Cytochrome P450 3A Enzymes Catalyze the O6-Demethylation of Thebaine, a Key Step in Endogenous Mammalian Morphine Biosynthesis*

Background: Mammals synthesize endogenous morphine; the enzyme catalyzing thebaine O6-demethylation, a key late step, is uncharacterized. Results: Human cytochromes P450 (P450) 3A4 and 3A5 catalyzed thebaine O6-demethylation; the P450 3A-selective drug ketoconazole inhibited the reaction in human liver microsomes and rat brain homogenates. Conclusion: P450s 3A4 and 3A5 catalyze thebaine O6-demethylation in humans. Significance: Enzymes catalyzing all oxidations in the latter steps of mammalian morphine biosynthesis have been identified. Morphine, first characterized in opium from the poppy Papaver somniferum, is one of the strongest known analgesics. Endogenous morphine has been identified in several mammalian cells and tissues. The synthetic pathway of morphine in the opium poppy has been elucidated. The presence of common intermediates in plants and mammals suggests that biosynthesis occurs through similar pathways (beginning with the amino acid l-tyrosine), and the pathway has been completely delineated in plants. Some of the enzymes in the mammalian pathway have been identified and characterized. Two of the latter steps in the morphine biosynthesis pathway are demethylation of thebaine at the O3- and the O6-positions, the latter of which has been difficult to demonstrate. The plant enzymes responsible for both the O3-demethylation and the O6-demethylation are members of the FeII/α-ketoglutarate-dependent dioxygenase family. Previous studies showed that human cytochrome P450 (P450) 2D6 can catalyze thebaine O3-demethylation. We report that demethylation of thebaine at the O6-position is selectively catalyzed by human P450s 3A4 and 3A5, with the latter being more efficient, and rat P450 3A2. Our results do not support O6-demethylation of thebaine by an FeII/α-ketoglutarate-dependent dioxygenase. In rat brain microsomes, O6-demethylation was inhibited by ketoconazole, but not sulfaphenazole, suggesting that P450 3A enzymes are responsible for this activity in the brain. An alternate pathway to morphine, oripavine O6-demethylation, was not detected. The major enzymatic steps in mammalian morphine synthesis have now been identified.

nism of action is as an agonist for the -opioid receptors that are distributed throughout the brain. Activation of these receptors is associated with analgesia, sedation, euphoria, physical dependence, and respiratory depression. The -binding sites are discretely distributed in the human brain, with high densities in the posterior amygdala, hypothalamus, thalamus, nucleus caudatus, putamen, and certain cortical areas (1).
A considerable body of evidence exists that morphine is present in the tissues of various animals that have not been medicated with morphine or other related opioids. In mammals, morphine has been detected in skin, lung, spinal cord, and, most notably, in the brain (2)(3)(4). The presence of morphine in brain is of particular interest due to the presence of the -opioid receptors. In the rat, morphine levels in structures of the brain have been quantified and range from 26 fmol/g of tissue (found globally in the brain) to 7.2 pmol/g of tissue measured in the hypothalamus (3,(5)(6)(7). One study quantified morphine in the temporal lobe from one human brain tissue at 3.4 ng/g of tissue (8). The presence of morphine in brain tissue in many mammals and the evidence from a human sample suggest that endogenous morphine is present in human brain.
Endogenous formation of morphine has been demonstrated in human cells in culture. In a seminal study, human neuroblastoma cells incubated in the presence of 18 O 2 produced 18 Olabeled morphine (9). Another study demonstrated morphine synthesis by human white blood cells and polymorphonuclear cells in a precursor-dependent manner (10,11). Taken together, these studies suggest that morphine biosynthesis can occur in the human brain. Humans have also been reported to excrete both codeine and morphine in urine, in the absence of treatment with these compounds (12).
The physiological role of endogenous morphine is currently unknown. The endogenous levels that have been quantified are below the plasma concentrations (M range) achieved during therapeutic use (13). Many hypotheses have been presented regarding the purpose of endogenous morphine that suggest a role in infection, sepsis, or inflammation, as well as neurological pathologies (Parkinson disease and schizophrenia) (14 -17).
