Abstract
Cytochrome P450 (P450) protein-protein interactions have been shown to alter their catalytic activity. Furthermore, these interactions are isoform specific and can elicit activation, inhibition, or no effect on enzymatic activity. Studies show that these effects are also dependent on the protein partner cytochrome P450 reductase (CPR) and the order of protein addition to purified reconstituted enzyme systems. In this study, we use controlled immobilization of P450s to a gold surface to gain a better understanding of P450-P450 interactions between three key drug-metabolizing isoforms (CYP2C9, CYP3A4, and CYP2D6). Molecular modeling was used to assess the favorability of homomeric/heteromeric P450 complex formation. P450 complex formation in vitro was analyzed in real time utilizing surface plasmon resonance. Finally, the effects of P450 complex formation were investigated utilizing our immobilized platform and reconstituted enzyme systems. Molecular modeling shows favorable binding of CYP2C9-CPR, CYP2C9-CYP2D6, CYP2C9-CYP2C9, and CYP2C9-CYP3A4, in rank order. KD values obtained via surface plasmon resonance show strong binding, in the nanomolar range, for the above pairs, with CYP2C9-CYP2D6 yielding the lowest KD, followed by CYP2C9-CYP2C9, CYP2C9-CPR, and CYP2C9-CYP3A4. Metabolic incubations show that immobilized CYP2C9 metabolism was activated by homomeric complex formation. CYP2C9 metabolism was not affected by the presence of CYP3A4 with saturating CPR concentrations. CYP2C9 metabolism was activated by CYP2D6 at saturating CPR concentrations in solution but was inhibited when CYP2C9 was immobilized. The order of addition of proteins (CYP2C9, CYP2D6, CYP3A4, and CPR) influenced the magnitude of inhibition for CYP3A4 and CYP2D6. These results indicate isoform-specific P450 interactions and effects on P450-mediated metabolism.
Introduction
Cytochrome P450s (P450s) are a superfamily of mono-oxygenases responsible for metabolizing 75% of commercial pharmaceuticals and other xenobiotics and endobiotics (Guengerich, 2006). They are found primarily in the liver, where they are membrane bound via their hydrophobic N terminus to the smooth endoplasmic reticulum (Szczesna-Skorupa and Kemper, 1993). Their mechanism of action is dependent on the sequential transfer of two electrons, each requiring cytochrome P450 reductase (CPR) (Guengerich and Johnson, 1997). The second transfer is facilitated by cytochrome b5 but cytochrome b5 is not required for P450-mediated metabolism (Yamazaki et al., 1997; Shimada et al., 2005; Locuson et al., 2006). In microsomes, P450 levels exceed those of CPR by at least 5-fold (Estabrook et al., 1971; Reed et al., 2011), suggesting that enzyme mobility is required (Kawato et al., 1982; Gut et al., 1983) or that multiple P450s use a single CPR (Peterson et al., 1976; Taniguchi et al., 1979; Eyer and Backes, 1992).
It has been suggested that P450s are clustered and not uniformly distributed in the endoplasmic reticulum (Matsuura et al., 1978) and that zonal differences may exist (Jungermann, 1995). In vitro, it has been shown that P450s interact (Kaminsky and Guengerich, 1985; Alston et al., 1991; Davydov, 2011), modulating P450-mediated metabolism (Backes and Kelley, 2003; Reed and Backes, 2012). This effect has been seen in many reconstituted enzyme systems, including human CYP2C9-CYP2D6 (Subramanian et al., 2009), human CYP2C9-CYP3A4 (Subramanian et al., 2010), human CYP2C9-CYP2C19 (Hazai and Kupfer, 2005), human CYP3A-CYP2E1 (Davydov et al., 2015), rabbit CYP1A2-CYP2B4 (Backes et al., 1998; Davydov et al., 2001; Reed et al., 2013), rabbit CYP1A2-CYP2E1 (Kelley et al., 2006), and human CYP3A4-CYP1A1 and CYP3A4-CYP1A2 (Yamazaki et al., 1997). P450-P450 interactions have also been observed in microsomal preparations (Kaminsky and Guengerich, 1985; Cawley et al., 2001), triple expression systems (Cawley et al., 1995; Tan et al., 1997; Li et al., 1999), and homomeric oligomerization of CYP2E1 (Jamakhandi et al., 2007), CYP2B4 (Davydov et al., 1992), and CYP3A4 (Davydov et al., 2005, 2010).
Although P450-P450 interactions in vitro have been found to affect metabolism, ambiguity remains because some studies have reported that P450-P450 interactions do not affect metabolism. Dutton, et al. (1987) showed no effect on P450-mediated testosterone metabolism in the presence of other P450s in rat microsomes (Dutton et al., 1987). Alston et al. (1991) showed the formation of heterodimers by some P450s, suggesting that there is some selectivity for the formation of heteromeric P450 complexes. Elucidating the underlying factors controlling P450-P450 interactions is essential for accurately predicting in vivo P450 substrate clearance rates from in vitro data. Predictions of clearance are often based on single-isoform reconstituted systems, which lack the possibility of heteroaggregate formation-based effects. In addition, although oligomerization does occur in solution, the extent and characterization of how P450s interact in vivo has yet to be fully examined. To this end, several groups have investigated P450 incorporation into membranes using molecular modeling approaches (Cojocaru et al., 2011; Liu et al., 2013). In addition, in vitro studies establishing P450-P450 membrane interactions using crosslinking (Alston et al., 1991; Hu et al., 2010), fluorescence resonance energy transfer (Szczesna-Skorupa et al., 2003), and luminescence resonance energy transfer (Davydov et al., 2013) have provided insight on the degree of P450 oligomerization in membranes.
