Evidence for a Homolytic Mechanism of Peroxide Oxygen-Oxygen Bond Cleavage during Substrate Hydroxylation by Cytochrome P-450*

As a “peroxygenase,” purified liver microsomal cy- tochrome P-450 catalyzes the peroxide-dependent hydroxylation of a variety of substrates in the absence of NADPH, NADPH-cytochrome P-450 reductase, and molecular oxygen, with the peroxide serving as the source of the oxygen atom in the hydroxylated substrate. The nature of the 0-0 bond-cleaving step has been studied with peroxyphenylacetic acid (PhCH2C03H), which would undergo decarboxylation upon homolytic cleavage because of the formation of the unstable carboxyl radical (PhCH2C020) but not upon heterolytic cleavage to yield phenylacetic acid and Compound I by the clas- sical peroxidase route. Isozymes of rabbit liver microsomal cytochrome P-450 induced by phenobarbital and 5,6-benzoflavone and camphor-hydroxylating cytochrome P-450 from Pseudomonas putida bring about the decarboxylation of peroxyphenylacetic acid with the formation of benzyl alcohol; toluene and other possible decarboxylation products were not detected. The peroxy acid was shown by anaerobic and experiments to be the source of the oxygen atom in the benzyl alcohol. Various substrates (cyclohexane, cumene, ethylbenzene, and cam- phor) undergo aliphatic

from the hydroxylatable substrate. Hydroxyl radical equivalent coordinated to the iron then combines with benzyl radical to form benzyl alcohol or with the carbon radical derived from the hydroxylatable substrate to generate the substrate alcohol.
In 1973, Kadlubar et al. (1) showed that liver microsomes were able to perform tertiary amine dealkylations in the presence of alkyl hydroperoxides in a manner similar to the NADPH-and O5-dependent dealkylations already known to be carried out by microsomal suspensions. The evidence supported the supposition that the peroxide reaction was mediated by cytochrome P-450. Later, Nordblom et al. (2), using P-450~~,,' and Sligar et al. (3), using P-450,,,, showed that cytochrome P-450 could, indeed, catalyze the hydroxylation of substrates in the presence of various hydroperoxide derivatives (Equation 1). NADPH, 02, and NADPH-cytochrome P-450 reductase were not required in these oxidations, in which cytochrome P-450 functions as what we are calling a "peroxygenase." RH + ROOH 4 ROH + R O H (1) Many groups (2-9) have suggested or assumed a mechanism for the peroxide-dependent reactions similar to that proposed for the peroxidases, in which an intermediate designated Compound I and equivalent to the perferryl ion (Fe"=O) dehydrogenates a substrate (AH,) in two 1-electron steps, yielding Fe"', HzO, and A. Compound I is formed by 2-electron oxidation of the heme, which, to occur in one step, requires heterolytic cleavage of the peroxide bond, RO-OH -+ RO-+ OH'. Cytochrome P-450 could hypothetically also form a Compound I with hydroperoxides. The P-450 Compound I would then hydroxylate rather than dehydrogenate the substrate to regenerate the ferric enzyme. However, several lines of evidence, such as facile hydroxylation by iodosobenzene diacetate (9), strong variation of regioselectivity of hydroxylation among different oxidants (9, lo), and the inability of stoichiometric quantities of hydroperoxides to convert P -4 5 0~~~ t o a Compound I (11) cast doubts on the veracity of the peroxidase-like mechanism, as discussed elsewhere (12).
We have devised an alternative to the classical peroxidasetype mechanism for P-45O-promoted, peroxide-dependent hydroxylation reactions (12). The alternative mechanism differs fundamentally from the peroxidase-type mechanism in that it proceeds through homolytic rather than heterolytic cleavage of the peroxide bond. Thus, rather than a perferryl ion, we propose an oxy radical and a thiyl ferric hydroxide complex (S'Fe"'OH-) as the active oxygen species, with some similarity to the Fenton reaction. The new mechanism postulates a discrete role of the thiolate ligand to the P-450 heme, namely as a 1-electron donor to the peroxide via iron. The oxy radical abstracts hydrogen from substrate and the nascent substrate carbon radical quickly recombines with the adjacent thiyl ferric hydroxide complex, which is 1-electron-oxidized relative to the native ferric enzyme and is, therefore, equivalent to peroxidase Compound TI. Dunford (13) has recently postu-' The abbreviations used are: P-450" and P -~~O L M , , isozymes of rabbit liver microsomal cytochrome P-450 induced by phenobarbital and 5,6-benzoflavone, respectively; P-450,,,, camphor-hydroxylating cytochrome P-450 from P. putida; PPAA, peroxyphenylacetic acid. lated the involvement of Compound I1 in P-450 reactions.
