Biotechnologically Relevant Enzymes and Proteins P450 Bm3 Fused to Phosphite Dehydrogenase Allows Phosphite-driven Selective Oxidations

To facilitate the wider application of the NADPH-dependent P450 BM3, we fused the monooxygenase with a phosphite dehydrogenase (PTDH). The resulting monooxygenase-dehydrogenase fusion enzyme acts as a self-sufficient bifunctional catalyst, accepting phosphite as a cheap electron donor for the regeneration of NADPH. The well-expressed fusion enzyme was purified and analyzed in comparison to the parent enzymes. Using lauric acid as substrate for P450 BM3 , it was found that the fusion enzyme had similar substrate affinity and hydroxylation selectivity while it displayed a significantly higher activity than the non-fused monooxygenase. Phosphite-driven conversions of lauric acid at restricted NADPH concentrations confirmed multiple turnovers of the cofactor. Interestingly, both the fusion enzyme and the native P450 BM3 displayed enzyme concentration dependent activity and the fused enzyme reached optimal activity at a lower enzyme concentration. This suggests that the fusion enzyme has an improved tendency to form functional oligomers. To explore the constructed phosphite-driven P450 BM3 as a biocatalyst, conversions of the drug compounds omeprazole and rosiglitazone were performed. PTDH-P450 BM3 driven by phosphite was found to be more efficient in terms of total turnover when compared with P450 BM3 driven by NADPH. The results suggest that PTDH-P450 BM3 is an attractive system for use in biocatalytic and drug metabolism studies.


Introduction
Cytochrome P450 monooxygenases (P450s) are versatile enzymes that catalyze a broad range of hydroxylation, epoxidation, sulfoxidation, deamination, and dehalogenation reactions and are attractive for applied biocatalysis (Sono et al. 1996;Bernhardt 2006). In vivo, the main functions include steroid hydroxylation, fatty acid hydroxylation, and xenobiotic detoxification, which consequently makes P450s of interest to the pharmaceutical industry (Gillam and Hayes 2013;Munro et al. 2013). Similar to other monooxygenases, catalysis requires a suitable substrate, dioxygen, and two electrons to be fed into the catalytic reaction cycle (Meunier et al. 2004). The electrons are transferred to the heme from a reducing cofactor via electron transfer proteins, which may be domains fused to the monooxygenase component or may occur as separate proteins transferring electrons in transient assemblies (Hannemann et al. 2007). The need for auxiliary redox proteins presents an obstacle for the efficient application of P450s because (a) complex cofactors have to be integrated into different proteins or protein domains to obtain a functional catalyst and (b) participating enzymes need to match each other in terms of concentration, stability, affinity, and activity to achieve maximal turnover. In addition, the dependency on NAD(P)H, slow electron transport, and coupling efficiency can limit activity (Lundemo and Woodley 2015).
Electronic supplementary material The online version of this article (doi:10.1007/s00253-016-7993-7) contains supplementary material, which is available to authorized users.
The highly active bacterial flavocytochrome P450 BM3 (CYP102A1 from Bacillus megaterium, EC 1.14.14.1) naturally circumvents some of these limitations. As a single polypeptide P450 (class VIII), it has an FAD-and FMN-containing NADPH-cytochrome P450 reductase (BMR domain) fused to the C-terminus of the P450 domain in a 119 kDa protein, that is active as a dimer (Ruettinger et al. 1989;Neeli et al. 2005). The enzyme can be solubly expressed in good yields in heterologous hosts such as Escherichia coli. It is one of the few P450s with reported activities above 1 s −1 , which is partly due to rapid electron transfer through the BMR domain (Munro et al. 1996;Munro et al. 2002). The enzyme efficiently hydroxylates medium-to long-chain fatty acids. Arachidonic acid was reported to be converted with the highest activity (285 s −1 ). Low uncoupling rates of approx. 3.8% were determined when using laurate as substrate (Noble et al. 1999).
After it was shown that the substrate spectrum could be modulated by various mutations, P450 BM3 was widely used as a surrogate for class II mammalian P450s. It also is used as a model P450 in mechanistic studies because of easy expression, high activity, and the availability of crystal structures of the separate domains. Whitehouse et al. reviewed recent progress on this enzyme and describe numerous mutants accepting non-natural substrates; many of them are bulky molecules like drugs and related compounds (Whitehouse et al. 2012).
