O₂ activation by non-heme iron enzymes

Abstract

The non-heme Fe enzymes are ubiquitous in nature and perform a wide range of functions involving O₂ activation. These had been difficult to study relative to heme enzymes; however, spectroscopic methods have now been developed that provide significant insight into the correlation of structure with function. This Current Topics article summarizes both the molecular mechanism these enzymes use to control O₂ activation in the presence of cosubstrates and the oxygen intermediates these reactions generate. Three types of O₂ activation are observed. First, non-heme reactivity is shown to be different from heme chemistry where a low-spin Fe^III-OOH non-heme intermediate directly reacts with substrate. Also, two subclasses of non-heme Fe enzymes generate high-spin Fe^IV=O intermediates that provide both σ and π frontier molecular orbitals that can control selectivity. Finally, for several subclasses of non-heme Fe enzymes, substrate binding to the Fe^II site leads to the one electron reductive activation of O₂ to an Fe^III-superoxide capable of H-atom abstraction and electrophilic attack.

Graphical abstract

Introduction

The non-heme iron enzymes perform a wide range of critical biological functions^1,2 that include the regulation of hypoxia,³ demethylation of DNA,⁴ antibiotic and natural product biosyntheses,^5,6 and bioremediation,^7,8 are related to disease states,^9,10 and include the anticancer drug bleomycin.¹¹ On a molecular level these functions involve electrophilic aromatic substitution,¹² H-atom abstraction,¹³ hydroxylation,¹⁴ mono-¹⁵ and dioxygenation,¹⁶ ring closure¹⁷ and expansion,¹⁸ ring cleavage,¹⁹ and halogenation.²⁰ As shown in Table 1, the mononuclear non-heme Fe enzymes divide into two classes dependent on whether an oxidized high-spin Fe^III site activates singlet organic substrates for the spin-forbidden reaction with triplet O₂ or a reduced high-spin Fe^II site activates O₂.

Table 1.

The classes of oxygen- and organic substrate-activating mononuclear non-heme Fe enzymes. Adapted from ref 2.

The organic substrate activating Fe^III sites have been considered in ref 21; here we focus on the dominant class where Fe^II activates O₂. These enzymes generally have a facial triad of three protein-derived ligands (2 His and one carboxylate),²² with additional H₂O-derived ligands completing the Fe^II coordination, although the halogenases have the carboxylate replaced by a halide²³ and several enzymes (including diketone-cleaving dioxygenase and cysteine dioxygenase) have the carboxylate replaced by a third His ligand.²⁴

The Fe^II enzymes in Table 1 are subdivided based on whether a cosubstrate or cofactor is required and the nature of the primary substrate.² In the pterin and α-ketoglutarate (αKG) dependent enzymes, the exogenous cosubstrate donates two electrons via the reactions given in Equation 1, while in the Rieske dioxygenases an endogenous 2Fe2S Rieske center donates one of the electrons for O₂ activation.

$$\text{Reduced pterin}+{}^{*}{\mathrm{O}}_2\to 2{\mathrm{e}}^{-}+{}^{*}4\mathrm{O}\mathrm{H}\text{ pterin}\to \text{oxidized pterin}+{}^{*}{\mathrm{H}}_2\mathrm{O}$$

$$\mathrm{\alpha }\mathrm{K}\mathrm{G}+{}^{*}{\mathrm{O}}_2\to 2{\mathrm{e}}^{-}+{}^{\mathrm{*}}\text{succinate}+\mathrm{C}{\mathrm{O}}_2$$ Equation 1

The next two categories of non-heme Fe enzymes in Table 1 only involve a substrate, where in the extradiol dioxygenases it is a catechol capable of two-electron oxidation to a quinone as part of the catalytic cycle, while in a fifth category the substrates are not intrinsically redox active. Finally, bleomycin is listed here as the Fe^II site of this glycopeptide antibiotic binds O₂ and is further reduced by one exogenous electron to generate activated bleomycin,²⁵ the intermediate kinetically competent to cleave DNA by H-atom abstraction in its anticancer function.