There are three main mechanisms by which a P450 can affect the metabolic rate of another, including competition for CPR, complex formation with altered affinity for CPR, and formation of a complex with altered substrate binding or turnover rate (Backes and Kelley, 2003; Reed and Backes, 2012). Controlled immobilization of P450s has recently been used to allow better understanding of metabolism (Davydov et al., 2005; Gannett et al., 2006; Mak et al., 2010; Fantuzzi et al., 2011; Panicco et al., 2011; Wollenberg et al., 2012; Bostick et al., 2015) and the formation of homomeric/heteromeric complexes (Backes and Kelley, 2003). We developed a scheme whereby soluble P450 is covalently attached to a self-assembled monolayer (SAM) on a gold film. The result retains full metabolic activity compared with the same amount of lipid-reconstituted P450 (Supplemental Fig. 1) and uses the normal protein partners and cofactors to metabolize substrates (Gannett et al., 2006). This platform models how P450 is bound in vivo and allows for control of P450 aggregation. Here, we assess the interaction of CYP2C9 with three major P450 isoforms (CYP2C9, CYP3A4, and CYP2D6) and with CPR, demonstrate the formation of homomeric/heteromeric P450 complexes in real time, and determine the consequences of these complexes on P450-mediated metabolism, all supported by molecular modeling studies of the P450 complexes.
Materials and Methods
Chemicals and Reagents.
All chemicals were used as purchased. Potassium phosphate, HEPES, sodium chloride, 11-mercaptoundecanoic acid (MUA), 1-octanethiol (OT), N-(3-dimethylaminopropyl)-N′-ethylcarbodiimide hydrochloride (EDC), N-hydroxysulfosuccinimide sodium salt (NHS), (S)-flurbiprofen (flurbiprofen), 4′-hydroxyflurbiprofen, dilauroylphosphatidylcholine (DLPC), NADPH, and cetyltrimethylammonium bromide were purchased from Sigma-Aldrich (St. Louis, MO). Truncated histidine-tagged CYP2C9 was prepared by expression in an Escherichia coli system, isolated, and purified as described previously (Cheesman et al., 2003; Hummel et al., 2005; Locuson et al., 2006). Human CPR was purchased from Invitrogen (Carlsbad, CA). All solutions were prepared in 18 MΩ-cm water unless otherwise noted. Phosphate buffer (40 mM, pH 7.4) contained 150 mM sodium chloride [phosphate-buffered saline (PBS)]. HEPES buffer (10 mM, pH 7.4) contained 150 mM sodium chloride, 3 mM EDTA, and 0.005% (v/v) Surfactant P20. Titanium and gold pellets were purchased from Kurt J. Lesker Company (Jefferson Hills, PA). Cryobuffer was made from PBS and glycerol [80:20 (v/v)].
Molecular Modeling.
Computational modeling was employed to determine possible enzyme binding orientations and the relative binding energy of the enzyme pairs CYP2C9-CYP2C9, CYP2C9-CYP3A4, CYP2C9-CYP2D6, and CYP2C9-CPR. To account for conformational changes affecting the binding strength and binding site, several CYP2C9 crystal structures were examined, including 1OG2 (without substrate), 1OG5 (with bound warfarin, closed) (Williams et al., 2003), and 1R9O (with bound flurbiprofen, partially open) (Wester et al., 2004). Structural similarity was assessed using the MultiSeq module (Eargle et al., 2006) of Visual Molecular Dynamics. P450 structures, crystallized without substrate, were used to match the experimental conditions: CYP2D6 [Rowland et al., 2006; Protein Data Bank (PDB) identifier 2F9Q] and CYP3A4 [Yano et al., 2004; PDB identifier 1TQN]. The crystal structure selected for CPR is locked into a conformation adopted when binding to a P450 enzyme due to a TGEE truncation in the hinge region (Hamdane et al., 2009; PDB identifier 3ES9 chain A). GRAMM-X (Tovchigrechko and Vakser, 2006) was used to computationally predict binding modes and the 10 lowest potential energy models per enzyme pair were selected for analysis. All homodimers and heterodimers were examined in Visual Molecular Dynamics (Humphrey et al., 1996) with respect to residues defining areas of interest as reported in the literature, including the active site and solvent tunnel entrances (Cojocaru et al., 2011), the CPR binding site (Shen and Kasper, 1995; Hazai et al., 2005; Hamdane et al., 2009), and the truncated N terminus. The relative binding energy for the lowest-energy GRAMM-X models for each binding pair was calculated using the molecular mechanics (MM)/generalized born surface area (GBSA) method (Miller et al., 2012) after minimization, equilibration, and 2 ns molecular dynamics at 300 K using the Amber software suite (Roe and Cheatham, 2013).
Gold-OT/MUA-CYP2C9 Chip Fabrication.
Two different materials were used as test platforms for the experiments described: gold-coated silicon wafers for generation of metabolites and gold-coated glass coverslips for surface plasmon resonance (SPR) detection of binding. Both were prepared for surface modification using a Temescal BJD-2000 system (Edwards Vacuum, Phoenix, AZ) with an Inficon XTC/2 deposition controller (Inficon, East Syracuse, NY). Evaporation occurred at a voltage of 10.0 kV. The current was set so that metal vaporization occurred at a rate of 0.03–0.10 nM/s (approximately 50 mA for titanium and 100 mA for gold). Thin films composed of titanium, 5-nm thick, and then gold, approximately 50-nm thick, were deposited on silicon chips or glass coverslips. Silicon wafer platforms were diced into 4 mm × 6 mm chips and then were sonicated in a Bransonic 220 (Bransonic, Danbury, CT) (5 minutes) and rinsed in succession with acetone, isopropyl alcohol, and water to ensure a clean surface. Glass platforms were cut into 10 mm × 12 mm rectangles, sonicated (15 minutes) with ethanol. A SAM was then formed on the gold surface by first rinsing it with ethanol, followed by soaking it in an ethanolic solution of OT (7.5 mM) and MUA (2.5 mM), overnight under argon. Excess thiol was removed by rinsing, in succession, with absolute ethanol, 95% ethanol, and water (three times each). Chips were dried under a gentle stream of nitrogen.
SPR Binding Analysis.
