Introduction

Top Abstract Introduction Materials and Methods Results Discussion References

In the early stages of the drug discovery process, compounds are subjected to a battery of screens to maximize the success rate in clinical trials (Lin and Lu, 1997; Bajpai and Adkison, 2000; White, 2000). Cytochromes P450 (P450s¹), a superfamily of drug-metabolizing enzymes, are recognized to play an important role in clinically relevant drug-drug interactions, as a result of which new chemical entities (NCEs) typically are subjected to high-throughput P450 inhibition screening (Lin and Lu, 1997; Crespi, 1999; Moody et al., 1999; Rodrigues et al., 2001). In vitro studies that use human liver microsomes containing multiple P450s or recombinant enzymes are widely used for this purpose, and fully automated high-throughput assays, such as those that use LC-MS/MS with short run times (Ayrton et al., 1998; Yin et al., 1999; Chu et al., 2000) or 96-well microtiter plates with fluorescent probes (Chauret et al., 1999; Crespi et al., 1999), have been developed. Because of the broad substrate specificity of CYP3A4, which catalyzes the metabolism of a wide spectrum of therapeutic agents, this isoform has been the focus of many screening efforts. However, recent studies have shown a poor correlation in inhibition data obtained with different CYP3A4 substrates. Moreover, many substrates are not mutually competitive, thereby making this simple approach unreliable (Kenworthy et al., 1999; Wang et al., 2000a,b). Therefore, in vitro CYP3A4 inhibition screening should be performed with caution.

CYP3A4 metabolizes structurally diverse substrates with a wide range of molecular size, shape, enzyme affinity, and turnover number. Unusual kinetic interactions involving CYP3A4 have been found to include sigmoidicity or substrate activation (Ueng et al., 1997; Shou et al., 1999), partial inhibition (Wang et al., 1997, 2000a; Kenworthy et al., 1999), substrate inhibition (Tang et al., 2000; Lin et al., 2001; Schrag and Wienkers, 2001), activation (Shou et al., 1994; Domanski et al., 1998; Ngui et al., 2000), and differential kinetics (Shou et al., 2001). These phenomena are thought to be due to the binding of multiple substrates within the active site of the enzyme (Shou et al., 1994, 2001; Koley et al., 1995; He et al., 1997; Harlow and Halpert, 1998; Korzekwa et al., 1998; Hosea et al., 2000). As a result, the interaction of an inhibitor with one CYP3A4 probe may differ from that observed with other CYP3A4 substrates. The atypical interactions observed in vitro can be attributed to differential binding and steric restriction for interacting agents (substrates and inhibitors), resulting in both substrate- and inhibitor-dependent inhibition. This may lead to the erroneous evaluation of drug interaction potential. In the present study, we used testosterone, 7-benzyloxyquinoline (BQ), and 7-benzyloxy-4-trifluoromethyl-coumarin (BFC), commonly used probes for CYP3A4 inhibition screening, to elucidate the mechanism of kinetic interaction of any two substrates using three-dimensional fitting of a two-binding site model. Our results demonstrate that the three CYP3A4 substrates can bind to different domains at the active site of CYP3A4 and that any pair can interact with each other in a manner that is distinct from simple competitive inhibition. These results show that the use of more than one substrate may be required to accurately predict the potential for drug-drug interactions that occur at the level of CYP3A4 inhibition.

	Materials and Methods

Top Abstract Introduction Materials and Methods Results Discussion References

Chemicals. Testosterone, 6-hydroxytestosterone, cortisone, and NADPH were purchased from Sigma (St. Louis, MO). BQ, BFC, and their metabolites were obtained from Ultrafine Chemicals (Manchester, England, UK).

