Evidence for a 1-electron oxidation mechanism in N-dealkylation of N,N-dialkylanilines by cytochrome P450 2B1. Kinetic hydrogen isotope effects, linear free energy relationships, comparisons with horseradish peroxidase, and studies with oxygen surrogates.

Many enzymes catalyze N-dealkylations of alkylamines, including cytochrome P450 (P450) and peroxidase enzymes. Peroxidases, exemplified by horseradish peroxidase (HRP), are generally accepted to catalyze N-dealkylations via 1-electron transfer processes. Several lines of evidence also support a 1-electron mechanism for many P450 reactions, although this view has been questioned in light of reported trends for kinetic hydrogen isotope effects for N-demethylation with a series of 4-substituted N,N-dimethylanilines. No continuous trend for an increase of isotope effects with the electronic parameters of para-substitution was seen for the P450 2B1-catalyzed reactions in this study. The larger value seen with the 4-nitro derivative is consistent with a shift in mechanism due to either a reversible electron transfer step preceding deprotonation or to a hydrogen atom abstraction mechanism. With HRP, the trend is to lower isotope effects with para electron-withdrawing substituents, due to an apparent shift in rate-limiting steps. Biomimetic model high-valent porphyrins showed reduction rates with variously 4-substituted N,N-dialkylanilines that were consistent with a positively charged intermediate; such relationships were not seen for anisole O-demethylation with P450 2B1. In contrast to the case with the NADPH-supported P450 reactions, high deuterium isotope effects (∼7) were seen in the N-dealkylations supported by the oxygen surrogate iodosylbenzene. With iodosylbenzene, colored aminium radicals were observed in the oxidations of aminopyrine, N,N-dimethyl-4-aminothioanisole, and 4-methoxy-N,N-dimethylaniline. With the latter compound, a substantial intermolecular deuterium isotope effect was observed for N-demethylation. In the N-dealkylation of N-ethyl,N-methylaniline by P450 2B1 (NADPH-supported), the ratio of N-demethylation to N-deethylation was 16. Although it is probably possible for P450s to catalyze amine N-dealkylations via hydrogen atom abstraction when such a course is electronically or sterically favored, we interpret the evidence to favor a 1-electron pathway with N,N-dialkylamines with P450 2B1 as well as HRP and several biomimetic models.

Many enzymes catalyze oxidations of amines (1,2). In some cases stable N-oxides are formed. If hydroxylation occurs at a carbon adjacent to nitrogen, the resulting product is (usually) an unstable carbinolamine, which decomposes to generate an amine and a carbonyl derivative. This process is widespread in the biochemistry of natural products and also occurs with many drugs and other xenobiotics (chemicals not naturally found in the body) (1). These enzymes include copper proteins, flavoproteins, non-heme iron proteins, and hemoproteins (2, 3). Efforts have been made to compare the mechanisms by which these enzymes catalyze N-dealkylations, and much of the interest has involved two major hemoprotein groups, the peroxidases and P450s. 1 At least four mechanisms are possible for the hemoproteins, as shown for an N-demethylation reaction: 3 3 R 2 NH ϩ HCHO ͑2͒R 2 N CH 3 3 ͑2͒RN . ϩ CH 3 3 R 2 N CH 3 ϩ R 2 N ϩ ϭ CH 2 3 R 2 NH ϩ HCHO R 2 N CH 3 3 R 2 N CH 2 ⅐ 3 R 2 N CH 2 OH 3 R 2 N H ϩ HCHO R 2 N CH 3 3 R 2 N . ϩ CH 3 3 R 2 N CH 2 ⅐ 3 R 2 N CH 2 OH 3 R 2 NH ϩ HCHO

REACTIONS I-IV
Reaction I has been considered for P450s (5,6) and shown to occur in some instances (7) but is probably not very general, since most N-oxides are relatively stable (8). Reaction II is generally accepted to be used by peroxidases, because oxidized peroxidases are able to accept electrons but are not generally recognized to be capable of transferring oxygen (9 -13). The second step in the pathway is nonenzymatic and generally rate-limiting; thus the aminium radical can be readily detected (13,14). Reaction II is not favored for P450s because the oxygen in the aldehyde is not derived from H 2 O (15). Reaction III is a classic P450 methyl hydroxylation, in which the oxygen in HCHO is derived from O 2 and the hy-droxylation step follows hydrogen atom abstraction. Reaction IV is a variant of Reactions II and III and has been our own preference (Scheme 1). It incorporates the oxygen labeling results with the known tendency of high-valent states of hemoproteins to accept a single electron from amines and other chemicals of low E 0 and has also been used to rationalize observed rearrangements (16), radicals trapped from 1,4-dihydropyridine oxidation (17), linear free energy relationships (18,19), several similarities with the electrochemical oxidation of amines (20), and low kinetic deuterium isotope effects (14,21). However, the basis of mechanistic conclusions in favor of Reaction IV over III (12,22,23) has been questioned because of a reported trend in the kinetic deuterium isotope effects of a series of four 4-substituted N,N-dimethylanilines (24 -26). The basis of conclusions regarding linear free energy relationships has also been questioned (25), in favor of Reaction III. We have examined several aspects of the kinetic hydrogen isotope effects seen with a set of 4-substituted N,N-dimethylanilines, using both P450 2B1 and HRP. Linear free energy relationships have been considered with P450 2B1 and biomimetic models for both N,N-dimethylanilines and anisoles. Oxygen surrogates (chemicals that directly activate P450s to high valent forms) (27) have also been used in studies on isotope effects and in production of visible aminium radicals with P450.

