Oxidation of N-Nitrosoalkylamines by Human Cytochrome P450 2A6

Cytochrome P450 (P450) 2A6 activates nitrosamines, including N,N-dimethylnitrosamine (DMN) and N,N-diethylnitrosamine (DEN), to alkyl diazohydroxides (which are DNA-alkylating agents) and also aldehydes (HCHO from DMN and CH3CHO from DEN). The N-dealkylation of DMN had a high intrinsic kinetic deuterium isotope effect (Dkapp ∼ 10), which was highly expressed in a variety of competitive and non-competitive experiments. The Dkapp for DEN was ∼3 and not expressed in non-competitive experiments. DMN and DEN were also oxidized to HCO2H and CH3CO2H, respectively. In neither case was a lag observed, which was unexpected considering the kcat and Km parameters measured for oxidation of DMN and DEN to the aldehydes and for oxidation of the aldehydes to the carboxylic acids. Spectral analysis did not indicate strong affinity of the aldehydes for P450 2A6, but pulse-chase experiments showed only limited exchange with added (unlabeled) aldehydes in the oxidations of DMN and DEN to carboxylic acids. Substoichiometric kinetic bursts were observed in the pre-steady-state oxidations of DMN and DEN to aldehydes. A minimal kinetic model was developed that was consistent with all of the observed phenomena and involves a conformational change of P450 2A6 following substrate binding, equilibrium of the P450-substrate complex with a non-productive form, and oxidation of the aldehydes to carboxylic acids in a process that avoids relaxation of the conformation following the first oxidation (i.e. of DMN or DEN to an aldehyde).

P450 3 enzymes are found throughout nature and catalyze many reactions, most of which are mixed function oxida-tions (4). The mammalian P450s are of considerable interest because of their roles in the metabolism of steroids, eicosanoids, drugs, chemical carcinogens, and other important molecules (5). The general mechanistic features of P450 reactions include substrate binding, reduction to the ferrous state, binding of O 2 , the addition of a second electron, protonation, and rearrangement to generate a reactive iron-oxygen complex poised near the substrate (6,7). The active complex can be described as a formal FeO 3ϩ entity, with similarity to Compound I of peroxidases, which can be used to rationalize most reactions (7)(8)(9), although some alternate possibilities can also be considered. A generally accepted mechanism for many P450 oxidations involves the abstraction of a hydrogen atom by the FeO 3ϩ entity, followed by "oxygen rebound" to yield a hydroxylated product (10).
Among the chemical carcinogens activated by mammalian P450s are N,N-dialkylnitrosamines (also called N-nitrosodialkylamines) (11), including those found in tobacco products and also the simple nitrosamines N,N-dimethylnitrosamine (DMN) and N,N-diethylnitrosamine (DEN) (12). The mechanism of activation is agreed to involve ␣-hydroxylation of the nitrosamine in most cases (Fig. 1). The process is generally accepted to involve hydrogen atom transfer instead of the alternate 1e Ϫ oxidation implicated for some amines (7,8,13,14) because of the high oxidation potential of the nitrogen atom due to nitrosation. Rearrangement of an ␣-hydroxynitrosamine results in the formation of an aldehyde and an alkyl diazohydroxide, the latter of which can alkylate DNA (possibly via an alkyl nitrenium ion or carbocation) ( Fig. 1) (11). This alkylation of DNA is generally accepted to be the basis of the tumor-initiating ability of these nitrosamines. In 1973, Keefer et al. (15) reported that perdeuterated DMN caused considerably fewer liver tumors in rats than protiated DMN. The deuterated material also yielded fewer methylated DNA adducts (16). Subsequent in vitro experiments with rat liver microsomes indicated that the conversion of DMN to HCHO was characterized by an increased K m but little change in V max (17)(18)(19).
One of the enzymes involved in the oxidation of DMN is P450 2E1 (20,21). This enzyme also oxidizes ethanol to acetaldehyde (22,23), characterized by a non-competitive intermolecular deuterium isotope effect of 5 on the K m but none on k cat with human P450 2E1 (24). These results were explained by the presence of a rate-limiting step following the formation of the product acetaldehyde, as clearly documented by the observed pre-steady-state kinetic burst of product formation. P450 2E1 indicated that the enzyme also oxidizes the first product, acetaldehyde, to acetic acid (25)(26)(27). Pulse-chase experiments and fitting to kinetic models indicated that although P450 2E1 does not have a high intrinsic activity for oxidizing acetaldehyde, the kinetic course of ethanol oxidation yielded limited exchange of the intermediate acetaldehyde with the medium (27).
We extended the previous work with ethanol to DMN because the original carcinogenesis studies on ethanol had been done with this compound (15,16). P450 2A6 was examined because of several general catalytic and other similarities with P450 2E1 (5,28,29) and the considerable amount of structural (30) and kinetic (31) information about this enzyme. P450 2A6 oxidizes both DMN and DEN, as does P450 2E1; the catalytic efficiency of P450 2A6 is higher than that of P450 2E1 for DEN oxidation (28,32). Several aspects of the design of the earlier P450 2E1 experiments with ethanol (24,27) were used as a framework for the present study, including the kinetic deuterium isotope effects.
Our results show kinetic deuterium isotope effects on both k cat and K m for the oxidation of DMN to formaldehyde, but the intrinsic isotope effect for DEN oxidation to acetaldehyde is considerably lower and is attenuated in non-competitive intermolecular experiments. As in the overall oxidation of ethanol to acetic acid by P450 2E1 (27), a fraction of the aldehyde (or ␣-hydroxynitrosamine) is not released from the enzyme prior to further oxidation to the carboxylic acid. These results have implications in consideration of the activation of carcinogenic N-nitrosodialkylamines by P450 2A6 but also more generally in multistep reactions catalyzed by P450 enzymes.  (27).

