Studies on the oxidative half-reaction of xanthine oxidase.

The oxidative half-reaction of xanthine oxidase is reexamined with regard to the generation of the superoxide anion. By using cytochrome c reduction to monitor superoxide, it is found that the stoichiometry of superoxide produced to enzyme reoxidized is 2:1, significantly greater than previously reported (Olson, J. S., Ballou, D. P., Palmer, G., and Massey, V. (1974) J. Biol. Chem. 249, 4350-4362). Furthermore, the kinetics of superoxide-dependent cytochrome c reduction exhibits a pronounced lag during the rapid phase of enzyme reoxidation and a limiting rate identical with that of the slow phase of enzyme reoxidation. This indicates that superoxide is generated only in the last steps of the sequential removal of reducing equivalents from the enzyme by molecular oxygen. Experiments with the two-electron-reduced enzyme indicate that it too produces two superoxide ions for each molecule of enzyme reoxidized, demonstrating that it is the last two electrons to leave the enzyme in the course of reoxidation that form superoxide. The sequential scheme for the oxidative half-reaction must therefore be 6 leads to 4 leads to 2 leads to 1 leads to 0, where the numbers reflect the number of reducing equivalents in each intermediate. Using this scheme, both enzyme reoxidation and cytochrome c reduction can be accurately simulated. Reasons for the different behavior of the two-electron-reduced xanthine oxidase compared to the six- and four-electron-reduced enzyme are discussed.

During turnover with xanthine and molecular oxygen, xanthine oxidase generates significant amounts of the superoxide anion in addition to hydrogen peroxide (1-3). The oxidative half-reaction of the catalytic cycle takes place at the flavin site, and the two iron-sulfur centers and molybdenum site of the enzyme do not appear to be directly involved (4). Instead, these sites become reoxidized by transferring their reducing equivalents to the flavin site for reaction with oxygen. This internal electron transfer appears to involve a rapid oxidationreduction equilibrium among the various sites of the enzyme, with reducing equivalents being transferred to the flavin site because of its high oxidation-reduction potential relative to those of the other sites ( 5 ) .
In the course of the reaction of fully reduced xanthine oxidase with oxygen, the steps involved in the production of superoxide are not known. Fully reduced enzyme contains a total of six electrons (two each at the flavin and molybdenum sites, and one each at the two iron-sulfur centers) and reoxi-* This work was supported in part by Grant GM11106 from the National Institutes of Health and by the Amax Foundation. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. $Recipient of financial support from the Michigan Society of Fellows. dation must therefore involve some sort of sequential removal of reducing equivalents by more than one oxygen molecule. It has been inferred that superoxide is produced late in the course of enzyme oxidation (5). For example, under pseudofirst order conditions, the reoxidation reaction is markedly biphasic (6). The spectral characteristics of the slower phase suggest that it is due to the reoxidation of the one-electronreduced form of the enzyme, a process which necessarily involves the production of superoxide. Furthermore, the two phases exhibit different behavior as a function of oxygen concentration (6), demonstrating that they are due to fundamentally different processes. The rate constant of the faster phase shows a hyperbolic dependence on oxygen concentration, indicating the existence of a preequilibrium step prior to the reoxidation event, while that of the slower phase remains directly proportional to oxygen concentration up to 625 pM.
In an effort to describe further the reoxidation reaction in terms of the generation of the superoxide anion, experiments have been undertaken that take advantage of the well established reduction of cytochrome c by superoxide (1) and the inhibition of this reduction by superoxide dismutase ( 7 ) . The results demonstrate that the superoxide is indeed produced only in the final steps of the reoxidation reaction.

MATERIALS AND METHODS
Xanthine oxidase was isolated from fresh milk by the method of Massey et al. (8). The enzyme used had activity flavin ratio values (see Ref. 8) in the range of 140-150, i.e. it contained 67-72% functional active sites (see Ref. 11). AU enzyme concentrations quoted refer to those of the molybdenum or flavin cofactors, of which there are 2/ enzyme molecule (see Ref. 8). Kinetic experiments were performed with a stopped flow spectrophotometer designed and built by Dr. David P. Ballou of the Department of Biological Chemistry, University of Michigan, and interfaced with a Data General Nova 2 minicomputer (9). The instrument was designed so that the wavelength could be automatically scanned over a preselected 400-nm range with a total scanning time of 10 s to allow spectra to be recorded. Concentrations of the reagents reported in kinetic experiments are those before mixing.