Glass platforms coated with gold-SAM as described above were loaded into a Biacore X100 SPR (GE Healthcare, Piscataway, NJ) for further surface modification and analysis. The surface was then activated with EDC (0.4 M) and NHS (0.1 M) in PBS running buffer as per Biacore instructions. CYP2C9 (100 nM) in PBS running buffer was flowed (10 μl/min) over the test surface (480 seconds), capped with ethanolamine (1 M) in PBS running buffer, and then rinsed with PBS running buffer alone (480 seconds). The control surface was prepared by directly capping it with ethanolamine. Immobilization levels reaching 70–200 response units above the control surface, after running buffer rinse, were considered acceptable. CYP2C9 was immobilized and the binding partner flowed over, without substrate, conditions that favor the partially open conformation. Analysis of binding between immobilized and soluble enzymes (0, 5, 10, 50, 100, and 500 nM) was conducted in PBS running buffer at 30 μl/min. After stabilization, contact times for soluble enzyme were 180 seconds followed by 900-second dissociation. Flow cells were regenerated with cetyltrimethylammonium bromide (3.9 mM). Two concentrations of the soluble enzyme (50 nM and 100 nM) were measured twice to ensure reproducibility of the data. Each P450 binding pair was run at least three times. Fitting of kinetic data was accomplished using both Biacore X100 Evaluation software and Scrubber 2.0 (BioLogic Software, Campbell, ACT, Australia). CYP2C9, CYP3A4, and CPR were fit using Biacore X100 Evaluation software. Biacore X100 Evaluation was unable to fit CYP2D6 curves, and Scrubber 2.0 was employed instead. Crossanalysis between fitting software demonstrated consistency.
Immobilization of CYP2C9 Enzyme on Gold/Silicon Platform.
To bond the CYP2C9 to the SAM, the SAM-coated slides were first activated by immersion in an aqueous solution of EDC (2 mM) and NHS (5 mM) [1:1 (v/v)], each in water, for 2 hours at room temperature. Activated chips were rinsed with PBS, then soaked (24 hours, room temperature) in a solution containing CYP2C9 (100 nM), flurbiprofen (40 µM), and dapsone (40 µM) in PBS, under argon, to covalently bond CYP2C9 to the NHS esters. We previously showed that inclusion of substrates flurbiprofen and dapsone preserves CYP2C9’s active site integrity and allows retention of enzymatic activity through immobilization (Supplemental Fig. 1) (Gannett et al., 2006). Chips bearing the bonded enzyme were rinsed with PBS to remove substrates from the active site prior to placement in cryobuffer solution and were stored at −80°C for at least 8 hours prior to use.
Reconstituted CYP2C9 Enzyme Incubation.
Reconstituted CYP2C9 incubations were run in triplicate. Incubations contained CYP2C9 (0.25 nM), saturating CPR (8 nM), DLPC (10 µg), and flurbiprofen (160 µM) in PBS (40 mM, pH 7.4) for a total volume of 250 µl. The amount of lipid per incubation is the same as other literature studies (Locuson et al., 2006; Subramanian et al., 2009, 2010). Because of the high amount of lipid to P450 in our study, controls were run at a DLPC/CYP2C9 ratio of 100:1, and the measured rates of metabolism were lower than those obtained with the conditions used in this study (data not shown). Immobilized CYP2C9 (CYP2C9i) incubations were conducted using two 4 mm × 6 mm gold chips placed back to back (gold coating facing outward) inside a 1.5-ml microcentrifuge tube. The total volume of 250 µl was sufficient to completely cover the chips and expose CYP2C9i to reaction components. No DLPC was added to incubations containing CYP2C9i, and additional nonimmobilized P450s added to these systems were not provided with lipids, creating a lipidless system in which metabolism only occurs through CYP2C9i. Coverage of CYP2C9 on gold chips was estimated based on previous work (Gannett et al., 2006; Yang et al., 2009). Ratios of CYP2C9i to CYP2C9 in solution (CYP2C9s) ranged from 1:1 to 1:4. Ratios of CYP2C9/CYP2D6 and CYP2C9/CYP3A4 ranged from 1:0.25 to 1:8 and 1:1 to 1:12, respectively, to mimic possible in vivo ratios of these enzymes (Meijerman et al., 2007). Incubations were placed on ice during the addition of precooled solutions of reactants and then preheated (3 minutes) at 37°C before initiation of the incubation by the addition of NADPH (200 µM, final concentration) and then incubated at 37°C overnight (16 hours). Reactions were terminated by the addition of 10 µl 500 ng/ml 2-fluoro-4-biphenyl acetic acid (internal standard) and 30 µl 50% phosphoric acid. Incubation mixtures were centrifuged (10 minutes, 13,400g) to pellet proteins, and supernatant (135 µl) was collected and loaded into liquid chromatography vials for analysis.
The effect of the order of mixing of P450s and CPR was investigated by 1) preincubating CYP2C9 (4 nM) or CYP2C9i (0.25 nM) with CPR (5 minutes) prior to the addition of CYP3A4 or CYP2D6 or 2) preincubating CYP2C9 (4 nM) or CYP2C9i (0.25 nM) with CYP3A4 or CYP2D6 (5 minutes) prior to addition of CPR. In studies investigating the order of addition in lipid-reconstituted systems, CPR was lowered with a P450/CPR ratio less than 1:2. CPR levels for CYP2C9i studies were the same as used for the previously explained immobilized metabolism setup, saturating CPR, to ensure sufficient metabolite generation. Incubations were placed on ice during the addition of precooled solutions of reactants and then preheated (3 minutes) at 37°C before initiation of the incubation by the addition of NADPH (200 µM, final concentration) and then incubated at 37°C for 1 hour for lipid-reconstituted CYP2C9 or overnight (16 hours) for CYP2C9i incubations. Reactions were terminated by the addition of 10 µl 500 ng/ml 2-fluoro-4-biphenyl acetic acid (internal standard) and 30 µl 50% phosphoric acid. Incubation mixtures were centrifuged (10 minutes, 13,400g) to pellet proteins, and supernatant (135 µl) was collected and loaded into liquid chromatography vials for analysis. Linearity of flurbiprofen metabolism to its 4-hydroxy metabolite as catalyzed by CYP2C9 was evaluated using study parameters for lipid-reconstituted enzyme with respect to time and was found to be nonlinear for the time courses used in our study (data not shown). Insufficient metabolite production resulted from the immobilized protein experiments to be measurable within any timeframe that might have resulted in linear formation. All experiments were performed three times on separate days.