EXPERIMENTAL PROCEDURES
Synthesis-Unlabeled formaldehyde was synthesized from paraformaldehyde (Eastman Kodak Co.) to avoid complications resulting from organic solvents contained in commercially available formaldehyde solutions. A round-bottom flask containing 10 g of paraformaldehyde was heated to 125°C and purged gently with dry N 2 gas; the exiting vapors were passed through a bubbler containing 100 ml of H 2 O. The aqueous trap solution, containing absorbed monomeric formaldehyde, was passed through a Bakerbond quaternary amine 3-ml SPE column, and the formaldehyde was quantified following derivatization with 2,4-dinitrophenylhydrazone and HPLC (see below).
The general approach to synthesis of deuterated nitrosamine substrates involved LiAlD 4 reduction of various precursors to amines, followed by nitrosation (supplemental Figs. S1 and S2). The water solubility and volatility of the intermediates and products were issues in limiting the scale of the reactions. Also, final purity of the labeled DMN and DEN was not acceptable without distillation (because of these limitations in the work up of materials after chromatography), limiting the scale of the synthetic work to Ն0.5 g. The identity and purity of all substrates were established by NMR and MS (supplemental Figs. S3 and S4). Several of the compounds show complex NMR proton splitting due to the E/Z character imposed by the nitroso group, consistent with earlier literature (33) (supplemental Figs. S3 and S4). All deuterated nitrosamine derivatives (see below) were obtained as mixtures of E-and Z-isomers and were Ͼ97% isotopically enriched at the site(s) of modification, as judged by MS and NMR spectroscopy (33). Detailed syntheses and analytical data for all nitrosamines are included in the supplemental material, including N-nitrosodi- [1,1Ј-  Enzymes-P450 2A6 was expressed from a pCW plasmid in Escherichia coli DH5␣ (31,34,35). The yield of whole cell P450 hemoprotein expression was ϳ200 nmol/liter, as determined by Fe 2ϩ -CO versus Fe 2ϩ difference spectra (36). P450 2A6 was purified to electrophoretic homogeneity from solubilized 2A6 membrane fractions as described previously (31,34,35) using a combination of ion exchange (DEAE-Sephacel) and Ni 2ϩ -nitrilotriacetic acid chromatography. Recombinant rat NADPH-P450 reductase was expressed in E. coli and purified as described elsewhere (37). Recombinant human cytochrome b 5 was expressed in E. coli JM109 cells from a plasmid (pSE420(Amp)) provided by Satoru Asaki (Takeda Pharmaceuticals, Osaka, Japan). The protein was purified to electrophoretic homogeneity using modification of the DEAEcellulose and other chromatography methods described elsewhere (31,38).
Spectroscopy-NMR spectra were recorded using a Bruker 300-MHz spectrometer in the Vanderbilt facility. UV-visible spectra were acquired using an OLIS/Aminco DW2a instrument (OLIS, Bogart, GA). Mass spectra of synthetic products were recorded in the Vanderbilt facility using a Thermo-Finnigan TSQ-7000 instrument (ThermoFinnigan, Sunnydale, CA). LC-MS work was done with a Thermo LTQ instrument or a Waters Synapt mass spectrometer using an Acquity UPLC (Waters, Milford, MA), except for the competitive experiments (inter-and intramolecular), which were done by LC-MS using the TSQ-7000 instrument.
General Assays-Aldehydes were quantified by HPLC/UV analysis of the 2,4-dinitrophenylhydrazones as described previously (24,39), with some modification. The sensitivity of the assays was improved by purification of some of the reagents and by removal of glycerol from enzymes (by dialysis immediately prior to use), because commercial glycerol was found to be contaminated with formaldehyde. Dansyl hydrazones were used in some LC-MS analyses because of the low (electrospray ionization and atmospheric pressure chemical ionization) ionization efficiencies of HCHO-and CH 3 CHO-derived DNPH hydrazone derivatives. Applying the analytical methods described here to the analysis of unlabeled substrates, we were able to quantitate acetic acid in extracts of enzyme product mixtures, but not formic acid, at the level of sensitivity required for this work. Therefore, we used 14 C-labeled substrates and measured [ 14 C]formic acid utilizing ion exchange SPE columns.
For the DMN time course experiments, 0.5 nmol of 2A6, 1.0 nmol of NADPH-P450 reductase, 0.5 nmol of cytochrome b 5 , 30 M L-␣-dilauroyl-sn-glycero-3-phosphocholine, NADPH (1 mM), and DMN (17 mM) were incubated in 0.2 ml of potassium phosphate buffer (100 mM, pH 7.4) at 37°C for varying amounts of time. Reactions were initiated by adding NADPH. Incubations were terminated by adding 200 l of cold CH 3 CN and centrifuged (2 ϫ 10 3 ϫ g) for 5 min. The supernatants were transferred to amber vials, and 20 nmol of DCDO and 600 l of a freshly prepared solution of dansylhydrazine (0.5 mg ml Ϫ1 ) in CH 3 CN containing 0.3% CH 3 CO 2 H (v/v) was added. The reaction mixture was incubated at room temperature for 30 min and dried under nitrogen. Finally, the residue was reconstituted in 100 l of a mixture of H 2 O/CH 3 CN (3:1, v/v). Dansylated products were analyzed by LC-MS on a Waters Acquity UPLC system connected to either an LTQ (Thermo Fisher, Santa Clara, CA) or a Waters Synapt mass spectrometer using an Aquity UPLC BEH C18 octadecylsilane column (1.7 m, 2.1 mm ϫ 100 mm). LC conditions were as follows. Buffer A contained 10 mM NH 4 CH 3 CO 2 and 2% CH 3 CN (v/v), and buffer B contained 10 mM NH 4 CH 3 CO 2 and 95% CH 3 CN (v/v). The following gradient program was used, with a flow rate of 300 l/min: 0 -6.5 min, linear gradient from 25% B to 100% B (v/v); 6.5-7.5 min, hold at 100% B; 7.5-8 min, linear gradient to 75% A (v/v); 8 -10 min, hold at 75% A (v/v). The temperature of the column was maintained at 25°C. Samples (20 l) were infused with an autosampler. MS analyses were performed in the electrospray positive ion mode. The results of steady-state kinetic experiments were fit to hyperbolic plots using GraphPad Prism (GraphPad, Dan Diego, CA), and parameters and S.E. values were obtained with this program using non-linear regression.