The regeneration of NADPH, which is stoichiometrically consumed in the monooxygenation reaction, remains a major challenge for the application of P450s. Several regeneration approaches have been investigated, of which the enzymatic regeneration of the cofactor is most commonly applied. One possibility is the use of dehydrogenases; popular examples being glucose-6-phosphate dehydrogenase (Ahmed et al. 1999;Falck et al. 2001;Volz et al. 2002;Murataliev et al. 2004;Chen et al. 2008), glucose dehydrogenase (Sulistyaningdyah et al. 2005;Schewe et al. 2008;Schewe et al. 2009), isocitrate dehydrogenase (Capdevila et al. 1996;Schwaneberg et al. 2001;Peters et al. 2003;Landwehr et al. 2006), and formate dehydrogenase (Maurer et al. 2003;Kühnel et al. 2007). Phosphite dehydrogenase (PTDH, EC 1.20.1.1) is one of the most cost efficient options; it regenerates NADPH while oxidizing cheap phosphite to phosphate with the simultaneous reduction of NADP + . The large change in energy of this reaction makes it thermodynamically almost irreversible and only one inhibitor (sulfite) has been described to date (Costas et al. 2001;Woodyer et al. 2003;Relyea and van der Donk 2005).
The recycling of NADPH with phosphite has been reported for fusions of PTDH with several Baeyer-Villiger monooxygenases (BVMOs) (Torres Pazmiño et al. 2009). A codon-optimized gene was used for expression of PTDH-BVMO fusions in E. coli. The employed PTDH contained 18 mutations which led to enhanced activity towards NADP + , soluble expression, and stability (Johannes et al. 2007). Watanabe et al. also reported a fusion of PTDH to a Pdx-PCNA2 construct which functioned as the BPCNA-utilized protein complex of P450cam and its two electron transfer-related proteins^(PUPPET) for the regeneration of NADH (Watanabe et al. 2013).
To facilitate the application of P450 BM3 as a practical biocatalyst, we present here a fusion of P450 BM3 to the optimized variant of PTDH for cofactor recycling: PTDH-P450 BM3 . The turnover of NADP + to NADPH by the fused PTDH domain should enable the use of phosphite as a cheap reductant for performing P450 BM3 -catalyzed oxygenations with phosphate as a byproduct (Scheme 1). The fusion protein is compared to the regeneration system with the free PTDH and P450 BM3 and the P450 BM3 without cofactor regeneration system.
Bacterial strains, plasmids, and cloning E. coli TOP10 (Invitrogen, Carlsbad, CA, USA) was used for genetic manipulations. For expression, a pBAD vector or a modified pBAD vector with an NcoІ restriction site replacing NdeІ were used. The plasmid pPTDH in which the codonoptimized ptxD gene is fused to an N-terminal histidine tag codes the 18-fold mutated PTDH without a fused Scheme 1 Schematic model of PTDH coupled to P450 BM3 facilitating phosphite-driven cofactor recycling monooxygenase. The vector pJOE-P450 BM3 encoding the wild-type gene cyp102A1 was a kind gift from Prof. B. Hauer (University of Stuttgart) (Scheps et al. 2013). For the creation of the fusion enzyme, the vector pCRE2-PAMO was used as a template. It carries the 18-fold mutant of the phosphite dehydrogenase from Pseudomonas stutzeri WM88 with N-terminal histidine tag and the pamO gene in fusion (Torres Pazmiño et al. 2009). The fusion construct pCre2-P450 BM3 was made implementing the In-Fusion technique from Clontech, resulting in the sequence coding for the fusion construct PTDH-P450 BM3 (KX768143). Using primers with overhangs matching the expression vector, cyp102A1 encoding P450 BM3 was amplified via PCR using pJOE-P450 BM3 as template and then hybridized with the respective expression vector, following the recommendations of the manufacturer. For comparison studies with the unfused enzyme, cyp102A1 was subcloned into pBAD. Implementing primers with an overhang matching the target vector, cyp102A1 was amplified via a PCR reaction. The PCR product was purified and utilized as a primer for a second PCR reaction, subcloning the gene encoding P450 BM3 into the respective expression vector, yielding pBAD-P450 BM3 .