High-Performance Liquid Chromatography Assay of 4′-Hydroxyflurbiprofen.
Formation of the flurbiprofen metabolite, 4′-hydroxyflurbiprofen, was measured by high-performance liquid chromatography with fluorescence detection. The high-performance liquid chromatography system consisted of an Alliance 2695XE pump/autosampler and a 2495 fluorescence detector (Waters, Milford, MA) set at an excitation wavelength of 280 nm and an emission wavelength of 310 nm. The mobile phase consisted of potassium phosphate (20 mM, pH 7.4) and acetonitrile [50:50 (v/v)], a 0.75 ml/min flow rate, and a 4.6 × 150 mm Zorbax C18 column (Agilent, Santa Clara, CA). The retention times of 4′-hydroxyflurbiprofen and the internal standard were 4.1 and 7.7 minutes, respectively. Metabolite concentration was determined using a calibration curve of 0–50 ng 4′-hydroxyflurbiprofen.
Statistical Analysis.
The results for metabolism studies are expressed as the means ± S.E. and were analyzed by a one-way analysis of variance followed by Dunnett’s test for multiple comparisons with the positive control group. One-way analysis of variance with Dunnett's post-test was performed using GraphPad Prism software (version 6 for Windows, www.graphpad.com; GraphPad Software, La Jolla, CA).
Results
Computationally Predicted Dimers.
P450s are known to form dimers and multimers and their formation is believed to affect metabolic activity. Here, modeling approaches were used to predict the likely binding poses of homodimers and heterodimers, to infer the consequences of dimer formation on metabolic activity, and to estimate the stability of the dimeric complexes relative to the monomer forms. The enzyme pairs examined were CYP2C9-CYP2C9, CYP2C9-CYP3A4, CYP2C9-CYP2D6, or CYP2C9-CPR. For each pair, only the CYP2C9 residues involved in binding were considered, including those involved in CPR binding (Shen and Kasper, 1995; Hamdane et al., 2009), the solvent access channel, and the substrate channel entrance (Cojocaru et al., 2007, 2011) (Supplemental Table 1).
It has been demonstrated that P450s exist in both open and closed conformations (Scott et al., 2004). The closed conformation is favored upon substrate binding (Williams et al., 2000, 2003; Yano et al., 2004) and alters the affinity toward enzyme partners (Hazai et al., 2005; Roberts et al., 2010). To account for the potential conformational changes on the modeling results, dimer formation was examined with an open and a closed form of CYP2C9. To select these structures, three CYP2C9 crystal structures were considered: 1OG2 (without substrate), 1OG5 (with bound warfarin) (Williams et al., 2003), and 1R9O (with bound flurbiprofen) (Wester et al., 2004). The 1OG2 and 1OG5 models, that represent open and closed states, respectively, were more conformationally similar to each other than to 1R9O, a partially open conformation (Supplemental Fig. 2). Therefore, further analysis was conducted only on the 1OG5 (closed) and 1R9O (partially open) models.
For the purpose of generating dimer structures, blind docking, using GRAMM-X was performed for each dimer combination. Each of the 10 models returned by GRAMM-X for the homodimers and heterodimers were examined for binding in defined regions of interest (Supplemental Table 1), and the occurrence of each interaction type and frequency is shown in Table 1. Dimerization involved the solvent or substrate channels 25% of the time and the CPR binding site approximately 7.5% of the time in predicted models (Table 1). Dimerization most frequently involved the solvent access channel region of CYP2C9. In the case of the closed CYP2C9 (1OG5), this location was heavily favored. In either case, the CPR binding site was the least preferred site. The distinctions between bindings for the three sites were greatest for the closed model, the top pairs for which are shown for in Fig. 1.
Dimer binding sites predicted by GRAMM-X
Each dimer was composed of either 1R90 or 1OG5 and the indicated binding partner.
Representation of the top-scored interaction predicted by GRAMM-X for 1R9O bound to CYP3A4, CYP2C9, CYP2D6, and CPR. Residues defining the CPR binding site (green), the substrate access channel entrance (orange), the solvent channel entrance (blue), and the N terminus (black) are highlighted, and other residues have been removed from view. Partner molecules are light gray. (A) 1R9O-CYP3A4 solvent channel interface area. Binding as depicted by this model could affect the substrate turnover rate. (B) 1R9O-CYP2C9 CPR binding site interface area. In addition, the partially blocked surface could affect the strength of the CYP2C9-CPR binding interaction for the CYP2C9 (white) molecule. (C) 1R9O-CYP2D6 active site area. Binding as depicted by this model could affect substrate entry and exit into the active site. (D) 1R9O-CPR substrate access channel area. Dimerization as depicted in this model could affect substrate recognition, binding, and turnover rate.
Binding Free Energy In Silico.
Computational modeling was employed to determine the relative binding energy between four enzyme pairs: CYP2C9-CYP2C9, CYP2C9-CYP3A4, CYP2C9-CYP2D6, and CYP2C9-CPR. Molecular dynamics were run on these structures (neutralized and fully solvated with explicit waters). Binding free energies were calculated using the MM/GBSA method, which has been shown to be more successful in ranking binding affinities than the MM/Poisson–Boltzmann surface area method, although it is less successful in calculating absolute energy values (Hou et al., 2011). Resultant energies are reported in kilocalories per mole in Table 2. Negative values for total binding free energy suggest favorable protein-protein complexes. The rank order of the dimers is, from strongest to weakest, CYP2C9-CPR, CYP2C9-CYP2D6, CYP2C9-CYP2C9, and CYP2C9-CYP3A4.