Time course reactions for DEN oxidation were performed as described for nitrosamine N-dealkylation assays (see above) with the exception that reactions were terminated by the addition of 200 l of cold CH 2 Cl 2 , To the reaction mixture 0.5 nmol of CD 3 CO 2 H was added, and the products were derivatized with 4-nitrophenacyl bromide, as described above. The derivatized products were analyzed by LC-MS on a Waters Acquity UPLC system connected to an LTQ mass spectrometer (Thermo Fisher, Santa Clara, CA) using an Aquity UPLC BEH C18 octadecylsilane column (1.7 m, 1 mm ϫ 100 mm). LC conditions were as follows: buffer A contained 2% CH 3 CN and 0.01% HCO 2 H (v/v), and buffer B contained 95% CH 3 CN and 0.01% HCO 2 H (v/v). The following gradient program was used, with a flow rate of 150 l min Ϫ1 : 0 -4.0 min, linear gradient from 95% A to 50% A; 4.0 -4.5 min, linear gradient to 100% B; 4.5-5.5 min, hold at 100% B; 5.5-6 min, linear gradient to 95% A; 6 -8 min, hold at 95% A. The temperature of the column was maintained at 50°C; the injection volume was 20 l. Samples were infused with an autosampler. MS analyses were performed in the negative ion mode. Electrospray ionization conditions were as follows: source voltage, 4 kV, source current, 100 A; auxiliary gas flow rate setting, 37; sheath gas flow setting, 16; capillary voltage, Ϫ4 V; capillary temperature, 380°C; tube lens voltage, Ϫ22 V.
[ 14  0.18 mCi mmol Ϫ1 ) were initiated with an acetate-free NADPHgenerating system (dialyzed enzymes), incubated for a specified time at 37°C, terminated by the addition of 100 l of 10% ZnSO 4 ⅐7H 2 O (w/v), and centrifuged (2 ϫ 10 3 ϫ g). The supernatants were loaded onto Bakerbond TM quaternary amine 3-ml SPE columns that had been washed with 6 ml of CH 3 OH and equilibrated with 10 ml of H 2 O. After loading, the columns were washed with 10 ml of H 2 O to remove residual aldehyde or nitrosamine substrate. The [ 14 C]formic acid was eluted with 1.5 ml of 0.1 M HCl, and radioactivity was measured by liquid scintillation spectrometry. Recovery was calibrated using a sodium [ 14 C]formate standard.
Pre-steady-state Kinetics-Pre-steady-state kinetics of oxidation were performed in a quench-flow apparatus (model RFQ- Reactions were performed at 37°C, and P450 2A6 (100 pmol) was reconstituted as described for the nitrosamine N-dealkylation assays (see above). Reactions were initiated with 1 mM NADPH for a period of time ranging from 5 ms to 10 s and quenched with CH 3 CN containing 1% CH 3 CO 2 H. Propionaldehyde was used as an internal standard. Aldehydes were derivatized with dansylhydrazine and quantitated by LC-MS as described for the DMN N-demethylation time course experiments (see above). In the case of [ 13 C]HCHO, the yields were corrected for the isotopic abundance of [ 13 C]HCHO, resulting from background [ 12 C]HCHO.
For measuring oxidation of acetaldehyde to acetic acid, [ 14 C]CH 3 CHO (10 mM final concentration, 5 mCi/mmol) was used. Reactions were performed as described for the [ 14 C]formic acid assays with the exception that 17% ZnSO 4 ⅐7H 2 O (w/v) was used to quench the reaction. The product [ 14 C]CH 3 CO 2 H was purified using Bakerbond TM quaternary amine SPE columns and detected by liquid scintillation spectrometry (see above).
Stopped-flow Kinetics-Analyses were done using an OLIS RSM-1000 stopped-flow spectrophotometer in the rapid scanning monochromator mode, using the general procedures described earlier (31).
Ligand binding assays were done at ambient temperature (23°C). P450 2A6 (final concentration in mixing cell, 2.0 M) was mixed with either 12 mM DMN or 1.1 mM DEN (final concentrations) in 50 mM potassium phosphate buffer (pH 7.4). Kinetic analysis was done using the absorbance changes at 420 nm (decrease).
Reduction assays were done with ferric P450 2A6 and a 2-fold molar excess of NADPH-P450 reductase under an anaerobic CO atmosphere using procedures described earlier (31,41). The rate of the increase in absorbance at 450 nm was used in measurements of rates. The ligands used (DMN, DEN, HCHO, and CH 3 CHO) are all volatile and therefore were added to the tonometers containing ferric P450 2A6 after degassing and equilibration (with CO) was finished by opening a tonometer port (under positive CO pressure) and adding the (undiluted) ligand from a 500-l syringe (Gastight number 1750, Hamilton, Reno, NV), fitted with a ground joint for sealing to a port on the tonometer, which had been fitted with a screw-type driver and calibrated to deliver 5.7 l per turn. A fitted glass cap was then placed back on the tonometer (under positive CO pressure), and the device was used to load one of the drive syringes of the stopped-flow instrument.
In the case of ligand binding reactions, the reaction time was either 15 s or 150 ms. For reduction, reaction times were 3 or 30 s. In both cases, 4 -7 individual mixing reactions were used to derive rate constants (using the OLIS software, single exponential fits) and averaged.