Mammalian morphine biosynthesis appears to be highly similar to the pathway in the opium poppy, with at least some common intermediates (18 -20). The multistep biosynthetic pathway begins with the amino acid L-tyrosine, and the final steps in this pathway include the conversion of thebaine to morphine via two critical demethylation steps at the O 3 -and O 6 -positions. Two parallel pathways exist for conversion of thebaine to morphine, in which O 3 -and O 6 -demethylation reactions occur, potentially in either order ( Fig. 1) (21). Both of these pathways are plausible, in that thebaine, codeine, and oripavine are present in both mammals as well as the opium poppy (5,(22)(23)(24).
Although the intermediates are known and there is evidence that some of the reactions occur in mammals (9,25,26), the enzymes that catalyze these demethylation steps have only recently been identified. The enzymes that catalyze the O 3 -and O 6 -demethylation steps in the plant pathway are members of the Fe II /␣-ketoglutarate-dependent dioxygenase enzyme family (27). Specifically, in plants, thebaine O 6 -demethylase catalyzes O 6 -demethylation of both thebaine and oripavine; codeine O-demethylase catalyzes O 3 -demethylation of both thebaine and codeine (Fig. 1). Independent of plant studies, the discovery that O 3 -demethylation of thebaine and codeine can be catalyzed by the human enzyme cytochrome P450 (P450) 2 2D6 (28) has focused our search for the O 6 -demethylase to P450s, but this enzyme(s) has remained unknown, and the Fe II / ␣-ketoglutarate dioxygenases can be considered.
Both the P450 and the Fe II /␣-ketoglutarate-dependent dioxygenase families are known to catalyze diverse sets of reactions and are candidates for mediating the O 6 -demethylation of thebaine and oripavine. The P450 superfamily catalyzes the oxidation of most organic substances, generally using NADPH as an electron donor (29). The general reaction catalyzed by P450s is mixed function oxidation, manifestations of which include carbon hydroxylation, dealkylation of heteroatomic substrates, heteroatom oxygenation, and the oxidation of unsaturated compounds to epoxides and phenols (30). Fe II /␣ketoglutarate-dependent dioxygenases constitute a large family of soluble, non-heme iron-containing oxidases that couple the decarboxylation of ␣-ketoglutarate to activation of dioxygen to catalyze a variety of oxidation reactions. This family of enzymes most frequently catalyzes carbon hydroxylation, but other formal two-electron oxidations such as desaturation, cyclization, and halogenation are also known (31,32). Except for the two dioxygenases involved in the plant morphine biosynthesis (27,33), O-demethylation by the Fe II /␣-ketoglutarate-dependent dioxygenases has not been previously reported, to our knowledge.
The present study was designed to identify the enzyme(s) responsible for O 6 -demethylation of thebaine in mammals, humans in particular. Previous attempts to characterize the enzymes that catalyze O 6 -demethylation of thebaine and oripavine have been unsuccessful, in large part due to instability of codeinone and morphinone in aqueous solution (34,35). In the present work, we devised and validated a coupled enzyme assay to circumvent this issue. We established that human P450s 3A4 and P450 3A5 and rat P450 3A2 catalyze O 6 -demethylation of thebaine. Using rat brain microsomes, we also provide evidence that P450 3A enzymes are also responsible for thebaine O 6 -demethylation in mammalian brain tissue.

Materials
Thebaine, hydrocodone, and hydromorphone were purchased from Sigma-Aldrich. Thebaine was recrystallized from hot acetone. Oripavine and [N-C 2 H 3 ]thebaine were provided by the late M. H. Zenk (Donald Danforth Plant Science Center, St. Louis, MO). CYP3cide was a gift from K. D. Hardy (Lipscomb University, Nashville, TN). All other inhibitors were purchased from Sigma-Aldrich. Rat brains were a gift from C. K. Jones (Department of Pharmacology, Vanderbilt University). Liver microsomes from human liver tissue (set of 10 individual donors) were prepared in this laboratory according to published protocols (36).

