Oxygen Activation and Radical Transformations in Heme Proteins and Metalloporphyrins

Abstract

As a result of the adaptation of life to an aerobic environment, nature has evolved a panoply of metalloproteins for oxidative metabolism and protection against reactive oxygen species. Despite the diverse structures and functions of these proteins, they share common mechanistic grounds. An open-shell transition metal like iron or copper is employed to interact with O₂ and its derived intermediates such as hydrogen peroxide to afford a variety of metal–oxygen intermediates. These reactive intermediates, including metal-superoxo, -(hydro)peroxo, and high-valent metal–oxo species, are the basis for the various biological functions of O₂-utilizing metalloproteins. Collectively, these processes are called oxygen activation. Much of our understanding of the reactivity of these reactive intermediates has come from the study of heme-containing proteins and related metalloporphyrin compounds. These studies not only have deepened our understanding of various functions of heme proteins, such as O₂ storage and transport, degradation of reactive oxygen species, redox signaling, and biological oxygenation, etc., but also have driven the development of bioinorganic chemistry and biomimetic catalysis. In this review, we survey the range of O₂ activation processes mediated by heme proteins and model compounds with a focus on recent progress in the characterization and reactivity of important iron–oxygen intermediates. Representative reactions initiated by these reactive intermediates as well as some context from prior decades will also be presented. We will discuss the fundamental mechanistic features of these transformations and delineate the underlying structural and electronic factors that contribute to the spectrum of reactivities that has been observed in nature as well as those that have been invented using these paradigms. Given the recent developments in biocatalysis for non-natural chemistries and the renaissance of radical chemistry in organic synthesis, we envision that new enzymatic and synthetic transformations will emerge based on the radical processes mediated by metalloproteins and their synthetic analogs.

1. Introduction

The highly exergonic four-electron reduction of oxygen to water has an enthalpy change of −80 kcal/mol. The biochemical processes that mediate this reduction provide the energy that sustains all aerobic organisms on earth.¹ Paradoxically, however, the utilization of O₂-derived energy in aerobic biology represents a significant challenge due to the thermodynamic stability of O₂ in its triplet ground state.¹⁻⁵ The first one-electron reduction of O₂ to superoxide ion is endergonic by 7.8 kcal/mol, which makes the stepwise, one-electron processes unfavorable (Figure 1). Direct 2e^– reactions between ground-state O₂ (S = 1) and many closed-shell organic substrates are spin forbidden unless a stepwise route via diradical intermediates is available. To overcome these difficulties of exploiting O₂, nature has evolved families of enzymes to transport O₂ and to catalyze, manipulate, and control its reduction. A majority of these enzymes contain transition metals such as iron, copper, and sometimes manganese with unpaired d-electrons.⁶ These open-shell systems with variable oxidation and electronic states can facilitate the activation of triplet ground-state oxygen and participate at different stages of O₂ reduction (Figure 1). The study of these processes, known collectively as oxygen activation, is a central theme in biology and has driven the development of metallobiochemistry and bioinorganic chemistry since their inception.

Figure 1.

Reduction sequence of O₂ and representative metalloenzymes related to O₂ activation. Reduction potentials shown are versus normal hydrogen electrode (NHE).⁵

Heme-containing proteins have assumed a key position in our understanding of O₂-activation mediated by metalloproteins. The general catalytic cycle for oxygen reduction is shown in Figure 2, which starts with the binding of O₂ to a ferrous heme to afford a ferric-superoxo adduct, followed by reduction to ferric–hydroperoxo, and high-valent, iron–oxo (ferryl) intermediates via O–O bond scission. The rich chemistry that occurs between the heme iron and dioxygen provides the basis for the diverse roles of heme proteins in O₂ metabolism, including O₂ transport (myoglobin and hemoglobin), reduction of O₂ to water (cytochrome c oxidases), degradation of reactive oxygen species (catalases and peroxidases), and catalysis of a variety of oxygenation reactions of organic substrates such as steroid hydroxylation, a monooxygenation typical of cytochrome P450s, and dioxygenation reactions (heme dioxygenases).⁷ The use of metal-cofactors to effect serial reductive activation of O₂ is not limited to heme proteins but represents a general strategy employed by nature to harness the oxidation power of dioxygen. The reaction sequence and reactive intermediates similar to that shown in Figure 2a are widely present in a variety of O₂-activating metalloenzymes such as nonheme iron oxygenases and copper-containing oxygenases and oxidases (Figure 2b).^6,8−11 Accordingly, in spite of their diverse structures and functions, these metalloproteins share many common mechanistic grounds in terms of O₂ activation and catalysis.^1,6,12

Figure 2.

(a) Interaction of dioxygen with heme and heme–oxygen intermediates in heme proteins. (b) Representative metal–oxygen intermediates in metalloproteins (PHM: peptidylglycine α-hydroxylating monooxygenase).

In this interpretive review, we will survey the range of O₂ activation processes mediated by heme proteins and related model compounds with a focus on recent progress in the characterization and reactivity of important iron–oxygen intermediates through mid-2017. Key aspects of background material in the field will also be presented to provide context and perspective. The underlying structural and electronic factors contributing to O₂ activation chemistry will be delineated. The topics of O₂-activation and oxidative transformations mediated by heme proteins and metalloporphyrins have been extensively reviewed previously.^{3,4,7,13−32} While most of these reviews focus on either heme proteins or synthetic models, this review will provide a comprehensive perspective of the spectrum of oxidative transformations mediated by the iron–oxygen species in both heme proteins and analogous synthetic model compounds. We will discuss the fundamental mechanistic features of these transformations and how a variety of oxidative pathways are developed from a common mechanistic foundation. For model compounds, we will mainly focus on metalloporphyrins. Other porphyrinoid ligands, such as corrolazines, porphyrazines, and phthalocyanines will not be discussed in detail. Readers can refer to several recent reviews on these families of model compounds.^29,33−35

2. Key Intermediates Involved in O₂ Activation by Heme Proteins and Model Compounds

2.1. Oxy–Ferrous (Superoxoferric) Intermediates

2.1.1. Structural Features

The first step of O₂ activation by heme proteins is the binding of dioxygen to the heme iron, forming an oxy–ferrous intermediate (or ferric-superoxo intermediate, see discussions in section 2.1.2). Heme proteins exhibit varied O₂ binding features. For instance, O₂ transport proteins like hemoglobin and myoglobin have very high O₂ association rate constants (10⁷ M^–1 s^–1) and moderate to high dissociation rate constants (0.1–20 s^–1).^36,37 Heme-containing NO or CO sensor proteins are tuned to coordinate these tighter-binding molecules and show little binding of O₂ even in O₂-saturated solutions.^38,39 In P450s, O₂ binds to ferrous heme with rate constants around 10⁶ M^–1 s^–1,⁴⁰ and the resulting ferric-superoxo intermediate is prone to undergo autoxidation to the ferric state if it is not further reduced to form the ferric hydroperoxo compound 0.⁴¹ It is important to note that heme in the ferric state has a very weak oxygen-affinity and will not bind oxygen at ambient temperatures. It is the ease of electron transfer from iron to O₂ that is important for the formation of the ferric-superoxo intermediate and thus O₂ binding. For instance, highly electron-withdrawing Fe(II) porphyrin such as Fe(TPFPBr₈) is very inert toward dioxygen (TPFPBr₈ = β-pyrrole brominated 5,10,15,20-tetrakis(pentafluorophenyl)porphyrin).⁴²

The structural and electronic properties of the ferric-superoxo intermediate are crucial for understanding the O₂ activation in heme proteins. In the 1930s, Pauling and Coryell found that oxygenated hemoglobin (oxy-Hb) was diamagnetic and proposed that dioxygen was bound to heme in a bent, end-on mode.^43,44 This geometric feature was confirmed by an oxy–ferrous model compound synthesized by Collman and by crystal structures of oxy-Mb and oxy-Hb solved by Phillips and Shaanan, respectively.⁴⁵⁻⁴⁷ Following these pioneering efforts, crystal structures of ferric-superoxo intermediates have been solved for a number of heme proteins (Figure 3).⁴⁸⁻⁵² All reported structures reveal an end-on coordination of dioxygen to the iron with a Fe–O bond distance around 1.8 Å, an O–O bond distance ranging from 1.22 to 1.30 Å, and an Fe–O–O angle ranging from 110° to 160°.

Figure 3.

Structural parameters of ferric-superoxo intermediates of representative heme proteins (oxy-myoglobin, PDB: 1A6M; HRP compound III, PDB: 1H57; P450cam oxygen complex, PDB: 1DZ8). ^aData presented are for the β subunit of oxyhemoglobin in structure 1HHO.

2.1.2. Nature of Fe–O₂ Bonding

The electronic structure of Fe–O₂ heme complexes is remarkable because it forms a species with a singlet electronic ground state arising from the combination of high-spin (S = 2) ferrous heme and triplet O₂. The nature of Fe–O₂ bonding has been a subject of active debate for decades.⁵³ There are three models to explain Fe–O₂ bonding (Figure 4).⁵⁴ In Pauling’s model, an S = 0 state was assigned to both a dioxygen and a ferrous center.^43,44 In contrast, Weiss proposed that the Fe–O₂ species would be best described as a ferric-superoxide complex in which a low-spin (S = 1/2) ferric center is antiferromagnetically coupled to a doublet superoxide anion O₂^–.⁵⁵ In a third model initially suggested by McClure and refined by Goddard, an intermediate-spin (S = 1) ferrous center antiferromagnetically coupled to ³O₂ in a 3-center-4-electron bond affords the singlet Fe–O₂ species (“Ozone model”).^56,57

Figure 4.

Three models for Fe–O₂ bonding in the ferric-superoxo intermediate. Adapted from ref (64). Copyright 2013 American Chemical Society.

Mössbauer studies have lent support to the Weiss model. The Mössbauer parameters of oxyhemoglobin (isomer shift δ = 0.2–0.3 mm/s, quadrupole splitting ΔE_Q = 2.00–2.20 mm/s), first measured by Lang and Marshali, are consistent with a low-spin ferric center.⁵⁸ Sharrock and co-workers have reported similar parameters for oxygenated reduced P450cam (δ = 0.31 mm/s and ΔE_Q = 2.15 mm/s at 4.2K).⁵⁹ The Weiss model is further supported by resonance Raman spectroscopic studies, in which a ferric center was indicated by the oxidation state marker band of oxyHb (ν = 1377 cm^–1).⁶⁰⁻⁶²

Through density functional theory (DFT) and valence bond (VB) approaches, Shaik has shown that Fe–O₂ contains a dative σ(Fe–O) interaction between vacant 3d_z² and π*(O₂^–) and a rather weak π interaction between d_yz(Fe) and π_⊥*(O₂).⁵³ The H-bonding from a distal histidine residue (His64) in oxy-Mb and the electrostatic polarization from protein backbone stabilizes the negative charges on O₂. Although Fe–O₂ primarily follows the bonding scheme of the Weiss model, the VB analysis showed a mixing of the other two models, whose contributions depend on protein hosts and active site environments, making Fe–O₂ difficult to describe according to canonical oxidation state formalisms. Recent Fe K-edge X-ray absorption spectroscopy (XAS) studies by Sarangi et al. have shown that oxy-Hb is best described as a ferric superoxide species (Weiss model) in solution. However, in crystalline oxy-Hb, the Fe K pre-edge feature indicates Fe²⁺S = 0 character, consistent with the Pauling model. This surprising change of electronic configuration with environment supports the multiconfigurational description.⁶³ The impact of local environment on the Fe–O₂ bonding was also revealed by an iron L-edge XAS study of a ferric-superoxo “picket fence” porphyrin with 1-methylimidazole (1-MeIm) axial ligand, Fe(TpivPP)(1-MeIm)O₂ (TpivPP = meso-tetra(α,α,α,α–O-pivalamidophenyl)porphyrin, Figure 5b).⁶⁴ The results showed a strong π-interaction between d_yz(Fe) and π_⊥*(O₂) as opposed to the weak π-interaction calculated for oxy-Mb. This difference is likely caused by the much weaker hydrogen bonding to peroxo distal oxygen (N···O distance ∼3.9 Å compared to 2.67 Å in oxy-myoglobin) and the lack of active site polarization in this model compound.

Figure 5.

(a) Structure of Im-pyrroheme and Py-mesoheme and their binding affinities to O₂ and CO. (b) Structural parameters of Fe(TpivPP)(2-MeHIm)(O₂) are from ref (67).

2.1.3. Metal-Dioxygen Intermediates of Metalloporphyrins

Studies of biomimetic synthetic iron porphyrins have greatly aided in the understanding of O₂ binding and activation in heme proteins. An early study by Traylor et al. showed that the O₂ affinity of imidazole-ligated iron(II) porphyrin (1, Figure 5a) was over 3800 times stronger than that of a pyridine-ligated one (2, Figure 5a), whereas both complexes showed similar CO binding affinities.⁶⁵ This difference was thought to originate from the large π basicity of imidazole, which could better stabilize the weak O₂ ligand. Much progress has been made with sterically protected porphyrin scaffolds like “picket–fence” porphyrins, which prevent the irreversible bimolecular condensation of Fe(II) porphyrins to form μ-oxo-bridged diiron(III) complexes upon reacting with O₂. This strategy was first implemented by Collman and led to the characterization of the first Fe–O₂ porphyrin compound in the 1970s.^45,66 This model binds O₂ reversibly with affinity similar to those of Hb and Mb. More importantly, the Fe–O₂ intermediate prepared by Collman is diamagnetic and showed a bent end-on binding geometry of dioxygen just as postulated by Pauling. With a similar approach, many synthetic Fe–O₂ complexes have been synthesized and characterized with varied O₂ binding affinities tuned by ligand environments.^22,24 Very recently, the Scheidt and Schulz groups have used multitemperature X-ray crystallography and Mössbauer spectroscopy to study three ferric-superoxo complexes based on “picket fence” porphyrin TpivPP with 1-methyl-, 1-ethyl-, or 2-methylimidazoles as axial ligand, respectively.⁶⁷ Their results showed a low energy for the rotation of the Fe–O₂ unit. An 80K structure of Fe(TpivPP)(O₂) (3, Figure 5b) with 2-methylimidazole (2-MeHIm) axial ligand allowed them to accurately determine the position of both oxygen atoms, whose structural details were vailed in previous studies because of the high thermal motion and O₂ disorder. They obtained a Fe–O distance of 1.811(5) Å, O–O bond lengths of 1.281(12) Å, Fe–O–O angle of 118.2°, and an off-axis tilt of Fe–O bond of 6.2°.

Heme-containing metal–organic frameworks (MOFs) have recently been used as a platform to study the Fe–O₂ adduct because of the spatial separation of heme centers in the MOF scaffold and the gas-absorption properties of MOFs. This approach allows for solvent–free generation of ferric-superoxo intermediates. A rare 5-coordinate heme–O₂ complex could be obtained by reacting a Zr-based porphyrinic MOF PCN-224Fe^II (PCN stands for porous coordination network) with O₂ at −78 °C.⁶⁸ The species revealed a characteristic η¹ end-on binding mode of O₂ with a Fe–O distance of 1.79(1) Å, O–O distance of 1.15(4), and an Fe–O–O angle of 118(4)^o. PCN-224Fe^II showed a low O₂-binding enthalpy of −8 kcal/mol, about half of those of other heme model compounds with imidazole axial ligands. This weak O₂ affinity suggests the importance of electron-donation from axial ligands (σ-donation in the case of imidazole) for the formation of Fe–O₂ complexes.

Transition-metals other than iron have also been explored for O₂ binding in both synthetic metalloporphyrins and heme proteins. A great number of studies have focused on cobalt dioxygen complexes, which have a doublet ground state (S = 1/2) and thus can be probed by EPR spectroscopy.⁶⁹⁻⁷³ Co–O₂ is best described as a Co(III) superoxo species with the unpaired electron largely on the O₂ moiety. Co–O₂ also adopts a η¹ end-on binding mode as revealed by the structures of both Co-Mb and model compounds (Figure 6).^71,74 Co^II–O₂ binds dioxygen reversibly but with much lower affinity compared to their iron analogs (P_1/2 around 60 Torr vs 0.49–1.1 Torr for iron).⁷² As in iron porphyrins, axial ligands are important for Co–O₂ binding. A Co–O₂ intermediate without an axial ligand was recently synthesized in a MOF scaffold PCN-224, which showed a much lower O₂-binding enthalpy (−3.6 kcal/mol) compared to that of model compound with an imidazole ligand (−13.3 kcal/mol).⁷⁵ In contrast, cobalt(II) octaethylporphyrin supported on a highly oriented pyrolytic graphite showed a favorable O₂ binding (ΔH^o = −16.7 ± 3.6 kcal/mol), owing to the partial electron donation from the graphite substrate.⁷⁶

Figure 6.

Bonding and electronic properties of various metal-dioxygen porphyrin complexes.

In addition to iron and cobalt, manganese, chromium, and ruthenium dioxygen adducts have also been investigated in porphyrinoid scaffolds.²⁴ We will only briefly summarize their structural and electronic features (Figure 6). Dioxygen adducts of manganese(II) porphyrins have unique features compared to other metal-dioxygen species. Mn(II) binds oxygen with the loss of the axial ligand that usually stabilizes the metal–O₂ moiety.^77,78 Mn–O₂ exhibits an intermediate spin (S = 3/2) and a large zero–field splitting (−2.4 cm^–1).⁷⁹⁻⁸¹ The 900 to 990 cm^–1 IR stretching frequencies ν(O₂) of Mn–O₂ are much lower than those of end-on Fe–O₂ and Co–O₂ complexes (1100–1200 cm^–1).⁸² These properties indicate that O₂ binds to Mn(II) in a symmetric η² side-on mode, a Mn^IV–O₂^2– formalism. Oxy-Ru(II) porphyrins are diamagnetic and have O–O stretching frequency of 1103 cm^–1, suggesting a low-spin Ru(III) superoxide formulation as described in the Weiss model of Fe–O₂.⁸³⁻⁸⁵ For Cr^II–O₂ porphyrins, a 1142 cm^–1 O–O stretching frequency and a S = 1 ground state also indicates a Weiss model, in which a high-spin S = 3/2 Cr(III) and an S = 1/2 superoxide anion are antiferromagnetically coupled.^79,86

2.2. Ferric-(hydro)peroxo Intermediates

2.2.1. Ferric-(hydro)peroxo Intermediates of Heme Proteins

The binding of dioxygen to ferrous heme facilitates its reductive activation. In many hemoproteins such as nitric oxide synthase and cytochrome P450s, an electron and a proton are transferred to the oxy–ferrous complex to afford a ferric–hydroperoxo intermediate [Fe(III)–O–OH], also known as compound 0 (cpd 0). Distinct from most heme proteins, cytochrome c oxidase (CcO) utilizes a copper cofactor complexed with active site residues, including an unusual His-Tyr cross-link, to facilitate the O₂ activation. This arrangement leads to the formation of an analogous ferric-peroxo-cupric intermediate. O₂ activation by CcO is discussed in detail in section 4.3 of this review.

Cpd 0 species have proven to be short-lived, posing significant challenges for their characterization. Cryogenic radiolytic reduction, as first demonstrated by Davydov and Symons,⁸⁷⁻⁸⁹ has proven to be very useful for the study of cpd 0.⁹⁰ This method utilizes the hydrated electrons generated through radiolysis to reduce the ferrous-dioxygen intermediate in frozen solvent matrices. The combination of radiolytic cryoreduction with spectroscopic methods, especially EPR and ENDOR spectroscopies employed by Hoffman et al., have afforded critical insight into the electronic and structural properties of cpd 0.⁹¹ By controlling the cryogenic conditions, ferric-peroxo anion species [Fe(III)–O–O^–] could be obtained and its protonation at the distal oxygen to form the conjugate acid [Fe(III)–O–O–H, cpd 0] upon thermal annealing could be monitored by EPR spectroscopy. It is now known that cpd 0 has a low-spin ferric, S = 1/2 ground state. The g₁ value of cpd 0 was found to be indicative of the protonation state of the distal oxygen, with a g₁ > 2.30 commonly observed for the ferric–hydroperoxo state.⁹² The thermal annealing experiments also showed that the ease of proton delivery varied widely across heme proteins. For instance, Mb ferric-peroxo is stable near 100 K, whereas the ferric-peroxo anion of P450cam is readily protonated above 55 K.⁹³⁻⁹⁵ This ease of protonation may be related to the presence of the proton relay channel in P450cam. Very recently, Kincaid et al. have characterized ferric-(hydro)peroxo intermediates of lactoperoxidase (LPO) via cryoradiolysis/resonance Raman (rR) spectroscopy.⁹⁶ Under alkaline conditions (pH 8.2), the ferric-peroxo intermediate is the major form for LPO and no protonation occurred via thermal annealing. Under acidic conditions (pH 5.6), however, a mixture of LPO-peroxo and LPO–hydroperoxo intermediates were formed. Intriguingly, upon thermal annealing to 170 K, LPO-peroxo did not convert to the hydroperoxo species, and its rR intensity decreased substantially, whereas LPO–hydroperoxo was more stable with peak intensities almost unchanged. Radiolytic cryoreduction has also enabled the structural characterization of the fleeting ferric-(hydro)peroxo species with X-ray crystallography. In 2007, Schlichting et al. reported a crystal structure of compound 0 by controlled radiolytic reduction of ferrous-dioxygen intermediate of chloroperoxidase (CPO).⁹⁷ Cpd 0 of CPO showed an O–O distance of 1.5 Å, a long Fe–O distance of 1.9 Å, and an Fe–O–O angle of 131° (Figure 7a). Soon afterward, Unno and co-workers reported the characterization of a ferric peroxoanion compound of myoglobin, which showed a Fe–O distance of 1.85 Å, Fe–O–O angle of 120°, and a short O–O distance of 1.33 Å (Figure 7b), which distinguishes it from an O-protonated hydroperoxo state.⁹⁸ Peroxymyoglobin was also characterized by Andersson et al. in 2008 and similar structural features were reported.⁹⁹

Figure 7.

(a) Crystal structure of CPO-0 (PDB: 2J5M) (left) and the QM/MM optimized structure of doublet CPO-0 (generated with molecular coordinates provided in ref (97)). (b) Crystal structure of the ferric-peroxo intermediate of Mb (PDB: 2Z6T) and the corresponding QM/MM optimized structure (doublet ground state) (generated with molecular coordinates provided in ref (98)).

2.2.2. Ferric-(hydro)peroxo Intermediates of Metalloporphyrins

In parallel with the progress with heme proteins, synthesis and characterization of cpd 0 analogs in model compounds have also been actively explored. Valentine and co-workers reported the first preparation of a ferric porphyrin peroxide complex Fe(TPP)O₂^– (6, Figure 8) by treating Fe^III(TPP)Cl (4) with 2 equiv of KO₂ or Fe^II(TPP) (5) with 1 equiv of KO₂ in aprotic solvents.^100,101 This species exhibited a characteristic EPR spectrum of a rhombic high-spin Fe(III) (S = 5/2) complex with a broad and intense resonance at g ∼ 4.2, which is in contrast to the low-spin S = 1/2 state observed for typical cpd 0 species (g values ∼2). Similar complexes can also be formed by 1e^– electrochemical reduction of ferrous dioxygen intermediates.¹⁰² The infrared O–O stretching vibration of 6 at 806 cm^–1 and the lack of an axial basic ligand suggested that this complex is a side-on η²-peroxo ferric species.¹⁰¹ Several η¹ end-on ferric–hydroperoxo models were synthesized by Tajima et al. via addition of H₂O₂ to ferric porphyrins at low temperatures and subsequently characterized by UV–vis and EPR spectroscopies at 77 K.^103−105 The EPR spectra of these complexes showed a low-spin ferric intermediate with g values ∼2, similar to those observed for cpd 0 in heme proteins. In Tajima’s conditions to prepare η¹ end-on ferric-peroxo complexes (11), methanol was used as a cosolvent, and hydroxide or imidazole was added as the axial ligand (Figure 8b), in contrast to the conditions of preparing the side-on η²-peroxo ferric species 10, in which an aprotic-solvent was used and no axial ligand was present. Recently, the impact of axial ligands and solvent on ferric-peroxo bonding mode has been extensively studied by Naruta et al. They successfully synthesized both ferric–peroxo and ferric–hydroperoxo complexes of Fe^II(TMPIm), an iron porphyrin complex bearing a covalently linked imidazole axial ligand (9, Figure 8).^106,107 The ferric-peroxo complex 10 was characterized to be a rare seven-coordinate, side-on ferric-peroxo species with axial association of the covalently linked imidazole. An EPR signal for 10 at g ∼ 4.2 and an rR O–O stretching vibration at 807 cm^–1, which is characteristic for a high-spin η²-peroxo–ferric heme species, supports this formulation. Addition of methanol to 10 led to the formation of a low spin end-on hydroperoxoferric complex, 12, as judged by changes in the EPR spectra from g ∼ 4.2 to a new set of signals at g = 2.31, 2.19, and 1.95. The identity of 12 was further confirmed by a 4 cm^–1 upshift of ν(O–O) in MeOD solvent. The Mössbauer spectrum of 12 showed an isomer shift of 0.25 mm/s and a quadrupole splitting of 2.16 mm/s, which is in agreement with those of the previously characterized low-spin hydroperoxo–heme species.¹⁰⁸

Figure 8.

