Use of a Multistaged Time-Dependent Inhibition Assay to Assess the Impact of Intestinal Metabolism on Drug-Drug Interaction Potential

Transcription

1 X/12/ $25.00 DRUG METABOLISM AND DISPOSITION Vol. 40, No. 3 Copyright 2012 by The American Society for Pharmacology and Experimental Therapeutics 43257/ DMD 40: , 2012 Use of a Multistaged Time-Dependent Inhibition Assay to Assess the Impact of Intestinal Metabolism on Drug-Drug Interaction Potential Michael Zientek and Deepak Dalvie Pharmacokinetics, Pharmacodynamics, and Drug Metabolism, Pfizer Inc., La Jolla, California Received October 7, 2011; accepted November 23, 2011 ABSTRACT: In early discovery, compounds are often eliminated because of their potential to undergo metabolic activation and/or cytochrome P450 time-dependent inactivation (TDI). The blockbuster drug raloxifene is an example of a compound that would have been eliminated in the current paradigm. Despite raloxifene s in vitro bioactivation and TDI of CYP3A4, it is well tolerated in patients with no drug-drug interactions. This discordance is attributed to its presystemic glucuronidation, thereby decreasing the amount of unchanged raloxifene available for CYP3A inactivation. The current study used raloxifene as a model to assess the effect of hepatic and intestinal glucuronidation on the kinetic parameters of CYP3A4 inactivation. Therefore, a simple multistaged time-dependent inactivation using UDP-glucuronosyltransferase-enabled and Introduction Most drug metabolism groups within pharmaceutical companies have well established screening paradigms in the early phases of drug discovery to assess the metabolic activation potential (as evaluated by formation of a glutathione adduct) or time-dependent inactivation (TDI) of major human cytochrome P450 (P450) enzymes by new chemical entities (Obach et al., 2007; Zientek et al., 2010; Zimmerlin et al., 2011). When the results of both these assessments are positive, that particular compound is most certainly dropped or redesigned to minimize these issues. However, does a positive signal in any of these assays warrant elimination of a compound? (Evista) is a good example of a compound that would have, in the current paradigm, provided a positive result in the bioactivation/tdi screening strategy and thus most likely would have been attrited or sent for chemical redesign. For raloxifene, data revealing its propensity to undergo CYP3A4-mediated metabolic activation, leading to the formation of glutathione and N-acetylcysteine conjugates (Chen et al., 2002; Yu et al., 2004) came after regulatory approval. Despite these observations of in vitro bioactivation of CYP3A4, raloxifene has been and still is very well tolerated in Article, publication date, and citation information can be found at -absent reactions was built to understand the impact of the gut metabolism on inactivation potential. The results of these experiments demonstrated a 2.7-fold change in inactivation efficiency of CYP3A4. Incorporation of these results into a simulated midazolam drug-drug interaction study showed very little change in the pharmacokinetic parameters of the victim drug. In contrast, the absence of glucuronidation resulted in a 4.1-fold increase in the area under the curve (AUC) of midazolam, when in the presence of raloxifene, hence providing an understanding of the impact of intestinal glucuronidation on raloxifene s time-dependent inhibition of CYP3A4 and also providing a validation of a simple in vitro experiment to assess the influence of gut metabolism on timedependent inhibitors at the discovery phase. patients. This disconnect between the safety of raloxifene and its in vitro bioactivation is attributed to its extrahepatic glucuronidation (Kemp et al., 2002), which has also been affirmed by studies in our laboratory (Dalvie et al., 2008). Chen et al. (2002) have also demonstrated that raloxifene is a potent time-dependent inhibitor of CYP3A4 with estimated K I and k inact values of 9.9 M and 0.16 min 1, respectively. This mechanismbased inactivation of the enzyme is possibly mediated through a reactive metabolite because the addition of GSH attenuates inactivation to a certain degree. Several studies demonstrating the modification of amino acids of the CYP3A4 (not CYP3A5) active site have also been published (Baer et al., 2007; Pearson et al., 2007). These studies suggested that inactivation by raloxifene occurs through alkylation of Cys239 by the diquinone methide intermediate. However, no significant drug-drug interactions due to inhibition of metabolism of CYP3A4 substrates have been reported, nor does the package insert provide an associated warning even though the K I and k inact values of raloxifene determined in the in vitro mechanism-based inhibition studies are similar to known mechanism-based inactivators of CYP3A4 that cause severe drug-drug interactions, therefore emphasizing the importance of dose and exposure (Zhou et al., 2005). The disparity between potent inactivation of CYP3A4 and the lack of drug-drug interactions could also be attributed to a decrease in raloxifene exposure due to presystemic glucuronidation. ABBREVIATIONS: TDI, time-dependent inactivation; P450, cytochrome P450; UDPGA, UDP-glucuronic acid; DDI, drug-drug interaction; HLM, human liver microsomes; UGT, UDP glucuronosyltransferase; AUC, area under the curve; PK, pharmacokinetic; HIM, human intestinal microsomes. 467