Enzymes
Human P450 3A4-and 3A5-containing Baculosomes (microsomal preparation from baculovirus-infected insect cells co-expressing human NADPH-P450 reductase) were purchased from Invitrogen/Life Technologies, and rat P450 3A2containing Baculosomes were purchased from BD Biosciences. A Pseudomonas putida morphinone reductase cDNA vector was provided by N. C. Bruce (University of York, York, UK), and the enzyme was heterologously expressed and purified according to French and Bruce (37); when subjected to SDS-PAGE on a 10% (w/v) acrylamide gel, it showed a single band at 41 kDa.

Preparation of Rat Brain Homogenate and Microsomes
Male Sprague-Dawley rats (350 -450 g, Harlan Laboratories, Indianapolis, IN) used in the present studies were housed under a 12-h light/12-h dark cycle and given ad libitum access to food and water. All animal experiments were approved by the Vanderbilt University Animal Care and Use Committee, and experimental procedures conformed to guidelines established by the National Research Council Guide for the Care and Use of Laboratory Animals. Rats were briefly anesthetized with isoflurane and decapitated. Whole brains were removed and immediately washed in ice-cold 150 mM NaCl to remove blood clots. The brain tissue was homogenized in four volumes of ice-cold 0.10 M potassium phosphate buffer (pH 7.6) containing 0.15 M KCl using 7-10 up-and-down strokes in a manually operated Teflon glass Potter-Elvehjem homogenizer. All steps used to obtain microsomal fractions were carried out at 4°C. Homogenates were centrifuged at 14,000 ϫ g for 15 min at 4°C (to remove debris, nuclei, and mitochondria), and the resulting supernatant was centrifuged at 100,000 ϫ g for 1 h at 4°C. The microsomal pellet was resuspended in 10 mM Tris acetate (pH 7.4) buffer containing 1.0 mM EDTA and 20% glycerol (v/v) (microsomes from 1 g of brain in ϳ0.2 ml of buffer).

Enzymatic Assays
In general, incubations were conducted in a total volume of 300 l in 50 mM potassium phosphate buffer (pH 7.4) containing 10 pmol of P450 enzyme (Baculosomes, Life Technologies or BD Biosciences) or 0.5 mg of protein/ml of tissue homogenate or microsomal preparation. Unless otherwise noted, negative control samples for each experiment contained the same components but were heated in a boiling water bath (5 min) before initiating the reactions.
Additionally, to trap the primary reaction products, morphinone reductase (0.5 M) and NADH (1 mM) were included to convert codeinone and morphinone to the stable products hydrocodone and hydromorphone, respectively (37) (see Fig.  2A). Because the substrates codeinone and morphinone are unstable and we did not have a standard, the conditions were validated by two preliminary experiments. A reported substrate of morphinone reductase is 2-cyclohexen-1-one (38). Therefore, to validate the activity of morphinone reductase, we repeated the experiment described by Barna et al. (38) and measured the rate of 2-cyclohexen-1-one reduction. Our results give a rate of 1.3 s Ϫ1 under anaerobic conditions, which is similar to that reported (k cat 0.8 s Ϫ1 ). Anaerobic assays were done using the basic apparatus design of Burleigh et al. (39) as subsequently modified in this laboratory (40,41). Additionally, morphinone reductase was titrated in an incubation with P450 3A5 (described below), and saturation was seen at 0.2 M morphinone reductase (see Fig. 2B).
Following temperature equilibration to 37°C for 5 min, the reactions were initiated by the addition of an NADPH-regenerating system (0.5 mM NADP ϩ , 10 mM glucose 6-phosphate, and 1 IU ml Ϫ1 yeast glucose 6-phosphate dehydrogenase (36)) to analyze P450-dependent activity (42). Reactions generally proceeded at 37°C for 20 min and were terminated by the addition of 300 l of ice-cold CH 3 OH. Quinidine (as internal standard, 0.5 mol) and sodium borate buffer, pH 9.5 (300 l of 0.5 M solution) were added. The samples were mixed using a vortex device and extracted with 1.0 ml of CH 2 Cl 2 , and the layers were separated by centrifugation (10 3 ϫ g for 10 min). A 0.8-ml aliquot of the CH 2 Cl 2 layer (lower phase) was transferred to a clean tube. An additional 1.0 ml of CH 2 Cl 2 was added to the residual layer to extract the products, followed by mixing and another centrifugation step at 10 3 ϫ g. The organic layers were combined, and the solvent was removed under an N 2 stream. All samples were analyzed by LC/MS/MS (see below).

