Three pairs of surrogate redox partners comparison for Class I cytochrome P450 enzyme activity reconstitution

Most P450s require redox partners for the electron transfer during catalysis. However, little information is available on cognate redox partners for P450s, which greatly limits P450 function exploration and practical application. Thus, the stategy of building various hybrid P450 catalytic systems with surrogate redox partner has often adopted to engineer P450 biocatalysts. In this study, we compare three pairs of frequently-used surrogate redox partner SelFdx1499/SelFdR0978, Adx/AdR and Pdx/PdR and in terms of their electron transfer properties. The three selected bacterial Class I P450s include PikC, P450sca-2 and CYP-sb21, which are responsible for production of high-value-added products. Here we show that SelFdx1499/SelFdR0978 is the most promising redox partner compared to Adx/AdR and Pdx/PdR. The results provide insights into the domination for P450-redox partner interactions in modulating the catalytic activity of P450s. This study not only produces a more active biocatalyst but also suggests a general chose for a universal reductase which would facilitate engineering of P450 catalyst.

and this truncated mutant will be presented by Adx for convenience. The three chosen Class I P450 enzymes were PikC (CYP107L1) from S. venezuelae, P450sca-2 (CYP105A3) from S. carbophilus, and CYP-sb21 (CYP107Z14) from Sebekia benihana. Further structural analysis and molecular docking studies led to more understandings of the protein-protein interactions in the P450-Fdx and Fdx-FdR complexes, which are critical for shuttling of reducing equivalents and affect P450 activity. We expect that the key findings in this work will provide more mechanistic and practical guidance for construction of an efficient Class I P450 catalytic system.

Results
Spectral properties of the purified P450s and surrogate RPs. The three selected P450s and three pairs of surrogate RPs were successfully expressed and purified. All these proteins were purified to homogeneity as shown by SDS-PAGE analysis ( Supplementary  Fig. 1). The CO-bound reduced difference spectra of PikC, P450sca-2, and CYP-sb21 showed a characteristic peak at 450 nm, indicative of functional P450 enzymes ( Supplementary Fig. 2). The absorption spectra of the three flavoproteins (SelFdR0978, AdR, and PdR) gave the signature peaks at around 410 nm, 455 nm, and a shoulder at 395 nm. The three Fe 2 S 2 cluster-containing proteins (SelFdx1499, Adx, and Pdx) displayed the typical UV-Vis spectra with two local maximal peaks at approximately 380-420 nm and 460 nm ( Supplementary Fig. 3). The extinction coefficients of the flavin-bound SelFdR0978 and the iron-sulfur containing SelFdx1499 were determined to be ε 454 = 19,120 M −1 cm −1 and ε 460 = 11,270 M −1 cm −1 , respectively. The P450 activities supported by different surrogate RPs. The choice of surrogate RPs (when the innate RPs are unavailable) is crucial for activity reconstitution of P450s. To determine the optimal surrogate RPs for Class I P450 enzymes, a small reaction matrix with three surrogate RP systems and three P450s (against 3 substrates) was constituted. Regarding the P450 catalytic activities, PikC is able to hydroxylate the 12-membered ring macrolactone YC-17 to produce the mono-hydroxylated products methymycin and neomethymycin, and two dioxygenation products novamethymycin and ketomethymycin (Fig. 1a) 17 . P450sca-2 is an industrially important enzyme that can hydroxylate mevastatin to pravastatin as the desired product and the side product 6-epi-pravastatin ( Fig. 1b) 18,19 . CYP-sb21 is capable of catalyzing the sitespecific hydroxylation of the immunosuppressant cyclosporine A (CsA), giving rise to the valuable hair-growth-stimulating agent γhydroxy-N-methyl-L-Leu4-CsA (CsA-4-OH) 15 (Fig. 1c).
We also determined the NAD(P)H coupling efficiencies of the 27 P450 reactions (Supplementary Table 1) by following the previously established procedure 24 . The three cognate combinations (SelFdx1499/SelFdR0978, Adx/AdR, and Pdx/PdR) exhibited higher coupling efficiencies than the corresponding hybrid RP combinations. SelFdx1499/SelFdR0978 demonstrated the highest coupling efficiency in all three P450 reaction systems. The CYP-sb21 system only achieved the highest coupling efficiency of 12.6% when partnered by SelFdx1499/SelFdR0978, which is in good agreement with the low activity towards CsA 15 . The low NAD(P)H coupling efficiency is an important limiting factor for the activity of a reconstituted P450 system 15 . Thus, an NADPH regeneration system (GDH/glucose) was used in this study to increase the P450 activities.
