Evidence for C–H cleavage by an iron–superoxidecomplex in the glycol cleavage reaction catalyzedby myo-inositol oxygenaseGang Xing*, Yinghui Diao*, Lee M. Hoffart*, Eric W. Barr*, K. Sandeep Prabhu†, Ryan J. Arner†, C. Channa Reddy†,Carsten Krebs*‡§, and J. Martin Bollinger, Jr.*‡§

Departments of *Biochemistry and Molecular Biology, †Veterinary and Biomedical Sciences, and ‡Chemistry, Pennsylvania State University,University Park, PA 16802

Edited by Christopher T. Walsh, Harvard Medical School, Boston, MA, and approved March 2, 2006 (received for review September 28, 2005)

myo-Inositol oxygenase (MIOX) activates O2 at a mixed-valentnonheme diiron(II�III) cluster to effect oxidation of its cyclohexan-(1,2,3,4,5,6-hexa)-ol substrate [myo-inositol (MI)] by four electronsto D-glucuronate. Abstraction of hydrogen from C1 by a formally(superoxo)diiron(III�III) intermediate was previously proposed. Useof deuterium-labeled substrate, 1,2,3,4,5,6-[2H]6-MI (D6-MI), hasnow permitted initial characterization of the C–H-cleaving inter-mediate. The MIOX�1,2,3,4,5,6-[2H]6-MI complex reacts rapidly andreversibly with O2 to form an intermediate, G, with a g � (2.05,1.98, 1.90) EPR signal. The rhombic g-tensor and observed hyper-fine coupling to 57Fe are rationalized in terms of a (superoxo)di-iron(III�III) structure with coordination of the superoxide to a singleiron. G decays to H, the intermediate previously detected in thereaction with unlabeled substrate. This step is associated with akinetic isotope effect of >5, showing that the superoxide-levelcomplex does indeed cleave a C–H(D) bond of MI.

mixed-valent � nonheme diiron enzymes � oxygen activation � superoxointermediate

The ring-opening, glycol-cleaving, four-electron oxidation ofmyo-inositol (MI) to D-glucuronate (DG) by a single equiv-

alent of dioxygen (Scheme 1) is catalyzed by MI oxygenase(MIOX) (1–4). Recent studies have suggested that the enzyme,which was shown nearly 50 years ago to require iron (1, 2),contains a coupled dinuclear nonheme iron cluster (5), makingMIOX the most recent addition to the nonheme diiron oxygen-ase�oxidase family that also includes bacterial hydrocarbonhydroxylases (e.g., soluble methane monooxygenase), plant fattyacyl desaturases (e.g., stearoyl acyl carrier protein �9 desatu-rase), and protein R2 of class I ribonucleotide reductase (R2)(6–10). Mossbauer and EPR spectra showed that treatment ofrecombinant Mus musculus MIOX isolated in its iron-free formfrom Escherichia coli with Fe(II) and O2 leads to formation ofan antiferromagnetically coupled diiron cluster in either theII�III or III�III oxidation state, depending on the O2�MIOXratio and the presence or absence of a reductant (ascorbate orcysteine). Binding of MI was shown to perturb the spectra of bothoxidation states in a manner consistent with direct coordinationof the substrate to the cluster (5).

All nonheme diiron oxygenases and oxidases characterizedbefore MIOX activate O2 with the II�II oxidation state of thecofactor (11, 12). For several of the reactions, (peroxo)di-iron(III�III) intermediates have been demonstrated. These com-plexes are generally proposed to undergo O–O-bond cleavage togenerate high-valent iron complexes that cleave strong C–H orO–H bonds of their oxidation targets (8, 11–14). Indeed, thediiron(III�IV) cluster, X (15, 16), and the diiron(IV�IV) cluster,Q (8, 13, 17), have been directly characterized in the R2 andsoluble methane monooxygenase reactions, respectively. In eachof the previously characterized diiron-oxygenase�oxidase reac-tions, a diiron(III�III) ‘‘product’’ state of the cluster is generatedat the end of the oxidation sequence. Subsequent events require

reduction of the cluster back to the diiron(II�II) state byadditional proteins, with electrons provided ultimately byNAD(P)H. This redox cycling of the cofactor and provision oftwo electrons by the nicotinamide ‘‘cosubstrate’’ ensure that atmost two electrons can be extracted from the substrate. TheMIOX reaction, a four-electron oxidation, would seem to re-quire a different mechanism.

Indeed, a recent study concluded that the mixed-valent,II�III state of the cofactor, rather than the conventional II�IIstate, activates O2 for DG production in the MIOX reaction(4). Single-turnover experiments showed that the fully reducedenzyme (MIOXII/II) reacts with limiting O2 in the presence ofsaturating MI to generate the mixed-valent enzyme as a stableproduct with unit stoichiometry and with only a low yield ofDG. By contrast, the MI complex of the mixed-valent form(MIOXII/III�MI) reacts with limiting O2 cyclically (i.e., regen-erating MIOXII/III�MI on consumption of the O2) and with ayield of DG similar to the theoretical value of one per O2. Thecycle is sufficiently rapid to account for the rate of steady-stateturnover and proceeds through at least two accumulatingintermediate states [H, with a resolved g � (1.92, 1.76, 1.54)

Conflict of interest statement: No conflicts declared.

