The Oxidation of N-Substituted Aromatic Amines by Horseradish Peroxidase*

The mechanism of N-dealkylation by peroxidases of the Ca2+ indicator quina and analogs was investigated and compared with the mechanism of N-dealkylation of some N-methyl-substituted aromatic amines. Nitro- gen-centered cation radicals were detected by ESR spectroscopy for all the compounds studied. Further oxidation of the nitrogen-centered cation radicals, however, was dependent upon the structure of the rad- ical formed. In the case of quina and analogs, a carbon-centered radical could be detected using the spin trap 5,5-dimethyl-l-pyrroline N-oxide. By using the spin trap ~-methyl-2-nitrosopropane(~ert-nitrosobut~e), it was determined that the carbon-centered radical was formed due to loss of a carboxylic acid group. This indicated that bond breakage most likely occurred through a rearrangement reaction. Furthermore, extensive oxygen consumption was detected, which was in agreement with the formation of carbon-centered radicals, as they avidly react with molecular oxygen. Thus, reaction of the carbon-centered radical with oxygen most likely led to the formation of a peroxyl radical. The peroxyl radical decomposed into superoxide that was spin trapped by 5,5-dimethyl- l-pyrro- line N-oxide and an unstable iminium cation. The iminium cation would subsequently hydrolyze to the monomethyl amine and for~aldehyde. In the case of N- methyl-substituted aromatic amines, carbon-centered radicals were not detected during the peroxidase-cat- was incu- bated with 1 ml of the Nash reagent (consisting of 15 g of NH~Ac, 0.2 ml of acetylacetone, and 0.3 ml of glacial acetic acid in 100 ml of H20) for 5 min at 58 "C. To correct for color formation that occurred during some reactions, a 1-ml sample was also incubated with the Nash reagent without acetylacetone. Absorption was measured at 412 nm. Oxygen consumption was determined using a Clark-type electrode. All these experiments were performed at room temperature, and incubation conditions are described in the figure legends.

The mechanism of N-dealkylation by peroxidases of the Ca2+ indicator quina and analogs was investigated and compared with the mechanism of N-dealkylation of some N-methyl-substituted aromatic amines. Nitrogen-centered cation radicals were detected by ESR spectroscopy for all the compounds studied. Further oxidation of the nitrogen-centered cation radicals, however, was dependent upon the structure of the radical formed. In the case of quina and analogs, a carboncentered radical could be detected using the spin trap 5,5-dimethyl-l-pyrroline N-oxide. By using the spin trap ~-methyl-2-nitrosopropane(~ert-nitrosobut~e), it was determined that the carbon-centered radical was formed due to loss of a carboxylic acid group. This indicated that bond breakage most likely occurred through a rearrangement reaction. Furthermore, extensive oxygen consumption was detected, which was in agreement with the formation of carbon-centered radicals, as they avidly react with molecular oxygen. Thus, reaction of the carbon-centered radical with oxygen most likely led to the formation of a peroxyl radical. The peroxyl radical decomposed into superoxide that was spin trapped by 5,5-dimethyl-l-pyrroline N-oxide and an unstable iminium cation. The iminium cation would subsequently hydrolyze to the monomethyl amine and for~aldehyde. In the case of Nmethyl-substituted aromatic amines, carbon-centered radicals were not detected during the peroxidase-catalyzed oxidation of these compounds. Thus, rearrangement of the nitrogen-centered radical did not occur. Furthermore, little or no oxygen consumption was detected, whereas formaldehyde was formed in all cases. These results indicated that the N-methyl-substituted amines were oxidized by a mechanism different from the mechanism found for quin2 and analogs.
In a previous paper, we have shown that the fluorescent Ca2+ chelators quin2 and its analogs are susceptible to peroxidase-mediated oxidation (1). QuinP and its analogs served as reducing cofactors for the hydroperoxidase activity of prostaglandin H synthase, undergoing oxidation in the process.
