Time-Dependent Inactivation of P450 3A4 by Raloxifene: Identification

954

Chem. Res. Toxicol. 2007, 20, 954-964

Time-Dependent Inactivation of P450 3A4 by Raloxifene: Identification of Cys239 as the Site of Apoprotein Alkylation1 Brian R. Baer, Larry C. Wienkers, and Dan A. Rock* Amgen, Department of Pharmacokinetics and Drug Metabolism, 1201 Amgen Court West, Seattle, Washington 98119 ReceiVed January 31, 2007

Time-dependent inactivation of cytochrome P450s is typically a result of substrate bioactivation to form reactive species that subsequently alkylate the heme group, apoprotein, or both. The chemical identity of many reactive intermediates is generally proposed based on the products of trapping reactions with nucleophilic agents as only a few P450-drug adducts have been directly characterized. We describe the use of mass spectrometry to show that a single equivalent of raloxifene is bound to the intact P450 apoprotein. Furthermore, mass analysis of peptides following digestion with proteinase K revealed that the covalently bound drug is localized to residue Cys239. A mass shift of 471 Da to the intact protein and peptide, relative to control samples, indicated that time-dependent inactivation of P450 3A4 occurred through the raloxifene diquinone methide intermediately prior to nucleophilic attack of the sulfur of Cys239. Association between raloxifene adduction to P450 3A4 apoprotein and the observed timedependent inactivation was further investigated with the use of cysteine-specific modifying reagents. When P450 3A4 was treated with iodoacetamide or N-(1-pyrene)iodoacetamide, which alkylated residue Cys239 exclusively, time-dependent inactivation of P450 3A4 by raloxifene was prevented. The change in protein mass of 471 Da combined with the protection from inactivation that occurred through prealkylation of Cys239 provided conclusive evidence that raloxifene-mediated P450 3A4 inactivation occurred through the bioactivation of raloxifene to the diquinone methide and subsequent alkylation of Cys239. Introduction In vitro metabolism studies are routinely used to predict the in vivo biotransformation pathways of new chemical entities. Of particular interest are the specific oxidative pathways that form reactive metabolites of which certain substructures are more prone to generate “bioactivated” intermediates that can irreversibly inhibit the oxidizing P450 enzyme (1). The basal enzyme activity of the affected P450 can only recover as fast as the natural synthesis rate of the P450 enzyme, and as a result the susceptibility to drug interactions increases in instances where clearance is mediated by the inactivated P450. To this end, identification and elimination of these events is an important step in the development of safe pharmaceutical drugs. The most common early indicator of bioactivation is with the indirect observation of reactive metabolites trapped with exogenous nucleophiles like glutathione (GSH) (2). These methods identify the presence of a bioactivation step that generates an unstable electrophilic intermediate and typically triggers time-dependent inhibition experiments to look for P450 inactivation. Timedependent inhibition experiments are designed to investigate the degree of P450 inactivation through the generation of kinact and an apparent KI, which are used to further context the potential in vivo significance (3). Best practice is to eliminate the structural motif in the compound that is responsible for * To whom correspondence should be addressed. Phone: (206) 2657139. Fax: (206) 265-1149. E-mail: [email protected]. 1 Abbreviations: BQ, 7-benzyloxyquinoline; P450, cytochrome P450; DLPC, L-R-dilauroylphosphatidylcholine; DLPS, L-a-dilauroyl-sn-glycero3-phosphoserine; DOPC, L-a-diloleoyl-sn-glycero-3-phosphocholine; IA, iodoacetamide; LC-MS, liquid chromatography mass spectrometry; SRM, selected reaction monitoring; MS, mass spectrometry; PIA, N-(1-pyrene)iodoacetamide; TFA, trifluoroacetic acid.

inactivation. However, trapping agents such as GSH may not reflect the species responsible for enzyme inactivation, complicating identification of the culprit structural motif. Measuring P450-adduct complexes with mass spectrometry provides a method whereby the enzyme inactivating species can be measured directly. Foremost, the basic characteristics of the drug-protein adducts can be deduced from the mass change associated with the adducted protein. For example, a protein adduct may reflect a change in mass consistent with the parent drug plus 16, reflecting an oxidized metabolic species as the precursor to protein adduct formation. The level of detail regarding the protein-drug adduct can be increased through enzymatic or chemical digestions which enable MSn fragmentation data to be generated on the adducted peptide(s). The latter step has proven to be exceptionally challenging. For example, an intact protein adduct of tienilic acid and P450 2C9 was generated from which a mechanism for the thiophene based adduct was proposed; however, the exact location of adduction was unattainable after tryptic digestion (4). A furan-containing structure, L-754,394, was postulated to bind to Glu307 of the apoprotein based on evidence from Tricine SDS-PAGE and MALDI-TOF-MS using radiolabeled material, but the intact adduct was not observed (5). While the culprit substructures may be intuitive in the aforementioned examples, these procedures do not provide detailed information necessary for less obvious bioactivation mechanisms and would fail to yield hypotheses for structural modifications needed to eliminate reactive chemistry from a chemical scaffold. Raloxifene represents a well-characterized mechanism-based inhibitor of P450 3A4 (6). In vitro, P450 3A4 mediated the metabolism of raloxifene to produce several electrophilic species

10.1021/tx700037e CCC: $37.00 © 2007 American Chemical Society Published on Web 05/12/2007

Raloxifene Cys Adduct of P450 3A4 Apoprotein

Chem. Res. Toxicol., Vol. 20, No. 6, 2007 955

Scheme 1. P450 3A4 Bioactivation of Raloxifene

including a diquinone methide intermediate that reacted with GSH at various locations on the molecule (6, 7) (Scheme 1). Presumably, one of these species is responsible for P450 3A4 inactivation. Addition of GSH to the P450 3A4 incubation does not fully protect the enzyme from inactivation (6). The raloxifene intermediates were synthesized and further characterized to have distinctly different half-lives. These differences may provide a rational basis for distinguishing the protein-based inactivating species from the remainder of the GSH adducts (7). In total, four GSH adducts have been characterized. In this report we identified a single P450 3A4-mediated alkylation product of raloxifene to the apoprotein of P450 3A4 via whole protein mass spectrometry. Following proteinase K digestion, MSn analysis of the alkylated peptide revealed the site of adduct formation. The adducted site was confirmed to be responsible for the time-dependent inhibition of P450 3A4 by raloxifene, and the location was confirmed in an additional experiment using iodoacetamide to block the electrophilic residue. The results are discussed with a focus on direct measurement of reactive metabolites, the mechanism of inactivation by irreversible inhibitors, and use of protein-adduct characterization of P450 mechanism-based inhibitors.

