Previous studies have shown that the presence of one P450 enzyme can affect the function of another. The goal of the present study was to determine if P450 enzymes are capable of forming homomeric complexes that affect P450 function. To address this problem, the catalytic activities of several P450s were examined in reconstituted systems containing NADPH–POR (cytochrome P450 reductase) and a single P450. CYP2B4 (cytochrome P450 2B4)-, CYP2E1 (cytochrome P450 2E1)- and CYP1A2 (cytochrome P450 1A2)-mediated activities were measured as a function of POR concentration using reconstituted systems containing different concentrations of P450. Although CYP2B4-dependent activities could be explained by a simple Michaelis–Menten interaction between POR and CYP2B4, both CYP2E1 and CYP1A2 activities generally produced a sigmoidal response as a function of [POR]. Interestingly, the non-Michaelis behaviour of CYP1A2 could be converted into a simple mass-action response by increasing the ionic strength of the buffer. Next, physical interactions between CYP1A2 enzymes were demonstrated in reconstituted systems by chemical cross-linking and in cellular systems by BRET (bioluminescence resonance energy transfer). Cross-linking data were consistent with the kinetic responses in that both were similarly modulated by increasing the ionic strength of the surrounding solution. Taken together, these results show that CYP1A2 forms CYP1A2–CYP1A2 complexes that exhibit altered catalytic activity.
Cytochrome P450 plays a major role in the metabolism of a wide variety of drugs and other xenobiotics as well as numerous endogenous compounds. Although the interactions between NADPH–POR (cytochrome P450 reductase) and P450 necessary for mono-oxygenase function are well characterized [1–6], the potential for P450 enzymes to interact with each other has only been examined more recently [7–10]. Many of these P450–P450 interactions have been shown to significantly alter mono-oxygenase function.
Several laboratories have now reported that one P450 enzyme can influence the behaviour of another [7–15]. Our laboratory demonstrated that the catalytic activity of CYP2B4 (cytochrome P450 2B4)-dependent 7-PR (7-pentoxyresorufin) metabolism was inhibited by the addition of CYP1A2 (cytochrome P450 1A2) to the POR/CYP2B4 reconstituted system . Inhibition of this CYP2B4 activity by CYP1A2 (cytochrome P450 1A2) was dependent on the substrate used [8,16], was more pronounced when POR was subsaturating and appeared to be related to a high-affinity association between CYP1A2 and POR. When CYP1A2 function was examined in the same reconstituted systems, the presence of CYP2B4 synergistically stimulated CYP1A2 activity. Again, the response was more pronounced at subsaturating POR concentrations.
Subsequent reports showed that this interaction was not simply competition between CYP1A2 and CYP2B4 for POR, but actually involved formation of a CYP1A2–CYP2B4 complex [7,11,13]. In mixed reconstituted systems (containing CYP1A2, CYP2B4 and POR), POR appears to bind with high affinity to the CYP1A2 moiety of the CYP1A2–CYP2B4 complex, whereas the POR–CYP1A2 complex has a much lower binding affinity in simple reconstituted systems containing a single P450 [6,7,11]. These results support the concept that the catalytic behaviour of a P450 enzyme is significantly influenced by the presence of another P450.
P450–P450 complexes were also reported for CYP1A2–CYP2E1 (cytochrome P450 2E1). Again the CYP1A2–CYP2E1 complex bound POR with high affinity . Both the CYP1A2–CYP2E1 complex  and the CYP1A2–CYP2B4 complex mentioned above  were sensitive to alterations in ionic strength, as the inhibitory/synergistic effects on enzyme activities were relieved using higher concentrations of Hepes buffer. These data are consistent with the heteromeric P450–P450 complexes being stabilized by electrostatic interactions, where an increase in the ionic strength of the surrounding solvent is capable of disrupting these P450–P450 complexes. Interestingly, evidence for significant interactions that affected mono-oxygenase function was not found with mixed reconstituted systems containing CYP2B4 and CYP2E1 . Taken together from this limited dataset, P450–P450 interactions capable of influencing mono-oxygenase function were observed when CYP1A2 was one of the protein components, but were less likely when this P450 was not involved.
The above results are consistent with CYP1A2 facilitating the formation of P450–P450 complexes. This raises the question, if CYP1A2 supports the formation of heteromeric P450 complexes, then is CYP1A2 capable of forming complexes with other CYP1A2 enzymes? Despite its simplicity, little is known about the potential for a particular P450 enzyme to interact with itself within a simple reconstituted system containing POR and a single P450. The goal of the present study was to determine if CYP1A2 is capable of forming homomeric complexes that influence P450 function. The results support the idea that homomeric CYP1A2 complexes do exist, and are consistent with the concept that at least some P450 enzymes exist in the endoplasmic reticulum as dimers or higher-order complexes.
