Time-Dependent Inactivation of P450 3A4 by Raloxifene: Identification of Cys239 as the Site of...

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 toform reactive species that subsequently alkylate the heme group, apoprotein, or both. The chemical identityof many reactive intermediates is generally proposed based on the products of trapping reactions withnucleophilic agents as only a few P450-drug adducts have been directly characterized. We describe theuse of mass spectrometry to show that a single equivalent of raloxifene is bound to the intact P450apoprotein. Furthermore, mass analysis of peptides following digestion with proteinase K revealed thatthe covalently bound drug is localized to residue Cys239. A mass shift of 471 Da to the intact proteinand peptide, relative to control samples, indicated that time-dependent inactivation of P450 3A4 occurredthrough the raloxifene diquinone methide intermediately prior to nucleophilic attack of the sulfur ofCys239. Association between raloxifene adduction to P450 3A4 apoprotein and the observed time-dependent inactivation was further investigated with the use of cysteine-specific modifying reagents.When P450 3A4 was treated with iodoacetamide orN-(1-pyrene)iodoacetamide, which alkylated residueCys239 exclusively, time-dependent inactivation of P450 3A4 by raloxifene was prevented. The changein protein mass of 471 Da combined with the protection from inactivation that occurred through pre-alkylation of Cys239 provided conclusive evidence that raloxifene-mediated P450 3A4 inactivationoccurred through the bioactivation of raloxifene to the diquinone methide and subsequent alkylation ofCys239.

Introduction

In vitro metabolism studies are routinely used to predict thein vivo biotransformation pathways of new chemical entities.Of particular interest are the specific oxidative pathways thatform reactive metabolites of which certain substructures aremore prone to generate “bioactivated” intermediates that canirreversibly inhibit the oxidizing P450 enzyme (1). The basalenzyme activity of the affected P450 can only recover as fastas the natural synthesis rate of the P450 enzyme, and as a resultthe susceptibility to drug interactions increases in instanceswhere clearance is mediated by the inactivated P450. To thisend, identification and elimination of these events is an importantstep in the development of safe pharmaceutical drugs. The mostcommon early indicator of bioactivation is with the indirectobservation of reactive metabolites trapped with exogenousnucleophiles like glutathione (GSH) (2). These methods identifythe presence of a bioactivation step that generates an unstableelectrophilic intermediate and typically triggers time-dependentinhibition experiments to look for P450 inactivation. Time-dependent inhibition experiments are designed to investigatethe degree of P450 inactivation through the generation ofkinact

and an apparentKI, which are used to further context thepotential in vivo significance (3). Best practice is to eliminatethe structural motif in the compound that is responsible for

inactivation. However, trapping agents such as GSH may notreflect the species responsible for enzyme inactivation, com-plicating identification of the culprit structural motif.

Measuring P450-adduct complexes with mass spectrometryprovides a method whereby the enzyme inactivating species canbe measured directly. Foremost, the basic characteristics of thedrug-protein adducts can be deduced from the mass changeassociated with the adducted protein. For example, a proteinadduct may reflect a change in mass consistent with the parentdrug plus 16, reflecting an oxidized metabolic species as theprecursor to protein adduct formation. The level of detailregarding the protein-drug adduct can be increased throughenzymatic or chemical digestions which enable MSn fragmenta-tion data to be generated on the adducted peptide(s). The latterstep has proven to be exceptionally challenging. For example,an intact protein adduct of tienilic acid and P450 2C9 wasgenerated from which a mechanism for the thiophene basedadduct was proposed; however, the exact location of adductionwas unattainable after tryptic digestion (4). A furan-containingstructure, L-754,394, was postulated to bind to Glu307 of theapoprotein based on evidence from Tricine SDS-PAGE andMALDI-TOF-MS using radiolabeled material, but the intactadduct was not observed (5). While the culprit substructuresmay be intuitive in the aforementioned examples, these proce-dures do not provide detailed information necessary for lessobvious bioactivation mechanisms and would fail to yieldhypotheses for structural modifications needed to eliminatereactive chemistry from a chemical scaffold.

Raloxifene represents a well-characterized mechanism-basedinhibitor of P450 3A4 (6). In vitro, P450 3A4 mediated themetabolism of raloxifene to produce several electrophilic species

* To whom correspondence should be addressed. Phone: (206) 265-7139. 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-glycero-3-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.

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

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

including a diquinone methide intermediate that reacted withGSH at various locations on the molecule (6, 7) (Scheme 1).Presumably, one of these species is responsible for P450 3A4inactivation. Addition of GSH to the P450 3A4 incubation doesnot fully protect the enzyme from inactivation (6). The ralox-ifene intermediates were synthesized and further characterizedto have distinctly different half-lives. These differences mayprovide a rational basis for distinguishing the protein-basedinactivating species from the remainder of the GSH adducts(7). In total, four GSH adducts have been characterized. In thisreport we identified a single P450 3A4-mediated alkylationproduct of raloxifene to the apoprotein of P450 3A4 via wholeprotein mass spectrometry. Following proteinase K digestion,MSn analysis of the alkylated peptide revealed the site of adductformation. The adducted site was confirmed to be responsiblefor the time-dependent inhibition of P450 3A4 by raloxifene,and the location was confirmed in an additional experimentusing iodoacetamide to block the electrophilic residue. Theresults are discussed with a focus on direct measurement ofreactive metabolites, the mechanism of inactivation by irrevers-ible inhibitors, and use of protein-adduct characterization ofP450 mechanism-based inhibitors.

