Revealing the moonlighting role of NADP in thestructure of a flavin-containing monooxygenaseAndrea Alfieri*, Enrico Malito*†, Roberto Orru*, Marco W. Fraaije‡, and Andrea Mattevi*§

*Department of Genetics and Microbiology, University of Pavia, Via Ferrata 1, 27100 Pavia, Italy; and ‡Laboratory of Biochemistry, Groningen BiomolecularSciences and Biotechnology Institute, University of Groningen, Nijenborgh 4, 9747 AG, Groningen, The Netherlands

Edited by David P. Ballou, University of Michigan, Ann Arbor, MI, and accepted by the Editorial Board March 7, 2008 (received for review January 28, 2008)

Flavin-containing monooxygenases (FMOs) are, after cytochromesP450, the most important monooxygenase system in humans andare involved in xenobiotics metabolism and variability in drugresponse. The x-ray structure of a soluble prokaryotic FMO fromMethylophaga sp. strain SK1 has been solved at 2.6-Å resolutionand is now the protein of known structure with the highestsequence similarity to human FMOs. The structure possesses atwo-domain architecture, with both FAD and NADP� well definedby the electron density maps. Biochemical analysis shows that theprokaryotic enzyme shares many functional properties with mam-malian FMOs, including substrate specificity and the ability tostabilize the hydroperoxyflavin intermediate that is crucial insubstrate oxygenation. On the basis of their location in the struc-ture, the nicotinamide ring and the adjacent ribose of NADP� turnout to be an integral part of the catalytic site being activelyengaged in the stabilization of the oxygenating intermediate. Thisfeature suggests that NADP(H) has a moonlighting role, in that itadopts two binding modes that allow it to function in both flavinreduction and oxygen reactivity modulation, respectively. Wehypothesize that a relative domain rotation is needed to bringNADP(H) to these distinct positions inside the active site. Localiza-tion of mutations in human FMO3 that are known to causetrimethylaminuria (fish-odor syndrome) in the elucidated FMOstructure provides a structural explanation for their biologicaleffects.

drug metabolism � trimethylaminuria � oxygen

Two effective families of enzyme systems have evolved ineukaryotes to oxygenate xenobiotics: cytochromes P450 and

flavin-containing monooxygenases (FMOs). They both helporganisms to deal with potentially toxic exogenous compoundsfrom natural sources and, most notably in humans, represent theenzyme systems involved in the metabolism of drugs and pol-lutants (1). FMOs belong to the flavoenzyme class of single-component flavoprotein monooxygenases (2); they use equiva-lents from NADPH to reduce the FAD cofactor, which, in turn,becomes capable of reacting with molecular oxygen to yield theC4a-hydroperoxyFAD intermediate (Fig. 1A). This intermediateis responsible of the insertion of one oxygen atom into asubstrate, with the production of the second intermediate C4a-hydroxyFAD. The loss of the second oxygen atom from theflavin in the form of a molecule of water restores the oxidizedcofactor (3). The first FMO isolated from pig liver was describedas liver microsomal FAD-containing monooxygenase (4) andthoroughly characterized in its kinetic properties (5–8). FMOsact on a variety of nonpolar substrates that contain soft nucleo-phile heteroatoms (9). The most remarkable feature of theircatalytic cycle is their ability to stabilize the crucial C4a-hydroperoxyFAD intermediate until a substrate gains access tothe active site (10). A prominent aspect is that the presence ofNADP� is essential for intermediate stabilization and, consis-tently, NADP� remains bound to the enzyme throughout thecatalytic cycle, being the last product to be released (Fig. 1 A) (6).

The human FMO (hFMO) gene family comprises five genesand some pseudogenes. Of the five functional isoforms, hFMO3

is the one most abundantly expressed in adult liver and largelyinvolved in drug metabolism (11, 12). All isoforms display �60%sequence identity, a moderate level of polymorphism (13–15),and the presence of tissue- and age-specific splice variants (16).Several mutations in hFMO3 have been reported to cause theinheritable disease known as trimethylaminuria (TMAU) or‘‘fish-odor syndrome,’’ because of accumulation of substratetrimethylamine in body fluids (17). All mammalian FMOs havea strong membrane association and poor water solubility (9), andno structure of a mammalian FMO is currently available.

In recent years, several plant (18) and prokaryotic enzymeshave been annotated as FMOs. The first discovered bacterialFMO was identified in the methylotrophic bacterium Methyl-ophaga sp. strain SK1 (mFMO) for the purpose of biocatalyticproduction of the blue dye indigo (19). mFMO is not membrane-associated. Its sequence (456 aa) scores �31% identity withhFMO3 and contains two dinucleotide-binding signatures(Rossmann folds) for FAD and NADP and the FMO-identifyingmotif FXGXXXHXXX(Y/F) (20). mFMO shares 27% sequenceidentity with Schizosaccharomyces pombe FMO, whose crystalstructure has been recently reported (21). The sequences ofMethylophaga, human, and S. pombe FMOs are aligned (22, 23)in Fig. 1B.

