Production of drug metabolites is one area where enzymatic conversion has significant advantages over synthetic chemistry. These high value products are complex to synthesize, but are increasingly important in drug safety testing. The vast majority of drugs are metabolized by cytochromes P450 (P450s), with oxidative transformations usually being highly regio- and stereo-selective. The PPIs (proton pump inhibitors) are drugs that are extensively metabolized by human P450s, producing diverse metabolites dependent on the specific substrate. In the present paper we show that single mutations (A82F and F87V) in the biotechnologically important Bacillus megaterium P450 BM3 enzyme cause major alterations in its substrate selectivity such that a set of PPI molecules become good substrates in these point mutants and in the F87V/A82F double mutant. The substrate specificity switch is analysed by drug binding, enzyme kinetics and organic product analysis to confirm new activities, and X-ray crystallography provides a structural basis for the binding of esomeprazole to the F87V/A82F enzyme. These studies confirm that such ‘gatekeeper’ mutations in P450 BM3 produce major perturbations to its conformation and substrate selectivity, enabling novel P450 BM3 reactions typical of those performed by human P450s. Efficient transformation of several PPI drugs to human-like products by BM3 variants provides new routes to production of these metabolites.
Replacing traditional synthetic routes to complex organic molecules using enzyme-catalysed reactions has many advantages. These include lower energy requirements, fewer synthetic steps and the ability to activate centres that cannot be readily activated by traditional chemical synthesis methods. A promising area for this research is in the production of fine chemicals and drug metabolites (e.g. [1–3]). Human drug metabolites are often complex molecules that are expensive to prepare chemically, and thus attractive candidates for the application of enzymatic synthesis, given their high value as standards and as reagents in drug testing for the biotechnology and pharmaceutical industries . Cytochromes P450 (P450s or CYPs) are a superfamily of haem-containing mono-oxygenase enzymes that are of particular interest for synthetic applications, due to their ability to activate dioxygen and to insert one of its oxygen atoms into an unactivated CH bond, with the second oxygen atom being reduced to water . P450s catalyse numerous physiologically important reactions (e.g. oxidation of arachidonic acid to epoxyeicosatrienoic acids that have vasodilatory and anti-inflammatory effects in mammals) , and are also increasingly exploited for natural and unnatural oxidative reactions with potential uses in, for example medicine, bioremediation and pharmacology. These include hydroxylation of fatty acids/steroids, epoxidation of styrene, dealkylation of various drugs, and applications in both catabolic (e.g. initiation of the breakdown of vitamin D3 by human CYP24A1) and anabolic (e.g. cholesterol synthesis by CYP51 enzymes) processes (e.g. [5,7,8]). Human P450s are responsible for most of the primary (phase I) xenobiotic metabolism in humans, producing oxidized metabolites of pharmaceuticals that are then more readily excreted, or else further modified and targeted for excretion by phase II enzymes such as glutathione transferases. However, P450s are also responsible for the generation of activated drugs from their prodrug forms (e.g. the oxidation of the anti-cancer drug ellipticine by human CYP3A4 to generate DNA-modifying products) .
In recent years, the synthetic potential of the P450s has been recognized, and various P450s have been targets for biotechnological exploitation and subject to the diversification of their substrate recognition and the regio-/stereo-selectivity of their substrate oxidation by protein engineering methods such as chimaeragenesis and directed evolution (e.g. [10,11]). The most promising P450 for biotechnological applications is the Bacillus megaterium P450/CPR (P450 reductase) fusion enzyme P450 BM3 (BM3; also known as CYP102A1), due to its high expression levels, soluble nature (compared with membrane-bound eukaryotic P450s) and convenient single component organization (P450 and CPR domains on the same polypeptide chain). Its efficient electron transfer system (for electron transfer both within the CPR domain and between the CPR and the P450 domain) results in BM3 having the highest reported catalytic rate of substrate oxidation among the P450s (e.g. 17000 min−1 with arachidonic acid) . Engineering of BM3 has generated variants capable of many different reactions, including studies to create an olefin cyclopropanation catalyst , and to enable production of oxidized steroids and the oxidation of short chain fatty acids [14,15]. Approaches to altering BM3 substrate selectivity have often relied on the use of random mutagenesis, mostly by error-prone PCR (e.g. ). Although this methodology has identified several mutants with novel catalytic properties, it can be a rather random approach to new catalyst generation. BM3 mutants generated by directed evolution approaches typically have numerous mutations dispersed across the P450 domain . However, through analysis of data from several preceding studies on BM3 enzymes evolved for diverse activities, it has become clear that a small number of P450 (haem) domain amino acids are mutated most frequently in variants with novel activities, including those at Ala82 and Phe87 . Previously, we showed that the A82F mutation causes structural destabilization of the BM3 haem domain, resulting in a lower melting temperature, but also in an altered substrate specificity profile. X-ray crystallography showed that the substrate-free A82F haem domain also occupied a conformation different to that of the substrate-free WT (wild-type) P450 . The human gastric PPI (proton pump inhibitor) drug OMP (omeprazole) became a good substrate for the A82F BM3 enzyme, and our structural data revealed the productive mode of binding for OMP that enables its oxidation by the BM3 A82F mutant at the same position as that catalysed by its major human metabolizing P450 CYP2C19 . Binding of OMP was further improved by the additional F87V mutation, where removal of the aromatic bulk of Phe87 frees space immediately in the vicinity of the BM3 haem iron in the P450 active site .
In the present study, we demonstrate that mutant BM3 P450s containing the Ala82 and Phe87 mutations also facilitate the binding and oxidation of a range of other PPI drugs [ESO (esomeprazole), LAN (lansoprazole), PAN (pantoprazole) and RAB (rabeprazole)] that are not effective substrates for the WT BM3. The products of oxidation of these drugs are diverse from those of OMP but, as found for OMP, in most cases mimic those products produced by the major human P450s that metabolize these drugs. Our data provide further evidence that mutations at a small number of residues which can impact significantly on the structural organization of BM3 (‘gatekeeper’ mutants) are sufficient to induce major alterations in the substrate selectivity profile of this biotechnologically important P450 enzyme. Thus we demonstrate that such BM3 mutants can produce human-like metabolites of a number of PPI pharmaceuticals, with potential uses for these metabolites as standards and as reagents for pharmaceutical testing and FDA (Food and Drug Administration) compliance.