The high-spin Fe^II active sites in these enzymes had been spectroscopically inaccessible due to their lack of the π to π* transitions of heme enzymes.²⁶ Thus a new spectroscopic methodology was developed (variable-temperature variable-field magnetic circular dichroism, VTVH MCD)¹ that elucidated a general mechanistic strategy utilized by many of the different non-heme Fe^II enzymes in Table 1 that is summarized in Section 1. O₂ intermediates have thus far been clearly observed in bleomycin,²⁵ the αKG and pterin dependent enzymes,^27,28 and the extradiol dioxygenases.^29–31 A focus of our research on these intermediates has been to use nuclear resonance vibrational spectroscopy (NRVS) to define geometric structure and VTVH MCD to define electronic structure and their contributions to reactivity. The results of these studies on the Fe^III-hydroperoxide in activated bleomycin, the Fe^IV=O in the αKG and pterin dependent enzymes, and the possible involvement of an Fe-O₂^•− intermediate in several subclasses in Table 1 are summarized in subsequent sections. These studies emphasize: 1) the difference in reactivity between non-heme and heme Fe enzymes, 2) the nature of the frontier molecular orbitals (FMOs) of Fe^IV=O S = 2 and how these can control the type of reactivity, and 3) substrate contributions to the thermodynamically difficult one-electron reduction of O₂ to form Fe-superoxo intermediates.

1. General mechanistic strategy of non-heme Fe^II enzymes

Non-heme high-spin Fe^II active sites do not have intense absorption features, but do have weak d → d, ligand field (LF) transitions, which are characteristic of the coordination environment. Their low intensity and near IR energy region (12,000 – 5000 cm⁻¹) make the study of LF transitions in metalloproteins inaccessible by absorption spectroscopy. However, their ground states have S = 2 and thus are paramagnetic, which leads to high intensity in MCD spectroscopy at low temperature. As shown in Figure 1A, the low temperature (LT) MCD spectra of Fe^II sites is characteristic of coordination number, with six-coordinate (6C) sites having two transitions at ~10,000 cm⁻¹ split by ~2000 cm⁻¹, 5C square pyramidal Fe^II sites having transitions at >~10,000 and ~5000 cm⁻¹, 5C trigonal bipyramidal having transitions at <10,000 and <5000 cm⁻¹, and 4C distorted tetrahedral having LF transitions only at low energy (5000–7000 cm⁻¹) in LT MCD. The temperature and field dependence of the MCD data for these provide significant additional insight into the ground state of these sites, and the reader is referred to ref 1 for details.

Figure 1.

Panel A: Representative low temperature MCD spectra for (from L to R) 6C octahedral, 5C square pyramidal, 5C trigonal bipyramidal, and 4C distorted tetrahedral non-heme Fe (NHFe) complexes. Panel B: MCD spectra for resting PAH (black) overlaid with spectra for (from L to R) PAH-L-Phe (blue), PAH-pterin (green), and PAH-L-Phe-pterin (red). Only coordination of both cosubstrates leads to a 5C site. Adapted from ref 2.

Figure 1B demonstrates how LT MCD provided significant insight into the molecular mechanism of the Fe^II active site in the pterin dependent enzyme phenylalanine hydroxylase (PAH, done in collaboration with Prof. John Caradonna).³² The resting Fe^II site shows two transitions at ~10,000 cm⁻¹ split by ~2000 cm⁻¹ (black MCD spectrum in Figure 1B), demonstrating that it is 6C. Addition of the primary substrate, phenylalanine, only slightly perturbs the spectrum (blue), while addition of pterin has no effect on the MCD spectrum (green, which demonstrates that the pterin cosubstrate does not bind to the Fe^II center). Importantly, upon binding both phenylalanine and pterin, there is a dramatic change in the MCD spectrum showing one band at ~10,000 cm⁻¹ with one at ~5000 cm⁻¹ (red), indicating that the Fe^II has now gone 5C, opening a coordination position on the iron for O₂ activation.

Parallel experiments to those described above have been done on members of all the six subclasses of non-heme Fe^II enzymes in Table 1 with parallel results^32–37 that have led to the general mechanistic strategy utilized by the non-heme Fe^II enzymes shown in Figure 2.

Figure 2.

The general mechanistic strategy for NHFe enzymes, where a coordination position for O₂ is only available when all cosubstrates are bound to the active site. Adapted from ref 2.

The resting Fe^II site, the site in the presence of primary substrate, or the site with cosubstrate bound are 6C, coordinatively saturated and relatively stable in the presence of O₂. This coordination saturation is important in the presence of the reduced cosubstrate, as all electrons are present and the site is primed for O₂ reaction. Importantly, the presence of all cosubstrates opens a coordination position of the Fe^II for O₂ activation, enabling coupled turnover (i.e. one molecule of product for every cosubstrate).

The need for both cosubstrate and primary substrate to be present to form an open coordination position is a key advantage of the general mechanistic strategy defined by the MCD data, as it minimizes cosubstrate turnover without primary substrate. A kinetic model for the αKG dependent dioxygenases is given in Scheme 1.

Scheme 1.

