Separate roles for FMN and FAD in catalysis by liver microsomal NADPH-cytochrome P-450 reductase.

Rat liver microsomal NADPH-cytochrome P-450 reductase was prepared free of detectable amounts of FMN by a new procedure based on the exchange of this flavin into apoflavodoxin. The resulting FMN-free reductase binds NADP in the oxidized state with the same affinity (Kd = 5 microM) and stoichiometry (1:1 molar ratio) as does the native enzyme. Both the native and FMN-free reductase catalyze rapid reduction of ferricyanide, but the ability to reduce th 5,6-benzoflavone-inducible form of the liver microsomal cytochrome P-450 (P-450LM4) is lost upon removal of FMN. The FMN-free enzyme was reconstituted with artificial flavins which, in the free state, have oxidation-reduction potentials ranging from -152 to -290 mV, including 5-carba-5-deaza-FMN and several FMN analogs with a halogen or sulfur substituent on the dimethylbenzene portion of the ring system. Enzyme reconstituted with 5-carba-5-deaza-FMN has catalytic properties which are not significantly different from those of the FMN-free reductase, and is unable to reduce P-450LM4. On the other hand, the ability to reduce P-450LM4 and the other FMN-dependent activities of the native reductase are restored by substitution of several other analogs for FMN, but the kinetics of P-450LM4 reduction, studied under anaerobic conditions by stopped flow spectrophotometry, are significantly altered. The oxidation-reduction behavior of enzyme reconstituted with 7-nor-7-Br-FMN is substantially different from that of the native enzyme, and less thermodynamic stabilization of the semiquinone is observed with this flavin analog. In contrast, the oxidation-reduction properties of enzyme containing 8-nor-8-mercapto-FMN are similar to those of the native enzyme, but the spectral properties are significantly different. As shown in a stopped flow experiment, reduction of this FMN analog precedes reduction of P-450LM4 when a complex of the flavoprotein and P-450LM4 is allowed to react with NADPH. Our experiments support a sequence of electron transfer in this enzyme system as follows: NADPH leads to FAD leads to FMN leads to P-450. We propose that the enzyme cycles between a le- and a 3e-reduced state during turnover and that electrons are donated to acceptors via the reaction, FMNH2 leads to FMNH ..

constituted with 'I-nor-7-Br-FMN is substantially different from that of the native enzyme, and less thermodynamic stabilization of the semiquinone i s observed with this flavin analog. In contrast, the oxidation-reduction properties of enzyme containing 8-nor-8-mercapto-FMN are similar to those of the native enzyme, but the spectral properties are significantly different. As shown in a stopped flow experiment, reduetion of this FMN analog precedes reduction of P -4 5 0~ when a complex of the flavoprotein and P-4501.~~ i s allowed to react with NADPH.
Our experiments support a sequence of electron transfer in this enzyme system as follows: NADPH -+ FAD + FMN -+ P-450. We propose that the enzyme cycles between a le-and a 3e-reduced state during turnover and that electrons are donated to acceptors via the reaction, F"NH2 --* FMNH-.
* This research was supported by Grants GM-20877, GM-11106, and AM-I0339 from the National Institutes of Health and by Grant PCM76-14947 from the National Science Foundation. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "nduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
f Present address, Department of Biochemistry, Rice University, Houston, Texas 77001.
The flavoprotein NADPH-cytochrome P-450 reductase is a component of the microsomal mixed function oxidase system which catalyzes hydroxylation reactions of primary importance in the metabolism of lipids, drugs, and ot.her foreign compounds (1,2). This liver microsomal system consists of the reductase, several different forms of cytochrome P-450LM,' and phospholipid (3)(4)(5). Each of the protein components, including two inducible (6-10) and a major constitutive form (11) of P -4 5 0~~, as well as the reductase (12-16), have been purified following detergent solubilization of the membrane and characterized by numerous investigators. Recent studies using purified enzymes have provided evidence that the reductase and cytochrome P -4 5 0~~ form a catalytically signscant complex when the two proteins are bound in a 1:l stoichiometric ratio (17-19), a result which implies that the concentration of reductase, which is generally present in the microsomal membrane in lower amounts than the cytochrome, may significantly affect the overall rate of metabolism of various substrates by P -4 5 0~~ under physiological conditions. Detailed studies of the kinetics of P -4 5 0~~~ reduction also suggest that the phenomenon of biphasic reduction, known for many years to be characteristic of the system as studied in the microsomal membrane (20, Zl), may result, at least under certain conditions, from a catalytic property of the flavoprotein (17). In addition, the reductase also transfers electrons to cytochrome b5, and it was recently demonstrated with purified enzymes that it may replace the NADH-cytochrome b5 reductase to support NADPH-dependent desaturation of fatty acids (22). Thus, an understanding of the mechanism of electron transfer by the reductase is of interest not only to those studying the P-450~~-cataiyzed hydroxylation reactions in purified and reconstituted systems, but also to those with a more general interest in the electron transport and associated reactions of the intact microsomal membrane.
The microsomal P-450 hydroxylase system differs from those found in P s e u d o~o n~~~u t i d u (23) and in adrenocortical mitochondria (24) in which an iron-sulfur protein is required as an intermediate electron carrier between the flavoprotein and the Cytochrome. The microsomal flavoprotein, which supplies the two reducing equivalents from NADPH required for the P-450-catalyzed hydroxylation reaction, without the involvement of an iron-sulfur protein, contains one molecule each of FMN and FAD (12,25), rather than a single molecule of FAD, is the case in the other two systems. The precise details of the mechanism by which the microsomal reductase carries out the electron transfer reactions which in the other systems are catalyzed by a 1-electron carrier (the iron-sulfur protein) is not yet entirely understood. Recent investigations with the purified flavoprotein, however, have led to the proposal that the two flavins have separate roles in catalysis (26) such that FAD is involved in the reaction with NADPH, and FMN is involved in the subsequent t.ransfer of electrons to P-4501AM (27,28). This hypothesis is attractive since it provides an analogy to the bacterial and mitochondrial systems, in which the role of the FAD-cont.aining flavoprotein is to accept reducing equivalents from NAD(P)H in a single 2-electron reduction step, and then to reduce 2 molecules of an ironsulfur protein, which in turn provide electrons to P-450 in successive 1-electron transfer steps (29,30). The roles of FAD and FMN in the NADPH-cytochrome P-450 reductase might be envisioned in a similar manner, with FMN performing the reactions carried out by the iron-sulfur protein.