MATERIALS AND METHODS
Mutagenesis and expression of WT and mutant BM3 enzymes
The CYP102A1 gene encoding intact WT flavocytochrome BM3 in the plasmid vector pET15b (Novagen) was used for mutagenesis to create A82F, F87V and F87V/A82F [DM (double mutant)] mutants as described previously . Intact BM3 enzymes were expressed as N-terminal His6-tagged enzymes either from pET15b (F87V and DM) constructs directly or after cloning the WT and A82F genes into pET14b using NdeI/BamHI sites. The WT and mutant haem domain genes were generated using the relevant pET14b/15b constructs as described previously . The haem domain genes (amino acids 1–473 of the 1048 amino acid flavocytochrome) were transferred as NdeI/BamHI fragments into pET20b to allow haem domain production in the absence of an N-terminal His tag and to enable improved protein crystallization. All genes were fully sequenced to confirm relevant mutations and to ensure no exogenous mutations were incorporated. The WT and A82F intact BM3, and WT and all mutant BM3 haem domains, were expressed in BL21 Gold (DE3) Escherichia coli cells (Agilent) in TB (terrific broth) medium with cells grown at 37°C, and with agitation at 200 rev./min in an orbital incubator. F87V and DM intact BM3 proteins were grown using autoinduction TB medium (Melford) from 4 litre transformant cultures and with cell growth for 24–36 h.
Purification of WT and mutant intact P450 BM3 and haem domains
Intact WT and mutant P450 BM3 enzymes and haem domains were purified essentially as described previously . Cells were collected by centrifugation at 4°C (6000 g for 10 min) and resuspended in ice-cold buffer B [50 mM KPi (potassium phosphate), 250 mM NaCl and 10% (v/v) glycerol, (pH 7.0)]. Protease inhibitors (EDTA-free Complete™ tablets; Roche) were maintained in all buffers used during purification. Cells were lysed by sonication on ice using a Bandelin Sonopuls sonicator (at 40% power, with 50 pulses for 5 s each and 25 s between pulses), and the supernatant containing soluble intact BM3 or haem domain protein was collected after high speed centrifugation (20000 g for 40 min at 4°C). The supernatant was collected again after a 30% ammonium sulfate cut on ice. P450 proteins were purified by Ni-IDA (Ni2+-iminodiacetic acid) chromatography (Qiagen), with bound proteins washed extensively at 4°C in buffer B plus 5 mM imidazole, then eluted with 200 mM imidazole in buffer B. Proteins thus purified were transferred into buffer A [50 mM Tris/HCl and 1 mM EDTA (pH 7.2)] and passed down a Sephacryl S-200 SEC column (GE Healthcare; 26×60 cm; AKTA purifier system). Pure BM3 fractions (checked by SDS/PAGE) were concentrated by ultrafiltration (Vivaspin sample concentrator with molecular mass cut-off at 30 kDa; Millipore) and stored in buffer A plus 50% (v/v) glycerol at −80°C. For non-tagged BM3 haem domains, a further 30–60% ammonium sulfate cut was applied. The P450-containing pellet was resuspended in buffer A and dialysed into the same buffer to desalt, then further purified by anion-exchange chromatography on an AKTA purifier system using a Q-Sepharose anion-exchange column (16×10 cm), with elution in a gradient of 0–500 mM KCl in buffer A. Haem domain fractions were desalted (GE Healthcare column; 26×10 cm) on the AKTA into 25 mM KPi (pH 7.0), loaded on to a hydroxyapatite column (Bio-Rad Laboratories; 16×11 cm) and eluted in a 200 ml gradient of 25–500 mM KPi (pH 7.0). Pure haem domains were concentrated by ultrafiltration as described above and used immediately for crystallography, or flash-frozen in liquid nitrogen and stored at −80°C. As described previously, intact BM3 and haem domain proteins with the A82F mutation were passed through a Lipidex 1000 column (PerkinElmer) in 25 mM KPi (pH 7.0) to remove any fatty acid retained during purification .
Concentrations of the LS (low-spin) forms of WT and mutant intact BM3 enzymes and haem domains were determined using molar absorption coefficients of ε418=105 mM−1·cm−1 and ε419=95 mM−1·cm−1 respectively, at the Soret maximum [19,20]. Fe(II)CO complexes were formed by bubbling sodium dithionite-reduced WT/mutant BM3 and haem domains (approximately 2–4 μM) with CO gas . WT and mutants showed near-complete formation of the P450 (thiolate-co-ordinated) state, with little of the P420 state (which probably results from cysteine thiol co-ordination) formed in any case [22,23].
Fatty acid and PPI drug binding to WT and mutant intact BM3 enzymes
Dissociation constants (Kd values) for binding of NPG (N-palmitoylglycine), ESO, LAN, PAN and RAB to WT/mutant intact BM3 enzymes were determined by absorption titrations (~1–2 μM protein) in 100 mM KPi (pH 7.0; assay buffer) at 25°C in 1 cm pathlength quartz cuvettes as described previously [19,24]. Titrations were continued until no further P450 haem spectral changes occurred. Difference spectra (produced by subtracting each successive ligand-bound spectrum from that of the ligand-free enzyme) were generated in each titration. Maxima and minima in each set of difference spectra were identified (using the same wavelength pair in each titration) and the overall absorbance changes Δ(Apeak−Atrough) were plotted against the substrate concentration. Data were fitted using a standard (Michaelis–Menten) hyperbolic function or (where the Kd value is ≤5× the P450 concentration) using the Morrison (quadratic) equation for tight-binding ligands (eqn 1), in order to determine Kd values [25,26]. UV–visible spectroscopy was done on a Cary 50 UV–visible spectrophotometer (Agilent). Data analysis and fitting was done using Origin Pro (OriginLab).