EXPERIMENTAL PROCEDURES
Chemicals-Mn III TPP-Cl and Cr III TPP-Cl were prepared by insertion of metals into TPP (Aldrich Chemical Co., Milwaukee, WI) and silicic acid chromatography as described (28). PhIO was prepared by alkaline hydrolysis of the diacetate (Aldrich) (29). CuOOH (Aldrich) was washed with aqueous NaOH as described (30).
N-Methylanilines were obtained from commercial sources when available. 4-Cyano-N-methylaniline was prepared from 4-aminobenzyl cyanide (Aldrich) by formation of the aminomethyl succinimide (m.p. 192-194°C, lit. 188 -190°C) and subsequent treatment with NaBH 4 / (CH 3 ) 2 SO (31), and the product was purified by preparative TLC (silica gel G containing fluorescent indicator, Sigma-Aldrich Techware, 1.0-mm layers, n-hexane-CHCl 3 -(C 2 H 5 ) 3 N, 5:4:1, v/v/v, 254-nm detection) and elution with acetone. N,N-Dimethylanilines were obtained commercially when available and purified by preparative TLC when necessary, using the above system for TLC and adjusting the ratio of hexane to (C 2 H 5 ) 2 O. The N-CD 3 (d 3 ) and N-CD 2 H (d 2 ) derivatives were generally prepared by treatment of the N-methylaniline with CD 3 I or CD 2 HI (Cambridge Isotopes, Cambridge, MA, Ն98% atomic excess abundance) in refluxing acetone/K 2 CO 3 (14,32) or, occasionally, tetran-butylamine chloride/benzene/H 2 O (33,34). The d 6 (N-(CD 3 ) 2 ) derivatives were prepared by similar methods. All of the deuterated preparations were purified by preparative TLC as described above and analyzed by analytical TLC in a similar system or by HPLC (see below). Repurification was done using preparative TLC until Ͼ99% purity was achieved. The atomic excess abundance was Ͼ98% expected for all deuterated N,N-dimethylanilines, as judged by mass (fast atom bombardment or GC introduction) or NMR spectroscopy. Most N,N-dimethylanilines were dissolved in (C 2 H 5 ) 2 O and converted to HCl or oxalate salts by the addition of dry (C 2 H 5 ) 2 O saturated with the acid, which were recovered by centrifugation and dried. The salts were dissolved in H 2 O for assays.
HRP reactions were done with H 2 O 2 as described (14). In reactions involving model metalloporphyrins, either Mn III TPP-Cl or Cr III TPP-Cl was dissolved in CH 2 Cl 2 and an excess of PhIO was added. When the color change was complete (1-2 min), the suspension was filtered in vacuo into a 10-fold excess volume of cold (Ϫ20°C) hexane, in order to precipitate the hypervalent metalloporphyrin complex (40). The material was collected by centrifugation (3 ϫ 10 3 ϫ g, 10 min), dried under an N 2 stream, and stored desiccated at Ϫ80°C. Portions were dissolved in CH 2 Cl 2 and held at Ϫ78°C until assays were to be done, when aliquots were diluted into CH 2 Cl 2 at 25°C.
Colorimetric determination of HCHO production was done using the Nash assay (39,42).