At the same time, arachidonic acid metabolism was stimulated. Oxidation of these compounds resulted in a decrease in * 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  quin2 fluorescence as well as a loss of its ability to bind calcium. These results indicated that one or more of the -N-CH&OOH groups, responsible for the binding of calcium, were oxidized by the peroxidase.
Little is known about the mechanism by which these Nsubstituted aromatic amines are oxidized by peroxidases. The oxidation of N-methyl-substituted aromatic amines, on the other hand, has been well documented. However, several mechanisms of N-demethylation have been proposed, and there is still some controversy about the exact mechanism. Galliani et al. (2) reported that horseradish peroxidase, the enzyme most commonly used in these studies, catalyzed the H2O2-dependent N-demethylation of a number of N,N'-dimethyl-and N,N'-dibutylaniline derivatives. They suggested a mechanism involving one-electron oxidation of the nitrogen followed by deprotonation (-H+) to form a neutral carboncentered radical (Scheme lA ). This carbon-centered radical can either be oxidized to the iminium cation, which is subsequently hydrolyzed to the monomethyl amine and formaldehyde (Scheme 1, pathway A ) , or it can react with oxygen to form a peroxyl radical, which then decomposes to the monomethyl amine and formaldehyde (Scheme I, pathway B ) .
Griffin et al. (3)(4)(5) showed that oxidation of N-methylsubstituted aromatic amines by horseradish peroxidase led to the f o~a t i o n of nitrogen-centered cation radicals and, in some cases, to a series of more complex reactions including dimerization to give a substituted benzidine. It was suggested that one-electron oxidation results in the formation of the nitrogen-centered cation radical, which loses a hydrogen atom to form an iminium cation. This leads to the formation of formaldehyde (Scheme 1, p a t h~a y C) or to radical dimerization (not shown), depending upon radicd stability of the parent compound and, most likely, also on enzyme, H202, and substrate concentration. Another possible mechanism has been proposed by Eling et al. (6,7) for the N-demethylation of aminopyrine. Aminopyrine was oxidized to a nitrogen-centered radical cation, which then ~s p r o p o~i o n a t e s t o t h e iminium cation and aminopyrine. The iminium cation is hydrolyzed to the monomethyl amine and formaldehyde (Scheme 1, pathway D).
In contrast, Wollenberg and co-workers (8)(9)(10)(11)(12) suggested that N-demethylation of N,N'-dimethylaniline by horseradish peroxidase involved hydrogen atom abstraction (-H. ) from the methyl group, leading directly to the formation of a neutral carbon-centered radical rather than to the formation of a nitrogen-centered radical cation (Scheme 1, pathway E ) .
One-electron oxidation would subsequently lead to the formation of the iminium cation and formaldehyde.
The mechanism of N-demethylation is obviously not clearly defined and might furthermore depend upon the compound studied. The mechanisms outlined in Scheme 1 provide various possibilities for the mechanism of the oxidation of the H-0 SCHEME 1. Mechanisms possible for the oxidation of N-substituted aromatic amines. R, substituted aromatic ring.

CH,
more complex N-substituted aromatic amines like quin2 and its analogs. We therefore decided to study the formation of the various free radical intermediates in the oxidation of these compounds by horseradish peroxidase and Hz02. We have used the technique of electron spin resonance spectroscopy, which enabled us to detect. moderately stable radicals directly (nitrogen-centered cation radicals) and to detect reactive radicals by spin trapping (carbon-centered radicals) in order to try to elucidate the mechanism of oxidation. termed half-dimethyl BAPTA, was synthesized according to Wojcik and Ostrich (13). Other chemicals were of the highest grade available. ESR spectra were obtained using an IBM ER-200 spectrometer operating at 9.7 GHz with a 100-kHz modulation frequency, equipped with an ER-4103 TM cavity. The solutions were transferred to the quartz flat cell by means of a rapid sampling device (14). Flow experiments were conducted with a quartz fast-flow mixing chamber flat cell obtained from Wilmad Glass Go., Buena, NY (type WG-Q4, modified flat cell, 17-mm width). Reagents were prepared in two 2liter bottles. Outlets at the bases of the bottles were connected to the inlets of the flat cell with Tygon tubing. Gravity flow from a height of 2 meters was regulated by Gilmont compact flow meters. All ESR experiments were performed at room temperature.