In eqn (1), Aobs is the observed absorbance change at ligand concentration S, Amax is the absorbance change at ligand saturation, Et is the P450 concentration and Kd is the dissociation constant for the P450–ligand complex.
Steady-state kinetic analysis of WT and mutant intact BM3 enzymes with fatty acid and PPI substrates
Steady-state kinetic studies were also done on a Cary 50 UV–visible instrument. Substrate (ESO, LAN, PAN and RAB)-dependent oxidation of NADPH was determined at 340 nm. BM3 concentration was kept constant (in range 25–150 nM) with the substrate concentration varied and a near-saturating NADPH concentration used (200 μM). Assays were done at 25°C in assay buffer with a 1 cm pathlength quartz cuvette. Enzyme rate constants for substrate-induced NADPH oxidation were determined in triplicate at each substrate concentration at 340 nm using Δε340=6.21 mM−1·cm−1. Rate constants were plotted against the substrate concentration. Data were fitted to the Michaelis–Menten equation to define the kcat and Km parameters for substrate-dependent NADPH oxidation and are reported in Table 1.
|BM3 enzyme||Substrate||HS haem (%)||Kd (μM)||kcat (min−1)||Km (μM)||kcat/Km (min−1·μM−1)|
|BM3 enzyme||Substrate||HS haem (%)||Kd (μM)||kcat (min−1)||Km (μM)||kcat/Km (min−1·μM−1)|
EPR spectroscopy (for ligand-free and drug substrate-bound WT and mutant BM3 enzymes) was performed using a Bruker ELEXSYS E500 EPR spectrometer, operating at X band, fitted with an ESR-900 liquid helium flow cryostat (Oxford Instruments) and a Super High Q (ER-4122SHQE) resonator. Spectra were recorded at 10 K with a microwave power of 0.5 mW and a modulation amplitude of 0.5 mT. Protein samples (200 μM) in KPi buffer (100 mM; pH 7.0) were prepared in (i) the absence of solvent; (ii) with addition of either 1.8 μl of methanol or DMSO solvent; and (iii) with 400 μM PPI drug dissolved in 1.8 μl of solvent. Samples were thus in a final volume of 250 μl in buffer or with buffer plus 0.72% solvent.
Enzymatic oxidation of substrates and product characterization
ESO, LAN, PAN and RAB turnover and analysis by LC–MS
Turnover reactions for oxidation of ESO, LAN, PAN and RAB were done in deep-well blocks at 37°C with shaking for 30 min. Reaction mixtures contained purified WT or mutant (F87V, A82F or DM) BM3 enzymes (1 μM), substrate (10 μM), NADPH regeneration system (7.76 mM glucose 6-phosphate, 0.6 mM NADP+ and 0.75 unit/ml glucose-6-phosphate dehydrogenase) in turnover buffer [50 mM KPi and 5 mM CaCl2 (pH 7.4)] in a final volume of 500 μl. On completion of the reaction, protein was mixed with an equal volume of ACN (acetonitrile) containing 1 μg/ml fluconazole IS (internal standard) by shaking the mixed samples at 800 rev./min for 10 min. Precipitated protein was filtered through protein precipitation plates (Phenomenex) into MS vials (FluidX) and clarified by centrifugation (4000 g for 25 min at 10°C). Analysis was done on a Thermo Exactive LC–MS with a CTC PAL auto sampler (Thermo Scientific) with a Kinetex 2.6U XB-C18 100 Å column (Phenomenex). A gradient of 0.1% formic acid to ACN was used to resolve products. Drugs and metabolites were run in positive mode with the molecular ion as M[H]+. All high intensity peaks were selected from the total ion chromatogram and analysed using Thermo Xcalibur quantification software, along with the fluconazole IS. This software then gave total ion readings for both the IS and the metabolites formed. The total ion data were then corrected for the IS and for any degradation of product that occurs non-enzymatically. Additional MS fragmentation analysis for PAN was carried out on an Agilent 6550 iFunnel Q-TOF LC–MS with 1290 Infinity LC system. A ZORBAX Eclipse Plus C18 (2.1×50 mm; 1.8 μm) Rapid Resolution HT column (Agilent) was used with a gradient of 0.1% formic acid to ACN to resolve products. Turnover reactions with the DM BM3 were done as described above, but in single vials with total reaction volumes of 5 ml. Products were extracted using Strata-X SPE columns (Phenomenex), dried under vacuum and eluted in 50:50 ACN/methanol. Fragmentation data analysis was performed with the MassHunter MSC (Molecular Structure Correlator) program (Agilent).
ESO, LAN and RAB turnover and analysis by NMR
Turnover reactions with ESO, LAN and RAB were done in 100–500 ml flasks at 37°C with shaking of reagents at 100 rev./min for 2 h. Reaction mixtures contained purified WT or mutant (F87V, A82F or DM) intact BM3 enzymes (1 μM), substrate (10–100 μM), and the NADPH regeneration system in 60–500 ml of assay buffer. Products were extracted using Strata-X SPE columns, dried under vacuum and eluted in 50:50 ACN/methanol, followed by drying under nitrogen and water removal by freeze drying. Analysis was done on a Bruker Avance 400 MHz NMR. 1H spectra were collected at 400 MHz and 13C spectra at 101 MHz. Spectra were baseline corrected and referenced to TMS (tetramethylsilane) standard by the residual non-deuterated solvent in the sample. δ values are given in p.p.m. and J values are in Hz. Full assignments were made by COSY, HMBC (heteronuclear multiple bond correlation) and HMQC methods. Signal splittings were recorded as singlet (s), doublet (d), doublet of doublets (dd) αβ system (AB) and multiplet (m). Processing was carried out using MestReNova Lite (Mestrelab Research, Santiago de Compostela, Spain) and ACD NMR Processor (Advanced Chemistry Development).