Support for this hypothesis has come both from studies on the oxidation-reduction properties of the purified protein and experiments with enzyme from which FMN was selectively removed. From the spectrophotometric and potentiometric measurements carried out by Iyanagi et al. (26) on the trypsinsolubilized form of the reductase, it is clear that the midpoint potentials of the two flavins, for which E ' o values of -190 and -328 mV were estimated, are quite different. Although these experiments were conducted with a form of the enzyme which had lost a membrane-binding segment as a result of proteolytic cleavage (31, 32), subsequent studies showed that the oxidation-reduction properties of the intact detergent-solubilized enzyme are essentially identical (15,16). The results of these static equilibrium experiments, in which the flavoprotein was titrated with reductants, indicate that the enzyme is reduced in four 1-electron steps and that the protein stabilizes the semiquinone of each flavin as a neutral, blue radical. Studies on enzyme from which FMN was selectively removed showed that FAD is the flavin of lower potential. It was also concluded that the flavin which remains partialky reduced after air-oxidation of the native enzyme is FMN in other words, the air-stable semiquinone form is FMNH.-FAD (28). Removal of FMN results in a loss of the ability of the enzyme to function with P -4 5 0~~ and to reduce certain artificial electron acceptors including cytochrome c and DCIP, yet the enzyme is still readily reduced by NADPH; ferricyanide reductase activity is only slightly affected by FMN removal and the ability to catalyze a transhydrogenase-type reaction is also retained (28). These observations are consistent with a role for FMN in the transfer of electrons to P-4 5 &~, but additional supporting evidence was sought in the studies to be described here.
Several implications of the working hypothesis outlined above may be addressed experimentally. If FAD serves as the sole acceptor for reducing equivalents from NADPH, there need be only one binding site for pyridine nucleotide, and for FMN to be directly involved in electron transfer to P-4501>", the two flavins must interact. Since it has been demonstrated that the air-stable semiquinone form of the enzyme (FMNH . -FAD) does not efficiently reduce P -4 5 0~~ (17) or cytochrome c (33) until reacted with NADPH, a mechanism involving electron transfer from FMNH2 rather than FMNH . would be favored. TO provide evidence for participation by FMN in electron transfer to P-450L~ we have taken advantage of the fact that FMN-free reductase binds a large number of artificial flavins very tightly and describe here both the catalytic and spect.ra1 properties of enzyme reconstituted with FMN analogs. In particular, the kinetics of reduction of p-450LM,, a reaction well characterized with the native enzyme in earlier work (17), are compared, and a detailed mechanism for the reduction of the oxidized cytochrome is proposed. In addition, data are presented on the stoichiometry of pyridine nucleotide binding.

EXPERIMENTAL PROCEDURES
Protein concentrations were estimated by the method of Lowry et a[. (34) using bovine serum albumin as the standard. Unless otherwise stated, all enzyme solutions were prepared in buffer containing 10% glycerol and phosphate buffers were the potassium salt. Spectra were recorded in a Cary 219 spectrophotometer with the temperature maintained at 10-12°C by use of a Lauda K2-R circulating water hath. Extinction coefficients of 21. 2 cm"' (15) and 10.2 mM" cm" were used to calculate the concentration of native and FMNfree reductase, respectively, at the wavelength of maximal absorbanee in the 450 nm region. Cytochromes P -4 5 0 1 .~~ and P -4 5 0 1 .~~ were purlfied from rabbit liver microsomes (9,351 and concentrations were determined from the absorbance of the reduced carbon monoxide complex using extinction coefficients reported by Haugen and Coon (9). The concentration of NADP solutions was determined following conversion to NADPH in the presence of isocit,rate dehydrogenase and isocitrate, using an extinction coefficient of 6.22 mM" cm" at 340 nm for the reduced pyridine nucleotide. Fluorescence measurements were made with a ratio-recording spectrofluorimeter (built by D. P. B.) which has been described in detail elsewhere (36).
Preparation of FMN-free Reductase-Detergent-solubilized reductase was purified from phenobarbital-induced rat liver microsomes as described previously (15) with a slight modification (28) to prevent proteolysis during subsequent treatment with high concentrations of KBr, Flavodoxin was purified from Megasphera ebdenii (37) and the apoprotein was prepared by dialysis versus 2 M KBr at pH 3.9 (38).
An extinction coefficient of 26.7 mM" cm" at 278 nm was used to calculate the apoprotein eoncentration (38). Reductase (2 ,UM) and apoflavodoxin (7 PM) were incubaced together in a solution containing 0.05 M sodium pyrophosphate, 0.2 mM EDTA, 2 M KBr, and 20% glycerol, which was adjusted to pH 8.3 with dilute HCl, for a period of 12 to 14 h. These conditions serve to lower considerably the affinity of FMN for the reductase but do not appreciably change the binding constant of FMN for apoflavodoxin. The result is that the FMN of the reductase is exchanged into flavodoxin. The protein solution was then dialyzed for 24 h against 1 0 0 volumes of 0.05 M phosphate buffer, pH 7.7, containing 0.2 mM EDTA and 208 gkycerol, with one change of the buffer solution. These operations were conducted at 4*C in the dark. Henex 690 was added to a final concentration of 0.1% and the solution was applied to a column of 2',5'-ADP-agarose. The column was washed with the same buffer solution to remove flavodoxin and the reductase was eluted with 2 mM 5"AMP; excess detergent was removed and the protein was concentrated in a manner identical with that described in the earlier procedure (28). Overall recovery was about 75% and the preparations had a specific activity of at least 57 pmol of cytochrome E reduced/min/mg when assayed in 0.3 M phosphate, pH 7.7, following preincubation with FMN.
Artificial Flavins-Iso-FMN and 5-deaza-FMN were gifts from Dr. P. Hemmerich and Dr. S. Ghisla, Universitat Konstanz, and 7-Br-FMN was a gift from Dr. S. G . Mayhew, University College, Dublin. The sources of materials and procedures for the preparation of 8-CI-and 8-mercapto-FMN are described elsewhere (39). Analogs were purified by the method of Mayhew and Strating (40) prior to use. Anaerobic Methods a n d Stopped Flous Apparatus-Procedures for the preparation of anaerobic solutions have been described previously (15,41). Photoreduction of the FMN-free reductase with 3,113dimethyl-5-deazaisoalloxazine as the catalyst (42) was performed under conditions outlined in earlier studies (28). Stopped flow spectrometry (system constructed by D. P. B.) was carried out with a single beam instrument designed for anaerobic work and interfaced to a Data General Nova 2 minicomputer. Routine use was made of the computer to program the high voltage to the photometer for spectra, as well as to correct for nonlinearities in the base-line.