Here, we present the biochemical characterization of mFMOwith kinetic data on a number of substrates and the 2.6-Åresolution structure obtained by x-ray crystallography. In addi-tion to the interest raised by the homology with the hFMOs, thecrystal structure of the bacterial enzyme highlights the directinvolvement of NADP� in the stabilization of the crucialC4a-hydroperoxyFAD intermediate, a unique feature of theseenzymes.

ResultsBiochemical Properties. mFMO is a 2 � 53-kDa dimeric proteincontaining one molecule of noncovalently tightly bound FADper monomer. The spectrum of the fully oxidized enzymeexhibits typical maxima at 273, 372, and 442 nm, a shoulder at462 nm, and a ratio Abs273/Abs442 of 9.9 (Fig. 2, trace A).Addition of NADPH rapidly bleaches the protein solution,accounting for an efficient reduction of the flavin cofactor. Likecommonly observed in mammalian enzymes (24), in the absence

Author contributions: A.A., M.W.F., and A.M. designed research; A.A., E.M., and R.O.performed research; A.A., E.M., R.O., M.W.F., and A.M. analyzed data; and A.A. and A.M.wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission. D.P.B. is a guest editor invited by the EditorialBoard.

Data deposition: The atomic coordinates have been deposited in the Protein Data Bank,www.pdb.org (PDB ID codes 2vq7 and 2vqb).

†Present address: Ben May Department for Cancer Research, The University of Chicago, 929East 57th Street, W423J Chicago, IL 60637.

§To whom correspondence should be addressed. E-mail: mattevi@ipvgen.unipv.it.

This article contains supporting information online at www.pnas.org/cgi/content/full/0800859105/DCSupplemental.

© 2008 by The National Academy of Sciences of the USA

6572–6577 � PNAS � May 6, 2008 � vol. 105 � no. 18 www.pnas.org�cgi�doi�10.1073�pnas.0800859105

Dow

nloa

ded

by g

uest

on

June

24,

202

0

of an organic substrate, mFMO shows a slight NADPH oxidaseactivity (Km,NADPH � 13 �M, kcat � 0.06 s�1 measured insteady-state conditions) producing hydrogen peroxide (Fig. 1 A).

The ability to stabilize the C4a-hydroperoxyflavin intermedi-ate, well documented in the early works on pig FMO (4, 6, 7), canalso be demonstrated in mFMO. Mixing the enzyme with anequimolar amount of aerated NADPH (in the absence of anorganic substrate) leads to the immediate formation of a C4a-hydroperoxyflavin intermediate characterized by the typicalabsorbance spectrum with a peak at �360 nm (refs. 6, 25–27; Fig.2, trace B). Over �30 min, the spectrum slowly reacquires thecharacteristic 442-nm peak of the oxidized flavin (Fig. 2, tracesC to G). Conversely, addition of excess substrate immediatelybrings the spectrum back to that of the fully oxidized flavin.

Steady-state kinetic assays performed at a fixed concentrationof NADPH with typical FMO substrates showed Michaelis–Menten behavior (Table 1). Substrates were selected on the basisof the following considerations: trimethylamine is a commonendogenous substrate for hFMO3; methimazole and (S)-(-)-nicotine are of pharmacological relevance (9); methimazole andN,N-dimethylaniline has been extensively exploited to test mam-malian FMO activity in literature; indole is the putative substrateof mFMO in the pathway for the production of indigo (19). Allthese compounds are good substrates, with trimethylamineexhibiting the highest catalytic efficiency (Table 1). An enzy-matic activity has also been unexpectedly appreciated on di-methyl sulfoxide initially used for dissolving some substrates,confirming that mFMO accepts a broad range of N- andS-containing molecules largely overlapping with the substrates of

the first Rossmann fold for FAD binding, the fingerprint sequence for FMOs(20), and the second Rossmann fold for NADP binding, respectively. The greenasterisks mark residues that are part of the active site, whereas the redasterisks indicate residues that are mutated in patients affected by TMAU. Thealignment was performed with ClustalW2 (22) and ESPript (23).