MM/GBSA binding free energies for CYP2C9 dimers
Calculated binding free energies for top-scored interactions predicted by GRAMM-X for 1R9O bound to CPR, CYP2D6, CYP2C9, and CYP3A4. Values in parentheses are standard deviations. All values are based on at least 50 snapshots from a single trajectory spaced by 10 ps to avoid correlation between models.
Binding Affinity In Vitro.
CYP2C9 was immobilized to a planar gold surface in a SPR flow cell via a SAM (Supplemental Fig. 3), reproducibly achieving immobilization levels of 70–200 response units relative to the ethanolamine-capped control surface. This low immobilization level was targeted to allow kinetic measurements of the system (Karlsson et al., 1991; Myszka, 1999). CYP2C9i was then exposed to soluble CYP2C9, CYP2D6, or CYP3A4 to analytically determine equilibrium dissociation constants for each dimer pair. A control flow cell was used to subtract out any nonspecific interactions between the solution enzyme and the SAM. A representative sensorgram obtained for CYP2C9i and CYP2D6 is shown in Fig. 2, along with the best fits, which assumed a dimer formation (1:1). Attempts to fit with higher-order oligomers (e.g., 1:2) resulted in poorer fits, implying that only dimers were formed in the SPR experiment. Therefore, the binding equilibrium dissociation constants were calculated based on a 1:1 binding model (Table 3). The rank order of binding affinities, from strongest to weakest, is the enzyme pairs CYP2C9-CYP2D6, CYP2C9-CYP2C9, CYP2C9-CPR, and CYP2C9-CYP3A4.
SPR sensorgram obtained from a system composed of soluble CYP2D6 binding to CYP2C9i. Both the observed response (black) and the Scrubber 2.0 fitted lines (orange) are shown for CYP2D6 concentrations of 5, 10, 50, 100, and 500 nM. Concentrations of 50 and 100 nM were repeated to ensure reproducibility. CYP2D6 in PBS was flowed over the surface from 0 to 180 seconds and then PBS only was flowed for 180 to 900 seconds at a flow rate of 30 μl/min. CYP2C9i was bonded to the surface of the sample cell and ethanolamine-capped SAM served as the control, in the control cell, at 30 μl/min.
SPR-derived CYP2C9 equilibrium dissociation constants
Average equilibrium dissociation constants (KD) ± S.E.M. are given for the binding of CYP2C9 with the indicated partner. Values were calculated by fitting the SPR sensorgrams with both Biacore Evaluation or Scrubber 2.0 software and assuming 1:1 stoichiometry. SPR runs were performed in triplicate on separate days.
CYP2C9 Effect on CYP2C9i-Mediated Flurbiprofen Metabolism.
We next investigated the effects of P450-P450 interactions on metabolism utilizing our immobilized platform. The amount of immobilized enzyme was estimated as previously described (Gannett et al., 2006; Yang et al., 2009) coupled with the SPR immobilization response. This gave a coverage of 41 pg/mm2. The effect of various amounts of solution CYP2C9 coincubated with CYP2C9i on flurbiprofen metabolism was then determined.
Figure 3 shows the amounts of metabolite formed from CYP2C9i-mediated metabolism of flurbiprofen in the presence of three concentrations of lipidless CYP2C9 in solution at a total P450/CPR ratio of 1:4. Metabolism by lipidless CYP2C9 alone did not significantly differ from the negative control (data not shown); thus, metabolite formation is due entirely to CYP2C9i. CPR was added prior to the addition of lipidless CYP2C9, followed by the addition of NADPH. With a ratio of CYP2C9i/CYP2C9 of 1:2, significant activation of flurbiprofen metabolism was observed. CYP2C9i-mediated metabolism of flurbiprofen was not significantly different from the positive control at ratios above 1:2 (Fig. 3).
Effect of solution CYP2C9 on CYP2C9i-mediated metabolism of flurbiprofen at three ratios of 2C9i to lipidless CYP2C9 and saturating CPR [P450(total)/CPR, 1:4], incubated for 16 hours. A significant activation (P < 0.01) is seen at a 2C9i/2C9 ratio of 1:2. Single-factor analysis of variance was used for statistical comparisons. 2C9, CYP2C9; 2C9i, immobilized CYP2C9. **P < 0.01.
Effect of CYP3A4 and CYP2D6 on CYP2C9-Mediated Metabolism of Flurbiprofen.
The effects of various amounts of CYP3A4 and CYP2D6 on CYP2C9i-mediated flurbiprofen metabolism are presented in Figs. 4 and 5. The presence of CYP3A4 had no effect on CYP2C9-mediated metabolism of flurbiprofen when CYP2C9 was in solution with lipid (Fig. 4A) or when CYP2C9 was immobilized (Fig. 4B) with P450(total)/CPR ratios of 1:8. In both cases, CPR was added prior to the addition of CYP3A4.
Effect of CYP3A4 on CYP2C9-mediated metabolism of flurbiprofen. (A) Metabolite formation in solution incubations containing CYP3A4 and CYP2C9, reconstituted in lipid, and with saturating CPR [P450(total)/CPR, 1:8], incubated for 16 hours. (B) Metabolite formation in incubations containing CYP3A4 (solution), CYP2C9i-mediated flurbiprofen metabolism at saturating CPR [P450(total)/CPR, 1:8], incubated for 16 hours. Single-factor analysis of variance was used for statistical comparisons. 2C9, CYP2C9; 2C9i, immobilized CYP2C9; 3A4, CYP3A4.