Pulse-Chase Experiments-Experiments were done with [ 14 C]DMN (17 mM, 0.5 mCi mmol Ϫ1 ) in the case of DMN and with N-nitrosodi-[2,2Ј-d 3 ]ethylamine (0.14 mM) in the case of DEN. In both cases, the reaction was initiated (37°C) with 2.5 and an NADPH-generating system (see above). After 1 or 2 min either (unlabeled) HCHO (2 mM in the case of DMN) or CH 3 CHO (1.5 mM in the case of the substrate DEN) was added, and the reaction was allowed to proceed another 18 or 19 min (at 37°C) (to 20 min total reaction time). In the case of DMN, the product [ 14 C]HCO 2 H was measured as described for the [ 14 C]formic acid assays (see above) with the exception that, after sample loading, the column was washed with 20% CH 3 OH (5 ml) and H 2 O (10 ml) to completely elute the residual [ 14 C]DMN substrate. The product [ 14 C]HCO 2 H was then eluted with 1 M HCl. In the case of DEN, the product CD 3 CO 2 H was derivatized with 4-nitrophenacyl bromide and analyzed by LC-MS, using the procedure described earlier. The non-deuterated CH 3 CO 2 H in the reagents served as an internal standard (ϳ10-fold higher level). Assays were run in triplicate, and the mean results were compared, with the extent of decrease due to the presence of the added aldehyde being indicative of the fraction of the unlabeled aldehyde (or its equivalent) that exchanged.

RESULTS
Preliminary Assays-P450 2A6-catalyzed rates of conversion of DMN and DEN to formaldehyde and acetaldehyde, respectively, were constant up to at least 15 min, and this reaction time was used in most subsequent assays. Both reactions were highly dependent upon the presence of cytochrome b 5 ; with DMN (used at 17 mM), the rate of HCHO production was 0.0030 Ϯ 0.0015 s Ϫ1 in the absence of cytochrome b 5 and 0.11 Ϯ 0.05 s Ϫ1 in its presence. For DEN (used at 140 M), the rates of acetaldehyde formation in the absence and presence of cytochrome b 5 were 0.04 Ϯ 0.002 and 0.15 Ϯ 0.68 s Ϫ1 , respectively. Accordingly, all further assays included cytochrome b 5 . Initial parameters of k cat ϭ 0.42 s Ϫ1 and K m ϭ 15 mM were estimated for DMN oxidation to formaldehyde, and k cat ϭ 0.13 s Ϫ1 and K m ϭ 0.15 mM for DEN oxidation to acetaldehyde. These results are consistent with literature indicating the preferential catalytic efficiency of P450 2A6 toward DEN (28,32). With this preliminary information, concentrations of 17 mM DMN and 1 mM DEN were used for the competitive and intrinsic kinetic deuterium isotope effect studies (see below).
Intramolecular Kinetic Deuterium Isotope Effects-The D k values (Table 1) form a basis for comparison with other isotope effects, which should show attenuation if specific steps contribute to rate limitation (3). In principle, comparisons of rates of C-H and C-D bond cleavage at a carbon atom substituted with both protium and deuterium should provide an estimate of D k, the intrinsic isotope effect. These values, estimated by MS of the aldehyde products, were 10.2 Ϯ 0.2 and 3.7 Ϯ 0.2 for DMN and DEN, respectively.
Both of the values have caveats. The D k value of 10 for the -CHD 2 group(s) of DMN has the potential contribution of a geminal secondary isotope effect on the C-H value (42). In the general literature, these are usually 1.0 -1.2 (42)(43)(44), including the few cases in which they have been estimated for P450s (45,46). Such secondary isotope effects are multiplicative, so with two deuteriums, the contribution could be as high as 1.4. Thus,  the true D k for DMN oxidation could be as low as 10/1.4 ϭ 7. The contribution of a geminal secondary kinetic isotope effect should be less in the case of the methylene in DEN (-CHD-) (i.e. Յ1.2), so the D k should be Յ3.5 and possibly ϳ3. 4 We also measured the apparent kinetic hydrogen isotope effects for unsymmetrically deuterated DMN and DEN (Table  1). In principle, the comparison of these values with the estimates of the intrinsic kinetic isotope effects (see above) can provide an estimate of the ability of a substrate to turn within the active site of the enzyme (48 -51). These values were as high as the values obtained with N-nitrosodi-[1-d 2 ,1Ј-d 2 ]methylamine and N-nitrosodi-[1,1Ј-d 2 ]ethylamine (Table 1), within experimental error, and do not provide evidence for restricted motion in the active site, which would have the expected effect of attenuating these values. However, even if effects are not seen in such experiments, they must be considered in light of the alternate interpretation that the substrates rapidly exchange with the medium, as opposed to rotating within the active site. The rates of binding of DMN and DEN to P450 2A6 were measured. Binding of either substrate to P450 2A6 yielded a "Type I" difference spectrum, with an apparent shift of low spin iron ( max 418 nm) to high spin ( max 390 nm). The apparent K d was 12 mM for DMN and 1.1 mM for DEN (supplemental Figs. S5 and S6). This change could be observed in a stopped-flow spectrophotometer and occurred at (first-order) rates of 80 Ϯ 18 s Ϫ1 with 12 mM DMN and 73 Ϯ 22 s Ϫ1 with 1.1 mM DEN (at 23°C). 5 These parameters are for the binding of ligands but do not directly reflect the "off" rates. The high K d values prevent estimation of the dissociation rate by extrapolation to zero substrate concentration (31,52). A dissociation rate was estimated indirectly by mixing a preformed P450 2A6-DEN complex with 4-phenylimidizole, which is presumed to occupy the same site but yields a different spectral complex. 5 The rapid release of DMN and DEN was indicated by these experiments and the competitive intermolecular isotope effect results (see below).