This paper was submitted directly (Track II) to the PNAS office.

Abbreviations: MI, myo-inositol; MIOX, MI oxygenase; DG, D-glucuronate; D6-MI,1,2,3,4,5,6-[2H]6-MI; FQ, freeze-quench; 2H-KIE, deuterium kinetic isotope effect; H6-MI,1,2,3,4,5,6-[1H]6-MI.

§To whom correspondence may be addressed. E-mail: jmb21@psu.edu or ckrebs@psu.edu.

© 2006 by The National Academy of Sciences of the USA

Scheme 1. Reaction catalyzed by MIOX and the mechanism used to simulatekinetics of MIOXII/III�MI, G, and H in the FQ EPR experiment depicted in Figs. 2and 3. The rate constants for MI binding and dissociation, decay of H, andrelease of DG are taken from a previous study (4). For rate constants lackingerror limits, more experiments are required to estimate such limits.

6130–6135 � PNAS � April 18, 2006 � vol. 103 � no. 16 www.pnas.org�cgi�doi�10.1073�pnas.0508473103

EPR signal, and MIOXII/III, with its optical absorption featurecentered at �495 nm]. A mechanism was proposed in which O2adds to the diiron(II�III) cluster to generate a formally(superoxo)diiron(III�III) intermediate that abstracts a hydro-gen atom from C1 of MI. Transfer of the resulting iron-boundhydroperoxide to the C1 radical was proposed to generate a1-hydroperoxy-MI intermediate, which Hamilton and cowork-ers argued could decompose to DG (18).

In the present study, the use of deuterium-labeled substrate[1,2,3,4,5,6-[2H]6-MI (D6-MI)] has allowed an additional inter-mediate state, G, to be detected by freeze-quench (FQ) EPRspectroscopy. Kinetic data establish that G converts to H, theintermediate previously detected in the reaction with the unla-beled substrate [1,2,3,4,5,6-[1H]6-MI (H6-MI)] (4), and that thisconversion is associated with a deuterium kinetic isotope effect(2H-KIE) of �5. These results imply that G cleaves a C–H bondof the substrate in converting to H. The EPR properties of G andthe fact that its formation is readily reversible are consistent withthe previously postulated (superoxo)diiron(III�III) formulationfor the C–H cleaving complex. The data thus provide directspectroscopic and kinetic evidence for C–H cleavage by ametal-superoxide enzyme intermediate, a phenomenon previ-ously proposed on the basis of less direct kinetic and computa-tional analyses to occur in the reactions of the uncoupleddicopper enzymes, dopamine �-monooxygenase and peptidylglycine �-amidating monooxygenase (19, 20).

ResultsStopped-Flow Absorption Evidence for a Substrate 2H-KIE in the MIOXReaction. Studies have shown that the MIOXII/III�MI complexexhibits both an absorption feature centered at 390 nm (4) andan axial g � (1.95, 1.81, 1.81) EPR signal (5). Mixing of thecomplex with limiting O2 in the presence of excess substrate wasshown to result in rapid decay and, subsequently, completeregeneration of both spectral features (4). To test for a substrate2H-KIE on this reaction, the kinetics of the change in absorbanceat 390 nm upon mixing of MIOXII/III�D6-MI with limiting O2were compared with those for the H6-MI reaction (Fig. 1A). Thetrace for the D6-MI reaction (solid line) shows an initial rise

phase preceding the decay and redevelopment phases charac-terizing the H6-MI reaction (dashed line). These data suggestthat an additional intermediate that absorbs more intensely thanMIOXII/III�MI accumulates exclusively (or to a much greaterextent) in the D6-MI reaction. Kinetic-difference spectra (Fig.1B) show that the positive difference absorption associated withthe presumptive new intermediate maximizes near 435 nm (solidred spectrum). Accumulation of a new intermediate specificallyas a result of the presence of deuterium in the substrate wouldstrongly implicate this species in effecting C–H(D) cleavage.

Detection of the C–H-Cleaving Iron Intermediate, G, by FQ EPRSpectroscopy. Indeed, the EPR spectra of samples frozen early inthe reaction of MIOXII/III�D6-MI with O2 (Fig. 2) exhibit a set ofthree transient features in the 3,250–3,600 G region (Fig. 2B)that were not previously detected in similar experiments withH6-MI (4). These features are indicative of a rhombic S � 1�2spin system having effective g values of 2.05, 1.98, and 1.90. Therhombic signal develops rapidly, reaching maximum intensitywithin �25 ms under these conditions (excess, 0.95 mM O2) anddecays completely within a few hundred milliseconds as the g �(1.92, 1.76, 1.54) signal of the previously detected intermediate,H, develops (marked in Fig. 2D). The intensities of resolvedfeatures in the derivative EPR spectra allow the kinetics of theMIOXII/III�D6-MI reactant, the g � (2.05, 1.98, 1.90) interme-diate, which we hereafter designate G, and H to be defined inrelative terms (see Fig. 6, which is published as supportinginformation on the PNAS web site). The kinetic data (Fig. 3)suggest that G is a precursor to H and imply that the first stepin the kinetic mechanism previously used (4) to analyze theH6-MI reaction, conversion of MIOXII/III�MI and O2 to H,actually comprises two steps, formation of G from the reactantsand decay of G to H (Scheme 1). In the H6-MI reaction, thesecond step is sufficiently rapid to have prevented detection of