Reduced enzyme was prepared by titration with sodium dithionite in an anaerobic tonometer equipped with a side arm cuvette. Reduction was monitored by optical spectroscopy in the 300-700-nm range. Spectra were recorded 20 min after each dithionite addition and carefully watched for increases in absorbance at 318 nm indicating the accumulation of excess dithionite. By ending the titration at the first sign of such absorbance increases, the concentration of excess reducing equivalents could be held to within 5% of the concentration of reducing equivalents in the enzyme.
Beef heart cytochrome c (type V) was purchased from Sigma. Bovine liver superoxide dismutase was purchased from Diagnostic Data, Inc., Mountainview, CA. Unless otherwise stated, experiments were performed in 0.1 M sodium pyrophosphate buffer, pH 8.5, with 0.3 m~ EDTA present, at 25 "C.
Computer simulations were performed with a Nova 2 minicomputer using a fourth order Runge-Kutte (10) routine in the Fortran programming language. Extinction coefficients for the intermediate oxidation states of xanthine oxidase were those calculated by Olson et al. (5). Fits to the data were obtained by manually adjusting the appropriate rate constants in a sequential scheme until the kinetic time course of the reoxidation was adequately described. In simulating the cytochrome c kinetic data, a second order rate constant of 1.6 X lo5 M" s-' for the reduction of cytochrome c by 0 2 was used (3).

Oxidation of Reduced Xanthine Oxidase with Molecular
Oxygen-The optical absorbance change at 450 nm on mixing reduced xanthine oxidase with oxygen is shown in Fig. 1 where it can be seen that the excursion is markedly biphasic. Furthermore, after subtracting out the slow phase and expanding the time scale, it is found that the fast phase is an exponential process only after a pronounced lag phase. Plots of the obsewed rate constants for the fast and slow phases as functions of oxygen concentration are shown in Fig. 2, A and B , respectively. The fast phase exhibits hyperbolic dependence on oxygen concentration and from the x and y intercepts of a double reciprocal plot, a K d of 5 x M and limiting rate, Kfmt, of 125 s-', can be determined. These results are in reasonable agreement with those of Olson et al. (6) who obtained values of 8.3 X M and 120 s-l for K d and k s t , respectively. The rate constant for the slow phase is directly proportional to oxygen concentration giving a second order rate constant of 1.0 X M" s-', in good agreement with the results of Olson et al. (6). These workers concluded from the wavelength dependence of the two phases that the faster involved oxidation of both flavin and iron-sulfur centers whereas the slower involved oxidation only of the iron-sulfur centers. The fact that the two phases exhibit different behavior when the oxygen concentration is varied indicates that two different mechanisms of reoxidation must be operative.
The Reaction of Reduced Xanthine Oxidase with Oxygen in the Presence of Cytochrome c-In an effort to determine which, if either, of the two phases of the oxidative half-reaction was responsible for the generation of superoxide, the oxidation of reduced xanthine oxidase by oxygen was carried out in the presence of oxidized cytochrome c, which is known to be reduced by superoxide (1,3). T o eliminate complications from absorbance changes due to the reoxidation of the enzyme itself, the reaction was carried out f i s t in the absence and then in the presence of a catalytic amount of superoxide dismutase, and the two traces subtracted from one another to give the kinetics of superoxide-dependent reduction of cytochrome c. Initially there was some concern that H202 also produced in the reaction would interfere by reoxidizing the cytochrome c reduced by superoxide, but control experiments  The time courses at 550 nm on mixing reduced xanthine oxidase with 200 p~ oxygen and 63 p~ oxidized cytochrome c, in the absence and presence of superoxide dismutase, are shown in Fig. 3A. The absorbance change in the presence of superoxide dismutase is due to enzyme reoxidation, not to reduction of cytochrome c (control experiments, results not shown). Thus, in agreement with earlier work (l), there was no observable superoxide-independent reduction of cytochrome c. It can be seen that the superoxide-dependent reduction of cytochrome c is characterized by three phases (Fig.  3B), a pronounced lag followed by what appears to be a biphasic time course. A comparison with the time course of enzyme reoxidation in the absence of cytochrome c reveals that the limiting rates of both time courses are identical. In addition, the lag in cytochrome c reduction extends over much the same time interval as the fast phase in enzyme reoxidation. In the first 100 ms of the reactio3, 70% of the absorbance change due to enzyme reoxidation had taken place, but virtually no cytochrome c had been reduced in this time. The kinetics of cytochrome c reduction is not profoundly sensitive to cytochrome c concentrations above 10 FM. The only effect of increasing the cytochrome c concentration f-om 10 to 60 p~ (after mixing) is a slight increase in the rate of superoxidedependent cytochrome c reduction immediately after the lag phase (data not shown). The limiting rate of the reaction is largely unaffected.