Fig. 1. Biochemical properties of FMOs. (A) Scheme of the catalytic cycle.NADP� that forms upon reduction of the flavin cofactor stays bound through-out the reaction and is essential for stabilization of C4a-hydroperoxyFAD(Bottom Right) (3–6). Enzyme-bound NADP� prevents FMO from working asan NADPH oxidase, which would produce H2O2 (dashed line). S and S-Oindicate substrate and monooxygenated product, respectively. (B) Alignmentof the sequences of Methylophaga FMO (MeFMO), Homo sapiens FMO3isoform 1 (hFMO3), and S. pombe FMO (SpFMO). Sequence identities tohFMO3 are 31% for mFMO and 27% for S. pombe FMO. The red bars indicate

Fig. 2. Absorbance spectra of mFMO. Trace A represents the fully oxidizedform in the absence of NADP�. Trace B was obtained as described in Materialsand Methods by the addition of an equimolar amount of NADPH underaerobic conditions and is consistent with formation of the C4a-hydroperoxy-flavin intermediate. The presence of a small amount of oxidized enzymebecause of the intrinsic NADPH-oxidase activity accounts for the shoulderobserved at �450 nm. Traces C to G were sequentially recorded after 25 sec,90 sec, 3 min, 10 min, and 18 min and show the slow decay of the C4a-hydroperoxyflavin intermediate, which is completed after 30 min.

Alfieri et al. PNAS � May 6, 2008 � vol. 105 � no. 18 � 6573

BIO

CHEM

ISTR

Y

Dow

nloa

ded

by g

uest

on

June

24,

202

0

mammalian FMOs. Moreover, as observed in steady-state ki-netics of mammalian enzymes (4), NADP� was competitive (Kiof 30 �M) with NADPH. Taken together, these observationssuggest that, similar to the mammalian FMOs, the bacterialenzyme behaves like a ‘‘cocked gun,’’ spending most of the timein the form of the hydroperoxyflavin intermediate. Any het-eroatom-containing compound gaining access to the stabilizedintermediate is monooxygenated, in accordance with the strikingsubstrate promiscuity of FMOs.

Overall Three-Dimensional Structure. Wild-type mFMO crystal-lized readily, but all crystals invariably suffered from severetwinning. This problem was addressed by exploiting a number oftechniques, among which surface mutagenesis proved to besuccessful (28). We targeted sets of two/three consecutivecharged residues (mainly Glu and Lys) supposed to be solvent-exposed based on alignment with homologous proteins [sup-porting information (SI) Table S1]. The mutant E158A/E159Agave crystals that turned out to be free from twinning (themutant has biochemical properties virtually identical to those ofwild type; Table 1). Addition of NADP� was essential for crystalgrowth. mFMO structure was solved by single-wavelength anom-alous dispersion method at 2.6-Å resolution (Tables 2 and 3) (29,30). Protein molecules are arranged to give two identical ho-modimers (Fig. 3) in the asymmetric unit, which is consistentwith the dimeric form determined by size-exclusion chromatog-raphy for both wild-type and E158A/E159A mFMOs. However,each of the active sites is exclusively built up by residuesbelonging to a single subunit. rmsd between equivalent C� atomsof the monomers are �0.1 Å.

The protein is made up of two distinct domains (Fig. 3A), alarger FAD-binding domain (residues 1–169 and 281–461) anda smaller NADP-binding domain (residues 170–280). Bothdomains display a typical dinucleotide-binding fold and areconnected through a double linker (around residues 170 and

280) that may act as a flexible hinge (Fig. 3A). There are threelong stretches interposed between the two domains; they lacksecondary structure elements (residues 44–80, 166–186, 276–306), suggesting the absence of any rigid block opposing thereciprocal movement of the domains. A relatively similar overallconformation (rmsd of 2.3 Å for 343 C� atoms) has beenobserved in the structure of S. pombe FMO in complex withNADP� (21).

The Active Site. Inspection of the active site reveals the existence ofa cleft leading to the NADP� molecule at the interface between thetwo protein domains (Fig. 4A). NADP� occupies the cleft, blockingthe access to the flavin and the catalytic core. The nicotinamide ringis sandwiched between the re side of the flavin and the single �-helixof the NADP-binding domain and is roughly coplanar with isoallox-azine at a distance of 3.0–4.0 Å. Importantly, the nicotinamideorients its reactive C4 away from the flavin N5 (C4-N5 distance of5.1 Å) so that the NADP� amide group is within hydrogen-bonddistances from the flavin O4 and N5. These potential hydrogen-bond interactions led us to tentatively assign the orientation of theNADP� carboxamide group as shown in Fig. 4B. Spanning inproximity of the flavin ring, a 161-Å3 cavity was detected by theprogram VOIDOO (31) by using a 1.4-Å radius sphere as probe(Fig. 4C). Its limits are defined by the isoalloxazine, the NADP�

ribose and several protein residues (Asn-78, Gly-79, Pro-80, Cys-83,Leu-84, Phe-402, Trp-405, Gln-323, Trp-324, Tyr-325, Ser-326,Tyr-212). The role of Tyr-212 is remarkable, because its side chaintogether with the ribose of NADP� shields the cavity from theoutside, forming the only barrier to the solvent (Fig. 4).