5,5'-
All spectra were acquired and analyzed using an H P 9000 computer interfaced to the spectrometer. Initial hyperfine splitting constants were obtained using the AUTOCORRELATION routine and were refined to their final values by the TUNE routine, as described previously (15). When necessary, hyperfine assignments were made using the A~TOSIMULATION routine (16) applied in Fourier space (17). The Fourier transform of the ESR signal produces a series of frequency vaIues and can be simuiated much more efficiently than t.he original spectrum. A first derivative spectrum that. has been centered in the magnetic field domain and exhibits little or no decay will produce an entirely imaginary component after a Fourier transform (17). We used this component in our analysis and display its plot in several figures. In addition, a Fourier transform reduces the The abbreviations used are: BAPTA, 1,2-bis(2-aminophenoxy)ethane-N,N,N',~'-tetraacetic acid; DMPO, 5,5-dimethyl-l-pyrroline N-oxide; t-NB, 2-methyl-2-nitrosopropane (tert-nitrosobutane); half-dimethyl BAPTA, N,N'-di(carboxyImethyI)-4-met~~ylaniline; W, watts; a, hyperfine splitting constant.

CH3
number of observable lines in the spectrum by compressing the information into a relatively low frequency range and thus simplifies comparison between experimental and simulated spectra.
Formaldehyde formation was measured as described previously (1). Briefly, after incubation with horseradish peroxidase and HzOp, the enzyme was precipitated using &SO4 and Ba(OHI2, and the sampfes were centrifuged. Subsequently, 1 ml of the supernatant was incubated with 1 ml of the Nash reagent (consisting of 15 g of NH~Ac, 0.2 ml of acetylacetone, and 0.3 ml of glacial acetic acid in 100 ml of H20) for 5 min at 58 "C. To correct for color formation that occurred during some reactions, a 1-ml sample was also incubated with the Nash reagent without acetylacetone. Absorption was measured at 412 nm.
Oxygen consumption was determined using a Clark-type electrode. All these experiments were performed at room temperature, and incubation conditions are described in the figure legends.  Fig, 1A. When N,NJ-dimethy~-p-toluidine, H202, or horseradish peroxidase was omitted from the incubation mixture, no ESR signals could be observed (results not shown). The spectrum was analyzed by using a computer correlation technique (15)(16)(17). Hyperfine splitting constants were found for all nuclei with spin in the radical cation, and they are listed in Table I In the case of N,N'-dimethylaniline, the nitrogen-centered cation radical could not be detected under the conditions used for N,N'-dimethyl-p-toluidine due to the instability of this radical. However, by using the ESR fast-flow technique, we were able to detect the nitrogen-centered cation radical of N,N'-dimethylaniline ( Fig. 2 A ) . No signal could be detected when N,N'-dimethylaniline, HzOz, or the enzyme was omitted from the reaction mixture (results not shown). The assignment of the hyperfine splitting constants was accomplished I?-Dealkylation of Aromatic Amines by computer simulation (Table I), and the simulation is shown in Fig. 2B. The Fourier transform of the experimental spectrum and the simulation are shown in Fig. 2, C and I), respectively.

