Liver microsomal lipid enhances the activity and redox coupling of colocalized cytochrome P450 reductase‐cytochrome P450 3A4 in nanodiscs

The haem‐containing mono‐oxygenase cytochrome P450 3A4 (CYP3A4) and its redox partner NADPH‐dependent cytochrome P450 oxidoreductase (CPR) are among the most important enzymes in human liver for metabolizing drugs and xenobiotic compounds. They are membrane‐bound in the endoplasmic reticulum (ER). How ER colocalization and the complex ER phospholipid composition influence enzyme activity are not well understood. CPR and CYP3A4 were incorporated into phospholipid bilayer nanodiscs, both singly, and together in a 1 : 1 ratio, to investigate the significance of membrane insertion and the influence of varying membrane composition on steady‐state reaction kinetics. Reaction kinetics were analysed using a fluorimetric assay with 7‐benzyloxyquinoline as substrate for CYP3A4. Full activity of the mono‐oxygenase system, with electron transfer from NADPH via CPR, could only be reconstituted when CPR and CYP3A4 were colocalized within the same nanodiscs. No activity was observed when CPR and CYP3A4 were each incorporated separately into nanodiscs then mixed together, or when soluble forms of CPR were mixed with preassembled CYP3A4‐nanodiscs. Membrane integration and colocalization are therefore essential for electron transfer. Liver microsomal lipid had an enhancing effect compared with phosphatidylcholine on the activity of CPR alone in nanodiscs, and a greater enhancing effect on the activity of CPR‐CYP3A4 nanodisc complexes, which was not matched by a phospholipid mixture designed to mimic the ER composition. Furthermore, liver lipid enhanced redox coupling within the system. Thus, natural ER lipids possess properties or include components important for enhanced catalysis by CPR‐CYP3A4 nanodisc complexes. Our findings demonstrate the importance of using natural lipid preparations for the detailed analysis of membrane protein activity.


Introduction
Cytochrome P450 3A4 (CYP3A4) is the most abundant human cytochrome P450 (CYP). Along with its redox partner, NADPH-dependent cytochrome P450 reductase (CPR), it is membrane-bound in the endoplasmic reticulum (ER) and found predominantly in cells of the liver and small intestine. CYP3A4 has great pharmacological significance since it is responsible for the metabolism of about half of all medically prescribed drugs and many xenobiotic compounds [1][2][3]. CYPs are haem protein mono-oxygenases whose catalytic activity depends on the integrity of a multicomponent electron transfer chain. Eukaryotic CYPs in the ER depend for activity on electron transfer from NADPH via flavin adenine dinucleotide and riboflavin-5 0 -phosphate (FMN) moieties of the diflavin protein CPR, and the coordinated activity of these two proteins likely depends on their close association within the ER membrane. Although the fundamentally conserved reaction mechanism of this coupled enzyme system is understood in some detail [4][5][6][7], the influence of the complex composition of the surrounding membrane on catalytic activity is poorly understood. Furthermore, comparative studies of CYP catalytic activity are complicated by the various effects of homotropic and heterotropic cooperativity of substrate binding, homomeric and heteromeric complex formation, and different CPR : CYP ratios [8][9][10].
Eukaryotic CPR and CYPs each have an N-terminal membrane anchor domain (MAD) containing hydrophobic amino acids by which they are tethered to the ER [11][12][13]. An intact MAD on CPR, as well as the association of phospholipid, have long been recognized as requirements for the coupled activity of the two enzymes in reconstituted mammalian systems. Early methods to isolate mammalian CPR involved treatment of microsomes with trypsin, yielding a proteolytically cleaved form of the enzyme which lacked the MAD. This form of CPR was able to reduce cytochrome c, but failed to interact functionally with its natural CYP redox partners (reviewed in ref. [14]). Only when CPR was purified intact from microsomes using detergent was it first possible to reconstitute a functional CYP-mediated mono-oxygenase system. The activity also required the inclusion of phospholipid, the essential component of which was shown to be phosphatidylcholine (PC) [15,16]. An intact MAD to tether the CPR to the phospholipid membrane therefore appeared to be essential for reconstitution of the activity. However, the MAD is not merely required to tether CPR to the membrane, since competitive peptide binding indicated that reconstitution of activity required the direct interaction of the MAD of CPR with CYP [11].
Although the MAD could be the sole determinant of CPR binding to the ER, this is almost certainly not the case for CYP. Molecular dynamics simulations and other studies suggest that CYP makes extensive contact to the membrane with the face distal to the side of CPR interaction, and that it could sit partly immersed within the membrane. This intimate embedding within the membrane is possibly essential to the integrity of access tunnels through the protein structure by which substrates gain entry to the active site [13,[17][18][19]. Many CYP substrates are lipophilic, thus their access to the CYP active site is likely to be through the membrane [20,21]. This embedding within the membrane may explain why integrity of the MAD of CYP is not necessarily essential for reconstitution of the coupled mono-oxygenase system [22][23][24].
The ER has a complex phospholipid composition and has structural heterogeneity, and studies of CPR and CYP1A2 from the ER of rabbit liver show that these enzymes copurify with lipid microdomain fractions enriched in certain phospholipids [25,26]. How and to what extent different phospholipids influence the activity of the mono-oxygenase systems is not clear. The influence of different phospholipids has been studied by reconstitution in phospholipid vesicles. Generally it has been found that increasing the proportion of anionic phospholipid in lipid mixtures stimulates activity, and that specific anionic phospholipids, such as phosphatidic acid (PA), can have greater stimulatory effect. Enhanced activity has been proposed to be due to better binding and deeper insertion of protein within the membrane, and closer interaction between CPR and CYP promoting more effective coupling for electron transfer [27][28][29][30][31]. Furthermore, the stimulatory effect of specific anionic phospholipids has been shown to be dependent on an intact MAD on the CYP [23,24].
When CYP3A4 was incorporated with CPR into vesicle membranes of varied composition, maximal product formation rate was observed in vesicles composed of liver microsomal phospholipid. Compared with PC alone, a PC/PS (phosphatidyl serine) mixture of 3 : 1 enhanced the rate by about twofold, whereas liver microsomal phospholipid enhanced the rate by about fivefold. Moreover, this maximal rate was accompanied by a high degree of electron transfer coupling [32]. Not all CYPs, however, show the same responses. Enhancement of CYP3A1 reconstituted activity, which is highly dependent on coincorporation of cytochrome b5 in membrane vesicles, varied with different phospholipids according to the inclusion of detergents in the reaction. Detergent possibly ameliorated the effects of protein aggregation. Activity in liver microsomal phospholipid was enhanced by detergent, but to no greater extent than with PC. Furthermore, the importance of unsaturated fatty acid in PC was shown [33].