Experimental Procedures

Materials. Iodoacetamide, reduced glutathione, Chaps, potassiumHepes, MgCl2, guanidine hydrochloride, raloxifene, CaCl2, ZnSO4,and NADPH were purchased from Sigma-Aldrich (St. Louis, MO).7-Benzyloxyquinoline (BQ), 7-hydroxyquinoline metabolite stan-dard, and pooled HLMs were purchased from BDGentest (Woburn,MA). N-(1-Pyrene)iodoacetamide (PIA) was purchased from In-vitrogen (Carlsbad, CA).L-R-Dilauroyl-sn-glycero-3-phosphocho-line (DLPC), l-R-diloleoyl-sn-glycero-3-phosphocholine (DOPC),and l-R-dilauroyl-sn-glycero-3-phosphoserine (DLPS) were pur-chased from Avanti Polar Lipids Inc. (Alabaster, AL). Sequencing-grade 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 aZorbax RX-C8 column (2.1× 150 mm) was from Agilent (PaloAlto, CA).

P450 Enzymes and Co-enzymes.Recombinant P450 3A4 wasexpressed inEscherichia coli DH5R in the pCW 3A4-His6expression vector, kindly provided by Dr. Ron Estabrook. In a 2.8L 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 buffercontaining 50 mM potassium phosphate (pH 7.4), 500 mM NaCl,20% glycerol, 50µM testosterone (stabilizing ligand), 20 mMâ-mercaptoethanol, 1% Emulgen 911, and Sigma protease inhibitorcocktail (0.5 mL/L of initial culture volume). Cells were lysed usinga French Press operated at 10 000 psi and then spun at 150 000g.The supernatant was loaded directly onto NTA-Ni resin, equilibratedwith 50 mM potassium phosphate (pH 7.4), 500 mM NaCl, 20%glycerol, 50µM testosterone, and 0.2% Emulgen 911. The columnwas washed with 200 mL of wash buffer containing 50 mMpotassium phosphate (pH 7.4), 20% glycerol, 20 mM imidazole,0.2% cholate, and 50µM testosterone. P450 3A4 was eluted fromthe column with buffer containing 50 mM potassium phosphate(pH 7.4), 20% glycerol, 500 mM imidazole, and 0.2% cholate. Theeluted protein was dialyzed against 100 mM potassium phosphate,pH 7.4, in 20% glycerol and stored at-80 °C. Rat cytochromeb5

and P450 reductase were expressed and purified as describedpreviously (9, 10).

Instrumentation. A LTQ mass spectrometer from Thermo (SanJose, CA) was connected online with an Agilent 1100 series HPLC(Palo Alto, CA) including a degasser, pumps, autoinjector, columnoven, and diode array detector. A Cary 6000 UV/vis spectrometerfrom Varian, Inc. (Palo Alto, CA) was used for measuring ligandbinding and ferrous CO-bound P450 spectra. A Safire2 fluorometerfrom Tecan (San Jose, CA) was used to measure turnover of BQto 7-hydroxyquinoline.

P450 3A4 Reaction with Iodoacetamide or Pyrenyliodoac-etamide. Pure recombinant P450 3A4 (500 pmol) was combinedwith 20 µg/mL liposomes [DLPC, DOPC, DLPS (1:1:1, w/w/wper mL)] and 0.1 mg/mL Chaps in 100µL of 50 mM potassiumphosphate (pH 7.4) buffer. IA (500µM) or PIA (50 µM) wasallowed to react with P450 3A4 for 1 h at room temperature.Initial experiments demonstrated P450 3A4 was not sufficientlyalkylated by treatment with 50µM of IA, a concentrationthat proved effective with PIA treatment, and therefore, 500µMIA was used in subsequent experiments. The reaction wasquenched with 5 mM reduced glutathione. The extent of cysteinealkylation in P450 3A4-IA or P450 3A4-PIA was determined bythe relative increase in the deconvoluted mass of the P450 3A4protein.

Incubations with Reconstituted P450 3A4 and Raloxifene.P450 3A4 (100 pmol) was combined with NADPH-P450 reductase(200 pmol), cytochromeb5 (100 pmol), 0.1 mg/mL Chaps, 20µg/mL liposomes [DLPC, DOPC, DLPS (1:1:1, w/w/w per mL)], 3mM reduced glutathione, 50 mM potassium Hepes, pH 7.4, and30 mM MgCl2 in a total volume of 0.5 mL. Raloxifene was addedto a final concentration of 50µM. The reconstituted P450 3A4was incubated for 2 min at 37°C prior to addition of 1 mM

Scheme 1. P450 3A4 Bioactivation of Raloxifene

Raloxifene Cys Adduct of P450 3A4 Apoprotein Chem. Res. Toxicol., Vol. 20, No. 6, 2007955

NADPH. The reaction was allowed to proceed for 20 min unlessotherwise noted.