Kinetic model including the general mechanistic strategy in the αKG dependent dioxygenases

The enzyme first reversibly binds cosubstrate (C) or primary substrate (S) to form 6C complexes, important in preventing activation of O₂ by the EC complex in the absence of primary substrate. The binding of both cosubstrates results in a 5C ECS complex that can now react with O₂. In TauD, binding of the primary substrate significantly enhances the binding affinity of αKG, reducing K_D by an order of magnitude.³⁸ Although it is unclear whether O₂ binds reversibly before the irreversible reaction in the ECS complex,³⁹ saturation behavior with increasing concentrations of O₂ is observed.⁴⁰ The only intermediate observed in the reaction with O₂ is the Fe^IV=O complex (I) that results from αKG decarboxylation, and its formation and decay to the product complex (EP) have been observed in multiple αKG and pterin dependent enzymes.^{27,28,41–43} The general mechanistic strategy has the advantage of requiring primary substrate to be present when I is formed, which minimizes unproductive decay channels that include self-hydroxylation and auto-oxidation reactions that inactivate the enzyme.⁴⁴ Indeed, KIE experiments that increase the accumulation of I show a marked increase in these unproductive decay pathways.²⁷ Note that there had been a “booby-trapped” mechanism proposed for DAOCS where the EC complex was thought to react with O₂ to generate a stable peroxy intermediate.⁴⁵ However, a number of calculations and experiments have demonstrated that the peroxy is at higher energy relative to the Fe^IV=O,^46,47 which is the only intermediate observed.²⁷ Also it has been shown that DAOCS does, in fact, form a ternary complex in line with the general mechanistic strategy presented above.⁴⁸

Insight has been derived into the molecular basis for the 6C → 5C conversion in the presence of cosubstrate and primary substrate and into the role of the facial triad and its second sphere residues in this conversion. As shown in Figure 2, the 6C → 5C conversion with primary substrate and cosubstrate involves loss of a H₂O ligand. The identity of the lost ligand upon binding of both cosubstrates was determined by crystallography for the αKG dependent enzymes⁴⁹ and with EXAFS for the pterin dependent NHFe enzymes.⁵⁰ The factors driving water ligand loss were evaluated in the αKG dependent enzyme, factor inhibiting HIF-1 (FIH) in collaboration with Prof. Mike Knapp. In the presence of only cosubstrate, electron donation by the αKG, which coordinates the Fe^II in a bidentate mode, weakens the Fe^II-OH₂ bond based on LF changes from VTVH MCD.⁵¹ (This strong donor interaction is also present in the extradiol dioxygenases that coordinate a catecholate substrate bidentate, as well as in the enzymes that react with non-redox active substrates like isopenicillin-N synthase, IPNS.) However, the water remains bound due to hydrogen bonding between the unbound oxygen of the monodentate carboxylate ligand of the facial triad and the water. Further addition of primary substrate, which does not directly coordinate Fe^II but binds in the active site pocket, destabilizes the 6C site due to steric interactions with the coordinated water. In FIH, primary substrate binding also stabilizes the 5C site through hydrogen bonding between the substrate and the unbound oxygen of the carboxylate thereby weakening the carboxylate-water H-bond. These effects are shown in Figure 3.

Figure 3.

Schematic showing the H-bonding and steric contributions to water loss and 5C site formation in FIH upon substrate (C- terminal transactivation domain, CAD) binding. Adapted from ref 51.

Second sphere residues are also important in preserving the reactivity of the 5C site with O₂. In FIH, a second sphere arginine (R238) maintains the monodentate coordination of the facial triad carboxylate ligand by H-bonding to the unbound oxygen. DFT calculations predict that removal of this H-bonding interaction would result in the facial triad carboxylate binding bidentate,⁵¹ which would eliminate the open coordination position for O₂ binding. Indeed, activity in the R238M FIH mutant is reduced to 5% of WT.⁵² Importantly, in enzymes that coordinate a cosubstrate or primary substrate bidentate to the Fe^II (αKG dependent enzymes and the extradiol dioxygenases), a second sphere residue is present that provides an H-bond to stabilize a monodentate carboxylate, critical in maintaining an open coordination position for O₂ activation. Note that in the NHFe enzymes that do not have the bound primary substrate or cosubstrate occupy two coordination positions (pterin dependent and Rieske dioxygenases), there is no second sphere H-bond and the carboxylate goes bidentate still leaving open coordination positions for O₂ activation, as shown in Figure 4.

Figure 4.