Experimental Procedures Materials. Iodoacetamide, reduced glutathione, Chaps, potassium Hepes, MgCl2, guanidine hydrochloride, raloxifene, CaCl2, ZnSO4, and NADPH were purchased from Sigma-Aldrich (St. Louis, MO). 7-Benzyloxyquinoline (BQ), 7-hydroxyquinoline metabolite standard, and pooled HLMs were purchased from BDGentest (Woburn, MA). N-(1-Pyrene)iodoacetamide (PIA) was purchased from Invitrogen (Carlsbad, CA). L-R-Dilauroyl-sn-glycero-3-phosphocholine (DLPC), l- R-diloleoyl-sn-glycero-3-phosphocholine (DOPC), and l-R-dilauroyl-sn-glycero-3-phosphoserine (DLPS) were purchased from Avanti Polar Lipids Inc. (Alabaster, AL). Sequencinggrade trypsin and proteinase K were purchased from Roche (Indianapolis, IN). A Poros R2 perfusion column (2.1 × 100 mm) was from Applied Biosystems (Cambridge, MA), a C18 column (2.1 × 250 mm) was from Grace Vydac (Hesperia, CA), and a Zorbax RX-C8 column (2.1 × 150 mm) was from Agilent (Palo Alto, CA). P450 Enzymes and Co-enzymes. Recombinant P450 3A4 was expressed in Escherichia coli DH5R in the pCW 3A4-His6 expression vector, kindly provided by Dr. Ron Estabrook. In a 2.8 L Fernbach flask, cells were shaken at 180 rpm at 27 °C for 48 h. Media and remaining expression conditions are as described by

Gillam et al. (8). Pelleted cells were resuspended in buffer containing 50 mM potassium phosphate (pH 7.4), 500 mM NaCl, 20% glycerol, 50 µM testosterone (stabilizing ligand), 20 mM β-mercaptoethanol, 1% Emulgen 911, and Sigma protease inhibitor cocktail (0.5 mL/L of initial culture volume). Cells were lysed using a French Press operated at 10 000 psi and then spun at 150 000g. The supernatant was loaded directly onto NTA-Ni resin, equilibrated with 50 mM potassium phosphate (pH 7.4), 500 mM NaCl, 20% glycerol, 50 µM testosterone, and 0.2% Emulgen 911. The column was washed with 200 mL of wash buffer containing 50 mM potassium phosphate (pH 7.4), 20% glycerol, 20 mM imidazole, 0.2% cholate, and 50 µM testosterone. P450 3A4 was eluted from the column with buffer containing 50 mM potassium phosphate (pH 7.4), 20% glycerol, 500 mM imidazole, and 0.2% cholate. The eluted protein was dialyzed against 100 mM potassium phosphate, pH 7.4, in 20% glycerol and stored at -80 °C. Rat cytochrome b5 and P450 reductase were expressed and purified as described previously (9, 10). Instrumentation. A LTQ mass spectrometer from Thermo (San Jose, CA) was connected online with an Agilent 1100 series HPLC (Palo Alto, CA) including a degasser, pumps, autoinjector, column oven, and diode array detector. A Cary 6000 UV/vis spectrometer from Varian, Inc. (Palo Alto, CA) was used for measuring ligand binding and ferrous CO-bound P450 spectra. A Safire2 fluorometer from Tecan (San Jose, CA) was used to measure turnover of BQ to 7-hydroxyquinoline. P450 3A4 Reaction with Iodoacetamide or Pyrenyliodoacetamide. Pure recombinant P450 3A4 (500 pmol) was combined with 20 µg/mL liposomes [DLPC, DOPC, DLPS (1:1:1, w/w/w per mL)] and 0.1 mg/mL Chaps in 100 µL of 50 mM potassium phosphate (pH 7.4) buffer. IA (500 µM) or PIA (50 µM) was allowed to react with P450 3A4 for 1 h at room temperature. Initial experiments demonstrated P450 3A4 was not sufficiently alkylated by treatment with 50 µM of IA, a concentration that proved effective with PIA treatment, and therefore, 500 µM IA was used in subsequent experiments. The reaction was quenched with 5 mM reduced glutathione. The extent of cysteine alkylation in P450 3A4-IA or P450 3A4-PIA was determined by the relative increase in the deconvoluted mass of the P450 3A4 protein. Incubations with Reconstituted P450 3A4 and Raloxifene. P450 3A4 (100 pmol) was combined with NADPH-P450 reductase (200 pmol), cytochrome b5 (100 pmol), 0.1 mg/mL Chaps, 20 µg/ mL liposomes [DLPC, DOPC, DLPS (1:1:1, w/w/w per mL)], 3 mM reduced glutathione, 50 mM potassium Hepes, pH 7.4, and 30 mM MgCl2 in a total volume of 0.5 mL. Raloxifene was added to a final concentration of 50 µM. The reconstituted P450 3A4 was incubated for 2 min at 37 °C prior to addition of 1 mM