2 468 ZIENTEK AND DALVIE Time points: 0, 3, 5, 7.5, 10, 13.5, 20 minutes 5 X KM reporter substrate MBI + HIM UDPGA + NADPH MBI + HIM +UDPGA+ NADPH Addi on of HLM 5 min 15 min MBI + HIM NADPH - No UDPGA MBI + HIM + +NADPH Addi on of HLM FIG. 1. Schematic diagram of the influence of UGTenabled human intestinal microsomes on the CYP3A4 inactivation kinetics of raloxifene. MBI, mechanismbased inhibition. Because cases similar to raloxifene could occur in the drug discovery process, it was our aim to build these routes of metabolism into the bioactivation/tdi screening strategy. Therefore, our primary goal was to design an in vitro assay to assess the impact of non-p450 and/or extrahepatic metabolism on the efficiency of inactivation for compounds that can be cleared by routes other than microsomal P450- catalyzed oxidation, using raloxifene as a model compound. Experiments were performed to estimate the kinetic parameters (K I and k inact ) of CYP3A4 inactivation after its preincubation with human intestinal microsomes (HIM) in the presence or absence of UDPGA. The secondary objective of the study was to predict the potential drug-drug interaction of raloxifene with midazolam, a known substrate of CYP3A4, using modeling and simulation with the estimates of the K I and k inact values obtained from the in vitro studies. The results obtained from the DDI predictions were then compared with those observed in vivo. Materials and Methods Acetonitrile, glucose 6-phosphate, glucose-6-phosphate dehydrogenase from baker s yeast, 0.1 M magnesium chloride solution, midazolam, NADP, alamethicin, 1 M potassium phosphate dibasic solution, and 1 M potassium phosphate monobasic solution were purchased from Sigma-Aldrich (St. Louis, MO). Pooled human liver microsomes (HLM) were prepared under contract by BD Gentest (Woburn, MA). Aliquots from the individual preparations from 56 individual human livers were pooled on the basis of equivalent protein concentrations to yield a representative microsomal pool with a protein concentration of 20.4 mg/ml (determined using the BCA assay; Thermo Fisher Scientific, Waltham, MA). The source reported that UGT1A1 (estradiol 3-glucuronide) and CYP3A4 (testosterone 6 -hydroxylation) activity in HLM was 1200 and 5400 pmol mg 1 min 1, respectively. Pooled HIM from 10 donors prepared to 20 mg/ml by enterocyte scraping of the jejunum and duodenum sections were purchased from Celsis (Baltimore, MD). The source reported that UGT (7-hydroxycoumarin glucuronide) and CYP3A4 (testosterone 6 -hydroxylation) activity in HIM was 774 and 17.8 pmol mg 1 min 1, respectively. A 10 mm stock solution of raloxifene was prepared using dimethyl sulfoxide and acetonitrile as solvents, and this was used for all incubations. Alamethicin was dissolved in ethanol at a concentration of 20 mg/ml. All other materials were of the highest quality attainable. (Note that raloxifene is a hazardous substance and should be handled with care.) 5 X KM reporter substrate Time points: 0, 3, 5, 7.5, 10, 13.5, 20 minutes Inactivation Kinetics. The procedure to assess the effect of intestinal glucuronidation on the inactivation of CYP3A4 by raloxifene is depicted in Fig. 1. All inactivation studies were performed using a Tecan Genesis laboratory automated system and were conducted in three parts. Pooled HIM (0.6 mg/ml) were treated with alamethicin (0.015 mg/ml) in 0.1 M KH 2 PO 4 (ph 7.4) and allowed to sit on ice for 10 min. ( M) was added to this mixture, and the incubations were warmed to 37 C. After 5 min, the incubation was started by addition of either MgCl 2 (1 mm) in potassium phosphate buffer (100 mm, ph 7.4) (control) or an NADPH-regeneration system containing MgCl 2 (1 mm) in potassium phosphate buffer (100 mm, ph 7.4), NADP (1 mm), glucose 6-phosphate (5 mm) and glucose-6 phosphate dehydrogenase (1 U/ml) or a mixture of an NADPH-regeneration system and UDPGA (1.0 and 3.0 mm, respectively) to achieve a total incubation volume of 0.14 ml. A 15-min incubation with human intestinal microsomes was chosen to coincide with the clinical T max of 15 to 30 min observed for raloxifene and, in fact, may have a conservative effect on the inactivation kinetics. After the 15-min incubation, 0.06 ml of (1.33 mg/ml) pooled human liver microsomes was added to the tube containing the above mixture. The final concentration of protein in the incubation mixture was 1 mg/ml in a total reaction volume of 0.2 ml. An aliquot (0.015 ml) of the incubation mixture was withdrawn at 0, 3, 5, 7.5, 10, 13.5, and 20 min and added to a final incubation mixture containing 10 M midazolam (5 times the observed K m of midazolam), MgCl 2 (1 mm), and an NADPH-regeneration system (1 mm) in potassium phosphate buffer (100 mm, ph 7.4) at 37 C (volume ml). This tertiary reaction was quenched with ml of chilled acetonitrile after each time point. Inactivation kinetic parameters were determined using the nonlinear leastsquares method. To determine k obs values, the decrease in the natural logarithm of the activity over time was plotted for each raloxifene concentration, and k obs values were described as the negative slopes of the lines. Inactivation kinetic parameters K I and k inact were fitted with nonlinear regression using eq. 1 (Tudela et al., 1987; Silverman, 1988; Hollenberg et al., 2008): k obs k inact I / K I I (1) in which [I] represents the concentrations of inactivator in the inactivation preincubations, k obs represents the negative values of the slopes of the natural logarithm of the percentage activity remaining versus inactivation incubation time at various [I], k inact is the limit maximal inactivation rate constant as [I] 3 infinity, and K I is the inactivator concentration yielding 0.5 k inact. HPLC Analysis. Samples were analyzed in the multiple reaction monitoring mode using a Sciex API 4000 mass spectrometer (Applied Biosystems, Foster City, CA) with a Shimadzu LC-10AD pump (Shimadzu Inc., Kyoto,