Estimation of competitive kinetic hydrogen isotope effects, expressed as D (V/K) using the convention of Northrop (43), was done using GCmass spectrometry. In the case of measurements with N-methylanilines produced in reactions, the products were extracted and converted to the trifluoroacetyl derivatives, which were analyzed using the chemical ionization mode (14). In all cases, selective ion monitoring was used and the reported ratios are of the integrals for the ions under consideration. In the case of the d 3 substrates, the competition is between two equivalent methyl groups, and D (V/K) is expressed simply as the ratio of the relative abundances (M ϩ 3/M), where M is the molecular ion (M ϩ 4/M ϩ 1 in the case of chemical ionization) (44).
When d 2 -labeled substrates are used there are four ways to break a C-H bond and two ways to break a C-D bond. When the N-methylani-SCHEME 1. Postulated pathway for P450-catalyzed tertiary amine N-dealkylation (shown for N-demethylation of an N,N-dimethylamine).
ine product is considered, three of the C-H bond cleavages will yield an M ϩ 2 product. One of the C-H cleavages and two of the C-D bond cleavages will yield the ion M. Thus, the ratio (M ϩ 2)/M can be expressed as where k is the rate of cleavage of a C-H or C-D bond. Solving for the ratio k C-H /k C-D gives which is considered as D (V/K). This approach is generally valid unless there is a high isotope effect, yielding high error due to a small denominator, or a high commitment to catalysis. The error in this approach is increased relative to the d 3 experiments because of the need to subtract, particularly when high isotope effects are considered, and this approach was not applied in the case of reactions catalyzed by HRP. Another approach was also used with the d 2 substrate. HCHO was trapped and derivatized with 4-nitrophenylhydrazine or 2,4-dinitrophenylhydrazine to form a hydrazone (45). Under the conditions used there was no detectable exchange of hydrogens with solvent when authentic d 2 -HCHO (Cambridge) was used in control experiments. Electron impact ionization was used in the GC-mass spectral analysis (46,47). In our preliminary experiments the contribution of the M ϩ 1 ion (near the theoretical contribution for 13 C) was considerably less than with the dimedone derivative (24). The M Ϫ 1 contribution was Ͻ0.2%, even at 70 eV, and calculations of atomic excess and D (V/K) were done as described by Biemann (48).
Rates of reduction of (MnO) 3ϩ TPP-Cl and (CrO) 3ϩ TPP-Cl complexes were measured by repetitive scanning in the wavelength range of 700 -500 nm, and the spectral changes were fit to first-order rate equations.

Kinetic Deuterium Isotope Effects in NADPH-supported P450
N-Demethylations-The isotope effects were measured with a series of 4-substituted N,N-dimethylanilines, using several approaches (Table I). The inherent error was lowest in the assays done with the d 3 substrates, and these values are considered most reliable. All of these values were in the range 1.56 -2.27 except for the 4-nitro derivative. The values recorded using the d 2 substrates were essentially identical, measuring the deuterium present in either the N-methylaniline or HCHO. No general trend in D (V/K) was noted over the range of -OCH 3 to -CN substituents, and the values obtained with 4-nitro-N,N-dimethylaniline were consistently higher (with both the d 2 and d 3 substrates) than any of the rest. Rates of these N-demethylation reactions of P450 2B1 in another study have been published elsewhere and fit to linear free energy relationships (18,19).
Kinetic Deuterium Isotope Effects in HRP-catalyzed N-Demethylations-HRP was considered as a model peroxidase because of its known high isotope effects in N-dealkylations (14,44) and the strong evidence that this hemoprotein catalyzes N-dealkylation reactions by sequential electron transfer (9 -13) but cannot catalyze oxygenation or hydrogen abstraction from unactivated carbon atoms (12,49,50). We have not carefully measured the rates of all of these reactions, but the apparent k cat values do vary considerably, with the same trend as for P450 reactions (18,19) but more markedly. For instance, the rate of N-demethylation of N,N-dimethylaniline was 1.6 ϫ 10 4 min Ϫ1 (14) and the rate for 4-nitro-N,N-dimethylaniline was 1.4 min Ϫ1 (this work).
As reported previously (14,44), D (V/K) values were high for N,N-dimethylaniline, using either the d 3 or d 2 method (Table  I). High values were also observed with the 4-methoxy, -methyl, -chloro, and -bromo derivatives. However, values dropped considerably for the 4-cyano and particularly the 4-nitro derivatives, a pattern opposite to that reported for NADPHsupported P450 reactions (Table I).