(a) Formation of a high-spin side-on η²-peroxo ferric intermediate (6). (b) Formation of low-spin end-on η¹-hydroperoxo ferric intermediate (8). (c) Conversion of a high-spin side-on η²-peroxo ferric intermediate (10) into low-spin end-on η¹-hydroperoxo ferric intermediate (12). Adapted with permission from ref (106). Copyright 2009 Wiley-VCH.

Addition of methanol to a 6-coordinated ferric-peroxo complex [(TMP)Fe^III(O₂^2–)] (13, Figure 9), however, led to a direct decomposition of 13. This striking difference highlights the importance of the imidazole axial ligand in stabilizing the ferric–hydroperoxo species. Soon afterward, the synthesis of an end-on ferric-peroxo anion intermediate was successfully achieved by the same group with a TMPIm ligand containing a bulky xanthene substituent hanging over the porphyrin (14, Figure 9).¹⁰⁹ The EPR spectrum of 14 is characteristic of a low-spin ferric complex with g values of 2.27, 2.16, and 1.96, consistent with those of end-on ferric-peroxo intermediates in heme proteins prepared via radiolytic reduction. Addition of methanol to this species made a new low-spin ferric complex with g values of 2.32, 2.19, and 1.95 and ν(O–O) of 807 cm^–1. The small increase in the spread of g values and a g₁ > 2.3 are indicative of an end-on ferric–hydroperoxo complex.

Figure 9.

Chemical structure of [(TMP)Fe^III(O₂^2–)] (13) and an end-on ferric-peroxo anion species containing a bulky TMPIm ligand (14).

The effect of the second coordination sphere on the formation of ferric hydroperoxo intermediates was recently explored by the same group with an iron(II) porphyrin containing both a covalently appended proton donor and an axial imidazole ligand (15, Figure 10).¹¹⁰ This complex can bind oxygen to afford the ferric-superoxo intermediate 17. Intriguingly, 17 can be reduced by 15 to afford a Fe^III-hydroperoxo intermediate 18 presumably via a PCET pathway. This process would be inhibited by converting the carboxylic acid moiety on xanthene to an ester. The corresponding iron(II) porphyrin 16 can still form a ferric–hydroperoxo intermediate 21 upon reacting with cobaltocene (CoCp₂) via a PCET pathway. Compared to 21, 18 showed a higher ν(Fe–O) (579 cm^–1 vs 576 cm^–1) and a lower ν(O–O) (807 cm^–1 vs 811 cm^–1) compared to 21, indicating that the intramolecular H-bond interaction enhances the Fe–O but weakens the O–O bonding.

Figure 10.

Impact of the second coordination sphere on the formation of a ferric-(hydro)peroxo intermediate. Adapted with permission from ref (110). Copyright 2016 Royal Society of Chemistry.

Ferric hydroperoxo intermediates have also been investigated using electrochemical methods.^{102,110−114} The use of electrodes modified with iron porphyrins was used to probe for reactive iron porphyrin intermediates under electrocatalytic conditions.^111,114 The electrochemical methods also represent a potentially “cleaner” way to generate iron porphyrin intermediates by circumventing the need of reducing agents. A recent example is the preparation of [Fe^III(TPFPP)(O₂)]⁻ and [(MeIm)Fe^III(TPFPP)(OOH)] intermediates via electrochemical reduction of the corresponding ferric-superoxo intermediate [Fe^III(TPFPP)(O₂^•-)] (TPFPP = 5,10,15,20-tetrakis(pentafluorophenyl)porphyrin).¹¹²

2.3. Heterolytic O–O Bond Cleavage of Ferric–Hydroperoxo Intermediates

2.3.1. O–O Bond Cleavage of cpd 0 to Generate Cpd I in Heme Proteins

The main role of the ferric–hydroperoxo species (cpd 0) in O₂ activation is to generate the key oxoiron(IV) porphyrin cation radical intermediate, which is the canonical compound I (cpd I). Historically, the terms compound I and compound II (the one-electron-reduction product of cpd I) stem from very early spectroscopic studies of the reaction of horseradish peroxidase (HRP) with hydrogen peroxide. In 1937, using an ocular spectroscope, Keilin and Mann observed a red intermediate with a Soret band near 420 nm upon addition of H₂O₂ to HRP.¹¹⁵ Four year later, Theorell identified a more short-lived green intermediate formed initially in the reaction between HRP and H₂O₂ (thus termed “cpd I”),¹¹⁶ which then converted into the red species (termed “cpd II”) observed by Keilin and Mann. Similar intermediates were also observed by Stern and Chance for catalase.^117−119 Jones and Dunford suggested the presence of an HRP-H₂O₂ precursor complex in preequilibrium with HRP and H₂O₂ (termed “cpd 0”) formed prior to the cpd I generation based on exacting kinetic measurements of this reaction.^120−123

It is now known that cpd 0 is an Fe(III)–OOH intermediate and is generated via H₂O₂ binding to the resting ferric heme. The subsequent cpd I formation involves the protonation of Fe(III)–OOH (cpd 0) at the distal oxygen and the subsequent heterolytic scission of the O–O bond to release water. Cpd 0 generation and the ensuing O–O bond cleavage in peroxidases are facilitated by residues at distal pocket acting as acid–base catalysts, as proposed first by Poulos and Kraut.¹²⁴ This proposal originated from their studies of cytochrome c peroxidase (CcP), in which a highly conserved histidine/arginine (His52/Arg48) diad in the distal pocket was found to be crucial for cpd I generation.^124,125 It has been suggested that upon binding of H₂O₂ to the ferric heme, His52 acts as a general base catalyst to deprotonate the proximal oxygen to form cpd 0 and then act as a general acid catalyst to protonate the distal oxygen of cpd 0, whereas Arg48 stabilized the leaving oxygen through hydrogen bonding. Recently, this mechanistic picture was further elaborated by Raven and Moody using neutron cryo-crystallography to visualize the crucial protons (deuterons) at the active site.¹²⁶ Surprisingly, their results showed that His52 remained protonated after the formation of CcP cpd I, suggesting that a second proton was delivered to cpd 0, presumably by proton relay through the H-bonded water network, for the O–O bond heterolysis of [Fe(III)–O–OH] (Figure 11).¹²⁷ In addition to the regulation of proton delivery to the peroxo distal oxygen to create a better leaving group (the “pull” effect), the O–O bond cleavage is further facilitated by the σ-donation (electron push) from an axial histidine ligand by providing electron density to the antibonding O–O orbital.¹²⁸ The low-spin electronic configuration of Fe–OOH is thought to facilitate O–O bond cleavage via interaction of the filled d_xz orbital on iron with the low-lying σ*(O–O) orbital.¹²⁹ This “push-pull” effect is a general mechanistic feature for the O–O bond to generate cpd I in heme proteins.^128,130

Figure 11.

Cpd I generation in cytochrome c peroxidase (CcP), PDB: 4CVJ.

The presence of residues acting as acid–base catalysts is common in heme proteins that utilize H₂O₂ to generate compound I. For instance, the histidine/asparagine residue pair found in catalases functions similarly to the His/Arg pair in peroxidases to promote compound I generation,¹³¹ while chloroperoxidase (CPO) and unspecific peroxygenases (UPO, also sometimes termed APO) employ a Glu/His or Glu/Arg acid–base pair. In CPO, as well as the UPO from M. rotula, glutamate hydrogen bonds to a histidine that stabilizes the negative charge built on the glutamate during O–O bond cleavage (Figure 12a),^132,133 whereas in UPO from A. aegerita, glutamate hydrogen bonds to an arginine to achieve this stabilization effect (Figure 12c).^133,134 CYP152, a family of hydrogen peroxide-utilizing P450 peroxygenases, lack the key histidine or glutamate residue to promote O–O cleavage. Instead, these proteins utilize the carboxyl group from a fatty acid substrate as an acid–base catalyst to rearrange the protons of H₂O₂ and aid the compound I generation (Figure 12b).^135,136 An alternative pathway for cpd I generation involves a O–O homolytic cleavage of the ferric-H₂O₂ adduct followed by a hydrogen-atom abstraction (HAT) from the Fe(IV)–OH intermediate by the incipient hydroxyl radical.^137−139 Recent computational studies of P450_SPα (CYP152B1) by Shaik et al. indicated that the carboxyl group of the fatty acid substrate keeps the HO· radical toward the H–O bond of the Fe(IV)–OH intermediate and facilitates the HAT and cpd I generation (dashed bracket in Figure 12b). Inspired by the role of substrate carboxyl group in CYP152, Watanabe et al. have used short-alkyl-chain fatty acids as cocatalysts and decoy molecules to expand the substrate scope of CYP152 to include a variety of substrates that lack the carboxyl acid groups.^140,141 In this case, short-alkyl-chain fatty acids served as the proton donor required for O–O bond activation.

Figure 12.

Proposed mechanism for generation of cpd I from cpd 0. The following structures are used: (a) Cpd 0 of CPO (CPO-0) from Leptoxyphium fumago (PDB: 2J5M) and the active site structure of MroUPO (PDB: 5FUJ); (b) P450_BSβ from Bacillus subtilis complexed with a fatty acid substrate (PDB: 1IZO); (c) crystal structure of peroxygenase from Agrocybe aegerita (PDB: 2YP1); and (d) oxygen complex of P450_cam from Pseudomonas putida (PDB: 1DZ8).

Unlike peroxidases, most P450s generate their compound 0 through reduction of iron-bound dioxygen, initially generating a ferric-peroxo anion species. Therefore, two protons are needed to produce cpd I, one for protonating the distal oxygen to generate cpd 0 and the other for facilitating O–O bond cleavage to generate water.²⁵ As revealed by the crystal structure of P450cam, these protons are delivered through a hydrogen bonding network of a water channel involving the key residue Thr252 that hydrogen bonds to the distal oxygen and forms an extended hydrogen bonding network with water molecules in the active site (Figure 12d).^49,50 This threonine is highly conserved in most P450 enzymes.¹⁴² Mutation of this site to alanine does not affect the formation of cpd 0 but diminishes the hydroxylation reactivity, affording H₂O₂ and the ferric heme, a so-called uncoupling process.^21,94

The conversion of O₂-dependent P450s into peroxygenases that can utilize hydrogen peroxide as a cosubstrate and terminal oxidant has generated considerable current interest. The low cost and high step-economy of using H₂O₂ as well as the increased process efficiency derived from not having to recycle reducing cofactors (e.g., NAD(P)H) make such peroxygenase-P450s highly attractive as biocatalysts.¹⁴⁰ Protein engineering techniques such as directed evolution have been extensively used to obtain variants that can efficiently utilize H₂O₂ for monooxygenation reactions.^143,144 The mutation of a conserved phenylalanine (F87 in P450_BM3, Figure 13) within a substrate recognition site to a smaller residue (alanine, glycine, or valine) has been shown to be crucial for the introduction of peroxygenation activity into O₂-dependent P450s.^144,145 Recently, Watanabe et al. found that the mutation of the highly conserved threonine (T268 in P450_BM3, Figure 13) into glutamic acid substantially enhanced the peroxygenase activity of a number of P450s, including P450_BM3, P450_cam, and CYP119.¹⁴⁶ The glutamic acid mutation is thought to mimic the role of the substrate carboxyl group in CYP152 for O–O bond activation.

Figure 13.

Key mutation sites (F87 and T268) for the induction of peroxygenase activity in P450_BM3 (PDB: 1BVY).

In addition to O–O bond heterolytic cleavage of iron-peroxo intermediates to generate cpd I, some heme proteins can interact with alkyl hydroperoxides to form alkoxyl radicals and ferryl intermediates (cpd II). A representative example is allene oxide synthase, which converts fatty acid hydroperoxides into allene oxides. This reaction is discussed in detail in section 4.1.

2.3.2. O–O Bond Cleavage in Metal-Peroxo Porphyrin Complexes

Metal-mediated O–O bond cleavage was extensively explored in model metalloporphyrin compounds.¹⁵ In particular, acylperoxy-iron(III) porphyrins have afforded many insights to the mechanistic details of this step.^147−151 It has been shown that the addition of m-chloroperbenzoic acid (mCPBA) to the solution of HO–Fe^III(TMP) (22, Figure 14a) led to the formation of acylperoxy-iron(III) complex 23. This species undergoes facile acid-catalyzed O–O bond heterolysis to afford cpd I 24 with a first-order rate constant (k_hetero) of 6.1 × 10^–3 s^–1 (−48 °C) and an E_a = 4 ± 0.4 kcal/mol, ΔH^‡ = 3.6 ± 0.4 kcal/mol, and ΔS^‡ > −25 cal mol^–1 K^–1.^149,151 In nonpolar solvents and the absence of acids, however, O–O bond homolysis occurred to afford an unusual iron(III) porphyrin N-oxide species, 26, and a diacylperoxide as products, whereas a ferryl species, 25, was formed for alkyl acylperoxy-iron(III) complexes due to more rapid decarboxylation of the alkyl acyl radicals. The impact of the porphyrin meso-substituents and axial ligands on O–O bond cleavage have been investigated by Watanabe et al.¹⁵² They found that both electron-donating meso-substituents and imidazole axial ligands enhanced the O–O heterolytic cleavage (Figure 14b). For instance, the k_hetero of the most electron-donating [(mCPB)Fe^III(TTMPP)] was found to be ∼55–fold faster than that of (mCPB)Fe^III(TDMPP) (mCPB = m-chloroperbenzoate). The presence of 1-methylimidazole (1-MeIm) axial ligand increased k_hetero by over 50-fold for [(mCPB)Fe^III(TDMPP)]. The effects of axial ligands were also investigated for O–O bond homolysis in toluene, in which imidazole ligation showed a much less rate enhancement (2.5-fold for 1-MeIm).

Figure 14.

(a) O–O bond heterolysis and homolysis of acylperoxy-iron(III) porphyrins. (b) Effect of axial ligands and meso-substituents on O–O bond heterolytic cleavage of acylperoxy-iron(III) porphyrins.

More recently, acylperoxy-iron(III) porphyrin intermediates have been further studied with different porphyrin ligand scaffolds and axial ligands under both stoichiometric and catalytic conditions.^14,153−157 It was found that the rates of oxygen atom transfer reactions mediated by acylperoxy-iron(III) porphyrins were orders of magnitude slower than those of corresponding cpd I species,^{153,156−158} demonstrating that cpd I is the active intermediate for the oxygenation reactivity. Woggon and van Eldik et al. synthesized an iron(III) porphyrin with covalently appended sulfonate axial ligand (27, Figure 15).¹⁵³ Upon reacting with mCPBA, this complex generated cpd I with k_hetero of 2.4 ± 0.1 s^–1 (238 K). The faster k_hetero determined in this case was attributed to the high polarity of acetonitrile solvent. Nocera et al. studied the impact of the secondary coordinating environment on O–O bond cleavage with so-called “hangman” iron porphyrins containing either a carboxylic acid (HPX-CO₂H) or a methyl ester (HPX-CO₂Me) functional group, respectively (Figure 15).¹⁵⁵ Both complexes afforded cpd I upon mCPBA oxidation in CH₂Cl₂ with similar rates, whereas in toluene, the O–O homolysis was inhibited for Fe^III(HPX-CO₂H), and only Fe^III(HPX-CO₂Me) afforded the O–O homolysis product porphyrin N-oxide. This result suggests that the presence of the acidic group facilitates the 2e^– pathway for O–O bond cleavage in the same manner as peroxidases. The nature of the leaving group O–X within the peroxo unit [Fe–O–OX] also affected the preference between 2e^– and 1e^– pathways. Alkyl hydroperoxides such as ^tBuOOH and cumene hydroperoxide, which have weaker alkoxy leaving groups, were shown to follow a homolytic O–O bond scission pathway upon reacting with Fe(III) porphyrins.^159,160

Figure 15.

Chemical structures of compound 27, Fe^III(HPX-CO₂H) (28), and Fe^III(HPX-CO₂Me) (29).

The O–O bond cleavage of ferric–hydroperoxo intermediates are mainly studied indirectly in oxidation reactions.^{148,150,161−163} It has been shown that iron(III) porphyrins could react with H₂O₂ in protic solvents to afford porphyrin cation radical intermediate.^164,165 Recently, van Eldik studied the reaction between H₂O₂ and an iron(III) octa-anionic porphyrin complex (30, Figure 16).¹⁶⁶ They found that cpd I was formed at pH < 9, whereas one-electron oxidized oxoiron(IV) intermediate was observed at pH > 9. At pH = 8, the rate constants for the coordination and dissociation of H₂O₂ was determined to be k₁ = 2259 ± 149 M^–1 s^–1, k_–1 = 1.1 ± 0.1 s^–1, and O–O heterolytic cleavage rate constant was measured to be ∼0.18 s^–1. In another study, the same group also synthesized two different ferric–hydroperoxo intermediates: five-coordinate high-spin [Fe^III(TPFPP)(OOH)] and six-coordinate low-spin [Fe^III(TPFPP)(OH)(OOH)]⁻ (31 and 32, Figure 16).¹⁶⁷ These two species displayed very different modes of O–O bond cleavage in which the former underwent O–O homolysis to afford an oxoiron(IV) porphyrin complex and the latter formed cpd I via O–O heterolysis.

Figure 16.

Chemical structures of compound 30, Fe(TPFPP)(OOH) (31), and [Fe(TPFPP)(OH)(OOH)]⁻ (32).

Manganese-peroxo porphyrin complexes have afforded a particularly informative example of metal-mediated O–O bond scission. Groves and co-workers have prepared an acylperoxomanganese(III) porphyrin [(mCPB)Mn^III(TMP)] (33, Figure 17a) by both direct acylation of manganese-dioxygen intermediate and direct addition of peroxyacid salts to manganese(III) porphyrins.¹⁶⁸ Warming a CH₂Cl₂ solution of 33 from −78 °C to −30 °C led to the formation of an oxoMn(V) intermediate 34. Weiss et al. prepared a manganese(III)-peroxycarbonate intermediate 35 by treating manganese-dioxygen complex with CO₂ at −70 °C, which readily converted to oxoMn(V) species 36 by elevating the temperature to −35 °C (Figure 17b).¹⁶⁹ Bruice et al. studied the effect of axial ligands on oxygen transfer from various percarboxylic acids and alkyl hydroperoxides to manganese(III) porphyrins. They found that the nitrogen base ligation lead to more than a 100-fold increase in O–O bond cleavage rate and subsequent oxygen transfer.¹⁷⁰ Recently, Groves et al. have determined the activation parameters of O–O bond heterolysis for a water-soluble manganese porphyrin cpd 0, [Mn^III(OH)(OOH)(TDMImP)] (37, Figure 17c).¹⁷¹ That work determined that the optimal pH was in the range of 8.5–11.5 for the reaction between H₂O₂ and Mn^III(TDMImP). The H₂O₂ oxidation was significantly slowed outside this range. Significantly, O–O bond cleavage was pH-independent over that range, proceeding with a first-order rate constant of 66 ± 12 s^–1. The activation parameters, ΔH^‡ = 4.2 ± 0.2 kcal/mol and ΔS^‡ = −36 ± 1 cal mol^–1 K^–1, revealed a remarkably low enthalpic barrier for this reaction. The pH independence suggests a concerted “push-pull” mechanism for O–O cleavage, in which the axial OH ligand is partially deprotonated to become more electron-donating (“push”) and the terminal oxygen of hydroperoxo is partially protonated to release a water molecule to the medium (“pull” effect) to afford the product trans-dioxomanganese(V) species (Figure 17c).

Figure 17.

(a) Conversion of [(mCPB)Mn^III(TMP)] into oxoMn(V) intermediate 34. (b) Conversion of manganese(III)-peroxycarbonate 35 into oxoMn(V) intermediate 36. (c) “Push-pull” mechanism for the O–O bond heterolysis of [Mn^III(OH)(OOH)(TDMImP)].

2.4. Compound I: Oxoiron(IV) Porphyrin Cation Radicals

2.4.1. Compounds I of Heme Proteins

Compound I (cpd I) is the primary intermediate that gives rise to the diverse oxidative reactivities of most heme proteins. The electronic structure of cpd I is best described as an S = 1 ferryl coupled to an S = 1/2 porphyrin π cation radical, which can afford either doublet or quartet states.^172,173 The net spin of cpd I is thus determined by the exchange coupling parameter (J) between the ferryl and the porphyrin cation radical (Table 1). Weak ferromagnetic coupling (J = 0–10 cm^–1) and a net spin S = 3/2 have been observed for cpd I of several peroxidases and catalase,^173−178 whereas for heme-thiolate proteins like CPO and P450s strong antiferromagnetic coupling (|J| > 40 cm^–1) was seen and a total spin of S = 1/2 was observed using EPR and Mössbauer spectroscopy.^18,179−182 The two-oxidizing-equivalent intermediate in cytochrome c peroxidase (CcP) obtained after H₂O₂ oxidation, known as compound ES, has unusual features compared to those of cpd I of most heme proteins. The UV spectrum of CcP-ES lacks the characteristic 690 nm absorbance of the porphyrin cation radical, and its EPR spectrum exhibited a unique free radical signal (g_⊥ = 2.006, g_∥ = 2.034).¹⁸³ It is recognized now that CcP compound ES is a ferryl porphyrin intermediate like compound II with an uncoupled protein radical at a tryptophan residue near the axial histidine (Figure 11).¹⁸⁴

Table 1. Structural and Spectroscopic Parameters of Representative Compound I Intermediates of Heme Proteins.

entry	Stotal	J (cm–1)	D (cm–1)	|J|/D	δ (mm/s)	ΔEQ (mm/s)	d(FeIV=O) (Å)	ref
HRP-I	3/2	|J| ≤ 2	26	∼0.1	0.08	1.25	1.64a	(173)
Catalase-I	3/2	∼10	∼25	0.4	–	–	–	(176)
LiP-Ib	3/2	∼10	∼33	0.29	–	–	–	(174)
APX-Ic	3/2	–	–	0.28	–	–	–	(175)
CPO-I	1/2	–37	36	1.02	0.15	1.02	1.661d	(179)
CYP119A1-I	1/2	–52d	40d	1.30	0.11	0.90	1.670d	(182)
SeCYP119–1	1/2	–	–	1.43	0.07	1.49	–	(201)
CYP158A2*-Ie	1/2	–	–	–	0.13	0.90	1.669d	(260)
P450cam-I	1/2	–	–	1.38	0.11	0.90	–	(180)

^aFrom ref (187).

^bLiP = lignin peroxidase.

^cAPX = ascorbate peroxidase.

^dFrom ref (181).

^eCYP158A2 * refers to CYP158A2 Y352F mutant.

While cpd I of horseradish peroxidase and chloroperoxidase have been well characterized for decades,^{179,185−187} the direct observation and characterization of a cytochrome P450 cpd I (P450-I) was much more challenging. Best known for their roles in drug metabolism and oxidative tailoring in biosynthetic pathways, P450s catalyze remarkably selective aliphatic C–H functionalization reactions.¹⁶ The oxygen rebound mechanism (Figure 18), first proposed by Groves in the 1970s,^188−190 invoked the oxoiron(IV) porphyrin cation radical, compound I, as the “active oxygen species.” Compound I was proposed to activate aliphatic C–H substrates via hydrogen atom transfer (HAT) to the ferryl oxygen. Subsequently, the incipient substrate radical recombines with the Fe(IV)–OH of compound II (oxygen rebound) to afford hydroxylation products. Because of the exceptional reactivity of P450 enzymes, efforts to understand P450-I have attracted sustained attention for four decades. Further, these pursuits have provided valuable insights for the development of new catalysts for selective aliphatic C–H functionalization.

Figure 18.

Oxygen rebound mechanism for C–H hydroxylation catalyzed by cytochrome P450s.