3 IMPACT OF SEQUENTIAL METABOLISM ON TIME-DEPENDENT INHIBITION 469 Japan) and a CTC PAL autosampler (LEAP Technologies, Carrboro, NC). An Onyx Monolithic C mm column (Phenomenex, Torrance, CA) was used for separation with a mobile phase composition of 0.1% formic acid in water (A) and 0.1% formic acid in acetonitrile (B). The following gradient was used at a flow rate of 3000 l/min): 0.01 to 0.4 min (15% B), 1.35 to 1.49 min (90% B), and 1.5 min (15% B). The sample injection volume was 20 l, and the flow was split postcolumn with 1 ml/ min going to the mass spectrometer. The API 4000 mass spectrometer used an electrospray ionization-positive mode TurboIonSpray, with the following settings: for 1 -hydroxymidazolam, the metabolite was monitored with transition (m/z) and a retention time of 0.98 min, and for buspirone (internal standard) m/z and retention time of 0.95 min were used. DDI Modeling and Simulations. -midazolam DDI simulations were performed using Simcyp (Simcyp Limited, Sheffield, UK), a computerbased modeling tool, to estimate the metabolic clearance of drugs and the effects of metabolic interactions in humans using in vitro metabolism and inhibition data (Howgate et al., 2006). Simcyp uses fundamental scaling procedures to predict in vivo hepatic clearance from in vitro metabolism data. The metabolic clearance of both raloxifene and midazolam and the effect of raloxifene on midazolam pharmacokinetics were estimated using the healthy volunteer population (sim-healthy volunteers). All in vitro parameters for metabolism and inhibition were obtained either from published reports or from the inactivation experiments described above (Tables 1 and 2). Simcyp uses scaling factors to extrapolate from microsomal fractions to in vivo clearance similar to the methods published in the literature (Obach et al., 1997; Cubitt et al., 2009). With extrapolation to in vivo clearance from in vitro, a key scaling parameter for HIM and HLM is milligrams of microsomal protein per gram of tissue. The mean value used by Simcyp for HIM is 2978 mg of microsomal protein/total intestine (%coefficient of variation 30%) (Paine et al., 1997; Cubitt et al., 2009), whereas for HLM the mean number is mg of microsomal protein/g liver (%coefficient of variation 30%) with adjustments for age (Pelkonen, 1973; Houston, 1994; Iwatsubo et al., 1997a,b). In the absence of any information on active drug uptake into the enterocyte, the fraction unbound in the enterocytes (f u, gut ) is generally set at a default value of 1 (which assumes that there is insufficient time for plasma protein binding equilibrium or erythrocyte uptake before the drug is removed from the basolateral side of the enterocyte). However, it may also be set at f u, p which assumes that there is sufficient time for plasma protein or erythrocyte uptake. TABLE 1 Simcyp input parameters In the absence of any data regarding this parameter, the latter scenario was used, because preliminary simulations indicated that to recover the observed fraction of raloxifene escaping the first-pass metabolism of the intestinal wall (f g ), f u, gut needed to be set to f u, p. These data provided, through simulation, an evaluation of the sequential pathways of the gut and the liver on the metabolism of raloxifene, as well as those pathways influencing raloxifene s DDI potential. In accomplishing this, the in vitro inactivation kinetics data described above could then be substituted for the in vitro human intestinal clearance value by adjusting the fraction unbound in the enterocytes (f u, gut )in a minimal Q gut model (Rostami-Hodjegan and Tucker, 2002; Yang et al., 2007) to a value equal to zero (eq. 2). This method forced the fraction of raloxifene escaping the gut metabolism to 100%. A similar method was used by Galetin et al. (2007) in exploration of the maximal inhibition of CYP3A4 intestinal first-pass metabolism. Q gut f g (2) Q gut f u,gut CL u,int,g where f u, gut represents the fraction unbound in the enterocytes, CL u, int, g is the intestinal clearance, and Q gut is the nominal blood flow in the small intestine. Q gut is a hybrid function and can be further decomposed into permeability clearance (CL perm) and enterocytic blood flow (Q ent ). In the raloxifene pharmacokinetic validation simulations, a population of 25 healthy volunteers, ages 18 to 65 years, was used in the assessment. In each of the studies conducted, single dose and multiple dose, raloxifene was dosed in the fasted state at 60 mg daily. The oral clearance (CL p.o. ) was calculated by dose/area under the curve in units of liter per hour. The area under the simulated plasma concentration time curve profile after a single dose was