Screen for soluble enzyme expression E. coli TOP10, BL21 (DE3), C43 (DE3) and SHuffle cells harboring pCre2-P450 BM3 were pre-grown overnight at 37°C in ampicillin-supplemented LB medium (5 mL; 50 μg/mL). TB-medium (50 mL, ampicillin-supplemented) in a non-baffled glass shaking flask (250 mL) was inoculated with an OD 600 of 0.08 and incubated at 37°C and 180 rpm. After an OD 600 of 0.75-0.85 was reached, expression of PTDH-P450 BM3 was induced by the addition of 0.02% arabinose (final concentration). The cells were incubated for 25 h at 30°C and 135 rpm. Cells were harvested by centrifugation (20 min, 8800×g, 4°C), washed with ice-cold sodium phosphate buffer (50 mM, pH 7.5) and frozen (−20°C) until cell lysis by sonication. The cell debris was separated by centrifugation and the content of PTDH-P450 BM3 in the cell-free extract was investigated via CO-difference spectroscopy. For the optimization of the arabinose concentration and the temperature for enzyme expression, TB-medium (2.5 mL) in 24square deep well plates with the sandwich cover system from enzyscreen®, was inoculated with a preculture of E. coli TOP10, harboring pCre2-P450 BM3 (5 μL). Cultures were grown at 17°C (48 h), 24°C (32 h), 30°C (16 h), and 37°C (16 h) and expression was induced directly after inoculation with 0% (as control), 0.002%, 0.02%, and 0.2% of final arabinose. Cells were harvested via centrifugation and cell pellets were frozen (−20°C, 16 h). Cell lysis was performed enzymatically (200 μL lysis buffer consisting of 2 mg/ mL lysozyme, 0.1 mg/mL DNasel, 5 mM MgSO 4 in 50 mM Tris/HCl (pH 8)) and with a freeze-and-thaw cycle (plate was shock frozen in liquid nitrogen and incubated in a shaker at 30°C for 1 h). Samples were taken and the cell debris was separated from the soluble fraction by centrifugation. Both the lyzed cell and the soluble fraction were analyzed for expression of PTDH-P450 BM3 via SDS-PAGE.
The PTDH was expressed as described before by Dudek et al. (Dudek et al. 2013). E. coli TOP10 cells were transformed with the plasmid and grown overnight at 37°C in 5 mL of LB supplemented with ampicillin (50 μg/mL). The next day cells were diluted (1:100) in TB supplemented with ampicillin (50 μg/mL) and arabinose (0.02% w/v) and incubated for 16 h at 30°C. Cells were harvested via centrifugation, washed once with ice-cold sodium phosphate buffer (50 mM, pH 7.5), and were stored until further use (−20°C).

Enzyme purification
For purification, the frozen cells were resuspended in sodium phosphate buffer (50 mM, pH 7.5, optionally supplemented with DNase) at RT and disrupted via sonication on ice. The resulting crude extract was centrifuged and the supernatant/ cell-free extract (CFE) diluted to a protein concentration of ∼20 mg/mL. The diluted CFE was then precipitated at an ammonium sulfate (AS) concentration of 55% (w/v) and collected by centrifugation (18,600×g, 15 min, 4°C). The resulting pellet was resuspended in sodium phosphate buffer (50 mM, 500 mM NaCl pH 7.5) and subjected to an affinity chromatography purification on a 5 mL HF HisTrap column on an Äkta purifier (GE Healthcare). Fractionated elution was carried out with imidazole (250 mM). The buffer was removed by a second AS precipitation with 65% AS. The resulting pellet was resuspended in ice-cold Tris/HCl. After the addition of sucrose (20 mg/mL) for stabilization during storage, the enzyme was shock frozen in liquid nitrogen and stored at −80°C (van Beek et al. 2015).
The PTDH was purified using Ni Sepharose High Performance (GE Healthcare) as described before (Dudek et al. 2013). Ten percent glycerol (v/v) was added to the purified enzyme, which was shock frozen and stored at −20°C.

Concentration determination
Concentrations of P450 BM3 and PTDH-P450 BM3 were determined by measuring CO-difference spectra (Omura and Sato 1964). After the addition of sodium dithionite (spatula tip), the enzyme solution was incubated (5 min) and a baseline spectrum from 400 to 500 nm was recorded. The samples were then bubbled with carbon monoxide (∼60 s) and incubated for 10 min before the UV spectrum from 400 to 500 nm was recorded. After subtraction of the baseline spectrum, the amount of P450 was calculated based on the maximum absorbance at 450 nm minus the absorption at 490 nm (ε 450nm = 91 mM −1 cm −1 ). This method is specific for the native P450-domain, therefore the concentration of native fused PTDH is an estimation that might vary slightly from the actual concentration.