Materials-The sources of materials used for the purification and assay of t.he reductase are the same as used in previous work (15,28). Potassium bromide was obtained from the J. T. Baker Chemical Co. and protocatechuate-3.4-dioxygenase from P. putida was generouslg provided by Dr. C. Bull. Water was glass distilled before use.

RESULTS
Preparation and Characterization of Reductase Free of FMN-The procedures previously described for the preparation of FMN-depleted reductase involved dialysis or ultrafiltration in the presence of 2 M KBr at slightly alkaline pH (27, 28); however, enzyme prepared by these methods still contained -10% residual native enzyme. The method for removal of free FMN from the solution apparently limits the efficiency of these procedures. In more recent studies, the use of apoflavodoxin to bind the flavin dissociated from the reductase was found to be much more effective. Although FMN may be dissociated from flavodoxin by treatment with KBr, acidic rather than alkaline conditions are required (38, 43); thus, a t pH 8.3, binding of free flavin to apoflavodoxin promotes the dissociation of FMN from the reductase. After incubation of a mixture of native reductase and apoflavodoxin for 12 h as described under "Experimental Procedures," more than 98% of the initial cytochrome c reductase activity (an FMN&pendent activity of the reductase) is lost. This activity is completely regained following incubation of the enzyme with FMN. Since the loss of FMN does not abolish the ability of the reductase to interact with pyridine nucleotide, separation of the reductase from the flavodoxin can be accomplished by affinity chromatography on 2',5"ADP-agarose, The flavin released from the final preparation of reductase shows an 11 r+_ 0.3-fold increase in fluorescence when hydrolyzed with Naja naja venom phosphodiesterase, the expected value for a sample containing only FAD (44). Furthermore, when the enzyme is reduced and exposed to air, no long wavelength absorbance attributable to the air-stable FMN semiquinone is detected in the spectrum of the reoxidized enzyme, an additional observation indicating that the preparations are virtually free of FMN.
Spectrophotometric Measurement of NADP Binding to Oxidized Native a n d FMN-free Reductase-Binding of oxidized pyridine nucleotide results in a perturbation of the flavin spectrum of both native and FMN-free reductase. The difference spectra generated by the addition of excess NADP to the oxidized enzyme are compared in Fig. 1A. The spectral changes are similar but not identical. Titration of either form of the enzyme with NADP generated a series of spectra with a single set of isosbestic points, as shown in Fig. 1B for the native reductase. From a plot of absorbance changes uersus NADP added, shown in Fig. 1, C and D, for native and FMNfree reductase, respectively, it is apparent that binding of pyridine nucleotide is essentially stoichiometric at the lower concentrations. Extrapolation of the linear portion of the titration curves indicated that, in both cases, the maximal absorbance change was produced after the addition of 1 mol of NADP/mol of enzyme. A dissociation constant of 4.4 p~, calculated from the experiment with the native enzyme, is very close to the K, of 6 PM which has been determined in kinetic experiments a t 30°C (15). A dissociation constant of 6.4 PM was calculated from the experiment with the FMN-free reductase.
For the native enzyme, values for the K , for NADPH, determined in 0.1 M phosphate buffer either in the reaction with P-4501.M or with cytochrome c, are in the range from 2 to 5 p~ (14,15), and inhibition by NADP is competitive with respect to NADPH. These observations all suggest that NADPH and NADP are bound with similar affinities and at the same site. Thus, it seems reasonable to conclude from the experiments summarized above that the reductase contains a single binding site for pyridine nucleotide and that the interaction is not Significantly altered when FMN is removed from the protein.
General Properties of Enzyme Containing Artificial Fla-uins-The artificial flavins used in reconstitution studies, with one exception, contain modifications in the dimethylbenzene portion of the flavin ring system, and have midpoint potentials ranging from -290 to -152 mV. The oxidation-reduction potentials of the free flavin and dissociation constants for binding to the FMN-free reductase are summarized in Table   I. Each analog is tightly bound by the FMN-free enzyme and cannot be removed by dialysis against phosphate buffer, pH 7.7. AS is the case with FMN, binding is accompanied by a significant perturbation of the flavin spectrum and, for fluorescent analogs, a dramatic quenching of their fluorescence. After reduction with excess NADPH under aerobic conditions, the analog-containing proteins underwent a reoxidation process resembling that seen with the native reductase. The FAD radical was transiently observed prior to reoxidation of FMN, and, with the exception of the 5-deaza-FMN-enzyme, oxidation of the FMN analog produced a semiquinone of low reactivity toward oxygen. In no case, however, was the partially reduced enzyme containing the analog semiquinone as resistant to further oxidation as was the native enzyme. The relative reactivity of the le-reduced semiquinone forms in air The catalytic activities of enzyme containing the artificiaI flavins are presented in Table 11 along with those of the FMNfree reductase. For the purpose of comparison, the data are expressed relative to the activity obtained for enzyme reconstituted with FMN; previous studies have already shown that the activities of the native enzyme altered or lost as a result of FMN removal by KBr treatment are restored when FMN is added back (28).
For the native enzyme, the rates of NADPH oxidation with the le-acceptors, cytochrome c and ferricyanide, and the 2eacceptor, DCIP, are very similar and about 20-fold greater than that obtained in the assay for P-450 reductase activity (15). It should be pointed out that, in the latter case, the reaction rate does not necessarily reflect the rate of P-450 reduction, but will be determined by the slowest step in the overall hydroxylation which includes the reaction of the reductase .P-450 complex with NADPH, substrate, and oxygen to yield products. It should also be emphasized that the data in the table were obtained using assay conditions known to be optimal only for the native enzyme.