Initial Activity Screens
Initial screening for O 6 -demethylation activity was performed as described above using human brain homogenate, human liver microsomes, or no enzyme source. In these experiments, P450-dependent activity was initiated by the addition of the NADPH-regenerating system, and dioxygenase-dependent activity was initiated by the addition of ␣-ketoglutarate (1 mM), Fe(NH 4 ) 2 (SO 4 ) 2 (100 or 20 M), and L-ascorbate (1.8 mM). In all cases, morphinone reductase and NADH were also present. Control samples contained the same components that were boiled before initiating the reactions. Control samples contained human liver homogenate, but no NADPH-regenerating system or ␣-ketoglutarate, Fe(NH 4 ) 2 (SO 4 ) 2 , or L-ascorbate.

Steady-state Kinetic Assays
Kinetic assays were performed as described above using Baculosomes containing P450 3A4 or 3A5. Additionally, thebaine concentrations of 1.6, 3.1, 6.3, 12.5, 25, 50, and 100 M were used. Results were fit to hyperbolic plots in GraphPad Prism (GraphPad Software, La Jolla, CA) to estimate k cat and K m values Ϯ S.E.

Oxidation of d 3 -Thebaine
Assays were performed as described above using pooled human liver microsomes or Baculosomes containing P450 3A4 or 3A5. In addition to using thebaine as a substrate, separate incubations were done with d 3 -thebaine ([N-2 H 3 ]thebaine) (100 M) as substrate.
P450 3A2 and Rat Brain Microsomes-Ketoconazole or sulfaphenazole (0, 0.2, or 2 M) was included in incubations containing P450 3A2 (4 pmol) or rat brain homogenate (800 g of protein) as described above. Control samples were incubated with and without vehicle present (0.5% CH 3 OH, v/v). Concentrations used were chosen based on published selectivity of inhibitors for rat P450s (49).
Identification of thebaine product analytes was based on retention times (t R ), the ratios of two different MRM transitions, and comparison with the expected values for standards. MRM transitions used were thebaine (m/z 3123351, m/z 3123381), oripavine (m/z 2983218, m/z 2983249), hydrocodone (m/z 3003199, m/z 3003215), and hydromorphone (m/z 2863201, m/z 2863242). Quantitation was done by constructing standard curves for each analyte (using the internal standard quinidine).and integrating the peak areas of one MRM transition with Xcalibur software (Thermo Scientific). The limit of quantitation for hydrocodone was 5 pmol, and the limit of detection was 1 pmol.
Identification of products of d 3 -thebaine was based on the t R of the non-deuterated standard and the expected m/z. The relevant masses of the compounds of interest were: for d 3 -thebaine, m/z 315; for oripavine, m/z 301; for hydrocodone, m/z 303; and for northebaine, m/z 298. Northebaine was not available as a standard, and therefore quantitation was based using a standard curve prepared using thebaine. The limit of detection for thebaine was 500 fmol, and the limit of quantitation was 2 pmol.

Results
Assays for O 6 -Demethylation-The O 6 -demethylation products morphinone (from oripavine) and neopinone/codeinone (from thebaine) are Michael acceptors and are inherently unstable, and preliminary attempts to demonstrate their formation in liver homogenates, microsomes, or purified P450s using LC-UV assays were unsuccessful. We also tried LC-MS assays, including ones in which we added NaBH 4 or NaBD 4 to trap the products, without success, probably because the approach only stabilized the product in solution at the completion of the incubation.
The use of coupled assays is well established in biochemistry. A bacterial morphinone reductase has been extensively studied as a model flavoprotein (34,37,50) and is known to rapidly reduce both morphinone and codeinone. The products, hydromorphone and hydrocodone, are stable and can easily be assayed (e.g. positive-ion ESI LC-MS). In preliminary assays, the concentration of morphinone reductase we established (0.5 M) was found to be saturating in detecting product formation (see below) (Fig. 2B).