Time-Dependent Inactivation of P450 3A4.Aliquots of theprimary reaction of P450 3A4 and raloxifene were removed at 1,5, 10, and 20 min and placed on ice. Time-dependent inactivationof P450 3A4 was determined by quantitating the remaining activitytoward BQ. The primary reaction was diluted 1:10 into thesecondary reaction containing 0.1 mg/mL Chaps, 20µg/mLliposomes [DLPC, DOPC, DLPS (1:1:1, w/w/w per mL)], 3 mMreduced glutathione, 50 mM potassium Hepes, pH 7.4, and 30 mMMgCl2, 1 mM NADPH, and 80µM BQ. The final volume of thesecondary reaction was 200µL and incubated in a 96-well plate(black with clear bottom). Production of 7-hydroxyquinoline wasmeasured every minute for 20 min at 37°C by fluorescence(excitation, 409 nm; emission, 530 nm). The 96-well plate wasshaken for 10 s prior to each measurement. P450 3A4 activity wasdetermined from a 10-min linear portion of the curve. Measurementswere made in triplicate to allow for calculation of standarddeviations. In control reactions, either raloxifene or NADPH wasomitted in the primary incubation.

Mass Analysis of Intact Protein.P450 3A4 protein (100 pmol)was separated from alkylation reagents (IA and PIA modified P4503A4) or incubation components (P450 3A4-raloxifene adduct) ona Poros R2 perfusion chromatography column (2.1× 100 mm) tofacilitate electrospray ionization. Protein in incubations withraloxifene was concentrated by precipitating with 1:10 dilution with15% ZnSO4, centrifuging at 13 000 rpm, and then resuspendingwith 100 µL of 6 M guanidine hydrochloride. Initial HPLCconditions 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 1min, 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 300µL/min from 6 to 10 min during data acquisition with the interfacedFinnigan LTQ mass spectrometer. The column was then washedby increasing solvent B from 65% to 85% in 1 min and holding at85% for 1 min at a flow rate of 1 mL/min. P450 3A4 eluted at 8min, 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 todetermine the standard deviation within data acquisition anddeconvolution calculations.

Proteinase K and Trypsin Digestion of P450 3A4.Prior toenzymatic digestion of P450 3A4, approximately 500 pmol ofprotein was purified on the Poros R2 column as described above.The peak corresponding to P450 3A4 in the 280 nm trace wascollected in a 1.5 mL tube and evaporated to 50µL. Buffercontaining 50 mM Hepes, pH 7.4, and 10 mM CaCl2 was added toeach sample for a total of 200µL. Samples prepared for proteinaseK digestion also contained 3 M guanidine hydrochloride to helpdenature the protein and increase proteolysis. Following additionof 5% w/w of proteinase K or trypsin, reactions were allowed toproceed for 2 h at 37°C. Reactions were stopped by placing onice until mass analysis.

UV-vis and Mass Analysis of Alkylated Peptides.Thepeptides resulting from digestion of P450 3A4 were injected ontoa 2.1× 250 mm C18 column (Grace Vydac) at a flow rate of 200µL/min. Initial conditions were 95% solvent A (0.05% TFA in H2O)and 5% solvent B (0.05% TFA in acetonitrile), and the followinggradient 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, acquiredwith an Agilent 1100 DAD detector, was used to locate the retentiontime of peptides alkylated with PIA or raloxifene. The interfacedFinnagin LTQ was used to first identify the modified peptides basedon parent mass and subsequently for MS2 experiments to confirmthe peptide sequence and the identity of the modified residue.Specific wavelengths (340 nm) and mass fragments were used toconfirm alignment of retention times between the UV and the MS.The MS was performed using a SRM scan with the defaultactivation time and collision energy of 35. Protein Prospector (http://prospector.ucsf.edu) was used to identify the sequences of the

modified peptides based on the experimentally determined molec-ular masses.

Measurement of Raloxifene Metabolites.Following a 20 minincubation of P450 3A4, P450 3A4-PIA, or P450 3A4-IA withraloxifene (described above) the reaction was quenched with 0.5mL of ice-cold acetonitrile containing 1µg/mL tolbutamide as aninternal standard. Control reactions were conducted simultaneouslyin which (i) GSH was reacted with IA or PIA for 1 h prior toincubation with P450 3A4 and (ii) NADPH was omitted from thereaction. In addition, pooled HLMs (0.4 mg) were incubated for20 min at 37°C for comparison of metabolic profiles. The sampleswere spun at 13 000 rpm at 4°C for 30 min to precipitate andpellet protein. The supernatant was evaporated to 200µL, and 20µL was injected onto an Agilent Zorbax RX-C8 column (2.1×150 mm) at a flow rate of 250µL/min. Initial HPLC conditionswere 90% solvent A (0.05% TFA in H2O) and 10% solvent B(0.05% TFA in acetonitrile). The elution gradient increasedfrom 10% to 50% B in 30 min and 50% to 95% B in 2 min andheld at 95% B for an additional 3 min. Two chromatographicallydistinct glutathionyl-raloxifene metabolites (m/z779) eluted at 17.7and 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 wasquantitated from the ratio of them/z 779 or 490 trace peak areas tothe m/z 271 trace peak areas representing the internal standard.Measurements were conducted in triplicate to determine standarddeviations.