Oxidations of N,N-Dimethylanilines and Anisoles by Biomimetic Models-High valent model metalloporphyrins can be prepared by treatment of metal TPP complexes with PhIO (40). 2 Reaction of (MnO) 3ϩ TPP-Cl with 4-cyano-N,N-dimethyl-2 The manganese and chromium compounds have advantages over iron in terms of stability. The high-valent complexes used here are designated (MnO) 3ϩ and (CrO) 3ϩ rather than defining sites of electron localization.  aniline in CH 2 Cl 2 led to the quantitative formation of HCHO as a product (after hydrolysis), and the kinetics of (MnO) 3ϩ TPP-Cl reduction and HCHO formation were identical (Fig. 1). We also demonstrated that (CrO) 3ϩ TPP-Cl could form HCHO in Ն90% yield (results not shown).
Rates of reduction of (MnO) 3ϩ TPP-Cl and (CrO) 3ϩ TPP-Cl by variously substituted N,N-dimethylanilines in CH 2 Cl 2 at 25°C yield free energy relationships that are very similar to those reported for P450 2B1-catalyzed N-demethylation reactions using either a Hammett (18) or a Marcus (19) relationship, consistent with the view that an intermediate with positive character (i.e. aminium radical) is formed in the reaction (Fig. 2). These relationships are very similar to those reported for accepted 1-electron oxidation processes, e.g. oxidation of aromatic amines by diphenylpicrylhydrazyl (51) and the oxidation of benzyldimethylamines by ClO 2 (52).
When these experiments were repeated with anisoles, the rates were much lower for both the (MnO) 3ϩ and (CrO) 3ϩ TPP-Cl complexes (results not shown), and no systematic relationship was seen for reduction. We also examined rates of anisole O-demethylation with P450 2B1 and found no obvious relationship between k cat and Hammett parameters (Fig. 3) (Table I). We considered the isotope effects in systems where more rapid generation of the highvalent P450 iron could be achieved, i.e. with oxygen surrogates instead of the usual NADPH-reductase/O 2 mode (19,54,55).
The isotope effects for both N,N-dimethylaniline and the 4-methoxy derivative increased in the CuOOH-supported system (Table II). In the system with PhIO, these values were increased to ϳ7. The isotope effect for the N-demethylation of 4-methyl-N,N-d 2 -dimethylaniline (-N-CD 2 H, CH 3 ) was found to be 6.32 Ϯ 1.05 in a P450 2B1/PhIO system, using deuterium analysis of HCHO.

Aminium Radical Formation in P450 Reactions
Supported by PhIO-To our knowledge, no direct detection of aminium radicals in a NADPH-supported P450 reaction has been reported, and we have not been successful in this regard either. However, such experiments have been done with HRP and other peroxidases and constitute some of the evidence for 1-electron oxidations with these hemoproteins (13). Griffin and her associates (56,57) have observed the formation of the characteristic aminopyrine radical in P450 reactions supported by CuOOH. The significance of this phenomenon has been unclear, however, since CuOOH and other alkyl hydroperoxides can be homolytically cleaved by some P450s to generate alkoxy radicals (58 -60).
We were able to reproduce the results of Griffin (56,57) with CuOOH, P450 2B1, and either aminopyrine, 4-methoxy-N,Ndimethylaniline, or N,N-dimethylamino-4-thioanisole (the N,N-dimethylaniline cation radicals without such anisole or thioanisole substituents have visible spectra (61) but are weak and less stable). Similar results could be obtained when CuOOH was replaced with PhIO, and the kinetic traces are shown in Fig. 4. With PhIO, there is no possibility of generation of alkoxy radicals. It should be emphasized that the level of aminium radical observed is a quasi-steady-state level, since production is continuing and heme is being destroyed. The colors disappeared after 1-2 min (or less, in the case of the thioanisole (Fig. 4B)). In other experiments that are not shown, the spectra were recorded and shown in all cases to match the broad blue ( max 650 nm) or purple ( max 550 nm) spectra observed with HRP. Only low levels of radical formation were seen in the absence of P450, and this low level could be further decreased by raising the pH to 8.0 and reducing the amine concentration (the small amount of radical is probably due to trace metals in the buffer system).
