The Ferrous-Dioxygen Intermediate in Human Cytochrome P450 3A4

The oxy-ferrous complex is the first of three branching intermediates in the catalytic cycle of cytochrome P450, in which the total efficiency of substrate turnover is curtailed by the side reaction of autoxidation. For human membrane-bound cytochromes P450, the oxy complex is believed to be the primary source of cytotoxic superoxide and peroxide, although information on the properties and stability of this intermediate is lacking. Here we document stopped-flow spectroscopic studies of the formation and decay of the oxy-ferrous complex in the most abundant human cytochrome P450 (CYP3A4) as a function of temperature in the substrate-free and substrate-bound form. CYP3A4 solubilized in purified monomeric form in nanoscale POPC bilayers is functionally and kinetically homogeneous. In substrate-free CYP3A4, the oxy complex is extremely unstable with a half-life of ∼30 ms at 5 °C. Saturation with testosterone or bromocriptine stabilizes the oxy-ferrous intermediate. Comparison of the autoxidation rates with the available data on CYP3A4 turnover kinetics suggests that the oxy complex may be an important route for uncoupling.

The oxy-ferrous complex is the first of three branching intermediates in the catalytic cycle of cytochrome P450, in which the total efficiency of substrate turnover is curtailed by the side reaction of autoxidation. For human membrane-bound cytochromes P450, the oxy complex is believed to be the primary source of cytotoxic superoxide and peroxide, although information on the properties and stability of this intermediate is lacking. Here we document stopped-flow spectroscopic studies of the formation and decay of the oxy-ferrous complex in the most abundant human cytochrome P450 (CYP3A4) as a function of temperature in the substrate-free and substrate-bound form. CYP3A4 solubilized in purified monomeric form in nanoscale POPC bilayers is functionally and kinetically homogeneous. In substrate-free CYP3A4, the oxy complex is extremely unstable with a half-life of ϳ30 ms at 5°C. Saturation with testosterone or bromocriptine stabilizes the oxy-ferrous intermediate. Comparison of the autoxidation rates with the available data on CYP3A4 turnover kinetics suggests that the oxy complex may be an important route for uncoupling.
Cytochrome P450 CYP3A4 is the most prevalent P450 monooxygenase in human liver and is responsible for the metabolism of almost half of xenobiotics encountered by man (1). This isozyme has a large and flexible active site and is able to bind and catalytically convert multiple substrates, often displaying homotropic and heterotropic cooperativity in substrate binding and product formation. As a central player in human drug metabolism, CYP3A4 is one of the most intensely studied P450s, either in isolated human liver microsomes (2-5), detergent-solubilized form (6), or purified soluble aggregates (7,8).
Recently, two groups have reported the crystal structure of a truncated CYP3A4 (9,10). Despite this structural information, precise chemical and biophysical characterization of the human P450s has lagged that of the monomeric, soluble P450 isozymes isolated from bacteria. In particular, the critical intermediates in the reaction cycle of human P450s have not been precisely documented, due in part to the inability to form a robust, monomeric, and soluble entity that is amenable to the rapid reaction and spectroscopic methodologies successfully applied to the bacterial P450s. This lacuna is significant, as CYP3A4 and other human P450s display complex aspects of substrate recognition and catalytic mechanism that are not present in the simpler P450s such as CYP101 from Pseudomonas.
The current version of the catalytic reaction cycle of cytochrome P450 monooxygenases involves intermediate reduction states of heme iron and atmospheric dioxygen and represents the result of over 30 years of intensive research (11,12). Traditionally, the cyclic process begins with the ferric low spin species and a water molecule occupying the sixth coordination site of the heme. In many cases the complementary fit of the native substrate into the pocket displaces this water ligand allowing the system to assume a predominantly high spin electronic configuration, although in some instances substrates may induce only partial spin shift (13,14). Electron transfer from a redox partner reduces the iron to the ferrous form which then can bind atmospheric dioxygen. The resultant ferrous-oxy complex in the P450 cytochromes is analogous to that of oxygenated hemoglobin and myoglobin. However, in the case of the P450s, the axial thiolate ligand and the facile interaction with redox partners allows a second electron input to elicit dioxygen bond scission, the production of water, and generation of an active heme-oxidant capable of substrate functionalization. Thus, the ferrous dioxygen intermediate is the key juncture between reversible dioxygen transport and oxygenase metabolism.