The O~idation
The Oxidation of N-CH2-COOH-substituted Aromatic Amines: Formation of Nitrogen-centered Cation Radicals-In this study, half-dimethyl BAPTA and 5,5'-dimethyl BAPTA were used as model compounds for the Ca'+ chelator quin2 and its analogs. Half-dimethyl BAPTA has a structure similar to that of N,N'-dimethyl-toluidine, with the two -CHz-COOH groups replacing the -CH3 groups. Incubation of halfdimethyl BAPTA with horseradish peroxidase and H202 at pH 4.0 resulted in the detection of a nitrogen-centered radical cation, as shown in Fig. 3A. When half-dimethyl BAPTA, horseradish peroxidase, or H20z was omitted from the reaction mixture, no ESR signal could be detected (results not shown). The radical detected for half-dimethyl BAPTA was rather unstable and decayed during the scan. To improve the signalto-noise ratio, several scans were accumulated. Although halfdimethyl BAPTA has a structure similar to that of N,N'dimethyl-p-toluidine, it was not possible to simulate the observed spectrum using the hyperfine splitting constants found for N , N ' -d i m e t h y l -~-~l u i d~e with four protons instead of six. We therefore decided to perform the experiment in 'H20 buffer, pH 4.0, to see whether one or more protons attached to the carboxylic acid groups could be detected under our experimental conditions. As can be seen in Fig. 4A, the spectrum changed considerably when 'HzO buffer was used, indicating that under these conditions, protons attached to the carboxylic acid groups could be detected. Using the computer correlation technique, we were able to obtain an additional hyperfine splitting constant of approximately 1.6 G for a deuterium ion, which implies a hyperfme splitting constant of approximately 10 G for the corresponding proton (Table  I). By using the hyperfine splitting constants found for N,N'dimethyl-~-~luidine and an additional hyperfine splitting constant of 10.0 G for two protons attached to the carboxylic acid groups, we were able to simulate the spectrum of halfdimethyl BAPTA (Fig. 3B). Fig. 4B shows the s i m~a t i o n for the spectrum detected in 'HZO buffer. The splitting constants are listed in Table I. The Fourier transform from the spectrum obtained in regular buffer is shown in Fig. 3C (experimental) and 3D (simulation) and from the spectrum in 'H20 buffer in Fig. 4C (experimental) and 4 0 (simulation).
Incubation of 5,5'-dimethyl BAPTA with H20a and horseradish peroxidase at pH 4.0 resulted in the formation of a nitrogen-centered radical cation as shown in Fig. 3E. No ESR spectrum was detected when 5,5'-dimethyl BAPTA, horseradish peroxidase, or HzOz was omitted from the reaction mixture (results not shown). Although the spectrum was similar to that found for half-dimethyl BAPTA, we were unable to analyze this spectrum. Comparison of the hyperfine splitting constants found for the protons on the phenyl ring shows that for all compounds studied, these hyperfine splitting constants are very similar (Table I). Variation occurs in the hyperfine splitting constants obtained for the nitrogen and for the protons from the -CH2/-CH3 groups attached t o the nitrogen. However, by using these hyperfine splitting constants, we were still unable to simulate the spectrum obtained for the nitrogen-centered radical from 5,5'-dimethyl BAPTA (Fig. 3E). The presence of the -0-CHZ-R group attached to the phenyl ring most likely causes significant changes in the hyperfine splitting constants.
In the presence of higher concentrations of 5,5'-dimethyl BAPTA, H20z, and enzyme, another spectrum could be detected (results not shown). Although we were unable to simulate this spectrum, we think it might be the result of a dimerization reaction, as was observed previously for N-   Table I. Line width, 0.6 G . C, Fourier transform of the experimental spectrum shown in Fig. 2A. D, Fourier transform of the computer simulation. methyl-substituted compounds by Griffin et al. (5). Incubation of BAPTA, 5,5'-difluoro BAPTA, or quin2 with horseradish peroxidase and H202 resulted in the formation of spectra similar to those detected when 5,5'-dimethyl BAPTA was incubated with high concentrations of enzyme and HzOz, we were not able to analyze these spectra (results not shown).
The ESR, we were able to detect the carbon-centered radical at lower substrate and enzyme concentrations. Furthermore, under the conditions used in our experiments, the addition of H202 was not required to produce the radical (Fig. 5D), although addition of H202 did enhance the intensity of the carbon-centered radical adduct spectrum. This signal increased in time (results not shown). No ESR signal was found when horseradish peroxidase, 5,5'-dimethyl BAPTA, or DMPO was omitted from the incubation mixture (Fig. 5, C, E, and F).  Table I. Linewidth, 0.35 G . C, Fourier transform of the experimental spectrum shown in Fig. 3A. 0, Fourier transform of the computer simulation. E, the sample contained 2.8 mM 5,5'-dimethyl BAPTA, 1.25 mM HzO,, and 3.75 pg/ml horseradish peroxidase in 0.2 M sodium acetate buffer, pH 4.0. The incubation was performed under a nitrogen atmosphere. Instrumental conditions were: microwave power, 20.9 mW; modulation amplitude, 1 G; time constant, 0.66 s; scan range, 160 G; scan time, 500 s. 14 scans were accumulated.
