Grapefruit juice has been found to significantly increase the oral bioavailability of several drugs metabolized by P450 3A4 such as cyclosporine, midazolam, felodipine, and saquinavir (Ducharme et al., 1995; Kupferschmidt et al., 1995, 1998; Bailey et al., 2000). Bergamottin (BG), shown in Fig. 1, and 6′,7′-dihydroxybergamottin (DHBG) are two of the major furanocoumarins found in grapefruit juice (Fukuda et al., 1997; Schmiedlin-Ren et al., 1997; He et al., 1998; Tassaneeyakul et al., 2000). BG and DHBG have been demonstrated to be effective mechanism-based inactivators of P450 3A4 and were thought to be the major active ingredients responsible for the grapefruit juice-drug interaction (Schmiedlin-Ren et al., 1997; He et al., 1998). Although the concentrations of two furanocoumarin dimers, GF-I-1 and GF-I-4, were much lower than that of BG or DHBG, these two dimers were shown to be very potent inhibitors/inactivators of P450 3A4 (Fukuda et al., 1997; Tassaneeyakul et al., 2000). In addition, epoxybergamottin and bergapten (5-methoxypsoralen) from grapefruit were proven to inhibit P450 3A4 activity (Ohnishi et al., 2000; Wangensteen et al., 2003). Furanocoumarin derivatives are also found in several herbal medicines from the families Umbelliferae, Rutaceae, and Leguminosae, and these compounds have shown a striking ability to inhibit the activity of P450 3A (Guo and Yamazoe, 2004). Thus, naturally occurring furanocoumarins from grapefruit juice as well as several plants may alter the pharmacokinetics of clinically used drugs and contribute to the grapefruit juice effect.

A previous study suggested that the mechanism by which BG mediated the inactivation of P450 3A4 was through modification of the apoprotein. However, this hypothesis has not been validated experimentally (He et al., 1998). P450s 3A4 and 3A5 are the major P450s in human liver and intestine microsomes and are responsible for the metabolism of approximately 60% of clinically relevant drugs (Wrighton et al., 1990). Although P450 3A5 is expressed at much lower levels and, in general, metabolizes drugs less efficiently than P450 3A4, P450 3A5 was chosen to elucidate the molecular mechanisms of P450 inactivation by BG in the present study based on the following: P450 3A5 shares 85% of amino acid sequence homology with P450 3A4 (Wrighton et al., 1990), P450 3A5 has been detected in a significantly higher percentage of children under 19 years of age and in African-Americans compared with Caucasians (Kuehl et al., 2001), and P450 3A5 is the primary member of the P450 3A family expressed outside the liver and intestine. P450 3A5 has been found in the adrenal gland, prostate, and the kidney, suggesting that this isoform may serve an important physiological and pharmacological function in these tissues (Kuehl et al., 2001; Koch et al., 2002).

Inactivation of P450s in phenobarbital-treated rats by various natural furanocoumarin derivatives (Letteron et al., 1986) and the mechanism-based inactivation of P450 2B1 by several furanocoumarins in umbelliferous vegetables such as bergapten, xanthotoxin (8-methoxypsoralen), and isopimpinellin has been reported previously (Cai et al., 1993; Koenigs and Trager, 1998a). Thus, it is likely that BG may also affect the catalytic activity of other members of the P450 2B subfamily. As expected, after characterizing the effect of BG on rat P450 2B1, rabbit P450 2B4, and human P450 2B6, it was found that the BG inactivation was concentration- and time-dependent in all three 2B isoforms (Table 1). A variety of widely used drugs including bupropion, efavirenz, methadone, ifosfamide, and cyclophosphamide are preferentially metabolized or stereoselectively metabolized by P450 2B6 (Faucette et al., 2000; Huang et al., 2000; Ward et al., 2003; Gerber et al., 2004). Moreover, P450 2B6 has been shown to be present not only in the liver but also in the small intestine, the kidney, and the brain (Gervot et al., 1999).

Both P450s 2B6 and 3A5 exhibit genetic polymorphisms and wide tissue distribution and play an important role in the metabolism of endobiotics and xenobiotics including clinical drugs routinely used in cancer chemotherapy, hormone therapy, inhibition of HIV protease, depression, immunosuppression, and calcium channel blockers (Kane and Lipsky, 2000; Ingelman-Sundberg, 2004). Knowledge concerning the effect of BG and related compounds on the function of P450s 2B6 and 3A5 is important for effective drug therapy and drug development.

