Biochemical Journal

Research article

Pivotal role of P450–P450 interactions in CYP3A4 allostery: the case of α-naphthoflavone

Dmitri R. Davydov , Nadezhda Y. Davydova , Elena V. Sineva , Irina Kufareva , James R. Halpert


We investigated the relationship between oligomerization of CYP3A4 (cytochrome P450 3A4) and its response to ANF (α-naphthoflavone), a prototypical heterotropic activator. The addition of ANF resulted in over a 2-fold increase in the rate of CYP3A4-dependent debenzylation of 7-BFC [7-benzyloxy-4-(trifluoromethyl)coumarin] in HLM (human liver microsomes), but failed to produce activation in BD Supersomes™ or Baculosomes® containing recombinant CYP3A4 and NADPH-CPR (cytochrome P450 reductase). However, incorporation of purified CYP3A4 into Supersomes™ containing only recombinant CPR reproduced the behaviour observed with HLM. The activation in this system was dependent on the surface density of the enzyme. Although no activation was detectable at an L/P (lipid/P450) ratio ≥750, it reached 225% at an L/P ratio of 140. To explore the relationship between this effect and CYP3A4 oligomerization, we probed P450–P450 interactions with a new technique that employs LRET (luminescence resonance energy transfer). The amplitude of LRET in mixed oligomers of the haem protein labelled with donor and acceptor fluorophores exhibited a sigmoidal dependence on the surface density of CYP3A4 in Supersomes™. The addition of ANF eliminated this sigmoidal character and increased the degree of oligomerization at low enzyme concentrations. Therefore the mechanisms of CYP3A4 allostery with ANF involve effector-dependent modulation of P450–P450 interactions.

  • cytochrome P450 3A4
  • luminescence resonance energy transfer (LRET)
  • microsome
  • oligomerization
  • protein–protein interaction
  • α-naphthoflavone


CYP3A4 (cytochrome P450 3A4, where CYP refers to cytochrome P450), the major drug-metabolizing enzyme in human liver, is the most prominent example of a P450 that exhibits so-called ‘atypical kinetics’. These kinetic ‘abnormalities’ are revealed in S-shaped (sigmoidal) dependencies of the rate of metabolism on substrate concentration (homotropic co-operativity) or activation of metabolism of one substrate by a second one (heterotropic co-operativity). The most common interpretation of all known examples of CYP3A4 co-operativity is that the large substrate-binding pocket of microsomal drug-metabolizing cytochrome P450 (also referred to in the present paper as P450) sometimes requires more than one substrate molecule to assure a productive binding orientation of at least one of them (for review, see [13]). The presence of at least two molecules of some substrates in the binding pocket of CYP3A4 is well established [46]. The above model also provides a reasonable explanation for most cases of homotropic co-operativity, but fails to explain complex instances of heterotropic activation in CYP3A4. Attempts to delineate the mechanism of additive effects of different effectors or lack of competition between two substrates, with each possessing homotropic co-operativity, have led researchers to propose complex and mutually incompatible models involving the presence of three or even more substrate-selective binding sites in one enzyme molecule [711].

In our view, one of the most important obstacles to understanding P450 activation is that all observations of co-operativity, whether homotropic or heterotropic, are considered together and presumed to have similar mechanistic grounds. However, the actual experimental observations diverge into two different categories. On the one hand, there are numerous observations of co-operative effects on enzyme–substrate interactions that are reflected in S-shaped dependencies of the rate of metabolism on substrate concentration (homotropic co-operativity) or increase in enzyme affinity for one substrate in the presence of another. In the latter kind of heterotropic co-operativity, the effector increases the activity of the enzyme at low substrate concentrations, but has no effect on turnover or even inhibits the enzyme at saturating substrate. Such examples are best represented by the effect of ANF (α-naphthoflavone; also known as 7,8-benzoquinone) on hydroxylation of steroids, such as progesterone or testosterone [1216].

On the other hand, there are numerous cases of strong effects of activators, such as ANF [1719], steroids [9,20] or quinidine [8,2124] on (Vmax or kcat) that do not necessarily involve any effect on enzyme–substrate interactions. Thus the addition of ANF to CYP3A4-containing microsomal systems was shown to cause a dramatic (12–45-fold) enhancement in metabolism of phenanthrene [17], phenacetin [18] or benzo[a]pyrene [19] measured under saturating concentrations. Similarly, quinidine causes a multifold increase in the rate of CYP3A4-dependent metabolism of substrates such as diclofenac, R-warfarin, piroxicam [23,24] and methoxicam [22].

From our perspective, the examples where no change in Vmax is observed are fundamentally different from the latter case of ‘true’ heterotropic activation. The first type of co-operativity reflects simple additive effects of multiple substrate molecules bound in a large substrate-binding pocket and, according to Sligar and coworkers [2,16], involves no specific allosteric effects. However, the observations of multifold increases in enzyme turnover may be better explained with a model involving effector-induced redistribution of a pool of P450 conformers with different activity and substrate specificity [1,19]. This hypothesis is supported by the modulatory effect of allosteric activators such as ANF on the partitioning between CYP3A4 conformers revealed in the kinetics of CO-rebinding after flash photolysis [19] or dithionite-dependent reduction [25] in enzyme oligomers.

Implicit in these results are slow transitions between persistent CYP3A4 conformers, so that the distribution of the enzyme between populations remains unchanged within the time frame of the experiments. The cause of such apparent ‘freezing’ of conformational states of the enzyme remains puzzling. From our perspective, oligomerization of cytochrome P450 in the microsomal membrane and in reconstituted systems (see [26,27] for reviews) is the most viable explanation for such persistent heterogeneity [1]. According to our hypothesis, the functional heterogeneity of the CYP3A4 pool is caused by diverse conformations and/or orientation of the subunits in the enzyme oligomer. Thus the instances of ‘true’ heterotropic activation in CYP3A4 reveal an allosteric transition associated with a change in the degree of oligomerization and/or alteration of the architecture of the oligomer.