Fig. 1. Stopped-flow absorption evidence for a substrate 2H-KIE in the MIOXreaction. Concentrations after mixing were �0.3 mM MIOXII/III�MI (0.5 mMtotal MIOX), 0.10 mM O2, and 25 mM D6-MI (solid traces) or H6-MI (dashedtraces). (A) The change in absorbance at 390 nm with time for the tworeactions. (B) Kinetic-difference absorption spectra (changes with respect tothe first reliable spectrum at 1.9 ms).

Fig. 2. X-band EPR spectra at 10 K of FQ samples from the reaction at 5°C ofMIOXII/III�D6-MI with excess O2. Concentrations after mixing were 0.22 mMMIOXII/III�MI, 0.95 mM O2, and 33 mM D6-MI. Reaction times were 0 ms (A), 24ms (B), 60 ms (C), and 260 ms (D). Note the different scaling of A.

Xing et al. PNAS � April 18, 2006 � vol. 103 � no. 16 � 6131

BIO

CHEM

ISTR

Y

G (4), but the substrate 2H-KIE allows G to accumulate to readilydetectable levels under the same reaction conditions. The kineticdata for the D6-MI reaction can be simulated in terms of Scheme1. The two crucial features of this kinetic mechanism are (i) alarge substrate 2H-KIE in conversion of G to H, which impliesthat G abstracts hydrogen from MI; and (ii) reversibility information of G from MIOXII/III�MI and O2, which is indicated bythe biphasic decay of MIOXII/III�MI and the coincidence of theslow phase with decay of its successor, G. This reversibility is animportant clue to the identity of G, as discussed in detail below.

Photolytic Lability of and Electron–Nuclear Hyperfine Coupling to 57Fein G. After trapping at �120 K, G is stable for at least a few weekswhen stored in the dark in liquid nitrogen (77 K), but theintermediate is very light sensitive, even at 77 K (Fig. 4).Exposure of a sample containing the intermediate (Fig. 4A) tolaboratory fluorescent light for �45 min at 77 K caused almostcomplete decay of the characteristic EPR signal (Fig. 4B)without engendering any obvious new signal. The difference ofthe EPR spectra taken before and after light exposure (Fig. 4C,solid line) is thus the resolved spectrum of G. The spectrum ofG obtained by repeating the photolysis experiment with a sampleprepared with 57Fe (Fig. 4D, solid line) shows hyperfine couplingto the I � 1�2 nuclei, demonstrating that G is, as expected, aniron complex. The EPR spectrum of 57Fe-coupled G can besimulated (Fig. 4D, dotted line) by using the parameters ob-tained from simulation of the 56Fe spectrum (Fig. 4C, dottedline) and isotropic hyperfine coupling tensors of �13 G and �25G for the two 57Fe nuclei.

Estimate for the 2H-KIE on Decay of G. To prove that the conversionof G to H involves abstraction of hydrogen from MI, it wasdeemed important to demonstrate formation of G also in theH6-MI reaction and to verify that the 2H-KIE on decay of G istoo large to be a secondary effect. To promote accumulation ofG in the H6-MI reaction at a reaction time accessible bytraditional FQ methodology, the reaction temperature wasreduced to 0.5°C. The EPR features of G are just discernible inthe spectrum of a 20-ms sample from the H6-MI reaction (Fig.5A, red spectrum). They are, as expected, much less intense thanin the spectrum of an equivalent sample from the D6-MI reaction(Fig. 5A, blue spectrum). The g � (1.92, 1.76, 1.54) signal of H(red arrows in Fig. 5 A and B) is correspondingly more intensein the spectrum of the H6-MI sample, confirming that the lesseraccumulation of G results from its faster decay to H. Photolysisof the intermediate in both samples (Fig. 5B) allowed the relativeconcentrations of G to be estimated from the difference spectra(Fig. 5C). G accumulates in the H6-MI reaction only to 0.13 timesits level in the D6-MI reaction at 20 ms under these conditions.

