Kinetic analysis of cytochrome P450 reductase from Artemisia annua reveals accelerated rates of NADPH-dependent flavin reduction



Cytochrome P450 reductase from Artemisia annua (aaCPR) is a diflavin enzyme that has been employed for the microbial synthesis of artemisinic acid (a semi-synthetic precursor of the anti-malarial drug, artemisinin) based on its ability to transfer electrons to the cytochrome P450 monooxygenase, CYP71AV1. We have isolated recombinant aaCPR (with the N-terminal transmembrane motif removed) from Escherichia coli and compared its kinetic and thermodynamic properties with other CPR orthologues, most notably human CPR. The FAD and FMN redox potentials and the macroscopic kinetic constants associated with cytochrome c3+ reduction for aaCPR are comparable to that of other CPR orthologues, with the exception that the apparent binding affinity for the oxidized coenzyme is ~ 30-fold weaker compared to human CPR. CPR from A. annua shows a 3.5-fold increase in uncoupled NADPH oxidation compared to human CPR and a strong preference (85 100-fold) for NADPH over NADH. Strikingly, reduction of the enzyme by the first and second equivalent of NADPH is much faster in aaCPR, with rates of > 500 and 17 s−1 at 6 °C. We also optically detect a charge-transfer species that rapidly forms in < 3 ms and then persists during the reductive half reaction. Additional stopped-flow kinetic studies with NADH and (R)-[4-2H]NADPH suggest that the accelerated rate of flavin reduction is attributed to the relatively weak binding affinity of aaCPR for NADP+.


cytochrome P450 reductase from Artemisia annua


Arabidopsis thaliana CPR (GenBank number ×66016)


cytochrome P450 reductase


ferredoxin NADP+-reductase


kinetic isotope effect








Artemisinin is an endoperoxidized sesquiterpene that is used in combination therapies to treat malaria, a global disease that continues to threaten the lives of > 200 million people [1]. Currently, artemisinin is sourced from the Chinese medicinal plant Artemisia annua, although poor extraction yields and infrequent harvests make the drug unaffordable, especially for afflicted individuals from poorer developing nations [2]. In an effort to reduce the cost, a method has been developed for the microbial production of the semi-synthetic precursor of artemisinin, artemisinic acid, by transforming Escherichia coli and Saccharomyces cerevisiae with enzymes from A. annua involved the biosynthetic pathway: amorpha-4,11-diene synthase, CYP71AV1 and cytochrome P450 reductase (CPR) [3, 4]. Amorpha-4,11-diene synthase catalyzes the first committed step of artemisinin biosynthesis: the conversion of farnesyl diphosphate to amorpha-4,11-diene [5]. CYP71AV1 is a cytochrome P450 monooxygenase that performs a three-step oxidation of amorpha-4,11-diene to artemisinic acid, and CPR functions in the sequential transfer of two electrons to CYP71AV1 [3, 6].

CPR contains two flavin cofactors, FAD and FMN, which mediate the flow of electrons from the obligate two-electron donor, NADPH, to single electron acceptors, such as the P450 haeme centre. Extensive studies of mammalian forms of CPR have revealed key mechanistic steps of catalysis [6-8]. In sequence, the steps include hydride transfer from NADPH to FAD, shuttling of a single electron from the FAD hydroquinone (FADH2) to the higher potential FMN cofactor, followed by the relay of a single electron from the reduced FMN to the P450 haeme [9-11]. Genome sequencing of several land plants (e.g. rice, poplar, grape and Arabidopsis) reveals that these organisms contain 250–450 individual P450s and one to three paralogues of CPR [12-14]. Similarly, yeast and animals have one gene for CPR and numerous P450s [15, 16]. It is envisioned that a single CPR passes electrons to multiple members of the microsomal P450 superfamily, thereby driving catalysis for an astounding array of metabolites. In eukaryotes, the CPR-P450 electron transfer complexes form on the cytoplasmic side of the endoplasmic reticulum where the P450s and CPR are localized via their respective N-terminal transmembrane anchors [12, 13, 17].

The crystal structures of mammalian CPR reveal that the protein is organized into functional motifs. The FMN domain is localized to the N-terminal half of the protein and is structurally homologous to bacterial flavodoxin [18, 19]. At the C-terminus, a ferredoxin NADP+-reductase (FNR)-like domain contains the binding sites for NADP(H) and FAD. The two flavin-binding motifs are separated by a connecting subdomain and a flexible hinge region that is ~ 15 amino acids in length [18]. Based on structural and solution studies of yeast and mammalian forms of CPR, the enzyme is considered to undergo large-scale domain motion in the process of transferring reducing equivalents from NADPH to the P450 haeme center [20-24]. Both in the crystalline-state and in solution, the enzyme adopts a compact state with the FMN domain docked adjacent to the FNR domain [25]. This conformation supports rapid interflavin electron transfer as it aligns the FAD and FMN cofactors, such that the isoalloxazine rings form a near continuous ribbon with the dimethyl benzene groups < 4 Å apart [18]. The compact conformation is not conducive for direct electron transfer from the FMN to the P450 haeme because the FMN cofactor is sequestered into the core of the protein and many of the negatively-charged groups that stabilize the FNR-FMN domain interface also mediate electrostatic interactions with the P450s [26]. Interprotein electron transfer requires the enzyme to adopt a more extended state in which the FMN domain is rotated away from the FNR module and is free to form a P450-complex.

Based upon ~ 50% amino acid sequence similarity (35% identity), it is anticipated that CPR from plants participate in a similar mechanism of flavin-mediated electron transfer and conformational motion (Fig. 1). That said, subtle differences in catalysis likely exist because aaCPR was shown to support higher turnover of CYP71AV1 compared to CPR from Saccharomyces cerevisea [4]. Moreover, CPR from Arabidopsis thaliana (ATR1), which has 76% sequence similarity with aaCPR, supported a higher efficiency and activity of rabbit P450 2B4 turnover compared to human CPR [27]. These initial findings suggest that plant CPR may be a more efficient reducing partner to native and non-native P450s. In the present study, we performed a kinetic and thermodynamic analysis of NADPH-dependent flavin reduction in CPR from A. annua in an effort to allow insight into these catalytic differences and to provide a basis for further optimization of the enzyme for microbial synthesis of bioactive natural products.

Figure 1.