The reductase activities and cofactor preference of FdRs. The reductase activities of FdRs were determined using 2,6dichloroindophenol (DCIP) as an electron acceptor. As a result, SelFdR0978 was more efficient in DCIP reduction than both AdR and PdR. Electron donor preference was investigated by changing NADPH to NADH in the reaction mixtures. All the three FdRs were able to use either NADPH or NADH as electron donor. SelFdR0978 showed a strong preference for NADPH over NADH (3-fold increase), while AdR only exhibited a slightly increased activity with NADPH than with NADH. As previously reported 25 , PdR preferred NADH to NADPH. When the preferred electron donor was used for the three FdRs, SelFdR0978 gave the highest DCIP reduction, which was approximately 2 and 9 times higher than PdR and AdR, respectively (Fig. 3a).
The electron transfer efficiency of different RP systems (NAD(P) H → redoxin reductase → redoxin → cytochrome c). Cytochrome c (cyt c) is routinely used as an electron acceptor to evaluate the electron transfer efficiency of different RPs 26,27 . The constructed nine (3 × 3) different RP combinations showed a broad range of NAD(P)H-dependent cyt c reductase activities (Table 1). Consistent with the observations made in previous studies 26,28,29 , the three FdRs were found to directly transfer electrons to cyt c more slowly in the absence of Fdx (Fig. 3b), thus having minor effects on cyt c reduction by Fdx. Among the tested three redoxins, as expected, the combinations with their cognate redoxin reductases gave a higher activity of cyt c reduction. SelFdx1499 was determined to be the most efficient redoxin with the prominent high k cat /K m values when coupled with SelFdR0978 (k cat /K m = 0.92 μM −1 s −1 ) or AdR (k cat /K m = 0.85 μM −1 s −1 ). Both combinations demonstrated about 4-fold higher k cat /K m values when compared with that of SelFdx1499/PdR (k cat /K m = 0.23 μM −1 s −1 ). Adx was also an efficient redoxin when paired with AdR (k cat / K m = 0.27 μM −1 s −1 ) and SelFdR0978 (k cat /K m = 0.16 μM −1 s −1 ). Adx and Pdx showed a minor difference when combined with SelFdR0978. The combinations of Adx/PdR and Pdx/AdR showed the lowest reducing efficiencies with k cat /K m values of 0.004 and 0.09 μM −1 s −1 , respectively (Table 1 and Supplementary Fig. 8). An overview of the K m values suggested that all three redoxins could form an efficient electron transfer chain with their corresponding cognate FdRs. SelFdx1499/AdR gave the highest k cat value (89.5 ± 25.5 s −1 ) among the tested RP pairs. All other RP pairs including the three cognate pairs showed >79% lower k cat values relative to that of SelFdx1499/AdR (Table 1). In general, the presence of SelFdx1499 or SelFdR0978 led to a relatively higher k cat value for unknown reasons.
Electrostatic surface analysis and interprotein docking of P450s and RPs. Alphafold2 was used to predict the unsolved crystal structures of P450sca-2, SelFdR0978, and SelFdx1499 30 . Then, the computer models for the P450-Fdx and FdR-Fdx ET complexes were generated by ZDOCK Server (ZDOCK 3.0.2) 31 . As expected, the three Fdxs unanimously bind onto the proximal side of P450s. To understand the recognition mechanism of the nine P450-Fdx pairs, the electrostatic surfaces were analyzed. PikC, P450sca-2, and CYP-sb21 have some conserved Lys, Arg and His residues that form a positively charged feature on their proximal surfaces (Supplementary Figs. 9 and 10). By contrast, Fdxs harbor a number of acidic amino acids surrounding the hydrophobic Fe 2 S 2 cluster (Fig. 4). The acidic and basic amino acids shown on the protein-protein interfaces were predicted to be involved in formation and stabilization of these P450-Fdx complexes (Fig. 5). Notably, because the three Fdxs have different electrostatic and interaction surfaces, they interact with the same P450 via different acidic amino acids, thus adopting different interacting orientations. For example, in the CYP-sb21-SelFdx1499 complex, E32 (Helix1 of Fdx) interacts with K49 (Helix B of CYP-sb21), while D60 and D63 interact with R106 and R110 (Helix C of CYP-sb21) through electrostatic contacts. In the CYP-sb21-Adx complex, E47 interacts with K107 (Helix C of CYP-sb21), while E73 and D76 (Helix3 of Adx) electrostatically interact with R347. SelFdx1499 and Adx demonstrate a more negatively charged surface than Pdx, which might have greater opportunities to form an ideal interface with various P450 enzymes, thereby achieving more efficient electronic transmission. For the three SelFdx1499-P450 complexes displaying high substrate conversions, the distances between the iron-sulfur cluster and heme-iron in the most likely docking modes range from 12.7 to 15 Å, which are shorter than the corresponding distances observed in the Adx-P450 and Pdx-P450 complexes. The results indicated that a shorter Fe 2 S 2 -to-heme-iron distance would likely increase the electron transfer efficiency. Furthermore, SelFdx1499 has a larger anionic convex surface than both Adx and Pdx (Figs. 4 and 5). Taken together, the intensity of negatively charged surface, the charge distribution of Fdxs, and the distance between iron-sulfur cluster and heme-iron appear to be the key factors to determine the electron transfer efficiency between redoxin and P450.