Replacement of 4-methoxy-N,N-dimethylaniline with its d 6 derivative (N-(CD 3 ) 2 ) gave a small change in the rate of aminium radical formation (Fig. 4). When the formation of 4-methoxy-N-methylaniline in this reaction was analyzed by HPLC, there was a considerable decrease in the rate (Fig. 5, apparent isotope effect ϳ10), although the final level of product was nearly identical. DISCUSSION Considerable evidence has been accumulated to support the role of 1-electron transfers in peroxidase reactions (9 -13). With HRP, it is possible to generate the individual compound I and II forms and to characterize products by visible and EPR spectroscopy. P450s are oxygenases with considerably more complex mechanisms, and direct evidence for radical intermediates has been difficult to obtain. There is general, but not universal, agreement that hydroxylations of conventional hydrocarbons occur via hydrogen atom abstraction and oxygen rebound mechanisms (12,27,62,63). In the case of N-dealkylation reactions, there are at least six major lines of evidence used in support of a role of 1-electron transfer mechanisms (Scheme 1). (i) Alkyl radicals can be trapped and characterized when 4-alkyl-1,4-dihydropyridines are oxidized by P450s (17). Although a role for adventitious iron in this process has been suggested (64), such oxidations can be demonstrated with purified P450s under conditions where free metals are removed (65). (ii) Strained cycloalkylamines are mechanism-based inactivators of P450s (16,66,67). Although the characterization of protein adducts has not been reported, known rearrangements of the substrates are most consistent with aminium radical intermediates (16). (iii) Amines and sulfur compounds, which both have low oxidation potentials, are oxygenated and dealkylated by P450s (34,68,69). All of these reactions can be rationalized by cation radical intermediates; e.g. sulfur cation radicals are more stable than aminium radicals and oxygenation is favored (22,34,69,70). Furthermore, many of the experiments done with amines (e.g. linear free energy relationships) have yielded similar results with sulfur compounds (34, 69 -71). Although the possibility exists that the N-oxygenation reactions occur via FeO 2 ϩ rather than (FeO) 3ϩ (41,72), no direct evidence for such an alternative has been reported, and we have been able to demonstrate several of these with PhIO, an oxygen surrogate capable of using FeO 3ϩ chemistry but not FeO 2 ϩ . 3 (iv) Evidence exists for rearrangements of non-amines that is diagnostic of 1-electron oxidation, e.g. strained cycloalkanes of low oxidation potential undergo characteristic rearrangements (68,73). Con-  versely, the effect of raising the oxidation potential of an amine (e.g. acylation) is consistent with a change in mechanism from a rate-determining electron transfer to a rate-determining proton or hydrogen abstraction mechanism (74). (v) Linear free energy relationships among para-substituted N,N-dialkylamines are consistent with the view of an intermediate bearing a positive charge (e.g. aminium radical). Both Hammett (18) and Marcus (19) relationships have been published for the P450 reactions. (vi) Kinetic deuterium isotope effects tend to be low for both alkyl and arylaminoalkylamine N-dealkylations, in contrast to the high intrinsic isotope effects seen for alkane hydroxylations (8,14,20,24,44,75). These results have been interpreted as evidence for an aminium radical intermediate (44), and some theoretical considerations have been offered in support of this view (76). A similar conclusion has been reached about the mechanism of the non-P450 hemoprotein Pseudomonas aminovorans secondary amine monooxygenase (3). The view that P450-catalyzed N-dealkylation proceeds via 1-electron oxidation has been questioned, essentially in regard to the last two points discussed above and relying on similarity in general patterns of isotope effects between P450s and one chemical model for hydroxylation (24,25). The interpretations presented in those articles are that (i) the intermediate RRЈ-N-CH 2⅐ should yield the same linear free energy relationships (with regard to phenyl 4-substitution, etc.) as a charged aminium radical (RЈR-N . ϩ -CH 3 ) and (ii) that a trend of reported isotope effects linked to increasing electronegativity of four phenyl para-substituents in P450-catalyzed N,N-dimethylaniline N-demethylation is consistent only with a hydrogen atom abstraction mechanism, the accepted mode for a t-butoxyl radical model system the authors use for comparison (25). The interpretation offered is that hydrogen abstraction occurs instead of 1-electron oxidation and, in particular, "This provides powerful evidence that P-450 reacts with all of the substituted N,N-dimethylanilines by a hydrogen abstraction mechanism" (italics from text) (25).