The 5-deaza-FMN-enzyme had catalytic properties which did not differ significantly from those of the FMN-free reductase; the ability to function with P -4 5 0~~~ or to reduce cytochrome c and DCIP was not restored. Each of the other replacements resulted in an enzyme which regained the FMNdependent activities of the native reductase, although in no with those obtained with FMN. With the iso-FMN-, 7-Br-FMN-, 8-Cl-FMN-, and 8-mercapto-FMN-enzymes, the rates of NADPH oxidation in the presence of P-45oLM,, with benzphetamine as substrate for the hydroxylation reaction, were 26 to 30% of the control with FMN. In an experiment not shown, the apparent K , of P-450" for the iso-FMN-reductase (which showed the smallest rate of NADPH oxidation) was determined and found not to differ significantly from that for the native enzyme; thus, it is unlikely that the slower '-:<-t :ons may be attributed to a reduced affinity of the ..til rial reductases for the cytochrome. It was also confwmed :I] other experiments that substrate hydroxylation occurs in ti-::. reaction with P-4501.~, when each of these artificial enzymes is substituted for the native reductase. Enzyme containing the halogen-substituted analogs reduced the le-acceptors, cytochrome c and ferricyanide, at rates equal to or slightly higher than with FMN-reconstituted enzyme, but i"' ..*,ante were the reaction rates with all acceptors identical

I1
Activity of reductase reconstituted with FMN analogs A solution of FMN-free enzyme (7.7 p~) in 0.1 M phosphate (pH 7.7) containing 10% glycerol and 0.1 mM EDTA was incubated with a 1.7-to 2.0-fold excess of the free flavin at 4°C for a t least 90 min prior to dilution to a concentration appropriate for assay. For the experiments with 8-mercapto-FMN, an incubation mixture containing an equivalent amount of Na2S but with FMN in place of the analog and a mixture containing Na2S but without added enzyme or flavin were also prepared to determine the effect of Na2S on the activity of the reconstituted enzyme and on the nonenzymatic reaction rates, respectively. Reaction rates were determined spectrophotometrically at 30°C in 0.1 M phosphate buffer. P -4 5 0 1 .~~ reductase activity was determined at pH 7.4 in the presence of phospholipid and with benzphetamine as the substrate for the hydroxylation reaction (15). Other activities were determined at pH 7.7 using assay conditions and extinction coefficients described previously (15). The data are expressed as per cent relative to that obtained for the FMN-free enzyme plus FMN, for which the following activities, per mg of protein, were obtained: P-45&~,, 1.55 pmol of NADPH oxidized/min; cytochrome c, 49.2 pmol reduced/min; DCIP, 26.2 pmol reduced/min; ferricyanide, 68.6 pmol reduced/min; and 3-AcPyADP, 1.32 pmol reduced/min. The values for the control incubation containing Na&, FMN, and enzyme were not significantly different from those obtained with FMN alone.  Table 111. A to C, reaction traces obtained with native, FMN-free reductase reconstituted with FMN, and the FMN-free reductase, respectively. FMN-free enzyme to a level as great as observed with FMN present.

P -4 5 0~~~
Reduction by Enzyme Containing Artificial Flavins-The results discussed above indicate that, in the complete hydroxylation system with P-450L~2, the turnover of NADPH is significantly altered when FMN is replaced with artificial flavins. Since the rate of NADPH oxidation in these experiments was always at least an order of magnitude smaller than the rate of reduction of other electron acceptors such as ferricyanide, this observation suggested that a reaction involving electron transfer to P-45oLM (rather than reduction of the flavoprotein by NADPH) might be rate limiting in the complete hydroxylation system and prompted a more direct study of the effect of FMN replacement on the ability of the enzyme to reduce P -4 5 0~~. T h e conversion of oxidized P-45oLM to the reduced form may be monitored spectrophotometrically under anaerobic conditions by following the formation of the reduced carbon monoxide complex at 450 nm. This reaction is the first of two le-transfers which must occur in the catalytic cycle under aerobic conditions. Although the experiments in Table  I1 were carried out with P-450LMz, comparison of the reduction reactions with the artificial enzymes was performed with P-450LM4. In the recent studies by Oprian et al. (17), a kinetic analysis for the reaction of the native reductase with this form of p-45oLM was presented and the optimal experimental conditions for the reduction reaction were determined. The presence of substrate is not required for rapid reduction of €' -450tM,, and the reaction, which is monitored in a stopped flow spectrophotometer following rapid mixing of a preformed complex containing phospholipid and stoichiometric quantities of flavoprotein and cytochrome with NADPH, is biphasic.
As shown in Fig. 2, reduction of P-450LM4 by the FMN-free reductase is extremely slow. When FMN is added back, how-ever, the ability to rapidly reduce the cytochrome is restored, and the kinetic trace obtained is virtually identical with that seen with the native enzyme. In experiments reported elsewhere (49), a similar result was obtained with P-450LM2 as the acceptor. Thus, the very low activity of the FMN-free reductase with P -4 5 0~~, (see Table 11) is clearly due to a loss of the ability to reduce the cytochrome.
The rate constants obtained in a series of experiments with native and FMN-free reductase reconstituted with various flavins are presented in Table 111. In Experiment 1, in which FMN was replaced with iso-FMN, the analysis indicated that this reaction is also biphasic, but that the rate constants are both about 25% of those obtained with the native enzyme. In Experiment 2, conducted with different preparations of reductase and p-45&M4, the reactions of FMN-free reductase reconstituted with FMN, 8-Cl-FMN, 7-Br-FMN, and 8-mercapto-FMN were compared; a representative kinetic trace obtained in each case is shown in Fig. 3. The difference in reaction rates with these analogs was quite pronounced, particularly for enzyme containing the halogen-substituted analogs, where the rate constant for the rapid phase was only about 10% of that observed with the FMN-reconstituted enzyme. The reason for the lower reaction rates obtained with FMN (Experiment 2) is not known. The 8-mercapto-FMN-reductase was the only Reduction of P -4 5 0~.~, by NADPH-cytochrome P-450 reductase reconstituted with FMN analogs Reduction of P-459M4 was folIowed a t 448 nm under anaerobic conditions after rapid mixing of a preformed enzyme complex containing the flavoprotein, cytochrome, and phospholipid with an equal volume of a solution containing NADPH and phospholipid. The gas phase was CO, and the reaction was carried out in a stopped flow spectrophotometer with a 2-cm light path at 25°C. Before the enzyme solution was made anaerobic, the FMN-free reductase was incubated with a 1.2-fold excess of the indicated flavin for 45 min at 4OC, and the cytochrome, lipid, and buffer were then added. The final concentrations after mixing the two anaerobic solutions were as follows: P-4501.~<, 1.1 p M ; flavoprotein, 1.1 p M ; dilauroylglyceryl-3-phosphorylcholine, 30 pg/ml; potassium phosphate, pH 7.4,O.lO M; and NADPH, 100 PM. In Experiment 11, both solutions contained an oxygen-scrubbing system consisting of protocatechuate dioxygenase (57 nM iron) and protocatechuate (50 PM). Absorbance changes and rate constants for the two phases were calculated assuming two first order reactions as described by Oprian et al. (17) with appropriate correction in the determination of kl for the slower reaction. The values shown represent the average and standard deviation for a t least four reaction traces. " Due to a pronounced lag at the beginning of this reaction, the absorbance change in the fast phase was determined by subtraction of the calculated absorbance changes in the slower phases (see text) from the observed total change, rather than by extrapolation as described in Ref. 17. one of the artificial enzymes studied which reduced P-4501.~~ in a rapid reaction that occurred on the same time scale as for the native reductase. This reaction was more complex than with the other enzymes; although only two rate constants are given, the reaction contained a slower third phase which accounted for about 10% of the total 448 nm absorbance change. As indicated in Table 111, the reaction with the 7-Br-FMN-and 8-Cl-FMN-enzymes, although biphasic, showed a substantially larger percentage of the total absorbance change in the slow phase when compared to FMN. These data establish that the kinetics of P-4501.~~ reduction are significantly altered when FMN is replaced with other analogs, and, in addition, there appears to be a correlation between the oxidation-reduction potential of the free flavin and the rate of reduction, with enzyme containing the flavins of higher potential exhibiting the slower reaction rates.