Sequence alignment of human (hCPR), A. annua CPR (aaCPR) and A. thaliana (ATR1) CPR. Conserved residues that form noncovalent interactions with the 2′,5′-ADP portion of NADPH are indicated by an asterisk (R322, T563, R595, S624, R625, K630, Y632, D666; aaCPR numbering). The FMN, FAD and NADPH binding determinants, defined by Wang et al. [18], are indicated with boxes and identical residues are highlighted in grey.


Protein purification, spectral properties and flavin content

Recombinant aaCPR was expressed in E. coli as an N-terminally hexahistidine tagged protein and purified to homogeneity in two chromatographic steps (Fig. 2, inset). Throughout the purification, aaCPR was yellow, indicating that the flavins were fully oxidized. This is in contrast to human CPR, which is initially green as a result of the presence of a partially air-stable FMN semiquinone. The apparent molecular mass of purified aaCPR (75 kDa) is similar to the value calculated from the amino acid sequence (73 275 Da). HPLC analysis revealed 0.95 mol of FAD and 0.72 mol of FMN per mol of polypeptide. The absorbance spectrum of aaCPR shown in Fig. 2 is typical of diflavin reductases with peaks at 380 and 454 nm and a shoulder at 474 nm. The addition of equimolar NADPH produced a broad absorbance peak from 500 to 650 nm associated with the formation of the flavin blue semiquinone [6].

Figure 2.

Recombinant CPR from A. annua. Showing the UV-visible absorbance spectra of purified 10 μm aaCPR in 50 mm Tris-HCl (pH 7.5) in the oxidized state (solid line), the partially reduced state after the addition of equilmolar amount of NADPH (dashed line) and fully reduced after the addition of dithionite (dotted line). Inset: SDS/PAGE of two-step purification of aaCPR from E. coli Rosetta(DE3) pLysS. Lane 1, molecular weight markers (250, 150, 100, 70, 50, 40, 30 and 20 kDa); lane 2, E. coli crude extract; lane 3, eluate from the nickel-nitriloactetic acid column, lane 4, eluate from Q-sepharose column.

Steady-state assays

Electron transfer to cytochrome c3+ is mediated through the FMN cofactor of diflavin reductases; therefore, the turnover number for the reaction is dependent on hydride transfer from NADPH to FAD and interflavin electron transfer from FAD to FMN. The steady-state kinetic parameters for aaCPR reduction of cytochrome c3+ are listed in Table 1. Although the experimentally determined kcat (66 s−1) marginally underestimates the true turnover number as a result of the sub-stoichiometric concentrations of FMN in purified aaCPR, the value falls within the range of values determined for CPR from yeast, A. thaliana (ATR1) and insects in phosphate-free buffers [27, 28]. The Michaelis constants for NADPH (1.2 μm) and cytochrome c3+ (0.87 μm) are also comparable to values obtained for yeast, Arabidopsis thaliana and insects (Fig. S1) [27, 28]. However, direct comparison of the Km and kcat/Km for NADPH with mammalian forms of CPR is hampered by the fact that previous steady-state assays for these homologues were performed in high ionic strength phosphate buffers [20, 27, 29-32]. These buffering conditions inflate the apparent Km for the coenzyme because the phosphate anion acts as a competitive inhibitor with a Ki of 33 μm [21, 33]. We repeated the steady-state assay in 300 mm potassium phosphate at pH 7.5, which are similar to the conditions used for rat CPR, and obtained significantly elevated Km (31.3 μm) and reduced kcat/Km (1.82 × 10m−1·s−1) for NADPH, suggesting that binding of the reduced coenzyme is significantly weaker for aaCPR compared to its mammalian counterpart [29-32].

Table 1. Macroscopic kinetic parameters for aaCPR reduction of cytochrome c3+.
Kinetic parameterValue
  1. a

    The values in parentheses were determined in 300 mm potassium phosphate buffer (pH 7.5).

  2. b

    The value in parenthesis is for human CPR.

kcat (NADPH) (s−1)65.9 ± 1.4 (58.1 ± 2.2)a
Km (NADPH) (m × 10−6)1.22 ± 0.11 (31.3 ± 3.3)
kcat/Km (NADPH) (m−1·s−1 × 106)54.1 ± 5.0 (1.82 ± 0.2)
kcat (NADH) (s−1)16.5 ± 0.4
Km (cyt c3+) (m × 10−6)0.87 ± 0.13
Km (NADH) (m × 10−6)26 760 ± 1360
kcat/Km (NADH) (m−1·s−1 × 103)0.615 ± 0.033
[kcat/Km (NADPH)]/[kcat/Km(NADH)]85 100 ± 1700
Ki (NADP+) (m × 10−6)1.59 ± 0.18
Ki (2′,5′-ADP) (m × 10−6)0.26 ± 0.02
Ki (NAD+) (m × 10−6)26.9 ± 2.5 (17.0 ± 1.8)b
Table 2. Midpoint potentials of flavin cofactors in CPR from A. annua referenced to the standard hydrogen electrode.
Wavelength (nm)FMNox/sq (mV)FMNsq/hq (mV)FADox/sq (mV)FADsq/hq (mV)
  1. a

    To obtain reasonable estimate of the FADsq/hq couple for Fig. 5C, the value of c was fixed at 0.025 absorbance units. For further details, see 'Experimental procedures'.

Σ450–460−94 ± 3−205 ± 11−296 ± 6−337 ± 5
427−112 ± 2 −280 ± 2 
500 −219 ± 9 −336 ± 49a

In light of the sub-stoichiometric amount of the FMN cofactor in purified aaCPR (0.72 mol of FMN per mol of polypeptide), the cytochrome c3+ reductase activity was measured after a 10-min incubation of the purified enzyme with the free FMN cofactor. Titration of the enzyme with 0–10 μm FMN did not lead to an increase in catalytic activity (Fig. S2), suggesting that a fraction of the enzyme is in the inactive state, unable to incorporate free cofactor.