REACTIONS 1 and 2
Competitive Intermolecular Kinetic Deuterium Isotope Effects-The competitive isotope effects for the mixtures of d 0 and perdeuterated nitrosamines (Table 1) were within experimental error of the intramolecular values, the pseudointrinsic isotope effects. In addition, the values seen in the "mixed" experiments were just as high, with d 0 substrate plus asymmetrically labeled nitrosamines (Table 1). If exchanges were slow, these values should also be lower. 5 Non-competitive Intermolecular Kinetic Deuterium Isotope Effects-Comparison of the non-competitive intermolecular deuterium isotope effects with estimates of the intrinsic isotope effect ( D k) is a generally accepted approach to assess the extent to which the C-H bond-breaking step contributes to limiting an overall catalytic reaction of an enzyme (2,3,54), particularly if physical steps such as substrate binding and release are not slow, as shown earlier for DMN and DEN.
Deuterium substitution of DMN affected both k cat and K m ( Table 2) Table 1 from a non-competitive to a competitive experiment. If there is a significant attenuation of the kinetic isotope effect in going to a competitive experiment, then the interpretation is that the rate of exchange of the substrate (DEN) is a factor. However, the values for the competitive isotope effects measured subsequently in Table 1 are nearly as large and within experimental error. 5 One approach to the rate of exchange of ligands with ferric P450 2A6 involved a more direct kinetic analysis. For substrates with high affinity, the equation k obs ϭ k on [S] ϩ k off can be used for a simple two-state system (52) and has been applied to the binding of coumarin to P450 2A6 (31). However, the K m for DEN was ϳ0.15 mM (Table 3) and the spectrally estimated K d (increase in absorbance at 390 nm, decrease at 420 nm) for binding DEN was ϳ1.1 mM (supplemental Fig. S6), so the k obs values would be expected to be too fast to measure by stopped-flow methods. We employed an alternative approach, with the assumption that DEN and the "Type II" (53) ligand 4-phenylimidazole occupy the same space, a concept supported by the available crystal structures of P450 2A6 bound to coumarin and 8-methoxypsoralen (30). A preformed P450 2A6⅐DEN complex was rapidly mixed (at 13°C) with an excess of 4-phenylimidazole, and the rate of the change in the spectrum (from a max 390 nm complex to a max 427 nm complex) was measured (0.86 Ϯ 0.08 s Ϫ1 ) and compared with the rate of the formation of the latter complex in the absence of DEN (0.86 Ϯ 0.12 s Ϫ1 ). Analysis of the data with the kinetic model shown in Reactions 1 and 2, using K d,DEN ϭ k Ϫ1 /k 1 ϭ 1100 M (from a spectral titration) and K d,4-phenylimidazole ϭ 0.51 Ϯ 0.10 M (from spectral titration) and an experiment involving 5 mM DEN (5-fold Ͼ K s ) and 10 M 4-phenylimidazole (20-fold Ͼ K s ) with the program DynaFit (assuming a simple competitive model) indicated that k 1 and k 2 (at 13°C) were ϳ5 ϫ 10 5 M Ϫ1 s Ϫ1 , and thus k Ϫ1 must be Ͼ500 s Ϫ1 , which is consistent with the K d (K d,DEN ϭ k Ϫ1 /k 1 ϭ 5 ϫ 10 2 M Ϫ1 s Ϫ1 /5 ϫ 10 5 M Ϫ1 s Ϫ1 ϳ 1 mM) (see supplemental Fig. S7). From these results, we conclude that the exchange of DEN and, by inference, DMN with ferric P450 2A6 is very rapid (ϳ100 s Ϫ1 ) and is not an issue in the interpretation of the kinetic deuterium isotope effects. b These values contain error because of the inability to saturate the enzyme, and the D (V/K) value is a better (upper) estimate (i.e. ratios of tangents in Fig. 2).
have less error based on visual observation of the tangents (Fig.  2), was ϳ13, within experimental error of the approximated intrinsic D k (Table 1). When DEN oxidation was measured, D V and D (V/K) were near unity (Table 3), within experimental error.
Limited Methyl (␤) Hydroxylation of DEN-One possibility to consider is that P450 2A6 can generate alternate products when a C-D bond replaces C-H, which could affect the interpretation of the kinetic isotope effects. P450s can catalyze a reaction generating NO 2 Ϫ (55), although this is a relatively minor pathway. Another possibility is that deuterium substitution of the ␣-carbon of DEN could favor hydroxylation of the methyl group. We synthesized 2-hydroxy-DEN, characterized it, and used it as an LC-MS standard to monitor its possible formation from DEN by P450 2A6. With d 0 -DEN, the rate was Ͻ0.004 min Ϫ1 (with a DEN concentration of 1 mM). With d 2 d 2 -DEN, the rate was ϳ0.011 min Ϫ1 . We conclude that methyl hydroxylation is not a major pathway, even when the ␣-carbon is deuterium-substituted.

Comparisons of Product Formation with d 0 -, d 3 -, and d 3 d 3 -DMN-
The lack of attenuation of kinetic isotope effects in the intermolecular competitive experiments (Table 1) and the results of the 4-phenylimidazole binding experiments 5 indi-cate that P450 2A6 exchanges substrates rapidly, consonant with the results of studies with P450 2A6 and coumarin (31). One question is whether such exchange can occur at the stage of the actual oxygen complex (putative FeO 3ϩ ).