P450-dependent O 6 -Demethylation of Thebaine-Incubations of thebaine and human liver microsomes in the presence of an NADPH-regenerating system and the morphinone reductase-NADH trapping system resulted in the formation of three products that were not seen in control reactions with heatinactivated enzymes. These metabolites were subjected to MS after separation by UPLC. Comparison of two of the compounds with standards of alkaloids that were expected to be formed allowed the identification of the unknown compounds as oripavine (via O 3 -demethylation, Fig. 1) (21) and hydrocodone (via O 6 -demethylation followed by morphinone reductase-catalyzed reduction, Figs. 1 and 2) (37) (Fig. 3). The third compound had an m/z of 298, suggesting the loss of a methyl group; however, the retention did not match that of any standards and was further investigated using [N-C 2 H 3 ]thebaine (see below). The identification of oripavine as a P450-dependent product is consistent with the previous findings that incubation of P450 2D6 with thebaine resulted in oripavine formation (51). Hydrocodone was also identified in samples that contained human liver homogenates, thebaine, ␣-ketoglutarate, ascorbate, and 100 M Fe II but not in a similar sample that contained only 20 M Fe II . The amounts of hydrocodone formed are compared in Fig. 4.
Iron-dependent O 6 -Demethylation of Thebaine-To further understand the iron concentration dependence of the ␣-ketoglutarate dioxygenase reaction observed with liver homogenates, thebaine was incubated in the absence of any enzyme but in the presence of ␣-ketoglutarate, L-ascorbate, and varying concentrations of Fe II (Fig. 5). The amount of hydrocodone produced appeared to be linearly dependent on the Fe II concentration at or above 100 M, suggesting that the hydrocodone seen in the initial screens was generated non-enzymatically and only due to the presence of the iron.
P450 3A-dependent O 6 -Demethylation of Thebaine-A panel of selective inhibitors was used to determine which P450s were responsible for the O 3 -and O 6 -demethylation of thebaine in human liver microsomes. Most notably, quercetin and ketoconazole decreased the production of hydrocodone by 40 and 73%, respectively, and quinidine decreased the production of oripavine by 91% (data not shown). Quinidine is known to be a selective inhibitor of P450 2D6, and that result is consistent with our previous findings (51). Concentration-dependent inhibition experiments were done with quercetin and keto-  or human liver microsomes that were boiled to inactivate enzymes and NADPH-generating system (G and H) and compared with authentic standards of oripavine (C and D) and hydrocodone (E and F). Traces shown in black are for oripavine (m/z 2983218, 249); traces shown in red are for hydrocodone (m/z 3003199, 215). Relative ion intensity is shown on the y-axis conazole. Quercetin (P450 2C8 inhibitor) did not show concentration dependence (results not shown) and therefore was not further investigated further.
Thebaine O 6 -Demethylation Catalyzed by Recombinant P450 3A4 and 3A5-Steady-state kinetic parameters for recombinant P450 3A4-and P450 3A5-mediated thebaine O 6 -demethylation were determined based on the resulting hydrocodone. The results displayed Michaelis-Menten kinetics, and the catalytic parameters were estimated (Fig. 7). P450 3A4 and 3A5 had similar apparent K m values (22 and 26 M, respectively); however, the maximum rate (k cat ) for the conversion of thebaine to codeinone by P450 3A5 (52 min Ϫ1 ) was 23-fold higher than that of P450 3A4 (2.3 min Ϫ1 ) (Fig. 7). For reference, the previously determined parameters for P450 2D6 conversion of thebaine to oripavine were k cat 4.6 min Ϫ1 and K m 48 M (38) (k cat /K m 0.1 M Ϫ1 min Ϫ1 ).