The main obstacles for studying P450-I included its very short lifetime and low yield.^18,191−193 Chloroperoxidase cpd I (CPO-I) has long been used as an analog of P450-I because it also contains an axial cysteine thiolate. However, CPO-I can only hydroxylate weak benzylic C–H bonds.¹⁹⁴ This lack of reactivity for C–H activation led to proposals of other possible intermediates such as an oxoFe(V) electromer or Fe(III)–OOH for P450 reactivity.^195−197 The long-sought P450-I was finally elucidated in 2010.¹⁸² Rittle and Green greatly improved the yield of cpd I of a thermophilic cytochrome P450, CYP119. Spectroscopic studies of CYP119-I revealed a broad UV absorbance around 690 nm typical of a porphyrin cation radical (Figure 19a) and a Mössbauer isomer shift of 0.11 mm/s and quadrupole splitting of 0.90 mm/s signaling iron(IV). An antiferromagnetic coupling between S = 1 ferryl and porphyrin cation radical with a |J/D| of 1.3 was determined by EPR spectroscopy. These data clearly showed that CYP119-I is an antiferromagnetically coupled (S = 1/2) oxoiron(IV) porphyrin cation radical. Unpaired electron density of the axial thiolate was also suggested. More importantly, CYP119-I was found to be highly reactive for C–H hydroxylation with apparent second-order rate constants ranging from 10⁴ to 10⁷ M^–1 s^–1 (Figure 19b). Following the characterization of CYP119-I, compounds I of CYP158A2 Y352F variant and P450cam have also been characterized.^180,181 All these P450 compounds I have similar spectroscopic features and exhibit high reactivity toward alkyl C–H bonds.

Figure 19.

(a) UV–vis transients observed upon 1:1 mixing of 20 μM ferric CYP119 with 40 μM mCPBA. Maximum yield of CYP119-I was ∼70% at 35 ms. (b) Observed first-order decay rates vs camphor concentration. (c) UV–vis transients observed upon 1:1 mixing of 13 μM ferric AaeUPO with 25 μM mCPBA. Maximum yield of AaeAPO-I was ∼70% at 30 ms. (d) Observed first-order decay rates vs p-ethylbenzoic acid concentration. (a) and (b) were adapted with permission from ref (182). Copyright 2010 American Association for the Advancement of Science. (c) and (d) were adapted from ref (200). Copyright 2012 American Chemical Society.

Unspecific peroxygenases (UPO, initially termed APO for aromatic peroxygenase) comprise a large and newly discovered family of P450-like, heme thiolate hydroxylases of fungal origin.¹⁹⁸ These enzymes are showing considerable promise as biocatalysts with high reactivity toward even aliphatic hydrocarbons.¹⁹⁹ Compound I of an unspecific peroxygenase from the common edible mushroom Agrocybe aegerita (AaeUPO) was recently characterized by Wang and Groves et al.²⁰⁰ UPOs are very significant and unusual because they can oxidize strong C–H bonds with H₂O₂ as oxidant and do not require any cofactors. AaeUPO-I displays similar UV–vis features as CYP119-I and fast rate constants for the hydroxylation of strong C–H bonds (up to 100 kcal/mol) in the range of 10–10⁵ M^–1 s^–1 (Figure 19c and 19d).²⁰⁰

Very recently, Green et al. reported the characterization of a selenocysteine-ligated cpd I of CYP119 (SeCYP119-I).²⁰¹ Given the structural similarity and the difference in electronic properties between selenocysteine and cysteine, SeCYP119-I provides a good opportunity to study the effect of electron-donation of axial ligands on cpd I reactivity. Compared to CPO-I and CYP119-I, SeCYP119-I exhibits a larger |J/D| value (|J/D| = 1.43) and a smaller isomer shift δ = 0.07 mm/s, suggesting a stronger electron-donation from selenolate [ R-Se^–] axial ligand. Reactivity studies showed that CYP119-I and SeCYP119-I were formed at similar rates (5 × 10⁶ M^–1 s^–1 and 7 × 10⁶ M^–1 s^–1 for CYP119 and SeCY119, respectively) upon treating corresponding ferric enzymes with mCPBA. But SeCYP119-I was more reactive toward C–H bonds than CYP119-I, leading to a significantly less accumulation of SeCYP119-I during reactions with substrates.

2.4.2. Compounds I of Iron Porphyrins

The first synthesis and characterization of an oxoiron(IV) porphyrin cation radical model compound was achieved by our group in 1981.²⁰² Treatment of a sterically hindered iron porphyrin, Fe^III(TMP)Cl (39, Figure 20a) with mCPBA in CH₂Cl₂/MeOH (4:1) at −78 °C produced a green compound formulated as [Fe^IV(O)TMP^•+]L (40, Figure 20a). This species had a Soret band at 406 nm and a broad absorption at 645 nm, typical of a porphyrin cation radical. EXAFS data showed a short Fe–O bond length of 1.6 Å. A ferryl iron–oxygen stretch, ν(Fe=O) of 801 cm^–1 (L = Cl) and 828 cm^–1 (L = MeOH), was obtained by resonance Raman spectroscopy.²⁰³ Magnetic susceptibility and EPR studies showed a total spin of 3/2 with strong ferromagnetic coupling between the ferryl and porphyrin cation radical (J > 40 cm^–1). These features stand in sharp contrast to the antiferromagnetic coupling observed in P450 cpd I.²⁰⁴ This green species [O=Fe^IVTMP^•+] was found to be reactive in oxygen atom transfer reactions.²⁰⁵ The direct oxidation of norbornene by [O=Fe^IVTMP^•+] in the presence of H₂¹⁸O led to the formation of an epoxide product with 99% ¹⁸O incorporation, indicating facile ferryl–oxygen exchange with water.

Figure 20.

(a) The synthesis of the first model cpd I (TMP^•+)Fe^IV=O and its structural and electronic parameters. (b) Depiction of the C–H activation mode of (TMP^•+)Fe^IV=O and its Mössbauer spectrum, which indicates high-valent iron. Adapted from ref (205). Copyright 1983 American Chemical Society.

A number of compound I analogues were synthesized with varied substitutions at meso- and pyrrole-β positions of the porphyrin ligand. These studies have been summarized in a number of reviews.^{15,26,27,31,32,206−208} In general, both meso- and pyrrole-β substituents affect cpd I electronic structures via altering the energy of porphyrin SOMO/HOMO orbitals a_1u and a_2u (Figure 21a). In unsubstituted porphyrins, a_2u is the higher-energy orbital and is singly occupied in cpd I. Since a_2u has a large spin density at meso-positions (as well as the pyrrole nitrogens), the increase of electron-deficiency of meso-substituents will substantially stabilize a_2u.²⁷ For instance, the most electron-withdrawing pentafluorophenyl derivative, Fe^IV(O)(TPFPP^•+), has an a_1u porphyrin radical configuration (Figure 21b). Fujii et al. showed that compounds I with pyrrole β-substituents are generally a_1u porphyrin radicals, and the electronic properties of pyrrole β-substituents were shown to have similar influences on a_1u and a_2u (Figure 21c). The nature of the porphyrin radical (a_1u or a_2u) has a large impact on its exchange interaction with iron center, in which large ferromagnetic coupling (J > 30 cm^–1) was observed for a_2u radicals due to large pyrrole nitrogen orbital coefficients and, by contrast, weak antiferromagnetic coupling (J < 10 cm^–1) for a_1u radicals.^209,210

Figure 21.

(a) Orbital diagrams of porphyrin a_1u and a_2u orbitals. (b) Effect of meso-substituents on the electronic structure of cpd I. (c) Effect of pyrrole-β substituents on the electronic structure of cpd I. (b and c) were adapted from ref (27). Copyright 2002 Elsevier B. V.

Although these cpd I analogs display a range of oxygen transfer reactions, their second-order rate constants for C–H activation (k) are generally less than 10 M^–1 s^–1.²¹¹ These values for the model compounds are 3 to 6 orders of magnitude slower than those recently measured for P450-I or UPO-I. A highly reactive synthetic compound I was reported by our group (Figure 22).²¹² This species was generated by reacting a cationic iron(III) porphyrin complex Fe^III(4-TMPyP) with mCPBA under rapid-mixing conditions. The resulting [Fe^IV(O)(4-TMPyP^•+)] could oxidize a range of C–H bonds with second-order rate constants, ranging from (2.85 ± 0.3) × 10⁶ M^–1 s^–1 for xanthene to (7.5 ± 0.1) × 10³ M^–1 s^–1 for ethylbenzoic acid. The rate constant of ethylbenzoic acid oxidation is comparable to those of CYP119 and CPO cpd I for ethylbenzene oxidation.²¹³ It was suggested that stabilization of the porphyrin a_2u HOMO and the increased electron deficit at the metal core due to the cationic porphyrin contribute to the very high activity of [Fe^IV(O)(4-TMPyP^•+)]. The significance of these findings is that no unusual properties of the protein architecture need to be invoked to explain high rates of C–H bond cleavage by heme proteins.

Figure 22.

Observed rate constant k_obs at 673 nm vs xanthene concentration for its oxidation by 41 at pH 4.7, 14.5 °C, yielding second-order rate constant k₃ = (2.8 ± 0.3) × 10⁶ M^–1 s^–1. Inset: correlation of k₃ with C–H bond dissociation energy. Adapted from ref (212). Copyright 2009 American Chemical Society.

Very recently, Sorokin and co-workers have reported the synthesis of a series of dimeric oxoiron(IV) cation radical species [(L)Fe^IV=N–Fe^IV(L^•+)=O] by treating μ-nitrido diiron porphyrins or phthalocyanines (Pc) with peroxides or mCPBA (42 and 43, Figure 23).^214,215 A high-valent μ-nitrido diiron–oxo porphyrin cation radical intermediate [(TPP)Fe^IV=N–Fe^IV(TPP^•+)=O] was successfully isolated and characterized (44, Figure 23).²¹⁶ This species has a green color similar to other porphyrin cation radicals and has a half-life of ∼20 min at −80 °C. The single narrow EPR signal at g = 2.001 suggested an S = 1/2 state with the unpaired electron localized on the porphyrin macrocycle. In line with the EPR results, magnetic susceptibility measurements revealed a strong antiferromagnetic coupling between two iron centers with −J_Fe–Fe > 600 cm^–1, which leads to a S_FeFe = 0. The Mössbauer spectrum showed a very small isomer shift (0.00(1) mm/s) and a moderate quadrupole splitting (0.75(2) mm/s), also consistent with iron(IV). These high-valent μ-nitrido diiron–oxo complexes have exceptional activity for C–H activation.²¹⁷ Under catalytic conditions, they were shown to activate C–H bonds of methane (BDE = 104 kcal/mol) to afford formic acid with total turnover numbers (TTN) up to ∼220.²¹⁸ Recent computational studies showed that the activation barriers of [(L)Fe^IV=N–Fe^IV(L^•+)=O] type intermediates for methane oxidation are over 10 kcal/mol lower than that of a model P450 cpd I.²¹⁹ The calculation also suggests that the donor properties of the bridged nitrido group dramatically increased the basicity of the oxo group, which is crucial for the high reactivity of this class of compounds.

Figure 23.

Structures of μ-nitrido diiron porphyrins and phthalocyanines (Pc) and formation of cpd I 44 from 42 and mCPBA.

2.4.3. Manganese Porphyrin Analogues of Compound I

Synthetic manganese porphyrin analogues of compound I have long been known to have high reactivity toward C–H bonds.²²⁰ While relatively stable oxoMn(IV) porphyrin species have been isolated and well-characterized,^221,222 the direct characterization of oxoMn(V) porphyrin complexes has been very challenging due to their high reactivity. The presence of reactive oxoMn(V) porphyrin complexes has been usually implicated from reactivity patterns and ¹⁸O-exchange into products from water.^223−225 The first Mn porphyrin compound I (MnPor-I) to be kinetically evaluated was generated by treating Mn^IIITMPyP with a peroxyacid such as mCPBA using rapid-mixing stopped-flow techniques (45 to 46, Figure 24).²²⁶ MnPor-I 46 rapidly converted to previously characterized oxoMn(IV) via 1e^– reduction with a first-order rate constant of 5.7 s^–1. Unlike compound I of iron porphyrins, MnPor-I was diamagnetic, displaying a well-resolved and unshifted proton NMR spectrum and lacked the characteristic broad absorbance between 650–700 nm of porphyrin cation radical, suggesting MnPor-I was an oxoMn(V) species.²²⁷46 catalyzed very fast olefin epoxidation with a second-order rate constant of 6.5 × 10⁵ M^–1 s^–1. A 35% ¹⁸O incorporation into the epoxidation products was also observed when carrying out the reaction in the presence of H₂¹⁸O, which is in accordance with an aqua–oxo interconversion via prototropy as suggested by Meunier (47 to 48, Figure 24).^223,225 By contrast, manganese-substituted heme proteins, such as Mn-HRP, form only manganese(IV) species and a protein radical.²²⁸ Apparently, the reduction potential of oxoMn(V)protoporphyrin IX is too high to be tolerated by the more redox sensitive protein.

Figure 24.

Generation of a cpd I manganese porphyrin analogue and the oxo-aqua interconversion of oxoMn(V) intermediate.

A related oxoMn(V) porphyrin was synthesized by treating cationic [Mn^III(TF₄TMAP)](CF₃SO₃)₅ (49, Figure 25) with two equivalents of alkaline hydrogen peroxide.²²⁹ This species transformed olefins to epoxides with good yields (∼50%). In line with the ¹⁸O exchange experiment of 46, the epoxidation mediated by 49 with H₂¹⁶O₂ in H₂¹⁸O water led to a 45% ¹⁸O incorporation. Newcomb et al. prepared a series of oxoMn^V porphyrins (Figure 25b) in organic solvents via laser flash photolysis (LFP) of corresponding Mn^III perchlorate porphyrin complexes.²³⁰ The second-order rate constants for olefin epoxidation by these complexes were in the range from 10⁴ to 10⁶ M^–1 s^–1, consistent with that of water-soluble oxoMn^V(TMPyP). For oxoMn^V(TPFPP), the second-order rate constant for ethylbenzene oxidation was determined to be 1.2 × 10⁵ M^–1 s^–1, more than 10⁵ times faster than the corresponding iron complex.²¹¹ A dinuclear oxoMn^V porphyrin complex has been described by Naruta et al. This species was prepared from the oxidation of a manganese(III) porphyrin dimer with mCPBA in the presence of excess tetrabutylammonium hydroxide (ⁿBu₄OH) (Figure 25c).²³¹ Interestingly, decomposition of this oxoMn^V dimer produced oxygen. A number of oxoMn^V complexes have been prepared with other porphyrinoid ligands such as corroles and corrolazines, as recently reviewed by Goldberg et al.²⁹

Figure 25.

(a) Chemical structure of Mn^V(O)(TF₄TMAP). (b) Generation of oxoMn^V intermediates via laser flash photolysis (LFP). (c) A dinuclear oxoMn^V intermediate 52.

In addition to these oxoMn^V complexes, an unusual trans-dioxoMn(V) porphyrin structure was deduced from the pH-dependent halide oxidation mediated by oxoMn^V porphyrins. Apparent second-order rate constants spanned a range of 5 orders of magnitude from ∼10⁷ M^–1 s^–1 at pH 5.2 to ∼10² M^–1 s^–1 at pH 9.²³² This profound pH dependence was analyzed to result from two acid–base equilibria between the trans-dioxoMn(V), oxo–hydroxo Mn(V), and oxo-aqua Mn(V) species, which displayed increasing oxygen-atom-transfer activities along the sequence. Proton ionizations with pK_a1 < 5 and pK_a2 = 7.7 were estimated for oxo-aqua species from the Nernst eq (Figure 26a). The synthesis and characterization of such trans-dioxoMn(V) porphyrin complexes were reported by Groves and Spiro. These complexes were synthesized by treating Mn(III) porphyrins with H₂O₂ in the presence of excess ⁿBu₄OH (Figure 26b).^233,234 These compounds exhibited Raman ν_sym(O=Mn^V=O) values of 741–744 cm^–1, which were decreased by 21–26 cm^–1 and 39–44 cm^–1, respectively, upon single and double ¹⁸O substitution on O=Mn^V=O moiety. Significantly, ν_assym(O=Mn^V=O) values detected by IR were predictably different, confirming the linear, triatomic structure of the dioxoMn(V) unit (Figure 26c). A similar Mn^V(O) intermediate was synthesized by Nam et al. and was assigned as a six coordinate trans-oxohydroxo Mn(V) complex.²³⁵ However, a Mn–O distance of 1.68 Å determined by EXAFS indicates a double rather than a triple bond between Mn and oxygen. It is very likely that this species is also a trans-dioxoMn(V) complex.

Figure 26.

(a) pH-dependent reactivity of Mn(III) porphyrins for halide oxidation and the acid-base equilibria between the trans-dioxoMn(V), oxo–hydroxo Mn(V), and oxo-aqua Mn(V) species. (b) Synthesis of a trans-dioxoMn(V) porphyrin 54. (c) Characterization of 54 via rR and IR spectroscopies. (d) Mn–O bond length vs 1/F^1/3(Mn–O). (c) and (d) were adapted from ref (233). Copyright 2007 American Chemical Society.

On the basis of the ν_sym and ν_assym data obtained for compound 54, the Mn–O bond force constant (F) and stretch–stretch constant (k) were determined to be 454 and 67.2 N/m, respectively. Using the recently reported EXAFS bond length of 54 and available data of various terminal monooxo-manganese complexes,²³⁵ a linear correlation was found between F^–1/3 and Mn–O bond lengths (Figure 26d), in line with predictions of Badger’s rule.^236,237 A similar correlation between F^–1/3 and Fe–O bond lengths has previously been reported by Green et al. for oxoiron(IV) porphyrins.²³⁶ Although trans-dioxo metal porphyrins have been characterized for second- and third-row transition metals such as Ru^VI and Os^VI,^238−242 manganese remains the only known first-row transition metal that can form a trans-dioxo species. The isolated trans-dioxoMn(V) porphyrins were less reactive toward typical substrates but could be activated via decreasing the pH. This pH-dependent reactivity supports the previously proposed acid–base equilibria between the trans-dioxoMn(V), oxo–hydroxo Mn(V), and oxo-aqua Mn(V) species and their increasing reactivity for oxidation reactions.^227,243 DFT calculations indicated that this trend correlated with corresponding energy gaps between unreactive singlet ground state and higher-energy triplet and quintet states.²⁴⁴

2.5. Compound II and Its Protonation State–ferryl or Hydroxoiron(IV)

The one-electron-reduction of oxoiron(IV) porphyrin cation radicals, cpd I, affords a ferryl porphyrin (cpd II), which also has an iron(IV) center (S = 1), similar to that of cpd I but lacking the porphyrin cation radical. The protonation state of cpd II has generated considerable research interest because of its importance in determining cpd I reactivity (see section 3.1).²⁴⁵ HRP-II, Mb-II, and CcP compound ES were among the earliest cpd II or cpd II-like species to be spectroscopically characterized.^{115,122,173,246−248} The first synthetic ferryl porphyrin model compound was described by Balch et al.²⁴⁹ X-ray absorption measurements of these intermediates showed a Fe–O bond lengths in the range of 1.64–1.70 Å, indicating a Fe=O unit.^{248,250−252} The Fe=O structure is also consistent with the resonance Raman studies in which Fe(IV)–O stretching vibrations were detected in the range from 740 to 870 cm^–1.²³⁶ These results, however, contradicted data obtained with X-ray crystallography, in which much larger (>1.80 Å) Fe–O bond distances were determined.^48,253,254 This discrepancy arose mainly from the radiolytic reduction caused by the hydrated electrons generated by the X-ray irradiation during data collection. For instance, Poulos and co-workers found that the Fe–O distance of CcP-ES changed linearly with the radiation dosage. The Fe–O distance measured at high-X-ray doses was 0.17 Å longer than that measured at low doses (1.9 Å vs 1.73 Å).²⁵⁵ Structures obtained by nonionizing neutron cryo-crystallography also support a Fe(IV)=O structure for CcP-ES with a Fe–O distance around 1.6 Å.¹²⁶ Since HRP-II, Mb-II, and CcP-ES exist in Fe(IV)=O state with pH as low as 4.5, their conjugate acid (Fe(IV)–OH) would have to be highly acidic (Table 2).²⁴⁵ Recently, a more accurate measurement of the Mb-II pK_a was performed by Green et al. An upper limit of 2.7 for pK_a(Mb-II) was determined by probing Mb-II over a wide pH range (3.9–9.5) with a combination of EXAFS, Mössbauer, and resonance Raman spectroscopies.²⁵⁶ No evidence of a protonated Mb-II was obtained.

Table 2. Structural and Spectroscopic Parameters of Representative Compound II of Heme Proteins^a.

entry	d(FeIV–O)b (Å)	δ (mm/s)	ΔEQ (mm/s)	pKa(FeIVO–H)	ref
HRP-II	1.70 ± 0.02 (6.0)	0.03 (6.9)	1.61 (6.9)	≤3.5	(245)
Mb-II	1.66 (3.9)	0.07 (3.9)	1.58 (3.9)	≤2.7	(256)
CcP-ES	1.67 (6.0)	0.05 (7.0)	1.55 (7.0)	≤4	(245)
Catalase-II	1.78 (5.3)	0.02 (5)	2.28 (5)	13.1	(262)
CPO-II	1.82 (6.5)	0.10 (6.5)c	2.06 (6.5)c	≥8.2	(250)(259),
CYP119-II	–	0.12 (8.0)	1.94 (8.0)	12.2	(260)
CYP158-II	1.84 (9.0)	0.10 (9.0)	2.05 (9.0)	12.0	(260)

^apHs are shown in parentheses.

^bFe–O bond distances obtained from EXAFS studies.

^cParameters for the major component. A minor component is present with δ = 0.11 mm/s and ΔE_Q = 1.59 mm/s.

The cpd II of chloroperoxidase (CPO) stands in sharp contrast to HRP-II, Mb-II, and CcP-ES. The EXAFS measurements by Green, Dawson, and Gray revealed a Fe–O length of 1.82 Å at pH 5.5 and pH 6.7.²⁵⁰ The visible spectrum of CPO-II remained unchanged from pH 3–7, suggesting a pK_a(CPO-II) ≥ 8.2 (assuming CPO-II is 95% protonated at pH = 6.9). The unusual basicity of CPO-II is explained by the strong electron donation by the axial thiolate, which leads to a more basic ferryl oxygen. This “push” effect of a cysteine axial ligand suggests that cpd II of other heme-thiolate proteins might also be basic.^257−259 In 2013, Green and co-workers studied the protonation event of a P450 cpd II (CYP158-II) using rapid-mixing pH-jump experiments coupled with a variety of spectroscopic techniques. Both deprotonated and protonated forms of CYP158-II were well-characterized, and a remarkably basic pK_a of 12 was determined.²⁶⁰ The ferryl protonation led to an elongation of Fe–O bond from 1.68 to 1.84 Å and a shortening of Fe–S bond from 2.36 to 2.27 Å along with an increase in quadrupole splitting and isomer shift, consistent with the generation of a Fe(IV)–OH species. The pK_a(cpd II) of another P450, CYP119, was also determined in the same study. Unlike CYP158, wild-type CYP119 did not accumulate cpd II. Mutational studies have shown that Y352 is important for generation of CYP158-II, as it can serve as an electron source to reduce cpd I to cpd II. In CYP119, a leucine residue (L316) is in the position corresponding to Y352 in CYP158. The redox inactive leucine might prevent efficient cpd I reduction to generate cpd II. Indeed, an L316Y mutation was found to substantially increase the yield of CYP119-II. Despite having a different active site environment, CYP119 displayed a very similar pK_a(cpd II) to that of CYP158. Very recently, the protonation of cpd II of a heme-thiolate peroxygenase, AaeAPO, has been investigated by Groves and co-workers.²⁶¹ A pK_a of 10.0 was obtained from pH-jump double-mixing stopped–flow experiments.

Very recently, Green et al. reported a pK_a of 13.1 for cpd II of Helicobacter pylori catalase (HPC).²⁶² The high pK_a value of HPC-II suggests that the tyrosinate axial ligand of catalase is electron-donating enough to maintain a highly basic ferryl, even though it forms a hydrogen bond with a conserved arginine residue. Another case of a basic cpd II without a thiolate axial ligand was recently reported by Raven and Moody. They showed that cpd II of ascorbate peroxidase (APX-II) exists as Fe^IV–OH.²⁶³ A neutron diffraction analysis of APX-II prepared via soaking APX crystals in m-CPBA solution revealed a long Fe–O bond distance of 1.84 Å, apparently confirming an earlier X-ray determination for that protein by the same authors.²⁶⁴ The cause of protonation in APX-II and the protonation state of compound II of other peroxidases remains to be determined.