A major step in the understanding of the P450 mechanism came in the early 1970s when two groups observed the ferrousoxy state in the P450 CYP101 camphor hydroxylase from Pseudomonas putida (15,16). Similar experiments carried out by Coon and co-workers (17) with P450LM2 and P450LM4 were less successful, because of the relative instability of this state and the aggregate nature of these purified preparations. In the ensuing 30 years, despite the importance of this intermediate in P450 catalysis, the observation and characterization of the oxy-ferrous intermediate in mammalian cytochromes P450 have been problematic. In several cases the oxy complex was reported with heterogeneous kinetic properties, which were variously attributed to an essential complex behavior of the enzyme, to the presence of multiple populations of aggregated forms of P450, or both. This ambiguity leaves many unanswered mechanistic questions about the specific roles of different substrates as determinants of cytochrome P450 kinetics and in the degree of coupling of redox equivalent consumption to product formation. Guengerich and co-workers (2, 18) have published several important works in the initial characterization of the rate-limiting steps of CYP3A4 catalysis, together with detailed kinetic modeling.
In this study, we make use of a self-assembled nanoscale phospholipid bilayer to stabilize human CYP3A4 as a soluble, robust, and monomeric entity. In this system, the cytochrome is present in its native membrane-bound configuration, and the robustness of the monodisperse preparation allows rapid mix stopped-flow measurements. We report here the isolation and detailed spectral characterization of the oxy-ferrous state of CYP3A4, document the dependence of the autoxidation rate on the substrate presence, and analyze these kinetics in terms of a model of redox coupled dioxygen and superoxide interaction with the ferric/ferrous hemoprotein.

EXPERIMENTAL PROCEDURES
Expression and Purification of CYP3A4-Cytochrome P450 3A4 was expressed from the NF-14 construct in the PCWoriϩ vector with a C-terminal pentahistidine tag generously provided by Guengerich and co-workers (19). The presence of the His tag does not perturb turnover parameters of CYP3A4 (20). Heterologous expression and purification from Escherichia coli were carried out using a modified procedure (21) as described in the supplemental material. CYP3A4-Nanodiscs were prepared in a substrate-free form and kept at 4°C.
The application of the Nanodisc system for solubilization of integral membrane proteins incorporated into nanoscale bilayers has been described in detail in several publications (21)(22)(23)(24)(25). Assembly of human CYP3A4 in Nanodiscs was accomplished using the scaffold protein MSP1D1 with the poly(histidine) tag (26) removed as described previously (21). Briefly, purified CYP3A4 from the E. coli expression system was solubilized by 0.1% Emulgen 913 and mixed with a disk reconstitution mixture containing MSP1D1, POPC, 2 and sodium cholate present in 1:65:130 molar ratios. Detergents were removed by treatment with Biobeads (Bio-Rad), which initiates self-assembly. The result of this self-assembly reaction is a monomer of CYP3A4 contained in a discoidal POPC bilayer ϳ10 nm in diameter stabilized by the encircling amphipathic membrane scaffold protein belt (see Supplemental Material).
Substrate Binding-Formation and decay of the oxy-ferrous complex was studied for CYP3A4 saturated with testosterone (TS, 200 M) or bromocriptine (BC, 3 M) and in the substratefree form. Substrates were added using stock solutions in methanol with the final concentration of methanol always less than 1%. Representative examples of spectral titration of CYP3A4 Nanodiscs with TS were published earlier (21) and with BC are shown in supplemental Fig. 2S. For both substrates, a spin con-version higher than 90% was reached at saturating concentrations, confirming the structural and functional homogeneity of CYP3A4 in Nanodiscs (21), and in agreement with the recent study of reduction kinetics of CYP3A4 (27).