The present study probes the above concept through investigation of the interrelationship between the activating effect of ANF and the degree of CYP3A4 oligomerization in microsomal membranes. We studied the dependence of the effect of ANF on kinetic parameters of O-debenzylation of 7-BFC [7-benzyloxy-4-(trifluoromethyl)coumarin] by CYP3A4 on the surface density of the enzyme in model microsomal membranes. These studies were complemented with an examination of the effect of CYP3A4 concentration in the membrane on the degree of its oligomerization assessed with a new LRET (luminescence resonance energy transfer)-based method. The results of the present study demonstrate a strong parallelism of the degree of P450 oligomerization with its susceptibility to activation by ANF. We also demonstrate a marked effect of ANF on the equilibrium of P450 oligomerization and architecture of the enzyme oligomers. Taken together, our results establish P450–P450 interactions as a central element in the mechanism of CYP3A4 allostery.



HFC [7-hydroxy-4-(trifluoromethyl)coumarin], glucose 6-phosphate, glucose-6-phosphate dehydrogenase from baker's yeast, protocatechuate 3,4-dioxygenase from Pseudomonas sp., protocatechuic acid, NADPH and l-α-PC (phosphatidylcholine) from egg yolk were from Sigma–Aldrich. l-α-PE (phosphatidylethanolamine) was from bovine liver and 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphate (PA; phosphatidic acid) were obtained from Avanti Polar Lipids. 7-BFC and CYP3A4 Baculosomes® Plus (catalogue number P2377) were obtained from Invitrogen. CYP3A4 BD Supersomes™ containing human recombinant CYP3A4 and NADPH-CPR (cytochrome P450 reductase) (catalogue number 456207) and control BD Supersomes™ containing rat recombinant NADPH-CPR (catalogue number 456514) were from BD Biosciences. The preparation of HLM (human liver microsomes) Xtreme-200, which represents a pool of 200 donors of both genders, was obtained from Xenotech. ERIA (erythrosine 5′-iodoacetamide) was from Anaspec and BODIPY-PC (2-decanoyl-1-{O[11-(4,4-difluoro-5,7-dimethyl-4-bora-3a,4a-diaza-s-indacence-3-propionyl)amino]undecyl}-snglycero-3-phosphocholine) was from Invitrogen/Molecular Probes. DYM (DY-731 maleimide) was obtained from Dyomics. Octyl glucoside (octyl-β-D-glucopyranoside) and ANF were from Fluka and Indofine Chemical Company respectively. All other chemicals were of ACS grade and were used without further purification.

Expression and purification of CYP3A4 and its mutants

Wild-type CYP3A4 and its cysteine-depleted mutants CYP3A4(C58,C64) [28] and CYP3A4(C468) [29] were expressed as His-tagged proteins in Escherichia coli TOPP3 cells and purified as described previously [30].

Modification with thiol-reactive probes

Before modification of the cysteine-depleted mutants with thiol-reactive probes, we eliminated TCEP [tris-(2-carboxyethyl)phosphine] contained in the storage buffer by two repetitive 1:10 dilution/concentration cycles with the use of a Centrisart I MWCO (molecular-mass cut-off) 100 kDa concentrator (Sartorius). The labelling was performed by incubation of a 15–20 μM solution of the protein in 0.1 M Hepes sodium buffer, pH 7.4, containing 150 mM KCl, 0.5 mM EDTA and 20% (v/v) glycerol (buffer A) with an thiol-reactive probe added at a 1.1:1 molar ratio to the protein with constant stirring under an argon atmosphere at ambient temperature (21–23°C) for 90–120 min. Saturation of the reaction media with argon was achieved with bubbling (2–3 bubbles/s) of the argon gas. After 5 min of bubbling, the protein and the label were injected into a closed reaction vessel with a gas-tight syringe. The process of modification was monitored by a decrease in the fluorescence of the label, which is due to FRET (fluorescence resonance energy transfer) to the haem of P450. The reaction was terminated by addition of DTT to a final concentration of 3 mM. The DTT adduct of unreacted probe was removed from the concentrated samples by incubation with Bio-Beads SM-2 (Bio-Rad Laboratories) followed by gel filtration on Bio-Spin 6 spin columns (Bio-Rad Laboratories) equilibrated with buffer A.

Preparation of proteoliposomes

Proteoliposomes were obtained by incorporation of CYP3A4 into pre-formed liposomes prepared with the octyl glucoside/dialysis technique described previously [31]. Specifically, we used a 2:1:0.6 mixture of PC, PE and PA with addition of 2 μg of BODIPY-PC per mg of lipid mixture. BODIPY-labelled phospholipid was included in the mixture for easy detection of the liposomes during their separation by gel filtration. Lipids (10 mg) were mixed as chloroform solutions, and the solvent was removed by evaporation under a stream of argon gas and subsequent drying in a vacuum for 2 h. The suspension of lipids in buffer A, containing 1.54% octyl glucoside, was prepared using a vortex-mixer and incubated for 30 min at room temperature (21–23°C) under argon atmosphere maintained as described above. The mixture was then diluted with the same buffer containing no detergent to a final octyl glucoside concentration of 0.43%. The mixture was dialysed at 4°C under constant gentle bubbling with argon gas against three changes of 1000 ml of buffer A, each containing 5 ml of Bio-Beads SM-2. After 72 h of dialysis (24 h per portion of the buffer), the mixture was concentrated on MWCO 300 kDa Diaflo membranes (Millipore) to a phospholipid concentration of 8–15 mM and stored at −80°C under argon conditions. To incorporate cytochrome P450 into pre-formed liposomes, a solution of purified CYP3A4 (100–150 μM) was added to an 8 mM suspension of the liposomes in buffer A containing 1 mM DTT to reach a desired protein-to-lipid molar ratio. The mixture was incubated overnight (~16 h) with constant stirring under an argon atmosphere at 4°C. Separation of unbound protein by gel-exclusion chromatography on Toyopearl HW 75F resin [32] demonstrated quantitative incorporation of the enzyme into the liposomes at an RL/P (lipid/P450 ratio) ≥70. The molar ratio of phospholipids to cytochrome P450 in the final preparation was estimated on the basis of the determination of total phosphorus in a chloroform/methanol extract according to the method described by Bartlett [33]. The proteoliposomes were stored at −80°C under argon atmosphere.