After obtaining the native dataset at 2.6-Å resolution, severalattempts were made to bind substrate molecules inside thecrystals by either soaking or cocrystallization but none of themwere successful. Nevertheless, in the electron density map cal-culated from soaked crystals (we refer to the experiment withN,N-dimethylaniline hereafter named ‘‘dataset from soaking;’’Tables 2 and 3), a well defined ellipsoidal peak of density wasfound in proximity to the catalytic site inside the above-describedcavity, halfway between Tyr-212 and the pyrimidine ring of theflavin (Fig. 4C). Alternative refinements with several candidatemolecules and atoms only yielded reasonable B-factors whendioxygen or monoatomic ions of comparable electron densitywere fitted inside. Based on these observations, we assigned theorphan electron densities to dioxygen, which is included in thecoordinate file deposited in the Protein Data Bank (32). How-ever, we do not rule out that, instead of molecular oxygen, asuperoxide molecule, possibly generated upon the x-ray expo-sure of the crystal, is bound at this position. Likewise, thepresence of a monoatomic ion (especially if positively charged,

Table 1. Data from steady-state kinetic assays on mFMO

Substrate Km, �M kcat, s�1 kcat/Km, M�1 s�1

Trimethylamine 7.3 � 0.6 6.1 � 0.1 8.4 � 105 � 0.7Trimethylamine

(E158A/E159A)7.1 � 1 6.4 � 0.3 9.0 � 105 � 1

Nicotine 130 � 30 3.0 � 0.2 2.3 � 104 � 0.6Methimazole 66 � 7 1.0 � 0.02 1.5 � 104 � 0.2N,N dimethylaniline 232 � 28 1.8 � 0.4 7.8 � 103 � 2Indole 90 � 14 0.7 � 0.03 7.8 � 103 � 1

Table 2. Data collection statistics

SeMet* Native Soaking†

a, c Ň 219.2, 126.7 219.6, 133.6 218.7, 130.7Resolution, Š3.0 2.6 2.8Rsym

§¶ 0.12 (0.51) 0.12 (0.45) 0.08 (0.43)Completeness, %¶ 99.0 (97.2) 97.3 (86.1) 99.7 (100.0)Redundancy¶ 5.1 (5.0) 3.0 (2.4) 3.2 (3.1)I/�(I)¶ 11.7 (2.1) 8.0 (2.3) 13.1 (2.1)Unique reflections 69,172 110,675 86,940

*Phasing was performed with the program SHARP (41). The overall figure ofmerit for centric and acentric and reflections was 0.2 and 0.4, respectively.†The dataset was measured from a crystal soaked for 30 min at 4°C in a solutioncontaining N,N-dimethylaniline 5 mM, NADP� 1 mM, PEG4000 25%(wt/vol),Na/Hepes 0.1 M, pH 7.5.

‡The crystals were obtained with the mutant E158A/E159A. They belong tospace group P61, and their asymmetric unit contains four mFMO monomers.

§Rsym � �h� i I(hkl)��I(hkl) /�h� iI(hkl).¶Data for the highest-resolution shell are given in parentheses.

Table 3. Refinement statistics

Native Soaking

Resolution, Å 2.6 2.8R factor (Rfree)* 0.248 (0.269) 0.225 (0.242)No. of protein atoms† 14,569 14,552No. of water molecules 124 0No. of FAD and NADP� atoms 404 404rmsd bond length, Ň 0.012 0.012rmsd bond angle, °‡ 1.42 1.43Ramachandran plot§

Most favorable region, % 89.2 88.2Disallowed regions, % 0.0 0.0

*Rfactor � ��Fobs � Fcalc�/� Fobs . Rfree is the Rfactor value for 5% of the reflectionsexcluded from the refinement.

†No electron density has been detected for residues 1–7 and 451–461.‡rmsd from ideal values calculated with REFMAC5 (29).§Figures from PROCHECK (30).

6574 � www.pnas.org�cgi�doi�10.1073�pnas.0800859105 Alfieri et al.

Dow

nloa

ded

by g

uest

on

June

24,

202

0

such as K�) cannot be excluded. A structural superposition of thetwo sets of atomic coordinates (native dataset and dataset fromsoaking) reveals that the accommodation of putative dioxygendoes not induce any significant shift in the position of proteinresidues and cofactors.

Location of hFMO3 Mutations and Polymorphisms. The remarkabledegree of sequence identity makes it interesting to locate mFMOresidues corresponding to hFMO3 positions affected by patho-logical mutations and polymorphisms (11, 17). All classicalTMAU-causing mutations target residues that are located eitherin the active site or in structurally relevant regions (Figs. 1B and3A). Examples are the human mutation Asn61Ser that targetsthe catalytically crucial Asn-78 in mFMO and the Pro153Leumutation (homologous to mFMO Pro-173) that targets a residueof the domain linker. Conversely, hFMO3 polymorphisms com-monly observed in human populations concern positions that in

mFMO structure fall in solvent-exposed areas far apart from theactive site (Fig. 3A).