Dead-end and product inhibition studies reveal that NADP+ (Ki = 1.6 μm) binds with slightly weaker binding affinity compared to 2′,5′-ADP (Ki = 0.26 μm), a phenomenon that has also been observed in spinach ferredoxin-NADP+-reductase and methionine synthase reductase [34, 35]. The six-fold increase in the Ki for the oxidized coenzyme is attributed to the bipartite binding mode of the coenzyme. The 2′,5′-ADP portion of the cofactor is initially anchored to the active site through a number of conserved polar residues [34]. Subsequent binding of the ribityl-nicotinamide portion of the coenzyme is partially disfavored by a conserved FAD aromatic stacking residue (W704; all residue numbers correspond to those of aaCPR). This residue must be displaced before hydride transfer from the C4 of the nicotinamide to the N5 of the FAD isoalloxazine ring. It is considered that steric repulsion between the nicotinamide mononucleotide and W704 contributes to the weaker binding affinity of the coenzyme [29]. Interestingly, human CPR shows tight and near equivalent binding for 2′,5′-ADP (Kd = 50 nm) and NADP+ (Kd = 53 nm) [21]. The same is also true for CPR from Anopheles gambiae, except that this enzyme elicits an eight-fold higher Kd (~ 0.35–0.4 μm) for the oxidized coenzyme and its analogue compared to human CPR [36].

To identify structural alterations in the coenzyme binding cleft that likely account for the differences in the NADP+ and 2′,5′-ADP binding affinities, a homology model of aaCPR using swiss model was generated using human CPR (Protein Data Bank code: 3QE2) as a reference (Fig. 1B) [37]. Inspection of the active site reveals that polar residues that coordinate to the 2′,5′-ADP moiety in human and rat CPR are also present in the aaCPR model (Fig. 3). These conserved residues (R322, T563, R595, R625, K630, Y632, S624 and D666) are also noted in a sequence alignment between aaCPR and human CPR. Based on the model and the sequence alignment, weaker NADP+ binding is not attributed to the absence of a conserved residue in the aaCPR active site.

Figure 3.

A model of the coenzyme binding cleft of aaCPR generated by swiss model protein modelling server [37]. Human CPR (PBD code: 3QE2) was used as the reference structure. The coenzyme binding determinates along with the conserved FAD-stacking aromatic tryptophan are shown as grey sticks and the FAD and the 2′,5′-ADP portion of NADP+ are shown in cyan.

Diflavin oxidoreductases show a high specificity for NADPH versus NADH as the reducing coenzyme. A cluster of residues that coordinate to the 2′-phosphate of the cofactor (S624 R625, K630 and Y632) and the conserved FAD aromatic stacking residue act in concert to control the specificity of the coenzyme [29, 34, 38]. Despite the universal presence of these residues amongst FNR-like proteins, the degree of coenzyme preference [ratio of kcat/Km (NADPH) to kcat/Km (NADH)] does vary. For example, spinach FNR, rat CPR, methionine synthase reductase, nitric oxide synthase and P450BM3 prefer NADPH over NADH by 67 500 23 250, 19 400, 4987 and 8571-fold, respectively [35, 39-42]. This ratio is important for assessing the potential of using aaCPR to perform regio- and stereospecific P450-mediated reactions outside the cell (e.g. on immobilized platforms), where it is more cost-effective to use the less expensive NADH over NADPH. Given the role of aaCPR as a redox partner for a P450 involved in artemisinin biosynthesis, and the wider potential of using plant CPRs to reduce P450s implicated in natural product biosynthesis, it was of interest to determine the degree that this enzyme preferentially binds and oxidizes NADPH versus NADH. As shown in Table 1, aaCPR elicited a 85 100-fold preference for NADPH compared to NADH, the highest coenzyme preference ratio recorded to date amongst FNR-like flavoproteins.

The NADPH oxidase activity of the enzyme, characterized by single electron transfer from the reduced flavins to O2, derails the flow of electrons from NADPH through the flavins to the P450 haeme. Given that uncoupled NADPH oxidation products (H2O2 and superoxide) potentially lead to cellular oxidative stress and low production yields of the bioactive natural product, it was of interest to compare the rates of NADPH oxidase activity in aaCPR with that of human CPR. As shown in Fig. 4, the steady-state rate of NADPH oxidation is two- to three-fold higher in aaCPR compared to human CPR and, in both enzyme systems, an increase in ionic strength leads to increased rates of NADPH turnover.

Figure 4.

Rates of uncoupled NADPH oxidation by aaCPR (squares) and human CPR (circles) at increasing ionic strength. Conditions for the reaction are described in the Experimental procedures.

Determination of the flavin midpoint potentials

Spectroelectrochemical redox titrations of aaCPR were performed to determine the midpoint potentials of the flavin cofactors. The titrations were performed under anaerobic conditions, starting with fully oxidized enzyme. After the addition of a small aliquot of dithionite, the voltage was allowed to equilibrate before the absorbance spectrum was recorded. Figure 5A shows select absorbance spectra acquired during the redox titration of aaCPR. As the first electron equivalent is added to the protein, the absorbance maxima at 454 nm decreases, whereas a broad absorbance band centred at 600 nm appears. During this initial redox phase, the higher potential FMN cofactor is converting from the oxidized to semiquinone state. An isosbestic point at 500 nm denotes this transition. This isosbestic point disappears upon the addition of the second and third reducing equivalents. During this phase of the titration, the FMN semiquinone is converting to the hydroquinone form and the FAD is transitioning from the oxidized to semiquinone state, resulting in multiple redox pairs and the absence of a clear isosbestic point. The absorbance maximum at 454 nm continues to diminish during this phase, whereas the semiquinone absorbance band at 600 nm enlarges and then remains relatively fixed. The final reduction phase, indicated by the presence of an isosbestic point at 427 nm, involves conversion of the FAD semiquinone to the hydroquinone form, and a further reduction in the optical signals at 454 and 600 nm.

Figure 5.

Spectroelectrochemical titration of aaCPR. (A) Spectral properties of aaCPR during anaerobic redox titration. The spectra were recorded at several points during the redox titration. The black lines denote spectra recorded during the addition of the first electron (oxidized to semiquinone transition), with an isosbestic point at 500 nm. The blue lines indicate spectra recorded between ~ −110 and −325 mV and the red lines are spectra recorded at potentials more negative than −325 mV, with an isosbestic point at 427 nm. (B) A plot of summed absorbance values between 450 and 460 nm against the reduction potential. The data were fit by a four-electron equation (Eqn 2) and the midpoint potential values are listed in Table 2. (C, D) Plots of absorbance at 500 and 470 nm, respectively, against reduction potential. Both of these data sets were fit by Eqn 1 and the midpoint potential values, referenced to the standard hydrogen electrode, are listed in Table 2.