information about the trapped carbon-centered radical could be obtained by isotope labeling of the compounds at specific positions. 5,5'-Dimethyl BAPTA was 13C-labeled at the CH, moiety in the acetate side group, and it was shown that the radical was formed at this 13C-labeled position (1). Incubation of 13C-labeled 5,5'-dimethyl BAPTA with horseradish peroxidase and Hz02 at p H 4.0 gave similar results (Fig. 5B). The radical is trapped at the same position, and the presence of The second small radical adduct signal that was detected is the DMPO-superoxide radical adduct (Fig. 6A, marked by arrows). Addition of superoxide dismutase (130 pg/ml) inhibited the formation of the DMPO-superoxide radical adduct  Table I. Line width, 0.5 G. C, Fourier transform of the experimental spectrum shown in Fig. 4A. D, Fourier transform of the computer simulation. almost completely (Fig. 6C). Hyperfine splitting constants for the DMPO-superoxide radical adduct are (aN = 14.3 G, a: = 11.2 G, and a: = 1.4 G. The third species that was detected in the spectrum comes from the DMPO-hydroxyl radical adduct ( Fig. 6 A , marked by *). The DMPO-hy~oxyl radical adduct (aN = 15.0 G and a2 = 15.0 G) is most likely a decomposition product of the DMPO-superoxide radical adduct (19) since the addition of superoxide dismutase also inhibited the formation of this radical adduct. The very weak superoxide dismutase-insensitive component may be due to the trapping of a peroxyl radical, which decomposes to superoxide (20). The complete simulation is shown in Fig. 6B.
Incubation of N,N'-dimethyl-p-toluidine or N,N'-dimethylaniline with horseradish peroxidase and H202 in the presence of DMPO under identical conditions did not result in the detection of a carbon-centered radical. The experiment was also performed using a wide range of substrate and enzyme concentrations at pH 4.0 (results not shown). When higher enzyme and Hz02 concentrations were used, no carboncentered radical could be detected in the reaction between these compounds and horseradish peroxidase, although nitrogen-centered radicals could be detected in some cases. These results indicated that these aromatic amines were oxidized by horseradish peroxidase to nitrogen-centered radicals, with little or no formation of carbon-centered radicals.
Identification of the Carbon-centered Radical Using the Spin Trap t-NB-A major disadvantage of using a nitrone spin trap like DMPO is the lack of structural information obtained about the trapped radical. Although the site at which the carbon-centered radical is formed can be determined by using specifically labeled compounds, additional information about the structure of the trapped radical is required. Further infor- mation can be obtained using the nitroso spin trap t-NB, Nitroso compounds have a distinct advantage over nitrones because the reactive free radical attaches directly to the nitrogen atom of the spin trap, which will give rise to additional hyperfine splitting constants (21). Incubation of 5,5'-dimethyl BAPTA with horseradish peroxidase and HzOz in the presence of t-NB at pH 7.4 resulted in the spectrum shown in Fig. 7A. These experiments were performed at pH 7.4, because t-NB was more soluble at higher pH. When H20z was omitted from the reaction mixture, a small signal was detected, when 5,5'-dimethyl BAPTA, horseradish peroxidase, or t-NB was omitted from the reaction mixture, no ESR signal could be detected (results not shown). The spectrum shown in Fig. 7A is the result of two different radical species. The first species is due to the trapped carboncentered radical of 5,5'-dimethyl BAPTA, and its structure is shown in Fig. 7A. The large nitrogen-coupling constant arises from the nitroxide, and this hyperfine splitting constant of 16.1 G is typical for a spin adduct of a carbon-centered radical. Furthermore, a hyperfine splitting constant of 2.2 G for the nitrogen of 5,5'-dimethyl BAPTA is observed. The hyperfine splitting constants for the two protons (a21 = 10.3 G and u p = 7.8 G) are inequivalent, due to the conformational inequivalence of the two methylene protons. The simulation for this spin adduct is shown in Fig. 7C. The second spin adduct comes from t-butyl hydronitroxide, a species that can be formed due to the reduction of t-NB, with splitting constants uN = 14.3 G and u; = 14.9 G and is shown in Fig. 7D. The composite simulation is shown in Fig. 7B.
