Articles pubs.acs.org/acschemicalbiology
Biocatalytic Characterization of Human FMO5: Unearthing Baeyer− Villiger Reactions in Humans Filippo Fiorentini,†,‡ Martina Geier,‡ Claudia Binda,† Margit Winkler,‡ Kurt Faber,§ Mélanie Hall,*,§ and Andrea Mattevi*,† †
Department of Biology and Biotechnology, University of Pavia, via Ferrata 9, 27100 Pavia, Italy Austrian Centre of Industrial Biotechnology, c/o Institute of Molecular Biotechnology, Graz University of Technology, Petersgasse 14, 8010 Graz, Austria § Department of Chemistry, University of Graz, Heinrichstrasse 28, 8010 Graz, Austria ‡
S Supporting Information *
ABSTRACT: Flavin-containing mono-oxygenases are known as potent drugmetabolizing enzymes, providing complementary functions to the well-investigated cytochrome P450 mono-oxygenases. While human FMO isoforms are typically involved in the oxidation of soft nucleophiles, the biocatalytic activity of human FMO5 (along its physiological role) has long remained unexplored. In this study, we demonstrate the atypical in vitro activity of human FMO5 as a Baeyer−Villiger monooxygenase on a broad range of substrates, revealing the first example to date of a human protein catalyzing such reactions. The isolated and purified protein was active on diverse carbonyl compounds, whereas soft nucleophiles were mostly non- or poorly reactive. The absence of the typical characteristic sequence motifs sets human FMO5 apart from all characterized Baeyer−Villiger mono-oxygenases so far. These findings open new perspectives in human oxidative metabolism.
D
including drugs, pesticides, and dietary components, ranging from rather small molecules, such as trimethylamine and dimethyl sulfoxide, to bulkier and pharmacologically active compounds, like benzydamine and moclobemide.3,9,10 FMO5, on the contrary, proved to be non- or poorly active on such classical substrates.7,11 This, combined with its secluded gene location (outside the 220 kb cluster on chromosome 1 hosting all other FMOs12), led to the description of this isoform as pseudoactive FMO. It is however known that FMO5, besides showing good expression in the small intestine, kidney, and lung, accounts for more than 50% of total FMO transcripts in the human liver.8 Aiming at revealing unknown biocatalytic properties of this long neglected isoform, we first set up an effective and reliable protocol for expression, detergent-mediated extraction, and purification of human full-length FMO5 (hFMO5) in a stable and active form. We employed thermal-stability and gelfiltration analysis to identify optimal conditions for maximal stability of the hFMO5-detergent complex and, accordingly, to study its activity toward NADPH and oxygen via analysis of its FAD spectral behavior. Most relevant, the availability of purified hFMO5 enabled the performance of extensive substrate profiling, which, besides confirming poor acceptance of classical FMO substrates, clearly indicated a strong in vitro oxygenating
etoxification and metabolism of potentially harmful xenobiotics in humans generally take place via the oxidative action of two finely evolved enzyme families: cytochrome P450 mono-oxygenases and flavin-containing mono-oxygenases (FMOs).1 Though long overlooked, the number of oxidative reactions specifically attributed to FMOs in the human body is quickly increasing.2 By directly modifying bioactivity and bioavailability of multiple and diverse exogenous compounds, FMOs have been gradually framed as main players in oxidative phase I metabolism of xenobiotics.2,3 Moreover, the firmly emerging relationship between a number of metabolites specifically produced by FMOs (e.g., trimethylamine N-oxide) and the control of key biological processes (health and longevity,4 atherosclerosis,5 and cholesterol regulation6) highlight unexpected physiological functions of these enzymes. Five FMO isoforms are expressed in humans (FMO1−5). They are all associated with the endoplasmic reticulum membranes through a C-terminal trans-membrane α-helix and share as much as ca. 60% sequence identity with each other. Human FMOs exhibit different tissue expression profiles among main organs involved in xenobiotic metabolism (liver, kidney, small intestine, lungs, and brain),7,8 hinting at possible differences in their substrate preferences. They, however, all share their dependence on NADPH as an external electron source, which is required to reduce the FAD cofactor and in turn allows oxygen activation. So far, human FMOs have mostly been found active in the oxidation of compounds containing soft nucleophilic heteroatoms, mainly nitrogen and sulfur. “Typical” FMO substrates have been gradually identified, © XXXX American Chemical Society
Received: December 10, 2015 Accepted: January 15, 2016
A
DOI: 10.1021/acschembio.5b01016 ACS Chem. Biol. XXXX, XXX, XXX−XXX
Articles
ACS Chemical Biology
Figure 1. hFMO5 expression and purification. (A) Size-exclusion chromatography shows elution occurring well after the column void volume (7 mL), excluding that hFMO5 is purified in an aggregated or misfolded state. Moreover, the sharpness and monodispersity of the peak prove the sample homogeneity (black line = Abs280; dark-gray line = Abs370; light-gray line = Abs450). (B) Native- (left) and SDS-PAGE (right) analysis further confirmed the lack of aggregation/misfolding and the homogeneity of the purified enzyme (hFMO5 = 60 kDa). (C) The UV/visible absorbance spectrum of the fully oxidized TX100R-purified enzyme (solid line) exhibits maxima at 274, 375, and 454 nm and a Abs274/Abs454 ratio of 8.7. The extinction coefficient of FAD bound to hFMO5 (ε454 = 12.050 mM−1 cm−1) was determined by recording the spectrum after denaturation of the protein (dotted line).