956 Chem. Res. Toxicol., Vol. 20, No. 6, 2007 NADPH. The reaction was allowed to proceed for 20 min unless otherwise noted. Time-Dependent Inactivation of P450 3A4. Aliquots of the primary reaction of P450 3A4 and raloxifene were removed at 1, 5, 10, and 20 min and placed on ice. Time-dependent inactivation of P450 3A4 was determined by quantitating the remaining activity toward BQ. The primary reaction was diluted 1:10 into the secondary reaction containing 0.1 mg/mL Chaps, 20 µg/mL liposomes [DLPC, DOPC, DLPS (1:1:1, w/w/w per mL)], 3 mM reduced glutathione, 50 mM potassium Hepes, pH 7.4, and 30 mM MgCl2, 1 mM NADPH, and 80 µM BQ. The final volume of the secondary reaction was 200 µL and incubated in a 96-well plate (black with clear bottom). Production of 7-hydroxyquinoline was measured every minute for 20 min at 37 °C by fluorescence (excitation, 409 nm; emission, 530 nm). The 96-well plate was shaken for 10 s prior to each measurement. P450 3A4 activity was determined from a 10-min linear portion of the curve. Measurements were made in triplicate to allow for calculation of standard deviations. In control reactions, either raloxifene or NADPH was omitted in the primary incubation. Mass Analysis of Intact Protein. P450 3A4 protein (100 pmol) was separated from alkylation reagents (IA and PIA modified P450 3A4) or incubation components (P450 3A4-raloxifene adduct) on a Poros R2 perfusion chromatography column (2.1 × 100 mm) to facilitate electrospray ionization. Protein in incubations with raloxifene was concentrated by precipitating with 1:10 dilution with 15% ZnSO4, centrifuging at 13 000 rpm, and then resuspending with 100 µL of 6 M guanidine hydrochloride. Initial HPLC conditions were 58% solvent A (0.05% TFA in H2O) and 42% solvent B (0.05% TFA in acetonitrile) at a flow rate of 1 mL/min. The following gradient elution profile was utilized: 42% B for 1 min, 42-50% solvent B in 5 min, 50-65% solvent B in 1 min, 65% solvent B for 5 min. The flow rate was reduced to 300 µL/ min from 6 to 10 min during data acquisition with the interfaced Finnigan LTQ mass spectrometer. The column was then washed by increasing solvent B from 65% to 85% in 1 min and holding at 85% for 1 min at a flow rate of 1 mL/min. P450 3A4 eluted at 8 min, and mass spectra were averaged over the entire peak width. The protein mass was deconvoluted using MagTran (version 1.02, Amgen). Each P450 3A4 protein sample was run in triplicate to determine the standard deviation within data acquisition and deconvolution calculations. Proteinase K and Trypsin Digestion of P450 3A4. Prior to enzymatic digestion of P450 3A4, approximately 500 pmol of protein was purified on the Poros R2 column as described above. The peak corresponding to P450 3A4 in the 280 nm trace was collected in a 1.5 mL tube and evaporated to 50 µL. Buffer containing 50 mM Hepes, pH 7.4, and 10 mM CaCl2 was added to each sample for a total of 200 µL. Samples prepared for proteinase K digestion also contained 3 M guanidine hydrochloride to help denature the protein and increase proteolysis. Following addition of 5% w/w of proteinase K or trypsin, reactions were allowed to proceed for 2 h at 37 °C. Reactions were stopped by placing on ice until mass analysis. UV-vis and Mass Analysis of Alkylated Peptides. The peptides resulting from digestion of P450 3A4 were injected onto a 2.1 × 250 mm C18 column (Grace Vydac) at a flow rate of 200 µL/min. Initial conditions were 95% solvent A (0.05% TFA in H2O) and 5% solvent B (0.05% TFA in acetonitrile), and the following gradient was used to elute the peptides: 5% B for 2 min, 5-95% B in 28 min, and 95% B for 5 min. The 340 nm trace, acquired with an Agilent 1100 DAD detector, was used to locate the retention time of peptides alkylated with PIA or raloxifene. The interfaced Finnagin LTQ was used to first identify the modified peptides based on parent mass and subsequently for MS2 experiments to confirm the peptide sequence and the identity of the modified residue. Specific wavelengths (340 nm) and mass fragments were used to confirm alignment of retention times between the UV and the MS. The MS was performed using a SRM scan with the default activation time and collision energy of 35. Protein Prospector (http:// prospector.ucsf.edu) was used to identify the sequences of the

Baer et al. modified peptides based on the experimentally determined molecular masses. Measurement of Raloxifene Metabolites. Following a 20 min incubation of P450 3A4, P450 3A4-PIA, or P450 3A4-IA with raloxifene (described above) the reaction was quenched with 0.5 mL of ice-cold acetonitrile containing 1 µg/mL tolbutamide as an internal standard. Control reactions were conducted simultaneously in which (i) GSH was reacted with IA or PIA for 1 h prior to incubation with P450 3A4 and (ii) NADPH was omitted from the reaction. In addition, pooled HLMs (0.4 mg) were incubated for 20 min at 37 °C for comparison of metabolic profiles. The samples were spun at 13 000 rpm at 4 °C for 30 min to precipitate and pellet protein. The supernatant was evaporated to 200 µL, and 20 µL was injected onto an Agilent Zorbax RX-C8 column (2.1 × 150 mm) at a flow rate of 250 µL/min. Initial HPLC conditions were 90% solvent A (0.05% TFA in H2O) and 10% solvent B (0.05% TFA in acetonitrile). The elution gradient increased from 10% to 50% B in 30 min and 50% to 95% B in 2 min and held at 95% B for an additional 3 min. Two chromatographically distinct glutathionyl-raloxifene metabolites (m/z 779) eluted at 17.7 and 18.1 min, hydroxyl-raloxifene (m/z 490) eluted at 22.1 min, and the internal standard tolbutamide (m/z 271) eluted at 27.4 min. The relative production of raloxifene-derived metabolites was quantitated from the ratio of the m/z 779 or 490 trace peak areas to the m/z 271 trace peak areas representing the internal standard. Measurements were conducted in triplicate to determine standard deviations. Measurement of Type I Ligand Binding. A solution of 0.5 µM P450 3A4, P450 3A4-IA, or P450 3A4-PIA and 20 µg/mL of liposomes [DLPC, DOPC, DLPS (1:1:1, w/w/w per mL)] in 0.5 mL of 50 mM potassium phosphate (pH 7.4) buffer was equilibrated at 25 °C in a 0.7 mL cuvette. A baseline spectrum was acquired from 500 to 350 nm at a scan rate of 300 nm/min. Midazolam was titrated in to final concentrations ranging from 0 to 35 µM from 0.5 and 3 mM stock solutions in methanol, and spectra were acquired following each addition. Cyclosporin A was titrated into the P450 3A4 solution to final concentrations ranging from 0 to 16 µM from 0.5 and 3 mM stock solutions in methanol. The final organic solvent concentration did not exceed 2% of the total volume. The ∆Abs(390-420 nm) was plotted versus ligand concentration, and the curve was fit assuming one-site saturation in SigmaPlot (version 9.0, Systat Software, Inc.) to determine the KS for ligand binding to P450 3A4. P450 3A4 Ferrous CO Binding Spectra. Solutions of P450 3A4, P450 3A4-IA, or P450 3A4-PIA in 50 mM potassium phosphate (pH 7.4) were reduced with a few grains of sodium dithionite, and baseline spectra were acquired from 500 to 400 nm at a scan rate of 300 nm/min. Samples were then bubbled with CO for approximately 30 s, and spectra were acquired.