Crystallization of the DM BM3 haem domain and its ESO-bound complex and determination of its structure
Crystallography was performed using the sitting-drop method using a seeding protocol at 4°C. Crystals obtained for the DM haem domain during initial screens were used to create microcrystal screen stocks and consecutive screens (Molecular Dimensions) were made with drops that consisted of 150 nl of the DM haem domain protein (230 μM), 50 nl of seed stock and 200 nl of well solution using a Mosquito liquid handling robot (TTP LabTech). For the ESO–DM haem domain complex, the protein was saturated with ESO ligand before crystallization. ESO was titrated into the DM haem domain until no further change in haem iron spin state towards HS (high spin) was observed. Thereafter samples were concentrated by ultrafiltration in the presence of saturating ESO. Microseeding was also used to produce diffraction quality crystals. Crystals were obtained under a range of conditions and flash-cooled in liquid nitrogen before data collection. The mother liquor was supplemented with 10% PEG 200 where an additional cryoprotectant was required. Data were collected at Diamond synchrotron beamline IO2 (Harwell, U.K.) and were reduced and scaled using XDS . Structures were solved by molecular replacement with the previously solved BM3 haem domain structure in complex with OMP (PDB code 4KEY) using PHASER . Structures were refined using REFMAC5  and Coot .
Oligonucleotide primers for mutagenesis were from Eurofins MWG Operon. ESO and LAN were from Sigma–Aldrich, and PAN, RAB and all standards were from Santa Cruz Biotechnology. Bacterial growth medium (TB) was from Melford. Unless specified, other chemicals were from Sigma–Aldrich and of the highest purity available.
UV–visible spectroscopic binding studies of BM3 ‘gatekeeper’ mutants with diverse PPI drugs
In an previous study, we were able to show that the fatty acid hydroxylase P450 BM3 underwent a conversion of specificity into an OMP-binding/oxidizing P450 on introduction of one or both of the A82F and F87V mutations in the P450 active-site channel . Structural studies showed that the A82F mutation altered the conformational status of the BM3 haem domain to facilitate OMP binding, whereas the F87V mutation improved OMP access to the haem active site for effective catalysis . To investigate whether these mutations also facilitated binding of other PPI drugs in clinical use, we examined P450 haem spectral perturbation with a series of other PPI drugs of diverse structures (Figure 1). We selected the major PPI drugs ESO (the pharmacologically active S-enantiomer of OMP), LAN, PAN and RAB. P450 spectral binding studies were done to ascertain whether these PPIs could bind to mutant (and WT) BM3 P450s, and if binding induced a LS to HS shift in the ferric haem iron spin-state equilibrium that is typical of P450 substrate association. This is usually observed as a haem Soret band maximum shift from ~418 to ~392 nm when fatty acids bind to BM3. Optical binding studies were done with each PPI and the WT, A82F, F87V and F87V/A82F (DM) intact P450 BM3 proteins, and binding data compared with those collected previously with NPG, a tight-binding fatty acid-derived substrate for BM3 [19,31]. Control studies showed no discernible haem spectral perturbation caused by the solvents (DMSO and methanol) in absence of PPIs. NPG gives a near-complete haem iron HS shift for WT and all mutants, with Kd values less than 1 μM (Table 1). However, the binding studies with the selected PPIs showed no spectral perturbation with WT BM3, suggesting negligible productive binding. The F87V mutant also showed no discernible spectral shifts upon titration with the PPIs. However, both the A82F and DM BM3 mutants bound all four PPIs with substantial HS shifts, thus indicating that the A82F mutation is a primary determinant of altered selectivity and productive binding mode for these drugs. In each case, there was an additive effect on PPI affinity by inclusion of the F87V mutation, with the DM showing a greater proportion of HS shift than the A82F mutant and with lower PPI Kd values for the DM BM3 in each case. This is probably due to the influence of the F87V mutation in increasing the size of the active-site cavity and allowing the drugs to move further towards the haem. ESO shows the highest affinity, with an A82F Kd value of 23.9 μM and a DM Kd value of 2.89 μM, associated with a near-complete HS conversion. LAN gives an ~70% shift to the HS form in the DM BM3, with Kd values of 140 μM for A82F and 58.6 μM for the DM. PAN gives up to an ~50% shift with the DM, and with an A82F Kd value of 25.6 μM and DM Kd value of 8.5 μM. RAB showed relatively little HS shift with A82F (~10%) and comparatively weak binding (Kd=159 μM); however, RAB binding was greatly improved in the DM with ~40% HS accumulated at saturation and a Kd value of 43.9 μM (Table 1). Binding spectra for the DM BM3 enzyme with each of the PPI drugs are shown in Figure 2. These UV–visible spectroscopic data indicate that diverse PPI class compounds bind the A82F mutation-containing BM3 gatekeeper mutants. Binding occurs in proximity to the haem catalytic site to displace the haem iron's distal (sixth) water ligand, and thus the PPIs occupy a productive mode similar to that for the fatty acid substrates of P450 BM3.
PPI drug structure and functional groups
Binding of diverse PPI drugs to the P450 DM BM3 enzyme
Steady-state kinetic analysis of PPI turnover by BM3 mutants
Steady-state kinetic studies (in presence of near-saturating NADPH) were performed on intact WT and BM3 mutants, using each of the PPIs at a range of substrate concentrations and comparing the kinetic data with those with NPG as the substrate . As shown in Table 1, the efficient NPG lipid substrate has high kcat values (4050–5130 min−1) and low Km values (26.3–1.9 μM) with WT and the F87V, A82F and DM BM3 enzymes, leading to high catalytic efficiency in all cases. With WT BM3, only LAN showed significant substrate-stimulated NADPH oxidation over background levels (kcat=481 min−1), with a LAN Km value of 61.8 μM. In contrast with the data for the WT BM3, all of the BM3 mutants showed substrate-stimulated oxidation with the four PPIs. DM BM3 showed greatest affinity (lowest Km values) for each of the PPIs, and the highest (for RAB) or joint highest activity, within the margin of error, along with A82F for ESO and PAN (Table 1). However, F87V BM3 proved the fastest LAN-dependent NADPH oxidase (kcat=1560 min−1) (Figure 3). Indeed, despite little evidence for PPI-dependent HS conversion of its haem iron, the F87V BM3 kcat values for PPI-dependent NADPH oxidation were >1000 min−1 in all cases. The superior PPI-binding affinity exhibited by DM BM3 in its Km values (and mirrored in the Kd values) underpins DM BM3 having the highest catalytic efficiency (kcat/Km ratio) for PPI-dependent NADPH oxidation in all cases. DM BM3 also exhibits enhanced binding to the NPG lipid substrate compared with WT BM3 (Km values of 1.91 and 13.9 μM respectively; Table 1).