Incorporation of CYP3A4 into model microsomes

Incorporation of CYP3A4 into model microsomes was performed by incubation of an undiluted suspension of various commercial preparations of Baculosomes® or Supersomes™ (5–7 mg/ml protein, 2–3.5 mM phospholipid) containing purified CYP3A4 and phospholipid at molar ratios of 1:2000 to 1:100 for 16 h at 4°C with continuous stirring. After incubation, the suspension was diluted 8-fold with 100 mM potassium phosphate buffer (pH 7.4) and centrifuged at 35 000 rev./min in an Optima XL-80XP ultracentrifuge (Beckman Coulter) with a SW50L rotor for 90 min at 4°C. The pellet was resuspended in the same buffer (200 μl per 500 μl of initial microsomal suspension) and briefly sonicated (two rounds of 10 s at 40% power) with a BioLogics Model 30000 sonicator with a micro-tip. The amount of incorporated cytochrome P450 was calculated from the difference between the haem protein added to the incubation medium and the enzyme found in the supernatant. The content of phospholipids was estimated as described above for proteoliposomes.

Activity measurements

The rate of 7-BFC O-debenzylation was measured by a real-time continuous fluorimetric assay. A suspension of microsomes was added to 300 μl of 0.1 M Hepes sodium buffer (pH 7.4), to a final P450 concentration of 0.02–0.1 μM and placed into a 5 mm×5 mm quartz cell with continuous stirring and thermostatically controlled at 25°C. An aliquot of a 20 mM stock solution of 7-BFC in acetone was added to attain the desired concentration in the range 1–100 μM. The reaction was initiated by the addition of 10 μl of a cocktail of NADPH, glucose 6-phosphate and glucose-6-phosphate dehydrogenase, which resulted in final concentrations of 200 μM, 10 mM and 2 units/ml of these ingredients respectively. The increase in the concentration of the product (HFC) was monitored with a Cary Eclipse (Varian, Agilent Technologies) or computerized Hitachi F2000 (Hitachi) spectrofluorimeter equipped with a thermostatically contolled cell holder and a magnetic stirrer [34], using an emission wavelength of 500 nm and excitation at 404 nm. The rate of formation of HFC was estimated by determining the slope of the linear part of the kinetic curve recorded over a period of 3–6 min. Calibration of the assay was performed at the end of each day by measuring the intensity of fluorescence in a series of four to five samples of the same reaction mixture containing HFC at concentrations from 0.5 to 3 μM.

Phosphorescence spectroscopy measurements

Phosphorescence spectroscopy measurements were performed with the use of a Cary Eclipse spectrofluorometer equipped with a Peltier four-position cell holder or with a FLS920 fluorescence spectrometer (Edinburgh Instruments) equipped with Model 63501 Xenon flash lamp illuminator (Oriel Instruments, Newport Corporation) and custom software for acquisition of delayed phosphorescence spectra. The excitation of donor phosphorescence was performed with monochromatic light centred at 540 nm with 20 nm bandwidth (Cary Eclipse) or with a broadband light passed through a 560 nm short-pass filter (FLS920). The spectra in the 620–850 nm wavelength region were recorded with 18–20 nm emission bandwidth and 100 μs (Eclipse) or 175 μs (FLS920) delay time.

Interactions of DYM-labelled CYP3A4 with erythrosine-labelled protein

Interactions of DYM-labelled CYP3A4 with erythrosine-labelled protein were judged from kinetics of the increase in the amplitude of LRET assessed from a series of spectra of delayed emission taken at 4-min time intervals during co-incubation of DYM-labelled proteins with membranes containing the erythrosine-labelled enzyme. The incubation was performed at 4°C under anaerobiosis, which was achieved by addition of the oxygen-scavenging enzyme protocatechuate 4,5-dioxygenase (0.5 unit/ml) and protecatechuic acid (20 mM). The experiments with liposomes were performed in 0.1 M Hepes sodium buffer (pH 7.4), containing 150 mM KCl and 20% glycerol. In the case of Supersomes™, the incubation buffer was 0.1 M potassium phosphate (pH 7.4). The total haem protein concentration in the incubation medium was in the range 2–5 μM.

Analysis of series of spectra

Analysis of series of spectra obtained in fluorescence spectroscopy experiments was done by PCA (principal component analysis) [35,36]. To resolve the changes in the fluorescence of the donor and acceptor, we used a least-squares fitting of the spectra of the first principal components by a combination of the standard spectra of emission of CYP3A4 labelled with ERIA and DYM. All data treatment procedures and curve fitting were performed using a 32-bit version of our SPECTRALAB software [35] running under Windows XP™.

P450 structure and binding pocket analysis

P450 structure analysis was performed with the ICM (Internal Coordinate Mechanics) software package [37] using data collected in the Pocketome database [38]. For the analysis of P450–P450 interactions, each P450 enzyme molecule in the PDB was studied in the context of its crystallographic neighbours including other P450 molecules in the ASU (asymmetric unit) as well as the symmetry-related molecules. Binding pocket analysis was performed using the ICM PocketFinder tool [39]. The algorithm is based on contouring of a transformed Lennard-Jones potential calculated from a 3D protein structure. The method predicts the likelihood of small-molecule binding in a particular location from the size and shape of the calculated envelope. Compound docking was performed with the ICM Ligand Editor tool. Fully flexible ligands were docked in the receptor-binding pocket represented as a set of potential grid maps. The obtained poses were combined with the full-atom model of the binding pocket and re-scored using the ICM full-atom ligand-binding score.