CYP3A4 Expression in Baculovirus. Plasmids pGem-7/CYP3A4 and pGem-3Z/oxidoreductase (OR) containing the full-length cDNAs for CYP3A4 and OR (15) were provided by Dr. Frank J. Gonzalez (National Cancer Institute, Bethesda, MD). The entire coding region of each cDNA was excised from the vectors and inserted into baculovirus shuttle vector, pBlueBac (CYP3A4, XbaI/KpnI and OR, EcoRI). Recombinant virus was constructed according to the manufacturer's procedure (Invitrogen, Carlsbad, CA). CYP3A4 and OR proteins were coexpressed in Sf21 insect cells (Invitrogen). The multiplicity of infection ratio of viruses encoding CYP3A4 and OR was 1:0.1. Total P450 content was measured by the CO-difference spectrum and the molar ratio of CYP3A4 to OR in microsomal preparation was 1:2.8. The specific activity of CYP3A4 in testosterone 6-hydroxylation was 42 nmol/min, nmol.

Metabolism of CYP3A4 Substrates. Incubations of each substrate pair (testosterone-BFC, BQ-testosterone, and BQ-BFC) were performed using 96-well microtiter plates. The incubation mixture (usually a 0.2-ml final volume) consisted of 5 pmol of microsomal CYP3A4 protein, 0.1 M potassium phosphate buffer (KP_i, pH = 7.4), and substrate in the presence or absence of the second substrate/effector. The concentrations of the two substrates varied from 3.1 to 100 µM in a total of 64 incubations. The incubation mixture was preincubated for 5 min at 37°C in a shaking water bath, and the reaction was initiated by the addition of 1 mM NADPH and incubated further for 15 min. The reactions were terminated by the addition of 2 volumes of acetonitrile containing the appropriate internal standards (cortisone for testosterone metabolism and 7-OH-4-methylcoumarin for BQ and BFC metabolism). Samples were centrifuged at 4000 rpm for 10 min, and supernatants were divided into two aliquots for metabolite quantitation from each substrate. Standard curves for the metabolites of interest were prepared using the corresponding authentic reference materials. The consumption of each substrate at all concentrations studied was limited to less than 10%, and the rates of product formation were expressed as nanomoles of product formed per minute per nanomoles of P450.

HPLC Analysis. HPLC was performed using an HP1100 liquid chromatograph equipped with an HP model 1100 autosampler, a ternary solvent delivery system, and a fluorescence detector controlled by the Hewlett-Packard HPLC^2D ChemStation software. Metabolites of BQ, BFC, or both were separated on a Zorbax SB-C₁₈ column (5 µm, 4.6 mm × 15 cm; MAC-MOD Analytical, Chadds Ford, PA), eluted using a mobile phase with a linear aqueous acetonitrile gradient (6 min from 15-40% and then 4 min from 40-90% acetonitrile in water containing 20 mM NH₄Ac) at a flow rate of 1 ml/min. Samples were detected at emission and excitation wavelengths of 420 nm and 503 nm, respectively. Retention times of BFC and BQ metabolites and of internal standard were 10, 5.3, and 7.2 min, respectively. The metabolites were quantified by normalization against the internal standard.

LC-MS/MS. Quantitative analysis of testosterone and its metabolite, 6-hydroxytestosterone, was achieved using a PerkinElmer HPLC system, comprised of a Series 200 LC pump and LEAP autosampler coupled to a Sciex API 2000 mass spectrometer operated in the positive ion mode. Aliquots (20 µl) of incubation mixtures were chromatographed on a BDS hypersil C₈ column (5-µm particle; 2 × 50 mm; Keystone Scientific, Inc., Bellefonte, PA) with a mobile phase consisting of solvent A (90% water/10% MeOH/0.1% trifluoroacetic acid) and solvent B (10% water/90% acetonitrile/0.1% trifluoroacetic acid). Samples were eluted using a 1.9-min linear gradient from 100% solvent A to 50% solvent B at a flow rate of 1.5 ml/min. Metabolite and internal standard were detected under atmospheric pressure chemical ionization conditions and identified using the transitions m/z 305.5  269 (6-OH-testosterone) and m/z 361  163.2 (internal standard). The rate of 6-hydroxytestosterone formation was determined according to a standard curve generated concurrently.