The summed absorbance values from 450 to 460 nm were plotted against the potentials values normalized to the standard hydrogen electrode (Fig. 5B). The data were fit by Eqn (2) to give midpoint potentials for the FMNox/sq (−94 mV), FMNsq/hq (−205 mV), FADox/sq (−296 mV) and FADsq/hq (−337 mV) redox couples. The midpoint potentials values were also determined by plotting the absorbance values at the two isosbestic points that appear in the redox titration (Fig. 5C, D). At 500 nm, there is near-zero absorbance change occurring in the flavin oxidized to semiquinone transition; thus, the data at this wavelength can be used to determine the midpoint potentials of the sq/hq couples. Similarly, the change in absorbance at 427 nm can be used to determine the midpoint potentials of the FMNox/sq and FADox/sq couples. A fit of the data in Fig. 5C,D gave midpoint potential values of FMNox/sq (−112 mV), FMNsq/hq (−219 mV), FADox/sq (−280 mV) and FADsq/hq (−336 mV), which are in close agreement with the values determined from changes in the flavin absorbance maxima. In general, the flavin midpoint potential values are comparable to that of human and rat CPR, with the exception that the FADsq/hq couple is ~ 20 mV more positive in aaCPR [43, 44].

Stopped-flow absorbance spectroscopy with NADPH

NADPH-dependent reduction of aaCPR was followed under anaerobic conditions by both single and multi-wavelength spectroscopy. Initially, stopped-flow experiments were performed at 25 °C, although inspection of the time-resolved photodiode array spectra revealed that the enzyme was partially reduced by NADPH within the dead-time of the stopped-flow. In an effort to monitor the rate of the flavin reduction, the temperature for the experiments was reduced to 6 °C. Figure 6 shows the photodiode array spectra of the oxidized enzyme overlaid with spectra collected over 1 s after mixing the enzyme with a ten-fold excess of NADPH. The superimposed spectra reveal that NADPH-dependent reduction is relatively fast (i.e. > 500 s−1) because the flavin absorbance maxima is reduced by ~ 40% < 3 ms after the mixing event. By comparison, the initial flavin reduction event in mammalian CPR has an associated rate constant of 20 to 30 s−1 at 25 °C [7, 45]. A second unusual feature of Fig. 6 is the rapid formation of a broad absorbance band extending from 532 to > 700 nm with a minor peak at 600 nm. For full-length human CPR and rat CPR, bleaching of the flavin absorbance maxima is accompanied by the formation of a more ‘up and down’ spectral signature indicative of the blue semiquinone (Fig. 3), with a prominent peak at 600 nm and minimal absorbance > 675 nm [45, 46]. The profile of the absorbance band shown in Fig. 6 suggests that, in addition to disemiquinone formation, there is also rapid accumulation of charge-transfer species arising from electronic interactions between the FAD isoalloxazine and NADP(H) nicotinamide rings [46]. The minor peak at 600 nm, which appears and then disappears in < 0.2 s, suggests transient formation of the disemiquinone intermediate in a fraction of enzyme population. By contrast, the minimal absorbance change > 675 nm over 1 s indicates that the charge-transfer species fully forms within the dead-time of the stopped-flow and then persists during the reductive half-reaction (Fig. S3). To ensure that the broad absorbance band in the long wavelength region is not a consequence of baseline drift, we repeated the experiment with human CPR and, in this case, we did not observe a significant absorbance increase > 650 nm (data not shown).

Figure 6.

Reduction of aaCPR (25.2 μm) with NADPH (0.25 mm) monitored by stopped-flow photodiode array spectroscopy. Conditions:  50 mm Tris-HCl (pH 7.5) at 6 °C. The spectra generated upon mixing aaCPR with buffer were overlaid with spectral changes generated upon mixing NADPH with the enzyme.

In an attempt to extract rate constants for the different kinetic phases of flavin reduction NADPH-dependent reduction of aaCPR was also followed at 454 and 600 nm in single-wavelength mode where the dead-time for data acquisition decreases to 1 ms. The initial starting absorbance at 454 nm was determined by mixing oxidized aaCPR (28.8 μm) against an equal volume of buffer in the stopped-flow. After acquisition of the initial absorbance value at 454 nm, the same enzyme solution was rapidly mixed with NADPH (0.29 mm) or (R)-[4-2H]-NADPH (0.29 mm), and the change in absorbance was recorded over 0.5 s. Figure 7A reveals that aaCPR, is still partially reduced by the coenzyme within the dead-time of the stopped-flow. As anticipated, the loss of absorbance in < 1 ms is less pronounced with the deuterated substrate. For both NADPH and (R)-[4-2H]-NADPH, reduction of aaCPR was biphasic over 0.2 s, and a fit of the averaged traces to a double exponential equation from 0.001 to 0.25 s produced a kobs1 of 499 s−1 and kobs2 of 18 s−1 for NADPH and a kobs1 of 297 s−1 and 17 s−1 for (R)-[4-2H]-NADPH. Based on previous kinetic analysis of CPR homologues, it is likely that the initial fast kinetic phase (kobs1) encompasses the first hydride transfer event, whereas kobs2 (18 s−1) reflects further reduction of the enzyme by a second equivalent of NADPH. The value of kobs1 is an underestimate of the fast kinetic phase given that the expected starting absorbance at 454 nm (0.32) of the fully oxidized enzyme is slightly higher than the value of 0.28 recorded at the dead-time of the stopped-flow (1 ms). Accordingly, the kinetic isotope effect (KIE) on the first phase of flavin reduction (Dkobs1) is > 1.7, whereas the KIE on kobs2 is ~ 1. At 25 °C, the majority of the amplitude change for the first fast kinetic phase is lost, although a fit of the data by a double exponential equation gave a kobs2 of 65 s−1, which is ~ 3.6-fold faster than at 6 °C (Fig. S4). Figure 7A (inset) reveals that aaCPR continues to reduce over a 500-s time frame, although the majority of the flavin reduction occurs in < 0.2 s. The final slow kinetic phase, also observed for rat CPR, is multiphasic and likely represents the nonphysiological redox equilibration of the enzyme/coenzyme solution [45].

Figure 7.