Characterization of 5-Deaza-FMN-reductase-The activ-
ities of this artificial enzyme were essentially the same as seen with the FMN-free reductase, suggesting that 5-deaza-FMN is catalytically inert. The 5-deaza-FMN-enzyme was also the only one of the artificial enzymes studied which failed to produce a stable radical species following air oxidation of NADPH-reduced enzyme. In fact, the spectra obtained during turnover in air and subsequent reoxidation all retained the 400 nm absorption band characteristic of oxidized 5-deaza-FMN; thus, there was no indication that the artificial flavin was ever reduced. The experiments in Fig. 4 show that the inability of this flavin to restore FMN-dependent activity results from the fact that it is chemically unable to react with FAD. In Fig. 4 A_, an anaerobic solution of FMN-free reductase was reduced photochemically and a stoichiometric amount of 5-deaza-FMN was then added. Oxidized 5-deaza-FMN has absorption maxima at about 340 and 400 nm, and only very slight spectral changes were observed in these regions during the 30-min period following mixing. The difference between the final spectrum and that of the starting FMN-free reductase was obtained by subtraction, and, as judged by the well defined shoulder at 425 nm, is clearly the spectrum of enzymebound 5-deaza-FMN in the oxidized state. When the solution was subsequently exposed to air, the reoxidation process was identical with that previously described for the FMN-free reductase (28), an additional observation which confirmed that electron transfer from FAD to 5-deaza-FMN had not occurred. When an experiment of this nature is carried out with FMN or 7-Br-FMN, oxidation of FAD by the added flavin can be demonstrated (50). Although it was not expected that the oxidation-reduction potential of the bound 5-deaza-FMN would be low enough to prohibit electron transfer from  Interaction of 5-deaza-FMN in the oxidized or reduced form with the FMN-free reductase. A, an anaerobic solution of FMN-free reductase (7.1 PM) in 0.15 M phosphate, pH 7.4, containing 10% glycerol, 3 mM EDTA, and 0.1% deoxycholate in a final volume of 1.00 ml was photoreduced and 5-deaza-FMN (6.9 nmol in 0.1 M phosphate, pH 7.0; 0.07 ml) was then added from a side arm of the sealed vessel. Curve 1 (solid line) is the photoreduced FMNfree reductase and Curve 2 (dashed Ziine) was recorded 30 min after the addition of 5-deaza-FMN. The dotted spectrum was obtained by subtraction of Curve 1 from 2, with correction for the volume change.
FAD, the experiment was performed in reverse (Fig. 4B) to unambiguously rule out this interpretation. If the problem is thermodynamic rather than kinetic in nature, the transfer in the reverse direction should proceed readily.
In this experiment, a stoichiometric amount of free, reduced 5-deaza-FMN was added to a solution of oxidized FMN-free reductase under anaerobic conditions. Slow spectral changes occurred at 320 nm, which is the absorption maximum for reduced 5-deaza-FMN (47, 51), but there was no evidence for reduction of FAD. The difference between the final spectrum and that of the starting FMN-free reductase was similar to the spectrum of reduced 5-deaza-FMN, but with a slightly shifted absorbance maximum and an -20% decrease in extinction. A decrease in fluorescence relative to that of the free, reduced deazaflavin was also noted. These results indicated that the artificial flavin was bound by the enzyme and remained in the reduced form. When exposed to air, the enzyme-bound 5deazaflavin still remained reduced, an observation which is not particularly surprising in view of the extremely poor reactivity of reduced 5-deazaflavins with oxygen (51). Subsequent addition of NADPH resulted in reduction of FAD without any indication of a change in the oxidation-reduction state of the 5-deaza-FMN.