Results Characterization of a Raloxifene Adduct on P450 3A4 Apoprotein. In comparison with the mass spectrum of P450 3A4 alone, the apoprotein from the reaction with raloxifene showed an additional ion envelope, which contributed 30-40% of the total ion intensity (Figure 1A and 1B). The deconvoluted spectra of P450 3A4 revealed a single mass of 57 266.4 ( 0.7 Da, whereas the apoprotein from the raloxifene reaction had a second deconvoluted mass of 57 738.7 ( 1.6 Da. The mass shift of 472 Da indicated that one equivalent of the parent raloxifene substrate was bound to the P450 3A4 apoprotein. Alkylation of P450 3A4 by PIA and IA. Previous studies have shown that raloxifene can be bioactivated by P450 3A4 to a reactive arene oxide or diquinone methide, which can each react with the cysteine residue of glutathione. Therefore, this ‘soft’ nucleophilic residue was the suspect residue for the drugprotein adduct observed with P450 3A4 apoprotein. P450 3A4 was alkylated with 100 equiv of PIA in 50 mM potassium



Figure 1. LC-MS spectra and deconvoluted spectra for P450 3A4 apoprotein. Spectra are shown for unmodified P450 3A4 (A), P450 3A4 following the reconstitution and incubation with raloxifene (B), P450 3A4 reacted with PIA (C), and P450 3A4 reacted with IA (D). (Inset) Expanded spectrum between m/z 900 and m/z 1000. Bold print represents the masses and charge states of the adducted protein.

phosphate (pH 7.4) buffer containing liposomes, allowing for the potential alkylation of all seven cysteine residues. The extent of alkylation was monitored by LC-MS, and it was determined

that the majority of apoprotein contained a single adduct after incubation for 1 h at room temperature. This result was demonstrated by the shift in the mass-to-charge ratios in the

958 Chem. Res. Toxicol., Vol. 20, No. 6, 2007

Baer et al.

Figure 2. Chromatographic separation of peptides following proteinase K digestion of raloxifene-adducted P450 3A4: (A) UV 214 nm trace, (B) UV 340 nm trace, and (C) SRM m/z 660-1092 trace.

ion envelope and the larger molecular mass of 57 524.4 ( 0.3 Da in the deconvoluted spectrum relative to the mass of unmodified P450 3A4 (Figure 1C). The mass shift of 258.0 Da is near the predicted mass increment of 257.1 Da for a single alkylation by PIA. Solvent-accessible cysteine residues in P450 3A4 were also alkylated using 100 equiv of IA and analyzed by LC-MS. Although it was difficult to distinguish the small increment shift of 57.0 Da due to alkylation by IA, the apoprotein appeared to be largely modified with a single equivalent of IA as indicated by the molecular mass of 57 323.6 ( 0.6 Da in the deconvoluted spectrum (Figure 1D). The massto-charge ratios and deconvoluted spectra both contained broad peaks, suggesting a mixture of P450 3A4 with 0, 1, or possibly 2 alkylated residues, but the relative distribution is difficult to determine. Glutathione was used to quench the remaining alkylating agent at the 1-h time point so that P450-reductase and cytochrome b5, which were added to the reconstitution mix in subsequent experiments, would not be modified. Increasing the concentration of PIA or IA did not affect the extent of alkylation at the 1-h time point (data not shown), so the levels were kept low to minimize effects in metabolic studies. Ferrous CO-bound difference spectra were utilized to monitor the stability of P450 3A4 following alkylation for 1 h at room temperature. PIA or IA treatment decreased the P450 content to 87% and 71% of control, respectively, although a corresponding increase at 420 nm was not observed for the IA-treated preparation (data not shown). Identification of a P450 3A4 Residue Modified by Raloxifene and Cysteine Alkylating Agents. The peptides resulting from the proteinase K digestion of P450 3A4, subsequent to reaction with raloxifene, were separated on a C18 column and analyzed by UV and ESI-MS. The UV 214 nm trace shows the elution profile of all peptides (Figure 2A), and the UV 340 nm trace shows the elution of raloxifene-derived species, which contained a chromophore that uniquely absorbs in this region, at 18.5 min (Figure 2B). An initial MS scan revealed a m/z value of 660.4 corresponding to the single dominant UV 340