Stopped-flow absorbance traces tracking the reduction of aaCPR by NADPH. (A) A stopped-flow trace at 454 nm after mixing 28.8 μm aaCPR with 290 μm NADPH or 290 μm (R)-[4-2H]-NADPH at 6 °C under the same conditions as shown in Fig. 6. A fit of the data by a double exponential generated observed rate constants of 499 ± 4 s−1 and 18.1 ± 0.04 s−1 for NADPH and 297 ± 5 s−1 and 16.7 ± 0.3 s−1 for (R)-[4-2H]-NADPH. The flat line at the top was generated by mixing 28.8 μm aaCPR with an equal volume of buffer. (A, inset) Extended time course of an absorbance trace at 454 nm (500 s) after mixing 25.0 μm aaCPR and 0.25 mm NADPH. (B) Absorbance trace obtained at 600 nm after rapid mixing of 0.25 mm NADPH and aaCPR (25.2 μm). A fit of these data by a double exponential generated observed rate constants of 54.3 ± 1.0 s−1 (‘up’ phase) and 14.6 ± 0.4 s−1 (‘down’ phase). The absorbance trace at 600 nm extended over 500 s (B, inset). The grey line is the fitted trace.

The ‘up-down’ absorbance trace at 600 nm is likely only tracking formation and decay of the disemiquinone intermediate and not the charge-transfer species. This assumption is based on the lack of an absorbance change at 675 nm over 1 s that would otherwise signal a change in the concentration of the latter species. A fit of the data in Fig. 7B gave observed rate constants of 54 s−1 (‘up’ phase) and 15 s−1 (‘down’ phase). As with the single wavelength 454 nm traces, the value of kobs1 is likely an underestimate because a larger absorbance change is anticipated for this initial kinetic event judging from the photodiode array spectra in Fig. 6. The second rate constant is comparable to that for kobs2 from single wavelength experiments at 454 nm, indicating that the decay of the disemiquinone intermediate is kinetically coupled to the second hydride transfer event. Stopped-flow traces over a 500-s time domain (Fig. 7B inset) show a gradual increase in absorbance at 600 nm reflecting reformation of the flavin semiquinone during equilibration of the enzyme/coenzyme solution.

Stopped-flow absorbance spectroscopy with NADH

The faster observed rate of flavin reduction may be linked to the 30-fold difference in the Kd for NADP+ between the plant and human enzyme. Weaker binding suggests that the rate constants associated with the release of the coenzyme (k−1 and/or k5 in Fig. 8) are significantly faster in aaCPR. Elevated values of k−1 and/or k5 would result in faster kobs1 and kobs2, given that the observed rate constants are a summation of the forward (k1, k2, k3) and reverse rate constants (k−1, k−2, k−3) of the mechanism. To test this hypothesis, the pre-steady-state rate of NADH-dependent reduction of human CPR and aaCPR was measured by stopped-flow spectroscopy. Given that human CPR has a Ki value for NAD+ that is 1.6-fold lower than that of aaCPR, we expect that the rate of NADH-dependent flavin reduction to be modestly faster for the plant enzyme (Table 1). Indeed, Fig. 9 shows that NADH reduction of aaCPR is approximately two-fold faster compared to hCPR. The stopped-flow traces at 454 nm for both enzymes were biphasic over 10 s with a kobs1 and kobs2 of 3.3 s−1 and 0.62 ± 0.01 s−1 for aaCPR and 1.5 ± 0.1 s−1 and 0.22 ± 0.01 s−1 for hCPR (Fig. 9). The photodiode array spectra acquired under the same experimental conditions show that the high concentrations of NADH do not obscure the flavin absorbance spectra (Fig. S5).

Figure 8.

NADPH-dependent reduction of cytochrome P450 reductase. NADPH binds to the diflavin enzyme with the flavin cofactors in the fully oxidized state (FAD and FMN). The NADPH nicotinamide ring displaces a conserved tryptophan shielding the FAD isoalloxazine ring. Transfer of the hydride ion from NADPH to FAD, producing the FADH2 hydroquinone, is coupled to internal electron transfer from FADH2 to FMN, generating the flavin semiquinone, which has an absorbance signal at 600 nm. The enzyme is further reduced by a second equivalent of NADPH.

Figure 9.

Reduction of human CPR and aaCPR by NADH. Stopped-flow absorbance trace (black line) at 454 nm generated upon mixing of 22 mm NADH with 30 μm of aaCPR and human CPR. A fit of the data by a double exponential equation (grey line) generated an observed rate constant of 3.28 ± 0.01 s−1 and 0.62 ± 0.01 s−1 for aaCPR and 1.47 ± 0.01 s−1 and 0.22 ± 0.01 s−1 for human CPR. The reactions were performed in 50 mm Tris HCl (pH 7.5) at 25 °C under anaerobic conditions.


Metabolic engineering of microbes is increasingly viewed as a cost-effective and environmentally friendly approach to the synthesis of plant-based pharmaceuticals [47, 48]. This heterologous production platform has been used for the biosynthesis of functionalized terpenoids, isoflavones and benzylisoquinoline alkaloids, confirming that it can be a viable alternative to chemical synthesis or natural host extraction [3, 4]. The cytochrome P450 family (one of the largest within the plant genome) is involved in the formation of a large proportion of plant-based natural products, including those listed above [49]. The abundance of P450s and the scope of the chemical reactions that they perform have helped realize the broad repertoire of chemically diverse metabolites in plants [49-51]. Because plant P450s are increasingly used as tools for the production of pharmaceuticals and other complex chemicals, it is important that the mechanism of flavin-mediated electron transfer to plant P450s is fully understood. In the present study, we performed the first detailed kinetic and potentiometric analysis of CPR from a plant species and found marked differences compared to those of the mammalian and yeast homologues.

The most striking difference between the CPR orthologues is the accelerated rate of NADPH-dependent reduction and the detection of a charge-transfer species in aaCPR. The reductive half reaction of full-length CPR occurs in three resolvable kinetic phases over 500 s [45]. The ‘fast’ phase, with a rate constant of 20 s−1 (human CPR), 30–60 s−1 (rat CPR) or 48 s−1 (ATR1), involves concomitant bleaching of the flavin absorption maxima at 454 nm and an absorbance increase at 600 nm [23, 27, 45]. These simultaneous optical changes reveal that hydride transfer from NADPH to FAD is kinetically coupled to interflavin electron transfer. At the end of this first kinetic phase, 30% of the two-electron reduced form of the enzyme is in the disemiquinone state (species IV in Fig. 8), whereas the remaining 70% exists with both electrons on the FMN cofactor (species V in Fig. 8) [45]. The second kinetic phase involves bleaching of absorbance at 454 and 600 nm because transfer of a second hydride ion from NADPH to a population of the two-electron reduced CPR results in further flavin reduction and quenching of the flavin semiquinone intermediate. The observed rate constant associated with oxidation of the second NADPH decreases to ~ 3 s−1 in mammalian CPR as a result of the reduced thermodynamic pull of the high potential FMN [44, 45]. A final slow kinetic event involves equilibration of the coenzyme/enzyme solution reflected in minor absorbance changes at 600 and 454 nm. At the end of the time course, there is a mixed population of partially and fully reduced forms of CPR.