DLPC (dilauroylphosphatidylcholine), 7-ER (7-ethoxyresorufin), resorufin, NADPH, glucose 6-phosphate, glucose-6-phosphate dehydrogenase and formaldehyde were purchased from Sigma–Aldrich. BZP (benzphetamine) was a gift from Upjohn. 7-EFC (7-ethoxy-4-trifluoromethylcoumarin) and 7-HFC (7-hydroxy-4-trifluoromethylcoumarin) were obtained from Molecular Probes. Escherichia coli C41 cells were purchased through Avidis SA.
HEK (human embryonic kidney)-293T/17 cells (CRL-11268) were purchased from the A.T.C.C. (Manassas, VA, U.S.A.). Lipofectamine™ 2000, DMEM (Dulbecco's modified Eagle's medium) and PBS were purchased from Invitrogen. Coelenterazine 400a was purchased from Gold Biotechnology, and coelenterazine h was purchased from Promega. The BRET vectors were obtained from PerkinElmer (Waltham, MA). BRET (bioluminescence resonance energy transfer) measurements were performed on a TriStar LB 941 plate reader (Berthold Technologies).
CYP2B4 was expressed in E. coli C41 cells and purified according to standard procedures with minor modifications . Recombinant rabbit NADPH cytochrome P450 reductase [plasmid pSC-POR, provided by Professor Lucy Waskell (University of Michigan, Ann Arbor, MI, U.S.A.) and constructed from the plasmid pCWori-rabbit POR, utilizing a T7 promoter] was expressed in E. coli C41 cells, solubilized and purified according to a modification of previously described methods [18,19]. CYP1A2 was isolated from βNF (betanaphthoflavone)-treated rabbit liver microsomes as described previously . CYP2E1 was expressed and purified according to .
Measurement of mono-oxygenase activities
The catalytic activities of CYP1A2, CYP2E1 and CYP2B4 were determined using reconstituted systems where POR and CYP1A2, CYP2E1 or CYP2B4 were combined in DLPC. The reconstituted systems were pre-incubated at room temperature (22°C) for 2 h before the addition of the other assay components . After pre-incubation, the enzyme activities were measured at 37°C. All of the fluorimetric products were monitored on a SpectraMax M5 plate reader (Molecular Devices).
For EROD (7-ethoxyresorufin O-dealkylation) activity, the final assay conditions were: the reconstituted system [containing various concentrations of P450 (ranging from 0.025 μM to 0.2 μM), varying concentrations of POR (ranging from 0.025 μM to 0.4 μM) and 8 μM DLPC], and 4.5 μM substrate, in 50 mM Hepes buffer (pH 7.5). The reactions were initiated by the addition of NADPH to a final concentration of 0.5 mM. The accumulation of resorufin was monitored fluorometrically at an excitation wavelength of 530 nm and an emission wavelength of 585 nm. A hydroxyresorufin standard curve was calculated each time to determine the level of product formation [23,24].
For 7-EFC O-deethylation, the final assay conditions were: the reconstituted system (as described above), and substrate (0.1 mM 7-EFC), in 100 mM potassium phosphate buffer (pH 7.25). Reactions were initiated by the addition of NADPH to a final concentration of 0.5 mM. The fluorescent product, 7-HFC, was monitored over a 2-min time course using excitation/emission wavelengths of 410 nm/510 nm [25,26].
Demethylation of pNA (p-nitroanisole) was examined by directly measuring the formation of the product p-nitrophenol at 405 nm. The final assay conditions were: the reconstituted system (as described above), and pNA (0.4 mM), in 100 mM potassium phosphate buffer (pH 7.25). Reactions were initiated by the addition of NADPH to a final concentration of 0.5 mM. Formation of p-nitrophenol was determined using a molar extinction coefficient of 12.9 mM−1·cm−1 as described previously [27,28].
A fluorometric method was used to determine BZP and NDMA (N-nitrosodimethylamine) demethylation activity [29,30]. The final assay conditions were: the reconstituted system [containing P450 (ranging from 0.05 μM to 0.4 μM), POR (ranging from 0.05 μM to 0.8 μM) and 16 μM DLPC, as described above] and substrate (either 1 mM benzphetamine or 2 mM NDMA) in 100 mM potassium phosphate buffer (pH 7.25). The reactions were initiated by the addition of NADPH to a final concentration of 0.5 mM. After incubation at 37°C for times ranging from 0 to 8 min, the reaction was terminated with 0.11 vol. of a mixture of 25% (w/v) ZnSO4 and 0.5 M semicarbazide (10:1). The samples were combined with 0.1 vol. of saturated Ba(OH)2, placed on ice for 10 min and then centrifuged at 14000 g for 10 min. An aliquot (0.5 ml) of supernatant was mixed with 0.5 ml of Nash reagent [0.3% acetic acid, 0.2% acetylacetone and 15.4% (w/v) ammonium acetate]. The formaldehyde product was measured fluorometrically (excitation wavelength 410 nm, emission wavelength 505 nm) and was quantified using a standard curve. The initial rates were determined from the time course.