nm absorbing peak (Figure 3A). The SRM trace for the m/z 660.4-1092.2, a fragment ion, confirmed that the species with this mass aligns with the peak in the UV 340 nm trace (Figure 2C). Protein Prospector was used to search for raloxifeneadducted peptides with the mass equal to 660.4 and 1319.8 Da, considering that m/z 660.4 contained two charges. Assuming that a diquinone intermediate reacted with a cysteine residue of P450 3A4, as observed in the reaction with glutathione, the mass increase of an adducted peptide should be 471.2 Da, equal to the parent mass of raloxifene minus two hydrogens. This search in Protein Prospector produced the cysteine-containing peptide 237-NICVFPR-243. In the MS2 spectra the fragment ion for the raloxifene moiety at m/z 506 was observed, as previously seen in mass spectra of raloxifene-GSH adducts (Figure 3B) (6). The major fragment at m/z 651.9 represents the loss of water from the parent, double-charged, peptide, and the fragment at m/z 1092.2 represents the mass of the singlecharged y5 ion, assuming that the cysteine residue is the site of adduct formation. Subsequently, a MS3 spectrum of the y5 ion was acquired, and all of the predicted b fragment ions (b1 ) m/z 547, b2 ) m/z 674, b3 ) m/z 821, b4 ) m/z 918) were observed (Figure 3C). Fragment ion m/z 1075 represents the loss of water from the y5 ion, and m/z 587 represents the y5 ion following loss of the raloxifene moiety. In summary, these results confirmed that the primary sequence of the peptide is 237-NICVFPR-243 and that raloxifene was adducted to specifically to Cys239. The same protocol for P450 3A4 isolation and digestion by proteinase K was used to prepare P450 3A4-PIA, and the sample was loaded onto a C18 column for peptide separation. Effective digestion was confirmed by the elution profile of peptides in the UV 214 nm trace (Figure 4A). The UV 340 nm trace revealed a single dominant peak at 21 min, representing a peptide with the pyrene moiety of PIA (Figure 4B). Additionally, the complete UV-vis spectrum of the peak, acquired with the online diode array detector, was similar to the spectrum of PIA alone (data not shown). The mass of the peptide eluting at 21



Figure 3. Mass spectra of the raloxifene-adducted peptide, 237-NICVFPR-243. (A) Spectrum acquired at 18.5 min corresponding to the UV 340 nm absorbing peak shown in Figure 2B. (B) MS2 spectrum of m/z 660.5. (C) MS3 spectrum of m/z 1092.2.

min was m/z 606.0 (Figure 5A), and the SRM m/z 606.0-379.0 shows the aligned retention time in relation to the UV chromatogram (Figure 4C). Protein Prospector was used to identify the alkylated peptide with this mass as 237-NIC-239

(alkylation by PIA adds 257.1 Da). Subsequent MS2 experiments revealed the y1 (m/z 379.0) and y2 (m/z 492.1) fragment ions of this peptide as well as loss of water (m/z 589.1) (Figure 5B). Mass analysis of free cysteine alkylated by PIA also gives a


Baer et al.

Figure 4. Chromatographic separation of peptides following proteinase K digestion of P450 3A4-PIA: (A) UV 214 nm trace, (B) UV 340 nm trace, and (C) SRM m/z 606-379 trace.

m/z value of 379.0, in agreement with the cysteine residue being the C-terminal residue of the observed peptide (data no shown). From these results it can be concluded that the same cysteine residue at position 239 in P450 3A4 that is modified by raloxifene under turnover conditions is also assessable to alkylation by PIA. Purified P450 3A4-IA was digested with trypsin, which hydrolyzes the protein at specific residues and results in predictable peptides. Without a chromophore on the alkylating agent, modified peptides could not be traced following digestion with proteinase K, which nonspecifically digests the protein. In preliminary experiments, peptides containing cysteine residues, with or without alkylation, could not be detected after digestion of P450 3A4-IA by proteinase K. The only P450 3A4IA peptide from the trypsin digest that displayed an increase of 57.0 Da relative to control was peptide 213-FDFLDPFFLSITVFPFLIPILEVLNICVFPR-243, which eluted at 29.6 min (data not shown). The sequence of the peptide was confirmed by MS2 experiments, and carbamidomethylation of Cys239 was verified by the mass increase of 57.0 Da in the y5 ion (m/z 678.3) but not in the y4 ion (m/z 518.3). Time-Dependent Inhibition by Raloxifene Is Prevented by Pretreatment with PIA or IA. The protective effect of Cys239 alkylation, by either PIA or IA, against raloxifene timedependent inactivation was evaluated in a secondary reaction with the probe substrate, BQ. As expected for a time-dependent inhibitor, the rate of inactivation is greater for the reactions containing raloxifene (kobs ) 9.4 × 10-2 ( 1.3 × 10-2 min-1) as compared to the control incubation (kobs ) 3.8 × 10-2 ( 0.2 × 10-2 min-1). When P450 3A4-PIA is analyzed under identical conditions, the control reaction itself displays an increase in the initial rate constant (kobs ) 7.6 × 10-2 ( 0.6 × 10-2 min-1) as compared to the non-alkylated P450 3A4 preparation. Interestingly, the reaction with raloxifene does not significantly alter the initial rate constant (kobs ) 8.2 × 10-2 ( 0.7 × 10-2 min-1) relative to its control reaction, indicating that PIA offers a protective effect. P450 3A4-IA does not show