Phospholipid bilayer nanodiscs are disc-shaped membrane bilayers bound round the circumference by a double protein belt derived from apolipoprotein A1. As a means to study membrane protein incorporated in a native-like membrane in an aqueous environment, they offer the advantage over lipid vesicles of being more stable, and of being monodisperse and homogeneous in size, therefore affording greater control of stoichiometry and of avoiding the possibly complicating effects of aggregation [34][35][36]. The properties of human CPR and CYP3A4 have been studied in nanodiscs, both singly [37][38][39][40][41] and together as a functional pair [42,43]. However, nanodiscs have been little used to study the influence of different phospholipid composition on the activity of the enzymes. When CPR was incorporated monomerically in nanodiscs, changing the phospholipid composition from pure PC to a 50% PS/PC mixture caused flavin redox potential changes that were suggested could favour electron transfer from NADPH to CYP3A4. The redox potential difference across the flavin couple was increased in the PS/PC mixture compared with PC alone, providing greater impetus (driving force) for electron transfer. Thus, an increase in the content of anionic phospholipid was proposed to have an enhancing effect [37]. Membranes of mixed PS/PC (30% PS), furthermore, enhanced the activity of the CPR-CYP3A4 complex in nanodiscs, improving both the coupling and the NADPH-dependent substrate conversion rates (for two of the three substrates tested). It was suggested that this enhancement could also be due to effects on redox potential [43]. A drawback of these studies, however, was that none of these phospholipid mixtures resembled the natural composition of the ER.
No studies of CPR/CYP function so far have been made using nanodiscs of natural ER membrane composition. In our investigation we have chosen to exploit nanodiscs to study the influence of physiologically relevant membrane composition on the activities of CPR and CYP3A4. Through better control of the stoichiometry of membrane protein insertion into the nanodiscs, we hoped to avoid some of the inconsistency and variability due to aggregation and multimeric complex formation possibly inherent in previous studies involving lipid vesicles. For the assay of CYP3A4 mono-oxygenase activity, we have confined our investigation to a single substrate, 7-benzyloxyquinoline (7-BQ), the conversion of which to 7-hydroxyquinoline (7-HQ) can be measured conveniently by fluorimetry [44]. We have studied the effects of different membrane compositions on the activity of CPR, both alone, and in combination with CYP3A4 as a functional monodisperse 1 : 1 complex. In order to be physiologically relevant, we have compared enzyme activities within membranes composed of lipid extracted directly from liver microsomes, with PC alone, and in the case of the coupled enzyme complex with a mixture of component phospholipids designed to mimic the known composition of the ER. We have determined steady-state kinetic parameters in the various combinations, and investigated coupling in the dual enzyme system. We find that, compared with PC alone, liver microsomal lipid enhances the activities of both CPR and of the coupled enzyme pair, and furthermore liver microsomal lipid increases the extent of redox coupling during electron transfer between partner proteins.

Assembly of phospholipid bilayer nanodiscs containing CPR and CYP3A4
Phospholipid bilayer nanodiscs were assembled essentially as previously described [45], incorporating CPR or CYP3A4 either singly, or together as described by Denisov et al. [42]. Each nanodisc consists of two protein molecules derived from apolipoprotein A1, referred to as 'membrane scaffold protein' (MSP), arranged antiparallel to each other in the form of a double belt, within which the phospholipid bilayer membrane forms like the skin of a drum. Nanodiscs are assembled by mixing the constituent components together in appropriate ratio in detergent, then removing the detergent by adsorption to polystyrene beads. Combinations of gel filtration, affinity purification and nondenaturing and denaturing gel electrophoresis were used to demonstrate assembly and to purify nanodiscs containing inserted membrane protein. Premixing the assembly components in the appropriate ratio is important. Too great an excess of phospholipid can lead to the formation of lipid vesicles that can sequester membrane protein, and too little lipid can lead to protein aggregation.
The assembly of enzymatically active CPR in nanodiscs was demonstrated by native (nondenaturing) gel electrophoresis. Complexes within the gel which contained active CPR were identified by developing a staining procedure using the dye thiazolyl blue tetrazolium bromide (also known as methylthiazolyldiphenyl-tetrazolium bromide or MTT). MTT is readily soluble and forms a pale yellow solution. NAD (P)H-oxidoreductases catalyse the reduction of MTT to purple insoluble formazan [46]. When gels were soaked in MTT solution to which NADPH was added, formazan formed in the bands that contained CPR. Nanodisc complexes containing CPR gave rise to formazan staining, confirming that active CPR was present in these complexes (visible half way down the gel, Fig. 1A, first panel). However, when an N-terminally truncated form of CPR lacking the MAD was used in the assembly mixture instead of full-length CPR, formazan staining was not associated with the nanodisc complexes but was associated with a faster migrating band corresponding to free CPR (visible near the bottom of the gel, Fig. 1A, third panel); nanodiscs were still able to form in the latter mixture (shown by Coomassie Blue staining of the MSP half way down the gel, Fig. 1A, fourth panel) but they were devoid of any active CPR. These results therefore show not only that active CPR could be assembled within the nanodiscs but also that the MAD of CPR is essential for assembly of CPR in the nanodiscs. Nanodiscs were purified from the assembly mixture by adsorption to nickel-resin of the N-terminal polyhistidine tag (His-tag) contained on the MSP. CPR remained associated with the nanodiscs adsorbed to the resin, and was eluted together with the nanodiscs in 0.3 M imidazole (Fig. 1B). Further purification of the CPR-nanodiscs from nanodiscs lacking CPR was not necessary for the kinetic studies described below.
CYP3A4-loaded nanodiscs, and doubly loaded nanodiscs containing both CPR and CYP3A4, were assembled by similar methods. Purification was achieved by a combination of gel filtration and enrichment by nickel affinity, making use of the C-terminal His-tag of the CYP3A4, having first removed the Histag from the MSP by cleavage with tobacco etch virus protease. Doubly loaded nanodiscs were assembled after first preincubating CPR and CYP3A4 together at a 2 : 1 molar ratio prior to mixing with the MSP and lipid. This possibly facilitated complex formation between the two proteins, and in our hands tended to result in final ratios of CPR to CYP3A4 of about 1 : 1 in the purified assembled nanodiscs. Figure 2 shows protein extracted from assembled nanodisc mixtures separated according to particle size by gel filtration, for both CYP3A4-nanodiscs ( Fig. 2A) and CPR-CYP3A4-nanodiscs (Fig. 2B). Figure 2C shows the carbon monoxyferrous bound/unbound difference spectrum of nanodisc fractions containing CYP3A4 (those indicated in Fig. 2B), of which the characteristic prominent Soret peak at 450 nm is an indication of enzymatically active CYP [47]. Figure 2D shows how, following enrichment by nickel-affinity, CPR-CYP3A4-nanodiscs resolve as a single homogeneous peak by gel filtration containing all three constituent proteins: CPR, CYP3A4 and MSP.