The basicity of metal–oxo moieties in model compounds has been much less studied. Recently, Boaz and Groves investigated the protonation of cpd II of several water-soluble sulfonated iron porphyrins.²⁶⁵ A two-proton equilibrium with an apparent pK_a^obs = 5.5 was determined for the protonation of [O=Fe^IV(TMPS)(OH₂)] (56, Figure 27a). The resulting bis-aqua complex is an iron(III) porphyrin cation radical species 55 that is at the same net oxidation state. The absence of the accumulation of monoprotonated species 57 indicated that the first ferryl protonation with pK_a1 (∼4) must produce a much more basic intermediate, pK_a2 ∼ 7. The increase in basicity is apparently due to a change in iron oxidation state from a protonated ferryl to a hydroxoferric porphyrin cation radical through an electromeric equilibrium after the first protonation. The effect of this two-proton electromeric equilibrium was shown to have a significant effect on the C–H scission reactivity of the corresponding compound I analog, [O=Fe^IV(TMPS^•+)(OH₂)]. In terms of the Bordwell equation, such a two-proton equilibrium leads to an increase in the overall driving force of 5–6 kcal/mol because two O–H bonds are being formed at the transition state (defined as D(OH₂) in Figure 27a), with one proton coming from the substrate and the other from the medium.

Figure 27.

(a) Basicities of model hydroxo- and oxoiron(IV) porphyrin complexes. The second protonation provides an additional 5 kcal/mol driving force for HAT. (b) Acidities of several Mn^IV–OH complexes.

Several oxoMn(IV) porphyrins have been synthesized and characterized in the presence of strong base.^221,222 Fujii et al. successfully prepared and characterized a series of Mn^IV=O and Mn^IV–OH complexes with salen ligands via direct deprotonation of corresponding Mn^III–OH₂ salen cation radical complexes.^266,267 But the pK_a’s of these Mn–OH species have yet to be reported. Since manganese is less electronegative than iron, Mn(IV)–OH should be less acidic than Fe(IV)–OH. There are several acidity data reported for Mn^IV–OH complexes with corrolazine or nonheme ligands (Figure 27b).²⁶⁸ For instance, a lower limit around 15 was estimated for a Mn^IV–OH corrolazine complex 58 (Figure 27b).²⁶⁹ A similar Mn^IV–OH pK_a (∼15) was also estimated for a nonheme Mn^IV–OH species 59.²⁷⁰ Busch et al. synthesized a dihydroxomanganese(IV) complex 60 based on a cross-bridged cyclam ligand. pK_a of 6.86 and 10 were determined for the first and the second deprotonation, respectively.²⁷¹

2.6. Preventing Protein Oxidative Damage by Cpd I

Compounds I of P450s and UPOs have remarkable activity and can activate very inert C–H bonds in substrates. It is not a coincidence that they also exhibit the highest compound II basicities among heme proteins. As discussed in detail in section 3.1.1, the thermodynamic driving force for C–H activation by cpd I depends on the reduction potential of cpd I and the basicity of cpd II. According to the Bordwell equation,^272−274 an increase of one pK unit in cpd II basicity would enhance the O–H bond strength of cpd II by 1.37 kcal/mol. Therefore, the high cpd II basicity of P450s and UPOs would allow them to break strong C–H bonds with relatively low reduction potentials, which, at least in part, prevents the oxidative damage of protein superstructure by cpd I. The presence of chains of near-by tyrosine and tryptophan residues could provide further protection of enzymes during these oxidative transformations.^275,276 These redox-active residues have been proposed to transport oxidizing equivalents of cpd I to protein surfaces. This protection machinery could be important for P450 catalysis, in which compounds I that are not consumed due to inefficient substrate oxidation could be effectively quenched. For instance, a five-residue Tyr/Trp chain has been identified in CYP11A1, which includes residues Trp87, Trp231, Tyr90, Tyr93, and Tyr94 (Scheme 1). Trp87 is critical for triggering the protection machinery. Once Trp87 gets oxidized by cpd I to form a cation radical, the oxidizing equivalent on Trp87 will rapidly transit through Trp231 and Tyr90 to surface-exposed Tyr93 and Tyr94. Kinetic simulations of CYP11A1 have indicated a ∼3 μs survival time for cpd I before its oxidizing equivalents are transferred to the protein surface.²⁷⁵ In this regard, C–H activation by cpd I could be viewed as an inner sphere process that competes with a slower, long-range outer-sphere electron transfer.

Scheme 1. Chain of Redox Active Trp/Tyr Residues in CYP11A1 (PDB: 3N9Y).

3. C–H Functionalization Reactions Mediated by Oxoiron(IV) Porphyrin Cation Radicals (Compound I)

The majority of reactions mediated by cpd I are initiated via hydrogen atom transfer (HAT). These reactions not only include aliphatic hydroxylation as introduced above but also encompass a broad range of oxidative transformations such as desaturation, decarboxylation, C–C bond scission, and a variety of rearrangement reactions. Like C–H hydroxylation, these alternative reaction pathways involve an initial hydrogen atom transfer (HAT) via cpd I and the generation of an intermediate substrate radical. A common mechanistic feature shared by these reactions is the redirection of substrate to other reaction pathways leading to nonoxygenated products.^277−279 This strategy represents a general approach to develop novel catalytic systems for developing novel C–H functionalization reactions. In this section, we will first discuss fundamental aspects of HAT and oxygen rebound steps in cpd I-mediated reactions, as well as various factors that affect these two steps. Further, we will introduce several representative examples of nonconventional C–H activation transformations catalyzed by P450s and metalloporphyrins. In particular, we will delineate how substrate radical intermediates are controlled to achieve different types of reactions.

3.1. Factors that Affect Oxoiron(IV) Porphyrin Cation Radical Reactivity for HAT

3.1.1. Thermodynamic Properties of Cpd I

For a given C–H bond, the driving force of HAT reactions will be determined by the free energy of the O–H bond formed in cpd II [D_O−H(cpd II)]. Using thermochemical cycles to estimate bond dissociation free energies from measurable pK_a values and oxidation potentials has a long history,²⁸⁰ which was systematically developed by Bordwell and extended to metal–oxo species by Mayer.^272−274 In accordance with the Bordwell equation (eqs 1 and 2, Scheme 2).^272−274 The bond dissociation free energies of an X–H bond can be derived from the X–H pK_a and the X^– oxidation potential using a Hess cycle. Thus, an increase in cpd II ferryl basicity [higher pK_a for Fe(IV)O–H] would lead to a stronger O–H bond formed in cpd II and a more thermodynamic driving force (1.37 kcal/mol per pK_a unit) for the C–H abstraction by cpd I. For instance, the more electron-donating cysteine (thiolate) axial ligand in P450s can increase the basicity of compound II by over 7 pK units compared to the histidine-ligated myoglobin and horseradish peroxidase.^19,260,281 Considering the pK_a term in the Bordwell equation, this change in basicity corresponds to at least a 9 kcal/mol increase in O–H BDE of compound II.^260,261

Scheme 2. Determination of D(O–H) from the One-Electron Reduction Potential of Compound I and the pK_a of Compound II.

C_G is a constant depending on the solvent and electrode (57.6 kcal/mol for aqueous solution and the normal hydrogen electrode). When E′(cpd I) at a given pH is determined, eq 2 can be used to calculate D(O−H). D(O–H) obtained via eq 1 and eq 2 is the bond dissociation free energy (BDFE). If the entropic difference between protonated and unprotonated cpd II is negligible, the same equations can be used to calculate the bond dissociation enthalpies (BDE) of O–H bonds using constant C_H (55.8 kcal/mol for aqueous solution and normal hydrogen electrode).²⁷⁴

The reduction potential of cpd I to cpd II also plays an important role in determining cpd I reactivity. E′(cpd I/cpd II) of horseradish peroxidase (HRP) has been determined to be ∼0.95 V versus NHE at pH 6.0 (Table 3).²⁸² Intriguingly, HRP showed almost identical reduction potentials for cpd I and cpd II. It is much harder to determine E′(cpd I/cpd II) for P450s and CPO because of the short lifetimes of these intermediates. Instead, the two-electron reduction potential E′(cpd I/ferric) has usually been measured, which, in turn, can be used to estimate E′(cpd I/cpd II) with certain assumptions. For instance, assuming a similar redox potential for CPO-I and CPO-II, Green, Dawson, and Gray estimated the redox potential of CPO-I to be 1.3 ± 0.1 V at pH 6, which is 0.35 V higher than that of HRP, corresponding to an ∼8 kcal/mol increase in D_OH(cpd II).²⁵⁰ This enhancement together with the contribution of a basic CPO-II led to an overall over 11 kcal/mol increase in D_OH(CPO-II) compared to D_OH(HRP-II).

Table 3. pK_a and BDE_OH of Cpd II and Reduction Potentials of Cpd I.

entry	E′(cpd I/cpd II) vs NHE	pKa(cpd II)	BDEOH(cpd II)	ref
HRP	0.95 V (pH 6.0)	≤3.5	∼87 kcal/mol	(282)
CPO	1.3 ± 0.1 V (pH 6.0)	≥8.2	∼98 ± 4 kcal/mol	(250)
AaeUPO	1.4 V (pH 7.0)	10.0	∼100 kcal/mol	(283)

Very recently, Wang and Groves were able to determine the thermodynamic properties of the heme-thiolate peroxygenase AaeAPO-I.²⁸³ This approach utilized the rapid, reversible oxygen atom transfer between cpd I and halide ions to determine the two-electron cpd I reduction potential. The measured E′(cpd I/ferric) was in the range of 1.12–1.32 V versus NHE from pH 7.0 to 3.0. From plots of the rate for C–H abstraction versus substrate C–H BDE, the D_OH(CPO-II) was estimated to be 100 kcal/mol. This value leads to an E′(cpd I/cpd II) of 1.4 V and E′(cpd II/ferric) of 0.8 V at pH 7.0 (Figure 28). This estimate was further supported by the reaction of nitroxyl radicals with AaeAPO-I. 3-Carboxyproxyl (E^o ∼ 0.9 V) reacted rapidly with APO-I to form APO-II, which decayed much more slowly under these conditions. This pronounced reactivity difference shows that APO-II is a much milder oxidant. By contrast, the two analogous reduction potentials of HRP-I are both around 0.9 V. Thus, APO has increased the reduction potential of APO-I to perform the difficult C–H scission at the expense of APO-II, which needs only to interact with the incipient substrate radical. With nitroxyl radical as a selective reductant for APO-I, APO-II could be produced in high yield and the pK_a has been determined to be 10.0 via rapid-mixing, pH-jump spectrophotometry. With these values, the D_OH(APO-II) was calculated to be around 100 kcal/mol, consistent with the estimate above determined from kinetic measurements.

Figure 28.

Selective reduction of APO-I to APO-II by a nitroxyl radical with E^o = 0.9 V. Redox potentials are obtained from ref (283)

3.1.2. Effect of Axial Ligands and Porphyrin Peripheral Substituents on Cpd I Reactivity

It is clear from the above discussion that axial ligands can have a huge impact on cpd I reactivity. Such effects have been observed in a number of model compound studies. For example, a 650-fold rate enhancement was reported for cyclohexene hydroxylation by changing the axial ligand of Fe^III(TPFPP)L from CF₃SO₃^– to Cl^–.²⁸⁴ Subsequent DFT calculations showed that this ligand replacement was predicted to cause a 6 kcal/mol increase in Fe^IVO–H BDE, and that the main attribute of this BDE enhancement is the increase in cpd II basicity.²⁸⁵ The axial ligand effect on C–H hydroxylation was also investigated for Fe^IV(O)(TMP^•+) with para-substituted pyridine N-oxide (p-Y-PyO) axial ligands.²⁸⁵ In this system, the hydroxylation reactivity increased with the increasing electron-donating ability of para-substituent Y. A good correlation between calculated activation energies and Fe^IVO–H acidity was obtained, which, again, suggested that axial ligand could affect cpd I reactivity via tuning cpd II basicity.

While the increase in cpd II ferryl basicity via introducing electron-donating axial ligands has now been observed, the effect of axial ligands on cpd I reduction potential [E^o(cpd I)] is less clear. A number of theoretical calculations have shown that E^o(cpd I) decreased with increasing electron-donation from axial ligand and that cpd I with negatively charged ligands such as F^– and Cl^– exhibited lower E^o(cpd I) than those with neutral ligands such as imidazole.^286−289 However, Fujii et al. measured the reduction potential of a series Fe^IV(O)(TMP^•+)L and found an opposite trend (Figure 29). With L = ClO₄^– as a reference, a 0.08–0.16 V positive shift of E^o(cpd I) was determined for anionic axial ligands such as Cl^–, F^–, CH₃CO₂^–, and benzoate, whereas a negative shift (−0.08 to −0.12 V) was found for neutral imidazole-type ligands.²⁹⁰ The cause of this discrepancy between theory and experiments is not clear at this writing and further investigations are needed.

Figure 29.

Redox potentials (vs SCE^a) of cpd I models in CH₂Cl₂ at −60 °C. ^aUnder these conditions, the redox potential (E_1/2) of ferrocene was 0.460 V at −60 °C and 0.528 V at 20 °C. ^bValues in parentheses were obtained from differential pulse voltammetry. ^cIn CH₃CN at −35 °C. ^dIn CH₂Cl₂/CH₃CN (1:1). Adapted from ref (290). Copyright 2011 American Chemical Society.

Synthetic porphyrin ligands almost always contain substitutions at the pyrrole β-carbon and/or the meso-porphyrin positions. These substitutions, as discussed in previous sections, are important for the reactivity of synthetic metalloporphyrins. They can prevent undesired side reactions that diminish the reactivity of metalloporphyrin catalysts during the oxidative reactions, such as the formation of μ-oxo dimers and the destruction of porphyrin ligands. As discussed in section 2.4.2, the electronic properties of meso- and pyrrole-β substituents substantially affect the energy of porphyrin a_1u and a_2u π–orbitals, thus changing the singly occupied orbitals of porphyrin radicals of compounds I. Although the electronic structure of the porphyrin radical (a_1u or a_2u) apparently does not affect cpd I reactivity very much,²⁷ the introduction of electron-withdrawing groups at both positions were shown to greatly increase the reduction potential of cpd I (Figure 29)^290,291 and thus increase the thermodynamic driving force of HAT.^209,290 Indeed, the second order-rate constants for aliphatic hydroxylation mediated by an electron-deficient cpd I Fe^IV(O)(TPFPP^•+) were 2–10 times larger than those obtained with Fe^IV(O)(TMP^•+).²¹¹

3.1.3. Correlation of Thermodynamic Driving Force with Rate Constants of HAT

The thermodynamic driving force of HAT via cpd I is the net free-energy change resulting from the C–H bond that is broken and the O–H bond that is formed at the ferryl oxygen. In many C–H activation reactions mediated by iron–oxo intermediates, an empirical Brønsted-Evans–Polanyi (BEP) correlation has been observed,²⁹² in which the free-energy barriers (ΔG^‡), and thus the rates of hydrogen atom transfer (HAT) are negatively correlated with bond dissociation energies (BDE) of the scissile C–H bonds (eq 3, Scheme 3a).^273,293 We use BDE here instead of bond dissociation free energies for simplicity because comparative data is more readily available and the differences are small. The slope α (Brønsted coefficient, 0 < α < 1) is interpreted to reflect the position of hydrogen along the reaction coordinate in the linear transition state [Fe–O---H---C]^‡, with α = 0 referring to a reactant-like (“early”) transition state with a strong C–H bond character and α = 1 for a product-like (“late”) transition state with a strong O–H bond feature.²⁹⁴ While being broadly applied to a variety of HAT processes, the BEP relationship is less successful at correlating reactants with disparate molecular features. For instance, phenol O–H bonds have similar BDEs to benzylic C–H bonds but show much faster rates for HAT to iron–oxo intermediates.²⁶¹ Such limitations of the BEP equation have led Mayer et al. to apply the Marcus cross correlation to analyze HAT, in analogy to Marcus electron transfer theory.^295,296 Thus, for a HAT reaction, AH + B· → BH + A· (eq 4, Scheme 3b), the reaction rate for HAT (k_AH/B) is estimated from the equilibrium constant K_eq of the reaction and the two self-exchange rate constants (k_AH/A and k_BH/B) for the individual reactants A·/AH and B·/BH (eq 5 and 6). Since the rates of the self-exchange reactions can be experimentally determined in favorable cases (e.g., via NMR line broadening), Mayer’s method offers a powerful approach to predict kinetic properties of HAT from intrinsic properties of individual reactants.²⁷³ With this method, it was found that the self-exchange rate constants of phenols were ∼10⁸ larger than those of alkanes, which matched the observed difference in their HAT rate constants.²⁹⁵ The origin of the faster phenol rates is likely due to the fact that a phenol is a good hydrogen bond donor while an aliphatic C–H proton is not.

Scheme 3. (a) Brønsted–Evans–Polanyi (BEP) Correlation for HAT Reactions and (b) Cross Correlation of HAT Reactions.

BDEs (or BDFEs) are purely thermodynamic properties and should not necessarily mirror the structural and entropy effects that affect transition-state energies of HAT events. As recently summarized by Bietti et al. for aliphatic C–H activation by alkoxyl radicals, structural effects (such as polar effects, stereoelectronic effects, and strain-release) and medium effects can have very large impacts on the reactivity of C–H bonds.²⁹⁷ Thus, the thermodynamically weakest C–H bond is not necessarily the most reactive one. This phenomenon has also been observed in C–H activation reactions of heme proteins and metalloporphyrins. A good example is the deactivation of α C–H bonds near electron-withdrawing groups such as ketones, esters, and nitriles in metalloporphyrin-catalyzed C–H hydroxylations and halogenations,²⁹⁸ although these bonds have lower BDEs than other reactants. This is a typical polar effect. Another factor that can affect rates of HAT is the entropic contribution to the free-energy barriers. Entropy effects would be most evident for C–H activation of substrates with weak C–H BDEs. These reactions will have early transition states with the C–H bonds elongated to a minimal extent. Thus, activation enthalpies (ΔH^‡) are insensitive to C–H BDEs in this scenario, and overall free-energy barriers will be determined by the entropy term TΔS^‡. Tanko et al. have measured rate constants of C–H abstraction by ^tBuO·, for which the O–H bond formed during HAT has a BDE of 104 kcal/mol.²⁹⁹ A good BEP correlation was found for C–H BDEs greater than 92 kcal/mol, whereas for C–H BDEs below 92 kcal/mol, a flat log k_H versus BDE plot was observed. Further analysis showed similar ΔH^‡ values (∼2 kcal/mol) for the flat region, and the entropy contribution outweighs that of enthalpy for most hydrogen abstractions by ^tBuO·.

We have found a similar nonlinear BEP correlation for C–H hydroxylation by the fungal heme-thiolate peroxygenase AaeAPO compound I (AaeAPO-I) (Figure 30).²⁰⁰ A Brønsted coefficient of about 0.4 was determined for log k_H versus BDE for substrates with C–H bonds stronger than 90 kcal/mol, indicating a nearly symmetrical [Fe–O---H---C]^‡ transition state in this region. But for BDE < 90 kcal/mol, the rate constant was not sensitive to C–H BDEs. Since the observed rate constants in the flat region were shown to be slower than both the active site substrate exchange and the diffusion limit, HAT by AaeAPO-I for C–H BDEs less than 90 kcal/mol is entropy-controlled and is more dependent on factors such as orientation, accessibility, and trajectory toward the transition state. These results are consistent with a Fe^IVO–H BDE in AaeAPO-II of ∼100 kcal/mol. The entropy effect was also found to affect the chemoselectivity of oxygenation reactions catalyzed by metalloporphyrins. Fujii et al. determined the activation parameters for epoxidation and allylic hydroxylation for cyclohexene oxidation mediated by Fe^IV(O)(TMP^•+)Cl. By varying temperatures between 193 and 233 K, they determined the enthalpy and entropy of activation (ΔH^‡ and ΔS^‡) to be 8.0 ± 0.4 kcal/mol, −24.0 ± 1.6 cal mol^–1 K^–1 and 4.6 ± 0.6 kcal/mol and −40.3 ± 2.1 cal mol^–1 K^–1 for epoxidation and allylic hydroxylation, respectively.³⁰⁰ On the basis of these values, the contribution of the entropy term −TΔS^‡ would exceed that of enthalpy term in the free energy of activation (ΔG^‡) for allylic oxidation at temperatures above 114 K, while the entropy term provides the major contribution to ΔG^‡ for epoxidation at temperatures above 334 K.

Figure 30.

Nonlinear BEP correlation for C–H hydroxylation by AaeAPO-I. Adapted from ref (200). Copyright 2012 American Chemical Society.

3.1.4. Two-State and Multistate Reactivity

Compared to simple organic radicals such as ^tBuO·, a distinct feature of paramagnetic transition metal complexes is the existence of multiple spin states that can be close in energy. Thus, a given reaction may occur along different spin surfaces, and a higher-energy spin state could have the lowest-energy transition state. Such a situation can lead to spin crossover along the reaction coordinate from reactants to products. This concept was first proposed by Shaik, Schröder, and Schwarz and termed “two-state reactivity” (TSR) or “multistate reactivity” (MSR) (Figure 31a).^301−303 HAT by compound I was one of the first systems to be studied for TSR in transition metal catalysis (Figure 31b). DFT calculations of cpd I of iron porphine revealed that the quartet and doublet spin states, wherein the porphyrin cation radical is either ferromagnetically or antiferromagnetically coupled to the S = 1 ferryl, were very close in energy and had comparable activation barriers for C–H activation, suggesting important contributions of both spin states for HAT.³⁰⁴ QM/MM studies also showed similar mechanistic features for HAT by P450 compound I.^305,306

Figure 31.

(a) Schematic illustration of two-state reactivity (TSR). (b) Two-state reactivity for C–H activation via an iron porphine catalyst (energies in kcal/mol, obtained from ref (305)).

C–H activation mediated by oxoMn(V) porphyrins represents another example of potential TSR in metalloporphyrin-mediated reactions (Figure 32).^226,227 DFT calculations showed that the singlet ground state of oxoMn(V),²⁴³ which is known from the experiment,^227,233,243 had a very high energy barrier for C–H activation.^244,307,308 The triplet spin state, which is calculated to be only 3.4 kcal/mol higher than the singlet ground state, had the lowest-energy transition state (5.3 kcal/mol relative to the singlet ground state). The energy of the quintet state is much higher (11.1 kcal/mol) than the singlet state and is likely not accessible, despite the low activation energy of this spin state (2.2 kcal/mol). Therefore, during C–H activation, the unreactive singlet oxoMn(V) would undergo spin-crossing to the triplet energy surface.

Figure 32.

Energy profile for C–H activation mediated by an oxoMn(V) porphyrin complex.

Shaik recently introduced the concept of exchange-enhanced reactivity (EER) as a general principle to evaluate the preferred spin state for a given reaction state especially catalyzed by 3d metals.^309,310 The main idea is that transition states with an increased number of unpaired spin-identical electrons at metal center would be stabilized by enhanced electron-exchange. Such an exchange can be assessed by multiplying the exchange interaction of two unpaired electrons (K_dd) with the number of such interactions (n). The gained exchange-energy (ΔK_dd) versus the orbital promotion energy (ΔE_orb) needed for promoting an electron into a higher-lying orbital would determine the preferred spin state. This concept can be utilized to analyze the preference for the triplet state in oxoMn(V)-mediated C–H activation in Figure 32.³¹¹ The transition state on the S = 0 surface is an open shell singlet with no spin-exchange interactions, while for the triplet surface, three exchange interactions are present. The overall stabilization effect in the triplet state can be estimated by 3K_dd – ΔE_orb(d_xy → d_xz/yz), in which ΔE_orb(d_xy → d_xz/yz) is the energy needed for orbital excitation from d_xy to d_xz/yz. Although these energies have not been determined for manganese porphyrins, the K_dd of 11.1 kcal/mol and ΔE_orb(d_xy → d_xz/yz) of 26.4 kcal/mol have been estimated for an analogous oxoMn(V) corrolazine species, indicating a net stabilization of ∼6.9 kcal/mol. This stabilization energy is comparable to the calculated energy barrier difference between singlet and triplet pathways for this species.³¹¹

The kinetic isotope effect (KIE) for hydrogen/deuterium HAT has been proposed to be a good mechanistic probe to probe the reactive spin state of a reaction. Theoretical calculations have shown that different spin states can display very different KIEs.^312−314 For instance, in the demethylation reaction of para-substituted N,N-dimethylaniline (p-Y-DMA), a positive linear correlation was found between experimental KIE values (KIE_exp) and the Hammett substituent parameter of para-substituent Y.^315−317 The calculation showed that only the low-spin doublet state showed this trend, while the calculated KIEs of high-spin quartet state (KIE_HS) were much larger than the experimental values and were not sensitive to the para-substituent.³¹⁸ This result suggested that the low spin state is the primary state for cpd I reactivity of N-demethylation. A further analysis of KIE_exp using a blend of KIE_LS and KIE_HS showed that for Y = H and Cl, the reaction occurred exclusively on the LS surface, while for Y = CN and NO₂, the reaction occurred through both states with 70/30% and 50/50%, respectively.