DiscussionWe investigated mFMO for its ability to work as a good modelfor the study of human enzymes. Biochemical analysis revealedthat this prokaryotic enzyme shares many functional propertieswith mammalian FMOs, including the ability to stabilize C4a-hydroperoxyFAD for minutes. The well known role of NADPHcoenzyme in the catalytic cycle of FMOs is that of providingFAD with reducing equivalents (Fig. 1 A). The chemistry ofNADPH requires that reduction of the flavin cofactor is broughtabout through a hydride transfer from the C4 atom of thenicotinamide ring to the flavin N5 atom (Fig. 4B). The stereo-chemistry observed in mFMO structure does not reveal a closeproximity between these two crucial atoms and thus is clearly notcompatible with a hydride transfer. This implies that the struc-ture in the crystals does not correspond to the conformation inwhich NADPH is able to reduce the flavin. A modeling exper-iment (Fig. 5A) in which the hypothetical coordinates of theC4a-hydroperoxyflavin are superimposed onto the flavin coor-dinates of the mFMO structure places the two additional oxygenatoms of the C4a-adduct within hydrogen-bond distance fromO2 atom of NADP� ribose, which then would be perfectlypositioned to properly stabilize this oxygenating intermediate.The positioning of the two additional oxygens would require onlya small (1–2 Å) shift of the Asn-78 side chain, which in turn couldcontribute to their stabilization through polar interactions. Con-sistently, the homologous Asn-61 of hFMO3 is essential foractivity (refs. 11 and 17; Fig. 3A). These considerations suggestthat the conformation in mFMO crystals is likely to correspondto that promoting the stabilization of C4a-hydroperoxyFAD.This feature is known to depend on the presence of NADP� thatwas proposed to act as a ‘‘gate keeper,’’ preventing the decay ofthe intermediate possibly by protecting the flavin N5 fromsolvent attack (6). Our structure helps to explain these functionalproperties and is a clear observation of a second fascinatingfunctional property of NADP� in FMO catalytic cycle, i.e., thestabilization of C4a-hydroperoxyflavin. NADP� shields the ac-tive site and provides a proper H-bonding environment that canprolong the intermediate half-life. Therefore, NADP� is anintegral part of the enzyme catalytic machinery that promotesintermediate stabilization and substrate monooxygenation. Theapproach of a substrate to the intermediate may be promoted bya displacement of both the loop of Tyr-212 (NADP-bindingdomain) and the stretch of residues 407–415 at the end of a�-helix (FAD-binding domain, Fig. 3A), as suggested by thehigher B-factors of this protein region.

The crystal structure of mFMO is similar to that of S. pombeFMO (21). In particular, most of the catalytic residues areidentical in the two enzymes (Fig. 1B), which display similaractive sites and, most importantly, essentially the same confor-mation of bound NADP� (Fig. 5B). In their original article,Eswaramoorthy et al. (21) explained the observed NADP�-binding mode as indicative of a flavin-reduction mechanisminvolving hydride transfer from the nicotinamide C2 atom(rather than C4) to the flavin. This interpretation has beensubsequently revised as not compatible with known NADP(H)chemistry without any further comment (33). We believe that, inview of their structural similarity, Methylophaga and S. pombeFMOs are very likely to have a similar catalytic mechanism withNADP� being actively involved in the stabilization of theC4a-hydroperoxyflavin intermediate as described above. In thiscontext, we remark that also sequence-related Baeyer–Villigermonooxygenases, e.g., cyclohexanone monooxygenase and phe-nylacetone monooxygenase (34, 35), probably use the samecatalytic strategy. Indeed, for phenylacetone monooxygenase ithas been shown that NADP� is a prerequisite for efficient and

Fig. 3. Overall crystal structure of mFMO. (A) Ribbon diagram of themonomer. FAD-binding domain (residues 8–169 and 281–450) is orange andNADP-binding domain (residues 170–280) is green. FAD is shown as yellowsticks and NADP� as blue sticks. The positions of the long interdomain loop(residues 44–80), the hinge connecting the two domains, and the polypeptidestretch corresponding to residues 407–415 are outlined. mFMO residuescorresponding to TMAU-causing mutations (17) and polymorphisms in hFMO3(in parentheses) are in red and blue sticks, respectively. The position of a longinsert in hFMO3 (residues 238–299; Fig. 1B) is also indicated. It is expected tooccupy a surface-exposed position away from the active site. (B) Ribbonrepresentation of the mFMO dimer. One monomer is shown in the sameorientation and color as Fig. 3A; the other one is colored gray, with theNADP-binding domain in darker gray.