The molar ratio of CPR to CYP3A4 in the doubly loaded nanodisc preparations was determined spectroscopically. A convolution spectrum was constructed by the linear combination of the absorbance spectra of nanodiscs individually loaded with CPR or CYP3A4 (Fig. 3A). This convolution spectrum was then fitted to spectra of doubly loaded nanodiscs using leastsquares regression (Fig. 3B,C), with the relative amount of each individual (CPR and CYP34A) spectra fitted for in terms of the concentration of CPR and CYP34A. Extinction coefficients of 0.022 lM À1 Ácm À1 (CPR FMN maxima at~450 nm) and 0.10 lM À1 Ácm À1 (CYP34A Soret maxima) were used to determine individual protein concentrations and the preparations of doubly loaded nanodiscs used for kinetic analyses all had close to equimolar ratios of CPR to CYP3A4 (Table 1,   In this study, we assembled nanodiscs not only with POPC but also with a crude preparation of natural phospholipid extracted directly from bovine liver microsomes ('liver lipid'), and also a mixture of component phospholipids comprising the main constituents of liver ER, mixed in proportions approximately to match the natural composition of ER ('mixed lipid'). The temperature of nanodisc assembly is recommended to correspond to the melting temperature of the lipid used [45]; for liver and mixed lipids, we expected this to be 20°C or above. A series of trials were performed (with no inserted membrane protein) to compare different ratios of lipid to MSP at different temperatures. The efficacy of nanodisc assembly was assessed by native gel electrophoresis and by gel filtration. Clear differences were not readily discernible by native gel electrophoresis, but the best resolved nanodisc complexes tended to form with liver lipid at 20°C rather than at 4°C or 37°C, and at mass ratios of MSP to lipid similar to those used for assembly of nanodiscs with POPC at 4°C (data not shown). Gel filtration profiles of nickel affinity-enriched nanodiscs assembled at three different ratios of MSP to liver lipid at 20°C were indistinguishable from the profile of nanodiscs assembled with POPC at 4°C (Fig. 4). About 20°C was therefore chosen as the standard temperature for the assembly of liver and mixed lipid nanodisc preparations, using the same lipid mass concentration as for POPC.

CPR in liver lipid nanodiscs has enhanced activity with cytochrome c compared with CPR in POPC nanodiscs
We sought to analyse the activity of the CPR-CYP3A4 system by comparing the steady-state kinetic parameters of the enzyme components, both singly and in combination, and in membrane-integrated and in soluble form, in order to understand the significance of membrane integration and the influence of membrane composition. The results of the kinetic analyses are compiled in Table 1. A variety of assays were applied to compare the activities. First CPR was investigated alone, that is in the absence of its redox partner CYP3A4. In vivo, CPR serves as electron donor to CYPs. In vitro, however, it can also catalyse the reduction of cytochrome c from the reduced FMN domain of CPR. This reaction was assayed by two methods. First, the rate of formation of reduced cytochrome c was determined by measuring the increase in absorption at 550 nm. Second, the rate of oxidation of NADPH was determined by measuring the decrease in absorption at 340 nm. Thus, steady-state kinetic parameters for the reaction were determined both with respect to cytochrome c reduction and NADPH oxidation (Fig. 5). Furthermore, CPR activity was assayed in four different forms: (a) the full-length protein solubilized in detergent; (b) a truncated form of the protein (DN-CPR) lacking sixty N-terminal amino acids comprising the MAD; (c) the full-length protein embedded in nanodisc membranes composed of POPC; and (d) the full-length protein embedded in nanodisc membranes composed of liver lipid. Initial reaction velocities were fitted to the Michaelis-Menten equation and the measured parameters for the different forms of CPR are listed in Table 1, rows 1-4; those determined with respect to NADPH consumption are in columns B, C and D, and those determined with respect to cytochrome c reduction are in columns E, F and G. Human CPR is membrane-bound in its native state. No enhancement of CPR activity due to membrane-binding was observed, as indicated by the catalytic efficiencies (k cat /K M ) of the membrane-bound forms, which are similar to, or less than, those of the soluble forms of CPR (Table 1, columns D and G, rows [1][2][3][4]. Cytochrome c is a soluble protein and membrane integration of CPR might be expected to have little effect on, or even to impede, the interaction of cytochrome c with CPR. The aqueously soluble DN-CPR (lacking the MAD) exhibited the highest k cat values (Table 1, cells 2B, 2E), although its catalytic efficiency was less dissimilar to the other forms (cells 2D, 2G). Upon comparing the effects of membrane composition on the activity of CPR in nanodiscs, consistent differences were observed: according to both assay methods, the catalytic efficiency of CPR in liver lipid nanodiscs was approximately twofold higher than in POPC nanodiscs. This is reflected in higher k cat values in the liver lipid nanodiscs compared with POPC nanodiscs; the K M values in the two different membrane compositions are equal (Table 1, columns B-G, rows 3, 4). Thus, CPR in membranes composed of liver microsomal lipid has enhanced activity compared with CPR in membranes composed only of POPC.

Membrane incorporation enhances Odebenzylation of 7-Benzyloxyquinoline by CYP3A4
CYP3A4 catalyses the O-debenzylation of 7-BQ to yield a fluorescent product, 7-HQ, which can be quantitated by measuring emission at 515 nm upon excitation at 400 nm. In order to establish the assay, cumene hydroperoxide was used as an electron donor. The steady-state kinetic parameters were then determined for the O-debenzylation of 7-BQ by both CYP3A4 solubilized in detergent, and by CYP3A4 incorporated in POPC nanodiscs (Table 1, columns E, F, rows 5, 6). Although the K M values for the two forms of the protein were equal, the k cat for CYP3A4 in the POPC nanodiscs was 50% higher than for CYP3A4 solubilized in detergent. Thus, incorporation of CYP3A4 into POPC membranes in nanodiscs appeared to confer some enhancement of activity over the detergent-solubilized form. The reaction rates of CYP3A4 with cumene hydroperoxide as the electron donor were much lower than the rates with NADPH and CPR. For this reason, the comparison with liver lipid nanodiscs was not made, as it was considered to be of less physiological relevance.