Formaldehyde Formation and Oxygen Consumption-From the experiments with the spin traps DMPO and t-NB, it was clear that incubation of 5,5'-dimethyl BAPTA with horseradish peroxidase and HzOZ resulted in the oxidative cleavage of the -N-CH2-COOH group, releasing the carboxylic acid group. In order to investigate the fate of the carbon-centered radical, we decided to look for the formation of formaldehyde. Previously we showed that incubation of quin2 and its analogs with prostaglandin H synthase and arachidonic acid resulted in the formation of formaldehyde and that this formaldehyde formation could be used as an indication of the oxidation of the -N-CH2-COOH group (1). Formaldehyde can be measured using a colorimetric method developed by Nash (22). The Nash reagent contains acetylacetone, which reacts with formaldehyde in the presence of excess amounts of ammonium salts to form a yellow product, diacetyl-dihydrolutidine, that can be measured quantitatively at 412 nm. Incubation of 5,5'dimethyl BAPTA with horseradish peroxidase (concentration varied from 0 to 2.5 pg/ml) resulted in the formation of formaldehyde, as illustrated in Fig. 8. The addition of H202 was not necessary for formaldehyde formation to occur. The  aN ( t -N B ) = 16.1 G, uN (BAPTA) = 2.2 G, u p =  10.3 G , and a? = 7.8 G; for species 2: aN = 14.3 G and a7 = 14.9 G. C. computer simulation of species 1 . 0 , computer simulation of species amount of formaldehyde formed increased with enzyme concentration until a maximum was reached at approximately 1.5 pg/ml of horseradish peroxidase.
Oxygen consumption was measured under identical conditions, and the results are shown in Fig. 8. Oxygen consumption increased with enzyme concentrations, at approximately 1 pg/ ml horseradish peroxidase, oxygen consumption was complete in 5 min. Furthermore, the amount of oxygen consumed/min increased with enzyme concentrations. The total amount of oxygen consumed was always less than the total amount of formaldehyde formed, but both reached a maximum at an enzyme concentration of about 1.5 pg/ml. When oxygen consumption was complete, a red color appeared in the incubation mixtures. When DMPO was added to the incubation mixtures, formaldehyde formation was inhibited by approximately 80%, and oxygen consumption was completely inhibited (results not shown).
The experiments were repeated for N,N'-dimethylaniline, N,N '-dimethyl-toluidine, and aminopyrine; the results are shown in Fig. 9, A-F. Incubation of these compounds with horseradish peroxidase without the addition of H20z did not result in the formation of formaldehyde, as was observed in the case of 5,5'-dimethyl BAPTA. Incubation of N,N'-dimethylaniline with horseradish peroxidase and H202 resulted in the formation of formaldehyde, and this formation was most affected by an increase in the H20, concentration (Fig.  9A). However, no oxygen consumption could be detected when N,N'-dimethylaniline was incubated with horseradish peroxidase and H202 under these conditions (Fig. 933)-When N,N'-dimethyl-p-toluidine was incubated with horseradish peroxidase and H202, formaldehyde could again be detected in the incubation mixtures (Fig. 9C). Formaldehyde formation was not affected significantly by an increase in enzyme con-centration, but increasing amounts of HZO2 produced an increase in the formaldehyde formation. Oxygen consumption was determined under the same experimental conditions and amounted to a maximum of 30-40 nmol of oxygen/ml of sample, which was about 15% of the total amount of formaldehyde formed (Fig. 9B). Oxygen consumption was hardly affected by an increase in enzyme concentration, but an increase in HzOz concentration caused a small increase in oxygen consumption. In the case of aminopyrine, a smaller amount of formaldehyde was formed, but no oxygen consumption was detected under these conditions (Fig. 9, E and  F ) . For all these compounds, the addition of DMPO did not significantly affect formaldehyde formation (data not shown).