additional inactivation in the primary reaction (kobs ) 3.8 × 10-2 ( 0.3 × 10-2 min-1), suggesting that carbamidomethylation does not disrupt the stability of the protein. Yet, this alkylation maintains the protective effect against raloxifene timedependent inhibition (kobs ) 4.3 × 10-2 ( 0.7 × 10-2 min-1). Reactions were also conducted in which prereacted PIA or IA with GSH was subsequently added to P450 3A4 following timedependent inhibition measurement. The initial rate constants were not significantly different from the control reactions, indicating that the quenched alkylating agents do not disrupt substrate turnover (data not shown). Functional Characterization of PIA- and IA-Treated P450 3A4. To evaluate the proximity of the alkylated cysteine residue relative to the heme, ligand-induced spin-state changes were monitored using midazolam. The binding affinity for midazolam was similar for untreated P450 3A4 (5.3 ( 0.2 µM) as for the P450 3A4 sample pretreated with PIA (4.4 ( 0.1 µM). The difference spectra revealed a lower spin-state change for the PIA-treated P450 3A4 as compared to the control spectra for saturating concentrations of midazolam (Figure 6). This effect may be due to slightly less intact P450 following alkylation (87% of control as determined by the ferrous CO difference spectra). The relative production of glutathionyl-raloxifene and hydroxyl-raloxifene by PIA- and IA-treated P450 3A4 was measured to ensure that the alkylation procedure was actually preventing raloxifene adduction and not simply inhibiting turnover and therefore preventing the reactive intermediates responsible for protein adduction. Without standards for either metabolite, the analyte levels were normalized to the production by unmodified P450 3A4. PIA-treated P450 3A4 produced similar amounts of the two metabolites (117 ( 9% glutathionylraloxifene and 118 ( 6% hydroxyl-raloxifene) as compared to the control reaction (100 ( 8% glutathionyl-raloxifene, 100 ( 5% hydroxyl-raloxifene) (Figure 7). Interestingly, P450 3A4 pretreated with IA produced significantly more of each metabolite (200 ( 16% glutathionyl-raloxifene, 162 ( 18%



Figure 5. Mass spectra of PIA-alkylated peptide, 237-NIC-239. (A) Spectrum was acquired at 20.9 min corresponding to the UV 340 nm absorbing peak shown in Figure 4B. (B) MS2 spectrum of m/z 660.0.

hydroxyl-raloxifene). Control reactions in which GSH was reacted with IA or PIA for 1 h prior to incubation with P450 3A4 demonstrated that alkylated GSH did not greatly influence P450 3A4 turnover of raloxifene. Incubations in which NADPH was omitted produced relatively little signal for the two metabolites monitored (6.3 ( 0.9% glutathionyl-raloxifene, 1.5 ( 0.2% hydroxyl-raloxifene). HLMs were also included in this assay for comparison, and they demonstrated higher production of the hydroxyl-raloxifene relative to glutathionyl-raloxifene (Figure 7).

Discussion Direct characterization of the chemical intermediate(s) adducted to P450 apoprotein and presumably responsible for mechanism-based inhibition offers a distinct advantage over predictions made based on the identity of GSH-drug adducts that may or may not be related to the mechanism of inactivation. For example, raloxifene has been shown to form numerous GSH adducts from in vitro incubations (Scheme 1), yet it is unknown which if any of the GSH adducts represent the species responsible for inactivating P450 3A4. Therefore, the raloxifene P450 3A4 adduct was measured, and the deconvoluted mass spectral results produced an adducted mass difference of 471 Da, which is consistent with the adducting species arising from the diquinone methide intermediate (Scheme 1, Reaction A).

This measured difference in mass eliminates the o-quinone (Scheme 1, Reaction B) as the protein adducting species because the mass difference due to the raloxifene adduct would be 487 Da. However, the diquinone methide intermediate yields several electrophilic sites that could be quenched by P450 3A4 whereby the resultant protein adduct(s) would all produce a mass change of 471 Da. Thus, to further distinguish the specific site of raloxifene alkylation to P450 3A4 and provide insight to raloxifene P450 3A4 interactions a “bottom-up” analysis approach was used in an attempt to define the regioselectivity of alkylation. Identification of inactivating species and residues within the active site of P450 3A4 has been attempted previously using mechanism-based inhibitors. Historically, radiolabeled material is used to trace the adducted protein through numerous manipulations including dialysis, enzymatic digests, and LCMS conditions to facilitate the location of protein adducts. Despite the use of radiolabel tracers many of these attempts fail speculatively due to factors including incomplete digestion, poor ionization of the hydrophobic peptide to which the adduct is attached, complexity of the digestion mixtures, and low stability of the resultant alkylated peptide species. For example, Lightning et al. incubated P450 3A4 with a potent mechanismbased inhibitor [14C]-L-754,394 in order to trace the alkylated protein through gel electrophoresis and subsequent digestion


Baer et al.

Figure 6. Type I binding spectra acquired during the titration of midazolam into solutions of P450 3A4 (A) or P450 3A4-PIA (B).

of the protein with CNBr in an attempt to identify the adducted peptide(s). A single peptide fragment associated with the radioactivity was identified to be I257-X317 (X denotes a homoserine lactone) based on MS data, although the drug adduct was never directly observed. The authors speculate that because the drug adduct was unstable during preparation for mass analysis the covalent linkage must be an ester bond with a Glu or Asp residue and tentatively assigned Glu307 on the I-helix as the site of attachment (5). Radiolabeled compound provided a tracer of the alkylation product through a series of experiments to facilitate the tentative identification of Glu307, which can

be supported further with recent crystal structures that show that Glu307 lies within the P450 3A4 active site (11, 12). The GSH adducts of raloxifene suggest a high affinity of the bioactivated intermediates for ‘soft’ nucleophiles. Solventaccessible cysteine residues were probed with cysteine-selective alkylating agents, PIA and IA, to maximize the potential for locating nucleophilic sites in P450 3A4. PIA was chosen for the alkylating agent in an effort to increase the propensity for alkylation to occur near the enzyme active site given pyrene and several close analogs are known P450 3A4 substrates (1315). Second, pyrene is fluorescent, which should facilitate the



Figure 7. Relative production of glutathionyl-raloxifene and hydroxyl-raloxifene by P450 3A4 following treatment with PIA or IA.