The nature of the axial ligand can also impact the relative energies of different spin states and the spin-state crossing. For instance, the HAT activity of oxoMn^V porphyrins with axial ligand X (X = H₂O, OH^–, and O^2–) has shown to increase in the following order: H₂O > OH^– > O^2–.^227,233 DFT calculations indicated that this trend correlated with corresponding energy gaps between unreactive singlet ground state and higher-energy triplet and quintet states.²⁴⁴ In a related study of less reactive oxoMn^V corrolazine (Cz) complexes, de Visser and Goldberg showed that fluoride and cyanide axial ligands increased the rate of HAT by 10³–10⁵-fold compared to that with no axial ligand.³¹⁹ Computational studies showed 12 and 17 kcal/mol increases in the O–H BDE of Mn(IV)–OH for fluoride and cyanide axial ligands, respectively. The origin of the rate-enhancement exerted by axial ligands is a matter of debate. A computational study by Shaik indicated that fluoride and cyanide also stabilized the triplet transition state by 14–16 kcal/mol, due to a reduction in the σ_xy → π*_xz/yz excitation energy and an increase in exchange coupling of unpaired d electrons in the triplet transition state.³¹¹ However, DFT and CASSCF calculations by the Goldberg and de Visser groups on Mn^V(O)(Cz)-mediated sulfoxidation of thioanisole indicated small spin–orbit coupling constants, suggesting a singlet-to-triplet spin-state crossing before the rate-limiting step is unlikely.³²⁰ In support of this statement, only the singlet spin state can computationally reproduce a V-shaped pattern observed in a Hammett analysis with para-substituted thioanisoles, in which a negative slope was obtained for electro-donating para-substituents but a positive slope was observed for electron-withdrawing substituents.³²⁰

In addition to increasing the reduction potential of cpd I, the porphyrin peripheral substituents can also affect cpd I reactivity via regulating the spin state crossover. Groves et al. have studied the influence of two cationic meso-substituents 2-N-methylpyridyl (2-TMPyP) and 4-N-methylpyridyl (4-TMPyP) on oxygen transfer reactivity of oxoMn(V) porphyrins.²²⁷ Surprisingly, the less electron-deficient [Mn^V(O)(4-TMPyP)], with the N-methyl pyridium ions farther from the porphyrin ring, was 10–10³ times more reactive than Mn^V(O)(2-TMPyP). Computational studies showed that the 4-pyridyl isomer [Mn^V(O)(4-TMPyP)] had a smaller low-spin-high-spin promotion energy and was therefore kinetically favored in a two-state energy profile involving spin state crossing from the starting singlet oxoMn(V) species.²⁴³ Furthermore, the meso-substituents were also shown to affect the acid–base equilibria involving the axial water ligand. As discussed in section 2.4.3, Mn^V(O)(4-TMPyP) and Mn^V(O)(2-TMPyP) have three different prototropic forms in water depending on the protonation state of the axial ligand, oxo-aqua, oxo–hydroxo, and trans-dioxo.²³² The reactivity of these three states increased in the following order: trans-dioxo < oxo–hydroxo < oxo-aqua. The less electron-deficient Mn^V(O)(4-TMPyP) exhibited higher basicity for the axial aqua and hydroxo ligands (Figure 33). Thus, at the same pH, the kinetically more reactive protonated species are more accessible for Mn^V(O)(4-TMPyP), leading to the observed higher reactivity.²²⁷

Figure 33.

pK_a of the hydroxo ligand of [Mn^V(O)(OH)(4-TMPyP)] and [Mn^V(O)(OH)(2-TMPyP)].

3.1.5. Quantum Tunneling in Hydrogen Atom Transfer

Hydrogen is a small particle and as such can show quantum tunneling effects.³²¹ As a result, hydrogen can “tunnel” through the transition state barrier and thus increase the rate of HAT (Figure 34a). This energy reduction (ΔΔE^‡_tunneling) associated with proton tunneling is commonly between 1–4 kcal/mol.³²² The kinetic hydrogen isotope effect is a useful tool to probe the tunneling phenomenon in HAT. There are three criteria proposed first by Kreevoy and co-workers with respect to organic radicals that are widely used for assessing a tunneling process: (1) a KIE significantly larger than 6.4 at 20 °C; (2) an activation energy difference between deuterium transfer and hydrogen transfer (E_a(D) – E_a(H)) larger than >5.0 kJ/mol (1.2 kcal/mol); and (3) a ratio of pre-exponential factors A_H/A_D < 0.7.^323,324

Figure 34.

(a) Quantum tunneling (wavy arrow) leads to an effective energy barrier (ΔE^‡_eff) 1–4 kcal/mol below TS. (b) Reactivity pattern for [Fe^IV(O)(TMC)(L)]^z+ in HAT from alkanes (red line) and oxygen-transfer to phosphines (back line). Figure is adapted from ref (328). Copyright 2015 American Chemical Society.

Tunneling effects have been found in HAT processes mediated by a variety of metalloenzymes and related metal–oxygen model compounds.^325−328 Many of these systems are nonheme iron-containing enzymes and/or their model compounds.^325−329 In contrast, there are few examples of tunneling effects in cytochrome P450s and metalloporphyrins. Although KIEs (10–32) larger than usually adopted theoretical limit (∼7) for nontunneling reactions have been reported for several P450-catalyzed reactions,^330−334 the extent of hydrogen tunneling in these reactions is not conclusive because of the difficulties in determining the activation energies and pre-exponential factors of these systems. In P450 model compounds, an early example of a possible tunneling event was reported by Sorokin and Khenkin.³³⁵ They determined a KIE value of ∼22 (20 °C) for cyclohexane oxidation catalyzed by Fe(TMP)Cl/NaOCl in benzene. They further determined a [E_a(D) – E_a(H)] of 4.5 ± 0.5 kcal/mol and A(H)/A(D) = 0.01, indicating the presence of a tunneling effect. Similar effects have also been observed by Meunier et al. in a study of intramolecular KIEs of adamantane-1,3-d₂ hydroxylation catalyzed by a variety of metalloporphyrins.³³³ For an Fe(TMP)Cl/NaOCl system, they observed a KIE of 9 (10 °C), [E_a(D) – E_a(H)] of 1.83 kcal/mol, and A(H)/A(D) of 0.35. Recently, large KIE values indicative of tunneling effects were also reported by Newcomb et al. for benzyl alcohol oxidation mediated by Fe^IV(O)(TMP^•+)(ClO₄^–)³³⁶ and by Fujii et al. for C–H(D) oxidations of xanthene and tetralin promoted by Fe^IV(O)(TMP^•+)(L) (L = NO₃^–, Cl^–, imidazole).³³⁷ It is worth noting that the KIE and tunneling effects are very dependent on the position of transition state along the reaction coordinate, which will be dramatically affected by the properties of ligands, metal centers, solvents, and substrates.^326,327,338 For instance, while most studies of tunneling effect have been conducted on iron porphyrins with electron-donating tetramesityl porphyrin ligand, manganese porphyrins and electron-withdrawing iron porphyrins like Fe(TDCPP)Cl exhibited much smaller KIEs (3–7) and catalyzed the reactions through classical HAT pathways.^{333,335−337}

H-tunneling could potentially contribute to the reactivity difference between hydrogen atom transfer (HAT) and oxygen atom transfer (OAT), as OAT reactions will not be affected by H-tunneling effect. For example, in a series of nonheme iron(IV)–oxo tetramethylcyclam (TMC) complexes with different axial ligands L ([Fe^IV(O)(TMC)(L)]^z+ (z = 1, 2)), the reactivity for HAT increased with the electron-donating ability of L, while oxygen atom transfer (OAT) showed an opposite trend (Figure 34b).^339−341 This reactivity difference was attributed to a combined effect of spin inversion probability and quantum tunneling.^328,342 In addition to decreasing the triplet/quintet energy gap and facilitating spin crossover, the most electron-donating thiolate axial ligand (compound 63, Figure 34b) also decreased the energy barrier by 3.4 kcal/mol via a quantum tunneling effect, in comparison to the 1.4 kcal/mol obtained with the CH₃CN axial ligand. Such a quantum tunneling effect was absent in oxygen atom transfer reactions and is the main attribute of the reactivity difference between HAT and OAT. The reactivity difference between HAT and OAT reactions were also explored for cyclohexene oxidation mediated by Fe^IV(O)(TMP^•+)L. It was found that ratio of epoxidation/allylic oxidation products decreased dramatically with decreasing temperatures.³⁴³ This trend is further studied by Fukuzumi and Nam et al.³⁴⁴ They found that while epoxidation was preferred at higher temperatures (213–233 K), the allylic hydroxylation took over at lower temperatures (193–213 K). This switchover was accompanied by a drastic increase in KIEs for allylic hydroxylation (1.4 at 233 K vs 13 at 193 K). The large difference in activation energy [E_a(D) – E_a(H) = 5.0 kcal/mol] and a small ratio of exponential factors (A_H/A_D ∼ 4 × 10^–5) suggest a significant contribution from tunneling for the allylic hydroxylation pathway.

The enhancement of HAT reactivity by H-tunneling is also manifested in an unusually reactive hydroxoiron(III) porphyrazine complex 65 (Figure 35) reported very recently by Gao and Groves.³⁴⁵ This hydroxoferric species abstracts benzylic C–H bonds with rate constants of 2.22 × 10³ M^–1 s^–1 for xanthene and 81 M^–1 s^–1 for dihydroanthracene. These rate constants are 5–6 orders of magnitude faster than other typical Fe^III–OH complexes and more reactive than many ferryl complexes (Figure 35a).³⁴⁵65 displayed an unusually high reduction potential of 0.84 V versus NHE and a very small activation enthalpy of 6.1 ± 0.3 kcal/mol for C–H scission. A large tunneling component was inferred from the observed deuterium isotope parameters (Figure 35b),^327,346,347 KIE of 20.2, E_a(D) – E_a(H) = 2.5 ± 0.4 kcal/mol, and A(H)/A(D) = 0.28 ± 0.14 for xanthene oxidation, contributing to the unusual activity of 65.

Figure 35.

(a) Correlation of C–H HAT rate constants with O–H BDE for a variety of metal–oxo and metal–hydroxo oxidants at 298 K. The black dashed line passes through permanganate, t-butylperoxyl, and t-butoxyl radicals for reference. The red ■ is (PyPz)Fe^III–OH. Reported hydroxometal(III) complexes (green ▲), Fe(IV)=O complexes (blue ▲) and nitroxyl radicals HBTO and PINO (magenta ◆) are also shown for comparison. (b) A highly reactive hydroxo-iron(III) porphyrazine complex 65 for HAT. Adapted from ref (345). Copyright 2017 American Chemical Society.

3.2. Intermediacy of Substrate Radicals–Oxygen Rebound

In P450-catalyzed hydroxylations, hydrogen atom transfer (HAT) has been shown to generate substrate radicals with short lifetimes. The results can be best understood as leading to an active site caged substrate radical [Fe(IV)–OH ·R] that subsequently rebounds to the Fe^IV–OH intermediate to afford hydroxylation products. The intermediacy of substrate radicals in such processes was first implicated in model reactions related to Fenton’s reagent, a mixture of Fe²⁺ and H₂O₂. The use of cyclohexanol (66) and trans,trans-3,5-dideuterocyclohexanol (67) as diagnostic substrates clearly indicated a radical epimerization process after HAT (Scheme 4a).¹⁸⁸ Later, similar radical epimerization or rearrangement processes were found in the hydroxylation of exo,exo,exo,exo-2,3,5,6-tetradeuteronorbornane (68) and selectively deuterated cyclohexene (69) by phenobarbital-induced P450_LM2 (CYP2B4) as well as reactions mediated by synthetic iron porphyrins (Scheme 4b).^189,348 These results clearly indicated a stepwise, nonconcerted mechanism for C–H hydroxylation mediated by cpd I.

Scheme 4. (a) C–H Hydroxylation of Cyclohexanol and Dideuterocyclohexanol Revealed a Step-Wise Radical Mechanism for C–H Hydroxylation Mediated by Fe²⁺/H₂O₂ and (b) Radical Epimerization and Rearrangement Were Also Observed for Hydroxylation Catalyzed by P450s and Model Compounds.

Additional insights regarding this stepwise radical C–H cleavage event were obtained using mechanistically diagnostic substrates and radical clock probes, which form radicals that undergo characteristic stereochemical/structural changes at known rates.³⁴⁹ A highly informative diagnostic clock is the simple hydrocarbon norcarane, which will afford different rearranged products for pathways involving radical and cationic intermediates (Scheme 5).³⁵⁰ Thus, the 2-norcaranyl radical (72 in Scheme 5) will ring-open to a kinetically favored homoallylic radical (73), whereas formation of the thermodynamically more stable cyclohepten-4-yl cation (75) is the favored rearrangement pathway for the 2-norcaranyl cation (74). Other small-ring systems, such as bicyclo[2.1.0]pentane and a variety of substituted cyclopropanes are also common radical clocks for the study of HAT in P450s and model systems.¹⁶ Radical clocks provide a facile way to measure the average lifetime of radical intermediates from the ratio of unrearranged products (U) to rearranged products (R) (Scheme 5). It is now known that P450s and synthetic iron porphyrins generally have very short radical lifetimes, tens to hundreds of picoseconds, with corresponding rebound rate constants between 10¹⁰ and 10¹¹ s^–1.¹⁶ A similar time frame is also measured for hydroxylation catalyzed by several nonheme systems such as soluble methane monooxygenase (sMMO) and toluene 4-monooxygenase (T4moH).^351−356 Recently, with norcarane radical clock, the rebound rate of AaeUPO was determined to be 1.06 × 10¹¹ s^–1 (radical lifetime: 9.4 ps), which is much faster than the values of other functional P450s.¹⁹⁹ Notably, substrate radical lifetimes in the picosecond time regime provide time for many molecular vibrations and rotations in an enzyme active site or solution solvent cage. However, there would not be time for such a short-lived radical to diffuse away from the active site.

Scheme 5. Radical and Cation Rearrangement of Norcarane and Radical Lifetimes of Representative P450s and Nonheme Oxygenases As Determined by the Measured Ratios of Rearranged (R) and Unrearranged (U) Products.

Direct study of the radical rebound step is difficult because of the fast recombination rate between substrate radicals and Fe^IV–OH species (10¹⁰–10¹¹ s^–1) and the difficulty of preparing discrete Fe^IV–OH intermediates. Very recently, Green and Goldberg et al. have successfully prepared a stable monomeric Fe^IV–OH complex with a sterically hindered tris(2,4,6-triphenyl)phenyl corrole (ttppc) ligand.³⁵⁷ This complex (denoted as [Fe^IV(ttppc)(OH)]) exhibits a Fe–O bond distance of 1.857 Å, a resonance Raman vibration ν(Fe–O) of 555 cm^–1, a small Mössbauer isomer shift (0.13 mm/s) and a large quadrupole splitting (2.21 mm/s). These features are characteristic of a Fe^IV–OH species. Treating [Fe^IV(ttppc)(OH)] with stable radicals like the trityl radical (Ph₃C·) led to the formation of the hydroxylation product (Ph₃COH) and Fe^III(ttppc) (Scheme 6). Kinetic measurements of this radical recombination step with a series of para-substituted trityl radicals afforded second-order rate constants (k), ranging from 12.6 to 357 M^–1 s^–1. A Hammett analysis with σ⁺_para parameter showed a small slope (ρ = −0.55). A Marcus cross-relation between (RT/F) ln(k) and redox potentials of substituted trityl radicals (E_ox[(p-X-C₆H₄)₃C^+/•]) also yielded a small slope (−0.15). These results clearly indicate that, despite the low oxidation potential of trityl radicals, their recombination with [Fe^IV(ttppc)(OH)] follows a concerted pathway with small charge separation, as opposed to a stepwise mechanism in which the carbon-centered radical is first oxidized to cations and then recombines with hydroxide.

Scheme 6. Radical Recombination between Trityl Radicals and [Fe^IV(ttppc)(OH)]. Adapted from ref (357). Copyright 2017 American Chemical Society.

After HAT, substrate radicals R· in the resulting [Fe^IV–OH ·R] cage are not static. Factors such as tightness of the cage, ease of solvent reorganization, the rate of radical rebound, and the nature of the substrate radical can all dramatically affect the behavior of R·.¹³ An instructive manifestation of the interplay between these factors is the C–H hydroxylations catalyzed by nonheme diiron hydroxylase AlkB, in which the use of three clock substrates with very different rearrangement rates (k_r) showed similar R/U ratios and led to drastically different apparent radical lifetimes, all in the low nanosecond time regime (1–200 ns) (Scheme 7).^358,359 The cause of this enigmatic behavior is a competition between the relatively slow radical rebound and diffusive radical cage escape to generate a solvent-separated radical pair (RP_ss). If the rate of radical cage escape (k_e) is comparable to that of radical rebound k_R and larger than the radical rearrangement rates k_rearr, then R/U ratios will be insensitive to the differences in k_rearr but only reflect the competition between cage escape and radical rebound k_e/k_R. Significantly, measured radical lifetimes for AlkB hydroxylases were found to be in the nanosecond time regime, much longer than those obtained for heme proteins such as P450 and APO. An embodiment of the cage escape effect in AlkB system is a decrease in radical lifetimes by increasing the substrate concentration. Higher substrate concentration would lead to more substrate molecules present in the long and hydrophobic substrate channel in AlkB,³⁶⁰ which could displace and reorganize the water in the active site and create a crowded environment that reduces the radical cage escape.

Scheme 7. Oxidation of Radical Clocks with Different Rearrangement Rate Constants.

The similar ratios of rearranged/unrearranged products (R/U) result from the competition between cage escape and rebound of substrate radicals. Adapted with permission from ref (358). Copyright 2008 Wiley-VCH.

In addition to the environment of the radical cage, the nature of the metal center can also dramatically affect the radical lifetime. For instance, manganese porphyrin-catalyzed hydroxylation exhibited long radical lifetime (ns), whereas no rearranged product was observed for ruthenium porphyrin-catalyzed reactions.^220,361 The former might relate to a lower reduction potential of Mn^IV–OH compared to Fe^IV–OH, and the latter might be caused by an unusually high electrophilic nature of Ru^V=O.³⁶² Like HAT, radical rebound can also occur along different spin surfaces and are subject to spin crossover events (Figure 36). DFT calculations for methane hydroxylation by a simplified heme cpd I showed that the rebound barrier on the quartet transition state is much higher than that on the doublet spin state due to the difference in frontier orbital interactions.³⁰⁴ In the quartet state, the unpaired electron of the radical goes into a higher-energy 3d_z² orbital, whereas in the doublet state, the electron transfers into a lower-energy singly occupied d_xz (or d_yz) orbital.

Figure 36.

Oxygen rebound of different spin states (energies in kcal/mol, obtained from ref (305)).

Groves et al. have found a significant impact of the donor properties of the axial ligands on rates of radical rebound. Manganese porphyrins with anionic electron-donating ligands such as OH^– and F^– were found to afford mainly radical rearranged products in hydroxylation of norcarane, whereas radical unrearranged products prevailed in reactions catalyzed by Mn porphyrins with neutral axial ligands such as pyridine and CH₃CN (Scheme 8).^298,363 This effect is reminiscent of axial ligand effect on HAT. Electron-donating ligands would diminish the reduction potential of Mn^IV–OH and destabilize frontier orbitals (3d_z² and d_xz/d_yz), leading to a higher rebound barrier. The deceleration of rebound would facilitate the escaping of substrate radicals from the radical cage. Indeed, a large amount of radical dimerization products were observed in xanthene hydroxylation by manganese porphyrins in the presence of F^– with yield dependent on the concentration of F^–.

Scheme 8. Effect of Axial Ligands on Radical Rebound to Mn^IV–OH Porphyrins.

The rebound step in the hydroxylation of deuterated ethylbenzene catalyzed by iron and manganese porphyrins has been explored recently.³⁶³ The results showed that more electron-deficient porphyrins displayed less stereochemical inversion at the benzylic hydroxylation site (Figure 37). The extent of stereochemical inversion was found to be sensitive to the metal center, with only 8–20% retention of configuration for Cl–Mn^V(O)(Por), while Cl–Fe^IV(O)(Por^•+) showed 80–90% retention. Significantly, F–Fe^IV(O)(Por^•+) afforded only 60–70% retention, showing that the donor properties of the ligand can also affect iron porphyrin stereochemistry. These results are also consistent with the general idea that oxygen rebound is faster for stronger oxidants and more able to compete with radical cage escape. The radical cage escape event was recently explored in nonheme systems by Nam and Shaik.³⁶⁴ It was determined that such processes occurred in a number of nonheme metal–oxo systems such as Mn^IVO, Fe^IVO, Cr^IVO, Fe^VO, and Ru^IVO. DFT calculations revealed small dissociation energies and relatively large rebound barriers for these systems.

Figure 37.

Stereochemistry of the rebound process for iron and manganese porphyrins. Adapted from ref (363). Copyright 2017 American Chemical Society.

3.3. Redirecting Substrate Radicals to Reaction Pathways Other than Oxygen Rebound

3.3.1. Decarboxylation Mediated by P450 OleT_JE

P450 OleT_JE, found in Jeotgalicoccus sp., converts C_n fatty acids to C_n–1 terminal alkenes through a desaturative decarboxylation (eq 9, Figure 38).³⁶⁵ This enzyme has attracted considerable interest for its potential applications in biofuel production and fine chemical synthesis.^366−369 P450 OleT_JE belongs to the P450 peroxygenase family CYP152. Other notable members in this family are P450_SPα (CYP152B1) and P450_BSβ (CYP152A1),^135,140,370 which have 36% and 40% amino acid identity to P450 OleT_JE, respectively. All three enzymes utilize an active-site arginine and the carboxylate group provided by the substrate to effect the heterolytic cleavage of H₂O₂ to generate compound I. Despite the similarity in H₂O₂ activation and active site structures to P450 OleT_JE, P450_SPα and P450_BSβ catalyze hydroxylations of fatty acid substrates rather than decarboxylation (Figure 38).^{135,136,277,370} P450_SPα hydroxylates only at the C_α position (eq 10), whereas P450_BSβ hydroxylates both the C_α and C_β positions of fatty acid substrates (eq 11).

Figure 38.

Native reactions catalyzed by P450 OleT_JE (PDB: 4L40), P450_SPα (PDB: 3AWM), and P450_BSβ (PDB: 1IZO). All three enzymes have a conserved arginine residue that can form a salt bridge with the carboxylate group of fatty acid substrates.

The decarboxylation reaction catalyzed by P450 OleT_JE is initiated by a hydrogen atom transfer (HAT) from the C_β position, as indicated by the large primary kinetic isotope effect (KIE ≥ 8.1) for fully deuterated fatty acid substrates.³⁷¹ This finding indicates that the oxygen rebound that normally occurs after HAT in P450 catalysis is inhibited in P450 OleT_JE. The substrate radical is thus redirected to decarboxylation presumably via a single electron transfer to cpd II and a subsequent C–C cleavage with concomitant loss of CO₂ (Figure 39). It is worth noting that a similar C_β-H scission step has also been identified as the key step for the decarboxylative desaturation catalyzed by UndA, a nonheme iron oxidase that can synthesize medium-chain 1-alkenes from fatty acids.^372,373 Very recently, Makris and co-workers observed OleT compound I and compound II using stopped-flow spectroscopic techniques by employing deuterated substrates to slow the initial HAT step.³⁷⁴ Intriguingly, OleT compound II was remarkably stable, having a lifetime of hundreds of milliseconds after its preparation from compound I via substrate C–H abstraction. The unusual stability of OleT-II supports the proposal that this enzyme imposes an inhibition of oxygen rebound, which would normally occur with rate constants in the range of 10¹⁰–10¹¹ s^–1 for heme-thiolate proteins. What causes this remarkable slow-down of the radical rebound process is still not entirely clear. But the fact that the amount of hydroxylated side products varied substantially with fatty acid chain length suggests that the binding and the positioning of substrates play a crucial role in affecting the reaction pathway.^369,375 Evidence in favor of rebound inhibition by substrate–protein interactions has recently been provided by the observation that the radical clock probe norcarane is a substrate for OleT_JE.³⁷⁶ Further, the radical lifetime for the norcaranyl radical intermediate was typical for other P450 proteins, in the picosecond time regime. Apparently, the radical intermediate of a small hydrocarbon substrate cannot be restrained by the active site architecture and rebound proceeds normally.

Figure 39.

Proposed mechanism for the oxidative decarboxylation catalyzed by OleT_JE.