Characterization of 7-Br-FMN-reductase-Reconstitution of the FMN-free reductase
with 7-Br-FMN or 8-Cl-FMN resulted in an artificial enzyme which reducedP-450~.~, at a rate substantially lower than with FMN present (see Table  111). In order to provide evidence that the marked changes in reactivity toward P-4501.~~ can be attributed to altered oxidation-reduction properties of the flavin bound at the FMN site, the 7-Br-FMN-enzyme was studied in some detail. Oxidation of reduced 7-Br-FMN-reductase by oxygen produces a le-reduced form, 7-Br-FMNH --FAD, similar to the air-stable semiquinone form of the native enzyme (50). The spectrum of the oxidized and the semiquinone form of the enzymebound analog are very similar to those of the enzyme-bound FMN, and the spectrum of the semiquinone is that of a neutral radical with an absorption maximum in the long wavelength region a t 585 nm. The results of a titration of the 7-Br-FMN- FIG. 5. Titration of 7-Br-FMN-reductase with dithionite under anaerobic conditions. FMN-free reductase was treated with a 1.5-fold excess of 7-Br-FMN for 1 h at 4°C and then dialyzed against 0.15 M phosphate, pH 7.7, containing 10% glycerol and 0.1 mM EDTA to remove unbound flavin. A solution of the reconstituted enzyme (17.6 p~, estimated using an experimentally determined extinction coefficient of 21.7 mM" cm" at 452 nm) in the above buffer mixture plus 0.1% deoxycholate in a final volume of 1.00 ml was titrated with 5.25-pl aliquots of a solution of dithionite. Spectrum A is the oxidized enzyme and Spectra B through E were recorded at the stages representing approximately 25,50, 75, and 100% reduction, as judged from the plot of absorbance changes at 585 nm shown in the inset. Volume changes were taken into account in the plot of absorbance changes but the spectra have not been corrected. 700 enzyme with dithionite, however, differ significantly from those obtained with the native reductase and are shown in Fig. 5. There are four discrete phases in the titration, as judged from the plot, shown in the inset, of 585 nm absorbance uersus dithionite added; only the spectra at the end of each phase are shown in the figure. During the first phase of the titration, there was an increase in absorbance at 585 nm but the total amount of semiquinone produced after the addition of -1 reducing eq/molecule of enzyme was much less than is observed with the native enzyme. The 585 nm absorbance then decreased as the addition of dithionite was continued and, at the stage representing about 50% reduction of the enzyme, the spectrum resembled a mixture composed primarily of oxidized and reduced flavin. Further additions of reduc-tant produced a second, more substantial increase at 585 nm and the long wavelength absorbance observed a t 75% reduction shows the characteristic features of the FAD semiquinone. The 7-Br-FMN-enzyme may be fully reduced by dithionite as shown but, with NADPH, reduction does not proceed beyond the 3e-reduced (75% reduction) stage (50). This latter observation indicates that the oxidation-reduction potential of the enzyme-bound FAD, which is -328 mV (26) and very near that of the NADPH/NADP couple, is not significantly perturbed by the presence of the artificial flavin.
The dissimilarities between the titration of 7-Br-FMN-and native reductase occur in the first half of the experiment. For the native enzyme, the midpoint potentials for reduction to the 2e-and 3e-reduced stages are not widely separated (26) and reduction of FMNH. is not complete before titration of FAD begins. The 7-Br-FMN was, however, completely reduced before production of FAD semiquinone was observed; thus, the midpoint potentials for both half-reactions involving enzyme-bound 7-Br-FMN are considerably higher than those for FAD. For enzyme-bound FMN, the two half-reactions, FMN/FMNH. and FMNH /FMNH2, are widely separated in potential so that there is considerable thermodynamic stabilization of the semiquinone and nearly the maximal amount of radical is produced after the addition of 1 reducing eq. However, the degree to which the enzyme stabilized the semiquinone form of the analog radical was not nearly as great and it was estimated from the titration data that only 16% of the total 7-Br-FMN was present as radical a t 50% reduction. A similar result was obtained from a titration with NADPH (50).
Since experiments with other enzymes in which the normal flavin is replaced with analogs have indicated that the change in the flavin oxidation-reduction potential observed on binding to apoprotein is a relatively constant value (52), it might be predicted that the potential of enzyme-bound 7-Br-FMN would be elevated to the same extent as with FMN (from -207 mV for free FMN to -190 mV when enzyme-bound). This would result in an artificial enzyme with a midpoint for the FMN analog of -137 mV (compared with -154 mV for free 7-Br-FMN). The experimental data are consistent with this hypothesis. Calculations based on the estimate of the amount of 7-Br-FMN radical produced in the dithionite titration (see Ref. 53) indicate that the half-reactions for the enzyme-bound analog are separated by -65 mV, rather than +160 mV for the native enzyme; thus, in the case of the artificial enzyme, the 7-Br-FMN radical state is thermodynamically destabilized compared to the considerable stabilization of the FMN radical found in the native enzyme. The oxidation-reduction properties of the 7-Br-FMN enzyme deduced from such experiments (see also Ref. 50) are compared with those of native enzyme (15. 26) in Fig. fi.

Characterization o f 8-Mercapto-FMN-enzyme-
The spectra of 8-mercaptoflavins, unlike those of normal flavins, are characterized by a single visible absorption band centered at 523 nm with an unusually high extinction coefficient of 30 mM" cm" (48). Binding of 8-mercaptoflavin to the FMN-free reductase resulted in a shift in the absorbance maximum to 550 nm (Fig. 7). The spectral characteristics of the resulting enzyme are typical of those found with proteins which stabilize normal flavins as the neutral radical (39). The 8-mercaptoflavin semiquinone has an absorption maximum a t 720 nm with an extinction coefficient of about 8 mM" cm", very similar to that observed for 8-mercaptoflavodoxin (39). The spectra obtained by stepwise reduction of the enzyme with dithionite are shown in Fig. 7. The 8-mercaptoflavin is reduced prior to FAD, producing semiquinone, as judged by the increase in absorbance at 720 nm (Curue B). Continued addition of dithionite leads to a decrease in absorbance at this wavelength, a slight increase in absorbance in the 600 to 650 nm region, and the appearance of a new absorption band around 350 nm. At about 50% reduction of the enzyme (Curue C), it is apparent that the equilibrium mixture contains substantial amounts of both 8-mercapto-FMN and FAD semiquinones. At 75% reduction (Curue D ) the only long wavelength-absorbing species remaining is FADH. ; thus the relationship among the various half-reactions is similar to that of the native enzyme in which the midpoint potentials for the addition of the second and third electrons are very close. There is also considerable thermodynamic stabilization of the 8-mercapto-FMN radical. An estimate based on the absorbance at 720 nm indicates that 66% of the total analog flavin is present as radical at 25% reduction of the enzyme. This behavior is in marked contrast to that seen with enzyme-bound 7-Br-FMN. Admission of air to the fully reduced 8-mercapto-FMN-enzyme resulted in the rapid return of absorbance to a spectrum very similar to that of Fig. 7, Curve B. This form of the enzyme was further oxidized at a much slower rate and the radical species was clearly that of the artificial flavin. This observation provides direct support for the previous conclusion that the partially reduced flavin of the air-stable semiquinone form of the native enzyme is FMN (28).