Distinction of the homolytic from the usual peroxidase-type mechanism depends on the main difference, the peroxide bond-breaking step. If the original peroxide were a peroxy acid (RC03H), then the 0-0 bond lysis would produce a carboxyl radical (RC02.), which is more or less prone to decarboxylation, depending on the stability of the daughter alkyl radical. Since benzyl radical is resonance-stabilized, then peroxyphenylacetic acid (PPAA) might decarboxylate if phenylacetoxy radical (PhCH2C02 .) were a reaction intermediate. Possible decarboxylation of peroxy acids is predicted by the homolytic mechanism and not by the heterolytic mechanism of 0-0 bond scission. Therefore, we have investigated the P-450-promoted hydroxylation of substrates by peroxyphenylacetic acid. We now report the production of benzyl alcohol during the reaction, formed by decarboxylation of phenylacetoxy radicals. This observation is strongly suggestive of homolysis of the 0-0 bond as the first step in P-450-catalyzed substrate hydroxylation by peroxides.

EXPERIMENTAL PROCEDURES
Electrophoretically homogeneous rabbit liver microsomal P -4 5 0 , >~~ and P-4501.~~ were prepared by the method of Coon et al. (14). The preparations used had specific contents of 17.7 and 18.2 nmol of P-450/mg of protein, respectively. P-450,., was isolated from Pseudomonas putzda as described by Gunsalus and Wagner (15). Water containing 99% was obtained from Prochem and perdeuterocyclohexane was from Aldrich. Peroxyphenylacetic acid was synthesized by alkaline perhydrolysis (16) of phenylacetyl chloride (Aldrich). The crude reaction product in CH2C12 was extracted twice with 0.2 M potassium phosphate, pH 6.5, dried, and freed of solvent by evaporation to yield a waxy white powder in a yield of about 40%. Recrystallization from benzene-petroleum ether afforded mica-like crystals, m.p. 67-74°C (with decomposition). Iodometric titration (17) indicated this material to be greater than 97% pure. Peroxyphenylacetic acid has a penetrating, clinging stench, Reaction mixtures contained the cytochrome (2.2 nmol of P -4 5 0~~> , 5.2 nmol of P -~~O L M , , o r 10 nmol of P-450,,,), dilauroylglyceryl-3-phosphorylcholine (0.1 mg, deleted when P-450,,,, was present), potassium phosphate buffer (0.1 mmol, pH 7.4, with P-450LM2 and P-450LM4, or 0.05 mmol, pH 7.0, with P-450,.,), various substrates (sufficient to saturate the solution), and PPAA (variable amounts, added as a 100 mM solution in methanol) in a total volume of 1.0 ml. The vessels were capped and sealed during the incubation at 30°C. The reactions were started by injection of PPAA and terminated after 5 min by injection of 1 ml of chloroform. Residual peroxy acid was reduced by addition of 0.1 g of sodium bisulfite. After addition of internal standard, the organic layer was separated, concentrated to approximately 0.1 ml under an air stream, and submitted to gas chromatographic analysis. Quantitation of product was by internal standard, with manual peak integration. The identities of 2phenyl-2-propanol and benzyl alcohol were confirmed by gas chromatography-mass spectrometry, performed on a Hewlett-Packard 5985 instrument a t 160"C, using electron impact ionization a t 20 eV. For the determination of toluene, the reaction mixtures were extracted with 0.1 ml of chlorobenzene rather than chloroform, and the extract was not concentrated, but submitted directly to gas chromatography. Since toluene has a retention time shorter than chlorobenzene, no interference from the solvent was observed. Control experiments established that the recovery of known toluene by this procedure was essentially complete.