Li et al. have recently examined several key active site residues of OleT_JE and found that Arg254, His85, and Ile170 are all required for the decarboxylation activity (Figure 40).³⁷⁷ The roles of these residues are thought to be the formation of a salt-bridge with the substrate carboxyl group (Arg245), acting as a proton donor (His85), and regulating substrate positioning in the active site (Ile170). As discussed in section 2.3.1, the active site structure of the CYP152 family of enzymes is similar to that of AaeUPO in that both proteins have an arginine in the distal pocket that forms a salt bridge with a nearby carboxylate that facilitates hydrogen peroxide binding and O–O bond scission. In AaeUPO, the carboxyl group is provided by a glutamate residue, while in CYP152 the carboxyl group is from the fatty acid substrate. Makris et al. recently found that a mutation of Pro246 to Asp in OleT circumvents the requirement for fatty acid substrates for H₂O₂ activation.³⁶⁷ The resulting OleT P246D variant is not active toward native fatty acids but showed high activity for a broad range of hydrocarbon substrates.

Figure 40.

Key active site residues of OleT_JE (PDB: 4L40).

3.3.2. P450-Catalyzed Desaturation Reactions

In addition to obtaining olefins via decarboxylative elimination, P450 has also shown capabilities for forming C=C bonds via dehydrogenation. An early example is the formation of Δ⁴-valproic acid (Δ⁴-VPA) during the metabolism of the antiepileptic drug valproic acid (VPA) (Figure 41a).³⁷⁸ Other representative substrates that undergo P450-catalyzed desaturation include warfarin, ethyl carbamate, and retinol.^279,379 The fact that purified P450 catalyzed both desaturation and hydroxylation in most cases and the alcohol and desaturation products did not interconvert indicates a bifurcation of the usual rebound mechanism. Furthermore, almost identical values for the primary kinetic isotope effects were determined for hydroxylation and desaturation products of VPA.³⁸⁰ On the basis of these results, it is proposed that the substrate radical resulting from the initial hydrogen-atom transfer (HAT) by cpd I will undergo a second HAT or proton-coupled electron transfer process (PCET) by compound II at C–H bonds vicinal to the carbon-centered radical, affording the desaturation product and H₂O (Figure 41a). This mechanism is reminiscent of the desaturation pathway of fatty acid desaturases, in which a bis-μ-oxo diiron(IV) cluster mediates two sequential HAT to form C=C bonds in long-chain fatty acyl derivatives (Figure 41b).^381,382 A recent mechanistic study of the histidine-rich nonheme diiron enzyme AlkB using diagnostic substrates norcarane and deuterated norcarane revealed H atom abstraction occurred at both C2 and C3 (Figure 41b).³⁵⁹ While the C2 radical leads to the formation of 2-norcaranol and the radical-rearranged homoallylic alcohol product, desaturation and hydroxylation products were obtained from the C3 radical with similarly large KIEs (∼20). From these results, a bifurcation of the reaction pathways leading from the Fe–OH/R· intermediate appears to be a common feature for both heme- and nonheme-catalyzed desaturation reactions.

Figure 41.

(a) P450-catalyzed hydroxylation and desaturation of valproic acid (VPA) and the proposed mechanism. (b) Proposed mechanism of fatty acid desaturases and the hydroxylation and desaturation of norcarane catalyzed by nonheme diiron hydroxylase AlkB.

The factors that control the hydroxylation/desaturation pathways have been studied via computational methods.^383,384 In general, the desaturation pathway was found to be less exothermic than the hydroxylation pathway. The energy barriers of two reaction channels varied depending on the spin state and electronic state of the Fe^IVOH intermediate, the properties of the substrate radicals, and the active site environment. For instance, comparable energy barriers were calculated for desaturation and oxygen rebound on low-spin doublet surface, whereas the desaturation step is more unfavorable along quartet surfaces.^383,384 Shaik also suggested that Fe^IV–OH species can also exist as a Fe^III–OH porphyrin cation radical electromer, whose energy barrier for the desaturation pathway is 8–20 kcal/mol higher than that of oxygen rebound.³⁸⁴ The desaturation pathway also requires the approach of Fe^IV–OH to C–H bonds vicinal to the substrate radical. If this step is restricted by the substrate/enzyme interaction then the desaturation pathway will be inhibited. In the case of camphor, gas-phase calculations showed similar barriers for desaturation and oxygen rebound on both doublet and quartet surfaces, however, the desaturation pathway was over 15 kcal/mol higher than oxygen rebound in the enzyme active site. An energy scan along the reaction coordinate revealed that the tight binding of camphor in the active site disfavored the reactive conformation that leads to the desaturation product.³⁸⁴

3.3.3. Oxidative Rearrangement Catalyzed by PntM (CYP161C2)

CYP161C2 catalyzes the conversion of pentalenolactone F (77) to pentalenolactone (78), a sesquiterpenoid antibiotic synthesized in Streptomyces arenae (Figure 42a).³⁸⁵ This transformation features an oxidative cationic rearrangement, which is very unusual in P450-mediated transformations but similar to the cation rearrangement pathways typically encountered in terpene cycloases. The reaction involves an initial hydrogen atom abstraction of H_si at C1 followed by an electron-transfer oxidation of the substrate radical by compound II. The carbocation intermediate can then undergo 1,2-migration of a methyl group with a subsequent loss of a proton at C3 (Figure 42a). Cane et al. have studied the mechanism of PntM using X-ray crystallography, kinetic measurements, site-directed mutagenesis, and product analyses.³⁸⁶ The inhibition of the rebound step in PntM could arise from the active site residues that stabilize the carbon cation intermediate. Examination of substrate-bound crystal structure of PntM revealed three active-site residues (F232, M77, and M81) that are in close proximity to C1, C2, and C3 and are unique among all known P450s (Figure 42b). However, mutation of these sites still afforded the pentalenolactone 78 as the only product, albeit with lower activity. Another possible cause of the inhibition of the rebound step is steric hindrance at the carbon-centered radical. After hydrogen abstraction at C1–H_si, the subsequent rebound step can be retarded by methyl and vinylidene groups at the si face of C1, whereas the outer-sphere 1e^– transfer would not be affected by the steric bulkiness at C1. The hydroxylation of 1-deoxypentalenic acid (79, Figure 42), which is structurally analogous to pentalenolactone F, by CYP105D7 provides additional support for this hypothesis. CYP105D7 abstracts the hydrogen at the re side of C1 followed by a normal oxygen rebound process (Figure 42b).³⁸⁷ Compared to the si face, C1 re face has much smaller steric hindrance with only a hydrogen atom and a methyl group at neighboring carbons (Figure 42c).

Figure 42.

(a) Proposed mechanism for the oxidative rearrangement of pentalenolactone F catalyzed by PntM. (b) Active site structures of PntM with bound pentalenolactone F (PDB: 5L1O). (c) Steric environment of C1 si and re sides of pentalenolactone F and hydroxylation of 1-deoxypentalenic acid by CYP105D7.

3.3.4. Mn-Catalyzed C–H Halogenations

In the early studies of Mn porphyrin-catalyzed hydroxylations, chlorination side products were observed. The chlorine source in those side reactions was determined to be the chloride ligand on the Mn porphyrin catalysts.²²⁰ Radical clock studies indicated an unusually long radical lifetime for the chlorination side product (∼25 ns). A lifetime this long suggests that the substrate radical intermediate may be freely diffusing. A detailed mechanistic study later revealed that radical lifetimes in Mn porphyrin-catalyzed hydroxylation can be controlled by axial ligands, as noted above, in which the rebound can be slowed down by anionic electron-donating axial ligands.^298,388 Another important finding was that a Mn(IV)-X type of intermediate (X = halogens), such as the well-characterized trans-difluoromanganese(IV) porphyrins, can trap independently generated alkyl radicals to yield halogenation products in a process analogous to the oxygen rebound (Figure 43a).³⁸⁹ These two mechanistic findings enabled the development of a series of aliphatic C–H halogenation reactions catalyzed by manganese porphyrin or manganese salen complexes (Figure 43, panels b and c).^{298,390−394} The catalyst-dependent regioselectivity and large primary kinetic isotope effect (KIE > 5) unambiguously indicated the involvement of oxoMn(V) intermediate in the C–H activation step. The radical nature of the halogen transfer step was implied by a large amount of radical rearranged products observed with the probe substrate norcarane. Furthermore, moderate enantioselectivity (20%–70%) was observed when a chiral manganese salen catalyst was used (Scheme 9a). Kagan analysis has shown a linear relationship between enantiomer excess (ee) of products and ee of manganese salen chiral catalysts.³⁹⁴ These results indicate that halogen transfer proceeded through a catalyst-bound halogen intermediate, namely Mn(IV)-X.

Figure 43.

(a) Formation of fluorination products via trapping of in situ generated alkyl radicals by Mn^IV(TMP)F₂. (b) Concept of manganese-catalyzed C–H halogenation reactions via heteroatom rebound. (c) Mn-catalyzed C–H halogenation and pseudohalogenation reactions.

Scheme 9. (a) Enantioselective C–H Halogenation-Catalyzed by a Chiral Manganese Salen Catalyst and (b) Radical lifetimes of Mn-Catalyzed C–H Halogenation Reactions.

What determines the bifurcation of substrate radicals to halogen rebound instead of oxygen rebound in this case? Clearly the nature of metal center is important, as switching from manganese porphyrins to iron porphyrins afforded mainly oxygenation products.³⁶³ In addition, hydroxylation products prevailed if more electron-withdrawing manganese porphyrins such as manganese 5,10,15,20-tetrapentafluorophenylporphyrin (TPFPP) were used. These results are consistent with a deceleration of oxygen rebound, as both the iron center and electron-withdrawing porphyrins were known to cause a faster oxygen recombination. This analysis is consistent with radical clock analysis, as all C–H halogenations exhibit unusually long radical lifetimes (2–30 ns) (Scheme 9b). Although the identity of the Mn(IV)-X intermediate is still not fully elucidated, especially the nature of the axial ligand trans to the halogen. DFT calculations showed that halogen rebound of alkyl radicals to Mn(IV)-X has low energy barriers, comparable to or even lower than the oxygen rebound.^389,391 The facile halogen transfer is also indicated by the successful application of C–H fluorination for ¹⁸F labeling.³⁹⁰ Unlike normal catalytic conditions, [¹⁸F]fluoride is present only in minuscule amounts (pmol–nmol) and other reagents (substrate, oxidant, and catalysts) are all in vast excess. Thus, Mn(IV)-¹⁸F intermediates must be competing effectively with a large amount of Mn(IV)–OH intermediates to trap alkyl radicals via fluorine transfer. Therefore, the observed high ¹⁸F conversion into target molecules indicates a preferential radical rebound to Mn(IV)–F.

These Mn-catalyzed halogenations are reminiscent of C–H chlorinations catalyzed by α-ketoglutarate (αKG)-dependent nonheme iron halogenases, in which a chloroferryl intermediate (Cl–Fe^IV=O) has been detected spectroscopically. This chloroferyl intermediate abstracts a C–H bond from substrates, and substrate radicals recombine with the chlorine rather than the hydroxyl group cis to the chloride in the rebound intermediate, Cl–Fe^III–OH (Figure 44a).^395−401 It has been shown by Bollinger and Krebs that substrate orientation is crucial for the preference for chlorine rebound.^{395,396,402,403} Computational and structural studies revealed that the oxo group repositions itself from an axial position to the equatorial plane defined by the chloride and α-KG (Figure 44, panels a and b).^395,396,399 This geometry facilitates π pathways for both the C–H abstraction and the subsequent radical recombination step. Solomon et al. analyzed the frontier molecular orbitals of Cl–Fe^III–OH. Although the Fe–OH bond is intrinsically weaker, the d_π*(Fe–OH) was found to be 4.5 kcal/mol higher in energy than d_π*(Fe–Cl), leading to a lower energy barrier for chlorine rebound (Figure 44c).⁴⁰⁴ Further, their results showed that the chlorine rebound to a substrate radical is best described as an asynchronous electron transfer from the substrate radical to iron followed by late ligand transfer to the substrate in the same single transition state, instead of a direct ligand radical transfer. This description should apply as well for the fluorine transfer step in manganese porphyrin-catalyzed fluorinations, as the generation of fluorine radicals is unlikely due to the enormously high one-electron oxidation potential of fluoride.⁴⁰⁵

Figure 44.

(a) Mechanism of C–H chlorination catalyzed by αKG nonheme halogenases. (b) Active site structure of a nonheme halogenase, WelO5 (PDB: 5IQT), with a bound 12-epi-Fisherindole U substrate. (c) Orbital diagram of d_π*(Fe–OH) and d_π*(Fe–Cl).

3.4. Oxygen-Atom Transfer Reactions Catalyzed by Compound I

In addition to activating aliphatic C–H bonds, P450s and their model compounds could also catalyze a myriad of oxygen-atom transfer reactions. These reactions, including alkene epoxidation, aromatic hydroxylation, and N- and S-oxidation, have been extensively reviewed.^{25,26,92,306,379} Herein we will only briefly discuss their mechanistic features. O-, S-, and N-dealkylations, although not belonging to OAT reactions, will also be discussed in this section for comparison with corresponding OAT reactions.

Alkene epoxidations catalyzed by P450s are widely present in natural product biosynthesis and drug metabolism.⁴⁰⁶ Also, the infamous epoxidation of benzpyrene by CYP1A1 is an important initial step in the etiology of human cancer.⁴⁰⁷ The mechanism of this oxygen transfer event appears to begin with an electrophilic ferryl addition to a C=C bond to form an alkyl radical intermediate, which undergoes a subsequent radical rebound step to forge the epoxide ring (Figure 45a). DFT calculations have indicated that the addition and rebound steps occur nonsynchronously in a single transition state or in a stepwise fashion with a rate-determining C–O bond forming step followed by a low-energy-barrier ring closure step.^306,408,409 In several alkene oxidation reactions, a cationic intermediate, presumably generated via single electron transfer from alkyl radical to the ferryl center, was inferred from the formation of aldehydes, ketone, or heme N-alkylated side products.^{17,92,146,410,411} The ketone or aldehyde products were generated via a 1,2-hydride shift, while heme N-alkylation products were likely to be formed via direct electrophilic attack of the carbocation from a pyrrole nitrogen of the porphyrin ligand.⁴¹² Very recently, Hammer, Arnold, and co-workers have harnessed the side activity of P450s for aldehyde formation to develop the first olefin anti-Markovnikov oxygenase (aMOx) via directed evolution (Figure 45b).⁴¹³ aMOx contains 12 mutations from its parent P450_LA1, a P450 from the rhodobacterium Labrenzia aggregata, and exhibits high anti-Markovnikov oxidation activity (up to 4500 TTN) and aldehyde/epoxide selectivity (up to 16:1). Furthermore, aMOx can convert α-methylstyrene to (S)-2-phenylpropionaldehyde with 91:9 enantiomeric ratio (er), representing the first catalytic asymmetric aldehyde synthesis via direct anti-Markovnikov oxidation. This result combined with a deuterium-labeling experiment indicate a mechanism involving an initial oxo-transfer followed by a highly ordered 1,2-hydride migration. It is still not clear how the 12 mutations in aMOx benefit the anti-Markovnikov activity. Some of these mutations are in the distal pocket and might help lock the substrate conformation for 1,2-hydride transfer. Intriguingly, one mutation (M391L) is in the highly conserved proximal cysteine loop, which indicates that the proximal heme ligand might also play an important role for anti-Markovnikov activity, as changes in Cys loop are known to affect the electron donation ability of proximal cysteine thiolate.¹⁸¹

Figure 45.

(a) Proposed mechanism for alkene epoxidation catalyzed by P450s. (b) Olefin anti-Markovnikov oxidation catalyzed by an engineered cytochrome P450 (aMOx). The homology model of wild-type P450_LA1 is shown with mutation sites highlighted by red sphere. Adapted with permission from ref (413). Copyright 2017 American Association for the Advancement of Science.

Aromatic hydroxylation is a paradigmatic cpd I-mediated reaction that has attracted decades of research interest, because of its importance in drug metabolism and the potential applications in organic synthesis. The mechanism involves an initial addition of cpd I to the arene π-system to form a tetrahedral σ-complex (Figure 46).^414−419 While many mechanistic studies support a direct electrophilic attack of cpd I to arenes (step a, Figure 46),^417−421 a recent study by Fujii et al. using a Marcus-type analysis showed a significant contribution of an electron transfer process in the rate-limiting step. The results were interpreted to indicate a mechanism involving a fast electron transfer equilibrium between cpd I and the arene, which is coupled to a subsequent C–O bond formation (steps b and c, Figure 46).⁴²² The tetrahedral σ-complex could adopt two forms: a cationic species with a ferric center and a radical form with a ferryl center. DFT calculations suggest that two electromers are admixed in the σ-complex.^417−420 A common mechanistic feature of P450-mediated aromatic hydroxylation is the migration of the ipso–hydrogen to the adjacent carbon (“NIH shift”) and the partial retention of this hydrogen in the final product.^30,423,424 There are several proposed mechanisms for the hydrogen migration and the formation of the hydroxylation product. The “NIH shift” could occur directly at the cationic σ-complex and led to the formation of a ketone product, which undergoes keto–enol tautomerization to form the product with partial retention of the migrated hydrogen. This mechanism is generally supported by DFT calculations,²⁵ showing that the hydrogen migration can proceed through a 1,2-hydride shift (step d, Figure 46),⁴¹⁹ or through a proton-shuttle mechanism mediated by the pyrrole nitrogen (step e).^417,418 In an alternative mechanism, an arene oxide intermediate is formed from the σ-complex (step f), which then converts into the hydroxylation product via “NIH shift” and tautomerization.^30,423,424 There are also examples in which the complete loss of the ipso–hydrogen was observed, indicating the formation of aromatic ring via direct deprotonation of the ipso–hydrogen without migration or that the keto–enol rearrangement is mediated by the active site architecture.^425,426

Figure 46.

Proposed mechanisms for aromatic hydroxylation.

Molecules containing O-, S-, N-alkyl groups undergo dealkylation, which are commonly seen in xenobiotic and drug metabolism.^427,428 Dealkylation reactions are triggered by the formation of α-hydroxylation intermediates, which decompose to afford aldehyde and dealkylated products.²⁵ There are two mechanisms proposed for the generation of α-hydroxylation intermediates: H atom-transfer (HAT) and single-electron transfer (SET) followed by proton transfer (PT) (Figure 47a). The O-dealkylation reactions generally follow an HAT mechanism as evidenced by the large primary kinetic isotope effect.^429−431 The mechanism for N-dealkylation is a topic of long debate. In support of an HAT mechanism, the use of radical clock substrates such as N-cyclopropyl-N-methylaniline afforded the ring-intact products.^432−434 Additionally, a large kinetic isotope effect (KIE > 6) was observed for dealkylation of N-alkylamides.^435,436 However, the KIE values of dealkylation of N-alkylamines are much lower (generally <3).^437,438 Furthermore, the reaction rates of dealkylation of substituted N-alkylalniline correlate well with their redox potentials and Hammett electronic parameters with negative slopes.²⁵ All these results seem to favor an SET–first mechanism. DFT calculations have provided critical insight for the understanding of N-delakylation mechanism and support the HAT–first mechanism. It has been shown that the SET step is not energetically favorable. Moreover, the low KIE and the linear correlation with substrate oxidation potentials could be reproduced with calculations based on the HAT mechanism.^318,439 Further analysis showed that the nitrogen lone pair can interact with the radical character developed at the carbon center during the C–H bond cleavage. This interaction leads to partial positive charge accumulation at the amine moiety in the transition state, causing the experimentally observed substituent effect.

Figure 47.

Proposed mechanisms for (a) O-, S-, and N-dealkylation. (b) S- and N-oxidation.

In addition to dealkylation, N-alkyl and S-alkyl groups could undergo direct heteroatom oxidation to afford N-oxide and sulfoxide products (Figure 47b). The mechanisms for both reactions have been shown to be a concerted direct oxygen atom transfer.^{313,440−443} In theory, the heteroatom oxidation and dealkylation could compete with each other. Indeed, both products were observed in some examples.^17,444,445 Intriguingly, it seems that N-alkyl groups favor dealkylation while S-alkyl groups prefer sulfoxidation. A recent computational study reproduced this trend and found that the activation energies of dealkylation pathway was largely affected by the radical stabilization ability of the neighboring heteroatom.⁴⁴³ The nitrogen inversion barrier of tertiary amines was recently found to affect the N-oxidation. Olsen et al. studied a series of tertiary amines and found that those compounds with N-oxidation reactivity had lower N-inversion barriers.⁴⁴⁶

Computational studies of sulfoxidation and N-oxidation have shown that the energy of the quartet transition state was much higher than that of the doublet transition state. Thus, the reaction is predicted to occur along the low-spin pathway.313,440−443,446,447 This is in sharp contrast to aliphatic hydroxylation, epoxidation, and aromatic hydroxylation reactions,305,306 in which close-lying transition states were found for both doublet and quartet spin surfaces.408,409,421,448−451 The main distinction of S- and N-oxidation compared to other oxidation reactions is that they were indicated to proceed through a synchronous 2e– transfer process. Therefore, in the high-spin quartet transition state, one electron from sulfur or nitrogen has to migrate to a high-energy σz2* orbital,441 giving rise to the high energy barriers of quartet transition states. Primary and secondary amines can also be oxidized via N-hydroxylation. This reaction has been much less explored. DFT calculations support two possible mechanisms with comparable activation energies, a direct oxygen atom transfer (OAT)/isomerization pathway and a hydrogen atom transfer (HAT)/radical rebound pathway.452−455

The axial ligand effect has also been explored for oxygen atom transfer reactions (OATs). Initial studies by Gross et al. showed that electron-donating ligands (such as F^– and Cl^–) substantially increased the epoxidation reactivity of model cpd I complexes, whereas cpd I species with weakly coordinating ligands, such as CF₃SO₃^– and ClO₄^– displayed lower reactivity.^456−458 A decrease in Fe^IV=O bond strength for electron-donating ligands was indicated by their lower stretching frequencies (ν_Fe=O = 801 cm^–1 as determined by resonance Ramon spectroscopy) compared to those with weak-ligating ligands (ν_Fe=O = 834–835 cm^–1).^457,458 Recently, phenolate and electron-neutral imidazole axial ligands were found to accelerate epoxidations with second-order rate constants 10 times faster than that of chloride ligand.⁴⁵⁹ The effect of para-substituted pyridine N-oxides ligands (p-Y-PyO) were also explored for alkene epoxidation and aromatic hydroxylation, and a rate enhancement was observed for both reactions by increasing the electron-donating ability of para-substituent Y.²⁸⁵ DFT calculations were employed to study the impact of axial ligands on a number of OAT reactions, including alkene epoxidation,^409,449 aromatic hydroxylation,^448,451 and sulfoxidation.⁴⁴⁰ In all of these reactions, calculated activation energies were not dependent on the reduction potential of cpd I but did exhibit a good negative linear correlation with BDE_OH of cpd II. This correlation might suggest that the C–O or S–O bond formation step in OATs shares similarities with the O–H bond formation in HAT reactions. Alternatively, it might imply the importance of Fe^IV=O weakening in cpd I for OAT, since calculations showed that, in addition to increasing the cpd II basicity, the electron donation from axial ligand also weakened the Fe^IV=O bond in cpd I.²⁸⁵

3.5. O–O Bond Formation Mediated by Oxometalloporphyrins

3.5.1. Oxygen Evolution from Oxometal Complexes

A key step in photosynthesis is the oxidation of water to dioxygen. This four-electron, four-proton process generates the electrons finally used to convert CO₂ to carbohydrates.⁴⁶⁰ In photosystem II, the water oxidation occurs via multiple proton-coupled electron transfers (PCETs) at the Mn₄Ca oxygen-evolving complex (OEC).⁴⁶¹ The development of molecular catalysts that can catalyze this process under mild conditions and low energy cost is important for solar energy conversion and storage.⁴⁶¹ Metalloporphyrins, with their ability to form high-valent metal–oxo intermediates, are promising catalysts for water oxidation and oxygen evolution.^33,462,463 The major pathways considered for metal–oxo-mediated water oxidation include nucleophilic attack of water/hydroxide by oxoMn(V) and the coupling of two metal–oxo (or μ-oxo) units (Scheme 10). The former is formally the microscopic reverse of the O–O bond cleavage for cpd I generation, while the coupling between two μ-oxo units in the latter pathway will afford a (μ-η²:η²-peroxo)dimetal complex similar to the oxygen adduct of hemocyanine.^464−466 In a third mechanism, the O–O bond is formed via coupling of a metal-bound oxide with a metal-ligated μ-oxo bridge.^467−470 This mechanism is mainly proposed for O₂ evolution of photosystem II OEC and related polynuclear metal–oxygen complexes.

Scheme 10. Proposed Mechanisms for Water Oxidation Mediated by Metal–Oxo Intermediates.