Effect of CYP2D6 on CYP2C9-mediated flurbiprofen metabolism. (A) Effect of CYP2D6 on solution metabolism of CYP2C9 reconstituted in lipid with saturating CPR [P450(total)/CPR, 1:8], incubated for 16 hours. Significant activation is observed at a 2C9/2D6 ratio of 1:0.25 (P < 0.05) and for 2C9/2D6 ratios higher than 1:0.25 (P < 0.01). (B) Effect of CYP2D6 on CYP2C9i-mediated flurbiprofen metabolism at saturating CPR [P450(total)/CPR, 1:8], incubated for 16 hours. Significant inhibition (P < 0.05) is seen for a 2C9i/2D6 ratio of 1:1. Single-factor analysis of variance was used for statistical comparisons. 2C9, CYP2C9; 2C9i, immobilized CYP2C9; 2D6, CYP2D6. *P < 0.05; **P < 0.01.
In contrast, the presence of CYP2D6 significantly affected CYP2C9-mediated metabolism of flurbiprofen, as suggested by the data shown in Fig. 5. In particular, in solution, CYP2C9-mediated metabolism of flurbiprofen was enhanced by the presence of CYP2D6 and in a concentration-dependent manner when CPR was present at saturating concentrations [P450(total)/CPR, 1:8] and added prior to the addition of CYP2D6 (Fig. 5A). This activation saturated at a CYP2C9/CYP2D6 ratio of 1:4 and increased metabolite production by approximately 400% (Supplemental Fig. 4). In contrast, CYP2C9i-mediated metabolism of flurbiprofen in the presence of CYP2D6 resulted in a concentration-dependent inhibition of CYP2C9-mediated metabolism of flurbiprofen (Fig. 5B). CYP2D6 inhibition of CYP2C9i metabolism saturated at 75% inhibition and did not change up to a ratio of CYP2C9i/CYP2D6 of 1:8 (Supplemental Fig. 5).
Influence of the Order of Addition of Enzymes on CYP2C9-Mediated Metabolism.
Figure 6 displays the amount of metabolite formed as a function of the order of addition of either CYP3A4 or CYP2D6 and CPR from CYP2C9-mediated metabolite production. Figure 6A shows that there is no effect of CYP3A4 on CYP2C9-mediated metabolism when CPR is added first and allowed to preincubate, which is in agreement with the results presented in Fig. 4A. In contrast, when CYP3A4 is allowed to preincubate with CYP2C9, a significant inhibition (up to 34%) of CYP2C9-mediated metabolism of flurbiprofen at a CYP2C9/CYP3A4 ratio of 1:2 is observed. The effect on metabolite production by CYP2D6 is depicted in Fig. 6B, which shows that the presence of CYP2D6 strongly inhibits CYP2C9 metabolism (up to 98%) regardless of the order in which the components are added prior to initiating the incubation. There is no significant difference between metabolite production at the 1:2 CYP2C9/CYP2D6 ratio when CPR is added before or after CYP2D6.
Order of addition of CPR and CYP3A4 or CYP2D6 effect on CYP2C9-mediated flurbiprofen metabolism. (A) Effect of CYP3A4 on solution metabolism of CYP2C9 reconstituted in lipid-lowered CPR concentration [P450(total)/CPR, 1:1.5], incubated for 1 hour. The column labeled “1 2C9 : 2 3A4 CPR” represents preincubation of CYP2C9 and CPR for 5 minutes prior to CYP3A4 addition. All other columns represent CYP3A4 preincubation for 5 minutes prior to addition of CPR. A significant inhibition (P < 0.01) is seen at all 2C9/3A4 ratios when CYP3A4 is preincubated but not when CPR is preincubated. (B) Effect of CYP2D6 on solution metabolism of CYP2C9 reconstituted in lipid with lowered CPR concentration [P450(total)/CPR, 1:1.5], incubated for 1 hour. The column labeled “1 2C9 : 2 2D6 CPR” represents preincubation of CYP2C9 and CPR for 5 minutes prior to CYP2D6 addition. All other columns represent CYP2D6 preincubation for 5 minutes prior to addition of CPR. Significant inhibition (P < 0.01) is seen in the presence of CYP2D6. There was no significant difference between preincubating the CPR or CYP2D6 with CYP2C9 at a 2C9/2D6 ratio of 1:2. Single-factor analysis of variance was used for statistical comparisons. 2C9, CYP2C9; 2D6, CYP2D6; 3A4, CYP3A4. **P < 0.01.
In comparison, Fig. 7 represents the same order-of-addition studies as mentioned earlier using CYP2C9i. Figure 7A shows that there is a significant decrease (up to 60%) of CYP2C9i-mediated metabolism when CYP3A4 is allowed to preincubate prior to the addition of CPR compared with when CYP2C9i is preincubated with CPR prior to the addition of CYP3A4. This result is in agreement with what is observed in order-of-addition studies of CYP2C9 and CYP3A4 in solution (Fig. 6A). When CYP2D6 was allowed to preincubate, we also saw a significant decrease (up to 46%) of CYP2C9i-mediated metabolism. This result is in contrast with solution studies that showed that order of addition had no effect on CYP2D6 inhibition of CYP2C9-mediated metabolism (Fig. 6B), and these findings may be attributable to a different binding conformation of CYP2C9-CYP2D6 between solution and immobilized studies.
Order of addition of CPR and CYP3A4 or CYP2D6 effect on CYP2C9i-mediated flurbiprofen metabolism. (A) Effect of CYP3A4 on CYP2C9i metabolism in incubations containing CYP3A4 (solution) at saturating CPR [P450(total)/CPR, 1:8], incubated for 16 hours. The column labeled “1 2C9i : 8 3A4 CPR” represents preincubation of CYP2C9 and CPR for 5 minutes prior to CYP3A4 addition. The column labeled “1 2C9i : 8 3A4” represents CYP3A4 preincubation for 5 minutes prior to addition of CPR. Preincubation of CYP3A4 results in a significant inhibition (P < 0.05) compared with CPR preincubation. (B) Effect of CYP2D6 on CYP2C9i metabolism in incubations containing CYP2D6 (solution) at saturating CPR [P450(total)/CPR, 1:8], incubated for 16 hours. The column labeled “1 2C9i : 2 2D6 CPR” represents preincubation of CYP2C9 and CPR for 5 minutes prior to CYP2D6 addition. The column labeled “1 2C9i : 2 2D6” represents CYP2D6 preincubation for 5 minutes prior to addition of CPR. Preincubation of CYP2D6 results in a significant inhibition (P < 0.05) compared with CPR preincubation. Single-factor analysis of variance was used for statistical comparisons. 2C9i, immobilized CYP2C9; 2D6, CYP2D6; 3A4, CYP3A4. *P < 0.05.