Modelling of the data collected from enzymatic assays
The kinetic data for the hyperbolic curves were fit using a global analysis of a mass action binding of POR and P450 using the Dynafit program (Biokin), which applies the Levenberg–Marquardt algorithm to generate a least squares non-linear regression fit of the data. The data were fit using a single KD value for the POR–P450 complex and a single kcat value for each activity, according to:where A represents the P450 enzyme, R represents POR, KAR is the dissociation constant for formation of the POR–P450 complex, kj is the kcat for the reaction and P represents the product formed. Each of the assays described in the present paper was performed under conditions of saturating substrate concentration; therefore, A and AR actually represent the substrate-bound forms of these enzymes, with the assumption being that once product is released from the enzyme, it is rapidly replaced by another substrate molecule. The kcat value is a characteristic of the ability of the substrate-bound AR complex to turn over a substrate and generate product. This value will vary for different substrates. The Dynafit program used is included in Supplementary Online data (Program A at http://www.BiochemJ.org/bj/446/bj4460489add.htm).
The Dynafit program determines the KAR and kcat values that provide the best fit of the data at all of the P450 concentrations (i.e. all of the curves within a Figure), assuming a Michaelis–Menten type model for substrate binding and catalysis. Consequently, the data analysis generates a single KAR and kcat value, where the only difference among the curves is the P450 concentration.
Demonstration of CYP1A2–CYP1A2 complexes by cross-linking with BS3 [bis(sulfosuccinimidyl) suberate]
Reconstituted systems of DLPC, CYP1A2 and/or POR were prepared as described above. The conditions used for these reconstituted systems were selected to assess the aggregation of enzymes (by chemical cross-linking) at different concentrations of POR, salt and lipid. Before incubating at room temperature for 2 h, all of the reconstituted systems, containing variable amounts of enzymes and lipid as described in the Results section, were diluted to a final P450 (or POR if P450 was not present in the system) concentration of 11.5 μM with buffer containing 50 mM Hepes (pH 7.4), 100 mM NaCl, 5 mM EDTA and 20% (v/v) glycerol. The samples were stirred briefly with a 10 μl pipette tip before incubating for 2 h at room temperature. After the 2 h incubation, 10 pmol of P450 from each reconstituted system was diluted to a volume of 48 μl with 50 mM Hepes buffer (pH 7.5). The incubations without CYP1A2 contained 0.2 μM POR and 32 μM DLPC. The samples were heated to 37°C for 5 min before cross-linking was initiated with 2 μl of 6.25 mM aqueous BS3 (Pierce-Thermo). After adding BS3, the samples were stirred briefly with a pipette tip and incubated at 37°C for 20 s. The reactions were terminated with 25 μl of 3× loading buffer containing 75 mM Tris (pH 6.8), 9% (w/v) SDS, 0.015% Bromophenol Blue, 36% sucrose, 750 mM glycine and 1 M NaCl. The loading buffer was heated at 37°C for 30 min before being used to dissolve the buffer components. Immediately after adding the 3× loading buffer, the samples were vortex-mixed and immersed in ice for at least 15 min and stored at 4°C. The protein samples were denatured by the addition of 2-mercaptoethanol with boiling for 30 min. From each sample 15 μl was loaded into individual lanes of an SDS/PAGE (4–12% gel) which was performed at 4°C by initially running the gel at 50 V for 30 min. After the 50 V run, the samples were run for an additional 1.5 h at 200 V. Gels were then transferred on to nitrocellulose membranes and immunoblotted using anti-CYP1A2 and anti-POR antibodies.
Detection of CYP1A2–CYP1A2 complexes using BRET
Vectors for BRET assays were generated by ligating rabbit CYP1A2 cDNA into the pRluc-N2 and pGFP2-N1 vectors. A vector for the expression of unlabelled CYP1A2 was generated by inserting a stop codon into the pGFP2-N1/CYP1A2 vector using site-directed mutagenesis. Both the Rluc (Renilla luciferase) and GFP (green fluorescent protein) constructs were fused to the C-terminal region of the CYP1A2 gene in order to minimize interference with membrane binding of CYP1A2.