DISCUSSION
As stated in the Introduction, various mechanisms have been proposed for the oxidation of N-substituted aromatic amines by peroxidases (Scheme 1, A-E). Comparison of the pathways outlined in Scheme 1 shows that the major difference between pathways A-E is that the initial radical formed in the pathways A-D is nitrogen centered, whereas the initial radical formed in pathway E is carbon centered. The results in this paper clearly demonstrate that nitrogen-centered cation radicals are detected from both the -CH,-and -CHz-COOH-substituted aromatic amines during peroxidase-catalyzed N-de&lkylation (Table I). A carbon-centered radical, on the other hand, was detected only in the case of 5,5'-dimethyl BAPTA and its analogs and not for N,N'-dimethyl-p-toluidine or N,N'-methyla aniline (Figs. 5-7). From the experiments with 5,5'-dimethyl BAPTA, it was clear that if carboncentered radicals were formed during the oxidation of N,N'dimethyl-p-toluidine or N,N'-dimethylaniline, they should be detected with the spin-trapping technique. Extensive oxygen consumption was measured during the horseradish peroxidase-catalyzed dealkylation of 5,5'-dimethyl BAPTA and analogs, whereas in the case of N,N'-dimethyl-p-toluidine, a small amount of oxygen consumption was observed (Figs. 8 and 9). No oxgyen was consumed during the oxidation of ~,N'-dimethylaniline or aminopyrine. Thus, there is general agreement between the ESR and the oxygen consumption data, as should be expected since carbon-centered free radicals react with molecular oxygen at near diffusion-limited rates (23). Formaldehyde formation, on the other hand, was detected for all the compounds, which indicates that N,N'dimethyl-p-toluidine, N,N'-dimethylaniline, and aminopyrine are N-demethylated by horseradish peroxidase and HzO, (Fig. 9). If the oxygen consumption had been stoichiometric with formaldehyde formation, it would have been easily detectable (Fig. 9). However, neither the ESR data (qualitatively) nor the oxygen consumption data (quantitatively) indicate that carbon-centered radicals are formed during the oxidation of N-methyl-substituted compounds by horseradish peroxidase and H202, with the possible exception of N,N'dimethyl-p-toluidine when detectable but less than stoichiometric oxygen consumption was found. This clearly demonstrates that different mechanisms of deal alkylation occur for -N-CH,-and -N-CHz-COOH-substituted aromatic amines. Any proposed carbon-centered radical intermediate would have to decompose very rapidly since its reaction with oxygen is known to be nearly diffusion controlled (23).