For the oxygen-labeling experiment, the medium was enriched to 10 atom R excess ''0 by dilution of 99% H2'"O. The contents of ''0 in product alcohols was estimated from their mass spectra using the ion pairs m / e 121 versus 123 (2-phenyl-2-propanol) and m / e 107 versus 109 and 108 uersus 110 (benzyl alcohol). For the anaerobic experiment, a reaction mixture containing all components except cumene and PPAA was deoxygenated by repeated evacuation and flushing with nitrogen in which the oxygen concentration had been reduced to less than 0.5 ppm by passage over a column containing reduced RASF catalyst (Ace Glass, Inc.) maintained at 120°C. Solutions of cumene and of PPAA in methanol which had been similarly deoxygenated were then injected sequentially via a microliter syringe through a septum to begin the reaction. Finally, deoxygenated chloroform was injected to terminate the reaction.

RESULTS
Benzyl radical produced via radical decarboxylation of PPAA could, at least conceivably, lead to several products, including toluene; benzyl alcohol, benzaldehyde, 1,2-diphenylethane, and benzyl phenylacetate. In actuality, benzyl alcohol was present in the product mixture but there were no more than traces of the other products mentioned. Benzyl alcohol was accompanied by an alcohol resulting from hydroxylation of the particular substrate being used. 2-Phenyl-2-propanol was produced from cumene, a-methylbenzyl alcohol from ethylbenzene, cyclohexanol from cyclohexane, and 5-hydroxycamphor from camphor.
Three purified forms of the cytochrome were used in the experiments summarized in Table I: P -4 5 0~~, , P -~~O L M , , and P-450,,,. All three induced decarboxylation of PPAA, but to different extents (Experiments 1, 2, and 11). Assuming that benzyl alcohol is the only product derived from decarboxylation of phenylacetoxy radicals and that the sum of benzyl alcohol and the substrate alcohol represents the amount of PPAA utilized by P-450, then decarboxylation occurred with 14 to 54% of the PPAA activated by the cytochrome. The values in Table I represent yields of products, not initial rates of product formation, since the reactions were apparently completed within a few minutes or perhaps a few seconds; product yields were the same when the reaction mixtures were allowed to run for 5 or 15 min before termination. The short duration of the reactions was probably due to rapid and irreversible oxidation of the heme. The reported yields, then, cannot be compared from experiment to experiment as a measure of relative rates under various conditions. However, within a given experiment, the relative yields should be equal to the ratios of the appropriate rate constants.
By comparison of Experiments 2 versus 3 and 11 versus 12, one may assess the reproducibility of the results, since these represent identical experiments performed weeks or months apart. Various control experiments are reported, including  4), and addition of the terminating reagent before the PPAA (No. 14). With cumene and camphor, small amounts of the hydroxylated products were present in the substrates as contaminants. The experiment which was the most appropriate control was subtracted from the other experiments in each group before calculation of percentage of decarboxylation (see footnotes, Table I). In all cases, the controls are of sufficiently low magnitude that the conclusions do not depend on whether these yields are subtracted. With [zHHlz]cyclohexane compared to ordinary cyclohexane, the overall yield of both products decreased, but the percentage of decarboxylation increased (No. 7 and 8). In Experiments 9 to 11, the total product yield increased with increasing PPAA concentration, but the percentage of decarboxylation remained essentially constant, in accord with the idea that the entire reaction is limited to the P-450 active site. When molecular oxygen was absent from the reaction mixture (Experiment 13), the only effect was a moderate increase (30%) in the yield of the substrate alcohol. The yield of benzyl alcohol, on the other hand, was unaffected. This is taken to mean that the oxygen atom in benzyl alcohol is not derived from molecular oxygen. The source of the oxygen atom in benzyl alcohol was shown not to be water by an experiment identical with No. 1 in Table I except that HZ'*0 was present. Analysis by gas chromatography-mass spectrometry showed that neither the 2-phenyl-2-propanol nor the benzyl alcohol contained "0. However, because of the magnitude of the background correction for the mass spectra in this experiment, it is possible that a low incorporation of I8O (~2 0 % ) would have escaped notice. The product ratios determined by this analytical technique were the same as in Table I. Interestingly, when no substrate was added to the reaction mixture (Experiment 19), the yield of benzyl alcohol did not rise to the previous total of benzyl alcohol plus substrate alcohol and, in fact, was not even as great as the amount produced when substrate was present. This suggests that the phenylacetoxy or benzyl radicals produced during the reaction may add to a susceptible double bond on the porphyrin as the first step in irreversible oxidation with consequent loss of catalytic activity. When a hydroxylatable substrate is present, the rate of addition to the porphyrin is retarded by competition of this substrate for the radical and perhaps also by physical obstruction of the approach to the porphyrin double bond.