Materials and Methods

Chemicals. Catalase, NADPH, glutathione (GSH), testosterone, l-α-dilauroyl-phosphatidylcholine, l-α-dioleyl-sn-glycero-3-phosphatidylcholine, and l-α-phosphatylserine were purchased from Sigma-Aldrich (St. Louis, MO). BG, bergapten, bergaptol, and psoralen were purchased from Indofine Chemical Co., Inc. (Hillsborough, NJ). [¹⁴C]BG (55 mCi/mmol) with 99% purity was obtained from American Radiolabeled Chemicals (St. Louis, MO). DHBG was a gift from the Florida Department of Citrus (Lakeland, FL). 7-Ethoxy-4-(trifluoromethyl)coumarin (EFC) was obtained from Invetrogen Corp. (Eugene, OR). All other chemicals and solvents were of highest purity from commercial source.

Purification of Enzymes. Plasmids for P450 2B6 and P450 3A5 were generous gifts from Dr. J. R. Halpert (University of Texas Medical Branch, Galveston, TX) and were expressed as His-tagged proteins in Escherichia coli Topp 3 cells and purified to homogeneity (Domanski et al., 2001; Scott et al., 2001). Reductase and the other P450s including P450s 2B1, 2B4, 2E1, 2C9, 2D6, and 3A4 were purified after expression in E. coli cells, and cytochrome-b₅ (b₅) was purified from liver microsomes of phenobarbital-treated Long-Evans rat as previously described (He et al., 1998; Lin et al., 2002).

Enzyme Assay and Inactivation. For the P450 2B6 primary reaction mixtures, P450 (1 nmol) was reconstituted with reductase (2 nmol), 2 mM GSH, and 50 units of catalase in 1 ml of 100 mM potassium phosphate buffer (pH 7.7). For the P450 3A5 primary reaction mixture, P450 (1 nmol) was reconstituted with 60 μg of a mixture (1:1:1) of l-α-dilauroyl-phosphatidylcholine, l-α-dioleyl-sn-glycero-3-phosphatidyl-choline and l-α-phosphatylserine, reductase (2 nmol), b₅ (1 nmol), 2 mM GSH, and 50 units of catalase in 1 ml of 50 mM HEPES buffer (pH 7.5) containing 20% glycerol, 30 mM MgCl₂, and 0.5 mM EDTA. The reconstituted P450s were incubated at 22°C for 45 min. For the concentration- and time-dependent inactivation of P450s by BG, the reactions were initiated by adding 1 mM NADPH to the primary reaction mixtures containing various concentrations of BG. The samples were incubated at 30°C for P450 2B6 and 37°C for P450 3A5. At the indicated times, 10-μl aliquots of the P450 2B6 primary reaction mixtures were transferred into 990 μl of a secondary reaction mixture containing 100 μM EFC in 100 mM potassium phosphate buffer (pH 7.4), and the EFC O-deethylation activity was determined as described previously (Lin et al., 2003). For P450 3A5, 25-μl aliquots of the primary reaction mixture were transferred into 975 μl of a secondary reaction mixture containing 200 μM testosterone in 100 mM potassium phosphate buffer (pH 7.5) to determine the testosterone 6β-hydroxylation activity (Lin et al., 2002).

Partition Ratios. BG at concentrations from 0.2 to 250 μM was added to the primary reaction mixture containing 1 μM P450 2B6 or P450 3A5. The reaction mixtures were initiated with 1 mM NADPH and incubated at 37°C for 1 h, allowing the inactivation to reach completion. Duplicate aliquots were removed and assayed for residual enzymatic activity as described above.