With the exception of the long-wavelength charge-transfer band (discussed below), the spectral changes associated with flavin reduction are conserved between aaCPR and its mammalian counterparts. Similar to hCPR, flavin reduction is multiphasic; hydride and interflavin electron transfer appear to be tightly coupled steps; and saturating levels of NAPDH lead to incomplete flavin reduction. However, the observed rate constants associated with the oxidation of the first and second equivalent of NADPH are significantly faster in aaCPR. Indeed, at 25 °C, the majority of the amplitude changes associated with the first hydride transfer step are complete within the dead-time of the stopped-flow. The rate constant for the second hydride step is also markedly faster in aaCPR (65 s−1 at 25 °C). Reducing the temperature to 6 °C resulted in an approximately four-fold decrease in kobs2 for aaCPR and enabled us to capture more of the amplitude change associated with the initial hydride transfer step. Although ~ 500 s−1 at 6 °C is an underestimate for kobs1, if we assume that this value increases four-fold at 25 °C, then donation of the first hydride equivalent from NADPH can be estimated to occur at ~ 2000 s−1, which is 30–100-fold faster than the mammalian homologues. The rate of hydride transfer in aaCPR is also significantly faster than that of yeast CPR (185 s−1 at room temperature) and the biotechnology workhorse, P450 BM3, whose reductase domain elicited an observed rate of ~ 200 s−1 at 5 °C [52]. The redox properties of the enzyme cannot account for the increase rate of flavin reduction because the four flavin redox couples in aaCPR are similar to that of the mammalian homologues, with the exception of the ~ 20 mV more electropositive FADsq/hq couple [53]. However, accelerated flavin reduction may explain why aaCPR elicits a modest increase in the rate of uncoupled NADPH oxidation compared to human CPR.

A second notable difference is the detection of a charge-transfer band from 550 to 700 nm in multi-wavelength stopped-flow studies with aaCPR. Interestingly, this charge-transfer species is also observed in the isolated FNR-component of human CPR but not in the full-length form of the enzyme [46]. It is unclear why this species is not detected in the latter enzyme. A larger optical change associated with disemiquinone formation in human CPR may mask the charge-transfer signal. Alternatively, there may be less π-π stacking between the FAD and the coenzyme in the full-length human enzyme. This interaction likely occurs upon coenzyme-induced displacement of a conserved tryptophan side chain (W704 in aaCPR and W676 in human CPR), which shields the re-face of the FAD isoalloxazine ring. Binding of the coenzyme to the active site presumably establishes an equilibrium between two conformational states; one in which the conserved tryptophan is stacked against the FAD and another in which the indole ring is displaced and the nicotinamide is adjacent to the FAD. For aaCPR and perhaps the isolated FNR-component of hCPR, the equilibrium is potentially shifted towards the later conformational state, resulting in optical detection of the associated charge-transfer species. Perhaps this shift towards the displaced FAD-shielding Trp accounts for the higher pre-steady-state rates of hydride transfer.

Alternatively, accelerated flavin reduction may be linked to the relatively weak binding affinity of aaCPR for the oxidized coenzyme. Compared with human CPR, NADP+ binds with 30-fold less affinity [21]. This difference is based on the Ki for NADP+ (determined in phosphate-free buffer) for aaCPR and the Kd for human CPR, as measured by isothermal titration calorimetry in phosphate-free buffer [53]. Faster dissociation of NADP+ after hydride transfer enables a second equivalent of NADPH to bind, further driving the enzyme to the four-electron reduced state. This hypothesis is supported by stopped-flow studies comparing the NADH-dependent reduction of aaCPR and hCPR. Human CPR, which has a modestly higher affinity for NAD+, elicited a two-fold decrease in the rate of NADH-dependent flavin reduction compared to its plant equivalent. This experiment suggests that coenzyme binding affinity controls the rate of flavin reduction.

Indeed, Phillips and Langdon postulated that release of NADP+ was the ionic-strength sensitive rate-determining step for the reductive half-reaction [54]. Early kinetic studies of rat CPR showed masking of intrinsic KIE for hydride transfer in low ionic strength buffers because these experimental conditions promoted tight binding of NADP(H) [55]. However, conditions that promoted faster release of the coenzyme (i.e. a high ionic strength buffer) caused an increase in the KIE because hydride transfer became more rate-determining. In phosphate-free, low ionic strength buffer (i.e. 50 mm Tris-HCl), we measured a primary kinetic isotope effect of > 1.7 for aaCPR, whereas, for human CPR, using the same experimental conditions, we obtained a KIE of 1.1 (data not shown). The elevated primary KIE on kobs1 for aaCPR indicates that hydride transfer is more rate-determining for the plant enzyme, presumably because of the accelerated release of the oxidized coenzyme. By contrast, slow dissociation of NADP+ from human CPR masks the intrinsic KIE on kobs1 and impedes the overall forward rate of flavin reduction.

At this stage, it is unclear why aaCPR elicits weaker coenzyme binding. A sequence alignment and a homology-based model of aaCPR reveals that the network of conserved residues that coordinate to the 2′,5′-ADP portion of the coenzyme are conserved between the two proteins. This suggests that variations amongst the nonconserved residues that line the active site also contribute to coenzyme binding affinity. Resolving these structural variations will require a high-resolution crystal structure of aaCPR.

In summary, although aaCPR does show a strong coenzyme preference for NADPH and a modest increase in the rate of uncoupled NADPH oxidation (features that may compromise its biotechnology potential), it does elicit significantly faster rates of NADPH-dependent flavin reduction, which is likely linked to its relatively weak binding for the coenzyme. Faster product dissociation after the reductive half reaction may account for the enzyme's ability to support faster P450 turnover in vivo.