The presence of a high isotope effect in the non-competitive intermolecular experiments (Table 1) permits the application of another type of experiment, which we have applied previously with P450 1A2 (51). To address the question of whether such exchange can occur at the stage of the actual oxygen complex (putative FeO 3ϩ ) formation, we applied another type of experiment using DMN and two DMN analogues (DMN-d 3 and DMN-d 3 d 3 ). The presence of deuterium in the substrate is not sensed in an enzyme catalytic cycle until the chemistry of C-H(D) bond breaking begins. Because the C-D bond-breaking step is relatively more difficult, the enzyme could (if the step were slow enough relative to exchange) dissociate the deuterated substrate, bind the protiated substrate, and oxidize the protiated substrate. The intermolecular non-competitive isotope effect and the overall rate of product formation would not be attenuated. However, if no exchange of the substrate can occur after complete activation of the FeO 3ϩ complex, the rate of product formation should reflect the substrate isotopic composition (i.e. the amount of product formed from the oxidation of d 0 d 3 -DMN (HCHO plus DCDO) should be intermediate between the amounts formed from d 0 -DMN (HCHO) and d 3 d 3 -DMN (DCDO)). When the experiment was done, the latter result was obtained (Fig. 3). A similar set of experiments using deuterated DEN analogues could not be done because the non-competitive intermolecular isotope effect was near unity ( Table 2).
Oxidations of Aldehydes to Carboxylic Acids-Both HCHO and CH 3 CHO were substrates for P450 2A6 oxidation ( Table  3). The oxidation of CH 3 CHO was much more efficient than that of HCHO, by a factor of 20-fold. The efficiencies (k cat /K m ) for oxidations of both aldehydes were ϳ10-fold less than for the oxidations of the corresponding nitrosamines to the aldehydes ( Table 2).
A non-competitive intermolecular kinetic isotope effect of 2.0 was observed for the P450 2A6-catalyzed acetaldehyde oxidation D (V/K) parameter, but not D V; the isotope effect was only on K m . A similar observation for acetaldehyde oxidation was made with human P450 2E1 (27). The P450 2A6-catalyzed oxidation of formaldehyde was not analyzed for an isotope effect because the assay of formic acid was not sufficiently sensitive without the use of 14 C-labeled material.  DMN and DEN to aldehydes. A, DMN. B, DEN. The points are means Ϯ range for duplicate assays at each concentration indicated. In some cases, the range was within the size of the point and is not shown. See Table  3 for estimated parameters.

JOURNAL OF BIOLOGICAL CHEMISTRY 8037
Time Courses of Aldehyde and Carboxylic Acid Formation from DMN and DEN-Careful assays showed that DMN was oxidized to both HCHO and HCO 2 H in a linear course, over a period of 10 min (Fig. 4A). Similarly, oxidations of DEN to both CH 3 CHO and CH 3 CO 2 H were apparently linear (at least 15 min) (Fig. 4, B and C). Most notably, no lag was observed in the formation of the carboxylic acid in either case.
Pre-steady-state kinetic analysis of the conversion of DMN and DEN to the respective aldehydes showed small but reproducible bursts in both cases, with 2-4% product formed in the rapid phase (Fig. 5). A small burst was also detected in the oxidation of CH 3 CHO (supplemental Fig. S8).
Spectrally Determined Binding of Ligands to P450 2A6-The addition of DMN or DEN to (ferric) P450 2A6 produced a classic Type I heme Soret difference spectrum (see above), with the lower UV region obscured by the absorbance of the ligand (supplemental Figs. S5 and S6). The estimated K d values for DMN and DEN were 12 and 1.1 mM, respectively. The Type I spectral changes for the aldehydes were even weaker (supplemental Figs. S9 and S10), with apparent K d values of ϳ200 mM.
Rates of P450 2A6 Reduction-Rates of reduction of P450 2A6 have been shown to be slow in the absence of substrate and enhanced by the presence of the substrate coumarin (31). Preliminary anaerobic studies showed that all of the Na 2 S 2 O 4 -reducible P450 2A6, when DEN was present, was reduced within 60 s. Stopped-flow kinetics assays established that the rate of reduction of P450 2A6 was enhanced by the presence of DMN or DEN but not formaldehyde or acetaldehyde (supplemental Fig. S11).
NADPH Oxidation Rates and Coupling Efficiency-In the absence of any added substrate, the reconstituted P450 2A6 system oxidized NADPH at a rate of 0.63 s Ϫ1 (i.e. 0.63 nmol of NADPH/s/nmol of P450. With 17 mM DMN (K m concentration), the rate was 1.5 s Ϫ1 , and with 0.14 mM DEN (K m concentration), the rate was 0.75 s Ϫ1 . Under these conditions, the rates of formation of the aldehydes with the same reagents were 0.05 and 0.075 s Ϫ1 , respectively (Fig. 4). Thus, with DMN and DEN, the efficiency of coupling to generate nitrosamine reaction products was 3 and 10%, respectively.
Pulse-Chase Experiments-The lack of lag phases in the production of carboxylic acids (Fig. 4) suggested that the aldehyde products of the nitrosamines might be retained by P450 2A6 and not in exchange with the medium. Accordingly, pulse-chase experiments were done, in which reactions were initiated with labeled DMN or DEN (concentration ϳ K m ). A large excess of the appropriate unlabeled aldehyde was added after 1-2 min, the reaction was quenched after 20 min, and the isotopic incorporation of the carboxylic acid products was measured by liquid scintillation spectrometry or LC-MS. The final ratios of HCHO/ [ 14 C]HCHO and of CH 3 CHO/CD 3 CHO were ϳ16 and ϳ7, respectively, based on the total amount of aldehydes formed in 20 min. If all of the intermediate aldehyde was in equilibrium (in each case), this approach should have eliminated all but 5-10% of the label in the recovered carboxylic acid. However, 80 -90% of the formic acid and 40 -60% of the acetic acid was labeled (Fig. 6).  Kinetic Modeling-An overall scheme of the major events of the reactions is shown in Fig. 7, including known rates of some of the non-enzymatic reactions (56 -59). If the aldehyde product dissociates from P450 2A6, then the rates of formation of the carboxylic acids (measured in Fig. 4) can be described by a model of a coupled reaction with separate k cat and K m values for the two steps (Tables 2 and 3). The predicted time courses of formation of the aldehydes and carboxylic acids (see supplemental Fig. S12) were compared with the experimental values (from Fig. 4) in Fig. 8. Two striking features were clear: (i) the theoretical plots underpredict carboxylic acid formation, and (ii) the theoretical model predicts lag phases for carboxylic acid formation, which is intuitive due to the relatively high K m values for oxidation of the two aldehydes (Table 3).