Cointegration of CPR and CYP3A4 in the same nanodiscs is essential for NADPH-dependent CYP3A4 substrate conversion Cytochrome P450 reductase is the natural electron donor for CYP3A4-mediated reactions, and both proteins in their natural state are membrane-bound in the ER. We sought to study the coupled function of CPR and CYP3A4 and investigate the influence of membrane integration. To this aim we combined various forms of the two proteins to study the effects on Odebenzylation of 7-BQ. The only combination of CPR with CYP3A4 that supported conversion of 7-BQ upon addition of NADPH was when both proteins were incorporated together within the same nanodisc (Fig. 6A,B). No combination involving any of the (membrane free) soluble forms of the proteins supported 7-BQ conversion. Furthermore, no conversion was observed when separate preparations of CPRnanodiscs and CYP3A4-nanodiscs were mixed together. Coincorporation of CPR and CYP3A4 within the same nanodisc phospholipid membrane, therefore, appears to be essential for the NADPHdependent O-debenzylation activity. O-debenzylation of 7-BQ by CPR-CYP3A4 in nanodiscs exhibits cooperativity with respect to substrate concentration Having confirmed that 7-BQ could serve as a substrate in an NADPH-dependent manner for CYP3A4 in combination with CPR in nanodiscs (Fig. 6B), the steady-state kinetic parameters of the reaction were determined both with respect to product formation rate, determined by measuring 7-HQ fluorescence, and with respect to NADPH depletion rate, determined by measuring NADPH absorbance (illustrated schematically in Fig. 6A). When initial rates from the reaction curves were plotted against substrate concentration, the curve for NADPH consumption best fitted the Michaelis-Menten equation (Fig. 6C), whereas the curve for product formation best fitted the Hill equation (Fig. 6D). The latter indicated a cooperative effect on reaction rate due to binding of the substrate. The Hill coefficients for 7-BQ O-debenzylation by the coupled CPR-CYP3A4 system in the three different membrane compositions had similar values; all had a value close to 2, indicating positive cooperativity of substrate binding (Table 1, cells 7H, 8H, 9H). Membrane composition therefore had no effect on the extent of the positive cooperativity of substrate binding.
Liver microsomal lipid enhances both turnover and catalytic efficiency of the coupled CPR-CYP3A4 enzyme pair in nanodiscs The effects of varying membrane composition on the steady-state kinetic parameters of the CPR-CYP3A4 system in nanodiscs was investigated. CPR and CYP3A4 were coincorporated into nanodiscs of three different membrane compositions: (a) POPC; (b) 'mixed lipid', consisting (by % mass) of 62% PC, 20% phosphatidylethanolamine (PE), 10% phosphatidylinositol (PI), 5% sphingomyelin (SM), 2% phosphatidylserine (PS) and 1% phosphatyidic acid (PA), comprised to resemble the phospholipid composition of the ER; and (c) 'liver lipid', phospholipid extracted directly from bovine liver microsomes. The kinetic parameters are listed in Table 1, columns B-G, rows 7-9. In each case, the ratio of CPR to CYP3A4 in the nanodisc preparations, determined by deconvolution of absorbance spectra, was close to unity (Table 1, cells 7A, 8A, 9A). The most obvious trend, on comparing these parameters is the enhancement of the coupled enzyme activity in the liver lipid nanodiscs, compared both to the nanodiscs composed of the mixed lipid, and to those composed of POPC. Catalytic efficiency in the liver lipid nanodiscs, as determined both by NADPH consumption and by 7-BQ O-debenzylation, is at least sixfold higher than in nanodiscs composed of the other lipids (Table 1, columns D, G, rows 7,8,9). This is a function both of higher k cat and of lower K M in the liver lipid nanodiscs (columns B, C, E, F, rows 7,8,9). Thus, not only does liver lipid enhance the activity of CPR alone in nanodiscs but it also enhances the activity of the coupled enzyme system, and to a greater extent than CPR alone. Differences between the mixed lipid and POPC were less apparent: a clear trend of increasing K M from liver lipid => mixed lipid => POPC is apparent by both assay methods (columns C, F, rows 7,8,9), but there was little difference in catalytic efficiency between the mixed lipid nanodiscs and the POPC nanodiscs (columns D, G, rows 7, 8).
CPR-CYP3A4 decoupling occurs in nanodiscs, but is greater in POPC membranes than in liver or mixed lipid membranes is apparent that there is a significant discrepancy: NADPH consumption is several fold higher than 7-BQ O-debenzylation, in all three membrane compositions ( Table 1, compare cells 7B, 8B, 9B with 7E, 8E, 9E). The discrepancy is highest with the POPC membranes, for which k cat with respect to NADPH consumption is 20-fold greater than with respect to 7-BQ O-debenzylation. With the mixed lipid and the liver lipid membranes, k cat with respect to NADPH consumption is 7-and 11-fold greater, respectively, than for 7-BQ O-debenzylation. These discrepancies suggest a degree of inefficiency or decoupling in the transfer of electrons from NADPH via CPR to CYP3A4, and that the decoupling is greater in the POPC membranes than in either the mixed lipid or liver lipid membranes.

Superoxide production accompanies CPR-CYP3A4 decoupling in nanodiscs
It has long been known that decoupling of electron transfer can occur in CPR-CYP systems [48]. Decoupling in the presence of molecular oxygen gives rise to the formation of reactive oxygen species, notably superoxide (O À 2 ). In order to demonstrate the production of superoxide during 7-BQ O-debenzylation by the CPR-CYP3A4 system we used the hydroethidine-derived fluorescent dye Mito-HE. Reaction of Mito-HE with superoxide forms a specific red-fluorescent 2-hydroxyethidium product (2-OH-Mito-E À ), which serves as an indicator of superoxide [49,50] (Fig. 7A). Mito-HE was added to the CPR-CYP3A4 nanodisc reaction mixture, with 7-BQ as substrate. The reaction was initiated by addition of NADPH, and the change in red fluorescence was monitored by measuring 600 nm emission upon 527 nm excitation. A marked increase in red fluorescence was initiated by the addition of NADPH, which continued throughout the course of the reaction, indicating the generation of superoxide (Fig. 7B). When Mito-HE was omitted from the reaction, no increase in red fluorescence was observed, as expected. In the absence of 7-BQ as substrate, when Mito-HE was included, an increase in red fluorescence was also observed, indicating that superoxide was generated by the system even in the absence of substrate; this increase, however, was less than when substrate was included (Fig. 7B). 7-HQ product formation was monitored simultaneously by measuring 515 nm emission upon 400 nm excitation. In the absence of Mito-HE, 7-HQ formation proceeded as expected from previous kinetic measurements. However, in the presence of Mito-HE, the rate of 7-HQ formation was significantly reduced (Fig. 7C). This indicated that Mito-HE had an inhibitory effect on the reaction; the reaction still proceeded, but at a reduced rate.

Time (min) 7-HQ fluorescence
Thus, in the presence of Mito-HE, the induction of red fluorescence upon NADPH addition indicates that superoxide is produced by the CPR-CYP3A4 system in the absence of substrate, but that superoxide production is significantly enhanced during substrate conversion. These results therefore suggest that electron 'leakage' occurs at multiple stages during electron transfer, both on transit between the flavin cofactors of CPR, and during the CYP3A4 reaction cycle. This is consistent with what is already known about uncoupling in the CPR-CYP system, in that reactive oxygen species are known to be generated both in the presence and absence of the CYP substrate upon addition of NADPH [48].