HEK-293T/17 cells were maintained in DMEM supplemented with 10% fetal bovine serum under a humidified 37°C atmosphere of air and 5% CO2. At 1 day prior to transfection, the cells were seeded on to 6-well plates at 7.5×105 cells/well. Cells were transiently transfected with Lipofectamine™ 2000 and medium was replaced with fresh DMEM 4 h after transfection.
At 24 h after transfection, the cells were washed and resuspended in 750 μl of PBS per well. Aliquots (100 μl/well) of the suspension were pipetted into both a black clear-bottom 96-well plate for GFP quantification and a white opaque 96-well plate for Rluc quantification and BRET measurement. The BRET ratio was determined by measuring the ratio of GFP fluorescence (at 510 nm) to Rluc luminescence (at 410 nm) immediately following the addition of coelenterazine 400a to a final concentration of 5 μM. BRET measurement was performed on three wells for each sample. Untransfected cells were used to normalize the GFP and Rluc signals, and the resulting BRET ratio for each sample was normalized to that of cells expressing the Rluc construct alone.
The ratio of GFP- to Rluc-tagged protein was determined by comparing the GFP fluorescence and coelenterazine h-induced Rluc luminescence to that of a GFP–Rluc fusion protein, which was assumed to express GFP and Rluc at a 1:1 ratio. GFP fluorescence was measured at 510 nm with a SpectraMax M5 plate reader. Rluc expression levels were determined by measuring coelenterazine h-induced luminescence in a TriStar LB 941 plate reader at 480 nm.
Formation of homomeric P450 complexes would be a function not only of the affinity of the proteins for each other, but also of the concentration of the P450s (i.e. more complexation would be expected at higher P450 concentrations due to more protein crowding in the membrane). Therefore the behaviour of a P450 that does not form complexes can be readily predicted theoretically. If a P450 enzyme-mediated catalytic activity is measured as a function of POR, classical hyperbolic curves are expected at each P450 concentration if the activity follows Michaelis–Menten kinetics (Figure 1A). As the P450 concentration is increased, there is a shift to the right in the POR concentration necessary to achieve saturation and an increase in the maximal activity that is proportional to the P450 concentration. However, if P450s exist as a complex, and either the activity of the P450–P450 complex or its affinity for POR differs from that of the P450 monomer, then deviations from this predicted behaviour would be expected.
Effect of CYP2B4 concentration on the formation of homomeric P450–P450 complexes
The ability of three different rabbit P450 enzymes CYP2B4, CYP2E1 and CYP1A2 were examined to determine if these proteins formed kinetically detectable complexes. When examined as a function of POR concentration, CYP2B4-dependent benzphetamine N-demethylation produced a hyperbolic response as expected for a simple interaction between POR and P450 monomers (Figure 1B). All three of the curves can be fit to a typical Michaelis function having a KD value of 0.013 μM for the POR–CYP2B4 complex and a kcat value of 51 nmol/min/nmol CYP2B4. Qualitatively similar results were obtained when the substrate pNA was used (Figure 1C). CYP2B4-dependent pNA demethylation also showed results consistent with a typical Michaelis–Menten reaction between monomeric POR and CYP2B4; however, the KD value for the POR–CYP2B4 complex was 20-fold larger than that found for benzphetamine. The data do not provide any evidence for the formation of CYP2B4–CYP2B4 complexes that affect the function of this enzyme. However, these results demonstrate that substrate can have a profound effect on the ability of a P450 to interact with POR [31,32].
CYP2E1-dependent 7-EFC O-deethylation activity measured as a function of POR concentration also yielded typical hyperbolic curves with a KD value equal to 0.011 μM (Figure 2A). It was consistent with simple interaction between monomeric CYP2E1 and POR expected for Michaelis–Menten enzymes, suggesting that any CYP2E1–CYP2E1 complexes that may be formed do not appear to alter the binding of POR to P450. In contrast, a different response was observed when CYP2E1-mediated NDMA demethylation was examined. Rather than the hyperbolic response observed with 7-EFC demethylation with CYP2E1 and with the CYP2B4-mediated reactions, NDMA demethylation was shown to generate a sigmoidal response as a function of POR concentration (Figure 2B). Interestingly, at low POR concentrations, the overall reaction rate was independent of CYP2E1 concentration. This lack of a hyperbolic response suggests a more complex interaction between POR and CYP2E1, and the type of response is sensitive to the substrate used.