Hollenberg and co-workers (8-12) have studied extensively the oxidation of N-methyl-substituted aromatic amines by peroxidases. They measured formaldehyde formation as a result of the oxidation of N,N'-dimethylaniline and determined the effect of various spin traps on the formaldehyde formation (8). Although all the spin traps caused some inhi- bition, the magnitude of the inhibition was relatively small (10-20%), which agrees with our results. They were unable to detect spin adducts of nitrogen-centered radicals by ESR. Thus, they concluded that free radical intermediates like nitrogen-centered radicals did not play a role in the Ndemethylation reaction (8). However, in our hands, the direct detection of aromatic nitrogen-centered radicals with ESR was not affected by the addition of spin traps, implying that these delocalized aromatic free radicals do not react with spin traps. The results of Hollenberg and co-workers actually indicated that carbon-centered radicals, which do react with spin traps, were not involved in the reaction and do not exclude the formation of nitrogen-centered radicals, as they suggested (8). Thus, together with our ESR and oxygen uptake data, this indicates that initially, a nitrogen-centered radical and not a carbon-centered radical is formed upon the oxidation of N,N'-dimethyl-p-toluidine and N,N'-dimethylaniline by horseradish peroxidase and Hz02. This suggests that pathway 1E is an unlikely mechanism for the oxidation of these compounds. The results in this paper illustrate that the mechanism by which N-substituted aromatic amines are oxidized by horseradish peroxidase greatly depends on the structure of the nitrogen-centered cation radical formed. In the case of the nitrogen-centered cation radical of 53'-dimethyl BAPTA, bond breakage probably occurs through a rearrangement reaction (Scheme 2 ) . Identification of the carbon-centered radical was achieved by using the spin trap t-NB. These results indicate that the carbon-centered radical was formed due to loss of a carboxylic acid group, which, as carbon dioxide, is a good leaving group. Subsequently, the carbon-centered radical most likely reacts with oxygen to form a peroxyl radical. Peroxyl radicals, however, are difficult to detect by ESR, either directly or by spin trapping (24). The DMPO-superoxide spin adduct that was detected (Fig. 6) indicates that a peroxyl radical might be involved in the reaction since the superoxide probably arises from the decomposition of the peroxyl radical into an iminium cation and superoxide (20). The involvement of oxygen in the oxidation of 5,5'-dimethyl BAPTA is further illustrated by the consumption of oxygen during the horseradish peroxidase-catalyzed N-dealkylation. Incubation of 5,5'-dimethyl BAPTA with horseradish peroxidase led rapidly to oxygen consumption that was not dependent upon H202 addition (Fig. 8). The carbon-centered radical was also easily detected in the absence of H202 (Fig. 50). Trace amounts of H202, which are formed due to autoxidation of 5,5'-dimethyl BAPTA, are apparently sufficient to start the oxidation by horseradish peroxidase. The superoxide that is formed during the reaction will dismutate to H202 and thus provide H202 to support the oxidation further.
The iminium cation, which is formed upon decomposition of the peroxyl radical, is also unstable and will hydrolyze to the mono-substituted amine and formaldehyde. The amount of formaldehyde formed during the horseradish peroxidasecatalyzed oxidation of 5,5'-dimethyl BAPTA was always slightly higher than the total amount of oxygen consumed (Fig. 8), which indicates that formaldehyde formation might occur through more than one pathway. These results suggest that the major pathway for dealkylation requires oxygen (Scheme 2) but that some formaldehyde might be formed through a mechanism that is oxygen independent (pathway IA).
In the case of our N-methyl-substituted amines, rearrangement of the nitrogen-centered radical cation to a carbon-centered radical is unlikely to occur since there is no good leaving group available. Thus, there are several possible pathways for the further oxidation of the nitrogen-centered radical, two of which are outlined in Scheme 1, C and D.
Furthermore, studies by Slaughter and O'Brien (25) on the oxidation of N,N'-dimethylaniline by horseradish peroxidase and H2O2 indicated that the oxidation might proceed along three separate pathways: N-demethylation to N-methylaniline and formaldehyde, dimerization to N,N,N',N'-tetramethyl-p-p'-benzidine, and another dimerization to an unidentified water-soluble product, possibly an N-oxide. This suggests that further studies on the products, formed together with formaldehyde, might make it possible to elucidate the mechanism by which these N-methyl-substituted aromatic amines are oxidized.
In conclusion, the results in this paper clearly demonstrate that N-substituted aromatic amines are good electron donors and thus are easily oxidized in a system containing horseradish peroxidase and HzOz. One-electron oxidation leads to the formation of nitrogen-centered cation radicals. The subsequent mechanism of N-dealkylation, however, is dependent upon the structure of the nitrogen-centered radical formed. When the N-alkyl side group contains a good leaving group, like the carboxylic acid group in 5,5'-dimethyl BAPTA and analogs, bond breakage occurs through a rearrangement reaction, with the formation of a carbon-centered radical (Scheme 2). When a good leaving group is not available, the mechanism for the oxidation of the nitrogen-centered cation radical is still unknown, and further studies are required to elucidate this mechanism fully.