Kinetic Model for Drug-Drug Interaction. According to the observed changes in apparent K_m and V_max in the presence of the two substrates (Tables 1-3), we proposed the two-site model (Scheme 1) that states that the two different substrates bind specifically to their respective domains at the active site to form S₂E, ES₁, and S₂ES₁ complexes, respectively, which break down to product (P₁ or P₂). Since the two substrates within the active site interact kinetically, the dissociation (K_S) and rate constants (k_p) for the doubly bound enzyme complex (S₂ES₁) may no longer resemble those of the singly bound forms (S₂E and ES₁), such that the greatest changes in apparent kinetic constants are reflected by factors a, , and , as designated in the model (Scheme 1) when [E]_total = [S₂ES₁]. The kinetic scheme for an enzyme with two binding sites where the two substrates interact is depicted in Scheme 1, whereas the corresponding velocity equations (Eqs. 1 and 2) express the kinetic relationship during the course of metabolism. The definition of kinetic parameters in the equations is described in Scheme 1. We tested the hypothesis that the derived equations provide a reasonable fit to the experimental data and that the resulting kinetic parameters can be used to interpret the nature of the observed interactions of the particular CYP3A4 substrates within the active site. A more precise description of molecular events incorporating the binding of multiple substrate molecules can be achieved with a steady state and rapid equilibrium approach to the analysis of the interactions. The two domains in the model, which are bound singly and simultaneously by the substrates, can be distinguished from one another in such a way that a change in the binding affinity of the substrate and rate of product formation can be observed.

(1)

(2)

View larger version (18K): [in this window] [in a new window]   Scheme 1.   Proposed kinetic model for the two substrate interaction involving CYP3A4-mediated metabolism. The model contains two binding domains in the active site of the enzyme, which can be occupied by two different substrates (S1 and S2) to form ES1 and S2E, respectively. The two singly bound enzyme complexes are further bound with the second substrate molecule to give S2ES1. KS1, KS2, k1, and k2 are dissociation and rate constants for ES1 and S2E, respectively. Factors a, , and  represent changes of dissociation and rate constants when the second domains binds with an additional substrate molecule (e.g., S2ES1). Vmaxp1 or Vmaxp2 = k1[E]total or k2[E]total. Velocity equations for P1 and P2 are shown in the text.

Kinetic Derivation and Analysis. The values of all kinetic parameters in the equations were determined by nonlinear regression analysis of the experimental data using Mathematica 4.0 (Wolfram Research, Inc., Champain, IL). Each surface plot consists of both observed experimental data (scatter points) and predicted results (meshed plot) using either Eq. 1 or 2. An evaluation of goodness of the data fit was performed with statistical analyses (RSS and R²).

	Results

Top Abstract Introduction Materials and Methods Results Discussion References

Studies on the time dependence of product formation from testosterone, BQ, and BFC at their respective K_m concentrations were performed to conform linearity up to at least 20 min (data not shown). Less than 10% of the substrate was consumed, and secondary metabolism was minimal during the course of the incubation. The linearity of product formation as a function of incubation time suggests that CYP3A4-catalyzed reactions were at steady state. Thus, all enzyme-substrate complexes were formed at the fixed rate at which they decomposed (e.g., d[S₂E]/dt = 0, d[ES₁]/dt = 0 and d[S₂ES₁]/dt = 0). We assumed also that all the equilibria between enzyme and substrate(s) are very rapid as compared with the rate at which the enzyme-substrate complex breaks down. Therefore, velocity equations were derived to generate dissociation (K_S1 and K_S2) and rate constants (k_p).