Activity determination and kinetic measurements
NADPH oxidation or NADP + reduction rates were measured spectrophotometrically at 30°C by following the absorption at 340 nm over 2-5 min (ε 340 = 6.2 × 10 3 M −1 cm −1 ) in a final volume of 500 μL in a quartz cuvette. For the investigation of the properties of the P450 BM3 (fused or WT), typical reaction mixtures contained lauric acid as a substrate (2.5 mM from a 25 mM stock in ethanol/methanol (1:1)) and enzyme (100 nM) in Tris buffer (50 mM, pH 7.5). The reaction was started by the addition of NADPH (250 μM). For the investigation of the PTDH, the reaction mixture contained enzyme (100 nM) and phosphite as the substrate (5 mM). NADP + was added to start the reaction (250 μM). For the determination of kinetic parameters or concentration-based activity, the concentration of the respective component was varied. All measurements were performed in triplicate. Kinetic parameters were calculated using the Michaelis-Menten equation as a function of the respective substrate concentration in GraphPad Prism 6.

GC-MS analysis
Samples from biotransformations investigating cofactor recycling (120 μL) were taken, and the conversions stopped in Eppendorf tubes prepared with NaCl (6 mg) and H 2 SO 4 (50%, 10 μL). The reaction mixtures were extracted twice with twice the volume of MTBE/hexane (1:1); the internal standard, decanoic acid (1 mM in the final derivatization mixture) was added during the first extraction with the MTBE/hexane phase. The organic phases were collected and evaporated. Samples were resuspended in MTBE (60 μL), followed by the addition of 1% TMS in BSTFA (60 μL) and incubated at 75°C for 30 min for derivatization. Tubes were sealed with parafilm during derivatization (Lalman and Bagley 2004;Scheps et al. 2013).

UPLC-MS analysis
Samples were analyzed on a Waters ACQUITY UPLC liquid chromatography system coupled to a Waters Synapt HDMS mass spectrometer equipped with an electrospray (ESI) ionization source. Samples taken during biotransformations were placed in a 96-well Nunc plate (injection volume 5-10 μL). Chromatograph separations were carried out on an ACQUITY UPLC BEH C18 column (130 Å, 1.7 μm × 2.1 mm × 100 mm; Waters, Milford, MA, USA) applying a flow rate of 0.5 mL min −1 at a column temperature of 45°C. The mobile phase consisted of ultra-pure water supplemented with formic acid (0.1% v/v, phase A) and pure acetonitrile (phase B). The gradient applied for separation was 0.0-6.0 min (10-70% phase B), followed by a return to the initial mobile phase composition over 0.01 min.
The MS analysis was performed under the following conditions using the parameters stated. Positive electrospray ionization (ESI) conditions in V-mode. A generic method with two scan functions was used as follows: m/z 80-1000, cone voltage 20 V and 0.1 s scan time, the collision energy (CE) in function 1 was 20 V and in function 2 an energy ramp of 15-45 V was applied, the transfer cell CE is 12 V. Data was collected in centroid mode. Leucine-enkephalin was used as lock mass (m/z 556.2771) for internal calibration at a concentration of 250 pg mL −1 and a flow rate of 0.04 mL min −1 . The MS data was processed in MetaboLynx 4.1 (Waters, Milford, MA, USA) using both the mass defect filter (MDF) and the dealkylation tool. The list of proposed metabolites was reviewed manually. A semiquantitative estimation of substrate conversion and metabolite formation were assessed by calculation of the fraction of the substrate peak area or a product peak area to the total related peak area detected (Kulig et al. 2015).

Enzyme expression and isolation
The plasmid pCre2-PAMO described by Torres Pazmiño et al. was used as a template to create an expression plasmid for the production of P450 BM3 fused to the C-terminus of a codonoptimized and stabilized PTDH which carries a His-tag at its N-terminus (Torres Pazmiño et al. 2009). Using this construct, six amino acids (SRSAAG) link the two proteins into one fusion protein of 157 kDa. For producing the P450 BM3 as a reference for the non-fused enzyme, the P450 BM3 -encoding gene was cloned into a regular pBAD vector, resulting in expression of P450 BM3 with a C-terminal His-tag (119 kDa) (Fig. S1). The PTDH was expressed from a pBAD vector with an N-terminal His-tag (Dudek et al. 2013).