Measurement of CYP2E1 activities as a function of [POR]: effect of CYP2E1 concentration
The responsiveness of CYP1A2 as a function of POR concentration was similar to that observed with CYP2E1. 7-EFC metabolism produced a hyperbolic response that was proportional to the P450 concentration and was consistent with a simple mass-action binding of POR to CYP1A2 (Figure 3A). The KD value for this process was found to be 0.012 μM. In contrast, CYP1A2-dependent 7-EC (7-ethoxycoumarin) O-dealkylation activity exhibited a sigmoidal response, suggesting that a more complex interaction than can be explained by a simple mass action interaction of POR with monomeric CYP1A2 must be invoked to explain the data (Figure 3B). A similar sigmoidal response was observed when 7-ER demethylation was examined (Figure 4A). Interestingly, at lower concentrations of CYP1A2 (0.025 and 0.05 μM) the catalytic activities as a function of POR concentration yielded hyperbolic curves, consistent with simple interaction between monomeric CYP1A2 and POR expected for Michaelis–Menten enzymes. However, at higher concentrations of CYP1A2 (0.1 and 0.2 μM) the sigmoidal response was observed, leading to activities at subsaturating POR (at 0.025 and 0.05 μM) that were independent of the CYP1A2 concentration.
Measurement of CYP1A2 activities as a function of [POR]: effect of CYP1A2 concentration
Effect of ionic strength on CYP1A2-mediated EROD as a function of [POR]
These deviations from Michaelis–Menten behaviour were increasingly observed at elevated P450 concentrations when the P450 proteins were more crowded into the membrane. This raised the potential for the sigmoidal behaviour to be the result of homomeric P450–P450 interactions. Our laboratory previously reported that heteromeric CYP1A2–CYP2B4 and CYP1A2–CYP2E1 [11,12] complexes could be disrupted by increasing the ionic strength of the buffer. Therefore if the sigmoidal behaviour observed at elevated CYP1A2 concentrations was also due to electrostatic interactions, then an increase in ionic strength should lead to disruption of these homomeric complexes and allow the kinetics to revert to a hyperbolic response that can be described by a simple Michaelis–Menten type interaction between POR and CYP1A2. As shown in Figure 4(B), the inclusion of 300 mM KCl in the buffer caused a change in the kinetic response to a hyperbolic response that could be explained by a single KD value for POR binding, results consistent with the disruption of CYP1A2–CYP1A2 complexes. A similar response was observed when 7-PR dealkylation was examined (Supplementary Figure S1 at http://www.BiochemJ.org/bj/446/bj4460489add.htm).
Detection of CYP1A2–CYP1A2 complexes by chemical cross-linking
The kinetic data show that CYP1A2 and CYP2E1 exhibit non-Michaelis behaviour, particularly at high P450:lipid ratios. If these kinetic effects are indeed the result of homomeric P450 complexes that are stabilized by electrostatic interactions, we would expect to be able to cross-link these P450s and potentially alter the degree of complexation by changing the ionic strength. For these experiments we focused on CYP1A2, which has been shown to readily form complexes with CYP2B4 [7,11,13] and CYP2E1 , in addition to the non-Michaelis Menten behaviour shown in Figures 3 and 4. In these experiments we prepared several reconstituted systems in DLPC, cross-linked the proteins with BS3, separated the proteins using polyacrylamide gels and immunoblotted with anti-CYP1A2 and anti-POR antibodies. Cross-linking with BS3 (Figure 5) shows that CYP1A2 forms several complexes in simple reconstituted systems ranging in size from approximately 100 kDa to greater than 250 kDa, both in the absence or presence of POR (Figure 5, lanes 3 and 5). Overall, cross-linking produced complexes in the 100 kDa, 200 kDa, 250 kDa and 300 kDa ranges, representing dimers, tetramers, pentamers and hexamers. Interestingly we did not observe a single band in the 100 kDa region, representative of a CYP1A2–CYP1A2 dimer, but several bands in the 100–125 kDa region. This is possibly due to the formation of dimeric complexes that were cross-linked in different orientations. In separate experiments the addition of substrate alone did not appear to significantly affect complex formation (results not shown).
Detection of CYP1A2–CYP1A2 complexes by chemical cross-linking with BS3
When the reconstituted systems were placed in a high ionic strength buffer (containing 300 mM KCl), changes in the sizes of the complexes were observed (Figure 5, lanes 4 and 6). Generally, there were decreases in the intensity of the complexes, which is consistent with the disruption of CYP1A2–CYP1A2 complexes with increasing ionic strength. Ionic strength also caused a decrease in the formation of POR-containing complexes in a manner similar to that shown for CYP1A2 (Supplementary Figure S2 at http://www.BiochemJ.org/bj/446/bj4460489add.htm). These results are consistent with the kinetic data, where an increase in the ionic strength shifted the sigmoidal activity compared with [POR] curves to hyperbolas that follow Michaelis–Menten behaviour.