Experimental procedures


The reagents NADPH, NADP+, 2′,5′-ADP, cytochrome c3+, benzyl viologen, methyl viologen, 2-hydroxy-1,4-naphthoquinone, phenazine methosulfate, ethanol-d6 and alcohol dehydrogenase from Thermoanaerobium brockii were purchased from Sigma-Aldrich (Oakville, ON, CA, USA). High Performance Q-sepharose and nickel-nitriloactetic acid were purchased from GE Biosciences (Baie d'Urfe, QC, CA, USA). Rosetta(DE3)pLysS competent cells were obtained from EMD Biosciences (Billerica, MA, USA). All other chemicals and reagents were purchased from Fisher Scientific (Ottawa, ON, CA, USA). (R)-[4-2H]-NADPH (A-side NADPD) was synthesized and isolated as described previously [56].

Cloning and expression of the aaCPR construct

The plasmid encoding aaCPR (pCWori-A13AMO-aaCPRct; plasmid number 20117) was obtained from Addgene (Cambridge, MA, USA). The primers, 5′-GGA ATT CCA TAT GCG CCG TAG CAG CTC TGC GGC CAA G-3′ (forward primer) and 5′-CGG GAT CCT TAC CAG ACA TCA CGC AGG TAG CGG CC-3′ (reverse primer) were designed to amplify residues 67–704 of the aaCPR sequence (removing the N-terminal transmembrane sequence) and to subsequently clone the PCR fragment into the NdeI and BamHI sites of pET15b. DNA sequencing by the NAPS DNA sequencing laboratory at the University of British Columbia, (Vancouver, Canada) confirmed no PCR-induced errors had been introduced into aaCPR that would affect its amino acid sequence. The cloning strategy incorporated a hexahistidine sequence on to the N-terminus of the protein. The resulting plasmid, named pETaaCPR, was transformed into Rosetta2(DE3)pLysS strain of E. coli. A single transformed colony was used to inoculate 100 mL of LB medium containing 100 μg·mL−1 ampicillin and 35 μg·mL−1 of chloramphenicol. After growth for 16 h at 37 °C, 10 mL of the starter culture was used to inoculate Terrific broth (0.5 L) containing ampicillin (100 μg·mL−1) and chloramphenicol (35 μg·mL−1). The culture was grown at 25 °C with shaking (220 r.p.m.) until A600 of 1.0 was reached, at which time isopropyl β-d-1-thiogalactopyranoside was added to a final concentration of 0.1 mm. The temperature of the incubator was reduced to 20 °C, and the culture was allowed to grow for a further 16 h. Cells were harvested by centrifugation (4000 g for 15 min at 4 °C) and the cell pellet was stored at −80 °C until purification.

Flavin determination and spectral analysis

The flavin cofactor content in purified aaCPR was determined by HPLC using a fluorescence detector [57]. Flavins were released by boiling 18 μm aaCPR for 10 min and then cooling rapidly on ice. Denatured protein was removed by centrifugation at 20 000 g for 5 min. FAD and FMN were separated by injecting 20 μL of the supernatant onto a C18 reverse phase column using a gradient ranging from 80% 10 mm potassium phosphate (pH 6.0), 20% methanol to 50% 10 mm potassium phosphate (pH 6.0) and 50% methanol at a flow rate of 1 mL·min−1. Flavins were detected by exciting at 454 nm and recording the fluorescence emission spectra. The column was calibrated using known concentrations of FMN and FAD purchased from Sigma. Protein concentrations were determined by the Lowry assay from Bio-Rad (Hercules, CA, USA) using bovine serum albumin as a standard.

Purification of aaCPR

All purification steps were performed at 4 °C or on ice. Cells from 3 L of bacterial culture (~ 46 g wet weight) were resuspended in 250 mL of 50 mm Tris-HCl (pH 7.5) with 1 mm phenylmethylsulfonyl fluoride and benzamidine. The cells were lysed by sonication (8-s pulses for 30 min with a 55-s interval, power setting 22%). The resulting cell suspension was centrifuged at 25 000 g for 50 min to remove particulates and then imidazole (20 mm) and NaCl (0.5 m) were added to the supernatent. The crude cellular extract was applied to a 5-mL Ni2+-nitrilotriacetic acid column equilibrated with 50 mm Tris-HCl (pH 7.5), 0.5 m NaCl and 20 mm imidazole. The column was washed with 30 mL of 50 mm Tris-HCl, 0.5 m NaCl (pH 7.5) and 20 mm imidazole, and the protein was eluted with 300 mm imidazole. Fractions containing aaCPR were pooled and dialyzed against 50 mm Tris-HCl (pH 7.5), 0.1 m NaCl, 1 mm EDTA and 5 mm β-mercaptoethanol for 16 h at 4 °C. The dialysate was loaded onto a 58-mL Q-sepharose HP column (2.6 × 11 cm) equilibrated in 50 mm Tris-HCl (pH 7.5). The column was washed with 120 mL of 50 mm Tris-HCl (pH 7.5) and the protein was eluted with a 0.9-L linear gradient of 0–0.5 m NaCl at a flow rate of 2 mL·min−1. Fractions containing aaCPR, as judged by the flavin absorbance spectra and SDS/PAGE analysis, were pooled, concentrated by ultrafiltration and flash-frozen in liquid nitrogen and stored at −80 °C. A truncated form of human CPR in which the N-terminal transmembrane sequence was removed and substituted with a hexahistidine tag was expressed in Rosetta(DE3)pLysS and purified as described for aaCPR.

Steady-state turnover analysis

The initial rate of aaCPR reduction of cytochrome c3+ was determined by monitoring the change in absorbance at 550 nm over 1 min (Δε = 21.1 mm−1·cm−1) on a Lambda 25 UV-visible spectrometer (Perkin Elmer, Boston, MA, USA) at 25 °C. Reaction mixtures of 1 mL containing 50 mm Tris-HCl (pH 7.5), 10 μm cytochrome c3+, variable NADPH (0.25–100 μm), NADP+ (0, 1, 2.5 and 10 μm), 2′,5′-ADP (0, 1, 2.5 and 10 μm) or NAD+ (0, 0.25, 0.5 and 1 mm) concentrations were initiated with the addition of aaCPR or human CPR. Initial velocities (without the presence of inhibitor) were fit to the Michaelis–Menten equation. Product and dead-end inhibition data were fit to the equation for competitive inhibition by nonlinear least-squares analysis using origin, version 8.5 (OriginLab Co. Northampton, M.A., USA). Uncoupled NADPH oxidation was measured at 25 °C in 50 mm Hepes-HCl (pH 7.5) with the addition of 1 mm NADPH and 10 nm of human or aaCPR at increased concentrations of potassium chloride. The initial velocity was calculated from the change in A340 using an extinction coefficient of 6.22 mm−1·cm−1.