Alfieri et al. PNAS � May 6, 2008 � vol. 105 � no. 18 � 6575

BIO

CHEM

ISTR

Y

Dow

nloa

ded

by g

uest

on

June

24,

202

0

enantioselective catalysis (36) and its crystal structure predicts aNADP�-binding mode identical to that found in FMOs, sug-gesting a similar mechanism of intermediate stabilization (37).

To bring about the first of its two roles (reduction of FAD byhydride transfer; Fig. 1 A), NADPH coenzyme must occupy a

position different from that observed in our structure. A shift ofthe ribose ring of NADP� coenzyme might be enough to movethe nicotinamide ring to a position productive for hydridetransfer. However, our repeated observation that soakingmFMO crystals in a solution of fresh NADPH rapidly bleachesand cracks them rather suggests that a relative domain rotationcould be involved in promoting FAD reduction by NADPH. Asimilar movement has been hypothesized for proteins withsimilar topology (38) and in the catalytic mechanism of pheny-lacetone monooxygenase (37).

In conclusion, in FMOs and related monooxygenases,NADP(H) has two binding modes likely to be associated todistinct protein conformational states; the first binding modealong the catalytic cycle promotes FAD reduction throughhydride transfer, whereas the second mode exerts stabilization ofthe crucial catalytic intermediate, a sort of moonlighting activityof NADP(H).

Materials and MethodsCrystallization and Structure Determination. Gene cloning and protein expres-sion and purification were carried out by using standard protocols as de-scribed in SI Text. Wild-type mFMO crystallized in many conditions usinglow-molecular-weight PEGs as precipitant. However, these crystals were af-fected by twinning. Conversely, mutant FMO E158A/E159A (SI Text) gavecrystals of a different form in microbatch technique at 4°C by mixing equalvolumes of 8 mg of protein/ml in Tris�HCl 25 mM, pH 8.0; NaCl 250 mM; NADP�

1 mM and of PEG4000 20% (wt/vol) in Na/Hepes 0.1 M, pH 7.5. These crystalswere not affected by twinning and could be used for structure determination.

Data were collected at the European Synchrotron Radiation Facility(Grenoble, France) and processed with the CCP4 package (39). The structurewas solved by the single-wavelength anomalous diffraction method (Table 2)by using SHELXD (40) and SHARP (41). The resulting electron density map was

Fig. 4. The active site of mFMO. (A) Surface representation of the enzyme with the catalytic site viewed from outside using the coloring scheme andapproximately the same orientation of Fig. 3B. (B) Binding of the NADP� nicotinamide ring. Only crucial residues are displayed. Oxygen is red, nitrogen is blue,carbon is differently colored depending on its belonging to FAD (yellow), NADP� (cyan), FAD-binding domain (orange), NADP-binding domain (green). For thesake of clarity, only the backbone atoms of Trp-76 are shown. The reactive C4a and N5 atoms of the flavin and C4 of NADP� nicotinamide are labeled. (C)Stereoscopic view of the active site of mFMO in approximately the same orientation as Fig. 4B and with the same color code. The red-colored electron density(2Fo�Fc, contoured at 1.5 �) was found in the dataset resulting from a soaking experiment (Table 2) and tentatively assigned to a dioxygen bound inside thecavity located in front of the flavin (here shown as a light blue transparent surface).

Fig. 5. The role of NADP� in the stabilization of C4a-hydroperoxyflavinintermediate. (A) Modeling experiment in which the hypothetical structure ofC4a-hydroperoxyflavin was superimposed to the flavin in mFMO structure.The color code is the same as in Fig. 4B. Hypothetical hydrogen bonds involvingthe hydroperoxyflavin atoms are shown as blue dashed lines. The accommo-dation of the additional oxygen atoms of the C4a-adduct would require a shiftof �1.5 Å of Asn-78 side chain (whose conformation in the native structure isshown as thin black stick). (B) Comparison of the NADP�-binding mode in S.pombe (Protein Data Bank ID code 2gv8) and Methylophaga FMOs. Thepicture was obtained by superimposing the C� atoms of the two proteins andshows the FAD (yellow) and NADP� (blue) molecules of mFMO together withFAD and NADP� of the S. pombe enzyme (red). The N5 and C4a atoms of theflavin and C4 and C2 atoms of NADP� are labeled.

6576 � www.pnas.org�cgi�doi�10.1073�pnas.0800859105 Alfieri et al.

Dow

nloa

ded

by g

uest

on

June

24,

202

0

used in molecular replacement calculations to solve the structure of the native(nonselenomethionine-substituted) enzyme. Phases were further improvedby multiple-crystal density averaging with DMMULTI (39). All subsequentmodel-building and refinement calculations were done by using data fromthe native crystals. The initial model was traced with ARP/wARP (42). Theprograms COOT (43) and Refmac5 (29) were used for model rebuilding andrefinement (Table 3). Figures were generated by using PyMOL (DeLano Sci-entific; www.pymol.org).