PPI-dependent steady-state kinetic analysis for the F87V BM3 mutant
The EPR data for the WT P450 BM3 haem domain (Figure 4, A), show a rhombic signal with a single set of g values at gz=2.41, gy=2.25 and gx=1.92 (2.41/2.25/1.92). These are consistent with previously reported data for WT BM3 (2.42/2.25/1.92) and indicative of a LS ferric haem with axial co-ordination to the central iron provided by cysteine thiolate and water . There is no significant signal corresponding to a HS ferric state (results not shown). Addition of the solvent DMSO (at final concentration 0.72%) perturbs the EPR spectrum (Figure 4, B) and gives rise to a new set of LS ferric haem g values at 2.45/2.25/1.91. These data indicate that DMSO solvent affects the environment around the haem iron, probably influencing the locations of water molecules within the active site.
EPR analysis of WT and gatekeeper mutant BM3 haem domains and their ligand complexes
Addition of PPI substrates did not produce any significant formation of HS haem iron in the WT and mutant BM3 EPR spectra (results not shown). WT BM3 exhibits a new species in the LAN-bound form that is different from that induced by DMSO alone (Figure 4, C). The g values for this new LS species are at 2.40/2.25/1.92. A very similar EPR spectrum was observed for WT BM3 bound to RAB (results not shown). The F87V mutation has a significant effect on the EPR spectrum with the substrate-free enzyme showing new LS species (Figure 4, D). Three LS states are observed in the native form (2.48/2.25/1.87, 2.45/2.25/1.90 and 2.41/2.25/1.92), and these are almost unchanged on addition of methanol solvent at 0.72% (2.48/2.25/1.90, 2.43/2.25/1.91 and 2.40/2.25/1.92; results not shown). Thus the organization of water molecules in the haem iron environment is clearly perturbed in the F87V mutant, as a consequence of the removal of the aromatic residue in the vicinity of the haem iron. However, the F87V EPR spectrum is converted into only one major species after the addition of DMSO (2.44/2.25/1.91; Figure 4, E). PPI addition does not then lead to further major changes in the EPR spectra. The A82F haem domain also shows three LS states in absence of solvent/substrate (2.47/2.25/1.90, 2.42/2.25/1.92 and 2.39/2.25/1.92; results not shown), again indicative that this mutation perturbs active-site water organization around the haem iron. Addition of solvent reduces the signal complexity (2.44/2.25/1.91 for DMSO; 2.44/2.25/1.90 and 2.42/2.25/1.91 for methanol; results not shown), but addition of PPI drugs again has little further influence on the EPR spectra. The native BM3 DM haem domain has an even more complex LS EPR spectrum (2.53/2.25/1.87, 2.49/2.25/1.90, 2.44/2.25/1.91 and 2.41/2.25/1.91; Figure 4, F) indicative of the combined influence of the A82F and F87V mutations on the distal haem co-ordination environment. The spectral complexity is decreased slightly on addition of methanol/DMSO solvents, but binding of the PPI drugs again has only marginal influence on the EPR spectrum above that induced by solvent alone, e.g. g values of 2.46/2.25/1.90 and 2.43/2.25/1.91 for the ESO-bound DM haem domain (Figure 4, G).
Collectively, these EPR data demonstrate clearly that major alterations in the environment around the iron distal pocket are induced by F87V, A82F and F87V/A82F mutations in the respective haem domains. DMSO influences the distal water environment considerably in WT and all gatekeeper BM3 mutants, but the effects of methanol are less profound. However, the addition of PPI drugs to the BM3 mutants (despite them inducing substantial HS shifts in the haem iron spin-state equilibrium at ambient temperature for the A82F and DM enzymes) causes only very minor further changes to the EPR spectra at 10 K.
Oxidative turnover of PPI drugs by WT and mutant BM3 enzymes using LC–MS
To analyse the catalytic activity of the WT and mutant forms of BM3, in vitro turnover of each PPI was performed with intact WT, F87V, A82F and DM BM3 enzymes, and products were analysed by LC–MS. Oxidation of each of the drugs was observed, as detailed below.
The LC–MS analysis for the ESO substrate showed a M[H+] of 346.1224 for the starting material, with a natural fragmentation occurring between the sulfoxide sulfur and the methoxybenzimi-dazole moiety to give the corresponding 198.0587 species. Upon enzymatic turnover, a +16 increase in these m/z peaks to 362.1134 and 214.0536 was observed, indicating insertion of oxygen into the pyridinyl fragment, generating oxidized ESO. Data for ESO oxidation by the DM BM3 enzyme are shown in Supplementary Figure S1 (http://www.biochemj.org/bj/460/bj4600247add.htm). The product quantities for ESO oxidation were high, with ~90% of a single monohydroxylated product obtained for both the F87V and DM BM3 enzymes. The A82F BM3 enzyme produced ~40% of the same monohydroxylated product in the same time (Figure 5A). It is interesting to note that, in contrast with previous studies using OMP , the oxidation of ESO results in negligible (<1%) formation of the carboxylated ESO. Our studies with the OMP racemate showed ~15% of the overall product formed was 5-carboxy OMP, resulting from three successive P450-mediated oxidation reactions at the pyridinyl 5-methyl position. The remaining product was the 5-OH OMP . This suggests a different metabolism for the R- and S-isomers of OMP by BM3 mutants carrying the A82F mutation, with successive oxidations of R-OMP (leading to 5-COOH OMP) being much more efficiently catalysed than with S-OMP (ESO) as a substrate. Human CYP2C19 was also reported to show stereoselectivity in OMP oxidation, favouring 5-hydroxylation of R-OMP over ESO, but more efficiently catalysing 5-O-demethylation of ESO over R-OMP .