and induction conditions for heterologous expression was ascertained by extracting and assaying hFMO5 from the membrane fraction to which it is associated (primarily through a C-terminal trans-membrane α-helix3). All mammalian FMOs indeed have a strong membrane association and poor water solubility.3 The nonionic detergent Triton X-100 Reduced (TX100R) proved to be the most efficient for extracting the protein from E. coli isolated membranes and one of the most effective in promoting protein stability (Tm = 47 °C), as determined by thermal-shift assays17 (Figures 1 and 2, Table 1). Accordingly, this allowed us to develop an effective protocol for hFMO5-TX100R purification (Figure 1A,B). Importantly, the UV/visible absorbance spectra of all purified protein samples showed proper retention of FAD (Abs274/Abs454 ratio of 8.7), further confirming correct folding and cofactor incorporation by the recombinant enzyme (Figure 1C).18 It is noteworthy that this protocol enabled the successful heterologous expression of a human membrane-associated protein in a bacterial host and its purification in milligram amounts (yield ∼2 mg per liter of cell culture).
activity toward a broad set of carbonyl compoundsfrom aliphatic and cyclic ketones to aldehydesresulting in Baeyer− Villiger oxidation reactions. These findings, in line with Baeyer−Villiger reactions already observed among Rhodococcus mono-oxygenases bearing the typical FMO fingerprint motif,13,14 expand the recent data about hFMO5 participation in the metabolism of two pharmaceuticals bearing carbonyl functional groups.15,16 Taken together, our data demonstrate the not yet considered existence of a widely active Baeyer− Villiger mono-oxygenase in humans, suggesting a possible yet unknown detoxification route in human metabolism.
■
RESULTS FMO5 Expression and Purification. Access to homogeneous and pure protein samples was crucial to allowing proper and thorough characterization of hFMO5 catalytic properties. With this aim, the full-length human FMO5 gene was cloned into a vector suitable for expression in E. coli, designed to fuse a cleavable affinity-stabilization tag (8xHis-SUMO) to the Nterminal portion of the enzyme. Identification of optimal strain B
DOI: 10.1021/acschembio.5b01016 ACS Chem. Biol. XXXX, XXX, XXX−XXX
Articles
ACS Chemical Biology
Figure 2. hFMO5-detergent stabilization assays. (A, B) ThermoFAD screenings identified the buffer pH (8.5) and ionic strength (0 mM KCl) offering the best stability to the hFMO5-detergent complex. (C) hFMO5 shows different detergent-dependent elution profiles from gel-filtration analysis (hFMO5-DDM = orange line; hFMO5-C6NG = red line; hFMO5-LDAO = violet line; hFMO5-TX100R = black line; hFMO5-ZW3-12 = blue line).