Effect of ANF on O-debenzylation of 7-BFC in HLM and in model microsomes containing recombinant CYP3A4

Consistent with our previous studies in a micellar reconstituted system [14], we observed a clear effect of ANF on 7-BFC metabolism in HLM (Figure 1). Fitting the dependence of the reaction rate on substrate concentration in the absence of ANF with the Hill equation yielded an S50 value of 15.8 μM and a Hill coefficient (h) of 1.4. Addition of 25 μM ANF caused a notable decrease in S50 and eliminated co-operativity of 7-BFC metabolism (Figure 1a and Table 1). These changes were associated with a greater than 2-fold increase in kcat.

Figure 1 Effect of ANF on 7-BFC O-debenzylation in HLM and in CYP3A4 Baculosomes® and CYP3A4 Supersomes™

(a) HLMs, (b) CYP3A4 Baculosomes® (○ and ●) and CYP3A4 Supersomes™ (∆ and ▲). The titration curves obtained in the absence of effector and in the presence of 25 μM ANF are indicated by open and closed symbols respectively. The lines represent the approximations of the datasets with the Hill equation.

View this table:
Table 1 Effect of ANF on O-debenzylation of 7-BFC in various microsomal systems

The values given in the Table were obtained by averaging the results of two to four individual measurements and the ‘±’ values show the confidence interval calculated for P=0.05. The values in parentheses given in the columns for S50, h and kcat represent the P values of the Student's t test for the hypothesis of equality of the respective values measured in the presence of ANF with those observed without ANF. P values ≤0.05 are underlined to emphasize the effects with high statistical significance. *Ratio of molar content of (total) cytochrome P450 to the content of CPR estimated from the rate of NADPH-dependent reduction of cytochrome c. †Preparation of CYP3A4 Baculosomes® enriched with CYP3A4 by co-incorporation of the purified enzyme (see the Materials and methods section). +, values obtained in the presence of 25 μM ANF; −, values obtained in the absence of 25 μM ANF.

We also studied the effect of ANF on 7-BFC metabolism in two commercial preparations of insect cell microsomes containing recombinant human CYP3A4 and NADPH-CPR, namely BD Supersomes™ and CYP3A4 Baculosomes® Plus. In contrast with the increase in kcat in HLM, some inhibition by ANF was observed in both Baculosomes® and Supersomes™ (Table 1). Furthermore, in contrast with HLM, in model microsomes ANF elicited no effect on either the h or S50.

In order to probe whether this difference between HLM and the model microsomes is related to differences in P450–P450 interactions and the degree of CYP3A4 oligomerization in particular, we studied the effect of incorporation of additional amounts of purified CYP3A4 into the CYP3A4-containing Baculosomes® or Supersomes™. As shown in Table 1 and Figure 1(b), the addition of ANF to CYP3A4-enriched Supersomes™, but not Baculosomes®, eliminated co-operativity with 7-BFC and decreased S50, similar to what was observed in HLM. However, neither of the two CYP3A4-enriched preparations revealed any statistically significant increase in kcat in the presence of ANF (Table 1).

Effect of ANF on O-debenzylation of 7-BFC in model microsomes with incorporated purified CYP3A4

In order to develop a model system that better reproduces the allosteric properties of CYP3A4 observed in HLM, we incorporated purified CYP3A4 into commercially available ‘control’ Supersomes™ containing recombinant CPR, but not P450. This approach allowed us to obtain model membranes with an RL/P varying from 140 to 2700. As show in Figures 2 and 3, the membranes with a high surface density of CYP3A4 (RL/P<500) closely reproduced the behaviour observed with HLM. Specifically, the addition of ANF eliminated the homotropic co-operativity of 7-BFC debenzylation and caused up to a 220% increase in kcat. Interestingly, the dependencies of h and kcat on the CYP3A4 surface density are represented with S-shaped curves having an inflection point at an RL/P value of ~700. The dependence of the activation by ANF on the surface density of CYP3A4 in the membrane supports our initial concept that the heterotropic co-operativity involves modulation of the quaternary structure of the oligomer and/or the degree of oligomerization.

Figure 2 Effect of ANF on 7-BFC O-debenzylation in CPR-containing Supersomes™ with purified CYP3A4 incorporated into the microsomal membrane

Effect of ANF on 7-BFC O-debenzylation in CPR-containing Supersomes™ with purified CYP3A4 incorporated into the microsomal membrane at an L/P ratio of 1350 (○ and ●) to 175 (∆ and ▲). The titration curves obtained in the absence of effector and in the presence of 25 μM ANF are indicated by open and closed symbols respectively. The lines represent the approximations of the datasets with the Hill equation.

Figure 3 Dependence of the effect of ANF on 7-BFC metabolism in CPR-containing Supersomes™ with CYP3A4 incorporated on the surface density of CYP3A4 in the membrane

(a) Effect of surface density of CYP3A4 on the h of 7-BFC metabolism in the absence of effector (○, light error bars) and in the presence of 25 mM ANF (□, dark error bars). (b) Effect of the P450 concentration in the membrane on the activation by 25 μM ANF as assessed by kcat. The error bars represent the S.D. calculated for two to three measurements.

Design of a method for detection of P450–P450 interactions

We next sought to develop a direct approach for studying the degree of CYP3A4 oligomerization in the membranes and the effect of allosteric effectors on P450–P450 interactions. For this purpose we employed resonance energy transfer between donor and acceptor probes incorporated into two separate preparations of the enzyme. According to our design the formation of mixed oligomers of the labelled enzyme molecules would result in resonance energy transfer with an amplitude proportional to the degree of oligomerization.