The electron transfer efficiency between redoxin reductase and redoxin could also influence the overall electron transfer rate and the catalytic efficiency of P450s. Some progresses have been made in general understanding of the interactions of AdR-Adx as well as PdR-Pdx based on their crystal structures 32,33 . However, the crystal structure of SelFdx1499-SelFdR0978 complex has not yet been successfully obtained. Thus, we built a model of SelFdR0978-SelFdx1499 using ZDOCK software, in which the interface is~522.8 Å 2 of the surface area in each protein. For AdR-Adx and PdR-Pdx, the interface areas are 580 Å 2 and 365 Å 2 , respectively 32,33 . In the X-ray structure of the covalent cross-linked AdR-Adx and PdR-Pdx complexes, the electrostatic forces were found to play an important role in formation and stabilization of productive ET complexes. Among the interface residues in PdR-Pdx, there are only one bidentate Arg 310 PdR-Asp 38 Pdx salt bridge and five hydrogen bonds 33 . The interface of AdR-Adx is composed of His 28 AdR-Asp 41 Adx salt bridge and several interprotein hydrogen bonds. In the ZDOCK model of SelFdx1499-SelFdR0978, Asp 70 and Asp 63 of SelFdx1499 interact with Arg 117 and Arg 364 of SelFdR0978 by forming a salt bridge, respectively. In combination with other interprotein hydrogen bonds, a stable conformation convenient for ET is formed. The shortest FAD-Fe 2 S 2 distance for SelFdR0978 with SelFdx1499, Adx, and Pdx were determined to be 6.5 Å, 9.0 Å, and 20.4 Å, respectively. For AdR (or PdR), the shortest FAD-Fe 2 S 2 distance were found to be 7.5 Å (11.9 Å), 10.3 Å (15.5 Å), and 15.1 Å (11.5 Å), when coupled with SelFdx1499, Adx, and Pdx, respectively (Fig. 6). The docking results of the three pairs of RPs agree well with the experimental data and suggest that a proper protein-protein interaction (PPI) conformation would directly determine the distance between FAD and iron-sulfur cluster, which is one of the key factors affecting ET efficiency.
To validate the ZDOCK models, the amino acid residues on the predicted Fdx interaction interfaces were replaced by alanine in a grouped manner. The conversion of YC-17 by PikC was selected for activity evaluation of different Fdx mutants including SelFdx1499-M1−M4, Adx-M1−M4, and Pdx-M1−M4. As expected, a majority of reactions showed a decreased substrate conversion rate when partnered by different Fdx mutants ( Table 2). It is intriguing that some Fdx mutants with their PPI interface charges removed gave an even higher YC-17 conversion rate than the corresponding wild-type Fdxs. For example, first, compared with the parental SelFdx1499, SelFdx1499-M2, and SelFdx1499-M3 led to an improved YC-17 conversion when partnered with AdR and PdR, respectively. Structure modeling showed that a negative interaction surface can still form ( Supplementary Fig. 11a-c). Protein-protein docking analysis of SelFdx1499-M2/AdR and SelFdx1499-M3/PdR revealed that the mutations induce the change of their corresponding PPI interfaces with AdR and PdR. These alternative PPIs resulted in closer distances between the iron-sulfur cluster and heme-iron compared to those for wild-type SelFdx1499 ( Supplementary  Fig. 11d, e). Thus, we reason that Fdx may harbor redundant negatively charged residues on the surface to recognize and interact with different P450s and FdRs. This redundancy could explain the tolerance of the interface residues to multiple chargeremoving mutations. Second, Adx-M4 with a changed Fdx-P450 PPI interface gave a higher PikC activity than the parental Adx (76.3% versus 63.8%). Third, the double mutations in Pdx-M4 even enabled the activity of the PikC/Pdx/SelFdR0978 system. Taken together, these results strongly suggest that the prediction of PPI interface residues based on P450-Fdx and Fdx-FdR docking should be reliable and practical, and these amino acids are important for the efficient catalysis by P450s.