Measurement of Type I Ligand Binding. A solution of 0.5µM P450 3A4, P450 3A4-IA, or P450 3A4-PIA and 20µg/mL ofliposomes [DLPC, DOPC, DLPS (1:1:1, w/w/w per mL)] in 0.5mL of 50 mM potassium phosphate (pH 7.4) buffer was equilibratedat 25 °C in a 0.7 mL cuvette. A baseline spectrum was acquiredfrom 500 to 350 nm at a scan rate of 300 nm/min. Midazolam wastitrated in to final concentrations ranging from 0 to 35µM from0.5 and 3 mM stock solutions in methanol, and spectra wereacquired following each addition. Cyclosporin A was titrated intothe P450 3A4 solution to final concentrations ranging from 0 to16 µM from 0.5 and 3 mM stock solutions in methanol. The finalorganic solvent concentration did not exceed 2% of the total volume.The∆Abs(390-420 nm) was plotted versus ligand concentration, andthe curve was fit assuming one-site saturation in SigmaPlot (version9.0, Systat Software, Inc.) to determine theKS for ligand bindingto P450 3A4.

P450 3A4 Ferrous CO Binding Spectra.Solutions of P4503A4, P450 3A4-IA, or P450 3A4-PIA in 50 mM potassiumphosphate (pH 7.4) were reduced with a few grains of sodiumdithionite, and baseline spectra were acquired from 500 to 400 nmat a scan rate of 300 nm/min. Samples were then bubbled with COfor approximately 30 s, and spectra were acquired.

Results

Characterization of a Raloxifene Adduct on P450 3A4Apoprotein. In comparison with the mass spectrum of P4503A4 alone, the apoprotein from the reaction with raloxifeneshowed an additional ion envelope, which contributed 30-40%of the total ion intensity (Figure 1A and 1B). The deconvolutedspectra of P450 3A4 revealed a single mass of 57 266.4( 0.7Da, whereas the apoprotein from the raloxifene reaction had asecond deconvoluted mass of 57 738.7( 1.6 Da. The mass shiftof 472 Da indicated that one equivalent of the parent raloxifenesubstrate was bound to the P450 3A4 apoprotein.

Alkylation of P450 3A4 by PIA and IA. Previous studieshave shown that raloxifene can be bioactivated by P450 3A4to a reactive arene oxide or diquinone methide, which can eachreact with the cysteine residue of glutathione. Therefore, this‘soft’ nucleophilic residue was the suspect residue for the drug-protein adduct observed with P450 3A4 apoprotein. P450 3A4was alkylated with 100 equiv of PIA in 50 mM potassium

956 Chem. Res. Toxicol., Vol. 20, No. 6, 2007 Baer et al.

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

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

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


ion envelope and the larger molecular mass of 57 524.4( 0.3Da in the deconvoluted spectrum relative to the mass ofunmodified P450 3A4 (Figure 1C). The mass shift of 258.0 Dais near the predicted mass increment of 257.1 Da for a singlealkylation by PIA. Solvent-accessible cysteine residues in P4503A4 were also alkylated using 100 equiv of IA and analyzedby LC-MS. Although it was difficult to distinguish the smallincrement shift of 57.0 Da due to alkylation by IA, theapoprotein appeared to be largely modified with a singleequivalent of IA as indicated by the molecular mass of 57 323.6( 0.6 Da in the deconvoluted spectrum (Figure 1D). The mass-to-charge ratios and deconvoluted spectra both contained broadpeaks, suggesting a mixture of P450 3A4 with 0, 1, or possibly2 alkylated residues, but the relative distribution is difficult todetermine. Glutathione was used to quench the remainingalkylating agent at the 1-h time point so that P450-reductaseand cytochromeb5, which were added to the reconstitution mixin subsequent experiments, would not be modified. Increasingthe concentration of PIA or IA did not affect the extent ofalkylation at the 1-h time point (data not shown), so the levelswere kept low to minimize effects in metabolic studies. FerrousCO-bound difference spectra were utilized to monitor thestability of P450 3A4 following alkylation for 1 h at roomtemperature. PIA or IA treatment decreased the P450 contentto 87% and 71% of control, respectively, although a corre-sponding increase at 420 nm was not observed for the IA-treatedpreparation (data not shown).

Identification of a P450 3A4 Residue Modified by Ral-oxifene and Cysteine Alkylating Agents.The peptides resultingfrom the proteinase K digestion of P450 3A4, subsequent toreaction with raloxifene, were separated on a C18 column andanalyzed by UV and ESI-MS. The UV 214 nm trace shows theelution profile of all peptides (Figure 2A), and the UV 340 nmtrace shows the elution of raloxifene-derived species, whichcontained a chromophore that uniquely absorbs in this region,at 18.5 min (Figure 2B). An initial MS scan revealed am/zvalue of 660.4 corresponding to the single dominant UV 340