Kinetic Assays. Kinetic assays were performed in duplicate at 25°C on a Varianspectrophotometer (Cary 100 Bio) equipped with a thermostatic cell compart-ment. A 1-ml mixture containing mFMO 0.1–0.2 �M; Tris�HCl 25 mM, pH 8.5;NaCl 35 mM; 0.1 mM NADPH (dissolved in NaOH 20 mM) was equilibrated for

2 min; the reaction was triggered by addition of substrate and followed bydecrease in absorbance at 340 nm (�340 � 6.22 mM�1 cm�1 for NADPH). Allsubstrates (Table 1) were dissolved in water except N,N-dimethylaniline (dis-solved in ethanol). Km for NADPH was measured by using a fixed trimethyl-amine concentration (2 mM) and varying NADPH concentrations (0–230 �M).To generate the C4a-hydroperoxyflavin intermediate (Fig. 2, trace B), anaerated solution of NADPH 7 �M was added to a solution of mFMO (7 �M inKPi 100 mM, pH 7.4) in a final volume of 1 ml at 4°C.

ACKNOWLEDGMENTS. We thank Dr. Heinz Gut (European Synchrotron Ra-diatino Facility, Grenoble, France) for outstanding support during data col-lection. Financial support by the Italian Ministry of Science (PRIN06 and FIRBprograms), Regione Lombardia, and the American Chemical Society Petro-leum Research Fund (46271-C4) is gratefully acknowledged.

1. Cashman JR (2005) Some distinctions between flavin-containing and cytochrome P450monooxygenases. Biochem Biophys Res Commun 338:599–604.

2. van Berkel WJ, Kamerbeek NM, Fraaije MW (2006) Flavoprotein monooxygenases, adiverse class of oxidative biocatalysts. J Biotechnol 124:670–689.

3. Ziegler DM (2002) An overview of the mechanism, substrate specificities, and structureof FMOs. Drug Metab Rev 34:503–511.

4. Poulsen LL, Ziegler DM (1979) The liver microsomal FAD-containing monooxygenase.J Biol Chem 254:6449–6455.

5. Beaty NB, Ballou DP (1981) The reductive half-reaction of liver microsomal FAD-containing monooxygenase. J Biol Chem 256:4611–4618.

6. Beaty NB, Ballou DP (1981) The oxidative half-reaction of liver microsomal FAD-containing monooxygenase. J Biol Chem 256:4619–4625.

7. Beaty NB, Ballou DP (1980) Transient kinetic study of liver microsomal FAD-containingmonooxygenase. J Biol Chem 255:3817–3819.

8. Jones KC, Ballou DP (1986) Reactions of the 4a-hydroperoxide of liver microsomalflavin-containing monooxygenase with nucleophlic and electrophilic substrates. J BiolChem 261:2553–2559.

9. Krueger SK, Williams DE (2005) Mammalian flavin-containing monooxygenases: struc-ture/function, genetic polymorphisms and role in drug metabolism. Pharmacol Ther106:357–387.

10. Cashman JR (1995) Structural and catalytic properties of the mammalian flavin-containing monooxygenase. Chem Res Toxicol 8:166–181.

11. Hernandez D, Janmohamed A, Chandan P, Phillips IR, Shephard EA (2004) Organiza-tion and evolution of the flavin-containing monooxygenase genes of human andmouse: identification of novel gene and pseudogene clusters. Pharmacogenetics14:117–130.

12. Cashman JR, Zhang J (2006) Human flavin-containing monooxygenases. Annu RevPharmacol Toxicol 46:65–100.

13. Cashman JR (2003) The role of flavin-containing monooxygenases in drug metabolismand development. Curr Opin Drug Discov Devel 6:486–493.

14. Cashman JR (2004) The implications of polymorphisms in mammalian flavin-containingmonooxygenases in drug discovery and development. Drug Discov Today 9:574–581.

15. Koukouritaki SB, Hines RN (2005) Flavin-containing monooxygenase genetic polymor-phism: impact on chemical metabolism and drug development. Pharmacogenomics6:807–822.

16. Lattard V, Zhang J, Cashman JR (2004) Alternative processing events in human FMOgenes. Mol Pharmacol 65:1517–1525.

17. Zhou J, Shephard EA (2006) Mutation, polymorphism and perspectives for the futureof human flavin-containing monooxygenase 3. Mutat Res 612:165–171.

18. Schlaich NL (2007) Flavin-containing monooxygenases in plants: looking beyond detox.Trends Plants Sci 12:412–418.

19. Choi HS, et al. (2003) A novel flavin-containing monooxygenase from Methylophagasp strain SK1 and its indigo synthesis in Escherichia coli. Biochem Biophys Res Commun306:930–936.