Because of the extremely narrow spectral band width of the reduced cytochrome c spectrum at 550 nm, kinetically determined absorbance changes are not a very reliable way to quantitate the amount of superoxide-dependent cytochrome c reduction. Instead, spectra were recorded in the stopped flow spectrophotometer after shots taken in the absence and presence of superoxide dismutase. For these quantitative determinations, an experimentally determined E~~~-~~: , of 18.9 mM-' cm" was used, representing the difference between the peak at 553 nm and the trough at 585 nm in the difference spectrum between oxidized and reduced cytochrome e. This operational extinction change was obtained by measuring the difference spectrum for a known concentration of cytochrome c before and after reduction with sodium dithionite in the stopped flow spectrophotometer. The stoichiometry of cytochrome c reduced/xanthine oxidase oxidized (determined from the AA4wnm on mixing reduced enzyme with oxygen in changes in the absence (-) and presence (---) of 5 X 1O"j g/ml of superoxide dismutase. B, semilogarithmic plot of the difference between the two traces in A. The limiting rate is identical with that of the slow phase of enzyme reoxidation, 0.9 s". The solid line represents a simulation to the data as described in the text.
the absence of cytochrome c ) using this extinction was found to be 2.1 k 0.2 at cytochrome c concentrations above 20 PM. This value is considerably higher than that of 0.1 obtained by Olson et al. (6). These workers quantitated superoxide anion production in rapid quench experiments by integrating its electron paramagnetic resonance signal at g = 2.08. This method is both less sensitive and more prone to artifacts than the cytochrome c technique. The main advantages in using cytochrome c lie in its high extinction change on reduction and the fact that superoxide is scavenged rapidly on formation ( k = 1.6 X lo5 M" s-l) and is not allowed to accumulate in concentrations sufficiently high to support rapid dismutation.
The Reaction of Two-electron-reduced Xanthine Oxidase with Oxygen-Because the time course of superoxide-dependent cytochrome c reduction shows a pronounced lag in the first 200 ms of reaction while the faster phase of enzyme oxidation is taking place, it is likely that superoxide generation is indeed associated with the latter stages of enzyme oxidation. To further investigate this likelihood, the reoxidation of the two-electron-reduced enzyme was examined. In the following experiments, this putative intermediate in the reoxidation of fully reduced xanthine oxidase was generated by treating anaerobic enzyme with a substoichiometric amount of xanthine (approximately 1 5 on a molar basis) just prior to loading the enzyme into the stopped flow apparatus. All experiments were completed within 15 min of preparing the enzyme sample in an effort to avoid complications due to the slow comproportionation of the thermodynamically unstable two-electronreduced species ( 5 ) .
The time course for the reoxidation of two-electron-reduced xanthine oxidase by 200 PM oxygen is shown in Fig. 4A. It can be seen that the reaction is distinctly biphasic, with the slow phase contributing about 40% of the absorbance change. The rate constant for the slow phase, 1.0 s", agrees very well with that for the slow phase in the oxidation of the fully reduced enzyme (0.9 s-l). Significantly, when the slow phase of the reaction was subtracted out and the time scale was expanded, the fast phase did not exhibit a lag (Fig. 4A, squares). That the slow phase of the reaction is due to the reoxidation of oneelectron-reduced enzyme is demonstrated in Fig. 5, which shows a series of time courses taken at various times after the preparation of the two-electron-reduced species. It is readily seen that with time the faster of the two phases in the reaction disappears without effect on the rate of the slower phase. This experiment is directly analogous to the spectral experiment of Olson et al. (6) establishing that on prolonged incubation of the two-electron-reduced enzyme, cornproportionation with oxidized enzyme takes place to form a population of the thermodynamically stable one-electron-reduced enzyme.