Table 1. Detergent-dependent Stability of hFMO5a detergents nonionic surfactants Triton X-100 Reduced (TX100R) Hega-9 Cymal 6 Neopentyl Glycol (C6NG) heptyl thioglucoside octyl β-glucopyranoside octanoyl sucrose nonyl β-glucoside dodecyl β-D-maltopyranoside (DDM) zwitterionic surfactants foscholine 8 foscholine 10 foscholine 12 dimethyl decyl phosphine oxide dimethyl decyl amine oxide dimethyl dodecyl amine oxide (LDAO) Zwittergent 3-12 (ZW3-12)
speculation that FMOs may exist in the cells predominantly in this activated form, ready to react with suitable substrates, as documented in early works on pig liver FMO20−23 as well as in the more recent characterization of a bacterial FMO.24,25 With hFMO5, upon the addition of NADPH (from equimolar to a > 10-fold molar excess) to the protein solution (6 μM) in aerated conditions, the UV/visible absorbance spectrum showed parallel oxidation of NADPH (bleaching at 340 nm) and reduction of hFMO5-bound flavin (bleaching at 450 nm; Figure 3A). However, at each time lapse recorded (5 s), the ratio between NADPH oxidized and hFMO5 reduced was ≫1, suggesting that hFMO5 is unloading large parts of the reducing equivalents acquired from NADPH to oxygen to produce hydrogen peroxide. As a result of this behavior, a large molar excess of NADPH was needed to observe substantial reduction of hFMO5 under aerobic conditions. Accordingly, in the course of these experiments, we could never observe spectral features consistent with the formation of a stable (hydro)peroxy-FAD intermediate. Hence, in the absence of substrates, hFMO5 showed relatively slow but yet significant NADPH oxidase activity, releasing hydrogen peroxide (KM,NADPH = 59 ± 9 μM, kcat = 11.84 ± 0.56 min−1 measured in steady-state conditions; Figure 3B). These properties support the notion of hFMO5 displaying rather distinct catalytic properties from that of the other four human enzyme isoforms.11,26 FMO5 Substrate Profiling. The employment of a pure sample is of crucial importance for studying the enzymatic and substrate-specificity properties of uncharacterized enzymatic systems. Though generally providing good enzyme stability and
Tm (°C) 47 34.5 43 n.d. 36 34 41.5 48.5 n.d. n.d. n.d. 35.5 39.5 41 n.d.
a
The table reports the Tm of hFMO5 in the presence of different detergents determined by the ThermoFAD technique (n.d. = not determined due to precipitation/misfolding).
FMO5 Spectrophotometric Characterization. FMOs are class B flavoprotein mono-oxygenases19 that can generate a stable C4a-(hydro)peroxyflavin intermediate upon reaction of NADPH-reduced flavin with oxygen. The ability to stabilize this crucial intermediate even for several minutes has led to C
DOI: 10.1021/acschembio.5b01016 ACS Chem. Biol. XXXX, XXX, XXX−XXX
Articles
ACS Chemical Biology
Figure 3. hFMO5 reactivity with NADPH. (A) Upon molar excess addition of NADPH (80 μM) to the protein solution (6 μM) under aerated conditions, the UV/visible absorbance spectrum indicates oxidation of NADPH (bleaching at 340 nm) and partial reduction of hFMO5 (bleaching at 450 nm). The spectrum does not show features (above all a clear absorbance peak in the 360−380 nm range) consistent with the formation of a stable flavin-peroxide intermediate. This accounts for the high rate of uncoupling producing hydrogen peroxide. Using lower NADPH concentrations did not change the outcome of the experiment but simply resulted in a lower extent of reduced enzyme accumulation. Likewise, prior incubation of NADP+ did not alter the outcome of the experiment. (B) The plot shows the apparent kcat values as a function of NADPH concentrations (10−450 μM). Such NADPH-oxidase activity of hFMO5 shows Michaelis−Menten behavior.