nm absorbing peak (Figure 3A). The SRM trace for them/z660.4-1092.2, a fragment ion, confirmed that the species withthis mass aligns with the peak in the UV 340 nm trace (Figure2C). Protein Prospector was used to search for raloxifene-adducted peptides with the mass equal to 660.4 and 1319.8 Da,considering thatm/z 660.4 contained two charges. Assumingthat a diquinone intermediate reacted with a cysteine residueof P450 3A4, as observed in the reaction with glutathione, themass increase of an adducted peptide should be 471.2 Da, equalto the parent mass of raloxifene minus two hydrogens. Thissearch in Protein Prospector produced the cysteine-containingpeptide 237-NICVFPR-243. In the MS2 spectra the fragmention for the raloxifene moiety atm/z 506 was observed, aspreviously seen in mass spectra of raloxifene-GSH adducts(Figure 3B) (6). The major fragment atm/z 651.9 representsthe loss of water from the parent, double-charged, peptide, andthe fragment atm/z 1092.2 represents the mass of the single-charged y5 ion, assuming that the cysteine residue is the site ofadduct formation. Subsequently, a MS3 spectrum of the y5 ionwas 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) wereobserved (Figure 3C). Fragment ionm/z 1075 represents theloss of water from the y5 ion, andm/z 587 represents the y5 ionfollowing loss of the raloxifene moiety. In summary, theseresults confirmed that the primary sequence of the peptide is237-NICVFPR-243 and that raloxifene was adducted to specif-ically to Cys239.

The same protocol for P450 3A4 isolation and digestion byproteinase K was used to prepare P450 3A4-PIA, and the samplewas loaded onto a C18 column for peptide separation. Effectivedigestion was confirmed by the elution profile of peptides inthe UV 214 nm trace (Figure 4A). The UV 340 nm tracerevealed a single dominant peak at 21 min, representing apeptide with the pyrene moiety of PIA (Figure 4B). Additionally,the complete UV-vis spectrum of the peak, acquired with theonline diode array detector, was similar to the spectrum of PIAalone (data not shown). The mass of the peptide eluting at 21

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) SRMm/z 660-1092 trace.


min wasm/z606.0 (Figure 5A), and the SRMm/z606.0-379.0shows the aligned retention time in relation to the UVchromatogram (Figure 4C). Protein Prospector was used toidentify the alkylated peptide with this mass as 237-NIC-239

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

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


m/zvalue of 379.0, in agreement with the cysteine residue beingthe C-terminal residue of the observed peptide (data no shown).From these results it can be concluded that the same cysteineresidue at position 239 in P450 3A4 that is modified byraloxifene under turnover conditions is also assessable toalkylation by PIA.

Purified P450 3A4-IA was digested with trypsin, whichhydrolyzes the protein at specific residues and results inpredictable peptides. Without a chromophore on the alkylatingagent, modified peptides could not be traced following digestionwith proteinase K, which nonspecifically digests the protein.In preliminary experiments, peptides containing cysteine resi-dues, with or without alkylation, could not be detected afterdigestion of P450 3A4-IA by proteinase K. The only P450 3A4-IA peptide from the trypsin digest that displayed an increase of57.0 Da relative to control was peptide 213-FDFLDPFFLSITVF-PFLIPILEVLNICVFPR-243, which eluted at 29.6 min (data notshown). The sequence of the peptide was confirmed by MS2

experiments, and carbamidomethylation of Cys239 was verifiedby the mass increase of 57.0 Da in the y5 ion (m/z 678.3) butnot in the y4 ion (m/z 518.3).

Time-Dependent Inhibition by Raloxifene Is Prevented byPretreatment with PIA or IA. The protective effect of Cys239alkylation, by either PIA or IA, against raloxifene time-dependent inactivation was evaluated in a secondary reactionwith the probe substrate, BQ. As expected for a time-dependentinhibitor, the rate of inactivation is greater for the reactionscontaining 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 underidentical conditions, the control reaction itself displays anincrease in the initial rate constant (kobs ) 7.6× 10-2 ( 0.6×10-2 min-1) as compared to the non-alkylated P450 3A4preparation. Interestingly, the reaction with raloxifene does notsignificantly alter the initial rate constant (kobs ) 8.2× 10-2 (0.7 × 10-2 min-1) relative to its control reaction, indicatingthat 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 carbamidomethy-lation does not disrupt the stability of the protein. Yet, thisalkylation maintains the protective effect against raloxifene time-dependent inhibition (kobs ) 4.3× 10-2 ( 0.7× 10-2 min-1).Reactions were also conducted in which prereacted PIA or IAwith GSH was subsequently added to P450 3A4 following time-dependent inhibition measurement. The initial rate constantswere not significantly different from the control reactions,indicating that the quenched alkylating agents do not disruptsubstrate turnover (data not shown).

Functional Characterization of PIA- and IA-Treated P4503A4. To evaluate the proximity of the alkylated cysteine residuerelative to the heme, ligand-induced spin-state changes weremonitored using midazolam. The binding affinity for midazolamwas similar for untreated P450 3A4 (5.3( 0.2 µM) as for theP450 3A4 sample pretreated with PIA (4.4( 0.1 µM). Thedifference spectra revealed a lower spin-state change for thePIA-treated P450 3A4 as compared to the control spectra forsaturating concentrations of midazolam (Figure 6). This effectmay be due to slightly less intact P450 following alkylation(87% of control as determined by the ferrous CO differencespectra).