An early example of a water–oxidation catalyst based on metalloporphyrins is a dimeric manganese porphyrin complex 81 synthesized by Naruta et al. (Figure 48a).⁴⁷¹ This species can catalyze the anodic water oxidation in aqueous acetonitrile solution containing ⁿBu₄NOH with a potential range of 1.2–2.0 V (vs Ag/Ag⁺) and Faradaic efficiency of 5–17%. An dimeric oxoMn^V intermediate 52 was later prepared by treating the Mn(III) dimer complex with mCPBA as a potential reactive intermediate for water oxidation (Figure 48a).²³¹ This species is very stable under basic conditions but decomposes to Mn(III) and dioxygen upon neutralization with acid. Åkermark, Sun, and co-workers reported a monomeric manganese porphyrinoid system for water oxidation (Figure 48b).⁴⁷² A manganese corrole complex 83 was shown to efficiently mediate the oxygen evolution at a low potential ∼0.8 V versus Ag/AgNO₃. An oxoMn^V intermediate 84 was prepared from 83. The addition of hydroxide 84 led to the formation of O₂ presumably via the nucleophilic attack of OH^–. A Mn(IV)-peroxo intermediate 86 was detected during this process, which was likely formed from the further oxidation of the Mn(III)-hydroperoxo intermediate 85. Nam also reported the formation of a Mn^IV-peroxo intermediate upon adding base to a oxoMn(V) corrole complex, Mn^V(O)(TFMPC) (TFMPC = 5,10,15-tris(3,5-trifluoromethylphenyl)corrolato trianion).⁴⁷³ This step is reversible as the addition of acid to peroxo-Mn^IV(TFMPC) could regenerate the oxoMn(V) intermediate.

Figure 48.

(a) Water oxidation mediated by a dimeric manganese porphyrin complex. Adapted from ref (231). Copyright 2004 Wiley-VCH. (b) Water oxidation catalyzed by a manganese corrole complex 83. Adapted from ref (472). Copyright 2009 American Chemical Society.

In addition to manganese, cobalt is another base metal that has been extensively studied for water oxidation. Nocera et al. reported the first cobalt porphyrinoid system for water oxidation (87, Scheme 11a) using a cobalt “hangman” β-octafluoro corrole complex 87. When immobilized in Nafion films, 87 can catalyze water oxidation at 1.4 V versus Ag/AgCl at pH 7 with turnover frequency (TOF) of 0.81 s^–1.⁴⁷⁴ Sakai et al. reported several Co(III) porphyrins, including Co(TPP), Co(TMPyP), and Co(TPPS), for water oxidation via photon-initiation using Ru^II(bpy)₃²⁺/Na₂S₂O₈ (Scheme 11b). The maximum TOFs were in the range of 0.16–0.35 s^–1 at pH 11.⁴⁷⁵ A second-order dependence of the initial rate on catalyst concentration was observed, suggesting a bimolecular oxyl–oxyl coupling pathway for O–O bond formation. The degradation of the catalyst has been observed presumably due to the highly oxidizing sulfate radical anions and single dioxygen. More recently, the same group developed a fluorinated Co(III) porphyrin catalyst CoFPS for water oxidation with improved stability and efficiency (TOF = 1.1 s^–1).⁴⁷⁶ In contrast to their previous studies, a first-order catalyst concentration dependence was observed for CoFPS, suggesting a nucleophilic attack pathway for the water oxidation reaction catalyzed by CoFPS.

Scheme 11. (a) Cobalt Hangman Corrole Complex for Water Oxidation and (b) Several Cobalt Porphyrins Utilized for Water Oxidation.

Reprinted from ref (474). Copyright 2011 American Chemical Society.

Groves and Wang developed a series of electron-deficient cationic cobalt porphyrins for highly efficient electrocatalytic water oxidation.⁴⁷⁷ For Co^III(TDMImP) (88, Figure 49), at a potential of 1.3 V versus Ag/AgCl at pH 7, the O₂ formation rate was determined to be 170 nmol cm^–2 min^–1 (k_obs = 1.4 × 10³ s^–1) with a near 90% Faradaic efficiency. The authors conducted detailed mechanistic studies of these cobalt-based water–oxidation systems. They identified that a Co^IV–O porphyrin cation radical species P^•+Co^IV–O (92, Figure 49) is the active intermediate for water oxidation. A linear dependence between catalytic current and catalyst concentration suggested that the rate-limiting O–O bond formation occurred at a single cobalt site. Intriguingly, the buffer anion was identified to play multiple important roles in the catalytic cycle. First, it acts as a proton acceptor in the conversion of P^•+Co^III–OH to P^•+Co^IV–O, which is found to be a single PCET process with a slope of −54 mV pK_a^–1 for a plot of onset potential of water oxidation versus buffer pK_a. Furthermore, it acts as a base to deprotonate water in the rate-limiting O–O formation step, as indicated by a linear correlation between (i_cat/i_water)² versus buffer base concentration and a KIE of 2.8 measured in H₂O versus in D₂O. Finally, an inhibition of water oxidation was observed at high buffer concentrations, presumably resulting from the formation of an inactive bis-coordinated Co(III) intermediate 91 with buffer anions. The unchanged intensity and UV–vis spectroscopic features of 88 after bulk electrolysis, the absence of an inhibition effect of EDTA, and the clean electrode surface with no sign of heterogeneous cobalt phase clearly exclude the possibility of forming heterogeneous metal oxide films at electrode surface, which is known to be a potential issue for homogeneous water–oxidation catalysts.⁴⁷⁸ More recently, Cao et al. reported a nickel-based water–oxidation catalyst with a cationic porphyrin ligand.⁴⁷⁹ A TOF of 0.67 s^–1 was observed at a potential of 1.32 V (vs NHE). The reactive intermediate of this system is proposed to be a Ni(III)–oxyl radical species. As in the case of analogous cobalt cationic porphyrins, buffer base was shown to participate in the rate-determining O–O bond formation step. An inhibition effect of buffer base at high concentrations was also observed.

Figure 49.

Proposed mechanism for water oxidation catalyzed by Co-porphyrins (B = buffer anions). Adapted with permission from ref (477). Copyright 2013 National Academy of Sciences.

3.5.2. Chlorite Dismutase

The conversion of chlorite to chloride and dioxygen is another reaction that might involve the O–O bond formation mediated by cpd I. This process is important for the microbial perchlorate respiration, as it removes the toxic end-product chlorite.⁴⁸⁰ In nature, the decomposition of chlorite is catalyzed by chlorite dismutase (Cld), a heme-containing protein with a histidine proximal ligand.⁴⁸¹ The active site structure of Cld is shown in Figure 50, which is featured by an arginine positioned above the heme and the lack of other hydrogen-bonding residues that are essential for O–O bond heterolytic cleavage.⁴⁸² The proposed mechanism for functional Cld involves an initial formation of a ferric–chlorito intermediate (Fe^III–OClO), analogous to compounds 0 in P450s and peroxidases (Figure 50c).⁴⁸³ A subsequent O–Cl bond heterolytic cleavage will generate cpd I and hypochlorite, which will undergo nucleophilic attack at cpd I to generate Cl^– and O₂.⁴⁸⁴ Alternatively, the homolysis of Cl–O bond of Fe^III–OClO will afford compound II (cpd II) and chlorine monoxide radical (ClO·. Cl^– and O₂ are formed via a rebound process between ClO· and cpd II. The latter mechanism is analogous to the mechanism proposed for NO oxidation catalyzed by nitric oxide dioxygenase (see section 4.2). The distal arginine can adopt either “in” and “out” conformations (Figure 50a and 50b),⁴⁸² which is proposed to play an important role in positioning the transiently formed hypochlorite to ensure its efficient capturing by cpd I (or cpd II).^485−487 There are two phylogenetically distinct class of Clds (clades 1 and 2). Clade 1 Clds (including NdCld and DaCld shown in Figure 50) are pentameric and hexameric, whereas clade 2 Clds are dimeric. Very recently, Obinger and co-workers have carried out detail structural and kinetic studies of a dimeric clade 2 Cld from Cyanothece sp. PCC7425 (CCld).⁴⁸⁸ The spectral features of various intermediates of dimeric CCld appeared to be much better resolved than those of clade 1 Clds, which allowed for the presteady-state kinetic studies of CCld. Their results showed that mixing CCld with chlorite under stopped-flow conditions led to the formation of cpd II within milliseconds at pH > 5.0. Furthermore, no O₂ formation was observed by treating CCld cpd I with excess hypochlorite (HOCl/^–OCl). These results suggest a homolysis pathway for chlorite degradation by CCld.

Figure 50.

(a) Active site of chlorite dismutase (Cld) from Candidatus Nitrospira defluvii (NdCld) with an imidazole ligand (PDB: 3NN1). Catalytically important arginine (Arg173) exhibits an “out” conformation in this structure. (b) Active site of Cld from Dechloromonas aromatica (DaCld) with a nitrite ligand (PDB: 3Q09). Catalytically important arginine (Arg183) exhibits an “in” conformation in this structure. (c) The proposed mechanism for chlorite dismutase.

The chlorite decomposition reactions have also been explored with metalloporphyrins. In a Mn porphyrin-catalyzed alkane oxidation with chlorite as oxidant, Collman et al. found a catalytic side reaction that produced Cl^– and O₂ from chlorite.⁴⁸⁹ The disproportionation of ClO₂^– into Cl^– and ClO₃^– was also observed especially in the absence of hydrocarbon substrates. More recently, Abu–Omar and co-workers studied the reaction between chlorite and several water-soluble iron porphyrins (Figure 51a).⁴⁹⁰ They also identified two degradation pathways of ClO₂^–, in which one afforded ClO₃^– and Cl^– and the other yielded O₂ and Cl^–. The more electron-withdrawing fluorinated Fe(TF₄MAP) catalyzed both reactions, while less electron-deficient Fe(TMAP) and Fe(TPPS) only catalyzed the disproportionation of ClO₂^– to ClO₃^– and Cl^–. On the basis of isotope labeling studies, they proposed Cl^– and O₂ were formed via a cpd I/rebound pathway similar to that proposed for Cld (cycle A, Figure 51), whereas that ClO₃^–/Cl^– pathway proceeds through a Fe(II) porphyrin/cpd II catalytic cycle (cycle B, Figure 51).

Figure 51.

(a) Iron porphyrins to mimic Cld reactivity. (b) The proposed mechanism of ClO₂^– decomposition to form Cl^– and O₂ and that of ClO₂^– disproportionation to ClO₃^– and Cl^– catalyzed by iron porphyrins. Adapted from ref (490). Copyright 2009 American Chemical Society.

Recently, electron-deficient cationic manganese porphyrins have been shown to catalyze the generation of chlorine dioxide from chlorite by Groves and Abu–Omar groups, respectively.^491−493 The reaction proceeds rapidly under mild conditions. For Mn(TDMImP), TOFs were determined to be 0.47 s^–1 at pH 6.8 and 1.0 s^–1 at pH 4.7. The catalytic efficiency can be dramatically increased by changing Mn(TDMImP) to an electron-deficient manganese porphyrazine catalyst MnTM23PyPz (TOF = 32.0 s^–1 at pH 4.5) (Figure 52a).⁴⁹² A detailed mechanistic study by Groves et al. with Mn^III(TDMImP) catalyst revealed an initial generation of oxoMn(V) and ClO^–, which then diffuses to oxidize a second Mn(III) site to oxoMn(V) (Figure 52b).⁴⁹² This is in sharp contrast to Cld and iron porphyrins, in which an in-cage recombination between cpd I and ClO^– would afford O₂ and Cl^–. Each oxoMn(V) species could produce two ClO₂ molecules via successive out-sphere electron transfer to ClO₂^–. The oxidation of ClO₂^– to ClO₂ by oxoMn^V and oxoMn^IV showed comparable rates at pH 6.8 (k_app = 3.13–6.30 × 10³ M^–1 s^–1). The oxidation of ClO₂^– by oxoMn^IV is also reversible [k_app = (4.59 ± 2.37) × 10² M^–1 s^–1]. Furthermore, the evolved ClO₂ can be further oxidized to ClO₃^– by oxoMn^V and oxoMn^IV in a subsequent phase of the reaction. Similar mechanistic features but at much slower rates have been described by Abu–Omar et al. for a Mn(TF₄TMAP)/NaClO₂ system.⁴⁹⁴

Figure 52.

(a) Mn porphyrin and porphyrazine catalysts for generation of ClO₂ from chlorite and their turnover frequencies. (b) Mechanism of chlorite dismutation catalyzed by Mn(TDMImP) and the apparent second-order rate constants for the elementary steps. Hypochlorite is written in protonated form as the measurements were carried out at a pH range of 4.7–6.8. Adapted from ref (492). Copyright 2011 American Chemical Society.

3.6. Single Electron Transfer (SET) Reactions of Cpd I

In addition to participating in hydrogen atom transfer (HAT) and oxygen atom transfer (OAT) reactions, compound I can also display single electron transfer (SET) reactivity, which is commonly seen in heme-containing peroxidases. Unlike most P450s, peroxidase compound I is generated directly from the resting ferric heme and H₂O₂ or organic hydroperoxides via the “peroxide shunt” mechanism. Peroxidase cpd I can carry out two discrete single-electron transfers to two substrate molecules and regenerate the ferric heme state (Scheme 12). This three-step mechanism was first proposed by Dunford et al. for the oxidation of ferrocyanide by H₂O₂ and HRP.²⁴⁶ The resulting substrate radicals can undergo further transformations.⁴⁹⁵ Thus, substrates of heme peroxidases are generally good one-electron donors, such as electron-rich aromatics, ascorbate, ferrocytochrome c, etc.⁴⁹⁶ Even strongly oxidizing Mn³⁺ can be produced from Mn²⁺ in this way by manganese peroxidase.

Scheme 12. Heme Peroxidase Generates Cpd I with H₂O₂ and Catalyzes Two Discrete Single Electron Transfer (SET) Reactions to Substrates.

Heme peroxidases are not the only enzymes that utilize H₂O₂ to form compound I. Unspecific peroxygenases (UPO) and chloroperoxidase (CPO) employ a similar “peroxide shunt” pathway to generate compound I to catalyze HAT and OAT reactions.⁴⁹⁷ As discussed in section 3.1, a major cause of this reactivity difference is the much larger O–H BDEs of UPO-II and CPO-II (98–100 kcal/mol) compared to that of HRP-II, originating from the elevated reduction potentials of UPO-I and CPO-I and the decreased acidities of UPO-II and CPO-II. Furthermore, the 1e^– reactions in peroxidases are facilitated by the proper positioning of substrates.⁴⁹⁸ The local protein environment of peroxidases hinders the approach of substrates to the Fe^IV=O moiety. Therefore, many substrates bind in the heme periphery and participate in a hydrogen-bonding network with heme propionate and active site residues (Figure 53).⁴⁹⁹ This network can serve as a relay to mediate the proton and electron transfer.^500,501 Alternatively, substrates can be oxidized at a remote site through a chain of redox-active amino acids via long-range electron transfer.^502−505

Figure 53.

Examples of substrate binding to the heme periphery of peroxidases and forming hydrogen bonds with heme propionates and active site residues. (a) Ascorbate peroxidase (APX, PDB: 1OAF). (b) Manganese peroxidase (MnP, PDB: 1YYD).

In some peroxidases, the porphyrin cation radical form of compound I is transient and will rapidly transfer one electron to a nearby redox-active residue like tyrosine or tryptophan. One example is the compound ES of cytochrome c peroxidase (CcP), which is an iron(IV)-oxo intermediate with an uncoupled Trp191 cation radical near the proximal histidine.¹⁸⁴ Although the Trp191 radical in CcP just serves as a storage for one oxidation equivalent, in some heme proteins, such radicals can directly mediate reactions. For instance, in prostaglandin H synthase (PGHS), the compound I formed via the normal peroxide activation pathway will oxidize a tyrosine (Tyr385) about 10 Å away to give a tyrosyl radical, which abstracts a hydrogen atom from arachidonic acid (Figure 54).^506,507 The resulting substrate radical will further react with two O₂ molecules to give product PGG2 and finally prostaglandin H₂ (PGH₂). This reaction also represents an example of substrate activation for O₂ reactions in heme proteins.²

Figure 54.

Conversion of arachidonic acid (AA) into PGG2 by PGHS.

Another example of using protein radicals for substrate activation is the synthesis of heme b by coproheme decarboxylase (HemQ).⁵⁰⁸ HemQ catalyzes a H₂O₂-dependent decarboxylation of two propionates in coproheme (94, Scheme 13). Recent mechanistic studies by DuBois et al. showed that Tyr145, which hydrogen bonds with one of the cleaved propionates, is first converted to a tyrosyl radical by coproheme compound I (95, Scheme 13).^509,510 The Tyr145 radical then abstracts a hydrogen from the heme propionate group presumably at the C_β position (96). The resulting carbon-centered radical transfers one electron through porphyrin to Fe(IV) and reduces it to iron(III) with loss of CO₂ (97 → 98). The decomposition of the second propionate is thought to follow a similar mechanism, but the identity of the mediator of C_β-H cleavage is not clear.

Scheme 13. HemQ-Catalyzed Synthesis of Heme b from Coproheme III.

Adapted from ref (509). Copyright 2017 American Chemical Society.

4. Reactions Mediated by Heme–Oxygen Intermediates Other than Cpd I

In addition to cpd I, other heme–oxygen intermediates generated during O₂ activation also exhibit rich redox chemistries that are harnessed by heme proteins for a variety of oxidative transformations, giving rise to the diverse biological functions of heme proteins. In this section, we analyze several representative reactions catalyzed by heme proteins and metalloporphyrins to delineate their mechanistic features and to discuss how the fundamental properties of heme–oxygen intermediates and their active site environments impact reaction outcomes.

4.1. Homolytic O–O Bond Cleavage of Peroxides by Heme Proteins

In addition to generating compound I via O–O bond heterolysis, organic peroxides can also react with ferric hemes via single-electron-transfer processes, which will oxidize the ferric heme to ferryl with concurrent homolytic cleavage of O–O bond. These transformations play important roles in the synthesis of important bioactive molecules, especially oxylipins.⁵¹¹ Natural sources of peroxides include fatty acid hydroperoxides formed via O₂ trapping of substrate allylic radicals.⁵¹² Subsequent O–O bond homolysis leads to oxy radicals that can undergo a plethora of rearrangements, bond cleavages, and other transformations.

One example is the synthesis of prostacyclin from prostaglandin H₂ (PGH₂) via prostacyclin synthase (Scheme 14a).^507,513,514 This transformation starts with the interaction between endoperoxide of PGH₂ and ferric heme center of prostacyclin synthase (104, Scheme 14a), leading to the homolytic O–O bond cleavage of endoperoxide to afford the C9-alkoxyl radical and ferryl heme intermediate (105). The C9-alkoxyl radical then add to the C5–C6 alkene to forge the new five-membered heterocycle and C5 radical (106), which is later oxidized to carbocation by the ferryl heme to regenerate the C5–C6 alkene moiety via loss of a proton (107). The homolytic cleavage of endoperoxide of PGH₂ is also used in the synthesis of thromboxane A₂ (TXA₂) mediated by thromboxane A synthase (Scheme 14b).^515−517 In this reaction, the ferric heme center interacts with O9 and afforded C11-alkoxyl radical (109 and 110, Scheme 14b), which is later rearranged to an allylic radical at C12 (111). Subsequent electron transfer and ring cyclization would afford TXA₂.

Scheme 14. (a) Synthesis of Prostacyclin from PGH₂ Mediated by Prostacyclin Synthase and (b) Synthesis of Thromboxane A₂ (TXA₂) Catalyzed by Thromboxane A Synthase.

In another instructive example, 13S-hydroperoxyoctadecatrienoic acid [13(S)-HPOT] can be activated by the CYP74 family of enzymes, affording three different products: 12,13(S)-allene oxide, etherolenic acid, and an analogous hemiacetal (Figure 55a).^518,519 The first step in all of these three transformations is the homolysis of an iron-bound O–O bond to afford an epoxyallylic radical and a cysteine-ligated Fe^IV–OH intermediate similar to compound II in P450 monooxygenases. The reactions then diverge depending on how the incipient substrate radical interacts with the Fe^IV–OH intermediate. In allene oxide synthase (AOS), the C11 epoxyallylic radical is oxidized by electron transfer to Fe^IV–OH to generate a substrate-derived carbocation intermediate, which eliminates a proton to form the new C=C bond. In hydroperoxide lyase (HPL), by contrast, the radical undergoes rearrangement to form a radical at C13, which then recombines with Fe^IV–OH via an oxygen-rebound mechanism. In the divinyl ether synthase (DES) pathway, the product is thought to be formed via a hydrogen atom abstraction from C14 by Fe^IV–OH or via an electron-transfer process followed by cation rearrangement and proton elimination. Sequence analysis and mutagenesis studies revealed that, in all three enzymes, an asparagine positioned directly above heme is important for mediating the O–O bond homolysis.⁵²⁰ The crystal structure of Arabidopsis thaliana AOS (AtAOS) complexed with 12R,13S-vernolic acid, a mimic of the epoxyallylic intermediate, showed that C11 is in proximity to F137 (Figure 55b), whereas in HPL and DES, a similar position is occupied by a leucine. The aromatic ring of F137 is thought to stabilize the C11 carbocation and thus facilitate the single electron transfer in the AOS pathway. In support of this rationale, a F137L mutation induced HPL activity for AOS.

Figure 55.

(a) Reactions catalyzed by the CYP74 family of enzymes. The mechanisms of all three pathways involve a common epoxyallylic radical intermediate. (b) Active site structure of Arabidopsis thaliana AOS complexed with vernolic acid (PDB: 2RCL). (c) Key N321 and I328 residues of AtAOS (PDB: 3CLI). The cysteine loop region is highlighted in red. (d) Key threonine residue of P450_BM3 (PDB: 1BVY). The cysteine loop region is highlighted in yellow.

It is intriguing to compare the structural difference between CYP74 and P450 hydroxylases, since CYP74 has a typical P450 protein fold but does not perform O–O bond heterolysis. In addition to the critical Asn residue (N321) located above the heme, there are two more structural features that distinguish CYP74 from P450 hydroxlyases.⁵²⁰ As can be seen from the crystal structure of AtAOS, the highly conserved threonine in the proton relay channel of P450 hydroxylases, which is crucial for O–O bond heterolysis, is replaced by an isoleucine (I328) (Figure 55, panels c and d). Further, there is a nine-residue insert in the heme-binding loop of AOS, which seems to contribute to the long Fe–S bond (2.4 Å) of AOS. A long Fe–S distance has been shown to decrease the electron-donating ability of cysteine thiolate, which, in turn, would lead to a higher Fe reduction potential and a more facile SET reactivity of compound II.¹⁸¹

In addition to CYP74, there is another enzyme that can also catalyze the synthesis of allene oxides via homolytic cleavage of fatty acid hydroperoxides. This enzyme is structurally similar to catalase and thus termed catalase-related allene oxide synthase (cAOS).⁵²¹ cAOS converts 8R-hydroperoxyarachidonic acid (8R-HPETE) into corresponding 8,9-epoxy allene oxide in the biosynthesis of preclavulone A. The crystal structure comparison between cAOS and catalase showed very similar active site structures, including the proximal heme-ligating tyrosine and the critical distal histidine and asparagine that are important for the O–O bond activation in catalase (Figure 56). A key difference is identified at a residue adjacent to the distal histidine, which is normally a valine in catalases but is changed to a threonine (T66) in cAOS.⁵²² The histidine in cAOS hydrogen bonds with both T66 and asparagine. This hydrogen bonding network could function to dampen H₂O₂ activation in cAOS by inhibiting the hydrogen bonding between histidine and the hydroperoxo moiety of Fe^III–OOH (cpd 0). Indeed, a T66V mutation greatly increased the catalase activity of cAOS, albeit with short lifetime due to the heme degradation. cAOS is thought to proceed through a similar epoxy allylic radical intermediate. As in the case of CYP74, products other than allene oxide were also observed in reactions catalyzed by cAOS and its relatives.^523−528

Figure 56.

(a) Reactions catalyzed by cAOS and the active site structure of cAOS (PDB: 1U5U). (b) Active site structure of human catalase (PDB: 1QQW) and the mechanism of O–O bond activation in catalase-compound 0.