Detection of CYP1A2–CYP1A2 complexes using BRET
Resonance energy transfer methods have been used recently to measure the formation of complexes among membrane-bound proteins. These methods have been used to detect the formation of both homomeric and heteromeric P450 complexes [9,33–35] as well as other proteins such as haem oxygenase [36,37]. The potential for CYP1A2 to form homomeric complexes was examined by BRET where the C-terminal region of the CYP1A2 gene was fused with Rluc and another CYP1A2 gene was fused with GFP. These two fusion proteins were then transfected into HEK-293T cells at various CYP1A2–GFP:CYP1A2–Rluc ratios, and the total amount of transfected DNA remained constant.
As a first step, it was important to demonstrate that the transfected fusion proteins localized in the endoplasmic reticulum. Evidence for the co-localization of the CYP1A2–GFP fusion protein with the endoplasmic reticulum marker DsRed-ER is shown in Supplementary Figure S3 (at http://www.BiochemJ.org/bj/446/bj4460489add.htm).
Next, the formation of CYP1A2–CYP1A2 complexes was established using BRET. Both the CYP1A2–Rluc and CYP1A2–GFP genes were transfected into HEK-293T cells at a series of CYP1A2–GFP:CYP1A2–Rluc ratios. At 24 h after transfection of the fusion proteins, the luciferase substrate coelenterazine 400a was added to the HEK-293T cells, generating a luminescent signal at 410 nm. In the event that a complex between CYP1A2 molecules occurs, there is energy transfer from the luminescent signal to the CYP1A2–GFP fusion protein, leading to a fluorescent signal at 510 nm. The BRET signal represents a ratio of the 510 nm fluorescence/410 nm luminescence. At low CYP1A2–GFP:CYP1A2–Rluc ratios, the BRET signal is low, due to the low levels of GFP construct limiting the potential for energy transfer; however, as the relative amounts of the GFP-containing construct were increased there was a substantial increase in the total BRET (Figure 6, upper curve). The total BRET appeared to saturate at high levels of GFP, which is consistent with the formation of a specific CYP1A2–CYP1A2 complex. Transfection of CYP1A2–Rluc with a naked GFP vector showed only a minimal increase in the total BRET (Figure 6, lower curve).
Demonstration of CYP1A2–CYP1A2 complex formation using BRET: disruption of the BRET signal by co-transfection with native CYP1A2
As an additional demonstration of the specificity of this response, transfection of unlabelled CYP1A2 (i.e. CYP1A2 DNA containing neither Rluc nor GFP) was able to compete with the CYP1A2 fusion proteins and diminish the total BRET response. This response, as measured by the maximal BRET at high GFP:Rluc ratios, was dependent on the amount of unlabelled CYP1A2 transfected into the cells, where increasing unlabelled CYP1A2 caused a greater inhibition of the total response (Figure 6, middle curves).
There is a potential concern that the transfection of unlabelled CYP1A2 would affect the BRET response, by decreasing the total amount of expression of the GFP- and Rluc-fusion proteins. However, such a decrease in protein expression will not affect the maximal BRET response for a high-affinity complex. As shown in Figure 7, the total amount of transfected DNA was varied by a factor of more than 35-fold. Regardless of the substantial decrease in the expressed protein, the BRET response was not significantly reduced. These results demonstrate that the decrease in BRET response (shown in Figure 6) was not an artefact of a decrease in expression of the CYP1A2–GFP and CYP1A2–Rluc fusion proteins. In addition, the results illustrate that CYP1A2 forms a high-affinity complex and the observed results are not simply the consequence of protein overcrowding in the membrane. If the complexes were due to protein overcrowding, a significant decrease in the BRET response would be expected with a decrease in protein expression, due to dissociation of the BRET-sensitive complexes . Taken together, these results demonstrate that P450 enzymes are capable of forming homomeric complexes that can modulate mono-oxygenase function, that these are highaffinity complexes and that they occur not only in reconstituted systems, but also in living cells.
Effect of quantity of DNA transfected on the BRET response: interaction of CYP1A2–GFP with CYP1A2–Rluc
It is now generally accepted that P450 cytochromes form complexes in membranes and proteoliposomes [7,13–16,39–41]. Furthermore, it has been shown that in many cases, the physical interaction of different types of P450s can alter the catalytic activities of the associating enzymes [7,10,12–14,39]. Both inhibition and activation of metabolism by P450s have been reported to result from P450–P450 interactions.
The original hypothesis that P450–P450 interactions could influence enzyme function was formulated by comparing substrate metabolism by a single P450 with mixed reconstituted systems containing mixtures of P450s. Thus the early evidence for these effects was derived from kinetic studies pointing to heteromeric interactions between two different P450 enzymes. However, most of the physical evidence for complex formation (aggregation) of P450 (derived from techniques such as cross-linking , rotational diffusion studies using optical [43,44], fluorescent  or electroparamagnetic resonance  analysis, analytical ultracentrifugation [46–48], gel-filtration chromatography [42,49,50] and electron microscopy [47,49]) was demonstrated using a single P450. This raises the question of whether homomeric complexation also influences enzyme function.