Spectral and HPLC Analysis. After incubating the primary reaction mixture (–b₅) with 20 μM BG in the control samples (–NADPH) or the inactivated samples (+NADPH) at 37°C for 20 min, the reduced CO difference spectra of 0.2 nmol aliquots of each of the P450s were determined by scanning from 400 to 500 nm on a UV-2501PC spectrophotometer (Shimadzu, Kyoto, Japan). To investigate if any heme loss occurred during inactivation of the P450s by BG, 0.1 nmol each of the control and the inactivated samples were analyzed by HPLC under acidic conditions to separate the heme from the apoprotein as previously described (Lin et al., 2002). The samples were analyzed on a C4 reverse-phase column (10 μm, 4.6 × 250 mm, 300 Å; Phenomenex, Torrance, CA). The solvent system consisted of solvent A (0.1% trifluoroacetic acid in H₂0) and solvent B (0.05% trifluoroacetic acid in acetonitrile). The column was eluted with a linear gradient from 40 to 85% B over 40 min with a flow rate of 1 ml/min. The eluate was monitored at 220 nm for protein, at 310 nm for BG, and at 405 nm for heme. Aliquots from each reaction mixture were also removed to determine the catalytic activity. To test if the inactivation by BG was irreversible, control and inactivated samples were dialyzed overnight at 4°C against 1 liter of 50 mM potassium phosphate buffer (pH 7.5) containing 20% glycerol and 0.1 mM EDTA, The samples were then reanalyzed for enzymatic activity, reduced CO difference spectra, and heme content by HPLC analysis.

Stoichiometry and Specificity of Binding. The primary reaction mixtures of P450 2B6 and P450 3A5 were incubated with 20 μM [¹⁴C]BG in the control and inactivated samples at 37°C for 30 min. Each of the samples (0.1 nmol) was mixed with 10 mg of bovine serum albumin and precipitated by adding a 5-fold volume of 5% sulfuric acid in methanol (Chan et al., 1993). The precipitates were collected using centrifugation, and the resulting pellets were washed at least five times with the same solvent until the radioactivity in the supernatant was essentially at background levels. The pellets were dissolved in 1 N NaOH, incubated at 60°C for 1 h, and counted using liquid scintillation counting (Econo-Safe; Research Products International Corp., Mount Prospect, IL). For SDS-PAGE analysis, P450s 2B1, 2B4, and 3A4 in the reconstituted systems were included for comparison together with P450s 2B6 and 3A5. Control and inactivated samples (25 pmol of each P450 isoform) were resolved on a 10% polyacrylamide gel and dried on 3-mm chromatography paper. The dried gels were exposed to Kodak Biomax MS film (Eastman Kodak, Rochester, NY) at –80°C before developing.

ESI-LC-MS Analysis. P450 2B6 and P450 3A5 along with P450 2B1, P450 2B4, and P450 3A4 were reconstituted as described for the spectral and HPLC analysis and incubated with 20 μM BG in the control and inactivated samples for 10 to 20 min. Control or inactivated samples (50 pmol) were analyzed on a C3 reverse phase column (Zorbax 300SB-C3, 3.5 μm, 3.0 × 150 mm; Agilent Technologies, Palo Alto, CA) equilibrated with 40% CH₃CN and 0.1% trifluoroacetic acid. After 5 min, the column effluent was directed into the mass analyzer of a LCQ mass spectrophotometer (Thermo Finnigan, San Jose, CA) as previously described (Regal et al., 2000). The CH₃CN concentration was increased linearly to 90% over the next 25 min, and mass spectra were recorded. The m/z spectrum corresponding to the protein envelopes of the P450s were deconvoluted to obtain the associated masses using the Bioworks software package (Thermo Finnigan). The ESI source conditions were: sheath gas at 90 arbitrary units, auxiliary gas at 30 arbitrary units, spray voltage at 4.2 kV, and capillary temperature at 230°C.

Metabolism of BG. P450s 2B6 and 3A5 (0.5 nmol each) were reconstituted as described above and were incubated together with 20 μM BG or 30 μM [¹⁴C]BG in the presence or absence of NADPH. After 10 min of incubation at 37°C, the reaction mixtures were extracted with 3 ml of ethyl acetate. The organic phases was dried under N₂ and dissolved in 50% acetonitrile/0.1% acetic acid for HPLC analysis. The metabolites and BG were separated using HPLC on a C8 column (Zorbax XDB, 5 μm, 2.1 × 150 mm; MAC-MOD, Chadds Ford, PA) with 0.1% acetic acid in H₂O (solvent A) and 0.1% acetic acid in acetonitrile (solvent B) using a gradient of 35% B for 15 min and then 35 to 80% B within 40 min at a flow rate of 1 ml/min. The eluate was monitored at 310 nm. For the radiolabeled samples, fractions were collected every 0.5 min and counted by liquid scintillation counting (Econo-Safe; Research Products International Corp.). Authentic standards of BG, DHBG, bergapten, bergaptol, and psoralen were separated under the same conditions that were used to monitor the products.

