV region, and MCD analysis permitted observation of the highest-energy d-d transition at ~11,300 cm⁻¹. In addition, they revealed that substitution of neutral ^Ph2TIP with anionic ^Ph,MeTp⁻ destabilizes the Fe 3d manifold by ~0.13 eV and significantly reduces the covalency of the Fe–S bond.

Like similar complexes generated by the Limberg and Goldberg groups, 1 and 2 are functional TDO mimics that react with O₂ to yield the corresponding sulfinic acid product. Our kinetic studies determined that the identity of the L_N3 ligand (^Ph2TIP versus ^Ph,MeTp⁻) has little effect on O₂ reaction rates. DFT methods were then employed to evaluate the energetics of formation of the Fe/O₂ adduct and its conversion to the cyclic Fe-O-O-S intermediate; the latter reaction is the putative rate-determining step in the dioxygenation mechanism. These calculations indicated that O₂ binding to the high-spin Fe(II) centers of 2 and 4 is endergonic by 9–13 kcal/mol due to an unfavorable entropic contribution. The transition state leading to the putative Fe-O-O-S intermediate along the S = 2 surface is further uphill by ~12 kcal/mol for 2 and 4, resulting in an overall barrier of ~24 kcal/mol relative to the starting complexes and O₂ (Scheme 3). These computed values are roughly consistent with the experimental activation parameters of ΔH^‡ = 12(1) kcal/mol and ΔS^‡ = −25(4) cal/(mol•K) measured experimentally for the reaction of 1 with O₂. In both the Fe/O₂ adducts and transition-state structures, the charge of the supporting ligand modulates the amount of unpaired spin density on the S atom of the thiolate ligand, with the TIP-based models exhibiting greater S-radical character than the Tp-based models. This correlation offers a possible rationale for the preference of TDOs for the 3His triad instead of the 2H1C triad, since the development of (partial) thiyl radical character is thought to facilitate formation of the critical Fe-O-O-S species in the enzymatic mechanism.¹⁰²

Since our efforts to trap intermediates in the dioxygenation reactions were unsuccessful, we explored the reactivity of 1 and 2 with NO – an established proxy for O₂. As summarized in Scheme 4, initial NO binding generates a 6C, high-spin {FeNO}⁷ adduct (1^NO and 2^NO) that is metastable at reduced temperatures, but converts at RT to a more stable 4C DNIC (RT^NO) with concomitant formation of disulfide. Samples prepared at low-temperatures also revealed a minor S = 1/2 signal (1^NO-LS and 2^NO-LS) that was shown, through a combination of spectroscopic and computational analysis, to correspond to a 5C species featuring a κ²-TIP ligand. The dissociation of the trans imidazole donor may be required for the conversion of the 6C {FeNO}⁷ adducts into the 4C {Fe(NO)₂}⁹ species.

Scheme 4.

Interestingly, even though the first-sphere coordination geometries of 1 and 2 are nearly identical to those of CDO and ADO, there are dramatic differences in electronic structure, spectroscopic properties, and reactivity between the synthetic models and enzyme active sites. For example, the S→Fe(II) CT transitions observed for CDO are much higher in energy than those observed for complexes 1–4. In addition, whereas 1^NO and 2^NO are high-spin and decompose at RT to form a DNIC, the 6C {FeNO}⁷ adduct of CDO is low-spin and relatively stable.⁸⁸ These differences highlight the importance of the protein environment in fine-tuning the geometric and electronic structures of the nonheme iron active-site. The rotational freedom of the imidazolyl and pyrazolyl donors allows the ^Ph2TIP and ^Ph,MeTp ligands to adopt κ²-N,N binding modes, thereby encouraging the formation of 5C and 4C Fe/NO adducts. In contrast, QM/MM studies have determined that the binding of NO to substrate-bound CDO causes little change in Fe-N_His bond distances,¹⁰² suggesting that the positions of the His ligands are fixed within the active-site. Thus, despite the considerable advantages of scorpionate ligands, our future TDO modeling studies will employ more constrained ligands that prevent elongation or breaking of the Fe-N bond trans to the O₂/NO binding site. Moreover, the incorporation of second-sphere donors capable of hydrogen-bonding interactions would serve to modulate spectroscopic features and stabilize reactive intermediates in a manner similar to the protein environment.