calculated from time 0 to 120 h. The area under the curve for steady state was calculated from time 0 to 24 h after last dose on day 21. For the midazolam interaction studies, another population of 25 healthy volunteers, ages 18 to 65 years, was used. A 60-mg dose, daily, of raloxifene was used in this simulation until the steady-state plasma concentration was reached, regardless of the inclusion of the gut metabolism. Multiple scenarios were investigated (Table 2). These included midazolam administered alone, midazolam administered with raloxifene applying the UGT gut clearance of raloxifene, midazolam administered with raloxifene applying only the oxidative route of metabolism and the modified inactivation kinetics of raloxifene Parameter CL p.o. clinical and CL p.o. Simcyp CL HIM HLM Simcyp Source Dose: 60 mg q.d. Single and multiple dose Single and multiple dose Log P ACD prediction Blood/plasma ratio 1 1 Default value pk a Monoprotic base Monoprotic base ACD prediction (8.98 B, 8.67 A1, 9.56 A2 ) (8.98 B, 8.67 A1, 9.56 A2 ) f u, blood Physicians Desk Reference, 2005 f u, gut a Estimated f a Physicians Desk Reference, 2005 K a Calculated from 1/T max (Morello et al., 2003) Q gut (l/h) Predicted from Jeong et al., 2004 V ss (l/kg) Physicians Desk Reference, 2005 CL p.o. (l/h) 3087 N.A. Physicians Desk Reference, 2005 HIM UGT V max, pmol min 1 mg Dalvie et al., 2008 K m, M f u, mic Dalvie et al., 2008 HLM UGT V max, pmol min 1 mg Dalvie et al., 2008 K m, M f u, mic Dalvie et al., 2008 CYP3A4 V max, pmol min 1 mg Dalvie et al., 2008 K m, M f u, mic Dalvie et al., 2008 K I, M f u, mic Experimental k inact, min Experimental K i, M f u, mic Pearson et al., 2007 CL renal, l/h 0 0 Hochner-Celnikier, 1999 CYP3A5 K i, M f u, mic Pearson et al., 2007 ACD, Advanced Chemistry Development, Toronto, ON, Canada; B, base; A1, acid 1; A2, acid 2; N.A., not available. a f u, gut value set to f u, p (see Materials and Methods).

4 470 ZIENTEK AND DALVIE TABLE 2 Midazolam and raloxifene Simcyp input parameters for DDI simulations using the modified in vitro inactivation kinetics assay results in place of the observed gut metabolism kinetics Midazolam CL oxidative Simcyp Dose 5 mg (day 14) 60 mg q.d. for 21 days Log P Blood/plasma ratio pk a Zwitterionic (10.95 B, 6.2 A1 ) f u, blood f a f u, gut 1 0 a K a, l/h Q gut, l/h V ss, l/kg 1 47 CYP3A4 V max, pmol min 1 mg (1-OH) K m, M f u, mic 2.16 (1-OH) V max, pmol min 1 mg (4-OH) N.A. K m, M f u, mic 31.8 (4-OH) 1 N.A. K I, M f u, mic Monoprotic base (8.98 B, 8.67 A1, 9.56 A2 ) In vitro with UGT In vitro without UGT K i, M f u, mic CYP3A5 V max, pmol min 1 mg (1-OH) N.A. K m, M f u, mic 4.16 (1-OH) 1 N.A. V max, pmol min 1 mg (4-OH) N.A. K m, M f u, mic 38.4 (4-OH) 1 N.A K i M f u, mic CL renal, l/h k inact, min 1 In vitro with UGT In vitro without UGT B, base; A1, acid 1; A2, acid 2; N.A., not applicable. a F u, gut value set to zero to force the fraction (f g ) escaping the gut to 100%, thus removing the gut UGT elimination pathway. with UGT activity, and midazolam administered with raloxifene applying only the oxidative route of metabolism and the modified inactivation kinetics of raloxifene without UGT activity. In each scenario on day 14, midazolam was dosed as a single dose of 5 mg. The area under the curve of midazolam was then compared in the presence and absence of raloxifene (AUC i /AUC). Results Inactivaction Kinetics. The goal of the inactivation studies was to understand the raloxifene concentration effect on both the K I and k inact after raloxifene was preincubated with intestinal microsomes in the presence and absence of UDPGA. The expanded study using human intestinal and liver microsomes fortified with UDPGA provided K I and k inact values of 24.2 M and min 1, respectively (Table 2). If the ratio of k inact and K I (k inact /K I ) is an indicator of potency, then with UGT activity, the potency is equal to 7.4. The reaction without UDPGA and only active P450 due to the NADPH-regeneration system provided a K I of 7.96 M and a k inact of min 1, with an approximate potency of 19.8 (Tables 1 and 2). The inclusion of UDPGA and therefore activation of intestinal UGTs decreases the potency by approximately 2.7-fold (Fig. 2). The decrease in raloxifene concentration due to the preincubation with HIM and UDPGA was not measured directly but was assumed to be in accordance with the kinetic parameters acquired previously (Dalvie et al., 2008; Cubitt et al., 2009). Because raloxifene is a lipophilic amine, all data were adjusted for human intestinal and human liver microsomal binding when the inhibition kinetic parameters were used for each simulation (Table 1). K,Obs [] µm FIG. 2. Modified inhibition kinetics (K I and k inact ) determination experiment using intestinal microsomes in the presence ( ) and absence ( ) of UDPGA. DDI Modeling and Simulation. To validate the simulations of raloxifene pharmacokinetics, a comparison of clinical PK data, simulated PK data using CL p.o. obtained from the literature (Hochner- Celnikier, 1999; Morello et al., 2003; Physicians Desk Reference, 2005), and simulated PK data using in vitro data previously obtained by our laboratories was made (Table 1). The simulations were performed both as a single dose and multiple dose, and the results are found in Table 3. These data matched well with the published