Interaction of BFC with Testosterone. Eight concentrations each of BFC and testosterone (0-100 µM) were used to evaluate product formation, providing 64 incubation data points corresponding to various combinations of concentrations. Each incubation was divided into two aliquots that were analyzed by HPLC and LC-MS/MS for quantitation of BFC and testosterone metabolites, respectively. In the absence of effector, K_m and V_max values were determined to be 49.6 µM and 29.6 min¹ for testosterone and 48.5 µM and 49.3 min¹ for BFC, respectively (Table 1). The effect of one substrate on the metabolism of another substrate is shown in Fig. 1, A and B. The scatter plots indicate actual measurements of the reaction rate (z-axis, vertical) of a given substrate (x-axis, lateral) in the presence of the second substrate (y-axis, horizontal). Velocity equations were used to fit experimental data, and the meshed plots represent predicted results, which yielded estimates of kinetic parameters (Table 4). As seen in Table 1, the addition of BFC (3.13-100 µM) had a negligible effect on the apparent K_m of testosterone and resulted in a decrease (50%) in V_max. Similarly, testosterone had no effect on the apparent K_m but slightly decreased V_max values for BFC O-dealkylation. Hence, the pattern of the kinetic interaction of the two substrates did not conform to competitive inhibition in which no change in V_max and an increase in K_m should be observed. The results suggest that the two substrates have access to different binding domains within the active site. Thus, the two-site model was adapted to fit the experimental observations (Scheme 1) according to the observed kinetic interactions. Since the substrate interaction altered both apparent K_m and V_max, factors a, , and , which represent changes in dissociation and rate constants for one substrate in the presence of another, were taken into consideration. Surface plots (Fig. 1, A and B) and calculated kinetic values (Table 4) were used to assess the goodness of fit of predicted results with statistics (e.g., standard error, R², and RSS). K_S1, K_S2, V_max1, and V_max2 are dissociation constants and maximal velocities for singly bound ES₁ (S₁ = BFC) and S₂E (S₂ = testosterone), respectively. These values are consistent with apparent K_m and V_max values measured in the absence of the second substrate (Table 1). However, when the second molecule was present in the enzyme reaction ([E]_total = [S₂ES₁]), the changes for K_S and V_max were found to decrease by 51 and 28% for K_S (a = 0.49 and 0.72 from the two respective equations; Table 4) and by 75% for V_max1 (testosterone,  = 0.25) and 41% for V_max2 (BFC,  = 0.59), respectively. These results suggest that the two substrate molecules can bind simultaneously to two separate domains at the active site of the enzyme and subsequently interact with each other, leading to partial inhibition and changes in apparent kinetic constants (for example, decrease in both K_m and V_max).

View larger version (38K): [in this window] [in a new window]   Fig. 1.   Kinetic interaction between BFC and testosterone over the concentration range from 3.125 to 100 µM. A, effect of BFC on CYP3A4-mediated testosterone 6-hydroxylation; B, effect of testosterone on CYP3A4-catalyzed BFC O-dealkylation.

Interaction of BQ with Testosterone. Figure 2, A and B, shows the results of kinetic interaction studies between BQ and testosterone. In the absence of effector, the K_m values for testosterone and BQ were 58.6 and 76.2 µM, and the V_max values were 35.9 and 25.2 min¹, respectively (Table 2). In the presence of BQ, K_m and V_max for testosterone decreased slightly. On the other hand, testosterone decreased K_m and had no significant change on V_max. The inhibition was neither competitive nor noncompetitive. These findings indicated that the two substrate molecules must bind differentially to the two domains as in the case of BFC and testosterone. The predicted results (meshed plots) and the calculated parameters are shown in Fig. 2, A and B, and Table 4, respectively. The goodness-of-fit of the two velocity equations was evaluated by RSS (19.4 and 95.2) and R² (0.987 and 0.967). As a consequence, factor a values that define a maximal decrease in apparent K_m in the presence of the effector were calculated by the two equations to be 0.58 and 0.63, respectively. These imply that the inclusion of the second molecule can decrease the apparent K_m for the first substrate by up to 37 to 42%. Similarly, factors  and , which represent the changes in V_max1 and V_max2, when the second substrate is bound completely, were estimated to be 0.75 and 0.48, respectively, resulting in a maximal inhibition of 25% for V_max1 in BQ O-dealkylation and 52% for V_max2 in testosterone 6-hydroxylation.

View larger version (42K): [in this window] [in a new window]   Fig. 2.   Kinetic interaction of BQ and testosterone over the concentration range from 3.125 to 100 µM. A, effect of BQ on CYP3A4-mediated testosterone 6-hydroxylation; B, effect of testosterone on the CYP3A4-catalyzed BQ O-dealkylation.