We have addressed the issue of kinetic hydrogen isotope effects in a series of 4-substituted N,N-dimethylanilines varying in electron-withdrawing properties from methoxy to nitro (Table I). Our previous interpretations of the isotope effects seen with N,N-dimethylaniline (14) are still considered valid. That is, the low hydrogen isotope effects seen with P450s arise because the rate-limiting step in the oxidation process (once the high-valent P450 is made) is 1-electron oxidation, and the postulated base-catalyzed deprotonation of the high pK a aminium radical and oxygen rebound are rapid events (14,77).
Small differences among kinetic isotope effects are difficult to judge, e.g. a difference in the competitive isotope effects of 2.6 and 3.6 is only that of 5% excess abundance of deuterium. Due to the error in the values of estimates of aminium radical pK a values (78), the meaning of small differences in isotope effects is also difficult to ascertain, particularly when only a small set of compounds is examined (24,25). Some previous work in this area was done using crude microsomal preparations (composed of many P450 enzymes) or else P450s unlikely to catalyze these reactions were used, and no rates were reported (24,25). In this work, we utilized three independent mass spectral approaches in our studies and found that most of our patterns were consistent (e.g. Table I). The error analysis in this type of work is somewhat misleading, in that variance is usually reported in the literature for multiple incubations, but the day-to-day agreement is not as good. Small differences among isotope effects generated using different approaches have sometimes been considered evidence for attenuation due to commitment to catalysis (24), but we did not see any consistent trends in our work (Table I). If there is such an element, presumably related to lack of rapid equilibration of the equivalent methyl groups, one would expect the pattern to be constant among such a group of highly related compounds.
The N-demethylations catalyzed by the prototypic peroxidase HRP were addressed in detail, because the evidence for single electron transfer with this enzyme is compelling (9 -13). As reported earlier (14,44,79), the hydrogen isotope effects were high for N,N-dimethylaniline (Table I). Similar high isotope effects were measured for the electron-donating and halide para-substituted derivatives, which all have relatively low oxidation potentials (19). The aminium radicals all have relatively high pK a values (78). With the electron-withdrawing cyano and nitro substituents, the isotope effects drop (Table I), along with the rates. Although the suggestion has been made that the range of kinetic hydrogen isotope effects for N-demethylation should show a parabolic pattern due to transition state symmetry changes and a slight effect is seen in an inorganic model (25), this is certainly not the case in this established peroxidase mechanism, and a clear distinction can be seen for the model used and the enzyme.
We are of the opinion that the explanation that we and others have proposed to explain the observed hydrogen isotope effects for P450-and peroxidase-catalyzed N-demethylation is still valid and can be used to rationalize the changes seen with various 4-substituted N,N-dimethylanilines (14). With P450s (and model metalloporphyrins), 1-electron transfer from the amine is rate-limiting in the oxidation of the substrate by the high-valent FeO 3ϩ species, and the resulting FeO 2ϩ complex can catalyze the rapid deprotonation of the aminium radical, which is followed by (rapid) oxygen rebound to the incipient radical. As the oxidation potential of an N-alkylamine increases (e.g. 4-nitro-N,N-dimethylaniline), 1-electron oxidation is less favorable and alternative processes may compete with electron transfer. An obvious possibility is that the direct hydrogen atom abstraction mechanism becomes competitive with electron transfer and partially contributes to the rate and kinetic isotope effects that are observed. An alternative possibility is that the electron transfer step begins to exhibit increased reversibilty, such that deprotonation of the intermediate radical cation contributes importantly to rate determination. This mechanism is depicted below (with k ET indicating forward and reverse electron transfer and k H indicating deprotonation). Thus, depending on the relative values for k ET1 , k ET Ϫ 1 and k H , the rate of substrate oxidation could be primarily a function of either k ET or k H or of both steps.
Our data do not enable distinction between these alternatives, i.e. competitive electron transfer and hydrogen transfer processes, or a single mechanism involving increasingly reversible electron transfer and deprotonation of the resultant radical cation. Whatever mechanism underlies the enhanced contribution of C-H bond breaking in rate determination, it must contribute only partially, so that the resulting kinetic isotope effect seen for 4-nitro-N,N-dimethylaniline N-dealkylation increases toward that seen for amide N-dealkylation (74) or anisole Odealkylation (see above) (34). This mechanistic shift is not surprising, since the E 0 for 4-nitro-N,N-dimethylaniline is 1.27 V and that for 4-methoxyanisole is 1.47 V (saturated calomel electrode) (19,34). For example, a switch in mechanisms between hydrogen atom abstraction and 1-electron transfer has been proposed in a substituted FeTPP/PhIO system with a series of benzyl trimethylsilanes varying in E 0 (and the solvent) (80). This effect overcomes any enhanced acidity of the aminium radical due to a lower pK a (78).