A comprehensive but minimal kinetic model was developed utilizing KinTek Explorer software. The model is necessarily complex because it must include multiple phenomena, including (i) normal events known to occur in P450 reactions, including separate substrate binding and oxygen activation steps, (ii) the irreversible loss of reduced oxygen species from the activated complex (see above) (60), (iii) the partial burst kinetics (Fig. 5), (iv) the lack of affinity of the aldehydes for the enzyme (supplemental Figs. S9 and S10), (v) the use of the product as a substrate in the second reaction (Table 3), (vi) the non-competitive kinetic deuterium isotope effect seen with DMN but not DEN (Table 2), and (vii) the maximum rates possible for converting part of the substrate to product (Fig. 4). A minimal model was developed that is consistent with these data (Fig. 9), with S denoting the nitrosamine, P the aldehyde, and Q the carboxylic acid. Critical features included fitting the time courses of formation of the aldehyde and acid simultaneously, with a lack of a lag for the acid, as well as showing a partial kinetic burst and an isotope effect for DMN. The rate constants for ligand binding were set at 10 7 M Ϫ1 s Ϫ1 (consistent with other work with P450 2A6 (31) and accepted diffusion-limited rates of interaction of enzymes and ligands (44), with rate constants for ligand release balanced to fit the spectrally estimated K d values (supplemental Figs. S5, S6, S9, and S10). The rate constants for steps 1-4 were set not to be faster than the rapid formation of product by P450 2A6 (Fig. 5).
Fits are shown for DMN and DEN oxidations to aldehydes and acids in Fig. 10, using the rate constants shown in Table 4 for the steps in Fig. 9 (see also supplemental Fig. S13). The time courses of product formation fit the experimental data well, and partial bursts are predicted (as shown with the inset for DMN in Fig. 10B). Applying an intrinsic D k of 12 for DMN N-dealkylation (Table 1) reduced the rate of HCHO production (v) by 3.5-fold, and applying an intrinsic D k of 3 for DEN N-dealkylation (Table 1) reduced the rate of CH 3 CHO formation by only 1.2-fold (cf. Table 2).

DISCUSSION
The original impetus for this work was an in vivo study showing a strong deuterium isotope effect on the hepatocarcinogenicity of DMN (15). At that time, none of the mammalian P450s had been characterized. Subsequently, P450s 2A6 and 2E1 were identified as the major catalysts involved in the oxidation and bioactivation of short-chain N-nitrosamines, especially DMN and DEN (28,32,61). Nitrosamines should not be regarded only as xenobiotics, in that the in vivo nitrosation of secondary amines is a well established phenomenon (62)(63)(64) and considered to be of relevance in understanding human cancer etiology (65). Previous work on the  Table 4), and, after 1 or 2 min, the appropriate unlabeled aldehyde was added in large excess over the calculated yield of labeled aldehyde (Fig. 4).  stepwise oxidation of ethanol to acetic acid by human P450 2E1 identified an intrinsic kinetic deuterium isotope effect (expressed primarily in K m ), burst kinetics, and a processive pathway (24,27). In the oxidation of DMN and DEN by P450 2A6, some similar phenomena were seen, and some of the same mechanisms seem applicable, although several features are more complex. The best estimates for the intrinsic kinetic deuterium isotope effects for the oxidations of DMN and DEN by P450 2A6 are ϳ10 and ϳ3, respectively, as estimated from the noncompetitive intramolecular studies (Table 1, lines 1 and 2), with caveats about the contribution of secondary isotope effects. Similar kinetic isotope effects were measured in a variety of competitive experiments with labeled DMN and DEN (Table 1), arguing that exchange of the nitrosamine substrates is very rapid, a conclusion supported by direct measurements of on-rates (70 -80 s Ϫ1 with 12 mM DMN or 1.1 mM DEN at 23°C) and by an experiment in which the dissociation rate of a P450 2A6-DEN complex was estimated by displacement of another ligand (4-phenylimidazole) 5 (supplemental Fig. S7).
Major conclusions from the kinetic isotope effect work are that the isotope effect is much higher for oxidation of the methyl group (DMN) than the methylene group (DEN). Another major conclusion is that hydrogen atom abstraction is a rate-limiting step in the case of DMN but not DEN (Table  3 and Fig. 2). A simple explanation for the difference may be the inductive effect and the inherent energy of breaking a methylene C-H bond compared with a methyl (e.g. 94.5 versus 98 kcal/mol) (66). 6 Several lines of investigation argue that these substrates can tumble and/or exchange rapidly (Table 1) but not after the P450 is in the activated state and poised for C-H bond cleavage (Figs. 3 and 11).
One interesting aspect of this study is the observed processivity of oxidation. P450 2A6 oxidizes the nitrosamines to aldehydes ( Table 2) and oxidizes both aldehydes (HCHO and CH 3 CHO) to carboxylic acids (Table 3). However, the aldehyde oxidations are not particularly efficient (Table 3) (Fig. 6) indicate that a dissociative model cannot explain the results on formation of carboxylic acids (Fig. 8). However, the aldehydes were found to have only low affinity for (ferric) P450 2A6 (supplemental Figs. S9 and S10).