Results

Inactivation of P450s by BG. The BG-mediated inactivation of P450s 2B6 (Fig. 2A) and 3A5 (Fig. 2B) in the reconstituted system was time- and concentration-dependent and required NADPH. Linear regression analysis of the time course data were used to estimate the initial rate constants (K_obs) for the inactivation of the P450s by BG. From the double reciprocal plots (insets) of the values for K_obs and the concentrations of BG, the K_I, k_inact, and t_1/2 were determined. For comparison, the inactivations of P450s 2B1 and 2B4 by BG were also performed and the kinetic constants obtained for all four enzymes are shown in Table 1. The efficiency of inactivation (k_inact/K_I) in P450 2B subfamily is ∼10-fold better than P450 3A5.

Partition Ratios. P450s 2B6 and 3A5 were incubated with various concentrations of BG for 1 h allowing the inactivation to reach completion. The percentage of activity remaining was plotted as a function of the molar ratio of BG to each P450. The partition ratio was estimated from the intercept of the linear regression line obtained from lower ratios of BG to P450 with the straight line derived from the higher ratios of BG to P450 as described previously (Silverman, 1996). With this method, we estimated a partition ratio of ∼2 for P450 2B6 (Fig. 3A) and ∼20 for P450 3A5 (Fig. 3B). For 3A4, the extrapolated partition ratio was approximately 27 (data not shown). Attempts to obtain complete inactivation by further increasing the BG/P450 molar ratio, and the incubation times were not successful. The maximum achievable inactivation was approximately 80%. The inability to obtain complete inactivation has been observed with other mechanism-based inactivators and other P450s. One possible explanation is that binding of a reactive intermediate to an amino acid residue in the active site does not result in complete loss of activity but only wounds the enzyme so that it is less able to catalyze substrate metabolism. In these cases, the maximal loss of activity observed seems to be dependent on the probe substrate used (Kent et al., 2004).

Changes in the Reduced CO Difference Spectra and Heme Content. Although the inclusion of b₅ in the reconstituted system does not affect the determination of the reduced CO difference spectrum, the heme moiety from b₅ will interfere with the quantification of the heme content from the P450s when analyzed using reverse-phase HPLC. Because the same preparations were employed to measure the P450-reduced CO complex and the heme content, b₅ was omitted from the P450 reconstituted system for these studies (Lin et al., 2002). When P450 2B6 or P450 3A5 was incubated with 20 μM BG at 37°C for 20 min, the loss in enzymatic activity was accompanied by a loss of the spectrally detectable reduced CO complex in the inactivated samples compared with the control samples. Representative reduced CO difference spectra of the control and the inactivated samples are shown in Fig. 4. The maximal absorbance of the inactivated P450 2B6 shifted slightly to 452 nm.