N-Demethylation of d 3 -Thebaine by Human Liver Microsomes, P450 3A4, and P450 3A5-To confirm the identity of the unidentified analyte with m/z 298 (see above), d 3 -thebaine was used as substrate in incubations. An analyte loss of 14 atomic mass units indicates a loss of a methyl group. There are three methyl groups on thebaine; an O-methyl at both the 3-and the 6-positions and an N-methyl. The products of O 3 -demethylation (oripavine) and the O 6 -demethylation (hydrocodone after reduction by coupled enzyme system) were already identified, leaving the N-demethylation product, northebaine, as the likely third product. By using thebaine and d 3 -thebaine as substrates in parallel experiments and comparing the products, we were able to determine that northebaine was indeed the third analyte.
When thebaine and d 3 -thebaine were included as substrates in parallel experiments, with pooled human liver microsomes, three products were observed. The first product (t R 1.9 min), which had an m/z of 298 when starting with thebaine and an m/z of 301 when starting with d 3 -thebaine, was oripavine. The second (t R 2.05 min) had an m/z of 300 when starting with   thebaine and an m/z of 303 when starting with d 3 -thebaine, consistent with hydrocodone. The third, t R 2.4 min, had an m/z of 298 when starting with thebaine or with d 3 -thebaine, consistent with loss of the N-methyl group, giving northebaine. The northebaine gave fragment ions distinct from hydrocodone (m/z of 2983281, 266). Northebaine was also identified as a product when thebaine or d 3 -thebaine was included as substrates with P450s 3A4 or 3A5. Based on one time point and a standard curve of thebaine, rate estimates were calculated for the conversion of thebaine to northebaine by P450s 3A4 and 3A5 (88 and 27 pmol/pmol of P450/min, respectively.) Thebaine O 6 -Demethylation by P450 3A2 and Rat Brain Microsomes-We have been unsuccessful in demonstrating any P450-dependent activity in frozen brain samples obtained from cattle or rats. Stored human brain samples (Vanderbilt University tissue facility) or those provided by the Cooperative Human Tissue Network (Birmingham, AL; Nashville, TN) were also negative for known P450 reactions, e.g. 7-ethoxycoumarin O-deethylation. We were able to detect activities in microsomes prepared from freshly obtained rat brains, as reported previously (55). Thebaine O 6 -demethylation was detected with recombinant rat P450 3A2, a prominent rat Subfamily 3A P450 (Fig. 8A), or rat brain microsomes (Fig. 8B). Rates were estimated to be 0.75 fmol/min/pmol of P450 and 1.3 fmol/min/mg of total protein for P450 3A2 and rat brain microsomes, respectively (based on a single time point, 15 min). Additionally, O 6demethylation of thebaine was inhibited in both P450 3A2 and rat brain microsomes in the presence of either ketoconazole (Subfamily 3A-selective). Sulfaphenazole (P450 Subfamily 2C-selective (48)) did not inhibit O 6 -demethylation by either the P450 3A2 or rat brain microsomes.

Discussion
A wealth of complementary biochemical, molecular, and physiological studies from independent laboratories have provided evidence supporting the biosynthesis of morphine in a variety of animal cells and organs (see Ref. 21 for comprehensive review). These studies have characterized multiple enzyme-catalyzed reactions and chemically defined intermediate precursor molecules that share marked similarities to the plant biosynthetic scheme previously established in the poppy P. somniferum (3-5, 9, 10, 22, 24 -26, 51, 56 -58). The final steps of the pathway convert thebaine to morphine, and thebaine must undergo demethylation at both the O 3 -and O 6 -positions (Fig. 1). Thus, two parallel pathways are possible for the conversion of thebaine to morphine: one in which O 3 -demethylation precedes O 6 -demethylation and the other in which the order is reversed. In plants, both pathways occur and the O 3 -andO 6 -demethylationstepsarecatalyzedbycodeineO-demethylase and thebaine O 6 -demethylase, respectively, with the balance differing among species and strains (27). Because both of these enzymes are in the Fe II /␣-ketoglutarate-dependent  dioxygenase family (33), our initial hypothesis was that an enzyme from this family would also be responsible for these demethylation steps in mammalian systems. None of our results support the initial hypothesis; instead, it was shown that both human and rat P450 3A enzymes catalyze the O 6 -demethylation of thebaine. Additionally, the same reaction is catalyzed by a P450 in rat brain microsomes, apparently a member of the P450 3A Subfamily. These findings highlight a key difference in the biosynthesis of morphine in plants versus mammals. Oripavine and northebaine appear to be final products.