Schematic comparison of facial triad coordination for the classes of NHFe enzymes that have a second-sphere residue hydrogen bonding to the coordinated carboxylate (αKG dependent enzymes, shown at left, and the extradiol dioxygenases), leading to monodentate coordination, and those without a hydrogen bonding residue (pterin dependent enzymes, shown at right, and the Rieske dioxygenases), which leads to bidentate carboxylate coordination.

As mentioned above, three types of oxygen intermediates have been identified in the different subclasses of O₂ activating non-heme Fe enzymes. In the sections below, we consider each of these and its activation in terms of frontier molecular orbital (FMO) theory.

2. The low spin Fe^IIIOOH intermediate in bleomycin: non-heme vs heme reactivity

While activated bleomycin (ABLM) had been thought to be a low-spin Fe^III-hydroperoxide, a computational study raised issues that lead to its further elucidation.⁵³ Normally, resonance Raman spectroscopy would be used to identify the nature of an axial ligand; however, ABLM is unstable in the laser beam,⁵⁴ so NRVS was employed.⁵⁵ This experiment requires a third-generation synchrotron where the⁵⁷Fe nuclear Mössbauer transition is scanned to higher energy to observe vibrational sidebands.^56,57 The resultant NRVS spectrum gives intensity at a given energy reflecting the amount of Fe motion in a vibration of the metalloenzyme at that energy. The NRVS spectra of low-spin Fe^III BLM and ABLM are given in Figure 5, with DFT simulations of the NRVS data given below for the structure that best agrees with the data.

Figure 5.

A: NRVS spectra (top) and simulations (bottom) for low-spin ferric BLM showing an axial hydroxide bound to the Fe. B: NRVS data (top) and simulations (bottom) for ABLM. The three-peak pattern in the data can only be modeled by a low-spin Fe^III-OOH structure. C: Schematic depictions of vibrations assigned in B. Adapted from ref 55.

Ferric BLM shows a feature at 567 cm⁻¹ that shifts with H₂¹⁸O. The high energy of this stretch of the water derived ligand shows that low-spin Fe^III BLM has a hydroxide-Fe^III axial ligand. The intense feature at 400 cm⁻¹ is a degenerate pair of transaxial bends. In going to ABLM (Figure 5B), these split by 40 cm⁻¹ and there is a low-energy 328 cm⁻¹ peak. This can only be simulated with a low-spin Fe^IIIOOH where the 328 cm⁻¹ feature is the OOFe bend and this couples to the in-plane transaxial bend (Figure 5C), which leads to the split spectral features.

Thus, ABLM is a low-spin Fe^IIIOOH, which is the non-heme Fe equivalent of compound 0 in p450. As shown in equation 2A, it has been determined in p450 that the Fe^IIIOOH protonatively cleaves to generate compound I,^58,59 an Fe^IV=O with a one electron oxidized porphyrin ring.

$$\mathrm{F}{\mathrm{e}}^{\mathrm{I}\mathrm{I}\mathrm{I}}(\text{Por})(\mathrm{S}\mathrm{C}{\mathrm{H}}_3)-\mathrm{O}\mathrm{O}\mathrm{H}+{\mathrm{H}}^{+}\to \mathrm{F}{\mathrm{e}}^{\mathrm{I}\mathrm{V}}\mathrm{O}(\mathrm{P}\mathrm{o}{\mathrm{r}}^{•+})+{\mathrm{H}}_2\mathrm{O}\hspace{1em}\mathrm{\Delta }\mathrm{G}=-58\mathrm{k}\mathrm{c}\mathrm{a}\mathrm{l}/\text{mol}$$ Equation 2A

$$\mathrm{F}{\mathrm{e}}^{\mathrm{I}\mathrm{I}\mathrm{I}}(\mathrm{B}\mathrm{L}\mathrm{M})-\mathrm{O}\mathrm{O}\mathrm{H}+{\mathrm{H}}^{+}\to \mathrm{F}{\mathrm{e}}^{\mathrm{V}}\mathrm{O}(\mathrm{B}\mathrm{L}\mathrm{M})+{\mathrm{H}}_2\mathrm{O}\hspace{1em}\mathrm{\Delta }\mathrm{G}=+13\mathrm{k}\mathrm{c}\mathrm{a}\mathrm{l}/\text{mol}$$ Equation 2B

$$\mathrm{F}{\mathrm{e}}^{\mathrm{I}\mathrm{I}\mathrm{I}}(\mathrm{B}\mathrm{L}\mathrm{M})-\mathrm{O}\mathrm{O}\mathrm{H}+\mathrm{R}\mathrm{H}\to \mathrm{F}{\mathrm{e}}^{\mathrm{I}\mathrm{V}}\mathrm{O}(\mathrm{B}\mathrm{L}\mathrm{M})+{\mathrm{R}}^{•}+{\mathrm{H}}_2\mathrm{O}\hspace{1em}\mathrm{\Delta }\mathrm{G}=-7\mathrm{k}\mathrm{c}\mathrm{a}\mathrm{l}/\text{mol}$$ Equation 2C