Solutions of CYP3A4 in Nanodiscs were deaerated under the flow of argon gas and reduced by addition of a small excess of anaerobically prepared dithionite solution with the concentration determined using a molar absorption coefficient ⑀ 315 ϭ 8.05 mM Ϫ1 cm Ϫ1 (28). Complete reduction was confirmed by anaerobic absorption spectroscopy.
Stopped-flow Experiments-All stopped-flow experiments utilized an Applied Photophysics SX.18MV stopped-flow spectrophotometer using typical mixing volumes of 150 l in each syringe with a dead time of 1.5 ms. Anaerobic solutions of reduced CYP3A4 Nanodiscs (with or without substrate) in one syringe were mixed in 1:1 ratio with buffer saturated with pure oxygen gas containing the same concentration of substrate, if present. The resulting high oxygen concentration (ϳ690 M after mixing) was chosen in order to accelerate formation of the oxy complex. Replicate experiments were completed at different temperatures from 5 to 37°C. In each experiment from 400 to 800 spectra were collected on a logarithmic time scale for the slow reactions (total collection time 50 s) or linear time scale for faster processes (collection time up to 1 s). The integrity of CYP3A4 Nanodisc after stopped-flow mixing was tested in separate experiments by size-exclusion high pressure liquid chromatography, and the samples appeared indistinguishable from those before the stopped-flow experiments.
Data Analysis-The resulting data were analyzed according to Equation 1, where Fe 2ϩ and Fe 3ϩ denote the reduced (ferrous) and oxidized (ferric) state of CYP3A4 heme iron; k 1 is the apparent pseudo first-order rate constant of oxygen binding to ferrous cytochrome P450, and k 2 is the rate constant for autoxidation, i.e. formation of ferric CYP3A4 and release of superoxide. In most experiments oxygen binding was complete within the dead time of the instrument, and k 1 could only be approximated. In case of CYP3A4 saturated with BC at low temperatures, however, formation of the oxy complex was slow enough to resolve as a separate process completed during the first 30 -40 ms after mixing. Spectra collected in each kinetic experiment were arranged into a matrix with each column vector representing the spectra, and rows representing the absorption at each wavelength as a function of time. The number of independent spectrally distinguishable components and their time-dependent concentrations were determined using singular value decomposition (SVD) (29). For substrate-free CYP3A4 as well as for the experiments conducted in the presence of TS, only two spectral components corresponding to the spectra of oxy-ferrous and the final ferric P450 were derived from the spectral data sets. In the case of CYP3A4 saturated with BC, the initial oxygen binding was slower than in other samples, and in the third component, the initial spectrum of deoxy-ferrous CYP3A4 was also possible to discern. The spectra of pure species and rate constants were also calculated using the program SPECFIT (Spectrum Software Associates, Marlborough, MA).

RESULTS AND DISCUSSION
The ferrous-oxy complex of the cytochromes P450 is at the critical bifurcation point between a commitment to oxygenase catalysis and simple reversible dissociation of atmospheric dioxygen. This transient intermediate was isolated and characterized in the microbial P450 CYP101 from Pseudomonas by spectroscopic methods (30 -33) and x-ray crystallography (34). Similar structural and functional characterization of the ferrous-oxy complex in mammalian cytochromes P450 has met with significant frustrations due to the membrane-bound nature of these monooxygenases and the difficulty in generating a monomeric and monodisperse solution preparation for biophysical investigations.