Discussion
We have used phospholipid bilayer nanodiscs to investigate the significance of membrane insertion and the influence of membrane composition with respect to the activity of human CPR and CYP3A4. It has long been appreciated that trypsin-solubilized CPR isolated from liver microsomes lacks a portion of the N terminus required to tether the protein to microsomal membranes, and that this soluble form while retaining the capacity to catalyse NADPH-dependent reduction of cytochrome c, is unable to couple effectively with CYP and support CYP mono-oxygenase activity [11,51]. The detergent-solubilized full-length form of the protein could be incorporated spontaneously into microsomes or synthetic phospholipid vesicles, whereas the cleaved form could not [52]. In assembling CPR into nanodiscs, by observing the colour conversion of MTT in native polyacrylamide gels, we have been able to confirm in a very clear and direct way the essential requirement for the N-terminal domain in tethering CPR to the phospholipid membrane. The full-length heterologously expressed CPR clearly became incorporated into nanodiscs, whereas the 60-amino acid Nterminally truncated form (DN-CPR) did not (Fig. 1A). Upon incorporation of CYP3A4 into nanodiscs, and having developed a convenient fluorimetric assay for the enzyme using 7-BQ as substrate (initially using cumene hydroperoxide as electron donor), we tested which combinations of various forms of CPR with CYP3A4-nanodiscs would support reconstitution of the coupled NADPH-dependent activity. The only condition which supported reconstitution of the activity was when CPR and CYP3A4 were coincorporated in the same nanodiscs. No activity was detected when CPR-nanodiscs and CYP3A4-nanodiscs were assembled separately then mixed together, nor when either of the soluble forms of CPR (either the full-length detergent-solubilized or the DN-CPR form) were mixed with the CYP3A4-nanodiscs. The requirement for membrane insertion of CPR for reconstitution of the coupled activity has been previously inferred from, first, the requirement for an intact MAD on CPR [12,51,52], and second, the requirement for phospholipid in the reconstitution mixture [16]. Taken together, our results are a direct demonstration of the requirement for coinsertion of CPR and CYP3A4 within the same phospholipid membrane for reconstitution of activity. Interestingly, yeast CPR lacking the N-terminal MAD does support CYP-dependent sterol biosynthesis, both in vivo and in a reconstituted assay [53]. The explanation for this apparent qualitative difference between yeast and mammalian CPR is not known. It is possible that the N-terminal MAD of mammalian CPR is required to interact directly with CYP as well as to serve as a membrane anchor [11]. The requirement for similar interaction in yeast has not been demonstrated, and it may be that the residues required for this interaction are lacking from the yeast CPR Nterminal domain.
In looking at the effect of membrane insertion on CYP3A4 activity, we found that turnover (k cat ) was slightly enhanced by insertion into nanodiscs, compared with the nonmembrane-bound detergent solubilized CYP3A4 (Table 1, cells 5E, 6E). This result is not unexpected. It is consistent with molecular dynamics studies that suggest that one face of the protein could lie well embedded within the membrane surface and thus facilitate the access of amphiphilic substrates, which have the potential to accumulate within the membrane [13,[17][18][19].
When the activities of various forms of CPR alone were compared, with an assay using cytochrome c as a substrate, DN-CPR consistently showed the highest rates compared with the full-length CPR, with respect to both NADPH consumption and cytochrome c reduction (Table 1, cells 2B, 2E). It has been reported previously that trypsin cleavage (which results in loss of the N-terminal MAD [12]) enhances the activity of mammalian CPR with respect to cytochrome c reduction by a factor of up to 1.25 [51]. Our data show rate enhancements (k cat ) for the N-terminally truncated CPR of from 1.7-to 5-fold. It is not clear why the DN-CPR should exhibit higher rates, but presumably loss of the MAD makes CPR more freely accessible to interact with cytochrome c, which, unlike the natural CYP redox partner of CPR, is small, not membranebound and freely diffusible in aqueous medium. It is known that CPR adopts multiple conformations and that each will react differently with cytochrome c [54,55]. Truncation of the MAD in the DN-CPR is expected to affect the conformational landscape of CPR.
When inserted into nanodiscs, the activity of CPR with cytochrome c was consistently higher in membranes composed of liver microsomal lipid compared with membranes composed of POPC, in terms of both the catalytic constant (k cat ) and the catalytic efficiency (k cat /K M ), and according to both NADPH consumption and cytochrome c reduction (Table 1, rows 3, 4). As far as we are aware, this is the first report of enhanced activity of membrane-bound CPR alone in liver lipid compared with PC. The redox potentials of CPR incorporated alone in nanodiscs have been compared in POPC membranes and in membranes composed of 50% POPC/POPS, and have been found to differ [37]. The differences in potential between the various flavin states, it was proposed, were such as to promote more efficient electron transfer in the more anionic POPC/POPS mixture than in POPC alone. Whether the anionic phospholipid content of liver microsomal lipid could influence flavin redox potential so as to cause the enhanced CPR activity we observed in liver lipid nanodiscs is not known. The 50% POPC/ POPS mixture is not physiologically relevant, and the total content of anionic phospholipid in liver microsomes is substantially less than 50% [56,57]. The mechanism by which CPR reaction rates are enhanced by liver lipid, therefore, remains to be determined. Liver lipid substantially enhanced the coupled activity of CPR-CYP3A4 in nanodiscs, determined using the 7-BQ O-debenzylation assay. Furthermore, the enhancement of the activity of the coupled system with respect to 7-BQ O-debenzylation was greater than the enhancement of the activity of CPR alone with respect to cytochrome c ( Table 1, compare 4E/3E with 9E/7E, and 4G/3G with 9G/7G). With respect to NADPH consumption, the degree of enhancement was similar with regard to k cat ( Table 1, compare 4B/3B with 9B/ 7B) but markedly greater with regard to catalytic efficiency (compare 4D/3D with 9D/7D). This extra enhancement of the coupled enzyme activity suggests that we have detected an important mechanistic phenomenon relevant to the activity of the enzyme system in its natural state in the ER. It indicates that membranes composed of liver lipid have properties or contain components that are important for efficient catalysis by the coupled CPR-CYP3A4 system, which are lacking from membranes composed purely of POPC. The phenomenon affects CPR alone, but it affects the coupled CPR-CYP3A4 complex to a markedly greater extent. This enhancement by liver microsomal lipid is consistent with that previously observed for CPR-CYP3A4 in lipid vesicles, using different substrates [32]. Our results therefore confirm and expand the relevance of this previous study.