For p450, this is a highly exergonic reaction calculated to be −58 kcal/mol for a proton from solvent.⁶⁰ If the same calculation is done on ABLM (equation 2B), the reaction is 71 kcal/mol less favorable and is endergonic by at least 13 kcal/mol with the same source of proton.⁶¹ This energy difference reflects the fact that it is much harder to oxidize a non-heme relative to a heme ligand where the hole is delocalized over the porphyrin ring and further stabilized by its negative charge.

Alternatively, if the low-spin Fe^IIIOOH is directly reacted with the H-C bond of a substrate modeling DNA, the reaction is exergonic by 7 kcal/mol (equation 2C).⁶² This direct reaction of the Fe^III-hydroperoxide with DNA was studied experimentally,⁶² using CD to monitor the kinetics of the decay of ABLM. It was determined that DNA accelerated the rate of decay of ABLM and defined it as having a small primary and large secondary KIE. These experimental data were computationally modeled and found to reflect the transition state in Figure 6A, which is late in O-O cleavage and early in H-atom abstraction. This leads to consideration of the FMO of a low spin Fe^IIIOOH, Figure 6B, which is the peroxo σ* orbital. Upon elongation of the O-O bond to the late transition state, the σ bonding is essentially lost and the two electron holes in the σ* polarize, one on the distal O generating a hydroxyl directed toward the C-H bond of the substrate, and the other to the proximal O generating an Fe^IV=O capable of a second H-atom abstraction, also with a low barrier, leading to double strand cleavage, important in the anticancer activity of the drug.⁶⁴

Figure 6.

A: Transition state for the direct H-atom abstraction of a proton by ABLM, which is late in O-O cleavage and early in C-H abstraction. B: The hydroperoxo σ* FMO of the low-spin ABLM reactant (left), which polarizes into a hydroxyl radical and Fe^IV=O at the transition state (right). Adapted from ref 63.

3. Fe^IV=O Intermediates: frontier molecular orbital control of reactivity

Iron-oxo intermediates have been trapped and defined in a series of αKG and pterin dependent enzymes (syringomycin biosynthesis enzyme 2 [SyrB2],²⁰ taurine dioxygenase [TauD],²⁷ prolyl 4-hydroxylase [P4H],⁴³ CytC3,⁴¹ tyrosine hydroxylase [TH],²⁸ and PAH]⁴²). From EXAFS on the TauD intermediate J, the Fe-O bond length is 1.62 Å,⁶⁵ and from resonance Raman on this intermediate the ν_Fe-O is 821 cm⁻¹.⁶⁶ Mössbauer studies have been done on all the intermediates listed above giving δ = 0.22 – 0.30 mm/s and |ΔE_Q| = 0.76 – 1.27 mm/s, and importantly, from magnetic Mössbauer, the ground state is high-spin with S = 2.^67,68 In order to obtain detailed geometric structural insight into an Fe^IV=O intermediate in an enzyme system, NRVS was done on the halogenase SyrB2 in collaboration with Professors Krebs and Bollinger, where both the Cl⁻ and Br⁻-bound forms (replacing the carboxylate of the facial triad) could be studied, enabling a perturbation of the spectral features.⁶⁹ From the data in Figure 7, the NRVS spectrum of SyrB2 has three peaks where the intensity shifts from the high to the low energy region with Cl⁻ → Br⁻.

Figure 7.

Top: NRVS spectra for SyrB2-Cl (green) and SyrB2-Br (red). Middle: DFT NRVS simulations derived from the structure that best reproduces the experimental data. Bottom: Experimentally derived structure of the Fe^IV=O intermediate of SyrB2, which has a 5C trigonal bipyramidal geometry and an ~C₃ axis. Adapted from ref 69.

This is an interesting mass effect where the heavy Br⁻ shifts iron motion into the low energy vibrations. These data were correlated to a range of possible structural models using experimentally calibrated DFT methods, and one structure reproduced the three-peak pattern and shift of intensity with Br⁻. This is a 5-coordinate trigonal bipyramidal structure with the Fe-O bond along the approximate C₃ axis of this intermediate (Figure 7, bottom).