Using the Nanodisc self-assembled phospholipid bilayer system (22), we incorporated human CYP3A4 into a homogeneous prep-aration and characterized the fundamental substrate-binding isotherms. With both substrates used in this work, the spin conversion upon saturation was higher than 90%, indicating functional homogeneity in equilibrium titration experiments. Fig. 1A shows spectra and decay kinetics of the oxy complex of CYP3A4 in Nanodiscs without substrate present. The spectrum of the oxy complex shows a Soret band at 418 nm and a single broad maximum at 552 nm in the visible region. This spectrum looks very similar to the spectra of bacterial oxygenated cytochromes P450 (15,35,36). The rate of autoxidation of substratefree CYP3A4 in Nanodiscs is very high, k ϭ 20 s Ϫ1 at 5°C and ϳ140 s Ϫ1 at 29°C. Because of extremely fast decomposition of the oxy complex, it was impossible to measure the rate of this reaction at the physiological temperature of 37°C. This means that, even if reduction of the substrate-free CYP3A4 is possible in vivo, the oxy complex, once formed, would decompose much faster than the subsequent steps of enzymatic catalysis. Such fast autoxidation reactions have not been reported previously for other cytochromes P450 ( Table 1).
The results of stopped-flow experiments with CYP3A4 saturated with TS are shown in Fig. 1B. The oxy complex is formed within several milliseconds after mixing a deoxygenated solution of reduced CYP3A4 with oxygen. Even at low temperatures, e.g. 6°C, the apparent first-order rate of oxygen binding could only be estimated as 350 -400 s Ϫ1 , corresponding to a second-order rate of 5 ϫ 10 5 M Ϫ1 s Ϫ1 . At higher temperatures, oxygen binding is completed even faster and could not be measured with the current experimental methodologies. The spectra of the oxy complex of CYP3A4 saturated with TS shows a broad Soret band with a Soret maximum at 424 -425 nm, which is red-shifted as compared with the 418 -420 nm values observed for the oxy complex of other cytochromes P450 and resembles the spectra of oxygenated nitric-oxide synthase and chloroperoxidase, which also have thiolate as a proximal ligand to the heme iron. This shift may be caused by the specific perturbation of the heme and/or dioxygen ligand by substrate molecule bound in the active site of CYP3A4, as was suggested recently for nitric-oxide synthase (37)(38)(39). The shape of absorption spectra of the CYP3A4 oxy complex with TS bound and the positions of the main absorption bands do not depend on temperature in the range from 6 to 37°C.
Similar results were obtained with CYP3A4 saturated with BC. Representative spectra of the CYP3A4 oxy complex with BC bound are shown in Fig. 1C. For these samples oxygen binding was slower by an order of magnitude, with the apparent first-order binding rates ranging from 24 s Ϫ1 at 6°C to ϳ300 s Ϫ1 at 32°C (data not shown). At all temperatures the spectra of the oxy complex look very similar, with the Soret maximum at 420 nm.
Once formed, the oxygenated CYP3A4 quickly autoxidizes  via a single exponential process with no spectrally distinguishable intermediates. SVD analysis of experimental data shows that all data sets can be described with high precision with only two basis spectra corresponding to oxygenated and high spin ferric CYP3A4. At the lowest temperatures, several spectra measured at the first 10 ms with TS and within the first 20 -120 ms with BC were omitted from such analyses, because they revealed the presence of ferrous CYP3A4 before oxygenation was completed.
Kinetics of autoxidation of substrate-free CYP3A4 and of CYP3A4 in the presence of substrates at different temperatures is shown in Fig. 2 together with the corresponding fits to the single-exponential decay. Careful analysis of the data obtained for CYP3A4 in the presence of 3 M BC reveals the presence of the second process with a typical amplitude of 15-20% at all temperatures. This second process can be tentatively attributed to the presence of several conformers of CYP3A4 and/or multiple BC binding modes, which are in slow equilibrium on the time scale of our experiments. Because the major fraction of autoxidation kinetics could be approximated reasonably well with a singleexponential decay, we used the same simple model to fit the data obtained with BC to compare the influence of two different substrates on the autoxidation rates and to calculate the activation parameters of this reaction.