Interaction of BQ and BFC. The metabolism of BQ or BFC generated fluorescent O-dealkylation products that were detected by their fluorescence following an HPLC separation. BQ (3.13-100 µM) decreased significantly both the apparent K_m and V_max values for BFC from 42 to 21 µM (K_m) and from 52 to 14 min¹ (V_max). Similarly, when BFC was used as an effector, apparent K_m and V_max values for BQ decreased from 63 to 32 µM and from 19.2 to 11.8 min¹, respectively (Table 3). Hence, the interaction between the two substrates was not competitive. The two-site model was used to fit the data, resulting in three-dimensional surface plots (Fig. 3, A and B) and a solution of the predicted kinetic parameters (Table 4). Factor a values (0.3 and 0.21, respectively, from the two equations) indicate that one substrate can interact with another by diminishing apparent K_m by as much as 70 to 79%. Similarly, the apparent V_max for BQ decreased significantly by 91 ( = 0.09) or 77% for BFC ( = 0.21) in the presence of the respective effector.

View larger version (40K): [in this window] [in a new window]   Fig. 3.   Kinetic interaction between BQ and BFC over the concentration range from 3.125 to 100 µM. A, effect of BQ on CYP3A4-catalyzed BFC O-dealkylation; B, effect of BFC on the CYP3A4-mediated BQ O-dealkylation.

	Discussion

Top Abstract Introduction Materials and Methods Results Discussion References

Inhibition of a drug-metabolizing enzyme, typically a cytochrome P450 isoform, is a common cause of adverse clinical interactions between coadministered therapeutic agents. Consequently, high-throughput screening assays for P450 inhibition and drug-drug interactions have now become an integral part of drug discovery and development programs, where they serve to eliminate drug candidates at an early stage. Appropriate substrates that are used to probe a particular P450 enzyme can be selected to provide a rapid estimation of IC₅₀ values and, thus, an initial determination of the inhibitory capability for the NCEs. Because the majority of drugs metabolized by P450 are substrates for CYP3A4, this isoform has been the subject of the most intensive screening efforts.

Numerous investigators have developed assays for CYP3A4 activity with different probe substrates that are markers for this isoform. Unfortunately, recent experience has revealed that 1) poor correlations often are obtained with varying degrees of inhibition when different CYP3A4 probe substrates are used (Kenworthy et al., 1999; Wang et al., 2000a,b); 2) the presence of one substrate may result in partial inhibition of another substrate and vice versa (Kenworthy et al., 1999; Wang et al., 2000a); and 3) IC₅₀ values for a given inhibitor may differ widely between two CYP3A4 substrates (Wang et al., 1997; unpublished data). These issues are of particular concern with respect to the current trend for adopting high-throughput screening approaches for assessment of enzyme inhibition. The present study provides evidence that the multiplicity and complexity of CYP3A4-mediated kinetics is due to the existence of multiple binding domains at the active site of the enzyme. Accordingly, an inhibitor that inhibits the metabolism of one substrate (competitive inhibition) may not inhibit that of other substrates (multiple binding domains). Thus, in vitro inhibitory screens may be inconsistent and lead to erroneous prediction of drug-drug interaction potential.

In the present study, drug-drug interaction between two of three CYP3A4 substrates was studied using a wide range of substrate concentrations that varied from approximately 0.1 to twice the K_m for each substrate. The concentrations for both interacting substrates were selected to cover all possible concentration-related drug-drug interactions. Thus, an entire kinetic analysis was conducted with a sufficiently large number of data points to obtain reliable estimates of kinetic parameters. The key finding from the present study is that the interaction between two CYP3A4 substrates did not conform to Michaelis-Menten kinetics (i.e., the interaction was not competitive within the active site of CYP3A4), suggesting that the two substrates can coexist and also interact with each other at the active site. This is not surprising because there is abundant evidence that CYP3A4 exhibits non-Michaelis-Menten kinetics. Thus, there are particular kinetic features for several CYP3A4 substrates that cannot be interpreted within the context of the Michaelis-Menten model and that require the adoption of an enzyme model with multiple binding sites to explain the observed drug interactions.