Control and BG-inactivated samples from the same experiments were also analyzed by HPLC. The elution profile was monitored at 220 nm for protein and at 405 nm for intact heme. Figure 5A shows the separation of the components in the reconstituted system with catalase eluting at 19 min, reductase at 27 min, and BG-inactivated P450 3A5 at 42 min. Essentially all of the proteins could be recovered from the column after the inactivation by BG (Fig. 5B). Similar results were obtained for BG-inactivated P450 2B6 (data not shown). However, when the elution profiles of the P450 2B6 and P450 3A5 reaction mixtures were monitored at 405 nm, a significant loss in the intact heme was seen. No detectable adducted heme or cross-linked heme absorbing at 405 nm could be observed in either the P450 2B6 or the P450 3A5 samples inactivated by BG, indicating that BG inactivation may have resulted in heme fragmentation or may have generated a very unstable adducted heme. Representative HPLC elution profiles monitored at 405 nm for the heme content of P450 3A5 and P450 2B6 in the control and inactivated samples are displayed in Fig. 5, C and D, respectively. The loss in enzymatic activity observed upon inactivation by BG was associated with a loss in the intact heme. Similar results were also observed when P450s 2B1, 2B4, and 3A4 were incubated with BG in the presence of NADPH. However, less than 50% of the P450 3A4 apoprotein was recovered after BG inactivation. Similar losses in apoprotein have previously been described for BG-inactivated P450 3A4 inactivated by BG, mifepristone and 17α-ethynylestradiol (He et al., 1998, 1999; Lin et al., 2002). Table 2 summarizes the effects of BG on the enzymatic activity, the reduced CO difference spectrum, and the intact heme remaining for all the P450s tested. When the primary reaction mixture was incubated with BG and NADPH for shorter times to remain within the linear range of inactivation so as to minimize heme destruction due to secondary oxidation events (5 min for 2B6 and 10 min for 3A5), the catalytic activity, P450-CO complex, and heme content remaining after BG inactivation were: 42, 69, and 72%, respectively, for P450 2B6; and 51, 57, and 68%, respectively, for P450 3A5.

The removal of free BG by extensive dialysis from the control and inactivated samples did not lead to a significant recovery of the relative catalytic activity, the reduced CO difference spectra, or the heme content of P450 2B6 and P450 3A5 compared with the values obtained for the same samples before dialysis (data not shown). Thus, the BG inactivation of P450 2B6 and P450 3A5 is irreversible.

Stoichiometry and Specificity of Binding. The amount of BG that was covalently bound to the apoprotein was determined by precipitating the proteins from the control and the BG-inactivated samples that had been incubated with [¹⁴C]BG. The stoichiometry of binding of a reactive BG intermediate to the apoprotein was determined to be ∼0.5:1 for both P450 2B6 and P450 3A5, suggesting that approximately 0.5 mol of a reactive BG intermediate bound per mole of P450 inactivated. Lower than 1:1 stoichiometries for mechanism-based inactivation of P450s has been reported for N-benzyl-1-aminobenzotriazole inactivation of P450 2B1, for 8-methoxypsoralen inactivation of P450s 2B1 and 2A6, and for L-754,394 inactivation of P450 3A4 (Kent et al., 1997; Koenigs and Trager, 1998a,b; Lightning et al., 2000). The binding stoichiometry of 0.5:1 determined here could at least in part be due to the contribution of the heme modification during the inactivation of P450s 2B6 and 3A5 by BG.

In addition, the control and inactivated samples that had been incubated with [¹⁴C]BG were analyzed using SDS-PAGE. Following electrophoresis, the gel was dried and analyzed by autoradiography. P450s 3A4, 2B1, and 2B4 were also included for comparison. The data in Fig. 6 clearly show that BG-derived radioactivity was associated only with the P450 apoprotein of all the P450s in the NADPH-treated samples but not in the control samples exposed to labeled BG. Binding of a small amount of [¹⁴C]BG to reductase was also seen and may indicate that some of the reactive intermediate escaped from the P450 active site. Together, these studies on the radioactivity associated with the apoprotein as determined by SDS-PAGE and the stoichiometry of binding indicate that a reactive BG intermediate is covalently bound to the P450 apoproteins.

ESI-LC-MS Analysis. The P450 reconstituted systems were incubated with BG in the absence or the presence of NADPH and analyzed using ESI-LC-MS. A total ion chromatogram for P450 3A5 is displayed in Fig. 7A. As shown in Fig. 7B, the mass of the inactivated P450 3A5 increased by 387 Da from 56836 ± 3 to 57223 ± 5. This increase in mass would be consistent with the mass of BG (338 Da) plus three oxygen atoms. Similar results were observed with P450 2B1, where the mass of the apoprotein from samples incubated with BG in the absence of NADPH (55,883 ± 1 Da) increased by 390 Da in samples incubated in the presence of NADPH (56,275 ± 3) (mass spectra not shown). An increase in the mass of the P450 2B4 apoprotein by 380 Da from 55,724 ± 4 to 56,104 ± 5 was also seen after inactivation of the P450 by BG in the presence of NADPH (mass spectra not shown). Similarly, BG-inactivated P450 2B6 samples increased in mass by approximately 388 Da from 54,521 ± 20 to 54,909 ± 39 (Fig. 7C). P450 2B6, particularly the inactivated sample, was the most difficult to analyze and showed the greatest S.D. in the mass assignment. Because of the large error in the mass assignment of the BG-modified P450 2B6, it is not clear if the mass of the protein increased by BG plus two or three oxygen atoms. An attempt to elucidate the mass of the BG-modified P450 3A4 apoprotein under the same conditions was not successful. The amount of P450 3A4 apoprotein in the total ion chromatogram determined after BG inactivation was less than half of that of the control sample, suggesting that either the inactivated P450 3A4 molecule could not be eluted from the reverse phase column or that the protein failed to ionize or vaporize for mass analysis.