Identical analysis of samples quenched at 43 ms (data not shown)showed that the ratio of [G] is even smaller (�0.03) at this longerreaction time. To estimate the 2H-KIE on the G3H step fromthese observations, kinetic simulations for G in the H6-MIreaction were carried out according to Scheme 1, with only therate constant of the G3H step varied (see Fig. 7, which ispublished as supporting information on the PNAS web site). A2H-KIE of 2n was assumed, with n varied in integer steps from0 (black trace) to 5 (orange trace). The ratio of [G] in the tworeactions was calculated as a function of time at each 2H-KIE(solid lines). Under the assumption that the kinetics of thereaction are unchanged by reduction of the temperature from5°C to 0.5°C, the experimental ratios of [G] (circular points)suggest a 2H-KIE of between 8 (solid green trace) and 16 (solidpurple trace). Under the more reasonable assumption that therate constants at the lower temperature are smaller than butproportional to those at 5°C, meaning that any reaction time at0.5°C effectively corresponds to a shorter reaction time at 5°C,a larger 2H-KIE would be necessary to produce the observedratios of [G]. More quantitative assessment of the 2H-KIE willrequire delineation of the kinetics (as in Fig. 3) under conditions(e.g., lower temperature), favoring accumulation of G in theH6-MI reaction at reaction times that can be covered moreeffectively by FQ EPR. Nevertheless, the 2H-KIE is sufficiently

Fig. 3. Kinetics of MIOXII/III�D6-MI (black), G (red), and H (blue) determined (asdescribed in Supporting Text) from the FQ EPR experiment of Fig. 2. Concen-trations are expressed in terms of the fraction of the initial [MIOXII/III�D6-MI].The solid lines are simulations according to Scheme 1.

Fig. 4. Photolytic decay of G at 77 K and its use to resolve the EPR spectrumand 57Fe hyperfine coupling for G. (A) The spectrum is of the 0.043-s samplefrom the experiment of Fig. 2 and was obtained after the sample had beenstored in the dark in liquid nitrogen for 1 week. (B) The spectrum was acquiredafter subsequent exposure of the same sample to laboratory fluorescent lightfor 40 min in liquid nitrogen. (C) The solid line is the change caused by theexposure (A � B). The dotted trace is a simulation of this difference spectrum.(D) The difference spectrum from equivalent analysis of a G-containing sampleprepared with 57Fe (solid line) and simulations generated by applying to thetheoretical spectrum for 56Fe-G two isotropic 57Fe hyperfine-coupling tensorsof either 7 G (corresponding to the superoxide-bridged formulation for G;dashed trace) or 13 G and 25 G [corresponding to FeA and FeB, respectively, ofthe favored terminal (superoxo)diiron(III�III) formulation for G; dotted trace].The experimental spectra were acquired at 10 K.

6132 � www.pnas.org�cgi�doi�10.1073�pnas.0508473103 Xing et al.

large (�5, taking into account errors inherent in the FQ EPRmeasurements) to conclude that it is primary and therefore thatconversion of G to H involves abstraction of a carbon-boundhydron from MI. As the bond to C1 is the only C–H(D) bond thatis obviously cleaved during the reaction, G most likely abstractsthe hydron from C1.

Additional Evidence for Reversibility of G Formation. Fig. 5 alsocorroborates the second key feature of Scheme 1, the reversibilityof O2 addition to MIOXII/III�MI to produce G. Comparison of theintensities of the g � (1.95, 1.81, 1.81) signal of MIOXII/III�MI (bluearrows) reveals that more of the reactant remains in the D6-MIreaction at both 20 ms and 43 ms (data not shown). The 2H-KIE onthe G3H step allows G to accumulate, and the reverse of Gformation then ‘‘feeds back’’ to maintain a greater population ofMIOXII/III�MI until the accumulated G is ultimately depleted by itsforward conversion.

Steady-State 2H-KIE. The kinetic mechanism of Scheme 1 predictsa 2H-KIE on steady-state turnover of close to unity. The reasonis that decay of H is almost totally rate-limiting for turnover ofeven D6-MI (at 5°C). Examination of the steady-state velocitiesat 5°C (by the orcinol colorimetric assay; see ref. 21) with �0.3mM O2 and 87 mM MI gave a 2H-KIE of 1.35 � 0.35 (four trials;error limits calculated from mean � standard deviation ofindividual velocities for H6-MI and D6-MI reactions), which isconsistent with the prediction of Scheme 1. We note thatHamilton and coworkers (18) had reported larger C1-deuteriumand -tritium KIEs of 2.1 and 7.5 on turnover of the pig kidneyenzyme at 30°C. It is not clear whether differences between the

kinetic mechanisms of pig kidney MIOX and our recombinantmouse enzyme or differences in the reaction conditions used areresponsible for the observed difference in steady-state 2H-KIE.

DiscussionG Is Formally a (Superoxo)diiron(III�III) Complex. Addition of O2 tothe diiron(II�III) cluster in MIOXII/III�MI would produce anadduct with an odd number of electrons and a half-integer-spinground state. The association of G with the rhombic g � 2 EPRsignal establishes that it has an S � 1�2 ground state, meaningthat it is either isoelectronic with this initial adduct or morereduced by two or four electrons. The fact that it accumulates toa much greater extent in the D6-MI reaction and the corollarythat it precedes C–H cleavage in the mechanism would seem torule out reduction of the initial adduct by the substrate, andreduction by an exogenous agent can be ruled out on the basisof the reaction stoichiometry of one O2 per DG (2), whichrequires that all four electrons needed to reduce O2 be extractedfrom MI. Thus, G has the same number of electrons as the initialdiiron(II�III)-O2 adduct.