Overall, the model proposed here accounts for all possible interactions between two substrates over a wide range of concentrations. The resulting kinetic estimates are listed in Table 4, and the predicted K_S and V_max values for each of the substrates are consistent with the corresponding values observed in the absence of the second substrate. Factor a, which represents a change in K_S in the presence of the second substrate, generally was less than unity. This suggested that the addition of the second substrate could result in a binding affinity greater than that for the singly bound complex (a = 0.21-0.72). Factors  and , which reflect changes in V_max for the two singly bound complexes when the second substrate is introduced, also were predicted and found to range from 0.09 to 0.75, meaning that interaction between the two substrates can decrease apparent V_max by as much as 25-91%. It is clear that our study made use of sufficient data points to allow detailed examination of the effects of atypical kinetics. Comprehensive data sets are particularly important for substrates showing atypical kinetics and ensure a high degree of confidence in data fitting that leads to a precise solution of kinetic parameters in the model.

Our study indicates that all kinetic interactions observed between the two substrates resulted, to varying degrees, in a decrease in both apparent K_m and V_max values, suggesting that the two substrate molecules may bind in close proximity and thereby partially inhibit the metabolism of one another. Taken together with the results of previous studies (Kenworthy et al., 1999; Houston and Kenworthy, 2000; Wang et al., 2000a; Shou et al., 2001a,b), the findings of the present investigation support the view that the physicochemical properties of different CYP3A4 substrates give rise to their unique interactions (K_S) with particular amino acid residues within the CYP3A4 active site. As a result, the three substrates examined here (BQ, BFC, and testosterone), which did not compete with each other in combinations of any two substrates, probably occupy their own binding domains at the active site of the enzyme. The three domains allow for the simultaneous existence of at least two substrates for metabolism. With regard to the change in apparent kinetic constants (K_m and V_max), the binding of one substrate may interfere with that of another substrate. This probably is a result of steric restrictions at the active site accommodating the two substrate molecules. However, it would be difficult to demonstrate, kinetically, the simultaneous occupancy of binding domains by more than two substrate molecules due to the complexity of the kinetic expression of the multiple-site model and a poor solution of numerous variables in the resulting equations.

Although interaction of any pair of the substrates can be described by the observed kinetics, more complexity of enzyme kinetics (i.e., allosterism of steroids) has to also be taken into consideration. It has been reported that CYP3A4-mediated steroid and BFC oxidations were shown to be sigmoidal characteristics. Allosteric mechanism of the former was understood by mutagenesis study (Harlow and Halpert, 1998; Domanski et al., 2000), in that the sigmoidicity of steroid hydroxylation can be shifted to a hyperbolic kinetics when replacing residues Leu-211 and Asp-214 with the larger Phe and Glu (L211F & D214E), respectively. This suggests that an additional site (effector binding site), for which the residues 211 and 214 are responsible, may be required to conduct cooperativity of the enzyme. Thus, more complicated situations can exist, which probably involve two sites for the steroid responsible for the sigmoidicity of steroid metabolism and one site for either BFC or BQ, although an interaction between the effector (steroid) site and BFC (BQ) site has not been elucidated kinetically.

When both domains are occupied by the respective substrates, complexity of drug-drug interaction within the CYP3A4 active site can occur and can be expressed by the factors, such as a, , and , in a manner different from that of the singly bound site. Thus, evaluation of the test compound as a CYP3A4 inhibitor can be inconsistent with use of different probe substrates. Although there is no simple or alternative solution for accurately predicting from one substrate to other due to the multiple binding, it is necessary to categorize CYP3A4 substrates by a fully kinetic analysis of drug-drug interaction. The rationale includes 1) grouping of the CYP3A4 substrates to ensure the substrates in a given group are competitive in nature, suggesting that the substrates bind to the same site; and 2) probe substrates from each group, therefore, can be selected for screening of NCEs for P450 inhibition, which may correlate well between substrates in the same group but may vary with the substrates between the different groups. In addition, care should be taken when atypical kinetics occur in terms of test compounds used. The interaction can be both substrate- and inhibitor-dependent due to the multiple bindings. Therefore, at present, it forces the choice of investigation of as many CYP3A4 substrates as possible in inhibition screens, particularly for the leads, until the mechanism of the unusual drug-drug interactions is fully understood, and thorough data for the rationale are completely convinced.