location of any alkylated residues and peptides with UV and fluorescence detection. A 1:1 stoichiometry was produced between PIA and P450 3A4 as monitored by mass spectrometry (Figure 1C). The site of alkylation was then investigated by protein digestion using proteinase K. Proteinase K was chosen to digest the alkylated protein because this protease efficiently and nonspecifically cleaves proteins into small peptides and amino acids. Digestion of protein using proteinase K should allow for the separation and identification of a drug-peptide adducts from a complex mixture based upon the distinct physical chemical differences expected for hydrophobic drug-peptide adduct(s) compared to the mixture of smaller hydrophilic peptide and amino acid fragments (16). In fact, the proteinase K digest of the PIA-P450 3A4 adduct produced a hydrophobic peptide adduct distinctly separated from the other fragments (Figure 2). Despite the nonspecific nature of proteinase K cleavage only one detectable PIA-peptide fragment was produced from three separate experimental digests, suggesting this technique is reproducible and thus may be suitable for other adducts. Therefore, the same procedure was performed with the raloxifene-P450 3A4 and yielded a single alkylated peptide (Figure 4). This signified that not only is a single site of P450 3A4 alkylated by the diquinone methide intermediate but also the quenching of the intermediate occurred with a high degree of regiospecificity given the detection of a single peak in the UV and MS chromatograms. These data are supported by data from the diquinone methide quenched with GSH, which produce resolved adduct peaks with chromatography, suggesting that if multiple peptide adducts formed they would separate in a similar fashion (6). Similar to the data from the intact mass spectra the mass of the peptide adduct was 471 Da larger than the predicted mass for peptide 237-NICVFPR-243 alone (Figure 3). Inspection of the adducted peptide by MS2 produced a fragment m/z 506, which likely results from cleavage adjacent to the thioether moiety and agrees with the spectrum previously acquired for the reactive diquinone methide raloxifene metabolites trapped by GSH (6). Further structural information could not be inferred from MS2 data due to the ambiguous fragmentation of the

adducted peptide and is consistent with that observed for the GSH-raloxifene adducts with the o-quinone (6). These data confirmed the role of the diquinone methide as the intermediate responsible for P450 3A4 inactivation and are congruent with the short ∼1 s lifetime previously measured for the diquinone methide intermediate and with the reduced time-dependent inhibition of close structural analogs that cannot form the diquinone methide (7, 17). Specific alkylation of Cys239 by both raloxifene and the iodoacetamide reagents provides a unique opportunity to examine the mechanism of P450 3A4 inactivation by raloxifene. Pretreatment of P450 3A4 with PIA or IA should block raloxifene adduct formation at Cys239 and, presumably, prevent the associated inactivation. Although the kinetics of 7-hydroxyquinoline formation by P450 3A4-PIA indicated a higher basal rate of protein degradation as compared to the unmodified protein, pretreatment with PIA prevented additional timedependent inactivation due to raloxifene turnover. Similar levels of hydroxyl-raloxifene and glutathionyl-raloxifene were detected in incubations with P450 3A4-PIA as in the control reaction, confirming that raloxifene was still a substrate for the modified enzyme. A similar experiment was conducted with P450 3A4 treated with IA, and again, pre-alkylation of the protein prevented time-dependent inactivation by raloxifene. Unlike treatment with PIA, IA treatment alone did not destabilize P450 3A4 as determined by the kinetics of BQ metabolism. In addition, the relative levels of both hydroxyl-raloxifene and glutathionyl-raloxifene were higher for P450 3A4-IA than the untreated P450. Collectively, these experiments confirmed that the site of alkylation for the cysteine modifying reagents and the raloxifene diquinone methide are the same and that adduction of Cys239 is responsible for the observed time-dependent inactivation. Moreover, the increased rate of P450 3A4 inactivation by PIA treatment alone indicated that this reaction may serve as an experimental model to further explore the relationship between protein-drug adducts and enzyme inactivation. Multiple mechanisms for inactivation due to protein adduction can be envisioned: (1) the drug adduct binds in the active site


and prevents additional turnover by blocking access to the heme catalytic center, (2) the drug adduct binds to the surface of the enzyme and prevents critical interactions with P450-reductase, or (3) the drug adduct induces a conformational change in the protein structure that results in disruption of necessary interactions for substrate binding or displacement of residues involved in the catalytic cycle. The functional effects observed with IAand PIA-treated P450 3A4 compared to raloxifene inactivation can help differentiate between these mechanisms. Time-dependent inactivation experiments demonstrated that BQ, a common P450 3A4 probe substrate, can still be metabolized by the control PIA-treated P450 3A4, indicating that the substrate can still access the active site. Even more convincingly, there was similar turnover of raloxifene to both hydroxylraloxifene and glutathionyl-raloxifene by native P450 3A4 and the PIA-treated P450 3A4. In addition, midazolam could still induce a type I spectral shift upon binding to PIA-treated P450 3A4, and the ligand exhibited a similar affinity for the alkylated enzyme as for the untreated P450 3A4. Given the multiple binding sites of P450 3A4, Cyclosporin A binding was also conducted and produced similar spectral shifts with and without alkylation (data not shown). From the enzymatic data collected in these experiments we cannot rule out the possibility of the alkylation interfering with redox partners. Yet, disruption of P450-reductase binding and electron transfer is not anticipated based on the position of Cys239 in the solved X-ray crystal structure, located between the G and G′ helices and buried in the lipid membrane, and the significant amount of enzymatic turnover observed with the alkylated proteins. Therefore, the observed enzymatic activities and biophysical properties of P450 3A4-PIA as well as the location of the modified residue are consistent with a mechanism involving a conformational change in protein structure, albeit additional experiments would be required to confirm this hypothesis. In summary, LC-MS analysis of both intact protein and proteolytic digests of P450 3A4, subsequent to incubation with raloxifene, revealed that a single equivalent of drug binds to Cys239 of P450 3A4, formed from the reactive diquinone methide intermediate. The mechanism for the observed timedependent inactivation is likely protein unfolding to an inactive conformation, analogous to the decreased stability of P450 3A4 when Cys239 is alkylated by PIA. Furthermore, time-dependent inactivation, but not raloxifene turnover, can be prevented with pretreatment of P450 3A4 with PIA or IA, establishing a direct relationship between adduction of drug and reduction in enzymatic activity. This methodology for identifying the chemical species responsible for covalent binding to P450 and subsequent inactivation may show utility in the lead optimization stage of drug development and be used to distinguish a different mechanism of inhibition by different chemical functional groups.