literature (Hochner-Celnikier, 1999; Morello et al., 2003; Physicians Desk Reference, 2005). Once satisfied that the in vitro data could mimic the clinical PK parameters for raloxifene, multiple simulation scenarios were investigated to understand the pathways that would lead to the drug-drug interaction potential of raloxifene (Table 2). One of the scenarios predicts the accumulation of raloxifene by adjusting the fraction unbound in the gut to zero, thus forcing raloxifene gut clearance to zero and then using the modified K I and k inact data in the absence and presence of cofactor (UDPGA) for UGT activity (Fig. 3). The accumulation was a greater than 7-fold increase in maximal concentration (C max ) and AUC, and the accumulation still increased in the absence of UDPGA. When applied to a midazolam DDI prediction in Simcyp, the simulated midazolam plasma concentration profiles for each scenario from the materials and methods are depicted in Fig. 4, and results are summarized in Table 4. In short, the simulation of raloxifene with the addition of the route of gut metabolism showed very little change in the PK parameters of midazolam. A median 1.3-fold change in the C max i /C max and the AUC i /AUC ratios was found. This 30% increase in midazolam AUC supports the lack of a reported clinical drug-drug interaction when it was coadministered with raloxifene. When the UGT gut metabolism was substituted with TABLE 3 Predicted median pharmacokinetic parameters from a single-dose and 14-day multiple dose simulation of 60 mg of raloxifene q.d. compared with the clinically derived parameters Source for data: Hochner-Celnikier (1999); Morello et al. (2003); Physicians Desk Reference (2005). Parameter Single Dose Multiple Dose CL p.o. clinical CL p.o. Simcyp CL p.o clinical CL p.o. Simcyp CL HIM HLM Simcyp Duration of study, days 5 14 T max,h 0.5 a 0.7 C max, g/l 0.5 a a AUC 0, g 27.2 a a h 1 l 1 CL/F, l/h 3087 a a a Data matched well compared with the published literature.

5 IMPACT OF SEQUENTIAL METABOLISM ON TIME-DEPENDENT INHIBITION 471 FIG. 3. Simulated plasma concentration time profile derived using in vitro data of a multiple 60-mg dose of raloxifene with (light gray) and without (dark gray) gut metabolism. the modified K I and k inact incorporating the UGT gut metabolism preincubation, there too only a small increase in interaction was observed. A 1.4- and 1.7-fold change, respectively, was observed in the C max i /C max and the AUC i /AUC ratios. Although when the UGT gut metabolism was substituted with the modified K I and k inact in the absence of UGT gut metabolism preincubation, substantial increases in both the C max i /C max and the AUC i /AUC ratios were seen (2.7- and 4.1-fold, respectively). Discussion Many attempts to perform in vitro-in vivo extrapolation fail to address the impact of drug metabolism by enzymes in the gut wall, as part of the first-pass effect in concert with hepatic metabolism. One of our aims of our work was to identify these routes of metabolism by building on the mathematical and in silico successes of the past. These successes could then be applied to a relatively simple inhibition experiment to identify the impact of the mode of metabolism early in the discovery effort. A tool, Simcyp, was used to aid in placing the in vitro data in the context of the in vivo outcome. It is well established in the literature that the premise and procedure of Simcyp simulations to predict in vivo clearance from in vitro data are valid (Houston, 1994; Howgate et al., 2006; Rakhit et al., 2008; Jamei et al., 2009). It has also well established that in vitro inhibition data can help to predict in vivo DDI outcomes (Mayhew et al., 2000; Yang et al., 2005; Obach et al., 2006; Fahmi et al., 2008). With this confidence in the models and the in vitro data, a good assessment was thought to be possible and provided the means in which to validate the in vitro model. Based on preliminary work conducted by Kemp et al. (2002) identifying specific gut UGTs being responsible for a metabolizing raloxifene, our laboratories showed that the disparity observed between the in vitro bioactivation liability of raloxifene and its good safety record in humans was explained with a series of experiments (Dalvie et al., 2008). In those experiments, substantial proof of raloxifene s rapid detoxification via intestinal glucuronidation in vivo was displayed. In our current work, an effort was made to identify the implication of these sequential events in the detoxication of raloxifene; thus, we have broadened the accepted method for understanding inactivation kinetics. The process consists of multistage preincubation, first in human intestinal microsomes with and without UDPGA, and then second in human liver microsomes in the presence of the P450 cofactor NADPH. This multistage assay is based on well established practices of determining K I and k inact kinetic parameters. These modified studies provide an understanding of the influence that hepatic and nonhepatic UGTs may have in discriminating against rather potent time-dependent inhibitors. The 2.7-fold decrease in potency in the presence of the UGT cofactor UDPGA provides evidence along with what is known about the clinical application of raloxifene that such phenomena do occur and do buffer potential risk by affecting the exposure of the