Discussion
Protein aggregation between P450s occurs in solution and can alter P450-mediated metabolism. Here, we show the formation of homomeric (CYP2C9-CYP2C9) and heteromeric (CYP2C9-CYP3A4, CYP2C9-CYP2D6, and CYP2C9-CPR) complexes occurs with different binding affinities. We reveal for the first time activation of CYP2C9 metabolism by homomeric dimerization. Furthermore, we show that controlling the aggregation state of CYP2C9 changes the effect that P450-P450 interactions have on metabolism. The immobilization used here offers an artificial mimic of in vivo P450 tethering (Gannett et al., 2006), which is of particular importance for more accurate insight on bound P450 function. Finally, we show that the order of addition and the concentration of CPR both modulate inhibition of CYP2C9-mediated metabolism by CYP2D6 and CYP3A4 in lipid-reconstituted and immobilized metabolism studies.
Computational modeling was used to predict dimer formation and probe how the complex may affect P450-P450 modulation of metabolism (Fig. 1). Docking indicated that dimer formation rarely involved residues necessary for CPR binding (Table 1), implying that the CPR binding site is not involved in modulating metabolism as a result of P450-P450 interactions. Instead, CYP2C9 metabolism is more likely to be affected due to dimerization occurring at the solvent channel for dimers with CYP2C9, CYP3A4, and CYP2D6, thus possibly changing the binding or turnover rate. This finding is in agreement with previous literature findings of dimerization occurring with residues near the F-G loop (Ozalp et al., 2006; Hu et al., 2010).
MM/GBSA calculations suggest that, in all cases, dimer formation is energetically favorable and is most favorable for CYP2C9-CPR (Table 2). However, none of the models resulting from blind docking between CYP2C9 and CPR involved the residues required for a functional CYP2C9-CPR complex. Others have also found that blind docking fails to generate a functional P450-CPR complex and have resorted to directed docking or homology modeling to achieve a functional complex (Hazai et al., 2005; Hamdane et al., 2009). Thus, the inability to achieve a correct model by docking may be related to the differences seen in the rank order between the modeling and in vitro studies (Table 3). This demonstrates that although molecular modeling can be a useful tool, results must be confirmed in vitro.
Recent work has demonstrated that P450-P450 interactions occur in membranes utilizing crosslinking (Hu et al., 2010; Reed et al., 2010; Davydov et al., 2015), coimmunoprecipitation (Alston et al., 1991; Subramanian et al., 2010), and fluorescent resonance energy transfer studies (Davydov et al., 2001; Szczesna-Skorupa et al., 2003; Ozalp et al., 2005). It is important to quantify the rates associated with these interactions and determine whether they may be involved in modulating P450 activity. Here, we used SPR (Ivanov et al., 2001) to measure binding of an analyte P450 to CYP2C9i in real time. The KD values determined for the binding of CPR to CYP2C9 are in line with reported values of 2.3 ± 1.0 nM (Wei et al., 2007) and 32.8 ± 0.2 nM (Locuson et al., 2007), demonstrating the validity and accuracy of the method. The high affinity binding that occurs between CYP2C9-CYP2C9 and CYP2C9-CYP2D6 is likely relevant to the results of metabolism assays, because they bind to one another more strongly than either does to CPR (Table 3). We also show that CYP3A4 forms a heteromeric complex with CYP2C9, but its KD (2.5 times that of CPR; Table 3) suggests that CYP3A4 competition for the CPR binding site is unlikely to factor into metabolism effects, although it is possible that CYP3A4 may alter CYP2C9’s affinity for CPR. Furthermore, since the SPR immobilization scheme is the same as used for the metabolism studies, correlation of complex formation with metabolic effects is valid.
P450 aggregation can affect metabolism by altering CPR affinity, substrate binding, or turnover rate (Reed and Backes, 2012). Altered CPR affinity can occur by blocking association of CPR, or through conformational changes that result from dimerization. Here, we use our controlled immobilization platform to show, for the first time, the activation of CYP2C9i metabolism through homomeric complex formation (Fig. 3). Although it is possible that dimerization of CYP2C9 restores activity to CYP2C9 that is in solution (lipid free), we would expect metabolite generation to saturate at the highest level if this is the case; yet a decrease, at ratios greater than 1:2, is seen (Fig. 3). We do note, however, that the observed decrease could be a result of higher-order assemblages of CYP2C9 in solution at higher concentrations preventing interaction with CYP2C9i. Further studies are necessary to definitively prove that the increased metabolism results from activation due to CYP2C9 dimerization. Given work by Hu et al. (2010) demonstrating that CYP2C8 exists as a dimer in membranes, we believe that it is possible that dimerization of CYP2C9 is necessary for the most catalytically active species. Our modeling data (Table 1) indicate that interaction with solvent access channels may implicate dimerization promoting substrate binding and exit, which could lead to higher substrate turnover rates (Stepankova et al., 2013).
In solution, it has been shown that interactions between CYP2C9 and CYP3A4 give a protein concentration-dependent inhibition of CYP2C9 metabolism if incubated with CYP3A4 prior to CPR addition (Subramanian et al., 2010). Here, we show that CYP2C9 preincubation with CPR prevents inhibition by CYP3A4 (Fig. 4) regardless of CPR concentration (Fig. 6A). Furthermore, the work by Subramanian et al. (2010) showed that the CYP2C9-CYP3A4 interaction could be eliminated by truncation of the N terminus of either binding partner. We show that if both P450s are truncated, interaction can still occur (Table 3) and this binding can result in decreased CYP2C9 metabolism if preincubated with CYP3A4 (Fig. 6A; Fig. 7A). The order of addition of protein partners has been shown to be important when evaluating the effects of P450-P450 interactions (Gorsky and Coon, 1986; Subramanian et al., 2009). The order-of-addition effect on CYP3A4 inhibition points to the formation of a quasi-irreversible heteromeric complex (Fig. 6A; Fig. 7A).