Stopped-flow flavin reduction analysis

Double turnover pre-steady-state kinetic measurements of aaCPR flavin reduction by NADPH were performed using a SF-61DX2 stopped-flow (TgK Scientific, Bradford-on-Avon, UK). The sample-handling unit of the stopped-flow was placed in a customized glove box (Belle Technology, Weymouth, UK) under a nitrogen atmosphere with O2 concentration < 5 p.p.m. Absorbance measurements in single- and multi-wavelength mode were performed at 6 °C in 50 mm Tris-HCl (pH 7.5). The buffer was made anaerobic by extensive bubbling with nitrogen gas followed by > 16 h equilibration in the glove box. Concentrated samples of aaCPR (30–40 μm) were introduced into the glove box, and protein was made anaerobic by applying 2 mL of the concentrated aaCPR sample to a 10-mL size exclusion column equilibrated with anaerobic buffer. Solid NADPH and NADH was introduced into the glove box and dissolved in anaerobic buffer. The concentration of the reduced coenzyme was determined by monitoring A340 (using an extinction coefficient of 6.22 M−1·cm−1). For multi-wavelength analysis of flavin reduction, a photodiode array detector was used to record absorbance spectra from 380 to 700 nm after rapid mixing of aaCPR against NADPH or NADH. NADPH-dependent reduction of aaCPR was also monitored at 454 nm and 600 nm under pseudo-first order conditions. An average of five individual traces of the single-wavelength data were fit to a double exponential equation over a select time domain. Reported concentrations of enzyme and substrate are syringe concentrations (i.e. before mixing); the solutions are diluted two-fold after the mixing event.

Potentiometric titrations

The electrochemical redox titrations of aaCPR were preformed in a glove box maintained under a nitrogen atmosphere at 25 °C. The titration buffer, 50 mm Hepes-KOH (pH 7.0), was made anaerobic by extensive bubbling with nitrogen, followed by > 16 h of equilibration in the glove box. The redox titrations were performed in 50 mm Hepes (pH 7.0) to maintain a neutral pH and to prevent protein precipitation, which occurred during redox titrations with 100 mm potassium phosphate (pH 7.0). Concentrated aaCPR was introduced into the glove box and gel filtered over a 10-mL size exclusion column (Econo-Pac 10 DG column; Bio-Rad) equilibrated with anaerobic titration buffer. The protein was diluted to 10 μm in a total volume of 3.5 mL. Redox mediators were then added to the protein solution:  benzyl viologen (1 μm), methyl viologen (0.3 μm), 2-hydroxy-1,4-naphthoquinone (5 μm) and phenazine methosulfate (2 μm). The flavin absorption spectra were recorded in a specially designed cuvette holder placed on a stir plate to which the fibre optic probes of the stopped-flow were connected. A xenon lamp illuminated the sample and a diode array detector was used to record the absorbance from 280 to 700 nm using a photodiode array detector. A Mettler Toledo FiveEasy voltmeter (Mettler Toledo SARL, Paris, France) coupled to a platinum-Ag/AgCl electrode was used to measure the electrochemical potential. The potential and spectra were recorded after each addition of reductant, dithionite, in accordance with the protocol of Dutton [58]. The observed potentials were normalized to the standard hydrogen electrode with the addition of 197 mV to the potential values. During the titration the baseline drifted; therefore, all spectra were corrected to the same baseline absorbance at 703 nm.

The midpoint potentials of the FAD and FMN cofactors were determined by plotting the absorbance at various wavelengths against the potential. The isosbestic point at 500 nm and 427 nm appearing in the electrochemical spectral titration was plotted against the potential and fit to Eqn (1), which is derived from extension of the Beer–Lambert Law and the Nernst equation. The absorbance changes at 500 nm reflect the transition from the flavin semiquinone to the hydroquinone, whereas those at 427 nm record the transition from the fully oxidized flavin to the semiquinone. As such, the absorbance changes at these wavelengths reflect a two-electron reduction process, allowing Eqn (1) to be used to extract the associated midpoints potentials [44, 59]:

display math(1)

Equation (2), which describes a sum of a two, two-electron redox process was used to fit summed absorbance data from 450 to 460 nm:

display math(2)

In these equations, E is the observed potential; E1′ and E2′ are the oxidized/semiquinone and semiquinone/hydroquinone midpoint potentials of one flavin; and E3′ and E4′ are the corresponding midpoint potentials of the second flavin. A is the total absorbance; a, b, and c are component absorbance values contributed by one flavin in the oxidized, semiquinone and reduced states, respectively; and d, e and f are the corresponding absorbance values associated with the second flavin. To extract midpoint potentials from plots of summed absorbance of 450–460 nm verses potential, we set a = d and f, which assumes that the absorbance contribution of the oxidized and reduced forms of the FAD and FMN were equal. Estimates of E1′, E2′, E3′ and E4′ (based on previously published values for human CPR) were entered and initially fixed during the fitting routine [53]. The values of af were constrained based on the absorbance data. For example, a/d were allowed to vary between 3.5 and 4.5, whereas c/f varied between 0.5 and 2. After several iterations, reasonable estimates of the individual flavin absorbance values (parameters a–f) were obtained. Combinations of these absorbance values (i.e. a and d, b and e, and c and f) were fixed, and then the potential values were then allowed to vary during the fitting routine. This process was repeated several times. During the final fitting iteration E1′ – E4′, and the absorbance values a, c, d and f were allowed to vary, whereas the parameters b and e were fixed. This strategy prevented the data from becoming over parameterized. It should be noted that a fit of Eqn (1) to the data obtained at A500 gave an unreasonably low value for c (i.e. −0.06), implying a negative absorbance value for the flavin hydroquinone. We therefore fixed the value of c to 0.025, a more reasonable estimate based on the endpoint of the titration, and then repeated the fitting routine. The revised fit gave us a more positive value for the FADsq/hq, comparable to that obtained from fitting Eqn (2) to the summed the absorbance data at 450–460 nm.


We would like to thank Dr Susan Murch for her HPLC analysis of the flavin content in aaCPR. This work is supported by a grant from the Natural Sciences and Engineering Research Council of Canada.