The relative production of glutathionyl-raloxifene and hy-droxyl-raloxifene by PIA- and IA-treated P450 3A4 wasmeasured to ensure that the alkylation procedure was actuallypreventing raloxifene adduction and not simply inhibitingturnover and therefore preventing the reactive intermediatesresponsible for protein adduction. Without standards for eithermetabolite, the analyte levels were normalized to the productionby unmodified P450 3A4. PIA-treated P450 3A4 producedsimilar amounts of the two metabolites (117( 9% glutathionyl-raloxifene and 118( 6% hydroxyl-raloxifene) as compared tothe control reaction (100( 8% glutathionyl-raloxifene, 100(5% hydroxyl-raloxifene) (Figure 7). Interestingly, P450 3A4pretreated with IA produced significantly more of each me-tabolite (200 ( 16% glutathionyl-raloxifene, 162( 18%

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


hydroxyl-raloxifene). Control reactions in which GSH wasreacted with IA or PIA for 1 h prior to incubation with P4503A4 demonstrated that alkylated GSH did not greatly influenceP450 3A4 turnover of raloxifene. Incubations in which NADPHwas omitted produced relatively little signal for the twometabolites monitored (6.3( 0.9% glutathionyl-raloxifene, 1.5( 0.2% hydroxyl-raloxifene). HLMs were also included in thisassay for comparison, and they demonstrated higher productionof the hydroxyl-raloxifene relative to glutathionyl-raloxifene(Figure 7).

Discussion

Direct characterization of the chemical intermediate(s) ad-ducted to P450 apoprotein and presumably responsible formechanism-based inhibition offers a distinct advantage overpredictions made based on the identity of GSH-drug adductsthat may or may not be related to the mechanism of inactivation.For example, raloxifene has been shown to form numerous GSHadducts from in vitro incubations (Scheme 1), yet it is unknownwhich if any of the GSH adducts represent the speciesresponsible for inactivating P450 3A4. Therefore, the raloxifeneP450 3A4 adduct was measured, and the deconvoluted massspectral results produced an adducted mass difference of 471Da, which is consistent with the adducting species arising fromthe diquinone methide intermediate (Scheme 1, Reaction A).

This measured difference in mass eliminates theo-quinone(Scheme 1, Reaction B) as the protein adducting species becausethe mass difference due to the raloxifene adduct would be 487Da. However, the diquinone methide intermediate yields severalelectrophilic sites that could be quenched by P450 3A4 wherebythe resultant protein adduct(s) would all produce a mass changeof 471 Da. Thus, to further distinguish the specific site ofraloxifene alkylation to P450 3A4 and provide insight toraloxifene P450 3A4 interactions a “bottom-up” analysis ap-proach was used in an attempt to define the regioselectivity ofalkylation.

Identification of inactivating species and residues within theactive site of P450 3A4 has been attempted previously usingmechanism-based inhibitors. Historically, radiolabeled materialis used to trace the adducted protein through numerousmanipulations including dialysis, enzymatic digests, and LC-MS conditions to facilitate the location of protein adducts.Despite the use of radiolabel tracers many of these attemptsfail speculatively due to factors including incomplete digestion,poor ionization of the hydrophobic peptide to which the adductis attached, complexity of the digestion mixtures, and lowstability of the resultant alkylated peptide species. For example,Lightning et al. incubated P450 3A4 with a potent mechanism-based inhibitor [14C]-L-754,394 in order to trace the alkylatedprotein through gel electrophoresis and subsequent digestion

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 absorbingpeak shown in Figure 4B. (B) MS2 spectrum ofm/z 660.0.


of the protein with CNBr in an attempt to identify the adductedpeptide(s). A single peptide fragment associated with theradioactivity was identified to be I257-X317 (X denotes ahomoserine lactone) based on MS data, although the drug adductwas never directly observed. The authors speculate that becausethe drug adduct was unstable during preparation for massanalysis the covalent linkage must be an ester bond with a Gluor Asp residue and tentatively assigned Glu307 on the I-helixas the site of attachment (5). Radiolabeled compound provideda tracer of the alkylation product through a series of experimentsto facilitate the tentative identification of Glu307, which can

be supported further with recent crystal structures that showthat Glu307 lies within the P450 3A4 active site (11, 12).

The GSH adducts of raloxifene suggest a high affinity of thebioactivated intermediates for ‘soft’ nucleophiles. Solvent-accessible cysteine residues were probed with cysteine-selectivealkylating agents, PIA and IA, to maximize the potential forlocating nucleophilic sites in P450 3A4. PIA was chosen forthe alkylating agent in an effort to increase the propensity foralkylation to occur near the enzyme active site given pyreneand several close analogs are known P450 3A4 substrates (13-15). Second, pyrene is fluorescent, which should facilitate the