Possible Structures of G. G could be a (superoxo)diiron(III�III) or[(hydro)peroxo]diiron(III�IV) complex formed by one-electronor two-electron, respectively, oxidative addition of O2 to thediiron(II�III) cluster or a more advanced state formed byaddition of O2 and subsequent O–O cleavage with furtheroxidation of the Fe ions (e.g., to the IV�V state). The revers-ibility of its formation strongly favors formulations in which thedioxygen unit remains intact. Reversibility in O–O cleavage is, toour knowledge, unprecedented for diiron complexes, although ithas been demonstrated for an inorganic dicopper complex (22).We consider it unlikely that O–O-bond reformation by oxidativecoupling of oxo�hydroxo ligands of a hypothetical diiron(IV�V)intermediate would be sufficiently facile to account for thereasonably rapid (�40 s�1) O2 dissociation required by thekinetic data. Of the possible formulations with intact O–Obonds, we consider the [(hydro)peroxo]diiron(III�IV) possibilityless likely. To our knowledge, formation of FeIV enzyme inter-mediates has thus far been seen only in conjunction with O–Obond cleavage (8, 11–13, 23). It is not obvious that reduction ofO2 by two electrons to a peroxide would be sufficiently exergonicto drive oxidation of an Fe ion of the diiron(II�III) reactantto the high-valent state. Moreover, Que and coworkers (24)recently prepared and characterized by resonance Ramanspectroscopy an inorganic (superoxo)diiron(II�III) complexfrom a diiron(II�II) precursor and O2. A superoxide ligand to adiiron(III�III) cluster, as proposed for G, should be at least asstable with respect to further reduction to peroxide as thediiron(II�III)-coordinated superoxide of the model complex.Obviously, further structural characterization is required todetermine the effective oxidation states of the two iron sites ofG and which of the two formulations is most appropriate.

Because its EPR properties are currently the only structuralinformation available for G, it is appropriate to ask whether theg-tensor and 57Fe hyperfine coupling are consistent with thefavored (superoxo)diiron(III�III) formulation. Such a complexwould contain three paramagnetic centers, two high-spin FeIII

ions (S � 5�2) and the superoxide radical anion (S � 1�2). Theexchange interactions among them would determine the relativeenergies of the total spin states. The experimentally observedS � 1�2 ground state can be rationalized by assuming antifer-romagnetic coupling between the two FeIII ions and between(among) the superoxide and the FeIII site(s) to which it binds.For coordination to only one FeIII (designated FeA), exchangecoupling between the superoxide and the other FeIII (FeB)should be much smaller and negligible. As a consequence, thespin of FeA would align antiparallel to those of the superoxideand FeB, resulting in an S � 1�2 ground state. Alternatively, a

Fig. 5. Accumulation of G early in the H6-MI reaction (red spectra) andphotolysis to quantify it relative to [G] in an identical sample from the D6-MIreaction (blue spectra). The reaction was carried out at 0.5°C. Concentrationswere 0.20 mM MIOXII/III�MI, 1.0 mM O2, and 33 mM MI. The reaction time was0.020 s. (A) The samples before light exposure. (B) The same samples after lightexposure as in Fig. 4. (C) The difference spectra (A � B). Note the differentscaling in C. The spectra were acquired at 12.5 K. Ten scans were accumulatedfor each.

Xing et al. PNAS � April 18, 2006 � vol. 103 � no. 16 � 6133

BIO

CHEM

ISTR

Y

bridging superoxide would yield two similar antiferromagneticFeIII-superoxide exchange interactions, resulting in a spin-frustrated cluster. Under the assumption of equal exchangecoupling of the two FeIII ions with the superoxide, an S � 1�2ground state is predicted for JFe-Fe � 0.25 � JFe-superoxide (Ham-iltonian � J S1�S2).

In exchange-coupled systems such as that proposed for G, theg-tensors of the total spin states are linear combinations ofthe g-tensors of the individual paramagnetic centers (25). Theg-tensors of organic radicals and high-spin FeIII ions are gener-ally less anisotropic than is observed for the ground state of G.However, quantum mechanical mixing of excited spin states withthe S � 1�2 ground state increases g-anisotropy, as has beendescribed for valence-localized high-spin diiron(II�III) com-plexes (26–28). For the hypothetical (superoxo)diiron(III�III)complex, the extent of mixing would depend on the zero-fieldsplitting parameters of the FeIII ions [DFe(III), (E�D)Fe(III)] andthe energy gaps between the spin ground and excited states,which would be determined by the three J values. Spin–Hamiltonian simulations were conducted to assess whether theg-anisotropy observed for G is compatible with the proposed(superoxo)diiron(III�III) formulation. In these simulations, theenergies of the different spin states of the exchange-coupledsystem were calculated by diagonalization of the spin–Hamiltonian matrix with the following physically reasonableparameters, �DFe(A)� � 3 cm�1, �DFe(B)� � 3 cm�1, and gFe(A) �gFe(B) � gsuperoxide � 2.00 (11). Exchange coupling constants ofJFe(A)-Fe(B) � 15 cm�1, JFe(A)-superoxide � 80 cm�1 (fixed), andJFe(B)-superoxide � 0 cm�1 (fixed) were used to represent theterminal superoxide complex, and JFe(A)-Fe(B) � 27 cm�1,JFe(A)-superoxide � 80 cm�1 (fixed), and JFe(B)-superoxide � 80 cm�1