The catalytic lifetime of P-450, expressed as the average number of catalytic cycles or enzyme turnovers which take place before the occurrence of an irreversible event such as radical addition to the porphyrin, was calculated from the data in Table I. In the absence of hydroxylatable substrate, P -4 5 0~~~ underwent irreversible oxidation by PPAA after only about two turnovers, whereas in the presence of substrate (cyclohexane, cumene, or ethylbenzene) the catalytic lifetime increased dramatically (to 6,14, and 22, respectively), thereby demonstrating a protective effect. Upon examination of experiments in which both PPAA and a hydroxylatable substrate were present, we observed a roughly inverse correlation between catalytic lifetime and percentage of decarboxylation. Specifically, in cases where the enzyme underwent few turnovers, the percentage of decarboxylation was relatively high, and vice versa. If phenylacetoxy were the major radical adding to the porphyrin, then the longer the time between its generation and its interception by the substrate, the more chance it would have to add to the porphyrin or to decarboxylate. /"c ", ,  DISCUSSION We interpret these results in terms of the scheme shown in generates thiyl ferric hydroxide complex and phenylacetoxy radical. Since the energetics of this reaction are unknown, we represent it as reversible. In addition to the possible back reaction (k2), phenylacetoxy radical has three modes of reaction. It may hydrogen-abstract from substrate (k3), decarboxylate to benzyl radical (k4) or, in the absence of a substrate molecule in the active site, add to a porphyrin double bond, leading to heme oxidation (kh,). If hydrogen abstraction occurs, then the substrate carbon radical may combine with the thiyl femc hydroxide complex (equivalent to Compound 11), reducing it to the native oxidation state and forming the substrate alcohol (ks). If the phenylacetoxy radical decarboxylates, then the benzyl radical itself combines with Compound I1 (ks), since it should behave in the same fashion as the hydroxylatable substrate radical.
The scheme in Fig. 1 may be used to rationalize all of the present results. For instance, k3/k4 is derivable from the percentage of decarboxylation and, in the case of oxidation of cumene by P-450L~,, would be about 4. The increase in percentage of decarboxylation when din-cyclohexane is hydroxylated compared to cyclohexane requires that phenylacetoxy radical be the hydrogen abstractor, or that some other intermediate (e.g. Compound I), which is the true hydrogen abstractor, be reversibly formed. For Compound I to be reversibly formed is at variance with all previous observations with peroxidases, as noted by Blake and Coon ( l l ) , and we have been unable to formulate an intermediate other than phenylacetoxy radical which might be both a good hydrogen abstractor and reversibly formed. Thus, the cyclohexane hydroxylation results may be due to the operation of an internal deuterium isotope effect on the ratio k3/k4. With cyclohexane, k3/k4 = 1.7, while with perdeuterocyclohexane k3/k4 = 0.85; the required isotope effect (k?/k?) is about 2.
A 30% increase was noted in the yield of 2-phenyl-2-propanol when the system was anaerobic. This suggests that a minor fraction of a solution-free radical chain reaction process appeared under anaerobiosis. In the absence of 0 2 , radicals which diffused away from the active site could initiate chain reactions among free substrate and PPAA molecules. However, in the presence of 0 2 , these solution chain reactions would be strongly inhibited because of the efficient capture of carbon radicals by O2 to give nonpropagating hydroperoxy radicals (ROO. ). Thus, aerobic reactions probably have insignificant contributions from solution chain reactions to the total product.
Of all the possible radical products deriving from PPAA, only benzyl alcohol is observed. If the entire reaction were confined to the P-450 active site, as in Fig. 1, then the probability of occurrence of products containing two benzyl groups would be low. The lack of hydrogen abstraction by benzyl radical to form toluene indicates rapid quenching (as, for example, by combination with Compound 11) before a suitable hydrogen donor could be encountered. The lack of effect of anaerobiosis on the yield of benzyl alcohol and the absence of benzaldehyde make it unlikely that benzyl radical exists very long after its generation, since it would very rapidly combine with dissolved molecular oxygen to give benzyl hydroperoxide. The intermediacy of benzyl hydroperoxide should lead to significant amounts of benzaldehyde.