With HRP, the high isotope effect is the result of rapid formation of aminium radicals (evidenced by the accumulation of colored products, which have been quantified (14)) and an inability of the FeO 2ϩ complex (Compound II) to carry out base catalysis due to steric exclusion (14,49,81). Thus, intermolecular dismutation of the aminium radicals is rate-limiting, and the observed isotope effects are high (14). When electron-withdrawing substituents are attached to N,N-dimethylanilines, the rates of N-demethylation drop considerably and 1-electron oxidation, not deprotonation, becomes rate-limiting. Thus, the kinetic isotope effects decrease (Table I). The isotope effects for anisole O-demethylation are high, e.g. 10.1 Ϯ 1.4 for 4-nitroanisole in the usual NADPH-supported system (see above). 4 When linear free energy relationships are considered, the patterns seen with plots of k cat versus E 0 substrate (among the various dimethylanilines) for high-valent metalloporphyrins are similar to those seen for P450 2B1 (19) and qualitatively similar to those of HRP. The inverse relationship between k cat and E 0 substrate ) translates to negative values when log k cat is plotted against Hammett or ϩ values (18), and this relationship also applies to the model metalloporphyrins considered here. The conclusion has been made that such patterns cannot be offered as support for the involvement of aminium radical pathways, since the electronic character at the para position would be expected to be transmitted to a methylene radical as well (Ph-N-CH 2 ⅐ ) (24). If hydrogen atom abstraction is expected to show a negative value in a Hammett plot, then that should also be the case with anisole O-demethylation, a process considered by many to proceed via hydrogen atom abstraction, as evidenced by large isotope effects (see above) and not generally catalyzed by peroxidases (80,83). Through-bond transfer of electronic charge, if it occurs with Ph-N-CH 2 ⅐ , should also be expected with Ph-O-CH 2 ⅐ , although possibly attenuated. However, we have not observed either inverse dependence of log k cat of anisole O-demethylation with E 0 substrate ) or a linear Hammett relationship (with negative ) in any of our studies with P450 2B1 or with either of the two model metalloporphyrins used here (Fig. 3). Thus, linear free energy relationships that can be attributed to positively charged intermediates are seen in P450-and metalloporphyrin-catalyzed oxidation of amines (18,19) (Fig. 2) and thiols (34,70) but not in anisoles (Fig. 3), arguing for a distinctive mechanism.
Oxygen surrogates were utilized with P450 2B1 with the goal of overcoming the slow oxygen activation phase of the overall catalytic process. There is extensive evidence in the literature that most of the known P450-catalyzed reactions can be accomplished with a system in which P450 and PhIO are mixed (54,55,84), although differences in regioselectivity have been observed (55,85). The reactions are not easy to study because of the rapid degradation of the P450, but available evidence suggests that the same inherent mechanism of substrate oxygenation is operative in systems using PhIO (86). The mechanism of P450 does not shift to one of a peroxidase, because many oxidations at saturated carbon centers can be accomplished. In many cases, the substrate oxidation rates achieved in PhIOdependent systems are considerably higher than observed in NADPH-supported systems (18,19,54); linear free energy relationships are similar to those seen in NADPH-supported systems (18,19).
Kinetic isotope effects were considerably higher in P450catalyzed N,N-dimethylaniline oxidations supported by PhIO than NADPH, even with electron-donating para-methoxy derivatives (Table I). Furthermore, in an extension of the work of Griffin with CuOOH (56,57), the formation of colored radicals could be observed, as in the case of HRP (Fig. 4). The possibility exists that this formation of colored radicals is a side reaction and not associated with the normal N-demethylation mechanism. However, rates of formation are relatively fast (Fig. 4), even without any correction for enzyme destruction or radical decomposition. The demethylated product is formed with similar kinetics, and there is a large kinetic isotope effect on product formation (Fig. 5). The large isotope effects (Fig. 5, Table II) may be attributed to aminium radicals leaving the P450 active site (as observed) and dismutating to generate product, similar to the peroxidase mechanism. The question arises as to why the aminium radicals leave the active site and are not subject to base-catalyzed deprotonation/rearrangement, which we have postulated in the case of the normal, NADPH-supported reaction (14). The answer may lie in a P450-PhIO complex structure (Scheme 2) proposed by Ortiz de Montellano and Groves (87) to account for the rapid incorporation of 18 O from solvent water into products (84,88). A possible explanation is that activation of H 2 O-liganded P450 iron by PhIO may yield a slightly different compound I-like intermediate, already protonated (or possibly incorporating a H 2 O molecule), that is less capable of acting as a base to abstract an 4 A kinetic deuterium isotope of 1.9 has been reported for the electrochemical oxidation of ␣,␣-d 2 -dibenzyl ether (82), which necessarily involves a cation radical.