The dependence of the rate constants for the two phases on oxygen concentration is shown in Fig. 6. It can be seen that, as was the case for the reoxidation of fully reduced xanthine oxidase, the slow phase in the time course for reoxidation of the two-electron-reduced enzyme is directly dependent on oxygen concentration. The slope of that plot in Fig. 6B gives a second order rate constant of 1.0 X m"' cm", in good agreement with results from experiments with fully reduced enzyme. The faster phase in reoxidation of the two-electronreduced enzyme exhibits saturating behavior with K d and limiting rate equal to 4.7 x M and 160 s-', respectiveIy ( Fig. 6 A ) . Considering reduction ( B ) . Conditions were as described in Fig. 3, except that the enzyme was reduced just prior to reaction by substoichiometric xanthine (1:5 on a molar basis). Solid lines represent simulations to the data, as described in the text. the reoxidation of fully reduced xanthine oxidase in complex with alloxanthine should closely emulate the reoxidation of four-electron-reduced enzyme. The reduced xanthine oxidasealloxanthine complex can be conveniently generated by treating anaerobic enzyme with an excess of allopurinol, which is converted to alloxanthine by active enzyme, which becomes reoxidation of the two-electron-reduced enzyme, experiments with cytochrome c were undertaken in an effort to detect superoxide generation in the reaction of two-electron-reduced xanthine oxidase with oxygen.
The kinetics of superoxide-dependent reduction of cytochrome c on mixing the two-electron-reduced enzyme with 200 p~ oxygen and 60 ~L M cytochrome c (before mixing) is shown in Fig. 4B, for comparison with the time course for enzyme reoxidation in the absence of cytochrome c in Fig. 4A. Like enzyme reoxidation, cytochrome c reduction is biphasic, and the faster phase exhibits no lag. The stoichiometry of cytochrome c reduced/two-electron-reduced enzyme oxidized (the latter determined from the absorbance change at 450 nm using a for two-electron-reduced enzyme of 11,200 M" cm"; Ref. 5) was found to be 2 1 . The amount of cytochrome c reduced did not vary with time as the two-electron-reduced enzyme comproportionated to give a population of the oneelectron-reduced species. This observation provides strong support for the conclusion that both electrons in the twoelectron-reduced enzyme react to form superoxide in the course of the reoxidation reaction. Thus, all of the superoxide generated in the reaction of fully (six-electron-) reduced xanthine oxidase with oxygen is accounted for in the reoxidation of the two-electron-reduced enzyme.
Reoxidation of Fully Reduced Xanthine Oxidase in Complex with Alloxanthine-Because the fully (six-electron-) reduced xanthine oxidase is reoxidized to the two-electron-reduced enzyme without producing superoxide, two two-electron steps must be involved with the four-electron-reduced enzyme as an intermediate. Four-electron-reduced enzyme cannot be generated quantitatively for kinetic examination, but as far as the oxidative half-reaction is concerned, the complex of fully reduced enzyme with alloxanthine should react in very nearly the same way. Massey et al. have shown that alloxanthine binds very tightly to the fully reduced molybdenum site to form an inhibitory complex that is air stable (11). Furthermore, these workers established that alloxanthine perturbs the spectra of reduced and oxidized xanthine oxidase identically and there is therefore no change in the oxidized minus-reduced difference spectrum. Because two electrons must remain behind at the molybdenum site in the course of reoxidation, the overall process involves the removal of four electrons. The only difference between this reaction and the reoxidation of four-electron-reduced enzyme is that, in the former, 100% of the flavin is reduced at the onset of the reaction whereas in the latter only 96% of the flavin is reduced. Clearly, the effects of this discrepancy should be minimal, and reduced. It has been shown that this reaction is rapid compared to the formation of the reduced enzyme-alloxanthine complex, and the active enzyme becomes fully reduced (11). In the course of the reoxidation reaction, the inhibitory complex prevents turnover with the remaining allopurinol present.
The results of such an experiment are shown in Fig. 7A. It can be seen that as in the case of the fully reduced enzyme (minus alloxanthine) the reoxidation of reduced xanthine oxidase in complex with alloxanthine is markedly biphasic. In contrast to the case with the uncomplexed enzyme, however, the fast phase in the reoxidation of the alloxanthine complex

Oxidative Half-reaction of Xanthine Oxidase
exhibits accurately exponential behavior, with no detectable lag phase. This is to be expected if the number of sequential steps contributing to the fast phase is decreased as postulated. Furthermore, the production of superoxide (Fig. 7 B ) does not exhibit as pronounced a lag in the course of reoxidation of the alloxanthine complex as does the reoxidation of uncomplexed reduced enzyme, as expected if one of the steps prior to superoxide production is eliminated in the former case. The stoichiometry of cytochrome c reduced to enzyme reoxidized remains 2:1, as with the six-electron-and two-electron-reduced enzyme.