Studies have also shown there is a protein concentration-dependent inhibition of CYP2C9 metabolism when coincubated with CYP2D6 at subsaturating CPR concentrations (Subramanian et al., 2009). Remarkably, we note drastically different effects of CYP2D6 on CYP2C9 metabolism depending on study parameters. Our result of 94% inhibition of CYP2C9 in solution by CYP2D6 at a 1:1 ratio (Fig. 6B) is in reasonable agreement with the 50%–80% reported previously (Subramanian et al., 2009). The difference in maximum inhibition could possibly be due to slight differences in lipid/protein ratios between the two studies. However, we show that at saturating CPR concentrations, metabolism by CYP2C9 in solution is activated by CYP2D6 (Fig. 5A). Even more compelling is that when CYP2C9 is immobilized, CYP2D6 is once again inhibited despite an excess of CPR allowed to preincubate with the CYP2C9i (Fig. 5B).
One possible explanation for the solution metabolism results is a dual effect of CYP2D6 on CYP2C9 metabolism. If the CYP2D6-CYP2C9 complex has a faster substrate turnover rate, but lower CPR affinity, we would expect to see activation of metabolism when CPR is in excess and a faster turnover rate dominating the kinetics, but we would expect to see inhibition when the CPR concentration is lower and electron transfer becomes a limiting step. Another factor to consider is the difference between solution and immobilized platform results (Fig. 5). Immobilization of CYP2C9 controls its aggregation, with the majority being monomeric (Gannett et al., 2006; Bostick et al., 2015). Therefore, the inhibition of CYP2C9 metabolism could be a result of the lack of competition for CYP2C9 homomeric complex (CYP2C9-CYP2C9) formation with heteromeric complex (CYP2C9-CYP2D6) formation, evidenced by the similar KD values for both interactions (Table 3). On the immobilized platform, the binding modes and conformations may also be limited or otherwise different compared with solution and CYP2D6 may prevent CPR from associating with CYP2C9, thus limiting metabolism. In this case, the lack of oligomerization when immobilized leads to more catalytically inactive CYP2C9-CYP2D6 complexes. This agrees with the hypothesis of Davydov et al. (2015) that complexed P450 species bind and metabolize differently depending on the degree of oligomeriziation. Furthermore, in solution, the order of addition had no effect on CYP2D6 inhibition of CYP2C9 metabolism (Fig. 6B), but it did result in increased inhibition of CYP2C9i metabolism (Fig. 7B). This suggests, but does not conclusively prove, that CYP2D6 binds in a different conformation that results in a quasi-irreversible complex when CYP2C9 is immobilized. Although changes in CYP2C9 metabolism through complexation are clearly observed in our study, it is important to note that the mechanisms of activation (Fig. 3; Fig. 5A) and inhibition (Fig. 5B; Figs. 6 and 7) remain ambiguous since metabolite formation was nonlinear over the extended time course of the studies.
This work demonstrates the feasibility of using an immobilized platform to better understand P450-P450 interactions. We propose that the immobilization of CYP2C9 may offer a more complete functional assessment of CYP2C9 dimers, due to its bound nature, as seen in vivo, and control of aggregation state without detergents. Being able to immobilize an enzymatically active P450, allows study of both P450-P450 complex formation and the resultant metabolic modulation. We show that CYP2C9 forms both homomeric and heteromeric complexes and that these complexes may activate or inhibit CYP2C9 metabolism. Further study is necessary to fully analyze the conformation and activity of immobilized P450s to offer complete comparison of their functions to that of in vivo P450s. Future studies using this platform may further elucidate the intricacies of P450-P450 interactions and can be expanded to probe whether the effects are due to changes in CPR affinity or substrate turnover.
Acknowledgments
The authors thank A.R. Biundo and E.L. Dolan for their assistance in running experiments. They also acknowledge the West Virginia University Shared Research Facilities for support in completing this work.
Authorship Contributions
Participated in research design: Bostick, Hickey, Wollenberg, Gannett.
Conducted experiments: Bostick, Hickey, Wollenberg.
Contributed new reagents and analytic tools: Flora, Tracy.
Performed data analysis: Bostick, Hickey, Wollenberg.
Wrote or contributed to the writing of the manuscript: Bostick, Hickey, Wollenberg, Flora, Tracy, Gannett.
Footnotes
- Received October 2, 2015.
- Accepted March 9, 2016.
All authors are co-first authors.
This research was supported by West Virginia University via the National Science Foundation through the Experimental Program to Stimulate Competitive Research [Grant EPS-1003907]; the National Institutes of Health National Institute of General Medical Sciences [Grant R01-GM086891]; and the National Science Foundation Integrative Graduate Education and Research Traineeship Program [Grant DGE-1144676].
dx.doi.org/10.1124/dmd.115.067637.
↵
This article has supplemental material available at dmd.aspetjournals.org.
Abbreviations
- CPR
- cytochrome P450 reductase
- DLPC
- dilauroylphosphatidylcholine
- EDC
- N-(3-dimethylaminopropyl)-N′-ethylcarbodiimide hydrochloride
- GBSA
- generalized born surface area, MM, molecular mechanics
- MUA
- 11-mercaptoundecanoic acid
- NHS
- N-hydroxysulfosuccinimide sodium salt
- OT
- 1-octanethiol
- P450
- cytochrome P450
- PBS
- phosphate-buffered saline
- PDB
- Protein Data Bank
- SAM
- self-assembled monolayer
- SPR
- surface plasmon resonance
- Copyright © 2016 by The American Society for Pharmacology and Experimental Therapeutics
References
In this issue
P450 Immobilization to Probe P450-P450 Interactions