A kinetic model was developed that is consistent with most of the observed results. In order to account for the lack of dissociation of the aldehyde(s), a solution used in our earlier studies on P450 2E1 (27) was invoked, namely that the P450 undergoes a conformational change upon binding substrate (see reviews of P450 crystal structures for physical evidence (68,69)) and then relaxes its conformation after forming product. We postulate that the P450 2A6-aldehyde complex is left in a conformational state leading to catalysis, avoiding the need to release, rebind, and then undergo the activating conformational change. Thus, the rate constants for events leading to catalysis in the second cycle compete with others in the "first" cycle (Fig. 9A). Another required feature is an equilibrium of an activated P450 2A6-substrate complex with an unproductive complex, added to explain the (pre-steady-state) partial bursts observed for production of aldehydes (70 -72). The abortive generation of reduced oxygen species (Fig. 9) was included in light of the low efficiency of NADPH coupling (see above). Any model must also have a C-H bond-breaking step that can account for the high kinetic deuterium isotope effects (for DMN) ( Table 2). 6 The values in Ref. 66   Finally, the model incorporates the spectrally measured K d values for the nitrosamines and aldehydes. 7 Literature rates of non-enzymatic steps are shown in Fig. 7, including the rates of rearrangement of ␣hydroxynitrosamines to aldehydes and the hydration of aldehydes/ dehydration of hydrated aldehydes. One initial consideration was that these phenomena might help explain the kinetic isotope effect results, but they do not help with the partial bursts and processivity, particularly if they are reasonable approximations of rate constants in the enzyme active site. One issue is the time needed for the breakdown of the ␣-hydroxynitrosamine to the aldehyde (Fig. 7). Another issue, if one uses a classic hydrogen atom abstraction mechanism for P450 2A6 (Fig. 11A), is that the aldehyde would need to hydrate, a slow process, before hydrogen atom abstraction (Fig. 7). One means of circumventing this kinetic problem is to use the alternative peroxide mechanism (27,76,77) (Fig. 11B), avoiding the need for the slow hydration of CH 3 CHO (Fig. 7) (57,58). (A third option could be 1e Ϫ oxidation of one of the hydroxyl groups (13,78).) At this time, we do not have definite evidence as to which of these pathways (Fig. 11) is operative (a small kinetic deuterium isotope effect was observed in a non-competitive experiment (Table 4), but this, by itself, does not have a single mechanistic interpretation).
A model (Fig. 9) was constructed based upon the seven points listed under "Kinetic Modeling." Several strengths of the model are as follows: (i) the kinetic courses of aldehyde and carboxylic acid formation match the experimental results, with no lags observed for carboxylic acid formation in either case (Fig. 10); (ii) the same rate constants can be utilized for DMN and DEN (Table 4), modified only to match the apparent nitrosamine affinity (supplemental Figs. S5 and S6); (iii) the model yields substoichiometric bursts of product formation (Fig. 10) to match the observed results (Fig. 5); and (iv) a kinetic deuterium isotope effect was expressed in the case of DMN but not DEN (cf. Table 3). Several points should be made about the model. Eliminating JS (P450Ј Fe 3ϩ -S), the unproductive complex in Fig. 9, yielded only a full kinetic burst (1 product/enzyme), not the partial burst. JP (P450Ј Fe 3ϩ -P) mirrors JS (i.e. if there is an unproductive complex with the nitrosamine, one might be expected with the aldehyde as substrate), and if JP is eliminated (Fig. 9), the reaction is 7 The crystal structures of P450 2A6 show only space for a single substrate (30,75), although none of the ligands is as small as DMN. Harrelson et al. (73,74) have proposed models of P450 2A6 (and P450 2E1) with two ligands present (simultaneously) in the active site. These proposals are based upon some patterns of kinetic deuterium isotope effects seen with xylenes and also on abnormalities in Eadie-Hofstee plots of steady-state kinetic results. However, the data points of Fig. 5 (also supplemental   too fast, and rate constants must be attenuated. If the F form (P450* Fe 3ϩ ) of the enzyme (Fig. 9) is eliminated (F regenerated from a productive conformational change following substrate binding), then carboxylic acid formation is too slow, and the burst phase is not substoichiometric. The step GS 3 E ϩ O (P450 FeO 3ϩ 3 H 2 O 2 ϩ H 2 O) is logical, in light of the observed uncoupling (see above); if these steps (GS, GP 3 E ϩ O) are dropped, the model still fits, but some rate constants must be attenuated. Nevertheless, the current model does have some deficiencies, including the following: (i) the predicted burst is substoichiometric but does not completely fit the experimental results, and (ii) the predicted expressed non-competitive kinetic isotope effect ( Table 2) is larger ( D (V/K) 13 Ϯ 5) than predicted by the model (3.2; see above). Further improvement may be in order, although the mechanism is already fairly complex, and we hesitate to include additional steps without justification. Several P450s catalyze sequential oxidations in steroid metabolism (5,7). P450 19A1 oxidizes androgens to estrogens with the accumulation of low concentrations of intermediates (79). P450 11B2 converts deoxycorticosterone to aldosterone without the dissociation of intermediates from the enzyme, as judged by pulse and quench experiments (80). Similar approaches led to the conclusion that in the oxidation of pregnenolone to dehydroepiandrosterone by P450 17A1, about 20% of the intermediate 17␣-hydroxypregnenolone did not dissociate from the enzyme (81), although a study with P450 17A1 transfected into HEK-293 cells yielded a different conclusion (82). The literature with P450 19A1, the steroid aromatase, is controversial regarding processivity (83). Even less information is available about P450s that do not normally have defined physiological roles in steroid metabolism. Rat P450 2C11 oxidizes testosterone to 16␣-hydroxyandrostenedione via androstenedione, and Sugiyama et al. (84) estimated that ϳ15% of the androstenedione intermediate does not dissociate in the process. The only other previous work in this area comes from our own laboratory with P450 2E1 and the conversion of ethanol to acetic acid via acetaldehyde (27); pulse experiments suggested that ϳ90% of the acetaldehyde did not dissociate in the process (see above). The processivity of P450 2E1 in nitrosamine oxidations has not been reported in the literature to date.  Table 4, with the numbers of the reaction steps ( Fig. 9) corresponding to each rate constant.