(fixed) were used to represent the bridged complex. For eachparameter set, a ground state of S � 1�2 was verified and its gvalues were calculated from the Zeeman splitting of the twostates. The experimentally observed g-anisotropy could be re-produced for both coordination modes. Thus, the rhombic EPRsignal of G is consistent with the (superoxo)diiron(III�III)formulation.

The two modes of superoxide coordination would give rise todifferent hyperfine interactions of the electronic spin with the57Fe sites. The absolute magnitudes of the hyperfine interactionsdepend on the product of the intrinsic A-tensors, which aretypically ��29 MHz (29) and almost isotropic for high-spin FeIII

sites, and the spin-projection coefficients, c, of the two sites (25).For the superoxide-bridged complex, the spin projection factorsof the FeIII sites would be identical and relatively small, 2�3.¶For superoxide coordination only to FeA, the coefficients wouldbe cA � �14�9 and cB � 7�3. The magnitude of the overallbroadening of the EPR spectrum of G by 57Fe (Fig. 4D) isincompatible with the small couplings predicted for the bridgedcomplex (dashed trace) but compatible with the couplingspredicted for the terminal complex (dotted trace). We thereforefavor assignment of G as a (superoxo)diiron(III�III) complexwith either end-on (�1) or side-on (�2) coordination of thesuperoxide to a single FeIII. More definitive determinationshould be possible from electron-nuclear double-resonancespectroscopic experiments on 57Fe- and 17O-labeled G.

Possible Mechanisms of DG Production. We previously tentativelyrationalized the significant changes to the EPR and Mossbauerspectra of the diiron(II�III) cluster in MIOX caused by bindingof MI in terms of an increase in the exchange coupling between

the Fe ions (5). We postulated that substrate binding imparts abetter mediator of superexchange than is provided by the proteinin substrate-free MIOXII/III. We further speculated that theC1-oxygen of MI bridges the two irons as an alkoxide, on thegrounds that this coordination mode could increase J, positionthe C1–H for abstraction by a diiron–oxygen adduct, andpossibly activate the C1–H bond for cleavage. This hypothesisrequires rigorous evaluation by structural methods (crystallog-raphy and electron-nuclear double-resonance spectroscopy).Moreover, knowledge of the manner of interaction of MI withthe diiron cluster (if any) and the orientation of the substratewith respect to the cofactor would drastically limit mechanisticpossibilities. With these caveats in mind, we nevertheless depictMI binding to the cluster via a C1–O bridge and envisage twomain classes of possible mechanisms for conversion of G to theproduct complex, MIOXII/III�DG (Scheme 2). Both classes wouldbe initiated by abstraction of hydrogen, most likely as a hydrogenatom, from C1 (Scheme 2 A). The first class (Scheme 2B Upper)would involve formation of the 1-hydroperoxy-MI intermediateenvisaged by Hamilton and coworkers (18), who noted thechemical precedent that this species could break down to DGwith incorporation of a single O atom from O2 into the carbox-ylate (30), as occurs in the MIOX reaction (3). This class wouldnot require formation of high-valent complexes. The secondclass (Scheme 2B Lower) would involve formation of a (possibly)coordinated gem-diol(ate) intermediate by formal hydroxyla-tion. Formation of a high-valent complex [e.g., diiron(III�IV)]would, in this case, provide an electron-deficient center intowhich electrons would flow during C1–C6 cleavage. Becausenumerous variations of each class may be proposed, we deferdiscussion of further details until additional insight, specificallyinto the mode of MI binding (by x-ray crystallography and

¶When JFe-Fe is in the range from 0.25 � JFe-superoxide to 1 � JFe-superoxide, cA � cB � 2�3 isobserved. Values of JFe-Fe outside of this range are incompatible with the EPR data: forJFe-Fe � 0.25 � JFe-superoxide, the ground state is S � 3�2; for JFe-Fe � 1 � JFe-superoxide, the S �

1�2 ground-state has spin projection factors of 0, which would result in no 57Fe-hyperfinecoupling in the EPR spectrum.

Scheme 2. Possible mechanisms for conversion of MI to DG initiated by theformally (superoxo)diiron(III�III) intermediate G. Both pathways would beinitiated by abstraction of hydrogen, most likely as a hydrogen atom, from C1(A). (B Upper) The first class would involve formation of the 1-hydroperoxy-MIintermediate. (B Lower) The second class would involve formation of a (pos-sibly) coordinated gem-diol(ate) intermediate by formal hydroxylation.