Metabolism of BG. M3 and M4 are two predominant metabolites generated from the metabolism of BG by P450 2B6 (Fig. 8A). In addition to M3 and M4, M1, M2, and M5 were generated from P450 3A5 (Fig. 8B). No major metabolites were observed in the control samples of P450 3A5 or P450 2B6 (data not shown). Authentic standards of bergaptol, DHBG, and BG were analyzed under the same conditions (Fig. 8C). M1 exhibited the same retention time as bergaptol, and M2 eluted with the same retention time as DHBG. Psoralen and bergapten were eluted at 13 and 18 min, respectively. The minor metabolites do not correspond to these two standards. When the fractions from the radiolabeled samples were collected and analyzed by liquid scintillation counting, all the metabolites except M1 had radioactivity associated with them (data not shown). The [¹⁴C]BG used in these experiments was labeled at the geranyl-1-C. Therefore, the lack of radioactivity associated with M1 indicates that the geranyl group is removed from BG during metabolism to generate M1. This result further strengthens the conclusion that M1 is indeed bergaptol since it would be formed by loss of the geranyl group. The fact that M1 is in extremely low level and M2 is undetectable in P450 2B6 suggests that the orientation of BG in the active site of P450 3A5 is different from that in P450 2B6 and that P450 3A5 may preferentially remove the geranyl group of BG and hydroxylate this side chain at the 6 and 7 position. In addition, three radiolabeled peaks from the metabolites formed by P450 3A5 could not be detected by monitoring the absorbance at 310 nm (data not shown). This suggests that these products may have been formed as a result of the cleavage of the geranyloxy side chain or ring opening of the furanocoumarin moiety. This result is consistent with the 8-methoxypsoralen metabolites identified in rat, dog, and human livers: O-dealkylation, oxidation and hydrolysis of furan ring, and hydrolysis of the lactone ring (Mays et al., 1986). The metabolite profile and total product formation observed with P450 3A4 was essentially identical to that of P4503A5, suggesting that P450 3A5 can metabolize BG as efficiently as P450 3A4 (data not shown). Other human P450 isoforms such as P450 2C9, 2E1, and 2D6 were also examined for the BG-dependent metabolism and inactivation. The results from the incubations of BG with these P450s in the reconstituted system were: no major metabolites and no inactivation were observed for P450 2E1; M4 and some minor metabolites, but not M1 and M2, were generated by P450s 2C9 and 2D6; and BG inactivated P450 2D6, but not P450 2C9 (data not shown). It seems that the ability to catalyze the elimination of the geranyl side chain and to form DHBG is a common feature for members of P450 3A family, but not for the P450 2 family.

Discussion

The inactivation of two human P450s by BG has been characterized in detail. The major findings are as follows: the inactivation was concentration- and time-dependent and required the metabolic activation of BG in the presence of NADPH; the inactivation was irreversible for both P450s, with apparent K_I values of 5 μM (–b₅) for P450 2B6 and 20 μM for P450 3A5 (+b₅); the inclusion of b₅ in the P450 2B6 primary reaction mixture resulted in a dramatic decrease in the K_I to 0.2 μM, whereas the k_inact and the t_1/2 were not altered (data not shown); the partition ratio was ∼2 and ∼20 for P450s 2B6 and 3A5, respectively; a stoichiometry of binding of ∼0.5:1 was obtained for the inactivation of both enzymes; the loss in catalytic activity was associated with a decrease in the spectrally detectable P450-CO complex; the mechanism of inactivation involved both the destruction of native heme and the covalent binding of a reactive BG intermediate to the apoprotein; and five major metabolites of BG were generated by P450 2B6, with four additional metabolites seen with P450 3A5.