The expression of PTDH-P450 BM3 was optimized in terms of the type of E. coli host strain (Table S1), the expression temperature, and the concentration of arabinose added for induction (Fig. S2). Four E. coli strains with different features for recombinant protein expression were tested for soluble production of PTDH-P450 BM3 , namely E. coli TOP10, E. coli BL21(DE3), E. coli C43, and E. coli SHuffle. Expression of soluble and functional enzyme was determined by CO-difference spectroscopy. The most efficient strain for production of PTDH-P450 BM3 was E. coli TOP10, while almost no expression was detected in E. coli BL21(DE3). To optimize soluble expression in E. coli TOP10, the temperature and concentration of inducer were varied; the best yield was obtained at 17°C and induction with 0.02-0.2% arabinose. As judged by SDS-PAGE analysis, addition of δ-amino levulinic acid further improved expression. Recombinant production of PTDH-P450 BM3 and P450 BM3 using E. coli TOP10 in a culture volume of 50 mL yielded 9.5 mg PTDH-P450 BM3 and 9.8 mg unfused P450 BM3 in the cell-free extract after 42 h cultivation. This shows that the functional expression of both P450 BM3 variants is very efficient. By exploiting the His-tag on both P450 BM3 variants, both enzymes could be easily purified. The integrity of the purified P450 BM3 and PTDH-P450 BM3 was confirmed by the characteristic UV/Vis-spectrum (Fig. 1). The UV/Visspectral features of both proteins are nearly identical which confirms the incorporation of the essential heme and flavin cofactors.
To verify whether the obtained enzymes were functional, monooxygenase and dehydrogenase activities were measured. For testing P450 BM3 activity, conversion of lauric acid (1) by P450 BM3 was analyzed. GC analysis revealed formation of monohydroxylation products in positions 11 (2), 10 (3), and 9 (4) in similar amounts (Scheme 2). By monitoring the rate of NADPH consumption using 50 nM enzyme, monooxygenase activities of 6.7 s −1 for P450 BM3 and 8.9 s −1 for PTDH-P450 BM3 were found, while uncoupling rates (defined as NADPH consumption in the absence of substrate) were determined as 0.22 s −1 for PTDH-P450 BM3 and 0.19 s −1 for P450 BM3 . This shows that the fused enzyme displays a slightly higher monooxygenase activity (Fig. 2a). For PTDH activity, a similar trend was observed with a k obs of 0.71 s −1 for the non-fused PTDH and 2.5 s −1 for the fused PTDH (Fig. 2b). These data show that PTDH-P450 BM3 was fully functional, displaying both monooxygenase and dehydrogenase activities.

Enzyme concentration dependent activity
Neeli et al. demonstrated that inter-monomer electron transfer is required for efficient hydroxylation of lauric acid and therefore the dimeric form of P450 BM3 is the catalytically active variant (Neeli et al. 2005). To examine if the oligomerization behavior of the fused enzymes affected the reactivity of the fusion construct, the specific activities of the two non-fused enzymes and PTDH-P450 BM3 were determined by following the consumption or generation of NADPH at different enzyme concentrations (1-50 nM for the P450 reaction, 1-100 nM for the PTDH reaction). The observed monooxygenase activity Concentrations of (PTDH)-P450 BM3 were determined via CO-difference spectra, PTDH concentrations by Waddell's method. Experiments were performed in triplicate decreased sharply at enzyme concentrations below 10 nM. Hyperbolic fitting of k obs values versus enzyme concentration gave apparent half-saturation concentrations of 8.8 ± 2.1 and 3.3 ± 0.9 nM for P450 BM3 and PTDH-P450 BM3 , respectively (Fig. 2a). A similar enzyme concentration dependence was previously reported for native P450 BM3 and is assumed to reflect the dissociation constant K D for the dimerization of P450 BM3 , being active as a dimer (Neeli et al. 2005). The lower K D for the fusion enzyme suggests that the fusion partner promotes oligomerization.
The rate of phosphite-dependent NADP + reduction by PTDH (−P450 BM3 ) was determined as a function of the concentration of the fused or non-fused PTDH (Fig. 2b). The k obs for the PTDH reaction was relatively unaffected at enzyme concentrations of 5-100 nM. However, at an enzyme concentration of 1 nM, a decrease in the specific activity was observed for PTDH-P450 BM3 , while an increase was detected for the non-fused PTDH.