Both substrates were found to significantly stabilize the oxy complex against autoxidation. With BC this effect is more pronounced, the rates of spontaneous decomposition of oxygenated CYP3A4 vary from 0.12 s Ϫ1 at 6°C to 2.6 s Ϫ1 at 37°C. With TS the rates at the same temperatures are 0.37 and 20 s Ϫ1 , i.e. 3-8 times higher. The stabilizing effect of the substrate on the oxy complex in CYP3A4 is slightly more pronounced with BC and is the same with TS as compared with that in CYP101 with camphor (16). The kinetics of autoxidation of heme proteins at high oxygen concentration is determined mostly by the rate of direct escape of superoxide or hydroperoxyl radical from the binding pocket (40 -42), although the experimentally observed rate of decay of the oxy-ferrous complex may also depend on possible nucleophilic ligand displacement (42)(43)(44). The differences observed between the autoxidation rates measured in CYP3A4 saturated with BC versus TS imply that the presence of substrates could modulate the escape of superoxide from the distal binding pocket of the protein. The rates of autoxidation measured at 24°C, 3.6 s Ϫ1 for TS and 0.84 s Ϫ1 for BC, are of the same order as the dissociation rates of these substrates, 20 s Ϫ1 and 0.25 s Ϫ1 at 23°C (45). For substrate-free protein, autoxidation at 6°C is 60 times faster than with TS, and more than 150 times faster than with BC. However, autoxidation is very fast, even in case of substrate-saturated CYP3A4, and is much faster than in other cytochromes P450. This suggests that autoxidation may be a key factor that determines the uncoupling behavior and production of toxic superoxide by cytochromes P450 in humans.
The temperature dependences of autoxidation rates were analyzed according to the Eyring equation in order to derive the activation enthalpies and entropies of this reaction in the absence and in the presence of substrates (46). For CYP3A4, we obtained similar free energies of activation (Fig. 3) to those measured for CYP101 (16) and CYP11A1 (47,48). The mechanism of autoxidation in CYP3A4 is likely the same, but the rates are faster by 2-3 orders of magnitude both with and without substrate. Analysis shows that the activation entropy provides a more significant contribution to the temperature dependence of autoxidation rate measured for TS-saturated CYP3A4 than for the BC-CYP3A4 complex. This may indicate the difference in the active site volume available to the superoxide prior to its escape from the interior of the protein molecule.
In well characterized systems such as CYP101 (16,35) and CYP102 (49 -51), the overall stability of the oxy complex is  known to depend on the presence of the substrate. However, in these bacterial cytochromes P450, the consumption of pyridine nucleotide is effectively coupled with product formation, and this first branch point does not contribute significantly to the overall uncoupling because the overall turnover rates are typically 10 -40 s Ϫ1 , i.e. much faster than the autoxidation rates of the oxy complex, which are in the range of 0.002-0.05 s Ϫ1 in these enzymes. On the contrary, product formation in human cytochromes P450 is usually much slower, 0.02-0.25 s Ϫ1 , and significantly lower than the rates of spontaneous decomposition of the oxy complex in CYP3A4 measured in this work. The essential feature of these P450s is the overall high degree of uncoupling, which though well established in general, but usually is not assigned to any specific step of P450 cycle (for review see Ref. 52).
Our results provide new information on the properties of oxygenated CYP3A4 solubilized in monomeric form in a POPC bilayer. We show that CYP3A4 in Nanodiscs is monodisperse and kinetically homogeneous with respect to formation and oxidative decomposition of oxy-ferrous complex. Without substrate, the oxy complex in CYP3A4 autoxidizes very quickly and returns to the ferric low spin state. Saturation of CYP3A4 with TS or BC significantly stabilizes this oxy-ferrous state. This effect is similar to the pronounced stabilization of other cytochromes P450 by their substrates against autoxidation, and together with the dramatic acceleration of reduction rate by the same substrates, it may be attributed to a positive shift of redox potential caused by substrate binding to cytochromes P450.