References (1) Kalgutkar, A. S., Gardner, I., Obach, R. S., Shaffer, C. L., Callegari, E., Henne, K. R., Mutlib, A. E., Dalvie, D. K., Lee, J. S., Nakai, Y.,

Baer et al.

(2) (3)

(4)

(5)

(6)

(7)

(8)

(9)

(10) (11)

(12)

(13)

(14)

(15) (16) (17)

O’Donnell, J. P., Boer, J., and Harriman, S. P. (2005) A comprehensive listing of bioactivation pathways of organic functional groups. Curr. Drug Metab. 6, 161-225. Monks, T. J., Lau, S. S., and Gillette, J. R. (1984) Diffusion of reactive metabolites out of hepatocytes: studies with bromobenzene. J. Pharmacol. Exp. Ther. 228, 393-399. Zhou, S., Yung Chan, S., Cher Goh, B., Chan, E., Duan, W., Huang, M., and McLeod, H. L. (2005) Mechanism-based inhibition of cytochrome P450 3A4 by therapeutic drugs. Clin. Pharmacokinet. 44, 279-304. Koenigs, L. L., Peter, R. M., Hunter, A. P., Haining, R. L., Rettie, A. E., Friedberg, T., Pritchard, M. P., Shou, M., Rushmore, T. H., and Trager, W. F. (1999) Electrospray ionization mass spectrometric analysis of intact cytochrome P450: identification of tienilic acid adducts to P450 2C9. Biochemistry 38, 2312-2319. Lightning, L. K., Jones, J. P., Friedberg, T., Pritchard, M. P., Shou, M., Rushmore, T. H., and Trager, W. F. (2000) Mechanism-based inactivation of cytochrome P450 3A4 by L-754,394. Biochemistry 39, 4276-4287. Chen, Q., Ngui, J. S., Doss, G. A., Wang, R. W., Cai, X., DiNinno, F. P., Blizzard, T. A., Hammond, M. L., Stearns, R. A., Evans, D. C., Baillie, T. A., and Tang, W. (2002) Cytochrome P450 3A4-mediated bioactivation of raloxifene: irreversible enzyme inhibition and thiol adduct formation. Chem. Res. Toxicol. 15, 907-914. Yu, L., Liu, H., Li, W., Zhang, F., Luckie, C., van Breemen, R. B., Thatcher, G. R., and Bolton, J. L. (2004) Oxidation of raloxifene to quinoids: potential toxic pathways via a diquinone methide and o-quinones. Chem. Res. Toxicol. 17, 879-888. Gillam, E. M., Baba, T., Kim, B. R., Ohmori, S., and Guengerich, F. P. (1993) Expression of modified human cytochrome P450 3A4 in Escherichia coli and purification and reconstitution of the enzyme. Arch. Biochem. Biophys. 305, 123-131. Holmans, P. L., Shet, M. S., Martin-Wixtrom, C. A., Fisher, C. W., and Estabrook, R. W. (1994) The high-level expression in Escherichia coli of the membrane-bound form of human and rat cytochrome b5 and studies on their mechanism of function. Arch. Biochem. Biophys. 312, 554-565. Rock, D., Rock, D., and Jones, J. P. (2001) Inexpensive purification of P450 reductase and other proteins using 2′,5′-adenosine diphosphate agarose affinity columns. Protein Expr. Purif. 22, 82-83. Williams, P. A., Cosme, J., Vinkovic, D. M., Ward, A., Angove, H. C., Day, P. J., Vonrhein, C., Tickle, I. J., and Jhoti, H. (2004) Crystal structures of human cytochrome P450 3A4 bound to metyrapone and progesterone. Science 305, 683-686. Yano, J. K., Wester, M. R., Schoch, G. A., Griffin, K. J., Stout, C. D., and Johnson, E. F. (2004) The structure of human microsomal cytochrome P450 3A4 determined by X-ray crystallography to 2.05-A resolution. J. Biol. Chem. 279, 38091-38094. Shou, M., Grogan, J., Mancewicz, J. A., Krausz, K. W., Gonzalez, F. J., Gelboin, H. V., and Korzekwa, K. R. (1994) Activation of CYP3A4: evidence for the simultaneous binding of two substrates in a cytochrome P450 active site. Biochemistry 33, 6450-6455. Silvers, K. J., Chazinski, T., McManus, M. E., Bauer, S. L., Gonzalez, F. J., Gelboin, H. V., Maurel, P., and Howard, P. C. (1992) Cytochrome P-450 3A4 (nifedipine oxidase) is responsible for the C-oxidative metabolism of 1-nitropyrene in human liver microsomal samples. Cancer Res. 52, 6237-6243. Schrag, M. L., and Wienkers, L. C. (2000) Topological alteration of the CYP3A4 active site by the divalent cation Mg(2+). Drug Metab. Dispos. 28, 1198-1201. Manabe, S., Sassa, S., and Kappas, A. (1985) Hereditary tyrosinemia. Formation of succinylacetone-amino acid adducts. J. Exp. Med. 162, 1060-1074. Liu, H., Liu, J., van Breemen, R. B., Thatcher, G. R., and Bolton, J. L. (2005) Bioactivation of the selective estrogen receptor modulator desmethylated arzoxifene to quinoids: 4′-fluoro substitution prevents quinoid formation. Chem. Res. Toxicol. 18, 162-173.

TX700037E

Time-Dependent Inactivation of P450 3A4 by Raloxifene: Identification

Recommend Documents