Kinetic parameters of PTDH-P450 BM3
The steady state kinetic parameters of PTDH-P450 BM3 were determined and compared to those of the non-fused PTDH and P450 BM3 . For analyzing the P450 kinetic parameters, the lauric acid-dependent oxidation of NADPH was measured in the absence of phosphite for the fusion enzyme and for P450 BM3 (Table 1). The phosphite-dependent reduction of NADP + was measured for PTDH-P450 BM3 and PTDH in the absence of lauric acid ( Table 2). The kinetic properties of the fused enzyme were found to be in the same range as those of the separate enzymes. While the K m values differed marginally, PTDH-P450 BM3 displayed a higher k cat for both enzyme domains (for PTDH activity, 1.9 s −1 for PTDH vs. 2.3 s −1 for PTDH-P450 BM3 ; for P450 activity, 7.3 s −1 for P450 BM3 vs. 8.0 s −1 for PTDH-P450 BM3 ). These results confirm that the fused system is fully functional and that both domains are at least as efficient as their non-fused counterparts. When considering lauric acid, the PTDH showed a lower activity than the P450 BM3 component. Therefore, the phosphite-dependent generation of NADPH should be the rate-limiting step during bioconversions. Yet, for other P450 BM3 substrates, lower rates have been reported which makes PTDH a suitable NADPH regeneration partner.

Phosphite-driven drug conversion under screening conditions
The high catalytic activity makes P450 BM3 an attractive biocatalyst for drug metabolism studies. Variants of the enzyme can be used for the production of human metabolites (Reinen et al. 2011). The application of PTDH-P450 BM3 for exploring drug metabolism under screening conditions was investigated with omeprazole, a treatment for gastroesophageal reflux disease, and rosiglitazone, an antidiabetic drug. Both drugs are metabolized by P450 BM3 (Whitehouse et al. 2012).
To investigate the functionality of the PTDH-P450 BM3 system, conversion of both drugs was performed with NADPH or with phosphite plus NADP + as electron source. Furthermore, conversion by the fusion enzyme was compared with conversions catalyzed by P450 BM3 . For these conversions, 10 μM of either omeprazole or rosiglitazone were incubated at 30°C with purified P450 BM3 or PTDH-P450 BM3 in a final reaction volume of 100 μL. Samples were taken after 0 and 4 h and analyzed by UPLC-MS.
Both enzyme variants produced a single metabolite when using omeprazole (5) as substrate. A monohydroxylated Table 1 Kinetics of NADPH dependent lauric acid oxidation. Kinetic parameters of PTDH-P450 BM3 and P450 BM3 for NADPH and lauric acid were analyzed by following absorption of NADPH in a spectrophotometer at 340 nm. Specific activities were determined with 100 nM of the respective enzyme and varying concentrations of either NADPH or lauric acid. When fixed, NADPH had a concentration of 200 μM and lauric acid of 2.  (6) was evident from a mass shift from the parent compound of 15.996 Da. Hydroxylation of one of the methyl groups of the 4-methoxy-3,5-dimethylpyridine moiety in omeprazole was proposed on the basis of fragmentation data (Scheme 3a). As expected, conversion with P450 BM3 only occurred when NADPH was added as an electron source and not in reactions with phosphite and NADP + . Around 16% of the hydroxylated metabolite was formed in the NADPH-driven P450 BM3 reaction (Fig. 3a). In the NADPHdriven reaction with the fusion enzyme, 25% of the same metabolite was formed. Reactions driven on phosphite as the sole electron donor showed similar behavior in terms of product formation. Under the tested conditions, the bifunctional enzyme was more efficient compared to non-fused P450 BM3 .
In reactions with rosiglitazone (MH + 358.12; 7) two main products were formed: a demethylated metabolite (MH + 344.11; 8) and an N-oxide (MH + 374.12; 9) (Scheme 3b) were proposed based on the fragmentation data. The unfused P450 BM3 produced these metabolites in reactions with NADPH as electron donor with a yield of 26% for the demethylation and 36% for the N-oxidation product after 4 h. Very little conversion (<5%) occurred when NADP + and phosphite were used. The ratio of demethylation vs. N-oxidation shifted when using PTDH-P450 BM3 ; around 50% demethylation and 15% N-oxidation product were formed in reactions with the bifunctional enzyme (Fig. 3b). As observed for omeprazole, the total turnover was higher in reactions with the fusion enzyme. Furthermore, conversions with PTDH-P450 BM3 did not show a significant difference when changing the electron source (NADPH or phosphite).