Selection of a donor/acceptor pair applicable in studies of haem proteins is challenging due to overlap of the haem absorbance with the emission bands of most commonly used fluorescent probes. In addition, the steady-state measurements of FRET are often obscured by the usual overlap of the excitation band of the donor with that of the acceptor. An attractive alternative to FRET is LRET, which uses the long-lived triplet state of a phosphorescent probe as an energy donor. Registration of the delayed fluorescence ensures a thorough selectivity in monitoring of the emission that originates from energy transfer. In our design, we employed ERIA as thiol-reactive phosphorescent probe. The phosphorescence of this dye is centred at 695 nm and has a lifetime of 0.1–0.2 ms. An appropriate acceptor dye, DY-731 maleimide (DYM), was found among near-infrared fluorescent tandem dyes in a previous study [40]. The spectral properties of this new donor/acceptor pair are illustrated in Figure 4. The R0 distance for this pair calculated assuming random rotation of at least one of the fluorophores (κ2=2/3) is equal to 33.6 Å (1 Å=0.1 nm), which is long enough to ensure efficient LRET between the dyes attached to neighbouring subunits in the enzyme oligomer.

Figure 4 Spectral properties of erythrosine and DY-731

Erythrosine and DY-731 are indicated by dark and light lines respectively. The continuous lines show the spectra of extinction. The broken lines show the spectra of fluorescence normalized to area.

In order to ensure site-directed incorporation of the probes, we used cysteine-depleted mutants bearing only two CYP3A4(C58,C64) or one CYP3A4(C468) potentially reactive cysteine residue [28,29]. As shown in our previous study [28], Cys58 has very poor accessibility for modification with large thiol-reactive probes. Therefore the stoichiometric labelling of CYP3A4(C58,C64) with ERIA is presumed to result from site-specific incorporation of the probe at Cys64, whereas CYP3A4(C468) can only be labelled at Cys468.

Studies of CYP3A4 oligomerization in proteoliposomes

To probe the applicability of the LRET donor/acceptor pair for monitoring P450–P450 interactions, we began with studies of CYP3A4 oligomerization in proteoliposomes. We used either CYP3A4(C58,C64)-ERIA or CYP3A4(C468)-ERIA as LRET donors, whereas CYP3A4(C468)-DYM was used as the LRET acceptor. We prepared large unilamellar liposomes containing either one of the above-mentioned erythrosine-labelled preparations of the enzyme at RL/P varying from 150 to 3000 (see the Materials and methods section). Incubation of these preparations with equimolar amounts of CYP3A4(C468)-DYM was expected to result in co-incorporation of this protein into the membrane and formation of mixed oligomers of donor- and acceptor-labelled proteins.

As shown in Figure 5(a), this incubation resulted in a profound decrease in the donor emission accompanied by appearance of the delayed fluorescence of the acceptor. Kinetic curves of the time-dependent decrease in the intensity of donor fluorescence observed at two different RL/P values (Figure 5a, inset) show clearly that the amplitude of LRET decreased with decreasing surface density of CYP3A4. Those observations are consistent with the time-dependent formation of mixed oligomers of the two proteins signified by the appearance of LRET. The amplitude of LRET established at the end of the kinetics of subunit exchange is therefore proportional to the steady-state concentration of the mixed oligomers in the membrane.

Figure 5 Interactions of ERIA- and DYM-labelled cysteine-depleted mutants in proteoliposomes studied with LRET

(a) A series of spectra of delayed emission recorded during incubation of a 5 μM suspension of CYP3A4(C58,C64)-ERIA incorporated into proteoliposomes at RL/P=150 with 5 μM CYP3A4(C468)-DYM. [The RL/P after incorporation of CYP3A4(C468)-DYM is equal to 75]. The inset shows the time dependencies of normalized intensity of donor fluorescence obtained in the experiments at an RL/P of 75 and 1350. The lines represent the approximation of the kinetic curves with a bi-exponential equation. (b) Dependence of the LRET amplitude on the CYP3A4 concentration in the membrane for the interactions of CYP3A4(C468)-ERIA (∆, broken line) and CYP3A4(C58,C64)-ERIA (○ and ●, continuous lines) with CYP3A4(C468)-DYM. The results of the experiment with CYP3A4(C58,C64)-ERIA carried out in the absence and the presence of 50 μM ANF are indicated by ○, dark continuous line and ●, light continuous line respectively. The lines show the results of the fitting of the datasets to eqn (1).

As shown in Figure 5(a), completion of the processes of protein incorporation and subunit exchange took approximately 12–16 h. The kinetic curves of decrease in donor fluorescence may be approximated with a bi-exponential equation (Figure 5a, inset). The offset value (intensity of fluorescence extrapolated to infinite time) determined from these approximations was used to estimate the amplitude of LRET at equilibrium. Similar results were also obtained with the use of CYP3A4(C468)-ERIA as the LRET donor.

The dependencies of LRET amplitude on the CYP3A4 concentration in the membrane obtained with CYP3A4(C58,C64)-ERIA/CYP3A4(C468)-DYM and CYP3A4(C468)-ERIA/CYP3A4(C468)-DYM donor/acceptor pairs are shown in Figure 5(b). The x-axis of this plot represents the surface density of P450 in the membrane calculated on the basis of RL/P, assuming the average area of the membrane per one phospholipid molecule is 0.72 nm2 [41] and the footprint of membrane-bound CYP3A4 is equal to 16 nm2.

Theoretically, these dependencies may be used to estimate apparent dissociation constants of CYP3A4 oligomers. However, the theory of diffusion-influenced association in 2D space is complex and poorly developed [42,43]. In the absence of an appropriate formalism, we approximated these dependencies with eqn (1) for the equilibrium of dimerization in solution: Embedded Image (1)

Where [E]0, [EE] and Kd are the total concentration of the associating compound (enzyme), the concentration of its dimers and the dissociation constant respectively. Despite the provisional nature of these approximations, they resulted in reasonable fits (ρ2≥0.9) that did reveal any systematic deviations from the experimental results. The parameters resulting from this fitting may be used for an approximate estimation of the oligomerization state of CYP3A4 in the membrane and LRET efficiency in the oligomers. According to our analysis (Table 2), the lipid-to-protein ratio at which the enzyme is 50% oligomerized (R50) is approximately 700 for the CYP3A4(C58,C64)-ERIA/CYP3A4(C468)-DYM pair and approaches 900 for the CYP3A4(C468)-ERIA/CYP3A4(C468)-DYM pair. Relocation of the donor fluorophore from Cys64 to Cys468 caused no considerable changes in either the LRET efficiency or apparent Kd of the enzyme oligomers. Both pairs of proteins exhibited a LRET efficiency of ~25% (Table 2), so that the interprobe distance in the mixed oligomers may be estimated to be in the range 40–42 Å.