6134 � www.pnas.org�cgi�doi�10.1073�pnas.0508473103 Xing et al.

electron-nuclear double-resonance spectroscopy) and the natureof H (by additional kinetic experiments and spectroscopy), hasbeen obtained. Regardless of the structural and mechanisticdetails, the spectroscopic and kinetic characterization of Greported here lends credence to the notion recently proposed forthe uncoupled dicopper oxygenases, dopamine �-monooxygen-ase and peptidyl glycine �-amidating monooxygenase, that O2-derived metalloenzyme intermediates with intact O–O bonds(i.e., superoxide-level complexes) can cleave relatively strongC–H bonds (19, 20).

Materials and MethodsMaterials. 57Fe metal was obtained from Advanced Materials andTechnologies, Inc. (New York). It was converted to 57FeII bydissolution in 2 N H2SO4 (4 H�Fe). D6-MI was obtained fromC�D�N Isotopes, Inc. (Quebec). Unlabeled MI (H6-MI) wasobtained from Sigma.

Preparation of Apo-MIOX. M. musculus kidney MIOX was over-expressed in E. coli, purified in its iron-free form, and quantifiedas described in ref. 5.

Preparation of the MIOXII/III�MI Reactant Complex. The MIOXII/III�MIreactant solution for stopped-flow and FQ EPR experiments wasprepared by subjecting a solution containing 0.9–1.1 mM iron-freeMIOX, 2.0 eq FeII (either natural abundance or �97% 57Fe), and3.4–4 mM ascorbate to the ‘‘O2-diffusion treatment’’ described inref. 4, venting the protein solution to the anoxic chamber to disperseresidual O2, and adding either H6-MI or D6-MI to either 50 mM (forthe stopped-flow experiment) or 100 mM (for the FQ EPRexperiments). In each FQ experiment, a sample of the reactantsolution was analyzed by EPR to determine [MIOXII/III�MI], whichwas in all cases 0.6 equiv relative to total MIOX (4, 31). In thestopped-flow experiment, [MIOXII/III�MI] was estimated by com-parison of absorbance changes caused by addition of MI to thoseseen in preparation of the complex for the FQ experiments.

Stopped-Flow Absorption and FQ EPR Experiments. General proce-dures for stopped-flow absorption and FQ EPR experiments

have been described in refs. 32 and 33. In each experiment, anO2-free solution of MIOXII/III�MI was mixed with MIOX buffer(50 mM Bistris-acetate, pH 6.0�10% wt/wt glycerol) containingO2. In the stopped-flow experiment, the buffer had been broughtto equilibrium with air at 23°C and was mixed with an equalvolume of the protein reactant solution. In the FQ EPR exper-iments, the buffer was brought to equilibrium with 1.05 atm ofO2 (1 atm � 101.3 kPa) at either 0°C or 5°C and was mixed with0.5 equivalent vol of the MIOXII/III�MI reactant solution. Thereaction temperature and reactant concentrations after mixingare given in the appropriate figure legend. In the calculation ofreaction times for the FQ samples, a ‘‘quench time’’ of 0.015 s wasadded to the known transit time through the FQ reaction loop,as indicated by previous studies (34). In each FQ EPR experi-ment, a control sample was prepared by freeze-quenching aftera mix of the protein reactant with 2 vol O2-free buffer. Com-parison of the spectrum of this control sample to that of anequivalently diluted but manually frozen sample of the reactantsolution indicated an FQ ‘‘packing factor’’ of 0.55, which is inagreement with our previous studies (34).

EPR Spectroscopy. The spectrometer has been described in ref. 34.Spectra were acquired at the temperatures indicated in the figurelegends with the following spectrometer settings: power, 20 mW;modulation amplitude, 10 G; acquisition time, 168 s; conversiontime, 164 ms; and time constant, 164 ms. Simulations werecarried out with the program SYMPHONIA (Bruker, Billerica,MA). Measurement of peak intensities is described in SupportingText, which is published as supporting information on the PNASweb site.

Kinetic Simulations. Simulations were carried out with the pro-gram KINTEKSIM (KinTek Corp., Austin, TX). Details of thescaling of the raw EPR intensities in the preparation of Fig. 3 areprovided in Supporting Text.

We thank Prof. Brian Hoffman for a thought-provoking discussion ofpossible mechanisms. This work was supported by an Innovative Bio-technology Seed Grant from Johnson and Johnson and The HuckInstitute of Life Sciences (Pennsylvania State University, State College).

1. Charalampous, F. C. & Lyras, C. (1957) J. Biol. Chem. 228, 1–13.2. Charalampous, F. C. (1959) J. Biol. Chem. 234, 220–227.3. Moskala, R., Reddy, C. C., Minard, R. D. & Hamilton, G. A. (1981) Biochem.