We sought to investigate what the special properties of the liver lipid membrane might be by performing a further comparison which involved incorporation of CPR-CYP3A4 into nanodiscs composed of a mixture of purified phospholipids combined in proportions to reflect the natural composition of the ER (PC 62%, PE 20%, PI 10%, SM 5%, PS 2% and PA 1%). However, in this case there was no enhancement compared with POPC (Table 1, row 8). We conclude, therefore, that the properties of, or components within, natural liver lipid which enhance CPR-CYP3A4 activity must be in addition to, or not directly related to, the combination of phospholipids in our ER-mimic mixture. Natural liver microsomal lipid contains additional components other than those included in our mixture, some of which are known to influence enzyme activity. Cholesterol, for example, is known to have significant effects on the activities of a number of membrane proteins [58] and can constitute as much as 10% (by phosphorus content) of the lipid in rat liver microsomal preparations [56,57]. Also, the anionic phospholipid cardiolipin, which has been reported to have a specific effect in stimulating the activity of human CYP1B1 in a manner dependent on the protein having an intact N-terminal MAD [24], was not included in our phospholipid mixture. Cardiolipin is a relatively minor phospholipid component of the ER (~1% [57]) and the effects described with CYP1B1 were observed at unnaturally high concentrations. However, cardiolipin has been suggested to have a specific role as a cofactor in the regulation of the activity of the yeast cytochrome bc1 complex, in which it has been proposed to maintain the structural integrity of proton-conducting channels [59]. At present, however, the properties of liver microsomal lipid which enhance the activity of CPR-CYP3A4 in nanodiscs remain unknown and must remain the focus of future study.
The CPR-CYP3A4 activity in nanodiscs revealed a significant degree of decoupling. That is, NADPH was consumed at a rate higher than could be accounted for by substrate conversion (Table 1, compare 7B, 8B, 9B with 7E, 8E, 9E). This implied that the system was 'leaking' electrons and this leakage was expected to give rise to the generation of superoxide (illustrated schematically in Fig. 7A). It was possible to demonstrate the generation of superoxide using Mito-HE, which upon reaction with superoxide is converted to a product (2-OH-Mito-E + ) with diagnostic fluorescent signal. Superoxide was not only generated during NADPH-dependent CPR-CYP3A4-substrate conversion but also, although to a lesser extent, in the absence of substrate (Fig. 7B). Superoxide generation in the absence of substrate suggests electron leakage from the CPR flavin-coupled electron transfer, as has been described previously [60], while the enhanced superoxide generation observed in the presence of substrate presumably represents the sum of leakage both from the CPR flavin-coupled electron transfer and from the CYP3A4 reaction cycle [48]. Addition of Mito-HE to the coupled reaction mixture resulted in partial inhibition of 7-BQ conversion (Fig. 7C). The cause of this is not understood, but possibly Mito-HE competes with 7-BQ as substrate for CYP3A4.
The extent of decoupling was affected by the composition of the nanodisc membranes. Compared with POPC, liver microsomal lipid not only enhanced CPR-CYP3A4 activity but also tightened the coupling of electron transfer (Table 1, compare 9B/9E with 7B/ 7E). The mixed phospholipid (ER mimic) membrane also enhanced coupling, to the same extent as liver lipid (Table 1, 8B/8E), although unlike liver lipid it failed to enhance activity. This suggests that the membrane properties or components responsible for enhancing activity are not necessarily the same as those responsible for tightening redox coupling, since the mixed phospholipid appears to enhance one but not the other, whereas the liver microsomal lipid enhances both. Cardiolipin has been found to reduce the output of reactive oxygen species by human CYP2E1 in phospholipid vesicles in a manner dependent on an intact N-terminal MAD [23]. Also both cardiolipin and PA have similar MAD-dependent effects in enhancing the activity of human CYP1B1 [24]. It is conceivable, therefore, that the PA in the mixed phospholipid membrane could affect coupling in our study. However, further detailed comparisons under carefully controlled conditions will be necessary to clarify such influences. By studying CPR-CYP3A4 as a defined 1 : 1 coupled complex in nanodiscs, we have shown at least that such comparisons are feasible.
In conclusion, the properties of nanodiscs that enable the preparation of monodisperse particles of relatively even size and of defined composition have allowed us to demonstrate in a very direct way the importance of tethering CPR together with CYP3A4 within the same membrane in order to reconstitute NADPH-dependent mono-oxygenase activity. We have furthermore shown how natural liver lipid has properties which enhance this activity, compared with membranes of synthetic phospholipid composition. The use of nanodiscs has enabled us to define this phenomenon with precision, with greater control over the stoichiometry of the enzyme system. Our findings emphasize the importance of using natural lipid preparations for the detailed study of the activity of membrane proteins in order better to understand the regulatory effects imposed by the membrane environment.

Experimental procedures
Protein production and purification All protein purification steps were performed at 4°C or in ice slurry. Plasmids pMSP1D1 and pMSP1E3D1 for expression of MSPs were obtained from Addgene (www.addgene.org). MSPs were expressed and purified as described by Ritchie et al. [45], except that expression was induced in batch cultures of 0.5 L Terrific Broth (TB) in 2-L flasks shaken at 30°C and 200 r.p.m. for 5-6 h after induction.
Plasmid pPORh1 for expression of human CPR (derived by substitution of human for rat CPR cDNA in pOR263 [61,62]) was kindly provided by Bettie Sue Masters, University of Texas Health Science Centre at San Antonio, TX, USA, who also provided a method on which the following procedure was based. CPR from pPORh1 in Escherichia coli has an N-terminal E. coli outer membrane protein A (OmpA) leader to target the protein for periplasmic membrane insertion. A single colony of E. coli C41 (DE3) transformed with pPORh1 was used to inoculate 5 mL 'Lauria-Bertani' broth (LB) broth containing 50 lgÁmL À1 carbenicillin. After 5-6 h of growth at 37°C, the culture was diluted in 250 mL TB containing 50 lgÁmL À1 carbenicillin, and allowed to grow overnight at 37°C. Six 2 L flasks, each containing 0.5 L TB, 50 lgÁmL À1 carbenicillin, were each inoculated with 30 mL of the overnight culture and grown at 28°C at 180 r.p.m. to an optical density of 0.8-1.0 (1 cm light-path at 600 nm wavelength). The incubation temperature was then decreased to 21°C for 30 min, 0.5 mM riboflavin (a precursor of the flavin cofactor) was added, CPR synthesis was induced by addition of IPTG at 0.4 mM final concentration, and the culture was grown for a further 24-30 h at 21°C. Cells were harvested by centrifugation and stored at À80°C.
Thawed cells were suspended in two volumes of lysis buffer (100 mM Tris-HCl, pH 7.6, 1 mM EDTA, 10% glycerol, 1 mM DTT), the suspension was stirred until homogeneous, lysozyme and DNase I were added, each to final concentration 10 lgÁmL À1 , 'Complete EDTA-free' protease inhibitor tablets (Roche Products Ltd, Welwyn Garden City, UK) were added (2 per 100 mL), and the suspension was stirred gently for a further 30 min. Cells were disrupted by sonication, coarse debris was removed by centrifugation [17 000 g (max), 20 min], then microsomal membranes were sedimented by further centrifugation [180 000 g (max), 1 h]. The membrane-containing pellets were suspended to homogeneity in buffer C (50 mM Tris-HCl, pH 7.7, 0.1 M EDTA, 10% glycerol, 0.05 mM DTT, 1% Triton X-100) in an homogenizer with motor-driven rotating Teflon pestle, two extra protease inhibitor tablets were added, and the homogenized suspension was stirred gently overnight to solubilize the CPR. The extract was then clarified by a second centrifugation step at 180 000 g (max) for 1 h.