Only a single previous study has implicated a homomeric interaction that influences catalytic activity . This study used a global mathematical analysis to fit kinetic data on the metabolism of p-nitrophenol by CYP2E1 at various concentrations of both the P450 and POR. Their data suggested the existence of the following types of P450 aggregates: (i) the P450–P450 homodimer; (ii) the POR–P450 binary complex; (iii) a quaternary complex of two PORs and two P450s; and (iv) a putative ternary complex of two P450s and one POR. The best fit of the kinetic data was consistent with only the binary POR–CYP2E1 complex being catalytically active; higher-order complexes were inactive.
In the present study, the potential for homomeric complex formation to affect P450 function was examined using three P450 enzymes, CYP2B4, CYP2E1 and CYP1A2. The kinetic data obtained using CYP2B4 could be explained by a simple mass-action complex being formed between POR and CYP2B4. Although these results do not preclude homomeric complex formation, the results demonstrate that the CYP2B4–CYP2B4 complexes formed do not appear to affect function. The ability of CYP2B4 to form complexes has been shown by cross-linking of reconstituted systems containing only lipid and CYP2B4 (Supplementary Figure S4 at http://www.BiochemJ.org/bj/446/bj4460489add.htm). Despite the tendency for these proteins to form complexes, mono-oxygenase activities were not affected, at least for the substrates tested.
Both CYP1A2 and CYP2E1 exhibited an apparent co-operative response when several, but not all, of the activities were examined as a function of POR concentration. Interestingly, an increase in the ionic strength of the surrounding buffer was capable of disrupting these CYP1A2–CYP1A2 complexes and caused the conversion of the sigmoidal kinetics to a hyperbolic response.
It is important to emphasize that our data suggest that the functional effects resulting from homomeric P450 interactions are largely substrate-dependent. CYP1A2-mediated metabolism of 7-EC, 7-PR and 7-E were clearly influenced by its self-aggregation, as shown by the significant sigmoidal curves when activity compared with [POR] was plotted. However, metabolism of 7-EFC by CYP1A2 fit reasonably well to simple Michaelis–Menten kinetics (Figure 3C). Similar results were obtained with CYP2E1 activities (Figure 2).
The kinetic evidence for homomeric complex formation was corroborated by experiments demonstrating that CYP1A2 forms physical complexes. There are two lines of evidence to support this contention. First, cross-linking data showed that dimers and higher-order complexes were readily detected when the proteins were incorporated into membranes prepared from dilauroylphosphatidylcholine (Figure 5). These experiments also showed that these CYP1A2–CYP1A2 complexes were disrupted by increasing the ionic strength of the assay system. These results are consonant with the kinetic data where the ‘sigmoidal’ kinetics in 7-ER, 7-PR and 7-EC metabolism at low ionic strength were converted into hyperbolic responses that could be explained by Michaelis–Menten behaviour at high ionic strength. The results support the formation of kinetically unique homomeric complexes that are disrupted by elevated ionic strength.
A second line of support for the formation of physical CYP1A2–CYP1A2 complexes comes from analysis of the BRET data. These experiments provide evidence for the formation of a high-affinity complex between CYP1A2 molecules. The specificity of these complexes was reinforced by the ability of unlabelled CYP1A2 to disrupt the BRET signal.
Given the tendencies of these enzymes to associate in proteoliposomes, it seems likely that homomeric and heteromeric P450 interactions also occur when there are no observable functional effects. However, for reasons that have not been elucidated, it seems that some substrates influence the system in a way that limits the ability of POR to interact with the P450, whereas other substrates do not impede the ability of POR to deliver electrons. Another possibility is that substrates change the affinities by which P450s bind to themselves or to other P450s. In cases where substrates weaken P450–P450 binding affinity, POR may be able to displace P450 enzymes from the aggregates and readily deliver electrons. In cases where the substrates strengthen the P450–P450 binding affinity, the access of POR to the P450s within the aggregate may be limited.
Previous studies also have revealed that the functional effects attributable to heteromeric P450 interactions are specific to the types of P450 being studied. For instance, testosterone-6β hydroxylation by CYP3A4 was activated by the presence of CYP1A enzymes but was not affected by the presence of CYP2C9 (cytochrome P450 2C9), CYP2D6 (cytochrome P450 2D6) or CYP2E1 . In the present study, we have presented data supporting the possibility that rabbit CYP2B4 does not form homomeric complexes that influence its ability to carry out substrate metabolism. Although this assumption is clearly tentative because of the limited number of compounds tested, CYP2B4-mediated metabolism could be modelled by simple Michaelis–Menten kinetics with both substrates, unlike the results obtained with CYP1A2 where sigmoidal curves were observed. This sigmoidicity was much more apparent when CYP1A2 concentrations in the membrane were elevated, leading to protein crowding and greater opportunities for P450–P450 interactions.