Our results clearly demonstrate that BG is an effective mechanism-based inactivator of P450s 2B6 and 3A5. The very low K_I and partition ratio observed for P450 2B6 indicates that BG is much more efficient inactivator of P450 2B6 than the P450 3A family, which is considered to be the major P450 contributing to the grapefruit juice effect (He et al., 1998; Ohnishi et al., 2000; Tassaneeyakul et al., 2000). Ingestion of grapefruit juice leads to a rapid intracellular degradation of P450 3A4 following a suicide irreversible inactivation. This generally results in increased drug bioavailability (Lown et al., 1997; Ohnishi et al., 2000; Kane and Lipsky, 2000). BG is the parent compound of DHBG and epoxybergamottin, as well as other dimeric derivatives of BG (Schmiedlin-Ren et al., 1997; He et al., 1998). BG is more stable and more lipophilic than the related furanocoumarins. BG is abundant in grapefruit extract and segments and is easily absorbed within the intestinal wall (Bailey et al., 2000; Kane and Lipsky, 2000). Although P450 2B6 is not as abundant as P450 3A in liver and intestine, the high efficiency for inactivation by BG indicates that there may be important clinical ramifications for drugs that are primarily metabolized by 2B6. Moreover, the potencies of BG, DHBG, bergapten, and 8-methoxypsoralen for the inactivation of P450 2B6 are very similar (data not shown).

The inactivations by BG were associated with decreases in the P450 reduced CO spectra for P450 2B6 and 3A5 (Fig. 4). The intensities of the protein peaks monitored at 220 nm between the control and inactivated samples were similar, whereas the heme contents of the inactivated samples were markedly decreased (Fig. 5). However, no detectable heme-BG adducts or BG adducts to heme fragments could be observed in the HPLC eluates monitored at 405 nm for heme content or at 310 nm for the furanocoumarin moiety (chromatograms not shown). Direct evidence for the covalent binding of a reactive metabolite of BG to the P450 apoprotein was provided by two approaches: [¹⁴C]BG was associated with the P450 apoprotein in inactivated samples analyzed by SDS-PAGE (Fig. 6), and the masses of the inactivated apoproteins increased by ∼388 Da when analyzed by ESI-LC-MS (Fig. 7).

The mechanism-based inactivation of P450s by other furanocoumarins such as 8-methoxypsoralen, bergapten, bergaptol, and psoralen has been well studied in other systems (Mays et al., 1989; Koenigs and Trager 1998a,b). In general, these studies have demonstrated that: the oxidation of a double bond in the furan ring is responsible for inactivation, the irreversible inactivation of P450 catalytic activity is due to a mono-oxygenated reactive metabolite, GSH is conjugated to a mono-oxygenated metabolite of furanocoumarins, and a dihydrodiol of the furanocoumarin is formed during metabolism. The chemical structure of BG is similar to that of 5-methoxypsoralen except that the methoxy side chain is replaced by a geranyloxy side chain in BG. Therefore, it is likely that the pathway for the metabolic activation of BG may be similar to that for 5-methoxypsoralen and related furanocoumarins. The increase in the mass of the apoprotein of ∼388 Da observed following inactivation of P450s 3A5 could result from the formation of a covalent adduct with one molecule of BG (338 Da) plus three oxygen/hydroxyl groups. Based on metabolism studies of furanocoumarins in other systems, it is possible that BG may have been hydroxylated twice either on the geranyloxy side chain alone (e.g., to give DHBG) or in combination with hydroxylation of the phenyl ring of the furanocoumarin moiety. The subsequent oxidation of the furan ring would give either a furanoepoxide or γ-ketoenal intermediate as the reactive species responsible for the inactivation. Experimental evidence for the involvement of such a mechanism is: oxidation of the furan ring double bonds to furanoepoxide or γ-ketoenal intermediates resulting in P450 modification has been well documented by LC-MS/MS studies on the inactivation of P450s 2B1 and 2A6 by furanocoumarins and P450 3A4 by furanopyridine (Koenigs and Trager 1998a,b; Lightning et al., 2000). Preliminary LC-MS/MS studies in our laboratory (data not shown) have revealed that BG-GSH adducts isolated from reaction mixtures have a major molecular ion with an m/z of 661 Da that would be equivalent to the mass of GSH (307 Da) plus BG (338 Da) and one oxygen (16 Da). The predominant fragment ion with m/z of 526 Da corresponded to the furanocoumarin moiety (203 Da) plus mono-oxygenated GSH (323 Da), indicating that the furanocoumarin moiety of BG was bioactivated to a reactive intermediate and trapped by GSH. A previous study reported the metabolism of BG by P450 3A4 to give hydroxylated metabolites (He et al., 1998). Similarly, preliminary studies in our laboratory (data not shown) indicate the formation of mono-, di-, and trihydroxylated metabolites of BG by P450s 2B6 and 3A5. Studies on the inactivation of rat cytochromes P450 have also shown hydroxylation of the phenyl ring of the furancoumarin moiety (Mays et al., 1986). Further studies aimed at the identification of the structures of the BG metabolites and the elucidation of the structure of the reactive intermediate(s) trapped by reaction with GSH using ESI-LC-MS/MS analysis are in progress (U. M. Kent, H. Lin, and P. F. Hollenberg, unpublished data).