outline for the first time the substrate profile properties of hFMO5 (Figure 4 and Table 2). Linear aliphatic ketones turned out to be excellent substrates, with a slight preference for longer asymmetric ketones (maximum 79% conversion with 6-methylhept-5-en-2-one (2a), 49% with heptan-4-one (6a)). The degree of regioselectivity in the enzymatic conversions varied, depending on the substrate. For instance, 3-methyl-pent-3-en-2-one (8a) was selectively converted to the acetate, which is not surprising, as only few Baeyer−Villiger mono-oxygenases (BVMOs) promote the least favored migration of the methyl group in the Criegee rearrangement.31,32 Conversely, octan-3-one (4a) and hexan-3-one (5a) yielded mixtures of normal and abnormal Baeyer−Villiger oxidation products (1:1.2 and 1:1.3, respectively). Furthermore, also alicyclic ketones were thoroughly studied to provide a benchmark against prototypical BVMOs (e.g., cyclohexanone mono-oxygenase and cyclopentanone mono-oxygenase). Cyclopentanone and cyclohexanone derivatives were equally well accepted with minor influence of the substitution type (α- vs β-, alkyl- vs phenyl-) on their reactivity (conversion levels from 10 to 44%). Of notice among this series of substrates, hFMO5 displayed pronounced enantioselectivity on 2-methylcyclohexanone (rac-12a, E ratio = 21). Mixed regioselectivity on cyclopentanone derivatives was observed, with formation of the abnormal lactone as a major product with β-substituted analog rac-18a. Independent of the regioselectivity, the minor product was always formed with exquisite enantioselectivity (17c and 18b). Among the aromatic ketones, an excellent (and highest) conversion level was registered with phenylacetone (9a, 85% conversion), a typical substrate of BVMOs. A modest 10% conversion was obtained with the more challenging β-ionone (10a), yielding the corresponding acetate product. Nevertheless, this reaction is notable because BVMO-catalyzed oxygenation of α,β-unsaturated ketones with oxygen insertion near the sp2 C atom is not common and has
efficiency, the previously utilized recombinant FMO5 supersomes, pooled liver S9 fractions, human liver cytosolic fractions, and microsomes,15,16 all suffer from unpredictable sidereactions, inevitably affecting the unbiased determination of specific activities. Catalytic properties of hFMO5 were investigated using a NADPH-recycling system based on glucose/glucose dehydrogenase under experimental conditions that promoted maximal enzyme stability as indicated by enzymatic activity and thermal shift assays (Figure 2, Table 1). Given the rather high thermal stability of hFMO5, all samples could be analyzed for overnight product formation by GC or HPLC and GC-MS methodologies (see the Supporting Information and Table 2). The first issue was to confirm that the enzyme is poorly active on “classic” FMO substrates.10,11,27,28 We found that trimethylamine is indeed not converted by hFMO5. Likewise, hFMO5 showed no activity on phosphorus-containing compounds such as trimethyl- and triphenyl-phosphine. Moderate conversion (20%) was obtained with thioanisole 1a (forming the corresponding sulfoxide), thus indicating some degree of activity on soft sulfur nucleophiles, in agreement with the previously reported sulfoxidation of S-methyl-esonarimod29 and fenthion30 by FMO5. We next screened other types of functionally diverse molecules susceptible to oxidation/oxygenation reactions, i.e., compounds bearing various degrees of unsaturation (including aromatics) and/or carbonyl groups. Among them, hFMO5 showed prominent and selective activity as Baeyer−Villiger mono-oxygenase, catalyzing the oxygenation of both aliphatic and cyclic ketones to the corresponding esters and lactones, respectively. No activity could be detected on uracil, N,N′-dimethyl propylene urea, δ-valerolactam, 2pyrrolidinone, 4-chloro-α-methylstyrene, benzyl alcohol, cinnamyl alcohol, or trans-anethole (see Supporting Information Table 1). In light of these first hits, we started to rationally D
DOI: 10.1021/acschembio.5b01016 ACS Chem. Biol. XXXX, XXX, XXX−XXX
Articles
ACS Chemical Biology Table 2. Conversion of Reactive Substrates in the Oxidation Reaction Catalyzed by Purified hFMO5a
Reaction conditions: 5 mM substrate, 5 mM NADP+, 4 μM TX100R-hFMO5, 20 mM glucose, 10 U GDH, 50 mM Tris-HCl pH 8.5, 10 mM KCl, 0.05% (v/v) TX100R (total volume 0.4 mL); reactions run overnight at 30 °C and 120 rpm. n.a.: not applicable. n.d.: not determined. bTool for calculation of enantiomeric ratio (E)52 available (open access) at http://biocatalysis.uni-graz.at/enantio/cgi-bin/enantio.plConversion calculated from [eesubstrate/(eesubstrate+eeproduct)]. cRatio of products. dAbsolute configuration not determined. eConversion of pure (R)-13a: 25%. fAbsolute configuration of second regioisomer not determined (