As indicated in Table 111 Reduction of 8-mercapto-FMN and P -4 5 0~~, monitored by stopped Bow spectrophotometry following reaction of the 8-mercapto-FMN-reductase*P-450~~, complex with excess NADPH. The experiment was conducted under anaerobic conditions in the presence of carbon monoxide as described for Table 111, Experiment I, except that the concentration of the reductase. P-4501,~~ complex after mixing was 5 PM and the phospholipid concentration was 100 pg/ml. unique absorption properties of this artificial enzyme offer an advantage over the native reductase in that the spectral changes associated with the reduction of FMN may be distinguished from those of FAD by proper selection of the wavelength for observation. Fig. 8 shows the absorbance changes at 524 and 448 nm during reduction of theP-4501.~~ . reductase complex with NADPH as has been described above. At 448 nm the absorbance changes represent primarily those associated with the reduction of P -4 5 0~~~. At 524 nm, which is isosbestic for the conversion of the oxidized cytochrome to the reduced carbon monoxide complex and nearly maximal for oxidized 8-mercapto-FMN (see Fig. 7 ) , a large decrease in absorbance was observed; a rate constant of about 140 min" was estimated from several traces a t this wavelength. On a longer time scale, the changes at 524 nm continued to show a further but relatively small decrease. In contrast to the observations a t 524, the trace at 448 nm representing cytochrome P -4 5 0~~~ reduction shows a pronounced lag at the beginning of the reaction, an observation strongly suggesting that reduction of 8-mercapto-FMN precedes reduction of the cytochrome. Observations a t 720 nm failed to provide clear evidence for either the formation or disappearance of 8-mercapto-FMN semiquinone on the time scale shown in the figure; however, turbidity problems precluded a stronger conclusion regarding changes at this wavelength. The magnitude of the absorbance decrease at 524 nm was, however, greater than expected for the conversion of oxidized 8-mercapto-FMN to the semiquinone and closer to the predicted absorbance change for full reduction of the artificial flavin. Even though no significant absorbance changes at 524 nm subsequent to the initial reduction of the 8-mercapto-FMN were seen during the time period of reduction of the cytochrome, the results may be reconciled with a mechanism involving electron transfer from reduced 8-mercapto-FMN toP-450LM4, provided that subsequent reduction of the 8-mercapto-FMN radical by FADHz is rapid.

DISCUSSION
The experiments described above provide considerable support for the proposal that the sequence of electron transfer within the liver microsomal hydroxylase system is as follows: Evidence that the reductase contains a single binding site for pyridine nucleotide was obtained by monitoring the spectral perturbation of the oxidized enzyme when titrated with NADP. The dissociation constants obtained with both the native and FMN-free enzyme are very similar, demonstrating that removal of FMN has little effect on the ability of the enzyme to bind NADP. The dissociation constant obtained in both cases is in the micromolar range. Since the experimentally determined oxidation-reduction potential of enzymebound FAD is essentially the same whether the enzyme is reduced by NADPH or dithionite (26), there is no indication for a preferential binding of oxidized pyridine nucleotide to the reduced P-450 reductase as is the case for the NADPHadrenodoxin reductase of the mitochond~al oxidase system (54) and the microsomal NADH-cytochrome b:, reductase (55).
Although the rate of reduction of the native and the FMNfree reductase by NADPH has not been directly measured in this work, earlier experiments with the protease-solubilized form of the enzyme (33,56) and recent studies with the native enzyme (57, 58) suggest that flavin reduction is the ratelimiting step in the overall catalytic reaction with acceptors such as cytochrome c and ferricyanide. Since the rate of ferricyanide reduction is considerably faster than the rate of P-450LM4 reduction, and is decreased only 25 to 30% by FMN removal, it may be concluded that the very slow reaction of the FMN-free enzyme with P -4 5 0~~ is not due to impairment of the reaction of the flavoprotein with NADPH. The same conclusion may be drawn in the cases of enzyme reconstituted with artificial flavins since, in each instance, a high turnover with ferricyanide was observed.
Of the FMN analogs examined, only 5-deaza-FMN did not restore FMN-dependent activities to the enzyme. Additional studies showed that the enzyme-bound artificial flavin is not reducible by NADPH, and furthermore, that it is unable to react with enzyme-bound FAD in either the oxidized or reduced form. Our observations on the catalytic properties of the 5-deaza-FMN-reductase show that the reactivity of FAD is not altered simply by the presence of bound flavin at the FMN site, and establish that the role of FMN cannot be purely structural. The inability of 5-deaza-FMN to substitute for FMN is not particularly surprising since the chemical properties of the 5-deazaflavins resemble more closely those of pyridines rather than flavins, and in no case has activity been demonstrated for a 5-deazaflavin-enzyme whose normal function involves 1-electron transfer (59).
One of the most important results of our studies with artificial flavins, both in providing mechanistic information concerning the electron transfer process and as a potentially useful tool for future studies with the complete hydroxylation system, was the finding that the kinetics o f P -4 5 0~~~ reduction were significantly altered by substituting other flavins for FMN. The reason for the biphasic nature of the reduction Of p.450LM by NADPH and reductase is not fully understood; however, recent studies have shown that this behavior is exhibited with a single purified cytochrome (p-4501.~~) and, therefore, cannot be explained by the presence of multiple forms of P-45t)r,M (17). In the present studies, it was found that the rate constants for both the fast and slow phases, as well as the percentage of the total reaction occurring in the fast phase, may be altered by replacing FMN with other flavin analogs. These results lend further support for the suggestion (17) that a catalytic property of the reductase rather than the cytochrome is responsible for this phenomenon.
The simplest interpretation of the results described here is that FMN catalyzes the electron transfer to oxidized P-4501,~; since studies with the native enzyme have essentially ruled out FMNH. as the species responsible for this electron transfer (17), direct participation by FMN very likely occurs via the reaction FMNH2 + FMNH .. Indeed, when the oxidationreduction properties of the 7-Br-FMN-reduct.ase, which reduces P -4 5 0~~~ at the lowest rate, were examined quantitatively, it was found that the midpoint potential for this halfreaction was greatly elevated relative to that of the native enzyme (Fig. 6). On the other hand, the relationship among the four oxidation-reduction half-reactions for the 8-mercapto-F~N-e~zyme appear to be very similar to those for the native enzyme. This artificial reductase catalyzed P-4501.~~ reduction as efficiently as does the native enzyme. Because of the unusual absorption properties of &mercapto-FMN, it was also possible to demonstrate that reduction of the analog preceded reduction of P -4 5 0~~~. It should be emphasized t.hat the FMN-free reductase is incapable of rapidly reducing P-450~". This observation is taken as strong evidence that FADH2, the species which would be produced in the initial reaction of the flavoprotein with NADPH, is not involved in direct electron transfer to oxidized P -4 5 0~~.