Discussion
The three-component Class I P450 system consists of an iron-sulfur-containing redoxin, an FAD-containing redoxin reductase, and a P450 monooxygenase. The catalytic rate of such a system is often limited by the rate of ET from RPs to P450 13 . The two-electron carrier FdR is responsible for shuttling two electrons from NAD(P)H to Fdx. The reduced one-electron carrier Fdx sequentially delivers two electrons to P450 to support the activation of dioxygen for substrate oxidation. A rapid ET from RP proteins to P450 requires effective PPIs of both FdR-Fdx and P450-Fdx, and appropriate reductive potential differences 6 . The PPIs between FdR-Fdx-P450 during electron transport have not yet been fully understood primarily because of the lack of co-crystal structures for the two-or three-component complex. To date, only little structural information of RP couples is available, such as PdR-Pdx (PDB code 3LB8) and AdR-Adx (PDB code 1E6E) 32,33 . For these redox couples, the electrostatic/ hydrophobic interactions, salt bridge-forming residues, and peripheral hydrogen bonds play essential roles in formation and stabilization of productive ET complexes 34 . The AdR-Adx complex is remarkably similar to PdR-Pdx in terms of binding topology and geometry 32 . Although the sequence identities for Adx/Pdx and AdR/PdR are only 32.4% and 19.4% (Supplementary Fig. 10), respectively, Pdx and Adx share a highly conserved 86 SRLXCQ 91 motif. Interestingly, His54, which was identified to play a significant role for AdR recognition, is absolutely conserved in vertebrate-type and some bacterial-type ferredoxins (e.g., Pdx) 9 , but not found in SelFdx1499 (Supplementary Fig. 10). Although Adx and Pdx have some similar features when compared with SelFdx1499, the activities of three P450s were low when paired with AdR/Pdx or PdR/Adx. This suggests that except the key residues and common manner of RPs interactions, the interface residues and proper distance between FdR and Fdx are also important factors. Cognate RPs including SelFdx1499/ SelFdR0978, Adx/AdR, and Pdx/PdR were active in all the three tested P450 reactions, indicating that all the three redoxins can deliver electrons to the selected P450s. Among the 27 reactions with different Fdx/FdR combinations, 74.1% of combinations successfully supported the P450 activities. Besides the desired interaction between iron-sulfur cluster and P450, it is also worth noting that the recognition between RPs could affect the Fdx-to-P450 electron transfer efficiency.
Fdx-P450 recognition is primarily dependent on the electrostatic interaction and especially important in the catalytic cycle of P450. It was thought that proper Fdx binding would result in changes in the spectral and physical properties of P450 35 . In the well-studied P450cam catalytic system, the salt bridge between Pdx Asp 38 and P450cam Arg 112 plays a vital role in the binding and ET in the Pdx-P450cam complex 35 . The comparison reveals a number of Arg, His, and Lys residues that are conserved among these three P450 enzymes (Supplementary Figs. 9 and 10).
The effect of ionic strength on the interactions of P450/Fdx/FdR was investigated in a serial NaCl concentrations ranging from 100 to 1000 mM at a constant pH of 7.4. As expected, the conversion rates of YC-17 by PikC when coupled with all three cognate RP pairs gradually decreased along with the increasing NaCl concentrations ( Supplementary Fig. 12). These results support that electrostatic interactions play an important role in electron transfer from RPs to P450.