Cofactor recycling
To establish the efficiency of the engineered PTDH-P450 BM3 fusion enzyme with respect to cofactor recycling, conversions were analyzed in greater detail. Omeprazole and rosiglitazone are poorly soluble in aqueous solution and so lauric acid was used as a model substrate to investigate bioconversions with limited NADPH supply. Reactions were performed with the fused enzyme (PTDH-P450 BM3 ), the non-fused parent proteins (PTDH + P450 BM3 ), and P450 BM3 without any cofactor recycling enzyme. Reaction mixtures consisted of 2.25 mM lauric acid, 4 mM phosphite, 50 μM NADPH, and 1.0 μM enzyme. As a consequence, conversions of lauric acid greater than 2.2% relied on cofactor regeneration by the PTDH. Conversions were monitored over time by taking samples for the analysis of phosphate production and the quantification of lauric acid and its metabolites. The yield of product and remaining substrate was determined as a percentage of the respective peak area to the total related peak area. Reactions were performed in duplicate (5 = omeprazole, 6 = monohydroxylated omeprazole, 7 = rosiglitazone, 8 = demethylated rosiglitazone, 9 = monohydroxylated rosiglitazone) Uncoupling was investigated by comparing the amount of phosphate formed, as an indicator of the amount of NADPH regenerated, and the amount of lauric acid oxidized. The concentration of phosphate was measured via a blue reaction product formed with molybdate (Saheki et al. 1985). Formation of phosphate which exceeded the production of hydroxylated lauric acids indicated uncoupling was occurring. Both the conversion of lauric acid and the amount of products formed were quantified after TMS derivatization by GC-MS analysis (Scheps et al. 2013). Three monohydroxylation products of lauric acid were identified via the characteristic elution and fragmentation patterns of their TMS derivatized products ( Fig. S3 and Fig. S4) and were formed at varying concentrations. The products were identified as 11-hydroxylauric acid (2), 10-hydroxylauric acid (3), and 9-hydroxylauric acid (4).
Conversions of lauric acid by PTDH-P450 BM3 reached completion after less than 2 h with a yield of 2.2 mM of combined monohydroxylation products in a ratio of 32.8% 9-hydroxylauric acid (4), 30.7% 10-hydroxylauric acid (3), and 36.7% 11-hydroxylauric acid (2). Excess phosphate, compared to the total amount of hydroxylated products formed, was produced mainly after conversion of lauric acid was complete. After 4 h, 25% more phosphate than product was formed, but during the first 1.5 h, only 10% phosphate production was uncoupled from substrate hydroxylation and uncoupling was still less earlier in the reaction (<5%). From the amount of monohydroxylated products formed, it was calculated back that phosphite supplied electrons for 44 turnovers of the NADPH cofactor. Even more turnovers of NADPH occurred, if the uncoupling is taken into consideration (Fig. 4a, b).
Conversions performed with non-fused PTDH and P450 BM3 were less efficient than with the fusion enzyme (production of 1.6 mM monohydroxylated lauric acid after 2.5 h in comparison to 2.2 mM for the fusion enzyme). After 4 h, 1.86 mM of the combined hydroxylation products were formed with a similar ration between 9-, 10-, and 11hydroxylauric acid as observed for the fused enzyme. The reaction did not reach completion after 4 h. Thus, phosphate production was still coupled to the product formation (approx. 6% uncoupling). In the time frame analyzed, at least 36 turnovers of the NADPH cofactor occurred (Fig. 4c, d).
The P450 BM3 without cofactor regeneration system was not able to drive the reaction to a product yield near the concentration of cofactor added (only 27.6 μM hydroxylated product formed, with 1.94 mM of lauric acid remaining, when 50 μM NADPH were supplied). With 30.0% 9-hydroxylauric acid (4), 25.6% 10-hydroxylauric acid (3), and 44.4% 11-hydroxylauric acid (2), the latter was favored more than in the other two reactions (Fig. 4e, f). As expected, no significant phosphate was produced in reactions without PTDH (data not shown).