The geometric and electronic structure of this enzyme intermediate⁷⁰ are very similar to those of a structurally defined model complex, [Fe^IV(O)(TMG₃tren)]²⁺, of Que et al.^71,72 Variable temperature MCD spectroscopy on this model (and the enzyme intermediate) then provided detailed electronic structural insight, and in particular, combined with calculations, defined the FMOs available for electrophilic reactivity.⁷³ The correlation of the MCD data to the absorption data in Figure 8A revealed and assigned key excited states which correspond to the transitions in Figure 8B: the dπ* → dσ* ligand field transitions involving two key FMOs and an oxo π to dπ ligand to metal charge transfer (CT) transition defining a third oxo π* FMO.

Figure 8.

A: Absorption (top) and MCD data (bottom) for [Fe^IV(O)(TMG₃tren)]²⁺. B: LF and CT transitions of [Fe^IV(O)(TMG₃tren)]²⁺ C: FMOs of [Fe^IV(O)(TMG₃tren)]²⁺ at the transition state demonstrating Fe^III-O^•− character and the different channels for reactivity. Adapted from ref 73.

Multireference CASPT2 calculations were correlated to the spectral data in Figure 8A and extended to the elongated Fe-O transition state (1.81 – 1.84 Å) associated with its reactivity. From Figure 8C, these studies reveal three very reactive FMOs: a σ* FMO oriented along the Fe-O bond with significant oxo p_z σ* character for reacting along this bond and two π* FMOs with oxo pπ character activated for reaction perpendicular to the Fe-O bond. Further, it was found that near the transition state all three FMOs gain a great deal of oxo hole character, effectively becoming Fe^III-oxyl sites for enhanced reactivity.

These π and σ FMOs and their orientation dependence, present only in S = 2 Fe^IV=O non-heme intermediates, provide an electronic mechanism to control reactivity. Two systems that appear to utilize these FMOs to enable reactivity are presented below.

(4-hydroxyphenyl) pyruvate dioxygenase (HPPD) and (4-hydroxy) mandelate synthase (HmaS) both react with the same substrate, (4-hydroxyphenyl) pyruvate (HPP), but with different outcomes of the reaction: electrophilic attack on the ring in the former and H-atom abstraction of the benzylic H in the latter. Spectral studies have shown that HPP (where the α-keto acid is covalently linked to the substrate) binds to the Fe with different conformations of the ring.^74,75 Extension to the decarboxylated Fe^IV=O intermediate in Figure 9 showed that, for HPPD, the ring can be oriented for σ* FMO attack on its π electron cloud, while for HmaS with the ring oriented away from the site the Fe=O is still reactive in H-atom abstraction using its π* FMO, perpendicular to the Fe-O bond.⁷⁶

Figure 9.

FMOs of the Fe^IV=O intermediate in HPPD (left) and HmaS (right), demonstrating that different substrate orientations lead to different reactivities using different FMOs. Adapted from ref 76.

For the halogenase SyrB2 mentioned above, reaction of the native substrate leads to selective halogenation, while reaction with alternative substrates can lead to hydroxylation,⁷⁷ which is thermodynamically favored. Following the reaction coordinate in Figure 10, O₂ binding to the 5C Fe^II activates it for nucleophilic attack on the carbonyl C, the CO₂ is lost, and the O-O bond cleaves. For the native substrate this leads to the 5C ~C₃ trigonal bipyramidal Fe^IV=O structure as found from the NRVS data in Figure 7, importantly with the Fe-O bond perpendicular to the H-C bond of the substrate.⁶⁹ From this orientation, the Fe^IV=O is still capable of H-atom abstraction with a reasonable barrier due to its active π* FMO. This leads to an Fe^III-OH first intermediate with the OH group oriented away from the carbon radical of the substrate, but with the halide well-oriented for rebound halogenation. Thus the S = 2 of the Fe^IV=O intermediate in NHFe enzymes has both π and σ FMOs (in contrast to the Fe^IV=O intermediate in heme enzymes with an S = 1 ground state,⁷⁸ which has only a π FMO⁷⁹) that can provide important flexibility and control of oxo reactivity in biology.

Figure 10.

Reaction coordinate for O₂ activation of SyrB2, leading to a 5C trigonal bipyramidal Fe^IV=O intermediate (as in Figure 7) with an Fe-O bond perpendicular to the substrate C-H bond. The contour gives the π* FMO used for reactivity. Adapted from ref 70.