Cytochrome P450 reductase was purified from the clarified supernate by affinity chromatography on 2 0 ,5 0 -ADP Sepharose 4B, as follows. Protein was applied to a 15-mL bed-volume of resin pre-equilibrated in buffer C, washed further with buffer C, then with 100 mL of 5 mM adenosine in buffer C. A two-stage gradient of increasing concentration of 2 0 (3 0 )-adenosine monophosphate (mixed isomers) in buffer C was applied to elute the CPR: a first stage of 0-0.5 mM (100 mL), followed by a second stage of 0.5-5.0 mM (100 mL). The two-stage elution effectively separates the intact, full-length CPR from a proteolytically cleaved form which lacks the MAD. After excision of the appropriate band from a polyacrylamide gel, the full-length protein was sent for N-terminal amino acid sequence determination (Alta Biosciences Ltd, University of Birmingham, Birmingham, UK) and found to have the N terminus Gly-Iso-Pro-Gly-Ser, which corresponds to the last five amino acids of the OmpA leader; this corresponds to the same N terminus mapped for other OmpA fusions [63]. Fractions containing full-length CPR were pooled then concentrated with a 30-kDa molecular weight exclusion filter. CPR concentration was determined according to the bound flavin content by absorbance measurement at 454 nm, using an extinction coefficient e = 21.4 mM À1 Ácm À1 [64]. Purified CPR in buffer containing 10% glycerol (v/v) was frozen in liquid nitrogen and stored at À80°C. A His-tagged, soluble form of human CPR, lacking the 60 amino acids from the N terminus comprising the MAD (DN-CPR), was expressed and purified as described previously [65].
Plasmid pB84 for expression of C-terminally His-tagged human CYP3A4 was kindly provided by R. Wolf, University of Dundee, UK, who also provided a method on which the following purification procedure was based. pB84 is derived from a plasmid for membrane protein expression in E. coli, derived from pCW as previously described [66]. Like CPR, CYP3A4 was expressed with an OmpA leader sequence as periplasmic membrane targeting signal, which is cleaved off as part of the targeting process, leaving the protein associated with the membrane by its own N-terminal MAD. E. coli strain K12 JM109 was transformed with pB84 and a single colony was used to inoculate 50 mL LB culture containing 50 lgÁmL À1 carbenicillin and grown at 37°C overnight. Several 2-L flasks containing 0.5 L TB, 50 lgÁmL À1 carbenicillin, were inoculated each with 6 mL of the overnight culture and grown at 37°C, 180 r.p.m., to optical density 0.8-1.0 (1 cm light-path at 600 nm). The incubation temperature was then decreased to 28°C, 1 mM 5-aminolevulinic acid (a precursor to the haem cofactor) was added, then CYP synthesis was induced by addition of 1 mM IPTG, and culture was continued for a further 20-24 h at 28°C. Cells were harvested by centrifugation and stored at À80°C.
Thawed cells were suspended in up to five volumes of lysis buffer: 50 mM potassium phosphate, 500 mM potassium chloride, pH 7.4, 20% glycerol. Lysozyme and DNase I were added each to 10 lgÁmL À1 final concentration, the suspension was stirred in ice until homogeneous, protease inhibitors were added (as for CPR), and cells were disrupted by sonication. CHAPS detergent was added to 10 mM final concentration and the lysate was stirred gently overnight. Debris was removed by centrifugation [48 000 g (max), 50 min], then the supernate was applied to a Ninitrilotriacetic acid agarose column (25 mL resin), pre-equilibrated with 50 mM potassium phosphate, 0.5 M potassium chloride, pH 7.4, 20% glycerol. The column was washed with five bed-volumes of the same buffer, then sequentially with the same buffer containing: (a) 5 mM CHAPS (b) 30 mM imidazole; and (c) 50 mM imidazole. CYP3A4 was then eluted in 0.4 M imidazole in 40 mM Tris-HCl buffer. Fractions containing CYP3A4 were pooled and dialysed against 100 mM potassium phosphate, 0.1 mM DTT, 0.1 mM EDTA, 5 mM CHAPS, pH 7.4, 20% glycerol, then concentrated with a 30 kDa molecular weight exclusion filter. The concentration of haem in the purified CYP3A4 preparations after dialysis was determined by pyridine haemochromagen assay [67,68]. This corresponded well with the concentration determined by 420 nm absorption using an extinction coefficient of e 420 nm = 100 mM À1 Ácm À1 . It was valid, therefore, to use the absorbance spectrum of the oxidized form of CYP3A4 in the calculation of CYP3A4/CPR ratios in the doubly loaded nanodisc preparations (see below).

Phospholipids and phospholipid mixtures
Phospholipids (from Sigma-Aldrich Ltd, Gillingham, UK or Generon Ltd, Slough, UK) were stored under nitrogen, either desiccated or dissolved in methanol/chloroform (2 : 1), in Teflon-capped glass vials at À20°C. Nanodiscs were assembled either with 2-oleoyl-1-palmitoyl-sn-glycero-3-phosphocholine (POPC), or a mixture of phospholipids intended to mimic the composition of the mammalian ER, or with crude phospholipid extracted from bovine liver microsomes. The phospholipid mixture to mimic the ER was composed as follows (% mass): PC 62%, PE 20%, PI 10%, SM 5%, PS 2% and PA 1% [56,57]. Liver microsomal phospholipid was prepared as follows. Ox liver fresh from slaughter was chilled on ice, diced and homogenized in a blender with 2-3 volumes of 0.25 M sucrose, 0.1 M potassium phosphate, pH 7.4. Coarse debris was removed from the slurry by two sequential centrifugation steps of 15 min at 14 000 g (max), microsomes were then sedimented from the supernate by ultracentrifugation (140 000 g, 1 h). The resulting deep red, glassy, gelatinous pellets were stored at À80°C until required. Phospholipid was extracted by the method of Bligh and Dyer [69], whereby microsomes suspended in a minimal volume of buffer were homogenized in methanol and chloroform in such proportion that a miscible system was formed; dilution with chloroform and water then caused phase separation, with lipid retained in the chloroform phase and nonlipid in the aqueous phase. Lipid yield was estimated by weighing the residue from an aliquot dried under nitrogen and lyophilized under vacuum. Appropriate quantities of phospholipid were dispensed in methanol/chloroform, dried and dissolved in 0.1 M sodium cholate, 20 mM Tris-HCl, pH 7.4, at 38 mgÁmL À1 (equivalent to 50 mM POPC) immediately prior to requirement for nanodisc assembly.