In this study, we clearly showed that SelFdx1499/SelFdR0978 is the most effective and efficient RPs to support the three P450 hydroxylases. An interesting question arose whether this finding is also true for other types of P450 enzymes. To address this issue, we further examined the supporting activities of surrogate RPs towards the Class I P450 epoxidase EpoK (CYP167A1). EpoK catalyzes the epoxidation of C12-C13 double bond in epothilone D to form epothilone B 36 (Supplementary Fig. 13a). Among the nine combinations of surrogate RPs, the combinations of SelFdx1499/ SelFdR0978 and Adx/AdR were able to reconstitute the epoxidation activity of EpoK, with the substrate conversion rates of 97% and 6%, respectively ( Supplementary Fig. 13). Of note, SelFdx1499/ SelFdR0978 exhibited a higher supporting activity for EpoK than the previously reported spinach RPs 37 . In conclusion, SelFdx1499/ SelFdR0978 seems to be an optimal choice for Class I P450 activity reconstitutions.
In a P450 catalytic cycle, RPs often limit the overall reaction rate. Due to the weak PPI and quick inactivation of Fdx, it is difficult to obtain the co-crystallized P450-RP and redox complexes. Therefore, the lack of structures for both redox complexes has hindered the understanding of the details during RP-P450 interactions. In this work, we investigated three typical prokaryotic RP pairs in order to find a convenient way to guide the selection of RPs for a specific Class I P450 reaction. Based on our results, we suggest that SelFdx1499/SelFdR0978 should be utilized more generally in the future.

Methods
Materials. NADPH, NADH, dithionite, and horse heart cyt c were purchased from Sigma-Aldrich (St. Louis, MO, USA). YC-17 was extracted from Streptomyces venezuelans ATC 15439 ΔpikC and mevastatin sodium was prepared by saponification with 0.1 M NaOH in 96% ethanol-water at 50°C 38,39 . The obtained compounds were verified by LC-MS. All antibiotics were obtained from SolarBio (Beijing, China). DNA polymerase restriction endonucleases were purchased from TaKaRa (Dalian, China). ClonExpress II One Step Cloning Kit were bought from Vazyme (Nanjing, China). The kits for plasmid extraction and DNA purification were purchased from Omega Bio-Tek (Jinan, China). Mutants of Fdxs were synthesized by BGI Genomics Co., Ltd. and the corresponding mutation sites were shown in Table 2. His-tagged protein purification was performed by using Ni-NTA resin (Sangon Biotech, Shanghai, China). Amicon Ultra centrifugal filters and PD-10 desalting columns were obtained from Millipore (Billerica, MA, USA) and GE Healthcare (Piscataway, NJ, USA), respectively. Oligonucleotides synthesis and DNA sequencing were conducted by Sangon Biotech (Shanghai, China).
General experimental procedures. High-resolution Q-TOF mass spectrometry analysis was carried out with a maXis ultrahigh-resolution TOF system (Bruker Daltonik, Germany). The reaction samples were analyzed by HPLC/HRMS with a Triart C18 column (YMC Co., Ltd., Japan). Mevastatin, pravastatin, YC-17, and their derivatives were monitored at 230 nm using a linear mobile phase gradient ranging from 20% (v/v) acetonitrile to 70% (v/v) acetonitrile in 0.1% (v/v) TFA aqueous solution over 30 min. CsA and CsA-4-OH were monitored at 210 nm and analyzed in a two-buffer gradient system consisting of 25% methanol (buffer A) and 100% acetonitrile (buffer B). The gradient elution profile was as follows: 0-4 min, 40% solvent B; 4-15 min, 40-61% solvent B; 15-32 min, 61-100% solvent B; 32-35 min, 40% solvent B. The flow rate was set at 1 mL/min, and the injection volume was 20 μL. Protein expression and purification. The plasmids harboring certain enzymeencoding genes used in this study are listed in Supplementary Table 2. The adx and adR genes were PCR-amplified with pKKHC-Adx 4-108 and pBAR-bAdR as templates 40,41 and the corresponding primers are listed in Supplementary Table 3. The truncated form of Adx (4-108 amino acids) and AdR were expressed in pCwori and pET30a with a C-terminal His-tag to help streamline purification. The Adx(4-108) allowed us to purify a more stable and active version of Adx 42 . All the P450s and RPs were expressed in E. coli (DE3) and purified via His-tag 43 . Typically, the cultures of recombinant cells in LB medium (50 mg L −1 Kanamycin and the rare salt solution) were grown at 37°C (220 rpm) until OD 600 reached 0.6-1.0 and then induced with 0.4 mM β-D-1-thiogalactopyranoside (IPTG) and 0.5 mM δ-aminolevulinic acid (ALA). The cells were cultured at 20°C for another 20 h and centrifuged at 6000 × g for 10 min to pellet cells. The following protein purification was carried out using Ni-NTA affinity chromatography 38 . Purified proteins were flash-frozen by liquid nitrogen and stored at −80°C for later use. P450 concentration was determined from CO-bound difference spectra (ε 450-490nm = 91,000 M −1 cm −1 ). The concentrations of redoxins and redoxin reductases were determined by measuring the absorbance at the selected wavelengths. The extinction coefficients used for calculation of concentrations were ε 414 = 11,000 M −1 cm −1 for Adx 29 , ε 450 = 11,300 M −1 cm −1 for AdR 44 , ε 415 = 11,100 M −1 cm −1 and ε 455 = 10,400 M −1 cm −1 for Pdx, ε 378 = 9,700 M −1 cm −1 and ε 454 = 10,000 M −1 cm −1 for PdR 25,45 .