Interestingly, in controls without NADPH added to the reaction, PTDH-P450 BM3 and PTDH + P450 BM3 still formed small amounts of product (16.4 and 10.4 μM of total monohydroxylated products), whereas no monohydroxylated products were formed in the controls with P450 BM3 . Possibly, these conversions are driven by phosphite and minor amounts of NADPH and/or NADP + co-purified with the enzyme.

Discussion
While the natural fusion of the P450 oxygenase domain to the BMR domain in P450 BM3 already makes this P450 a convenient system for oxidative biocatalysis, a NADPH regenerating system is still required for its efficient application. With the design of PTDH-P450 BM3, a fully self-sufficient system for phosphite-driven conversion of substrates by P450 BM3 was engineered. It utilizes the reducing power of phosphite to generate NADPH from NADP + , which then donates electrons into the P450 reaction cycle for oxidation. This simplifies enzyme handling by providing the benefits of a single biocatalyst, without compromising the catalytic properties of either the PTDH or P450 BM3 .
Specific activity was fully retained at lower enzyme concentrations for PTDH-P450 BM3 compared to P450 BM3 , indicating a higher affinity for oligomerization. An increase in the specific activity was observed for both enzymes in the fusion. Fogle et al. postulated that product release is the rate-limiting step for the variant stabilized and optimized for the conversion of NADP + (referred to as 12x-PTDH, lacking mutation A176R compared to the variant fused to P450 BM3 ) (Fogle and van der Donk 2007). The C-terminal modification of the PTDH might lead to structural changes that promote the release of NADPH from the active site. For the P450 BM3 , the product release and the electron transfer from the BMR to the heme domain are described as rate limiting (Whitehouse et al. 2012). Considering that the PTDH is N-terminally fused to the heme domain of P450 BM3 , a structural change that influences product release seems likely. Additionally, the assembly of the BMR domain of one P450 BM3 with the heme domain of a second P450 BM3 may be promoted by the fusion, and therefore the electron transfer may improve.
Phosphite-dependent conversion of omeprazole and rosiglitazone as exemplary drugs by PTDH-P450 BM3 showed comparable turnover to NADPH-driven reactions. The total turnover was slightly higher than with conversions containing P450 BM3 and NADPH in excess. This indicates that the fused enzyme is more efficient despite the observation that PTDH showed a lower activity than P450 BM3 when testing the fused domains separately. This can be explained by the relatively low rate of turnover of these drugs by P450 BM3 . Butler et al. reported conversion of omeprazole by wild-type P450 BM3 implementing glucose-6-phosphate dehydrogenase as a cofactor recycling system. In reaction mixtures with purified enzyme (0.1 μM) and the same concentration of substrate as used here (10 μM), less than 1% product formation was observed within 30 min. Higher conversions are described for mutants in the heme domain of P450 BM3 (e.g., F87 V ∼50%) (Butler et al. 2013). In this study, we report a higher turnover for the wild-type P450 BM3 without any cofactor recycling system (16%) and a slightly higher turnover for the fusion enzyme (25-26%). The higher yields are most likely caused by the higher enzyme concentrations that we employed and a prolonged reaction time. Clearly, it would be attractive to incorporate known mutations in P450 BM3 in order to generate a powerful phosphite-driven PTDH-P450 BM3 capable of generating drug metabolites.
The p-hydroxylation of the pyridine ring and Ndemethylation of rosiglitazone are both major reactions when considering the metabolism of these drugs in humans while the N-oxidation is not listed as a relevant pathway (Cox et al. 2000). N-demethylation is mainly performed by the human P450 2C8 (Baldwin et al. 1999). Rosiglitazone conversions . e, f P450 BM3 . Samples were taken and analyzed by GC-MS to determine substrate depletion (a, c, and e; solid line) and product formation (a, c, and f) (9-hydroxylauric acid, dotted line; 10hydroxylauric acid, dotted dashed line; 11-hydroxylauric acid, dashed line). In b and d, uncoupling was examined by following phosphate production measured with the molybdate assay (solid lines) and comparing it to total product formation determined by GC-MS (double dotted dashed line). For P450 BM3 conversions, only a background of phosphate was detected (data not shown). All samples were measured in triplicate. Controls were reaction mixtures without enzyme and reaction mixtures without NADPH of up to 99% by mutants of P450 BM3 using glucose-6-