Proportions of PPI turnover products identified by LC–MS
The LC–MS analysis for the LAN substrate shows the substrate at M[H+] 370.0830 with a natural fragmentation between the benzimidazole and pyridinyl moieties, and the pyridinyl fragment seen as 252.0302. The main oxidized product from oxidation by the DM BM3 enzyme is seen at M[H+] 386.0782 without fragmentation or loss of the pyridinyl moiety, indicating a change in the structure of the central sulfoxide to render it less labile (Supplementary Figure S2 at http://www.biochemj.org/bj/460/bj4600247add.htm). With the LAN substrate no significant turnover occurs with WT BM3 (<1%). However, the F87V and DM enzymes show much more turnover of LAN, with 16% and 33% products formed respectively. DM BM3 is not only more active, but also more selective with almost all turnover producing a single oxidized product (LAN sulfone). The BM3 F87V mutant produced a mixture of the LAN sulfone as the main product and also a secondary oxidized product at ~2.8% of total product (Figure 5B). The second oxidized metabolite was at too low a concentration for NMR analysis, although the LC–MS analysis showed a +16 product peak at M[H+] 386.0782 and an oxidized pyridinyl fragment at 268.0247, indicating hydroxylation on the pyridinyl portion of the molecule (results not shown). With five possible sites of oxidation on this fragment and no standards available, identification based purely on LC–MS analysis was inconclusive. The A82F mutant only shows ~2.5% LAN turnover to two oxidized products, with one of these being the sulfone and the other a +16 product and putatively the same hydroxylated species on the pyridinyl group as formed by the F87V mutant.
The LC–MS analysis for the PAN substrate shows the substrate as M[H+] 384.0819 with a natural fragmentation between the benzimidazole and pyridinyl moieties, and with the pyridinyl fragment seen as 200.0375 (Supplementary Figure S3 at http://www.biochemj.org/bj/460/bj4600247add.htm). The main oxidized product is seen as M[H+] 400.0774 without fragmentation and loss of the pyridinyl moiety. Product(s) from PAN turnover were difficult to reconcile with observed MS data, as the oxidized product also appeared in the starting material and in control samples. This led us to conclude that it was likely a synthesis by-product. We screened potential by-products and found a match with PAN N-oxide. PAN shows no discernible turnover into the N-oxide with WT BM3 (above that seen in control reactions) and the A82F mutant showed little further conversion (~1.5%), whereas F87V produced slightly more of the PAN N-oxide product (~5%). The greatest amount of PAN conversion (again into the N-oxide) was seen with DM BM3 at ~12% over the background level (Figure 5C). Fragmentation analysis confirmed PAN N-oxide as the relevant product. With such low turnover levels, it was not possible to confirm products by NMR, although comparisons with available standards provided conclusive data for enhanced production of the PAN N-oxide by the mutant BM3 enzymes, and particularly for DM BM3.
The LC–MS analysis of the RAB starting material shows a M[H+] peak at 360.1369, and fragmentation at the sulfoxide again occurs with a pyridinyl fragment at 242.0844. Analysis of RAB turnover by LC–MS shows three major metabolites, which are products of oxidative dealkylation reactions (Supplementary Figure S4 at http://www.biochemj.org/bj/460/bj4600247add.htm). Some non-enzymatic degradation of RAB occurs at ~25% with the control samples, largely to the thioether. This is a naturally occurring process , although is apparently increased somewhat by the use of an NADPH regeneration system in the assays, probably due to the availability of electrons to reduce the sulfoxide. The three metabolites observed are the demethylated RAB, the demethylated RAB thioether and the thioether product with loss of the entire pyridinyl alkyl chain. The demethylated RAB shows this same characteristic fragmentation pattern as RAB itself, but with −14 for the methyl loss, and with M[H+] at 346.1220 for demethylated RAB and at 228.0692 for its pyridinyl fragment. The second metabolite, the demethylated RAB thioether, shows a M[H+] at 330.1269. Fragmentation now occurs at the benzimidazole ring, with a 163.1329 species for the fragment with the methyl group and the demethylated species at 149.0234. This altered pattern is probably due to the differing stability of the pyridinyl moiety following the loss of the oxygen. The third and major metabolite was identified as the dealkylated metabolite of the RAB thioether, involving loss of oxygen and then loss of the entire pyridinyl alkyl chain to form the demethoxypropyl RAB thioether. The fragmentation pattern is the same as that seen for the other thioether.
RAB was extensively metabolized by WT BM3 and by each of the mutant BM3 enzymes, with WT BM3 generating 30% of the major metabolite (the dealkylated RAB thioether) and ~10% of the demethyl RAB thioether under the standard conditions used in these assays (10 μM substrate). The A82F mutant showed >70% RAB conversion and a very different product profile. The primary A82F BM3 metabolite is the demethylated RAB thioether at ~45%, with the demethylated RAB at ~20% and the dealkylated thioether only comprising ~8% of the product. The F87V BM3 mutant again shows the dealkylated thioether metabolite as the primary product with ~60% conversion, and with the demethylated thioether as the secondary product (~20%), and the demethylated RAB at ~15%. Finally, DM BM3 showed the greatest total RAB turnover at >95%, with >65% dealkylated thioether, >15% demethylated thioether and ~15% demethylated RAB (Figure 5D). The major human metabolites of RAB are the non-enzymatically produced thioether, and the CYP-mediated metabolites being the demethylated RAB and the RAB sulfone, produced by CYP2C19 and CYP3A4 respectively [35–37].