Scheme for the Reaction of Fully Reduced Xanthine Oxidase with Oxygen-The removal of six electrons from xanthine oxidase is necessarily a sequential process involving the reaction of several oxygen molecules with enzyme at various intermediate levels of reduction. The data described in the previous sections on the behavior of two-and four-electronreduced enzyme suggest that the correct sequence is 6 3 4 + 2 + 1 + 0, where the numbers represent the number of electrons in each intermediate enzyme species. Hydrogen peroxide is produced in the fist two steps and superoxide anion is produced in the last two. Olson et al. (6) showed that such a scheme could be used to simulate the absorbance changes observed on reoxidation very well using calculated extinction changes for each intermediate, but they were unable to account for the discrepancy between the small amounts of superoxide they observed and the 20-fold greater amounts predicted by the above scheme. Our results using cytochrome c reduction to monitor superoxide generation indicate that two superoxide ions are indeed formed for each enzyme molecule reoxidized, as predicted by the scheme.
Simulations of the kinetics for superoxide-dependent cytochrome c reduction and enzyme reoxidation demonstrate that the fist one-electron step in the reoxidation scheme is the reaction of the two-electron-reduced intermediate with oxygen. The time courses expected from the above scheme for both enzyme reoxidation and superoxide generation are shown as the solid lines in Figs. 1 and 3. As observed by Olson et al. (6), the rate constants required to adequately fit the enzyme reoxidation data in the fast phase were approximately twice those determined graphically from a semilogarithmic plot of the data, due to the sequential nature of the reoxidation process. The individual rate constants used to simulate the absorbance change at 450 nm are given in Table I and are related to an intrinsic rate constant by the fraction of reduced flavin in the appropriate enzyme intermediate. It should be pointed out that, while each of the fist three steps have an associated Michaelis complex with oxygen, this need be included only to determine the effect of varying oxygen concentration on the intrinsic rate constant required to adequately fit the data. At a given oxygen concentration, each step in the reoxidation scheme is described by a single rate constant (12). Both the pronounced lag and the burst phase prior to the  limiting exponential process in the cytochrome c kinetics are faithfully reproduced in the simulations (Fig. 3 B ) . If the initial one-electron step is moved up in the scheme to the fourelectron-reduced intermediate, the simulation is noticeably poorer and attempts to improve the simulation by varying one or more of the rate constants only make the fit to the enzyme reoxidation data poorer (data not shown). In any case, such a scheme could not easily be reconciled with the results from experiments with two-electron-reduced enzyme, indicating that on oxidation both electrons in this intermediate produce superoxide.
The kinetics for reoxidation of the two-electron-reduced enzyme and the fully reduced enzyme complexed with alloxanthine can be submitted to the same type of kinetic analysis as described above. In each case, using the same set of rate constants as in the simulations of fully reduced enzyme, the fits to the data are excellent (Figs. 4 and 7, solid lines). In the case of the two-electron-reduced enzyme, the faster phase is simulated accurately by an exponential process with a rate constant of 20 s-', precisely 58% of the intrinsic rate (35 s") as expected from the fraction of reduced flavin in the twoelectron-reduced species (Fig. 4A). The fraction of slow phase (40%) agrees well with the ratio of extinction changes for twoand one-electron-reduced enzyme from Olson et al. (6). The kinetics of cytochrome c reduction are also accurately simulated, with half of the absorbance change in each of the two kinetic phases.
With the fully reduced enzyme complexed with alloxanthine, both the more exponential behavior and somewhat faster rate observed in the fast phase of reoxidation (relative to the case with fully reduced enzyme in the absence of alloxanthine) are fit well by the simulation (Fig. 7 A ) . In addition, the less extensive lag in cytochrome c reduction in the presence of alloxanthine is accurately reproduced in the simulation. These results argue strongly that the fully reduced enzyme in complex with alloxanthine behaves like the fourelectron-reduced enzyme during reoxidation, as expected from previous work (11) and that the four-electron-reduced enzyme is an intermediate in the reoxidation of the fully reduced enzyme.