5. Fe^III-O₂^•− intermediates: one electron reductive activation of O₂

High-spin Fe^III-O₂^•− species have been invoked as the intermediate that reacts directly with substrate in three subclasses of NHFe enzymes: the non-redox active substrate enzymes (that include IPNS), the extradiol dioxygenases, and recently, the Rieske dioxygenases.^80–82 Of these, direct experimental evidence for the formation of a Fe^III-O₂^•− intermediate has only been reported in the extradiol dioxygenase homoprotocatechuate 2,3-dioxygenase (HPCD)^29,81 and, recently, in IPNS.⁸³ Defining the factors that enable the one-electron reduction of O₂ by a high-spin Fe^II site is fundamental in understanding the reactivities of these subclasses of O₂-activating mononuclear NHFe enzymes.

It is important to note that, from MCD and crystallography, the resting Fe^II active site is five-coordinate square pyramidal (facial triad plus two H₂O ligands) in two type I extradiol dioxygenases, 2,3-dihydroxybiphenyl 1,2-dioxygenase (DHBD)^84,85 and catechol 2,3-dioxygenase (CTD).^86,87 Thus, even in the absence of substrate, the resting active site is coordinatively unsaturated, yet these resting Fe^II enzymes are relatively stable in the presence of O₂. A computational study gave results consistent with this observation, finding that O₂ binding to the resting 5C facial triad site of DHBD was highly endergonic.³⁵ An experimentally calibrated functional (BP86 with 10% Hartree-Fock exchange) gives an endergonic free energy of 19 kcal/mol (Figure 11 right, blue).³⁶ Three factors contribute to the reaction of O₂ with the 5C facial triad being unfavorable. First, the one-electron reduction potential of O₂ to superoxide is low (−0.16 V in aqueous solution),⁸⁸ and the Fe^II/Fe^III reduction potential for a facial triad is high (+0.19 V for PAH, which represents a lower limit as PAH is 6C)⁸⁹ due to the poor donor ability of the facial triad ligation. In addition, binding of O₂ to an open coordination site leads to an entropic penalty of 10–12 kcal/mol at room temperature.⁹⁰ Finally, the Fe^III-O₂^•− bond is relatively weak. This can be seen in the computational comparison of the energetics of the binding of O₂ and NO to a facial triad site (Figure 11). The one electron reduction of NO to NO⁻ (red line center, Figure 11) is more difficult than that of O₂ to O₂⁻ (blue line) by ~5 kcal/mol, yet NO binding is more favorable by 25.9 kcal/mol due to the strong Fe-N bond (length = 1.74 Å) relative to the weaker Fe-O bond (1.87 Å).³⁶

Figure 11.

Comparison of the free energies for the one-electron reduction of NO/O₂ by Fe^II (middle) and for the formation of a Fe^III-NO/O₂ complex (right) for a resting facial triad and for substrate-bound IPNS. Sulfur donation stabilizes formation of an Fe^III-O₂⁻ complex by 25.9 kcal/mol relative to a resting facial triad. Adapted from ref 36.

In all of the three subclasses of NHFe enzymes invoking superoxide intermediates, substrate binding to the active site turns on the O₂ reactivity. The origin of this substrate activation was evaluated in collaboration with Prof. John Lipscomb for a member of the non-redox substrate enzyme subclass, IPNS, where the substrate bound to Fe cannot provide additional electrons.³⁶ IPNS binds its substrate, δ-(l-α-aminoadipoyl)-l-cysteinyl-d-valine (ACV), monodentate through its cysteinyl S and catalyzes the formation of isopenicillin-N through an initial H-atom abstraction step.⁹¹ DFT calculations found that O₂ binding to the 5C IPNS-ACV site was now exergonic by 7 kcal/mol (Figure 11 right, purple line), relative to +19 kcal/mol for O₂ binding to the resting facial triad site (Figure 11, blue line). From MCD spectroscopy on the stable nitrosyl analogue of the putative Fe^III-ACV-O₂^•− complex, a new low energy transition is present that is assigned as a substrate thiolate π to Fe^III 3d charge transfer. This reflects strong sulfur donation that stabilizes the Fe^III-ACV complex relative to the resting Fe^III facial triad (Figure 11, middle), lowering the reduction potential and enabling the formation of an Fe^III-ACV-O₂^•− complex by 25.9 kcal/mol. This superoxide Fe^III species has recently been identified in IPNS by Mössbauer spectroscopy.⁸³ Computationally, this Fe^III-superoxo intermediate is found to be activated by its π* FMO (Figure 12) to perform H-atom abstraction from the β methylene C of the substrate.³⁶

Figure 12.

FMO of Fe^III-ACV-O₂. Inset: Schematic of the unoccupied FMO, showing its favorable orientation for H-atom abstraction from the substrate β methylene carbon. Adapted from ref 36.