Oxidation of PPIs by WT and mutant BM3 enzymes using NMR
NMR spectroscopy was used for the analysis of the substrates and the main oxidized products from ESO, LAN and RAB turnover by WT and gatekeeper BM3 mutants. Large-scale enzymatic turnover reactions were carried out for each substrate in order to enable use of NMR for identification of the products seen in the LC–MS analysis. Confirmation of the position of oxidation of ESO at the 5-methyl group was obtained from a combination of 1H, 13C and 2D NMR (Supplementary Figures S5–S8 at http://www.biochemj.org/bj/460/bj4600247add.htm). Confirmation of the position of oxidation of ESO at the 5-methyl group (as catalysed by human CYP2C19 ) was obtained from a combination of 1H, 13C and 2D NMR. Analysis of a LAN standard along with metabolites showed the sulfone to be the product (formed when a second oxygen atom is introduced to the central sulfur atom) (Supplementary Figures S9–S11 at http://www.biochemj.org/bj/460/bj4600247add.htm). The LAN sulfone is also the major human metabolite formed by the human CYP3A4 . COSY and HMBC spectra were used to make full assignments of LAN and the LAN sulfone. NMR studies of RAB metabolism by the BM3 DM enzyme provided evidence for RAB substrate inhibition, since the proportions of products formed at higher RAB concentrations (50 μM and 100 μM) differed from those at 10 μM RAB (as also determined by LC–MS analysis). The major product formed at 100 μM RAB was the RAB thioether, whereas at 50 μM RAB the demethylated RAB was most prevalent. LC–MS analysis provided evidence for its further oxidative breakdown to the dealkylated thioether (results not shown), which is the major product observed at 10 μM RAB (Figure 5D). Full NMR spectral assignments in each case are given in the Supplementary Online Data (http://www.biochemj.org/bj/460/bj4600247add.htm). Assignments for RAB and its oxidation products were made by 1H, COSY and HMBC NMR (Supplementary Figures S12–S16 at http://www.biochemj.org/bj/460/bj4600247add.htm).
The structure of ESO bound to the DM BM3 haem domain was solved to 1.83 Å resolution (PDB code 4O4P), using molecular replacement with our previously determined DM haem domain OMP-bound structure (PDB code 4KEY) (Supplementary Table S1 http://www.biochemj.org/bj/460/bj4600247add.htm) . As observed for the OMP complex, the ESO-bound structure is highly similar to the WT fatty acid complex , despite the dissimilarity in ligand nature. An overlay of the DM ESO-bound active-site ligand density with the OMP-bound DM structure (PDB code 4KEY) is shown in Figure 6. The ESO-bound DM structure reveals that the sulfoxide oxygen is lost in the PPI, either due to synchrotron X-ray irradiation or, more probably, from the breakdown of ESO in the aqueous environment . Thus, in both DM haem domain structures, OMP/ESO are bound as the non-chiral thioether forms. This prevents us from gaining a clear structural rationale as to why OMP binds more tightly to DM BM3 than does ESO (OMP Kd=0.212 μM compared with 2.89 μM for ESO) , but yet is metabolized less selectively than ESO.
Overview of the S-omeprazole (ESO)-binding pocket in the DM BM3 haem domain structure
Production of drug metabolites is of interest to the biocatalysis industry as these compounds are often expensive and difficult to produce using conventional chemical methods . Most human drug metabolites are formed directly or indirectly by P450 enzymes. Understanding the pharmacology and toxicology of these metabolites is crucial to develop a full understanding of the mode of action of the parent drug, and to identify toxicity issues. The FDA requires that drug metabolites formed to significant levels in vivo are subjected to similar rigorous testing as the parent drug . However, production of sufficient amounts of drug metabolites for testing is challenging, particularly where regio- and/or stereo-selective drug oxidation by P450s occurs. Recombinant human P450s may generate small amounts of metabolites, but this may be impractical at a large scale due to poor enzyme stability, slow rates and requirement for an exogenous (CPR) redox partner. In contrast, P450 BM3 has the highest rates of substrate oxidation across the P450 superfamily and is a soluble catalytically self-sufficient enzyme. BM3 has been a test bed for P450 mutagenesis, and its activity profile has been substantially altered by protein engineering approaches [18,43]. Our previous studies showed that single mutations in the BM3 haem domain dramatically alter substrate selectivity to enable binding and oxidation of the PPI omeprazole, with the gatekeeper A82F mutation (in particular) altering conformational stability of BM3 to facilitate its diversification of substrate selectivity .
In the present study, we report the binding and oxidation of four other members of the PPI drug class by WT, A82F, F87V and F87V/A82F (DM) BM3 enzymes. ESO, LAN, PAN and RAB all bind to and/or are oxidized by the BM3 gatekeeper mutants. PPI binding induces a substrate-like type I haem absorbance shift for the A82F and DM BM3 enzymes, indicating a change in the P450 ferric haem iron equilibrium from LS towards HS. The F87V mutation improves PPI binding in each case, with Kd values consistently being lower for DM BM3 than for the A82F BM3 point mutant (Table 1). EPR spectroscopy reveals (through differences in the LS haem signal) major differences in the distal environment of the haem iron in the BM3 mutants compared with the WT enzyme. These probably reflect altered water environments around the sixth co-ordination position of the iron that arise as a consequence of structural changes in the immediate haem environment (induced by the F87V mutation) and through altered conformational dynamics in the haem domain (in the A82F and DM enzymes). It is important to note that the addition of DMSO as the carrier for certain PPI drugs also influences the BM3 LS EPR spectrum, again probably influencing the solvation state around the distal haem iron site (Figure 4). X-ray crystallographic studies of WT and F87A mutants of the BM3 haem domain indicated that co-crystallization of the mutant with DMSO at 28% (v/v) resulted in a direct interaction of DMSO with the haem iron. However, the DMSO concentration used for EPR in the present study is substantially lower (0.72%) .