The possibility of essential uncoupling of NADPH consumption by microsomal cytochrome P450 through the direct decomposition of oxy-ferrous complex and the dependence of this process on the enzyme saturation by substrate has been suggested previously (53). For most of the well characterized cytochromes P450, the autoxidation rates are lower than the total rate of pyridine nucleotide consumption and the rate of catalytic turnover. For the case of CYP3A4 in the absence of substrate, however, the oxy complex decomposes extremely quickly and likely is not stable enough to be effectively reduced to the peroxo level as required for monooxygenase catalysis. The latter kinetic selection may serve as an additional mechanism of regulation of NADPH consumption through coupling with the substrate binding, together with the known thermodynamic trigger, spin shift, and acceleration of the first reduction rate (54).
Further experimental studies focused on other reaction steps are needed to complete the kinetic characterization of the full enzymatic cycle of CYP3A4 and to elucidate the relative roles in total uncoupling at three branching intermediates, namely the oxy-ferrous, peroxo-and hydroperoxo-ferric, and ferryl-oxo heme complexes. The kinetic partitioning between the productive pathway and abortive decay of these intermediates critically depends on their stability as well as on the efficiency of the electron transfer from cytochrome P450 reductase and possibly other components, for example cytochrome b 5 . The overall results of such in vitro studies depend on the stoichiometry of all proteins in the reconstituted system as well as on other factors. To our knowledge there are no available data on the rate of the reduction of oxy-ferrous intermediate for CYP3A4 or any other mammalian membrane-bound cytochrome P450. However, the rate of spontaneous oxidative decomposition documented in this study for the oxygenated CYP3A4 saturated with TS, 20 s Ϫ1 at the physiological temperature of 37°C, can be used as the reference value for the estimation of the efficiency of the second electron transfer and for the uncoupling ratio at the first branch point. If the reduction of the oxygenated CYP3A4 is also 20 s Ϫ1 at the same conditions, the uncoupling due to autoxidation will be at least 50%. If it is slower, the major fraction of redox equivalents from reductase will be consumed through this autoxidation pathway with concomitant release of superoxide. This latter option is in agreement with the direct comparison of NADPH oxidation rate and superoxide production by human microsomes enriched in CYP3A4 (4).
In summary, we report the formation and decay of the oxy complex in CYP3A4 with and without bound substrates. The use of Nanodiscs for solubilization and isolation of monomeric functional CYP3A4 incorporated in POPC bilayer provides a stable and homogeneous enzyme preparation. Previous work (21) noted that CYP3A4 in Nanodiscs undergoes almost 100% conversion upon saturation with TS at 293 K. In this work the same complete conversion to the high spin state was obtained upon saturation with TS and BC at the physiological temperature range. Oxygen binding to the reduced CYP3A4 in stoppedflow experiments generates a transient oxy complex, also with 100% yield, and autoxidation of CYP3A4 with and without substrate is well described by single exponential decay. In all cases the final reoxidized protein remains fully functional.
These results suggest that the autoxidation is a very important process for the overall uncoupling reactions of human liver cytochromes P450. The lifetime of the oxy complex in the absence of substrate is negligible as compared with the NADPH consumption rate measured for substrate-free CYP3A4 (2, 6,45) and with the total turnover rate. For the substrate-bound CYP3A4 autoxidation is also fast, and with TS bound at 310 K, it is much faster than turnover measured in vitro (55)(56)(57)(58). Our results suggest that the oxy-ferrous intermediate of the P450 cycle may be the most important natural determinant of CYP3A4 activity and human drug metabolism. It represents the first branching point between the productive pathway of the substrate turnover and the abortive decay with the formation of superoxide. The coupling of redox equivalent consumption with the substrate turnover at this step of the P450 catalytic cycle is determined by the lifetime of the oxy complex, which must be sufficiently stable to permit the second electron transfer from P450 reductase. The presence of substrate at this step is critically important for stabilization of oxygenated CYP3A4, for the efficient use of redox equivalents, and for minimization of the side reaction generating toxic superoxide and peroxide. These observations help to understand the multiple regulatory roles of substrates in the physiological functioning of CYP3A4.