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


location of any alkylated residues and peptides with UV andfluorescence detection. A 1:1 stoichiometry was producedbetween PIA and P450 3A4 as monitored by mass spectrometry(Figure 1C). The site of alkylation was then investigated byprotein digestion using proteinase K. Proteinase K was chosento digest the alkylated protein because this protease efficientlyand nonspecifically cleaves proteins into small peptides andamino acids. Digestion of protein using proteinase K shouldallow for the separation and identification of a drug-peptideadducts from a complex mixture based upon the distinct physicalchemical differences expected for hydrophobic drug-peptideadduct(s) compared to the mixture of smaller hydrophilic peptideand amino acid fragments (16). In fact, the proteinase K digestof the PIA-P450 3A4 adduct produced a hydrophobic peptideadduct distinctly separated from the other fragments (Figure 2).Despite the nonspecific nature of proteinase K cleavage onlyone detectable PIA-peptide fragment was produced from threeseparate experimental digests, suggesting this technique isreproducible and thus may be suitable for other adducts.Therefore, the same procedure was performed with the ralox-ifene-P450 3A4 and yielded a single alkylated peptide (Figure4). This signified that not only is a single site of P450 3A4alkylated by the diquinone methide intermediate but also thequenching of the intermediate occurred with a high degree ofregiospecificity given the detection of a single peak in the UVand MS chromatograms. These data are supported by data fromthe diquinone methide quenched with GSH, which produceresolved adduct peaks with chromatography, suggesting that ifmultiple peptide adducts formed they would separate in a similarfashion (6). Similar to the data from the intact mass spectra themass of the peptide adduct was 471 Da larger than the predictedmass for peptide 237-NICVFPR-243 alone (Figure 3). Inspectionof the adducted peptide by MS2 produced a fragmentm/z 506,which likely results from cleavage adjacent to the thioethermoiety and agrees with the spectrum previously acquired forthe reactive diquinone methide raloxifene metabolites trappedby GSH (6). Further structural information could not be inferredfrom MS2 data due to the ambiguous fragmentation of the

adducted peptide and is consistent with that observed for theGSH-raloxifene adducts with theo-quinone (6). These dataconfirmed the role of the diquinone methide as the intermediateresponsible for P450 3A4 inactivation and are congruent withthe short∼1 s lifetime previously measured for the diquinonemethide intermediate and with the reduced time-dependentinhibition of close structural analogs that cannot form thediquinone methide (7, 17).

Specific alkylation of Cys239 by both raloxifene and theiodoacetamide reagents provides a unique opportunity toexamine the mechanism of P450 3A4 inactivation by raloxifene.Pretreatment of P450 3A4 with PIA or IA should blockraloxifene adduct formation at Cys239 and, presumably, preventthe associated inactivation. Although the kinetics of 7-hydrox-yquinoline formation by P450 3A4-PIA indicated a higher basalrate of protein degradation as compared to the unmodifiedprotein, pretreatment with PIA prevented additional time-dependent inactivation due to raloxifene turnover. Similar levelsof hydroxyl-raloxifene and glutathionyl-raloxifene were detectedin incubations with P450 3A4-PIA as in the control reaction,confirming that raloxifene was still a substrate for the modifiedenzyme. A similar experiment was conducted with P450 3A4treated with IA, and again, pre-alkylation of the proteinprevented time-dependent inactivation by raloxifene. Unliketreatment with PIA, IA treatment alone did not destabilize P4503A4 as determined by the kinetics of BQ metabolism. Inaddition, the relative levels of both hydroxyl-raloxifene andglutathionyl-raloxifene were higher for P450 3A4-IA than theuntreated P450. Collectively, these experiments confirmed thatthe site of alkylation for the cysteine modifying reagents andthe raloxifene diquinone methide are the same and that adductionof Cys239 is responsible for the observed time-dependentinactivation. Moreover, the increased rate of P450 3A4 inactiva-tion by PIA treatment alone indicated that this reaction mayserve as an experimental model to further explore the relation-ship between protein-drug adducts and enzyme inactivation.

Multiple mechanisms for inactivation due to protein adductioncan be envisioned: (1) the drug adduct binds in the active site

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


and prevents additional turnover by blocking access to the hemecatalytic center, (2) the drug adduct binds to the surface of theenzyme and prevents critical interactions with P450-reductase,or (3) the drug adduct induces a conformational change in theprotein structure that results in disruption of necessary interac-tions for substrate binding or displacement of residues involvedin the catalytic cycle. The functional effects observed with IA-and PIA-treated P450 3A4 compared to raloxifene inactivationcan help differentiate between these mechanisms. Time-de-pendent inactivation experiments demonstrated that BQ, acommon P450 3A4 probe substrate, can still be metabolizedby the control PIA-treated P450 3A4, indicating that thesubstrate can still access the active site. Even more convincingly,there was similar turnover of raloxifene to both hydroxyl-raloxifene and glutathionyl-raloxifene by native P450 3A4 andthe PIA-treated P450 3A4. In addition, midazolam could stillinduce a type I spectral shift upon binding to PIA-treated P4503A4, and the ligand exhibited a similar affinity for the alkylatedenzyme as for the untreated P450 3A4. Given the multiplebinding sites of P450 3A4, Cyclosporin A binding was alsoconducted and produced similar spectral shifts with and withoutalkylation (data not shown). From the enzymatic data collectedin these experiments we cannot rule out the possibility of thealkylation interfering with redox partners. Yet, disruption ofP450-reductase binding and electron transfer is not anticipatedbased on the position of Cys239 in the solved X-ray crystalstructure, located between the G and G′ helices and buried inthe lipid membrane, and the significant amount of enzymaticturnover observed with the alkylated proteins. Therefore, theobserved enzymatic activities and biophysical properties of P4503A4-PIA as well as the location of the modified residue areconsistent with a mechanism involving a conformational changein protein structure, albeit additional experiments would berequired to confirm this hypothesis.