20. Fraaije MW, Kamerbeek NM, van Berkel WJH, Janssen DB (2002) Identification of aBaeyer-Villiger monooxygenase sequence motif. FEBS Lett 518:43–47.

21. Eswaramoorthy S, Bonanno JB, Burley SK, Swaminathan S (2006) Mechanism of actionof a flavin-containing monooxygenase. Proc Natl Acad Sci USA 103:9832–9837.

22. Chenna R, et al. (2003) Multiple sequence alignment with the Clustal series of pro-grams. Nucleic Acids Res 31:3497–3500.

23. Gouet P, Courcelle E, Stuart DI, Metoz F (1999) ESPript: multiple sequence alignmentsin PostScript. Bioinformatics 15:305–308.

24. Palfey BA, Massey V (1998) in Comprehensive Biological Catalysis (Academic, SanDiego, CA), Vol 3, pp 83–154.

25. Kemal C, Bruice TC (1976) Simple synthesis of a 4a-hydroperoxy adduct of a 1,5-dihydroflavine: preliminary studies of a model for bacterial luciferase. Proc Natl AcadSci USA 73:995–999.

26. Poulsen LL, Ziegler DM (1995) Multisubstrate flavin-containing monooxygenases:applications of mechanism to specificity. Chem Biol Interact 96:57–73.

27. Suh JK, Poulsen LL, Ziegler DM, Robertus JD (1996) Molecular cloning and kineticcharacterization of a flavin-containing monooxygenase from Saccharomyces cerevi-siae. Arch Biochem Biophys 336:268–274.

28. Derewenda ZS (2004) Rational protein crystallization by mutational surface engineer-ing. Structure 12:529–535.

29. Murshudov GN, Vagin AA, Dodson EJ (1997) Refinement of macromolecular structuresby the maximum-likelihood method. Acta Crystallogr D 53:240–255.

30. Laskowski RA, MacArthur MW, Moss DS, Thornton JM (1993) PROCHECK: a program tocheck the stereochemical quality of protein structures. J Appl Crystallogr 26:283–291.

31. Kleywegt GJ, Jones TA (1994) Detection, delineation, measurement and display ofcavities in macromolecular structures. Acta Crystallogr D 50:178–185.

32. Berman HM, Westbrook J, Feng Z, Gilliland G, Bhat TN, Weissig H, Shindyalov IN,Bourne PE (2000) The Protein Data Bank. Nucleic Acids Res 28:235–242.

33. Eswaramoorthy S, Bonanno JB, Burley SK, Swaminathan S (2006) Mechanism of actionof a flavin-containing monooxygenase. Proc Natl Acad Sci USA 103:9832–9837, anderratum (2007) 104:14543.

34. Sheng D, Ballou DP, Massey V (2001) Mechanistic studies of cyclohexanone monoox-ygenase: chemical properties of intermediates involved in catalysis. Biochemistry40:11156–11167.

35. Torres Pazmino DE, Baas JB, Janssen DB, Fraaije MW (2008) The kinetic mechanism ofphenylacetone monooxygenase from Thermobifida fusca. Biochemistry 47:4082–4093.

36. de Gonzalo G, Ottolina G, Carrea G, Fraaije MW (2005) [Cp*Rh(bpy)(H2O)]2� as acoenzyme substitute in enzymatic oxidations catalyzed by Baeyer-Villiger monooxy-genases. Chem Commun 29:3724–3726.

37. Malito E, Alfieri A, Fraaije MW, Mattevi A (2004) Crystal structure of a Baeyer-Villigermonooxygenase. Proc Natl Acad Sci USA 101:13157–13162.

38. Lennon BW, Williams CH, Jr, Ludwig ML (2000) Twists in catalysis: alternating confor-mations of Escherichia coli thioredoxin reductase. Science 289:1190–1194.

39. CCP4 (Collaborative Computational Project, Number 4) (1994) The CCP4 suite: pro-grams for protein crystallography. Acta Crystallogr D 50:760–763.

40. Schneider TR, Sheldrick GM (2002) Substructure solution with SHELXD. Acta CrystallogrD 58:1772–1779.

41. Bricogne G, Vonrhein C, Flensburg C, Schiltz M, Paciorek W (2003) Generation, repre-sentation and flow of phase information in structure determination: recent develop-ments in and around SHARP 2.0. Acta Crystallogr D 59:2023–2030.

42. Perrakis A, Morris R, Lamzin VS (1999) Automated protein model building combinedwith iterative structure refinement. Nat Struct Biol 6:458–463.

43. Emsley P, Cowtan K (2004) Coot: model-building tools for molecular graphics. ActaCrystallogr D 60:2126–2132.

Alfieri et al. PNAS � May 6, 2008 � vol. 105 � no. 18 � 6577

BIO

CHEM

ISTR

Y

Dow

nloa

ded

by g

uest

on

June

24,

202

0