The formation of mono-oxygenated GSH conjugates and trioxygenated BG-modified P450 3A5 may imply: the furanoepoxide or γ-ketoenal intermediate of the dihydroxylated BG is very reactive and reacts readily with a target residue in the active site of the P450 and avoids trapping by GSH, and the furanoepoxide or γ-ketoenal intermediate of BG is less reactive and can diffuse out from the site of formation to be trapped by GSH. A similar phenomenon has previously been described in studies of 8-methoxypsoralen in mouse liver microsomes. Cysteine reacted with 8-methoxypsoralen following demethylation, whereas P450 protein reacted with 8-methoxypsoralen that retained the methyl group (Mays et al., 1990). Regardless of whether or not BG is hydroxylated prior to the formation of reactive species, the site for bioactivation to form the reactive intermediate is the double bond of the intact furan ring. When the 3A5 reconstituted system was incubated with DHBG (372 Da), the mass increase in the inactivated P450 was ∼392 Da from the ESI-LC-MS analysis, indicating only one oxygen was incorporated to this dihydroxylated BG (data not shown).

The HPLC elution profiles of the BG metabolites showed the following: the generation of bergaptol and DHBG from P450 3A5, but not from P450 2B6, is in agreement with the hypothesis that the P450 3A5 has a larger active site to accommodate a number of different orientations of the BG molecule, whereas the furanocoumarin moiety of BG is favored to face the active site of P450 2B6; and M1, bergaptol, is not responsible for the inactivation because the mass increase of the adducted P450s greatly exceeds the mass of M1 (202 Da). Our results for the metabolism of BG, in conjunction with previous studies on the metabolism of furanocoumarins (Mays et al., 1989; Koenigs and Trager 1998a,b) suggest that BG may be metabolized by a variety of different pathways as illustrated in Fig. 9.

In conclusion, we have demonstrated that BG is a very efficient mechanism-based inactivator of two human P450s involved in the metabolism of numerous xenobiotics and that the BG-dependent inactivation of P450s 2B6 and 3A5 occurs by two mechanisms: heme destruction and covalent binding of a reactive intermediate of BG to the apoprotein. Due to the significant polymorphic expression in diverse tissues, interracial differences, and interindividual variability, the contribution of P450s 2B6 and 3A5 to the metabolism of clinically used drugs may be much higher in some individuals than previously thought (Kuehl et al., 2001; Ward et al., 2003). The consumption of grapefruit juice, the ingestion of certain vegetables, or the use of herbal medicines in conjunction with the use of drugs that are primarily metabolized by these two enzymes may seriously impact on the biotransformation and/or clearance of these drugs. Thus, warnings currently in place for possible drug interactions following consumption of grapefruit juice may not be limited to substrates for P450 3A4 alone and should also be evaluated for drugs metabolized primarily by P450s 2B6 and 3A5. In particular, the potential for drug-drug interactions with the antidepressant bupropion and the HIV protease inhibitor efavirenz, which are both preferentially metabolized by P450 2B6, needs to be considered.