CONCLUSIONS
The reduction of cytochrome c by superoxide anion has been successfully utilized to study the generation of the latter in the reaction of reduced xanthine oxidase with molecular oxygen. In agreement with expectations from a considerable amount of indirect evidence (5, 6), it is observed that superoxide is generated only in the last steps of enzyme reoxidation, with a stoichiometry of two superoxide ions produced for each enzyme molecule reoxidized. The overall sequence for the reoxidation of fully (six-electron-) reduced xanthine oxidase is 6 -P 4 -+ 2 + 1 + 0, where the numbers represent the number of electrons in each enzyme intermediate. The fist two steps represent two-electron oxidations to form hydrogen peroxide, and the last two steps represent one-electron oxidations to form superoxide. Our results resolve the earlier discrepancy between results predicted by such a scheme and the very low amounts of superoxide actually detected (6). Superoxide was quantitated in this earlier work by observation of its electron paramagnetic resonance signal in samples collected by the rapid freeze technique. This method could suffer from appreciable dismutation of superoxide, even at the relatively high pH at which the experiment was performed. The cytochrome c technique, on the other hand, maintains the free superoxide concentration at a sufficiently low steady state concentration to preclude dismutation. The determinations of stoichiometry appear to be accurate within 5%. Furthermore, the steady state superoxide levels remain SO low that cytochrome c reduction can be analyzed kinetically to obtain rates that accurately reflect superoxide generation.
In good agreement with the previous work (5), the individual rate constants for the four-step reoxidation scheme that best fit both enzyme reoxidation and cytochrome c data (reacting with atmospheric oxygen) were 35 s-', 33 s-*, 20 s -I , and 0.9 s-l for steps 1 through 4, respectively. These values accurately reflect the fractions of reduced flavin expected to be found in the six-, four-, two-, and one-electron-reduced enzyme based on a rapid equilibrium model (Table I). Each step but the last involves the reaction of oxygen with FADH2, a relatively rapid process. The last step is slow not only because of the small amount of electron density actually residing at the flavin site but also because the reaction necessarily involves the blue neutral semiquinone, a species generally found to be relatively unreactive toward oxygen (13).
The reason for the different behavior of two-electron-reduced xanthine oxidase compared to the six-or four-electronreduced enzyme is unclear. With respect to the reoxidation at the flavin site, the enzyme formally changes from an oxidase to an electron transferase (14). Olson et al. (6) proposed that the transfer of reducing equivalents takes place in one-electron steps and that rapid re-equilibration of electron density at the flavin nucleus in the nascent FADH.. . .OFto other parts of the enzyme (particularly the iron-sulfur centers) prevents further reduction of oxygen in the case of two-electron-reduced enzyme. Clearly, the oxidation state of the other sites plays an important role as this is the only difference between the six-and four-electron-reduced enzyme and that fraction of the two-electron-reduced enzyme population capable of reacting with oxygen (Le. the 60% containing FADH2), but in view of the current understanding of flavin activation of molecular oxygen an alternative explanation must be sought. It is likely that a covalent flavin-oxygen intermediate, similar or identical with the 4a-peroxyflavin formed with flavoprotein hydroxylases (9), is formed on reaction of FAD& with oxygen. The very similar dependence of the oxidation of six-and twoelectron-reduced enzyme on oxygen concentration argues that the same intermediate is formed in both cases and that only its subsequent breakdown is different in the two situations. In the six-and four-electron-reduced enzyme, the iron-sulfur centers would be reduced and the intermediate would be able to break down into oxidized flavin and Hz02 as in the case of simple flavoprotein oxidases. In the two-electron-reduced enzyme, however, the oxidized iron-sulfur centers could perhaps exert a sufficient influence on the electronic structure of the flavin hydroperoxide to cause it to undergo a homolytic rather than a heterolytic cleavage. The immediate products of the reaction would be flavin semiquinone and Oz-, with the reducing equivalent of the semiquinone rapidly equilibrating away from the flavin site. The subsequent slow reaction of flavin semiquinone with 02, typical of electron transferases, would account for the second equivalent of 0 ; produced. Alternatively, structural changes around the flavin might occur in going from the four-electron-reduced enzyme to the two-electron-reduced form, bringing about sufficiently large changes in the hydrogen-bonding arrangement between the protein and isoalloxazine ring to cause the flavin to alter its chemical behavior ( i e . switch from a strict oxidase to an electron transferase; Ref. 14). While these concepts are clearly speculative, they are more in line with current ideas on the mechanism of oxygen activation by flavins.