perpetrator drug. The results of both the UGT-enabled and the UGT-absent reactions provided a measure of the impact of intestinal UGTs on raloxifene s time-dependent inhibition of CYP3A4. In this case, the majority of the UGT metabolism was attributed to intestine-specific UGT1A8 and UGT1A10 and to a far lesser extent to the liver and intestinal enzyme UGT1A1 (Dalvie et al., 2008). From our previous work, it was observed that liver UGT1A1 had minimal impact on the clearance of raloxifene and thus concentration. This observed tissue-specific contribution may not always be the case for newly tested compounds and, therefore, should be taken into account when an in vitro study is designed. To provide a validation of the in vitro system and its applicability in vivo, the impact of gut metabolism on raloxifene s mechanismbased inhibition of CYP3A4 was further investigated for both the UGT-enabled and the UGT-absent reactions using Simcyp simulation studies. First, to validate the raloxifene pharmacokinetic profile, raloxifene pharmacokinetics were compared using previously published human liver microsomal and human intestinal in vitro kinetics clearance data from our laboratory and benchmarked against clinical pharmacokinetic parameters (CL p.o., AUC, C max, and T max ) obtained from the literature (Table 1) (Hochner-Celnikier, 1999; Morello et al., 2003; Physicians Desk Reference, 2005; Dalvie et al., 2008). These validations were compared with clinical data for both single dose and multiple doses of raloxifene at 60 mg q.d. Once satisfied with the in vitro to in vivo correlation of the model for raloxifene, the goal was to use the in vitro inactivation data with human intestinal microsomes as a surrogate for gut metabolism so that the Simcyp simulations could provide an understanding of the impact of intestinal UGTs on raloxifene s time-dependent inhibition of CYP3A4. The accumulation ) Systemic n (mg/ml) idazolam S ncentration Mi Con Time (h) FIG. 4. Mean plasma concentration profile of a single dose of midazolam on day 14 of a 21-day study when coadministered with a 60-mg dose q.d. of raloxifene dosed to steady state. Each of the four profiles denotes an influence of the presence or absence of gut metabolism on systemic raloxifene concentration and thus its influence on midazolam systemic concentration. Represented in blue is the mean concentration profile of midazolam without coadministration, the red trace represents the mean midazolam plasma concentration coadministered with raloxifene with gut kinetic parameters, the purple trace represents use of the modified K I and k inact with intestinal microsomes in the presence of UDPGA (in vitro with UGT), and the green trace is use utilization of the modified K I and k inact with intestinal microsomes in the absence of UDPGA (in vitro without UGT).

6 472 ZIENTEK AND DALVIE TABLE 4 Predicted median pharmacokinetic parameters from a 21-day study of 60 mg of raloxifene q.d. dosed to steady state with a single 5-mg dose of midazolam dosed on day 14 PK Profile: Midazolam CL oxidative Simcyp CL HIM HLM Simcyp Midazolam Interaction Study CL oxidative Simcyp With UGT K I /k inact Without UGT K I /k inact T max,h C max, g/l AUC 0 24 h, g h 1 l 1 Midazolam T max,h C max, g/l AUC 0 24 h, g h 1 l 1 CL/F, l/h C max i /C max AUC i /AUC of raloxifene was substantial in the presence and absence of UDPGA when the fraction escaping the gut was set to 100% (Table 4). However, when the data were applied in the Simcyp midazolam DDI modeling, virtually no change in either midazolam AUC or C max in the presence of the HIM with UDPGA-activated gut UGTs was observed. This lack of effect was due to the in vitro influence of the gut metabolism on the inactivation kinetics (Table 4). In contrast, there was a greater than 4-fold median increase in midazolam exposure (AUC and C max ) when HIM were not supplemented with UDPGA (Table 4). The lack of interactions predicted from the model supported the absence of drug-drug interactions by raloxifene in the clinic and also the potential an analogous phenolic compound could have in the absence of gut-mediated metabolism. Of interest were some observations of the simulated patient population used by Simcyp. There were three subjects with much larger simulated interactions occurring in both the UGT-enabled and UGT-absent conditions when coadministered with midazolam. When investigated, the overarching reasons were two characteristics, one being the degradation constant (k deg ) assigned to these subjects and the other being the genotype of the polymorphic enzyme CYP3A5. The accepted degradation half-life for CYP3A4 is 70 h in Simcyp version 8, equating to a degradation rate constant (k deg ) of approximately h 1. The variation seen in these subjects ranged from to , which equates to approximately a 312- to 113-h half-life, thus pushing out the recovery of the enzyme and increasing the observed interaction. Several laboratories have recently observed that the enzyme degradation half-life of approximately 24 to 36 h of CYP3A4 is a more clinically relevant value to extrapolate in vitro to clinical DDI prediction (Wang, 2010; Rowland Yeo et al., 2011). When a 36-h CYP3A4 half-life was implemented in these simulations a 2-fold median (3-fold mean) drug-drug interaction with midazolam was observed (data not shown), thus still