The thebaine and oripavine O 6 -demethylation reactions are somewhat unusual in that they both involve cleavage of an enol ether, as opposed to an aliphatic or a phenolic ether. This cleavage is apparently unprecedented in P450 chemistry (30) aside from aryl ethers. As mentioned earlier, initial efforts to directly demonstrate this reaction were unsuccessful, before implementation of the morphinone reductasecoupled assay.
Efforts were made to determine whether mammalian Fe II /␣ketoglutarate-dependent dioxygenases catalyze O 3 -and O 6demethylation of thebaine, based on the precedence in plants.
Our preliminary experiments were designed using typical conditions from the literature, which included exogenous iron concentrations ranging from 50 to 500 M (33,59,60). Although thebaine O 6 -demethylation was observed, we noted that the generation of the product was neither time-dependent nor enzyme concentration-dependent. Experiments done in the absence of any enzyme showed that O 6 -demethylation of thebaine was dependent on the iron concentration (Fig. 5). At concentrations Յ20 M iron, no (non-enzymatic) O 6 -thebaine demethylation was detected. Although total mammalian cellular iron concentrations are estimated between 50 and 100 M, iron homeostasis is highly regulated and free iron represents only a minor fraction (Ͻ5%); even 20 M iron is much higher than the biologically relevant concentration and would be highly deleterious (61). When experiments with brain homogenates were repeated with 20 M iron, no hydrocodone formation was detected. Although no further studies on the mechanism of the iron-catalyzed demethylation were done, it is well known that Fe II catalyzes nonenzymatic hydroxyl radical formation from H 2 O 2 via the Fenton reaction (62)(63)(64), and thus it is probably not surprising that iron-catalyzed (non-enzymatic) demethylation was observed (see also Ref. 65 for iron-catalyzed N-demethylation of oripavine). Because the O-demethylation that was initially observed was an artifact resulting from the high iron concentration, it calls into question the use of high iron concentrations that are not biologically relevant. We cannot speculate on the relevance of our results in the general field of Fe II /␣-ketoglutarate-dependent dioxygenases. Many dioxygenases in this field have been extensively characterized (32, 59, 60, 66 -68). However, our results do suggest caution in the assignment of activities to enzymes in this family in the absence of more extensive characterization.
We report here that thebaine O 6 -demethylation is catalyzed by P450 Subfamily 3A enzymes. The use of P450-selective inhibitors, in combination with human liver microsomes, implicated human P450s 3A4 and 3A5 in catalysis of O 6 -demethylation, confirmed by in vitro experiments with recombinant enzymes. Although both of these enzymes generate the same product, the catalytic efficiency of P450 3A5 was 20-fold higher than that of P450 3A4. When considering the relevance of P450s 3A4 and 3A5 in common substrate metabolism, the much higher liver concentrations and typical superior catalytic efficiency usually focus responsibility toward P450 3A4 (52, 69 -71). Here, however, we are considering P450 activity in the brain, and P450 3A5 is expressed in the brain with mRNA levels comparable with that of P450 3A4 (72). The higher catalytic efficiency and its measured expression in the brain lead to the proposal that P450 3A5 plays an important role in human morphine biosynthesis.