Assembly and purification of nanodiscs
Assembly of phospholipid bilayer nanodiscs was essential as previously described [42,45]. To constitute CPR into nanodiscs, CPR was mixed with the membrane scaffold protein MSP1D1 and cholate-solubilized POPC at a molar ratio of 1 CPR: 65 MSP1D1: 130 POPC in a solution containing 10 mM POPC, 20 mM sodium cholate, 0.1 M NaCl, 20 mM Tris-HCl, pH 7.4. For nanodiscs composed of liver microsome and mixed phospholipids, mass concentrations equivalent to POPC were used. After mixing by rotation for 1 h at 4°C, assembly was initiated by addition of an approximately equal volume of Amberlite XAD-2 beads (prewashed and equilibrated in buffer), and the total mixture was incubated on a roller for 20-24 h at 4°C. Assembly of nanodiscs with liver microsome or mixed lipid was performed at 20°C to account for the higher melting temperature of these lipid mixtures. Purification of the assembled nanodiscs was by Ni affinity to the His-tag of the MSP. After removal of the Amberlite beads, the assembly mixture was applied to Ni-nitrilotriacetic acid resin (or HisTrap HP column) pre-equilibrated in 40 mM Tris-HCl pH 7.4, 0.3 M NaCl, and washed successively with 10 and 50 mM imidazole, then eluted with 0.3 M imidazole in the same buffer. Purity of the eluted fractions was assessed by SDS/PAGE. Nanodiscs could also be enriched by gel filtration (Superdex 200, or Sephacryl S-200 or 300, GE Healthcare Life Sciences, Little Chalfont, UK), and analysed by nondenaturing PAGE [precast TGX gels (Bio-Rad Laboratories, Watford, UK) or 9% acrylamide (30 : 0.8 mono/bis) in 25 mM Tris-glycine, pH 8.3, at 100 V]. Nanodisc-containing fractions were pooled, dialysed against 20 mM Tris-HCl, 0.1 M NaCl, pH 7.4, concentrated by filtration and stored at À80°C.
The method to constitute CYP3A4 into nanodiscs was similar to that for CPR, but to purify CYP3A4-nanodiscs, the His-tag was first removed from the MSP by cleavage with tobacco etch virus protease, leaving the His-tag on CYP3A4 alone available for affinity purification. Ni-affinity purification was then performed as for CPR nanodiscs.
For simultaneous incorporation of both CPR and CYP3A4 into nanodiscs, the larger MSP1E3D1 was used, better to accommodate both proteins, the His-tag first having been removed to allow affinity purification via the His-tag attached to CYP3A4. CYP3A4 and CPR were premixed and incubated together for 30-60 min at room temperature to facilitate possible binding between the two proteins, then all components were mixed in a molar ratio of 0.1 CPR: 0.2 CYP3A4: 1 MSP1E3D1: 90 POPC: 180 cholate, as recommended [42]. Assembly then proceeded as described above for CPR-nanodiscs. Following Ni-affinity enrichment (via the His-tag attached to CYP3A4), CPR remained associated with the nanodisc at approximately equimolar ratio with CYP3A4; gel filtration was then sufficient to give an homogeneous preparation of CPR-CYP3A4-nanodiscs.
CPR : CYP3A4 ratios were determined by absorbance spectroscopy using a NanoDrop 2000 UV-vis spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA), with fivefold repetition of readings. The two singly loaded nanodisc spectra in Fig. 3A were used as basis spectra and a convolution spectrum was created by the linear combination of these spectra. Fitting of the convolution spectrum to the doubly loaded nanodisc spectra allowed the concentration and molar ratios of the two proteins in the doubly loaded nanodisc preparations to be estimated. The fitting method was based on an algorithm to minimize root-meansquare deviation between the experimental and convolution spectra. The fitted spectra matched the scanned spectra well, with mean root-mean-square deviation approximating to 0.025 across the wavelength range. Preparations with CPR : CYP3A4 ratios close to 1 : 1 were used for kinetic analyses.

Enzymic reactions
Detection of CPR activity in native gels by reduction of thiazolyl blue tetrazolium bromide Following electrophoresis, gels were rinsed twice in water, soaked in a minimal volume of freshly mixed 0.5 mM NADPH, 1 mM MTT, 0.1 M potassium phosphate, pH 7.6, incubated at 37°C for up to 30 min to allow colour development, then rinsed further in water and photographed.

NADPH-dependent cytochrome c reduction by CPR
Cytochrome P450 reductase was first treated with potassium ferricyanide to oxidize the flavin cofactors, then repurified by passage through a desalting column. Reactions were performed in disposable plastic cuvettes in a UV-visible spectrometer within an anaerobic glovebox. CPR and cytochrome c were mixed in 0.3 M potassium phosphate, pH 7.4, at 25°C, in a range of concentrations, and reactions were initiated by the addition of NADPH. Change in absorbance was recorded either at 550 nm to measure cytochrome c reduction, or 340 nm to measure NADPH consumption. Initial velocities were fitted to the Michaelis-Menten equation.

Fluorescence assay of O-debenzylation of 7benzyloxyquinoline by CYP3A4
O-debenzylation of 7-BQ by CYP3A4 gives rise to the fluorescent product, 7-HQ, which can be detected by measuring emission at 515 nm upon excitation at 400 nm. Enzymes or nanodisc preparations were mixed in 0.1 M potassium phosphate, 7.5 mM magnesium chloride, pH 7.4, with 7-BQ (0.1-200 lM) in fluorescence quartz cuvettes. After equilibration at 37°C, reactions were initiated by addition of 150 lM NADPH. For reactions of CYP3A4 in the absence of CPR, cumene hydroperoxide was substituted as the electron donor. Product formation rates were determined from the initial linear phase of fluorescence increase. A range of concentrations of 7-HQ in the same buffer were used to create a standard curve for product concentration, which gave the conversion factor of 16.75 relative fluorescence unitsÁlM À1 . Initial velocities were fitted to the Hill equation. NADPH consumption rates, calculated from the absorbance decrease at 340 nm, were measured under the same conditions, and initial velocities were fitted to the Michaelis-Menten equation.
Fluorescence detection of superoxide formation using Mito-HE Mito-HE (MitoSOX Red; Thermo Fisher Scientific) was dissolved at 5 mM in DMSO and diluted to 10 lM in 1 mL of 7-BQ O-debenzylation reaction mixture (described above), but also containing 1 mgÁmL À1 salmon sperm DNA to enhance fluorescence of the product, Mito-2-hydroxyethidium (2-OH-Mito-E + ), formed upon reaction of superoxide with Mito-HE. 7-BQ O-debenzylation reactions were then initiated by the addition of NADPH. Production of 2-OH-Mito-E + was monitored by measuring fluorescence emission at 600 nm upon excitation at 527 [49,70].