providing a prediction of a DDI. In addition, because midazolam has been reported to be metabolized by both CYP3A4 and CYP3A5, the genotype of each donor was investigated, and the three outlier subjects were of the CYP3A5 *3/*3 genotype. This genotype does not provide active CYP3A5 enzyme to assist in the metabolism of midazolam; thus, the fraction is metabolized completely by CYP3A4, which puts the subject at higher risk for a DDI. Also interesting were subjects who had instances of long CYP3A4 enzyme half-lives but had the CYP3A5 *1/*1 genotype, and abundance of active CYP3A5 showed a less severe interaction than the median interaction. This CYP3A5 genotype phenomenon would not have been evident if raloxifene was an equally potent inhibitor for both CYP3A4 and CYP3A5, which has been reported to not be true (Pearson et al., 2007). Similar to the in vitro assessment of raloxifene, ezetimibe has been invested by a group at XenoTech, LLC (Lenexa, KS) using human liver microsomes and NADPH and in the presence and absence of UDPGA (Parkinson et al., 2010). Ezetimibe (a cholesterol absorption inhibitor) was found to show a significant time-dependent inhibition of CYP3A4 in the presence of NADPH and to a much lesser extent when UDPGA was included. These findings are supported in the clinic where ezetimibe, when coadministered with CYP3A4-metabolized drugs, does not show a significant change in AUC and/or C max of those coadministered drugs (Kosoglou et al., 2005) and are further supported by ezetimibe being routinely coadministered with a statin (cholesterol synthesis inhibitor) such as simvastatin, a CYP3A4 substrate, without causing an observed DDI. The hypothesis is that ezetimibe is extensively cleared in the liver and intestine by UGT1A1, 1A3, and 2B15, thus minimizing the exposure to CYP3A4. Our assessment differs from the ezetimibe example in that our work demonstrates the impact of a nonhepatic influence on TDI; therefore, a sequential instead of a parallel process is needed to be considered in vitro and during modeling. The study presented here is another example supporting the use of a combination of modeling and simulation software and in vitro tools for better understanding of the implications of multiple pathways on the drug-drug interaction potential. The process has allowed for our proof of concept in vitro experiments to be affirmed, compared with clinical and in vitro derived clinical simulations. Therefore, a simple in vitro assessment can be established for future potent time-dependent inhibitors, for which in vitro data have revealed a potential issue with the compound and extensive characterization both clinically and in vitro has not been achieved, but the compound still shows promise in pharmacology models. Acknowledgments We thank Ying Jiang for bioanalytical assistance and Drs. Karen R. Yeo and R. Scott Obach for constructive feedback on this work. Authorship Contributions Participated in research design: Zientek and Dalvie. Conducted experiments: Zientek and Dalvie. Performed data analysis: Zientek and Dalvie. Wrote or contributed to the writing of the manuscript: Zientek and Dalvie. References Baer BR, Wienkers LC, and Rock DA (2007) Time-dependent inactivation of P450 3A4 by raloxifene: identification of Cys239 as the site of apoprotein alkylation. Chem Res Toxicol 20: Chen Q, Ngui JS, Doss GA, Wang RW, Cai X, DiNinno FP, Blizzard TA, Hammond ML, Stearns RA, Evans DC, et al. (2002) Cytochrome P450 3A4-mediated bioactivation of raloxifene: irreversible enzyme inhibition and thiol adduct formation. Chem Res Toxicol 15: Cubitt HE, Houston JB, and Galetin A (2009) Relative importance of intestinal and hepatic glucuronidation-impact on the prediction of drug clearance. Pharm Res 26: Dalvie D, Kang P, Zientek M, Xiang C, Zhou S, and Obach RS (2008) Effect of intestinal glucuronidation in limiting hepatic exposure and bioactivation of raloxifene in humans and rats. Chem Res Toxicol 21: Fahmi OA, Maurer TS, Kish M, Cardenas E, Boldt S, and Nettleton D (2008) A combined model for predicting CYP3A4 clinical net drug-drug interaction based on CYP3A4 inhibition, inactivation, and induction determined in vitro. Drug Metab Dispos 36: Galetin A, Hinton LK, Burt H, Obach RS, and Houston JB (2007) Maximal inhibition of intestinal first-pass metabolism as a pragmatic indicator of intestinal contribution to the drug-drug interactions for CYP3A4 cleared drugs. Curr Drug Metab 8: Hochner-Celnikier D (1999) Pharmacokinetics of raloxifene and its clinical application. Eur J Obstet Gynecol Reprod Biol 85: Hollenberg PF, Kent UM, and Bumpus NN (2008) Mechanism-based inactivation of human cytochromes p450s: experimental characterization, reactive intermediates, and clinical implications. Chem Res Toxicol 21:

7 IMPACT OF SEQUENTIAL METABOLISM ON TIME-DEPENDENT INHIBITION 473 Houston JB (1994) Utility of in vitro drug metabolism data in predicting in vivo metabolic clearance. Biochem Pharmacol 47: Howgate EM, Rowland Yeo K, Proctor NJ, Tucker GT, and Rostami-Hodjegan A (2006) Prediction of in vivo drug clearance from in vitro data. I: impact of inter-individual variability. Xenobiotica 36: Iwatsubo T, Hirota N, Ooie T, Suzuki H, Shimada N, Chiba K, Ishizaki T, Green CE, Tyson CA, and Sugiyama Y (1997a) Prediction of in vivo drug metabolism in the human liver from in vitro metabolism data. Pharmacol Ther 73: Iwatsubo T, Suzuki H, Shimada N, Chiba K, Ishizaki T, Green CE, Tyson CA, Yokoi T, Kamataki T, and Sugiyama Y (1997b) Prediction of in vivo hepatic metabolic clearance of YM796 from in vitro data by use of human liver microsomes and recombinant P-450 isozymes. J Pharmacol Exp Ther 282: Jamei M, Marciniak S, Feng K, Barnett A, Tucker G, and Rostami-Hodjegan A (2009) The Simcyp population-based ADME simulator. Expert Opin Drug Metab Toxicol 5: Jeong EJ, Lin H, and Hu M (2004) Disposition mechanisms of raloxifene in the human intestinal Caco-2 model. J Pharmacol Exp Ther 310: Kemp DC, Fan PW, and Stevens JC (2002) Characterization of raloxifene glucuronidation in vitro: contribution of intestinal metabolism to presystemic clearance. Drug Metab Dispos 30: Kosoglou T, Statkevich P, Johnson-Levonas AO, Paolini JF, Bergman AJ, and Alton KB (2005) Ezetimibe: a review of its metabolism, pharmacokinetics and drug interactions. Clin Pharmacokinet 44: Mayhew BS, Jones DR, and Hall SD (2000) An in vitro model for predicting in vivo inhibition of cytochrome P450 3A4 by metabolic intermediate complex formation. Drug Metab Dispos 28: Morello KC, Wurz GT, and DeGregorio MW (2003) Pharmacokinetics of selective estrogen receptor modulators. Clin Pharmacokinet 42: Obach RS, Baxter JG, Liston TE, Silber BM, Jones BC, MacIntyre F, Rance DJ, and Wastall P (1997) The prediction of human pharmacokinetic parameters from preclinical and in vitro metabolism data. J Pharmacol Exp Ther 283: Obach RS, Walsky RL, and Venkatakrishnan K (2007) Mechanism-based inactivation of human cytochrome p450 enzymes and the prediction of drug-drug interactions. Drug Metab Dispos 35: Obach RS, Walsky RL, Venkatakrishnan K, Gaman EA, Houston JB, and Tremaine LM (2006) The utility of in vitro cytochrome P450 inhibition data in the prediction of drug-drug interactions. J Pharmacol Exp Ther 316: Paine MF, Khalighi M, Fisher JM, Shen DD, Kunze KL, Marsh CL, Perkins JD, and Thummel KE (1997) Characterization of interintestinal and intraintestinal variations in human CYP3Adependent metabolism. J Pharmacol Exp Ther 283: Parkinson A, Kazmi F, Buckley DB, Yerino P, Ogilvie BW, and Paris BL (2010) Systemdependent outcomes during the evaluation of drug candidates as inhibitors of cytochrome P450 (CYP) and uridine diphosphate glucuronosyltransferase (UGT) enzymes: human hepatocytes versus liver microsomes versus recombinant enzymes. Drug Metab Pharmacokinet 25: Pearson JT, Wahlstrom JL, Dickmann LJ, Kumar S, Halpert JR, Wienkers LC, Foti RS, and Rock DA (2007) Differential time-dependent inactivation of P450 3A4 and P450 3A5 by raloxifene: a key role for C239 in quenching reactive intermediates. Chem Res Toxicol 20: Pelkonen O (1973) Drug metabolism in the human fetal liver. Relationship to fetal age. Arch Int Pharmacodyn Ther 202: Physicians Desk Reference (2005) Evista (raloxifene hydrochloride), in Physicians Desk Reference, 59th ed, Thomson Healthcare, Montvale, NJ. Rakhit A, Pantze MP, Fettner S, Jones HM, Charoin JE, Riek M, Lum BL, and Hamilton M (2008) The effects of CYP3A4 inhibition on erlotinib pharmacokinetics: computer-based simulation (SimCYP) predicts in vivo metabolic inhibition. Eur J Clin Pharmacol 64: Rostami-Hodjegan A and Tucker GT (2002) The effects of portal shunts on intestinal cytochrome P450 3A activity. Hepatology 35: ; author reply Rowland Yeo K, Walsky RL, Jamei M, Rostami-Hodjegan A, and Tucker GT (2011) Prediction of time-dependent CYP3A4 drug-drug interactions by physiologically based pharmacokinetic modelling: impact of inactivation parameters and enzyme turnover. Eur J Pharm Sci 43: Silverman RB (1988) Mechanism-Based Enzyme Inactivation: Chemistry and Biology, pp 3 30, CRC Press, Boca Raton, FL. Tudela J, García Cánovas F, Varón R, García Carmona F, Gálvez J, and Lozano JA (1987) Transient-phase kinetics of enzyme inactivation induced by suicide substrates. Biochim Biophys Acta 912: Wang YH (2010) Confidence assessment of the Simcyp time-based approach and a static mathematical model in predicting clinical drug-drug interactions for mechanism-based CYP3A inhibitors. Drug Metab Dispos 38: Yang J, Jamei M, Yeo KR, Tucker GT, and Rostami-Hodjegan A (2005) Kinetic values for mechanism-based enzyme inhibition: assessing the bias introduced by the conventional experimental protocol. Eur J Pharm Sci 26: Yang J, Jamei M, Yeo KR, Tucker GT, and Rostami-Hodjegan A (2007) Prediction of intestinal first-pass drug metabolism. Curr Drug Metab 8: Yu L, Liu H, Li W, Zhang F, Luckie C, van Breemen RB, Thatcher GR, and Bolton JL (2004) Oxidation of raloxifene to quinoids: potential toxic pathways via a diquinone methide and o-quinones. Chem Res Toxicol 17: Zhou S, Chan E, Li X, and Huang M (2005) Clinical outcomes and management of mechanismbased inhibition of cytochrome P450 3A4. Ther Clin Risk Manag 1:3 13. Zientek M, Stoner C, Ayscue R, Klug-McLeod J, Jiang Y, West M, Collins C, and Ekins S (2010) Integrated in silico-in vitro strategy for addressing cytochrome P450 3A4 time-dependent inhibition. Chem Res Toxicol 23: Zimmerlin A, Trunzer M, and Faller B (2011) CYP3A time-dependent inhibition risk assessment validated with 400 reference drugs. Drug Metab Dispos 39: Address correspondence to: Dr. Michael Zientek, Pfizer, Inc., Science Center Dr., La Jolla, CA. michael.zientek@pfizer.com