In summary, LC-MS analysis of both intact protein andproteolytic digests of P450 3A4, subsequent to incubation withraloxifene, revealed that a single equivalent of drug binds toCys239 of P450 3A4, formed from the reactive diquinonemethide intermediate. The mechanism for the observed time-dependent inactivation is likely protein unfolding to an inactiveconformation, analogous to the decreased stability of P450 3A4when Cys239 is alkylated by PIA. Furthermore, time-dependentinactivation, but not raloxifene turnover, can be prevented withpretreatment of P450 3A4 with PIA or IA, establishing a directrelationship between adduction of drug and reduction inenzymatic activity. This methodology for identifying the chemi-cal species responsible for covalent binding to P450 andsubsequent inactivation may show utility in the lead optimizationstage of drug development and be used to distinguish a differentmechanism 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.,

O’Donnell, J. P., Boer, J., and Harriman, S. P. (2005) A comprehensivelisting of bioactivation pathways of organic functional groups.Curr.Drug Metab. 6, 161-225.

(2) Monks, T. J., Lau, S. S., and Gillette, J. R. (1984) Diffusion of reactivemetabolites out of hepatocytes: studies with bromobenzene.J.Pharmacol. Exp. Ther. 228, 393-399.

(3) Zhou, S., Yung Chan, S., Cher Goh, B., Chan, E., Duan, W., Huang,M., and McLeod, H. L. (2005) Mechanism-based inhibition ofcytochrome P450 3A4 by therapeutic drugs.Clin. Pharmacokinet. 44,279-304.

(4) 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., andTrager, W. F. (1999) Electrospray ionization mass spectrometricanalysis of intact cytochrome P450: identification of tienilic acidadducts to P450 2C9.Biochemistry 38, 2312-2319.

(5) Lightning, L. K., Jones, J. P., Friedberg, T., Pritchard, M. P., Shou,M., Rushmore, T. H., and Trager, W. F. (2000) Mechanism-basedinactivation of cytochrome P450 3A4 by L-754,394.Biochemistry 39,4276-4287.

(6) 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-mediatedbioactivation of raloxifene: irreversible enzyme inhibition and thioladduct formation.Chem. Res. Toxicol. 15, 907-914.

(7) 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 toquinoids: potential toxic pathways via a diquinone methide ando-quinones.Chem. Res. Toxicol. 17, 879-888.

(8) Gillam, E. M., Baba, T., Kim, B. R., Ohmori, S., and Guengerich, F.P. (1993) Expression of modified human cytochrome P450 3A4 inEscherichia coli and purification and reconstitution of the enzyme.Arch. Biochem. Biophys. 305, 123-131.

(9) Holmans, P. L., Shet, M. S., Martin-Wixtrom, C. A., Fisher, C. W.,and Estabrook, R. W. (1994) The high-level expression in Escherichiacoli of the membrane-bound form of human and rat cytochrome b5and studies on their mechanism of function.Arch. Biochem. Biophys.312, 554-565.

(10) Rock, D., Rock, D., and Jones, J. P. (2001) Inexpensive purificationof P450 reductase and other proteins using 2′,5′-adenosine diphosphateagarose affinity columns.Protein Expr. Purif. 22, 82-83.

(11) 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) Crystalstructures of human cytochrome P450 3A4 bound to metyrapone andprogesterone.Science 305, 683-686.

(12) Yano, J. K., Wester, M. R., Schoch, G. A., Griffin, K. J., Stout, C.D., and Johnson, E. F. (2004) The structure of human microsomalcytochrome P450 3A4 determined by X-ray crystallography to 2.05-Aresolution.J. Biol. Chem. 279, 38091-38094.

(13) Shou, M., Grogan, J., Mancewicz, J. A., Krausz, K. W., Gonzalez, F.J., Gelboin, H. V., and Korzekwa, K. R. (1994) Activation ofCYP3A4: evidence for the simultaneous binding of two substrates ina cytochrome P450 active site.Biochemistry 33, 6450-6455.

(14) 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 theC-oxidative metabolism of 1-nitropyrene in human liver microsomalsamples.Cancer Res. 52, 6237-6243.

(15) Schrag, M. L., and Wienkers, L. C. (2000) Topological alteration ofthe CYP3A4 active site by the divalent cation Mg(2+). Drug Metab.Dispos. 28, 1198-1201.

(16) Manabe, S., Sassa, S., and Kappas, A. (1985) Hereditary tyrosinemia.Formation of succinylacetone-amino acid adducts.J. Exp. Med. 162,1060-1074.

(17) Liu, H., Liu, J., van Breemen, R. B., Thatcher, G. R., and Bolton, J.L. (2005) Bioactivation of the selective estrogen receptor modulatordesmethylated arzoxifene to quinoids: 4′-fluoro substitution preventsquinoid formation.Chem. Res. Toxicol. 18, 162-173.

TX700037E


Time-Dependent Inactivation of P450 3A4 by Raloxifene: Identification of Cys239 as the Site of...

Documents

Transcript of Time-Dependent Inactivation of P450 3A4 by Raloxifene: Identification of Cys239 as the Site of...