View this table:
Table 2 Oligomerization of CYP3A4 in model membranes studied by LRET

The ‘±’ values show the confidence interval calculated for P=0.05. *Amplitude of LRET-dependent decrease in the intensity of donor fluorescence observed at RL/P of 1:150. †(RL/P)diss at which the amplitude of the titration curves reaches 50% of maximal.

We also probed the effect of ANF on P450–P450 interactions detected by LRET. The dependence of LRET amplitude on surface density of CYP3A4 obtained with the CYP3A4(C58,C64)-ERIA/CYP3A4(C468)-DYM pair in the presence of ANF is shown in Figure 5(b) (closed circles) and a light continuous line. As seen from this plot and Table 2, the addition of ANF had no effect on the parameters of P450–P450 interactions in this system.

Studies of P450–P450 interactions in model microsomes

The above method of detecting P450 oligomerization was then applied to study CYP3A4 in the absence and presence of ANF in membranes of Supersomes™ with incorporated CYP3A4(C58,C64)-ERIA and CYP3A4(C468)-DYM proteins. The setup of the experiments was similar to that used with liposomes. It should be noted, however, that in this case we were forced to change the incubation buffer used for CYP3A4 incorporation. Our initial attempts to use 0.1 M Hepes buffer (pH 7.4), as with liposomes, did not result in any detectable incorporation of the proteins into the membrane. However, replacement of the incubation buffer with 0.1 M potassium phosphate buffer (pH 7.4) allowed us to obtain microsomes containing both CYP3A4(C58,C64)-ERIA and CYP3A4(C468)-DYM.

As shown in Figure 6(a), incubation of CYP3A4(C58,C64)-ERIA containing Supersomes™ with CYP3A4(C468)-DYM resulted in appearance of LRET. Similarly to liposomes, an increase in the surface density of P450 augmented the LRET amplitude, which tended to approach a limit at high P450 concentrations (Figure 6b, open circles) [44]. Approximation of this dataset with eqn (1) suggests that, the microsomal membrane is characterized by a considerably higher Kd value of CYP3A4 oligomers than the proteoliposomal system (Table 2). However, unlike nearly hyperbolic dependencies observed with liposomes, the plot of LRET amplitude on P450 concentration in Supersomes™ was distinctly S-shaped (Figure 6b, open circles) (to calculate the surface density of P450 in microsomal membranes, we used the value 0.95 nm2 as an estimate of the surface of the microsomal membrane corresponding to one molecule of phospholipid in a monolayer [44]), so that the fitting of this dataset to eqn (1) was poor and revealed large systematic deviations (Figure 6b, black continuous line). Approximation of this dataset with the Hill equation (h=3.1±1.5) resulted in a considerably better fit (Figure 6b, black broken line). The LRET efficiency of ~10% estimated with this arbitrary approximation corresponds to an interprobe distance of ~48 Å.

Figure 6 Interactions of CYP3A4(C58,C64)-ERIA and CYP3A4(C468)-DYM mutants in CPR-containing Supersomes™ studied with LRET

(a) Series of spectra of delayed emission recorded during incubation of a 3 μM suspension of CYP3A4(C64,C468)-ERIA incorporated into Supersomes™ at RL/P=300 with 3 μM CYP3A4(C468)-DYM. (The RL/P after incorporation of CYP3A4(C468)-DYM is equal to 150). The inset shows the time dependencies of normalized intensity of donor fluorescence obtained in the experiments at the RL/P of 150 and 685. The lines represent the approximation of the kinetic curves with a bi-exponential equation. (b) Illustrates the dependence of the LRET amplitude on CYP3A4 concentration in the membrane. The continuous lines show the results of the fitting of the datasets to the eqn (1). The broken lines show the approximations of the data with the Hill equation. ○, dark lines show the results obtained in the absence of effector. Data obtained in the presence of 50 μM ANF are shown with ●, light lines.

In contrast with our observations with proteoliposomes, where the effect of ANF on P450–P450 interactions (Figure 5) was insignificant, the addition of ANF to Supersomes™ caused a prominent change in the shape of the dependence of the degree of oligomerization on P450 concentration in the membrane (Figure 6b). Interactions of CYP3A4 with ANF eliminated the S-shaped character of this dependence. Fitting of the dataset obtained in the presence of ANF with eqn (1) (Figure 6b, light continuous line) resulted in an adequate approximation and did not reveal the large systematic deviations seen in the absence of ANF. According to the results of the present study, addition of ANF to our model microsomal system promotes oligomerization of the enzyme, which is reflected in a decreased Kd and increased (RL/P)diss (Table 2). Strong parallels between the effect of ANF on P450–P450 interactions (Figure 6) and the dependence of the activating effect of ANF on surface density of CYP3A4 in Supersomes™ (Figure 3) suggests that the ANF-dependent modulation of P450–P450 interactions is directly related to the mechanism of enzyme allostery.


The central hypothesis of the present study was that heterotropic co-operativity of CYP3A4 that results in an increased kcat reflects effector-induced changes in enzyme oligomerization. The basic premise of the experimental design of the present study was that the involvement of protein–protein interactions in CYP3A4 allostery should reveal itself in a dependence of co-operativity on enzyme concentration (surface density) in the membrane. Rigorous testing of the hypothesis required two methodological advances: (i) a catalytically active membranous system in which the lipid/protein ratio could be varied systematically, and (ii) a reliable method for monitoring CYP3A4 oligomerization. Studies with a model system in which purified CYP3A4 was incorporated into commercially available ‘control’ Supersomes™ containing recombinant CPR, but no P450 provided decisive support for the effect of CYP3A4 concentration in the membrane on the allosteric properties of the enzyme. Specifically, model membranes with a high surface-density of CYP3A4 (RL/P<700) closely reproduce the behaviour of HLM, as evidenced by loss of homotropic co-operativity of 7-BFC debenzylation and a greater than 2-fold increase in turnover in the presence of ANF.