Despite limited PPI-induced HS haem development, the F87V BM3 enzyme proved an effective catalyst for oxidation of each of the PPI drugs, albeit with lower catalytic efficiency than DM BM3 (Table 1). Steady-state kinetic analysis of the BM3 gatekeeper mutants indicated PPI substrate-dependent NADPH oxidation at rates up to ~50% of those for turnover of the lipid substrate NPG. However, PPI product formation was more extensive for ESO and RAB (and for the racemate OMP) than for LAN and PAN, indicating that electron transfer was less efficiently coupled to substrate oxidation for LAN/PAN . The altered conformational flexibility of the A82F-containing BM3 enzymes underpins their ability to bind the different PPI drugs, with the F87V mutation altering accessibility to the haem active site and playing an important function in enhancing enzymatic efficiency in most cases. However, the F87V mutation alone is sufficient to facilitate substantial improvements in oxidation of certain PPIs, and particularly in the case of ESO, where product formed in unit time is similar to that from DM BM3.
Although structural similarity in the PPI drug ‘core’ (Figure 1) in part explains the ability of diverse PPIs to bind the different BM3 gatekeeper mutants, altered modes of binding for the tested PPIs is also evident from the different positions at which oxidation occurs on these molecules (Figure 5). Importantly, our data reveal that in many cases the PPI products formed by BM3 mutants replicate the major metabolites formed by their human P450 counterparts with the same drugs (Figure 7). More specifically, ESO is transformed by BM3 (A82F, F87V and DM) to 5-OH ESO, which is the major human metabolite produced by CYP2C19 . The data for ESO differ from those for the oxidation of the racemate OMP by the BM3 mutants, since there is negligible formation of the 5-COOH ESO product. These data suggest that the primary 5-OH ESO (5-OH S-OMP) metabolite does not bind productively to the BM3 variants, whereas the 5-OH R-OMP component from the racemate may be able to do so to enable formation of the 5-COOH OMP product. These results further suggest that, as with the human P450s, OMP metabolism is specific for each enantiomer . The major metabolites seen in humans for LAN are the sulfone and the 5-OH LAN, which are produced by CYP3A4 and CYP2C19 respectively . Our data show the primary product with our mutant BM3 enzymes (with up to 33% conversion by the DM) is the LAN sulfone. In the case of PAN, the extent of product formation is the lowest among the PPIs tested, but the product is exclusively the PAN N-oxide. In this case, the enzymatic reaction (catalysed most effectively by the DM BM3) produces a metabolite that is also observed in lower amounts as a by-product of PAN synthesis. The major pathway for RAB metabolism is its non-enzymatic (reductive) conversion into the thioether, with up to 50% conversion in aqueous solution in 1 h [47,48]. This product is also observed in the present study, and control reactions showed that BM3 enzymes alone had little effect on the extent of this reaction, but that the inclusion of an NADPH cofactor regeneration system approximately doubled the amount of thioether product formed, probably due to providing a more reducing environment for the reaction. The other major human P450 metabolites of RAB are the demethylated and sulfone forms, as produced by CYP2C19 and CYP3A4 respectively . Using the BM3 DM enzyme, the RAB product profile showed dependence on substrate concentration. At 100 μM RAB the thioether was the major product observed, whereas lowering the RAB concentration favoured the demethylated RAB (at 50 μM) or a novel metabolite involving the loss of the entire ether chain, producing a dealkylated thioether of RAB (the demethoxypropyl RAB thioether at 10 μM RAB). Substrate inhibition and substrate concentration dependency on product outcome is frequently seen in human P450 reactions, e.g. in the case of CYP2C9-catalysed methyl-hydroxylation of the non-steroidal anti-inflammatory drug celecoxib and for the CYP2D6-catalysed O-demethylation of the antitussive dextromethorphan .
Reactions schemes outlining pathways of P450 metabolism of PPI drugs
In conclusion, the results of the present study demonstrate that relatively limited mutagenesis of P450 BM3 can generate considerable diversity in substrate selectivity, with the utility of the BM3 gatekeeper mutants A82F, F87V and A82F/F87V now expanded to their oxidation of a range of PPI drugs in clinical use. More importantly, in most cases the products formed from these PPIs are the same as those generated by the major human drug-metabolizing P450s, notably CYP3A4 and CYP2C19. By lowering the energetic barrier to conformational reorganization and transition to a substrate-bound state, the mutations facilitate recognition of diverse PPI substrates that are disfavoured by WT BM3. The diversity in substrate recognition and oxidation among the major human hepatic drug-metabolizing P450s is considered crucial for the ability of a small number of such enzymes to metabolize and detoxify a wide range of xenobiotics . To achieve this end, evolutionary pressure probably selected conformationally flexible P450 variants able to accommodate molecules of diverse size and chemical character. The ability to radically change the substrate selectivity of P450 BM3 with point mutations that are structurally destabilizing in addition to altering the active site environment thus points to routes by which the hepatic P450s may also have evolved efficiently to deal with a plethora of environmental toxins.
heteronuclear multiple bond correlation
proton pump inhibitor
Christopher Butler, Caroline Peet, Kirsty McLean, David Leys, Michael Voice and Andrew Munro conceived and designed the experiments. Christopher Butler, Caroline Peet, Kirsty McLean, Richard Blankley, Karl Fisher, Stephen Rigby, David Leys and Michael Voice performed the experiments. Christopher Butler, Caroline Peet, Kirsty McLean, Michael Baynham, Richard Blankley, Stephen Rigby, David Leys, Michael Voice and Andrew Munro analysed the data. Christopher Butler and Andrew Munro wrote the paper. Kirsty McLean, Stephen Rigby and David Leys contributed to writing the paper.
We thank Dr Robert Šardzík for helpful discussions on NMR analysis, and Dr Colin Levy for assistance with synchrotron X-ray data collection.
The work was supported by the UK Biotechnology and Biological Sciences Research Council (BBSRC) [grant number BB/K001884/1 (to A.W.M./D.L.)] and an Industrial CASE studentship (BB/G01698/1) with Cypex Ltd to A.W.M./M.W.V., supporting C.F.B.
The structural co-ordinates reported for ESO bound to the DM BM3 haem domain have been deposited in the PDB under code 4O4P.