Biophys. Res. Commun. 99, 107–113.4. Xing, G., Barr, E. W., Diao, Y., Hoffart, L. M., Prabhu, K. S., Arner, R. J.,

Reddy, C. C., Krebs, C. & Bollinger, J. M., Jr. (2006) Biochemistry, in press.5. Xing, G., Hoffart, L. M., Diao, Y., Prabhu, S. K., Arner, R. J., Reddy, C. C.,

Krebs, C. & Bollinger, J. M., Jr. (2006) Biochemistry, in press.6. Vincent, J. B., Olivier-Lilley, G. L. & Averill, B. A. (1990) Chem. Rev. 90,

1447–1467.7. Nordlund, P. & Eklund, H. (1995) Curr. Opin. Struct. Biol. 5, 758–766.8. Wallar, B. J. & Lipscomb, J. D. (1996) Chem. Rev. 96, 2625–2657.9. Kurtz, D. M., Jr. (1997) J. Biol. Inorg. Chem. 2, 159–167.

10. Tshuva, E. Y. & Lippard, S. J. (2004) Chem. Rev. 104, 987–1011.11. Solomon, E. I., Brunold, T. C., Davis, M. I., Kemsley, J. N., Lee, S. K., Lehnert,

N., Neese, F., Skulan, A. J., Yang, Y.-S. & Zhou, J. (2000) Chem. Rev. 100,235–349.

12. Krebs, C., Price, J. C., Baldwin, J., Saleh, L., Green, M. T. & Bollinger, J. M.,Jr. (2005) Inorg. Chem. 44, 742–757.

13. Merkx, M., Kopp, D. A., Sazinsky, M. H., Blazyk, J. L., Muller, J. & Lippard,S. J. (2001) Angew. Chem. Int. Ed. 40, 2782–2807.

14. Fox, B. G., Lyle, K. S. & Rogge, C. E. (2004) Acc. Chem. Res. 37, 421–429.15. Bollinger, J. M., Jr., Edmondson, D. E., Huynh, B. H., Filley, J., Norton, J. R.

& Stubbe, J. (1991) Science 253, 292–298.16. Sturgeon, B. E., Burdi, D., Chen, S., Huynh, B. H., Edmondson, D. E., Stubbe,

J. & Hoffman, B. M. (1996) J. Am. Chem. Soc. 118, 7551–7557.17. Lee, S. K., Fox, B. G., Froland, W. A., Lipscomb, J. D. & Munck, E. (1993)

J. Am. Chem. Soc. 115, 6450–6451.18. Naber, N. I., Swan, J. S. & Hamilton, G. A. (1986) Biochemistry 25, 7201–7207.19. Evans, J. P., Ahn, K. & Klinman, J. P. (2003) J. Biol. Chem. 278, 49691–49698.

20. Chen, P. & Solomon, E. I. (2004) J. Am. Chem. Soc. 126, 4991–5000.21. Reddy, C. C., Swan, J. S. & Hamilton, G. A. (1981) J. Biol. Chem. 256,

8510–8518.22. Halfen, J. A., Mahapatra, S., Wilkinson, E. C., Kaderli, S., Young, V. G., Jr.,

Que, L., Jr., Zuberbuhler, A. D. & Tolman, W. B. (1996) Science 271,1397–1400.

23. Sono, M., Roach, M. P., Coulter, E. D. & Dawson, J. H. (1996) Chem. Rev. 96,2841–2887.

24. Shan, X. & Que, L., Jr. (2005) Proc. Natl. Acad. Sci. USA 102, 5340–5345.25. Bencini, A. & Gatteschi, D. (1990) EPR of Exchange Coupled Systems (Springer,

Berlin).26. Guigliarelli, B., Bertrand, P. & Gayda, J. P. (1986) J. Chem. Phys. 85,

1689–1692.27. McCormick, J. M., Reem, R. C. & Solomon, E. I. (1991) J. Am. Chem. Soc. 113,

9066–9079.28. Fox, B. G., Hendrich, M. P., Surerus, K. K., Andersson, K. K., Froland, W. A.,

Lipscomb, J. D. & Munck, E. (1993) J. Am. Chem. Soc. 115, 3688–3701.29. Munck, E. (2000) in Physical Methods in Bioinorganic Chemistry, ed. Que, L.,

Jr. (Univ. Sci. Books, Mill Valley, CA), pp. 287–319.30. Hamilton, G. A., Reddy, C. C., Swan, J. S., Moskala, R., Mulliez, E., Naber,

N. I., Nozaki, M., Yamamoto, S. & Ishimura, Y. (1982) in Oxygenases andOxygen Metabolism (Academic, New York), pp. 111–123.

31. Aasa, R. & Vanngård, T. (1975) J. Magn. Reson. 19, 308–315.32. Bollinger, J. M., Jr., Tong, W. H., Ravi, N., Huynh, B. H., Edmondson, D. E.

& Stubbe, J. (1995) Methods Enzymol. 258, 278–303.33. Price, J. C., Barr, E. W., Tirupati, B., Bollinger, J. M., Jr., & Krebs, C. (2003)

Biochemistry 42, 7497–7508.34. Baldwin, J., Krebs, C., Ley, B. A., Edmondson, D. E., Huynh, B. H. & Bollinger,

J. M., Jr. (2000) J. Am. Chem. Soc. 122, 12195–12206.

Xing et al. PNAS � April 18, 2006 � vol. 103 � no. 16 � 6135

BIO

CHEM

ISTR

Y