Furthermore, the LRET method that we developed to link the surface density of CYP3A4 in the membrane to the oligomerization state provided a powerful complement to the activity studies. Strong parallels between the degree of CYP3A4 oligomerization in the membrane and the extent of activation by ANF suggest that the effector-dependent modulation of P450–P450 interactions is directly related to the mechanism of allostery. A remarkable finding is the S-shaped character of the dependencies of the degree of CYP3A4 oligomerization (Figure 6) and the effect of ANF on enzyme turnover (Figure 3) on the surface density of CYP3A4 in Supersomes™. Addition of ANF eliminates the S-shaped character of dependence of LRET amplitude on CYP3A4 concentration and causes a multifold increase in the degree of CYP3A4 oligomerization at low enzyme concentrations. This observation reveals ANF as a potent modulator of P450–P450 interactions.

In the process of refining our model system, we discovered an intriguing difference between HLM and insect cell microsomes containing recombinant CYP3A4 and CPR. Specifically, the increase in the rate of 7-BFC metabolism caused by addition of ANF to HLM was not observed with either of the two brands of CYP3A4-containing insect cell microsomes, namely BD Supersomes™ or Baculosomes®. In Supersomes™, the levels of CYP3A4 are four to five times higher than that of CPR (similar to HLM), whereas in Baculosomes®, CPR is present in 2–3-fold molar excess over the haem protein. The findings of the present study suggest that the contrast between HLM and the recombinant systems is not related to the differences in CYP3A4/CPR ratio and ensuing degree of CYP3A4 saturation with the reductase. Rather, a likely explanation for the difference between HLM and the model microsomes is the high content of P450 apoprotein (haem-free enzyme) in the insect cell microsome [4547]. Interactions between holo- and apo-CYP3A4 may have an important effect, as noted previously in [48].

In addition, the S-shaped dependence of the oligomerization equilibrium on protein concentration was not observed with proteoliposomes (Figure 5). This contrast may indicate that the oligomerization of CYP3A4 in microsomal membranes is affected by its interactions with other proteins, in particular CPR. In support of this inference, a decrease in the degree of oligomerization of cytochrome P450 1A2 caused by incorporation of CPR into the membrane of proteoliposomes was demonstrated in studies of rotational diffusion of the haem protein [49]. It may be hypothesized that the oligomerization of CYP3A4 at low protein concentration and low P450/CPR ratios is hindered by the enzyme interactions with the reductase. This factor becomes less important at high concentration of the haem protein, when the concentration of CPR in the membrane becomes insufficient to prevent oligomerization.

A structural explanation for the relationship between CYP3A4 allostery and its oligomerization may be on the basis of the observations of peripheral ligand-binding site at the interface of the crystallographic dimer of some microsomal drug-metabolizing cytochrome P450 enzymes. Thus binding of two molecules of palmitic acid was observed in a hollow formed in the vicinity of the F′ and G′ helices of two interacting molecules of CYP2C8 (PDB code 1PQ2 [50]). Similar interactions with two molecules of peripherally bound progesterone were also detected in the crystallographic dimer of CYP3A4 (PDB code 1W0F [51]). Our recent studies of the interactions of Fluorol 7GA with CYP3A4 by FRET also suggested that the same region is the most probable site for high-affinity peripheral binding [29]. The role of peripheral substrate binding at this site in the mechanisms of CYP3A4 allostery was hypothesized by Williams et al. [51] and was advocated in our studies [28,29], as well as in studies of Roberts and Atkins [52] and Sligar and co-workers [3,53].

Detailed analyses of peripheral ligand binding in the crystallographic dimers of CYP3A4 and CYP2C8 [50,51] indicate that the high-affinity interactions at this site require oliogomeric protein. Many structures of CYP3A4 feature a dimer of subunits interacting through their F′ and G′ helices. The dimer is observed either as a part of an ASU in the crystal structure (e.g. PDB codes 2VOM and 2JOD [54]) or between the monomer in the ASU with a crystallographic symmetry neighbour (e.g. PDB code 1TQN [50]). Moreover, the dimer geometry is compatible with the proposed membrane orientation of the enzyme and is consistent across multiple structures of CYP3A4 and found in structures of other P450 enzymes: CYP1A2, CYP21A, CYP2A6 and CYP2C8. CYP3A4 and CYP2C8 dimer structures are unique in that the dimer interface forms a cavity that can accommodate small-molecule ligands. However, ligands are unlikely to bind with high affinity in the corresponding pockets in the monomeric subunits owing to their small size and low degree of ‘buriedness’. In other words, the peripheral site in the CYP3A4 structures appears to represent a ligand-binding cavity formed by the surfaces of two interacting molecules of the enzyme. Dissociation of the enzyme dimer will eliminate this cavity and thus compromise ligand binding.

Location of this peripheral site in the dimer of CYP3A4 (PDB code 1W0F) is shown in Figure 7. In this Figure, we also show one of the best poses of two ANF molecules docked into this site. It should be noted that this hypothetical CYP3A4 dimer structure is consistent with LRET efficiencies observed in experiments carried out in the present study. These efficiencies suggest that the inter-probe distances for Cys468-ERIA/Cys468-DYM and Cys468-ERIA/Cys64-DYM pairs are in the range 40–48 Å. Given the large sizes of ERIA and DYM probes (10–12 and 15–20 Å respectively), these estimates are in reasonable agreement with the distances between the sulfur atoms in Cys468/Cys468 and Cys468/Cys64 pairs in the CYP3A4 dimer (54.8 and 67.7 Å respectively).