Interestingly, two substrates did have dramatically different effects on the apparent binding affinity of POR and CYP2B4. The apparent Km value for POR and CYP2B4 was 13 nM in the presence of benzphetamine and 290 nM in the presence of p-nitrophenol. The putative effects of substrates on the binding of P450s and POR are variable. Using systems in which CYP2B4 and POR were reconstituted in separate lipid vesicles, Backes and Eyer  clearly showed that the rate of association of the enzymes was strongly influenced by substrate. Consistent with our current data, the earlier study showed that the rates of association were slightly stimulated by p-nitroanisole, but greatly stimulated by benzphetamine relative to the rate in the absence of substrate . Later studies presented findings that showed the effects of substrate on the binding affinity of POR varied widely among different types of P450 [32,50]. In other words, substrates affected the POR-binding affinities of some types of P450, but not others . Similarly, substrates did not affect the binding of POR to a variety of P450s that were affixed to the wells of a polystyrene plate . However, from seeing the various effects of substrates for P450 enzymes in the present study, it could be questioned whether more comprehensive sampling would have revealed that the POR binding to all of the P450s could be influenced by some compounds, but not others. Furthermore, it can be questioned whether the nature of binding between the proteins is profoundly altered after affixing one of the enzymes to the wells of a polystyrene plate. Conceivably, the influence of substrate on the binding of POR and P450 may be manifest when the proteins are reconstituted in a lipid milieu, but not when the enzymes are affixed to the plate.
The present study provides compelling evidence that the homomeric aggregation of P450s can influence the rate of substrate metabolism by a P450 enzyme, and that the functional effects of homomeric P450 complex formation are both substrate- and P450 type-specific. These homomeric complexes appear to be sensitive to alterations in the ionic strength where the non-Michaelis–Menten kinetics observed at low ionic strength can be converted into typical hyperbolic kinetics at elevated salt concentrations. The kinetic data correlated well with studies showing that physical CYP1A2–CYP1A2 complexes exist both in reconstituted systems (chemical cross-linking) and in cellular systems (BRET). Finally, the sensitivity of the kinetic data to ionic strength effects were consistent with those obtained by chemical cross-linking showing a decrease in the formation of CYP1A2–CYP1A2 complexes at elevated ionic strength.
The data demonstrating the existence of homomeric P450 complexes are supportive of previous reports of heteromeric P450 complexes that affect mono-oxygenase function. These heteromeric interactions have been reported not only in reconstituted systems [7,11–13], but also in cellular systems where P450s are expressed [34,35,53]. Additionally, heteromeric interactions that affect P450 function were shown to occur in rabbit liver microsomes . Although we have been able to show that these complexes occur, we currently do not know the actual size of the complexes in vivo. Cross-linking data show that P450s can form not only dimers, but also larger complexes (Figure 5). Consequently, we currently do not know the size of the average complex, or even whether they contain other non-P450 proteins as constituents. Additionally, we have several lines of evidence that these complexes occur in vivo. First, the kinetic data in rabbit liver microsomes providing evidence for heteromeric P450–P450 interactions are consistent with the effects reported using defined reconstituted systems . Secondly, the BRET data show that CYP1A2–CYP1A2 complexes can form when expressed in HEK-293T cells (Figure 6), and that these complexes are not the result of protein overexpression (Figure 7). Taken together, the data strongly supports the idea that P450s exist in the endoplasmic reticulum membrane as a mixture of homomeric and heteromeric complexes. Changes in the relative amounts of a particular P450 enzyme, such as alterations caused by exposure to a P450 inducer, would be expected to change the composition of these complexes and potentially lead to alterations in mono-oxygenase function.
bioluminescence resonance energy transfer
cytochrome P450 1A2
cytochrome P450 2B4
cytochrome P450 2E1
Dulbecco's modified Eagle's medium
green fluorescent protein
human embryonic kidney
cytochrome P450 reductase
James Reed and Wayne Backes conceived and designed the study. The experiments were conducted by James Reed, Patrick Connick, Dongmei Cheng and George Crawley. All authors were involved in the data analysis. James Reed, Patrick Connick, Dongmei Cheng and Wayne Backes wrote the majority of the paper.
This work was supported by the National Institute of Environmental Health Sciences [grant numbers R01 ES004344 and P42 ES013648].