Numerous searches for thebaine O 6 -demethylation using microsomes derived from human brain tissue were attempted but were not presented here because no enzymatic activity was observed for any P450-dependent reaction. This was likely due to the difficulty in obtaining human brain samples that undergo careful treatment necessary to retain enzymatic activity. Thus, microsomes were prepared from freshly acquired rat brains. In the presence of an NADPH-generating system, O 6 -demethylation was observed when thebaine was incubated with these microsomes (Fig. 8B). Additionally, the inhibition of this activity in the presence of ketoconazole (a P450 3A inhibitor that also inhibits rat P450 2C9) and the retention of the activity in the presence sulfaphenazole (a selective P450 2C inhibitor) (49) indicate that thebaine O 6 -demethylation in the brain microsomes is due to one or more of the rat P450 3A enzymes. (P450 activity was also lost in our rat brain tissues upon frozen storage (see also Ref. 73) consistent with the lack of activity in all frozen human brain samples we obtained.) Thus, in addition to showing that the human P450 3A enzymes can catalyze thebaine O 6 -demethylation, we report that this demethylation reaction occurs in rat brain, catalyzed by one or more Subfamily 3A P450 enzymes. The results presented here are consistent with a study that measured NADPH-dependent thebaine metabolism in rat brain microsomes (26). In that study, codeine was generated at a rate of 0.20 Ϯ 0.04 pmol/h/mg of protein (as estimated by a radioimmune assay instead of LC-MS), as compared with the present study in which hydrocodone was generated at a rate of 0.075 pmol/h/mg of protein in the coupled LC-MS assay.
As described, two parallel pathways lead from thebaine to morphine (Fig. 1). The first, in which thebaine is demethylated at the 3-position, yields oripavine, not thebaine, as the substrate for O 6 -demethylation. When oripavine was incubated with either human liver microsomes or recombinant P450 3A4 or P450 3A5, no O 6 -demethylation was observed (data not shown). Thus, the results presented in this study support the second pathway, in which thebaine is first demethylated at the O 6 -position, as the only pathway that contributes to morphine synthesis, at least in human liver.
No clear function has been attributed to endogenous morphine, but suggestions include infection, sepsis, and inflammation, as well as major neurological pathologies (Parkinson disease, schizophrenia) (reviewed in Ref. 21). Our findings indicate that P450 3A5 is most active in the O 6 -demethylation of thebaine in human biosynthesis of morphine (Fig. 7). P450 3A5 is highly polymorphic, and the wild-type allele, CYP3A5*1, is expressed with a frequency of only 5-15% in Caucasian populations, 25-40% in various Asian ethnic groups, and ϳ40 -60% in African and African-American populations (69,74,75). Thus, individuals with P450 3A5 polymorphisms that affect morphine biosynthesis might be expected to have different physiological outcomes related to this. Mikus et al. (12) measured urinary excretion of codeine and morphine in 40 people who had been phenotyped for P450 2D6 status, utilizing an HPLC separation/radioimmune assay for detection. They reported no difference in morphine levels due to P450 2D6 status and concluded that other P450s might be involved in the synthesis of endogenous morphine. The relevance of genetic variation of P450s 3A4 and 3A5 to the P450 2D6 work is unknown. In the work of Mikus et al. (12), considerable interindividual variability was reported (10 3 -fold), but 10-fold dayto-day intra-individual variabilities where also seen. Whether these variations are real or the result of analytical deficiencies is not clear.
Our results can be placed in context of the overall scheme of mammalian synthesis of morphine (Fig. 9). Administration of (R,S)-tetrahydropapaveroline to mice yielded reticuline, salutaridinol, thebaine, and oripavine in urine (24). All of the oxidation steps following the formation of tetrahydropapaveroline have now been characterized in humans in vitro. P450 2D6 or the two Subfamily 3A P450s (3A4 and 3A5) have roles in all oxidation steps examined, including the early step of oxidation of tyramine to dopamine (7). A caveat is that many of the demonstrations of catalytic activity have been done with individual enzymes and the possibility exists that some other P450s may also be found to contribute. Although many of the studies involved in elucidating the steps in Fig. 9 have been done in liver systems (26,51), they do not speak to the localization near the neural targets, which may play an important role in endogenous morphine biosynthesis in the brain. It is possible that an endocrine mechanism is more relevant to the function of endogenous morphine. In conclusion, our results now implicate several major human P450s in the scheme of Fig. 9, and these are found in a number of sites in the body (36). Author Contributions-V. M. K. conducted most of the experiments, analyzed the results, and wrote most of the paper. M. A. R. conducted experiments on the roles of P450s 3A4 and 3A5. T. K. conducted experiments searching for dioxygenase function. F. P. G. conceived the idea for the project, conducted the anaerobic assay of morphinone reductase activity, and wrote the paper with V. M. K.