Even before the physiological role of the reductase was fully realized, studies on the reaction with cytochrome c and other artificial acceptors were initiated by Masters et al. (33,56) as a result of general interest in the mechanism of flavin-catalyzed electron transfer. Since it contains no known oxidationreduction groups other than flavin, the enzyme falls into the category of a simple flavoprotein of the dehydrogenase-electron transferase class (59). Another well studied example of an enzyme of this type with nonequivalent flavins is a component of the bacterial sulfite reductase. This protein also e / FMNH. FADHZ I contains FMN and FAD and has a similar electron transfer function, but in a large multienzyme complex (60) containing other oxidation-reduction centers. Although there are several differences between this flavoprotein and P-450 reductase in both the catalytic and oxidation-reduction properties, the mechanism proposed by Siegel et al. (61,621, which involves separate roles for FMN and FAD, may also apply to the reaction catalyzed by P-450 reductase. This mechanism dictates that the enzyme cycle between a le-and a 3e-reduced state, with reducing equivalents entering the enzyme in the reaction of NADPH with FAD. The earlier studies on the reaction of the protease-treated P-450 reductase (33,56) have already provided evidence that, in reduction of cytochrome c, the enzyme is never fully oxidized during turnover but cycles between two different reduced states; it is now understood that the more oxidized state is the le-reduced species, FMNH -FAD, referred to as the air-stable semiquinone and that this form of the enzyme will not reduce P -4 5 0~~ until it first reacts with NADPH (17). Fig. 9A is a representation of a turnover event for a two-flavin enzyme in which the role of FAD is to accept reducing equivalents from NADPH, with subsequent transfer of electrons to an oxidized acceptor in a reaction carried out by FMN. This mechanism is essentially that proposed by Siegel et al. (61,62) and, as these authors point out, provides a means by which reducing equivalents from NADPH may ultimately be transferred in two equipotential le-transfer steps. Fig. 9B shows an alternate scheme which accomplishes the same overall process, but the role of FMN in this case is indirect; transfer of electrons to oxidized acceptors is carried out by FADH-, which is produced in the intramolecular disproportionation of FADH2 by oxidized FMN. Electron transfer reactions catalyzed assuming this mechanism would show an FMN-dependence even though it is the FAD which interacts directly with acceptors. As a mechanism for catalysis by P-450 reductase, Scheme I3 (Fig.   9B) falls into disfavor in light of the following considerations. There is no reason to expect that, once formed, the interme- Proposed mechanism for the reduction of P -4 5 0~~ by reaction of a fully oxidized reductase-P-450~~ complex with diate 2e-reduced species which contains FAD in the oxidized state is not readily reduced by NADPH to yield the fully reduced enzyme. The fate of this Be-reduced species is indicated in both schemes by dashed lines. Under the restrictions of Scheme B, the fully reduced enzyme would be unreactive in electron transfer and represent a dead end. In the mechanism outlined in Scheme A (Fig. 9A), however, the 4e-reduced enzyme should be perfectly capable of reducing acceptors and re-entering the cycle as shown. Oprian et al. (17) have shown that the reaction of the P-4501,~,.reductase complex with NADPH in the concentration range from 2 ,~LM to 1 mM does not alter either the extent or the kinetics of the reaction; thus, reduction is not inhibited at high levels of NADPH. This behavior is incompatible with Scheme B, making it very probable that it is indeed the FMN which is directly responsible for electron transfer to cytochromes. Since the anaerobic experiments with P-450LM4 were performed with oxidized reductase rather than the 1-electron reduced form and monitor only one of the two electron transfers to P-4501.~~ which are required under aerobic conditions to support the hydroxylation reaction, the intermediates produced in these experiments are not necessarily those depicted in the schemes representing an entire turnover event. Fig. 10 shows the sequence of reactions envisioned during 1-electron reduction of P -4 5 0~~~ by a stoichiometric amount of oxidized reductase when the complex is allowed to react with excess NADPH in the absence of oxygen. The restrictions described for Fig. 9A were applied; thus, the only species capable of electron transfer to the oxidized cytochrome is FMNH2, and the reaction with NADPH involves only FAD. The Be-reduced species, FMN-FADH2, produced in the initial reaction with NADPH, is shown to be in rapid equilibrium with the two other 2-electron reduced forms of the enzyme. Based on the midpoint potentials for the various flavin half-reactions, FMNH*-FAD will predominate. This form of the enzyme may reduce P"i50LM, or react with a second molecule of NADPH. Since the potentials of the enzyme-bound FAD and the pyridine nucleotide couple are nearly the same, it is important to realize that reduction of 2e-reduced enzyme to the fully reduced form is also a thermodynamic equilibrium. Although two pathways may be followed in this scheme, in both cases the species remaining after reduction of P-450I,M, is the 3e-reduced enzyme. This mechanism predicts that, for reduction of the stable complex of reductase and P-45oLM, the stoichiometry should be 2 mol of NADPH oxidized/mol of P"k5oLM, reduced the experimentally determined ratio is 1.7 (17). We have considered the possibility that biphasic reaction kinetics result from the fact that two species of reductase are produced which are capable of reducing P-4501.~~. However, this hypothesis is weakened considerably by the fact that the reduction reactions with the complex containing either oxidized or le-reduced reductase are kinetically indistinguishable (17). The possibility that FADH. also reduces P-450~" has not yet been rigorously ruled out, however, and may still be considered as a possible explanation for a biphasic reaction.
The experiments reported here with reductase reconstituted with FMN analogs are all consistent with the mechanism of electron transfer for the reductase which is shown in Fig. 9A. We have provided evidence that the enzyme contains a single binding site for pyridine nucleotide and that the rate of reduction by NADPH is only minimally affected by removal of FMN. Preparation and characterization of FAD-free reductase will be required to fully rule out a possible role for FMN in reactions with NADPH. In addition, our preliminary experiments with the 8-mercapto-FMN-reductase suggest that study of this artificial enzyme, which is as efficient in the reduction of P-450r.M4 as the native enzyme, may be one of the few experimental approaches which allow the reactions of each flavin to be monitored independently.