4.2. Reactions Initiated by Ferric-Superoxo Intermediates

In most heme proteins, ferric-superoxo complexes either serve as stable complexes, such as in oxymyoglobin and gas sensing proteins such as heme-nitric oxide/oxygen (H-NOX) binding proteins,^529−531 or rapidly convert to high-valent iron–oxo intermediates via reduction and O–O bond scission. Tryptophan 2,3-dioxygenase (TDO) and indoleamine 2,3-dioxygenase (IDO), however, utilize ferric-superoxo intermediates to initiate the oxidation of substrates.^532,533 TDO and IDO are histidine-ligated heme-containing dioxygenases involved in tryptophan oxidation.^534−536 TDO employs dioxygen to cleave the 2,3-bond of l-tryptophan to form N-formyl kynurenine, whereas IDO catalyzes the same reaction but has a much broader substrate scope, including a variety of indoleamine derivatives such as d-tryptophan, melatonin, and serotonin. IDO has recently gained substantial research interest because of its immunomodulatory potential for cancer treatment.^537,538

The initially proposed mechanism for TDO and IDO-catalyzed tryptophan oxygenation involved a histidine-catalyzed proton abstraction of the indole N–H, followed by electrophilic attack at C3 of indoleamine by the distal oxygen of ferric-superoxo intermediate (Figure 57a).⁵³⁹ But this mechanism is not consistent with later findings showing that N-methyl-tryptophan was an active substrate and that human IDO lacked the proposed proton-abstracting histidine.^540−542 DFT calculations indicated high energy barriers for the formation of either a dioxetane intermediate or a Criegee-type rearrangement.⁵⁴³ Further studies by Yeh et al. using resonance Raman and UV–vis spectroscopies revealed the presence of a ferryl species during catalytic turnover of IDO.^544−546 Similar ferryl species can also be generated by treating IDO and TDO using H₂O₂.^{541,547−549} On the basis of these results, a new mechanism has been proposed (Figure 57b), in which the ferric-superoxo complexes directly attacks the indoleamine ring via either electrophilic addition or radical addition to afford an indoline-2,3-epoxide intermediate and a ferryl compound II. Subsequent reactions between the epoxide and Fe(IV)=O are currently speculative but do lead to the formation of kynurenine type product and regenerate the ferrous heme. While the ferryl intermediate of IDO accumulates under steady-state conditions, TDO cpd II has yet to be observed during turnover. Rather, the [Fe(II)–O₂, l-Trp] ternary complex accumulates under catalytic conditions. Raven et al. have recently suggested that this difference is caused by a change of the rate-limiting step from compound II decay in IDO to oxygen addition to formation of the 2,3-epoxide intermediate in TDO.⁵⁵⁰

Figure 57.

(a) Previously proposed base-abstraction mechanism followed by either Criegee rearrangement or dioxetane formation. Adapted from ref (539). Copyright 2011 American Chemical Society. (b) A recently proposed mechanism involves the formation of an Fe^IV=O intermediate and a 2,3-indoline epoxide intermediate. Adapted from ref (550). Copyright 2016 American Chemical Society.

Another enzyme that utilizes an ferric-superoxo intermediate for oxygenation is nitric oxide dioxygenase (NOD).⁵⁵¹ As shown in Figure 58a, NO first reacts with an ferric-superoxo intermediate to form a proposed ferric-peroxynitrite intermediate. This intermediate then undergoes homolytic cleavage to generate NO₂ and compound II.⁵⁵² The recombination between NO₂ and compound II yields nitrate and ferric heme. This recombination process is reminiscent of the oxygen rebound of carbon-centered radicals in P450s. The study of peroxynitrite decomposition mediated by met-myoglobin offered critical insight into the radical rebound mechanism, especially the radical cage escape effect.^553,554 Unlike carbon-centered radicals, the diffusion of NO₂ can be easily detected via fluorogenic probes such as fluorescein. Kinetic simulations revealed that 10% of the NO₂ escaped from the radical cage [MbFe^IV=O ·NO₂] formed after peroxynitrite decomposition. Cytochrome P450 TxtE from Streptomyces scabies also employs a ferric-superoxo intermediate to generate NO₂ from NO, which is then used to nitrate l-tryptophan to form 4-nitro-l-tryptophan.^555,556 The proposed mechanism involves the formation of a ferric-peroxynitrite intermediate by the trapping of NO by the ferric-superoxo intermediate (Figure 58b). The NO₂ generated after peroxynitrite decomposition can then add to the aromatic ring, which is further oxidized by compound II to afford the nitration product. An alternative mechanism involves heterolytic cleavage to generate nitronium ion (NO₂⁺) for direct electrophilic nitration.⁵⁵⁵

Figure 58.

(a) Proposed mechanism for NO oxidation catalyzed by NOD. (b) Proposed mechanism for tryptophan nitration catalyzed by P450 TxtE.

4.3. O₂ Reduction Catalyzed by Heme/Cu Terminal Oxidases and Related Model Compounds

Heme/Cu terminal oxidases such as cytochrome c oxidases (CcO) catalyze the 4e^–/4H⁺ reduction of O₂ as a final step in respiration with concomitant membrane proton-translocation leading to ATP synthesis.⁵⁵⁷ The active site structure of CcO is unusual compared to many heme proteins as it contains a binuclear heme a₃/Cu_B center (BNC) with a Fe–Cu separation of around 5 Å (Figure 59a). Cu_B is coordinated to three histidine residues, one of which is covalently cross-linked to a tyrosine residue.^558−560 During O₂ reduction by CcO, electrons from cyt c are delivered one at a time to BNC via a dicopper complex (Cu_A) and a low-spin heme a. The mechanism of CcO action has been extensively studied. The catalytic cycle involves four consecutive proton and electron transfer steps and multiple intermediates are involved. Figure 59b showed a simplified mechanism of CcO-catalyzed oxygen reduction.⁵⁶¹ The Fe(III)–Cu(II) resting state of BNC (compound O) is converted to an E state and then to a Fe(II)–Cu(I) reduced state (R) via two one-electron reductions and two proton transfer steps. The ferrous heme a₃ in R will bind oxygen rapidly (∼1.4 × 10⁸ M^–1 s^–1), generating a low-spin end-on Fe–O₂ intermediate (A) similar to other oxygen-binding heme proteins. O–O bond cleavage in A, which can be viewed as an O–O hemolysis in a bridged Fe(III)–O–O–Cu(II) intermediate, yields a four-electron oxidized intermediate P_M, containing a ferric heme a, a Fe(IV)=O, Cu(II)–OH, and a tyrosyl radical. Four subsequent 1e^– reductions and proton transfers will lead sequentially through the F, O and E states, poising the protein for another oxygen-binding event. It is worth noting that in Figure 59b, we only showed intermediates formed after both electron- and proton-transfer steps. Furthermore, depending on the status of copper ligation and redox potentials of copper center, the cross-linked tyrosine could exist as tyrosyl radical with Cu(II) center being reduced to Cu(I).^561,562 Each of these reduction/protonation steps (scalar protons) are also coupled to one vectorial proton pumped across the membrane, thus maintaining the proton gradient required for ATPase phosphorylation.⁵⁶¹

Figure 59.

(a) Cofactor arrangement in a cytochrome c oxidase (PDB: 2YEV) and the structure of binuclear heme a₃/Cu_B center (BNC). (b) General catalytic cycle of CcO and important intermediates involved (R, A, P_M, F, O, and E). This mechanism is produced according to the catalytic cycles shown in refs (561) and (562).

While many aspects of the CcO mechanism have been clarified, the details of the O–O cleavage to generate compound P_M and the role of the Cu_B His-Tyr unit in this step remains elusive. As noted above, a putative ferric peroxo intermediate (Fe^III–O–O–Cu^II) is thought to be generated via 1e^– reduction of ferric-superoxo intermediate by Cu(I) center (Figure 60a).⁵⁵⁸ A subsequent proton-coupled electron transfer (PCET) from tyrosine to Fe^III–O–O–Cu^II could promote the O–O cleavage and generate the ferryl center and tyrosyl radical. Yoshikawa et al. have examined the binding of diatomic molecules (CO, NO, and CN^–) to the reduced bovine CcO via X-ray structural analysis.⁵⁶³ Compared to CO- and NO-bound structures, CN^– binding leads to the displacement and dissociation of a copper-bound histidine and the coordination of cyanide nitrogen to the copper. Furthermore, a water molecule is incorporated to form a hydrogen bond with both CN^– and covalently ligated tyrosine (Tyr244) (Figure 60b). These structural features were thought to be highly relevant to the ferric-superoxo intermediate of CcO, where O₂ is negatively polarized similarly to CN^–. The hydrogen-bonding network in the active site could provide a channel for PCET from Tyr244 to the peroxo moiety. Another intriguing feature of this CN-bound CcO structure is the significant bending of the Fe–C–N angle (96°). This bent structure is in line with previous resonance Raman characterizations and might be caused by the weak π-accepting ability of cyanide, which makes it easier to adopt a bent geometry compared to the isoelectronic Fe-CO complexes.^564−567

Figure 60.

(a) The proposed ferric-peroxo-cupric intermediate for the generation of P_M state. (b) A CN-bound CcO might share similar structural features to the putative ferric-peroxo-cupric intermediate (PDB: 3AG4).

Model compounds have offered critical insights to the O–O heterolytic cleavage in CcO.^23,568 The study of CcO model compounds has led to the development of electrocatalytic systems for the four-electron reduction of O₂ to water, which would have potential applications in fuel cells.^{20,33,569−571} These systems have been reviewed in detail very recently.³³ Common strategies for preparing CcO models include treating ferrous porphyrins with covalently appended copper(I) with O₂ or adding Cu(I) complexes to preformed, discrete ferric peroxo porphyrins. In early attempts, μ-oxo heme–Fe^III–O–Cu^II species were obtained, presumably from the disproportionation of the corresponding complexes.²³ The oxygen of Fe^III-μ-oxo-Cu^II species are basic and can undergo reversible protonation (Scheme 15). A pK_a of 8 has been reported for [(F₈TPP)Fe^III-(OH)-Cu^II(TMPA)]²⁺ (F₈TPP = tetrakis(2,6-difluorophenyl)porphyrinate, TMPA = tris(2-pyridylmethyl)amine).⁵⁷²

Scheme 15. Formation of Fe^III-μ-oxo-Cu^II Complexes by Treating Fe(II) Porphyrin/Cu^I Complex with O₂.

The first well-characterized heme-peroxo-copper adduct was reported by Collman with a complex “capped” porphyrin scaffold with an appended triazacyclononane (TACN) ligand for copper binding (114, Figure 61a).⁵⁷³ This species has a high spin (S = 2) ground state with a ν(O–O) of 758 cm^–1. More recently, a variety of heme-peroxo-copper adducts were characterized and shown to display a rich coordination chemistry. The peroxo moiety can adopt μ-η²:η¹, μ-η¹:η¹, or μ-η²:η² bonding modes depending on the precise copper ligand environment and heme axial ligand (Figure 61, panels b and c).^20,568,574 Karlin has described a μ-η²:η¹ heme-peroxo-copper compound obtained by treating a ferric peroxo complex [(F₈TPP)Fe^III(O₂^2–)]⁻ with a tetracoordinated Cu^II complex Cu^II(TMPA) (115, Figure 61c).^575,576 The compound showed a high spin (S = 2) electronic state with a ν(O–O) of 804 cm^–1. Naruta has successfully obtained the crystal structure of a μ-η²:η¹ heme-peroxo-copper adduct, [(TMP)Fe^III-(O₂)-(5MeTPA)Cu^II)BPh₄ (116, Figure 61c).⁵⁷⁷ The μ-η²:η¹ peroxo core showed two Fe–O bond distances of 2.031(4) and 1.890(6), respectively. Like [(F₈TPP)Fe^III-(O₂)-Cu^II(TMPA)]⁺, [(TMP)Fe^III-(O₂)-Cu^II(5MeTPA)]BPh₄ is EPR silent with an S = 2 spin state resulting from a strong antiferromagnetic coupling between high spin Fe^III (S = 5/2) and Cu^II (S = 1/2).

Figure 61.

(a) A heme-peroxo-copper model compound based on “capped” porphyrin with an appended TACN ligand for copper binding. (b) Different peroxo bonding mode for heme-peroxo-copper complexes. (c) Several representative heme-peroxo-copper model compounds.

The peroxo bonding mode can be regulated by copper coordination and axial ligand of iron porphyrins. For example, addition of a tridentate Cu(II) complex Cu(AN) (AN = 3,3′-iminobis(N,N-dimethylpropylamine)) to [(F₈TPP)Fe^III(O₂^2–)]⁻ afforded a μ-η²:η² peroxo coordination (117, Figure 61c), whereas adding imidazole-type axial ligand to [[(F₈TPP)Fe^III-(O₂)-Cu^II(AN)] shifted the peroxo coordination to the μ-η¹:η¹ mode (118, Figure 61c).^578−581 The end on μ-η¹:η¹ heme-peroxo-copper complex is diamagnetic with a low spin S = 1/2 Fe(III) antiferromagnetically coupled to Cu(II) (S = 1/2).^581,582 Very recently, Karlin et al. reported a [Fe^III-(O₂^2–)-Cu^II(MeTHF)₃]⁺ (MeTHF = 2-methyltetrahydrofuran) complex (119, Figure 61c). The weak coordination of the MeTHF ether oxygen allows for facile ligand exchange, providing ready access to a variety of heme-peroxo-Cu^II species.⁵⁸³ In addition to the heme-peroxo-Cu^II intermediate, isoelectronic Fe^III-superoxide-Cu^I complexes were synthesized using picket fence porphyrins.^584,585 These Fe^III-superoxide-Cu^I complexes are relatively stable at room temperature, suggesting that Cu^I could also play a nonredox role in stabilizing iron dioxygen adducts.

The reactivity of CcO models toward phenols was also explored. Collman et al. found that an Fe^III-superoxide-Cu^I complex could react with phenol both intra- and intermolecularly to afford a ferryl species, a tyrosyl radical, and Cu(II)–OH,^586,587 a process that resembles the transition from CcO A state to P_M state (Figure 62). Solomon and Karlin have tested the reactivity of both low- and high-spin heme-peroxo-copper complexes by treating them with phenols (Figure 63).⁵⁷⁸ While high-spin [(F₈TPP)Fe^III-(O₂^2–)-Cu^II(AN)]⁺ was not reactive toward phenols, the low-spin [(DCHIm)(F₈TPP)Fe^III-(O₂^2–)-Cu^II(AN)]⁺ (DCHIm = dicyclohexyl imidazole) reacted with phenols readily, even at −80 °C, to afford a ferryl intermediate, Cu(II)–OH, and a tyrosyl radical.⁵⁷⁸

Figure 62.

Reactions between Fe^III-superoxide-Cu^I complexes and phenols.

Figure 63.

Comparison of reactions with phenols between low-spin and high-spin Fe^III-peroxo-Cu^II complexes.

The impact of copper complex structure and the nature of the proton source on the O–O bond activation step was recently investigated.⁵⁸⁸ Two Fe^III-peroxo-Cu^II complexes with tridentate and tetradentate copper complexes, respectively (LS-3DCHIm and LS-4DCHIm), were prepared and examined for their reactivity toward different acids or electron donor decamethylferrocene (Fc*) (Figure 64). While LS-4DCHIm alone was not reactive toward Fc*, the addition of 4-nitrophenol led to the formation of an [LS-4DHCIm···4-NO₂-phenol] association complex, which underwent facile O–O bond cleavage via proton-coupled electron transfer from 4-nitrophenol and Fc* to peroxo moiety. Intriguingly, under similar conditions, LS-3DCHIm was not reactive. Further examination showed that LS-3DCHIm could not form the association complex with phenol presumably due to a more compressed and sterically inaccessible peroxo core. Strong acids like triflic acid led to the direct decomposition of both LS-3DCHIm and LS-4DHCIm and the release of hydrogen peroxide, regardless of the steric environment of peroxo group.

Figure 64.

(a) Phenol association will promote the O–O bond cleavage of LS-4DCHIm in the presence of electron donor Fc*. (b) LS-3DCHIm could not form the phenol association complex and is not reactive toward Fc*. DFT calculation showed a more compressed structure of LS-3DCHIm. Adapted from ref (588). Copyright 2017 American Chemical Society.

In addition to small molecule models of CcO, redesigned metalloproteins that mimic the reactivity of CcO represents another powerful approach to study its mechanism.⁵⁸⁹ Lu et al. found that the introduction of two mutations (L29H and F43H) would create a copper binding site in the heme pocket of myoglobin (Figure 65a).⁵⁹⁰ The resulting copper-bound myoglobin (Cu_BMb) showed heme oxygenase reactivity that degrades heme to verdoheme. In 2012, Lu and Wang generated a Cu_BMb with a cross-linked Tyr-His ligand via genetic incorporation of the unnatural amino acid imiTyr at the F33 position (Figure 65). The imiTyr-Cu_BMb exhibited enhanced oxidase reactivity (2/min).⁵⁹¹ In the same year, Lu et al. found that the introduction of a tyrosine mutation to Cu_BMb at F33 also improved O₂ reduction activity (24/min). Intriguingly, the observed oxidase activity is not copper-dependent. A subsequent study showed that a ferric–hydroperoxo species was formed for F33Y–Cu_BMb upon cryoreduction at 77K, whereas a peroxo–ferric state was formed for wild-type Mb.⁵⁹² The X-ray crystal structure of F33Y–Cu_BMb revealed an extensive H-bond network with two water molecules (W1 and W2), in which W1 is H-bonded to H43 and Y33 and W2 is stabilized by H-bonds to W1 and H29 (Figure 65b).⁵⁹³ This hydrogen-bonding network could polarize the O–O bond and facilitate its reduction and protonation. Very recently, Lu and co-workers studied the effect of nonheme metals on oxidase activity of the repurposed Mb.⁵⁹⁴ While Zn-Mb formed a stable ferric-superoxo intermediate, both Cu-Mb and Fe-Mb showed oxidase reactivity with Cu-Mb almost three times more reactive than Fe-Mb. The higher reactivity of Cu^I is attributed to the higher Cu^II/Cu^I potential and an enhanced weakening of O–O bond through dπ-pπ back bonding, as revealed by the longer O–O bond length (1.391 Å) in Cu-Mb than that in Fe-Mb (1.351 Å).

Figure 65.

(a) Crystal structure of oxy-Mb (PDB: 1A6M). (b) Crystal structure and the hydrogen bonding network in the active site of oxy–F33Y–Cu_BMb (PDB: 5HAV).

5. Conclusions and Outlook

The central goal of this review was to survey the range of redox reactions mediated by heme proteins. Parallel developments with model metalloporphyrins and computational approaches are also reviewed and have served both to elucidate the nature of reactive enzymatic intermediates and to discover new chemistry, some of which is unknown in the repertoire of heme protein reactions. Another goal was to provide context for the recent advances with key discoveries from the formative years of the field. This coverage took inspiration from a course we developed at Princeton University called Metals in Biology: From Stardust to DNA. It is interesting to consider that heme and heme proteins apparently evolved long before oxygen was present in the environment in significant quantities. From this perspective, it is clear that oxygen binding and oxygen activation are relatively recent innovations in biology. Accordingly, it is likely that access to hydrogen peroxide allowed heme proteins to be repurposed by exploring the rich chemistry made possible by the ferryl group as well as ferric and iron(IV) hydroxides. In this context, heme and nonheme iron proteins are essentially similar, tied together by the sequential reduction of oxygen to superoxide, hydrogen peroxide, and ultimately water. As we have seen in the examples discussed above, the porphyrin ring can be oxidized at ∼1 V versus NHE to a porphyrin cation radical. Accordingly, there is a direct mechanistic analogy between heme iron and nonheme diiron proteins such as sMMO and AlkB, wherein the second oxidizing equivalent resides on the second iron rather than the porphyrin ligand. There are also nonheme, monoiron systems such as TauD and naphthalene dioxygenase that exploit Fe(IV)/Fe(II) and Fe(III)/Fe(V) couples, respectively, in their substrate oxidations. Studies of heme protein and metalloporphyrin reaction mechanisms and the insights gained from that work have in recent years led to significant developments in drug design. Clopidogrel (Plavix) and similar drugs are effective at lowering the risk of heart attack and stroke due to the nature of their oxidative metabolism by liver cytochrome P450s. Likewise, inhibitors of P450 aromatase proteins, such as anastrazole, are emerging as key agents for the treatment of certain human breast cancers. New methods of C–H functionalization, such as C–H fluorination with fluoride ion-manganese systems, and their applications have offered novel and practical routes to more efficacious drugs and to the development of potential PET imaging agents. More generally, the recent upsurge of radical chemistry in organic synthesis has provided an explosion of methods for generating substrate radicals under mild conditions.^595−597 One can well imagine that a merger of these radical-generating processes (such as photoredox catalysis and electrocatalysis) with metalloenzymes and their model systems could lead to the development of entirely new synthetic transformations that are inaccessible today. These developments are further strengthened by recent advancements in biocatalysis for non-natural transformations, in which existing enzymes or proteins can be adapted and repurposed to catalyze abiological chemistry via directed evolution.^598,599 We envision that an orchestration between metalloenzymes/metal–oxygen systems and these emerging fields in catalysis have the potential to push the frontiers of biocatalysis and organic synthesis as well as the conceptual underpinnings of the field that continue to provide fresh air.

Glossary

Abbreviations

AOS

allene oxide synthase

APX

ascorbate peroxidase

CcO

cytochrome c oxidase

Cld

chlorite dismutase

Cpd 0

compound 0

Cpd I

compound I

Cpd II

compound II

DES

divinyl ether synthase

HAT

hydrogen atom transfer

Hb

hemoglobin

HRP

horseradish peroxidase

Mb

myoglobin

OAT

oxygen atom transfer

TPFPBr₈

2,3,7,8,12,13,17,18-octabromo-5,10,15,20-tetrakis(pentafluorophenyl)porphyrin

TpivPP

5,10,15,20-tetra(α,α,α,α-O-pivalamidophenyl)porphyrin

TPP

5,10,15,20-tetraphenylporphyrin

TMP

5,10,15,20-tetramesitylporphyrin

TPFPP

5,10,15,20-tetrakis(pentafluorophenyl)porphyrin

TDMPP

meso-tetrakis(2,6-dimethylphenyl)porphyrin

TTMPP

meso-tetrakis(3,4,5-trimethoxypbenyl)porphyrin

4-TMPyP

meso-tetrakis(4-N-methylpyridyl)porphyrin

2-TMPyP

meso-tetrakis(2-N-methylpyridyl)porphyrin

TF₄TMAP

meso-tetrakis(2,3,5,6-tetrafluoro-N,N,N-trimethyl-4-aniliniumyl)porphinato dianion

TFMPC

5,10,15-tris(3,5-trifluoromethylphenyl)corrolato trianion

PHM

peptidylglycine α-hydroxylating monooxygenase

TDO

tryptophan 2,3-dioxygenase

IDO

indoleamine 2,3-dioxygenase

Biographies

Xiongyi Huang received his B.S. degree in chemistry in 2010 from University of Science and Technology of China (USTC), where he performed undergraduate research in computational chemistry under the guidance of Prof. Yao Fu and Prof. Jing Shi. He received his Ph.D. in 2016 at Princeton University with Prof. John T. Groves. His Ph.D. work mainly focused on developing new synthetic strategies for the construction of aliphatic C–halogen, C–F, and C–N bonds using biomimetic manganese catalysts. While at Princeton, Xiongyi was a recipient of the HHMI International Student Research Fellowship (2013–2015) and the 2016 Chinese Government Award for Outstanding Students Abroad. He is currently an NIH Ruth L. Kirschstein postdoctoral fellow in the laboratory of Prof. Frances H. Arnold at Caltech, where he works on developing new enzymatic reactions via directed evolution.

John T. Groves is the Hugh Stott Taylor Chair of Chemistry at Princeton University. He received an S.B. in Chemistry at MIT and a Ph.D. in Organic Chemistry at Columbia University, where he worked with Ronald Breslow. Groves’ early academic career was at the University of Michigan where he was director of the Michigan Center for Catalysis and Surface Science and received the Phi Lambda Upsilon Award for Teaching and Leadership. Moving to Princeton University as Professor of Chemistry, he became Department Chair in 1988. He has received the Arthur C. Cope Scholar Award, the Alfred Bader Award in Bioorganic and Bioinorganic Chemistry, the Ira Remsen Award and the Award in Inorganic Chemistry, all from the American Chemical Society. He was co-Chair of the ACS-NSF Committee and Workshop on the Molecular Basis of Life Processes. He is a fellow of the American Academy of Arts and Sciences, the American Association for the Advancement of Science, the Japan Society for the Promotion of Science, and the Royal Society of Chemistry. Groves is a member of the U.S. National Academy of Sciences. International awards include the Grand Prix, Maison de la Chimie (Paris), the Frontiers in Biological Chemistry Award from the Max Planck Institute (Mülheim), and the Hans Fischer Award in Porphyrin Chemistry. Groves’ research has focused on the roles of metal–oxo species in chemical and biological oxidations.

The authors declare no competing financial interest.