Extinction coefficient determination. The extinction coefficient of FAD-containing ferredoxin reductase was determined by quantitating the FAD released from the protein following guanidinium chloride treatment. The concentration of FAD was calculated at 450 nm using the extinction coefficient of 11,800 M −1 cm −1 46 . PdR was used as a control to validate the method. The ferredoxin extinction coefficient was calculated by denaturing the ferredoxin with 0.3 volumes of 12 M HCl and heating at 100°C for 15 min. Precipitation was removed and the iron concentration was quantited at 535 nm using the extinction coefficient of 22,140 M −1 cm −1 against a blank without iron. The iron content was calculated from a (NH 4 ) 2 Fe(SO 4 ) 2 •6H 2 O standard curve 47 . Pdx was used as a control to validate the method.
Ferredoxin reductase activity assay. Some flavoenzymes have the ability to reduce artificial electron acceptors with NAD(P)H (diaphorase activity). The redox compound 2,6-dichlorophenol-indophenol (DCIP) can replace Fdx as an electron acceptor in vitro. Ferredoxin reductase activity was assayed by monitoring the reduction of DCIP as the decrease of absorbance at 600 nm (ε 600 = 21,800 M −1 cm −1 ) for 1 min 48 . One unit of ferredoxin reductase activity was defined as the amount of enzyme which required to reduce 1 μM of DCIP per sec at 30°C. Electron donor preference was also investigated by using NADH to replace NADPH for the initiation of reactions. NAD(P)H coupling efficiency. NAD(P)H coupling efficiencies of P450 reactions were determined by following the previously established procedure 24,50 . Briefly, NAD(P)H consumption was monitored at 340 nm with a SpectraMax plate reader (Molecular Devices), and calculated with the extinction coefficient of 6.22 mM −1 cm −1 24,51 . The substrate consumption was measured by HPLC when NAD(P)H was depleted. All measurements were performed in triplicate. The coupling efficiency was calculated as the percentage of NAD(P)H used for product formation over the total NAD(P)H consumption.
P450 enzymatic assay. Three bacterial P450 enzymes were chosen to analyze the reactivity profile of the three Fdxs and three FdRs. The standard assay contained 1 μM P450, 10 μM Fdx, 5 μM FdR, 100 μM substrate, 0.5 mM NAD(P) + , and NAD(P)H regeneration system (10 mM glucose and 2 U glucose-6-phosphate dehydrogenase) in a final volume of 100 μL reaction buffer (50 mM potassium phosphate buffer, pH 7.4). The ratio of 1:10:5 for P450/Fdx/FdR was used to ensure the adequate electron supplies and hence the efficient P450-mediated conversions. CsA is poorly soluble in water. To improve the solubility and conversion rate, 10% methanol was added to the CYP-sb21 enzymatic assay mixtures containing an alternatively optimized P450/Fdx/FdR content, namely, 2 μM CYP-sb21, 20 μM Fdx, and 10 μM FdR. The reactions were incubated at 30°C for 2 h (YC-17 and mevastatin) or 12 h (CsA) and an equal volume of MeOH was added to quench reactions and precipitate proteins. After high-speed centrifugation, the supernatants were used for HPLC and HRMS analysis.
Statistics and reproducibility. All enzyme catalytic assays were carried out in triplicate. Statistical significance was determined by a one-factor analysis of variance (ANOVA) 52 .
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
The cystal structural information for proteins are available from the Protein Data Bank under the associated PDB code described in this paper. All data needed to evaluate the conclusions in the paper are present in the paper. Additional data related to this paper may be requested from the corresponding author.