In the extradiol dioxygenases, understanding the one-electron activation of O₂ is complicated by the presence of the redox active catecholate substrate. In these enzymes, an alkylperoxo bridge to the substrate is formed after O₂ is activated through binding to the Fe^II site.⁹² Due to the non-innocence of the catecholate, the source of the electron for O₂ activation and thus the electronic structure of the species that attacks the aromatic ring is an open issue. On the basis of crystallographic⁸¹ and computational⁹³ studies, it was proposed that the substrate activates O₂ to the superoxy level, leading to a Fe^II-semiquinone-O₂^•− species (Scheme 2A) that would then radical couple to form the alkylperoxo bridged species. Another possibility is that the strong charge donation from the catecholate tunes the donor potential of the Fe^II in a manner similar to that defined above for IPNS, leading to a Fe^III-catecholate-O₂^•− species that would attack the ring (Scheme 2B). This is supported by computational study⁹⁴ and by the isolation of a Fe^III-catecholate-superoxo species in the H200N variant of HPCD using a slow substrate,²⁹ though this intermediate is not active in extradiol cleavage. Another possibility is that both Fe^II and substrate donate an electron to O₂, leading to a Fe^III-semiquinone-peroxy or hydroperoxy active species (Scheme 2C), supported by Mössbauer and EPR data that suggest the formation of such an intermediate in the H200C and H200N mutants of HPCD that are active in extradiol cleavage.^30,31 More experimental data are needed to evaluate the nature of the active species in the extradiol dioxygenases and define the role of substrate in their activation of O₂.

Scheme 2.

Possible electronic structures for the activated Fe-O₂ species in the extradiol dioxygenases.

In the Rieske dioxygenases (RDOs), which catalyze the cis-dihydroxylation of aromatic rings, the electrons necessary for the reaction are provided by the catalytic mononuclear NHFe Fe^II site and by a 2Fe2S Rieske cluster ~12 Å away capable of one electron donation.⁹⁵ In a peroxide shunt reaction of the RDO benzoate 1,2-dioxygenase with both the mononuclear Fe and Rieske sites oxidized and substrate bound, a high-spin Fe^III-(hydro)peroxy intermediate was isolated,⁹⁶ and has also been observed by crystallography.^97,98 However, the involvement of this peroxy intermediate in the O₂ activation mechanism has been called into question by a substrate dependence of the rate of Rieske oxidation, implying that an Fe^III-O₂^•− species that reacts with substrate to form an Fe^III-peroxo-aryl radical species is formed prior to oxidation of the Rieske site.⁸² In the RDOs, substrate does not bind directly to the Fe site (and thus cannot tune its reduction potential), and it is not redox active. It is possible that the bidentate coordination of the facial triad carboxylate that occurs in this class of enzymes (vide supra) could help stabilize the formation of an Fe^III-O₂^•− intermediate. It is also possible that the reaction is driven by a thermodynamically favorable later step in the mechanism. The relevance of a superoxide-Fe^III intermediate in the RDOs must still be understood in the context of the peroxide shunt reaction as this produces the same diol product.

6. Concluding comments

Our understanding of the non-heme iron enzymes has greatly advanced through detailed mechanistic, model, spectroscopic, and crystallographic studies. However, there are still important issues to be considered that include details of the molecular mechanism of O₂ activation, the nature of key intermediates, including the superoxo, and their electronic and steric contributions to reactivity and specificity. It is also important to understand how the mononuclear NHFe enzymes relate to the binuclear NHFe enzymes and to the heme enzymes in O₂ activation and their abilities to perform significant biochemistry.

Abbreviations

5C

5-coordinate

6C

6-coordinate

ABLM

activated bleomycin

ACV

δ-(l-α-aminoadipoyl)-l-cysteinyl-d-valine

αKG

α-ketoglutarate

BLM

bleomycin

DFT

density functional theory

EXAFS

Extended X-ray absorption fine structure

FIH

factor inhibiting HIF-1

FMO

frontier molecular orbital

HmaS

(4-hydroxy) mandelate synthase

HPCD

homoprotocatechuate 2,3-dioxygenase

HPPD

(4-hydroxyphenyl) pyruvate dioxygenase

IPNS

isopenicillin-N synthase

LF

ligand field

LT

low temperature

MCD

magnetic circular dichroism

NHFe

non-heme Fe

NRVS

nuclear resonance vibrational spectroscopy

PAH

phenylalanine hydroxylase

RDO

Rieske dioxygenase

SyrB2

syringomycin biosynthesis enzyme 2

TauD

taurine dioxygenase

VTVH MCD

variable-temperature variable-field magnetic circular dichroism