Up to this point, we have considered only radical decarboxylation of PPAA. However, at least in theory, it is possible for decarboxylation of the PhCH2C02 group to occur by a radical, carbanion, or carbonium ion mechanism. The benzyl carbanion should lead exclusively to toluene. The benzyl carbonium ion should lead to benzyl alcohol, and the benzyl radical would produce benzyl alcohol and/or toluene. Since toluene is not observed, only carbonium ion and radical decarboxylations need be further considered. These can be distinguished by determining the source of the oxygen atom in benzyl alcohol. If it comes from water, the decarboxylation must involve benzyl carbonium ion. If the oxygen atom does not come from water, then it must have come from the peroxy acid, and the mechanism of decarboxylation must be radical. The experimental observation was that no "0 from labeled water was incorporated into the benzyl alcohol. We conclude, therefore, that the source of the oxygen atom in benzyl alcohol is the peroxy acid and that the decarboxylation is a radical process.
These results prove the occurrence of benzyl radical in the system and strongly suggest phenylacetoxy radical as the precursor. Phenylacetoxy radicals could result from either peroxide homolysis or I-electron oxidation of phenylacetate anions. The latter reaction is known to occur during the Kolbe electrolytic oxidation of carboxylic acids (18). Experimental as well as theoretical estimations of the standard reduction potential for acyloxy radicals agree on a value of +2.2 to 2.4 V (19). Since the corresponding value for hydroxyl radicals is +2.33 V (ZO), it seems unlikely that there will exist any oxidant in this system capable of removing an electron from phenylacetate anion. Phenomenologically, we may note that peroxidase Compound I does not oxidize carboxylate anions to the corresponding radicals, since Compound I may be generated by the stoichiometric equivalent of peroxy acid. Thus, the only reasonable source of phenylacetoxy radical is through homolysis of the peroxy acid. However, it is possible that benzyl radical arose by concert,ed I-electron reduction and decarboxylation of PPAA, so that phenylacetoxy radical was not a discrete intermediate. In that case, hydrogen abstraction would have to be performed by some other radical species (e.g. Compound I) in a separate reaction, since benzyl radical does not hydrogen-abstract (no significant amount of toluene is observed). Since this possibility cannot be ruled out for the present results, we have also included in Fig. 1 an alternate route in which Compound I is formed with rate constant kl. This heterolytic reaction is followed by successive hydrogen abstraction-radical recombination in an "oxygen rebound" scheme (21, 22). However, with this peroxy acid, despite the presence of the excellent carboxylate leaving group, the ratio kI/kI can be no greater than 1 in some cases, since 50% of the total product is benzyl alcohol ( Table I, Experiments 1 to 3). It is expected that the maximum potential contribution of the heterolytic branch when the leaving group is alkoxide, as with alkyl hydroperoxides, will be drastically reduced. Furthermore, the problems in postulating the involvement of Compound I, as alluded to earlier, still apply.
In summary, the present results on substrate hydroxylation and peroxy acid decarboxylation by cytochrome P-450 may be rationalized by a single homolytic scheme as shown in Fig.  1. In addition, this basic mechanism may be easily adapted to accommodate hydroxylation by iodosobenzene diacetate. Importantly, the actual hydrogen abstractor ( R O . ) is different from peroxide to peroxide; hence, the observed variation of regioselectivity (IO) is a natural consequence of the Fentonlike mechanism. While the present results cannot be considered definitive, they restrict the applicability of alternatives such as the classical peroxidase-type mechanism. The homolytic mechanism may be readily applied to the NADPH/02dependent reactions, since they almost certainly involve a peroxide intermediate (12). Indeed, recent work by Sligar et al. (23) strongly hints a t acylation of iron-bound dioxygen, so that the normal mechanism of 0 2 activation by P-450 may actually involve a protein-bound peroxy acid. Thus, it is possible for the peroxide-and NADPH/Oa-dependent reactions to give different product distributions (24) even though the mechanisms are quite similar.