Figure 7 CP3A4 dimer geometry observed in the structure of the CYP3A4 complex with progesterone (PDB code 1W0F) and several other X-ray structures of the enzyme

(a) A view from the side along the membrane plane; (b) a zoom-in on the binding site at the dimer interface; (c) shows a view perpendicular to the membrane plane; and (d) a zoom-in on the binding site at the dimer interface. The ligand-binding envelopes identified by ICM PocketFinder in the monomeric structures are shown as solid red and blue envelopes; sites identified in the dimeric complex are represented by magenta wire mesh. Cys468/Cys468 and Cys468/Cys64 distances are shown in (a) and (c). (e) Progesterone binding to the site at the CYP3A4 dimer interface as observed by X-ray crystallography (PDB code 1W0F). (f) Potential binding mode of ANF at the CYP3A4 dimer interface predicted by molecular docking to the site.

The interactions of ligands with a binding site of this kind are likely to affect the architecture of the oligomers and/or the equilibrium of oligomerization. We hypothesize therefore that the P450–P450 interface in CYP3A4 oligomers in the membrane is similar to that observed in the crystallographic dimers [51], and the peripheral progesterone binding in the vicinity of the F′ and G′ helices of two interacting CYP3A4 molecules observed in the X-ray structure is physiologically relevant. This site may have an important functional role in allosteric regulation of the enzyme by some yet unknown physiological ligand (e.g. a steroid or its derivative). The interactions of this site with some specific ligands, such as progesterone, testosterone, ANF or quinidine, may cause a change in the activity and substrate specificity of the enzyme via a ligand-induced re-organization of the enzyme oligomer.

According to this hypothesis, the divergence of the instances of heterotropic co-operativity in CYP3A4 into two categories (with and without effector-induced increases in kcat) discussed in the Introduction may be caused by differences in the affinity of different substrates for a peripheral binding site. Substrates such as testosterone and progesterone, for which there is no increase in kcat value in the presence of effectors, may be hypothesized to have high affinity for the peripheral site. Therefore these substrates may serve as allosteric activators by themselves and do not require other effectors for efficient metabolism. In this case the heterotropic activation that is observed at sub-saturating substrate concentrations may only be explained by the binding of multiple molecules of both substrate and the effector in the active site of the enzyme, as analysed by Denisov and Sligar [3]. In contrast, substrates exhibiting a considerable increase in kcat value, in the presence of activators such as phenanthrene, benzo[a]pyrene or 7-BFC, are presumably unable to interact with the peripheral site and therefore require an additional effector for enhanced metabolism. An important corollary to this hypothesis is that the increases in kcat value with the latter group of substrates will only be observed in the enzyme oligomers formed in the membranes or in micellar reconstituted systems, but will not be reproducible in monomeric systems such as Nanodisks.

A plausible scenario for allosteric interactions involving a conformational transition in the enzyme oligomer in response to a peripheral binding of an allosteric activator has been discussed in our previous paper (see Figure 2 and relevant discussion in [1]). The hypothetical mechanism of heterotropic activation of CYP3A4 introduced in that paper is in good agreement with the results of the present study. According to the hypothesis, the instances of heterotropic activation of CYP3A4 reveal a complex allosteric mechanism that involves effector-induced modulation of protein–protein interactions in the enzyme oligomers. An additional complexity to the allosteric properties of the ensemble of cytochrome P450 co-existing in the membrane of the endoplasmic reticulum is a possible formation of mixed oligomers between different P450 species (see [26,27] for a review), which may disrupt the hypothetical peripheral binding site or modify its functional properties. Therefore the allosteric properties of each individual cytochrome P450 may be modified considerably by changes in the composition of the P450 pool during cell differentiation, aging or disease.

The results of the present study emphasize the functional importance of P450–P450 interactions in the microsomal membrane. Detailed exploration of these interactions and elucidation of the mechanisms of their modulation by allosteric ligands represent an ultimate requirement for delineating the key factors that dictate the catalytic properties of the ensemble of drug-metabolizing cytochrome P450 in human liver. New experimental approaches for the studies of P450–P450 interactions and their functional consequences introduced in the present study, and the use of model microsomes with variable surface density of the haem protein in particular, provide a solid methodological foundation for further progress.


Dmitri Davydov designed the study. Nadezhda Davydova and Dmitri Davydov performed most of the experimental work with the help of Elena Sineva, who participated in LRET experiments and activity assays. Irina Kufareva analysed the structures of crystallographic dimers of cytochromes P450 and built a model of peripheral interactions of ANF with the CYP3A4 dimer. Dmitri Davydov analysed the data and wrote the paper. James Halpert contributed to the project development, data interpretation and writing of the paper.


This work was supported by the National Institutes of Health [grant numbers GM054995 and GM071872].

Abbreviations: ANF, α-naphthoflavone (7,8-benzoquinone); ASU, asymmetric unit; 7-BFC, 7-benzyloxy-4-(trifluoromethyl)coumarin; BODIPY-PC, 2-decanoyl-1-{O[11-(4,4-difluoro-5,7-dimethyl-4-bora-3a,4a-diaza-s-indacence-3-propionyl)amino]undecyl}-sn-glycero-3-phosphocholine; CPR, cytochrome P450 reductase; CYP, cytochrome P450 (also referred to as P450); DYM, DY-731 maleimide; ERIA, erythrosine 5′-iodoacetamide; FRET, fluorescence resonance energy transfer; HFC, 7-hydroxy-4-(trifluoromethyl)coumarin; HLM, human liver microsomes; RL/P, lipid/P450 ratio; LRET, luminescence resonance energy transfer; MWCO, molecular-mass cut-off; octyl glucoside, octyl-β-D-glucopyranoside; PA, phosphatidic acid; PC, phosphatidylcholine; PE, phosphatidylethanolamine; S-shaped, sigmoidal


View Abstract