We conclude that all of the evidence presented here is consistent with and in favor of a sequential 1-electron transfer pathway of oxidative N-demethylation of many N-alkylamines by P450 and, by extension, thiols (69,70). No clear experimental evidence exists that a hydrogen atom abstraction mechanism should be favored (over 1-electron oxidation) for model metalloporphyrins, HRP, or P450 2B1. Changes in kinetic hydrogen isotope effects due to addition of para electron-withdrawing groups can all be interpreted for P450s and HRP (Table I). It is certainly possible for P450s to participate in hydrogen atom abstraction mechanisms with N-alkylamines when the juxtaposition and the stereoelectronic features of the substrate are more favorable (thus effectively raising the E 0 for P450-catalyzed electron transfer (19)) or when electron-withdrawing groups directly increase the E 0 substrate (e.g. 4-nitro-N,N-dimethylaniline, Table I). It is also possible that electron transfer may become reversible as the potential between the enzyme and substrate diminishes or becomes unfavorable, thereby making deprotonation partially rate-limiting (see above). We feel that a mechanistic commonality exists among all porphyrin systems considered to operate via FeO 3ϩ (or (metal-O) 3ϩ ) intermediates and that HRP would display similar properties if steric hindrance to substrate in the protein were removed (49). Whenever there exists a mechanistic choice with a substrate of comparatively low E 0 substrate , whether it be an amine (see above), thioether (69,70), phosphine (89), or strained hydrocarbon (68), 1-electron transfer is the preferred mechanism of oxidation by a high-valent hemoprotein or biomimetic model (22).
Recently Baciocchi et al. (90) reported that benzylaziridine, when oxidized either electrochemically, with FeTPP/PhIO or Ce IV or Co III salts, or with NADPH-fortified rat liver microsomal P450 preparations, yielded only products due to dimerization or tetramerization of a cation radical. The authors conclude that electron transfer is the dominant pathway in all of these systems.
Finally, there is another line of evidence strongly in favor of 1-electron oxidation mechanisms that has sometimes been overlooked. In earlier model studies, Shono et al. (20) noted not only the similarity of the kinetic deuterium isotope effects in electrochemical and microsomal systems but also the alkyl specificity, with demethylation being favored due to the preferential acidity of the methyl group in the putative aminium radical. This pattern was repeated in work with purified P450 2B1, where the substrate N-ethyl,N-methylaniline showed a ratio of demethylation:deethylation of 16 (41). This high ratio, repeatable and consonant with Shono's results (20), is not incidental. The result cannot be attributed to steric preference, since the total rates of N-dealkylation measured with N,Ndimethylaniline and N,N-diethylaniline were very similar (to each other and to N-ethyl,N-methylaniline). N-Demethylation would be preferred if an aminium radical were an intermediate (20) because of the enhanced acidity of a methyl proton compared with an ethyl (78). If hydrogen atom abstraction were the preferred route, N-deethylation should be favored because the inductive effect should stabilize a substituted alkyl radical (91), and such patterns of C-hydroxylation have long been recognized for P450s (92). Indeed, in our studies of the oxidative cleavage of esters by P450, a reaction that occurs with a very high isotope effect and is considered to probably involve hydrogen atom abstraction, deethylation and deisopropylation were preferred over demethylation (93). In various model chemical systems known to operate via 1-electron transfer chemistry, a preference for N-demethylation over dealkylation of a more substituted alkyl group is consistently seen (20, 52, 94 -97).
However, in the t-butoxyl radical system, the rate of hydrogen atom abstraction of an N-ethyl group was reported to be twice that for an N-methyl in a direct comparison (98). In both cases the patterns are interpreted in terms of stereoelectronic effects. This experiment (20,41) is compelling and constitutes another line of evidence (see above) in favor of a 1-electron mechanism for P450-catalyzed amine oxidation.