- Email: [email protected]
Structural Basis for Substrate Recognition and Specificity in Aklavinone-11-Hydroxylase from Rhodomycin Biosynthesis
doi:10.1016/j.jmb.2009.09.003
J. Mol. Biol. (2009) 393, 966–977
Available online at www.sciencedirect.com
Structural Basis for Substrate Recognition and Specificity in Aklavinone-11-Hydroxylase from Rhodomycin Biosynthesis Ylva Lindqvist 1 , Hanna Koskiniemi 1 , Anna Jansson 1 , Tatyana Sandalova 1 , Robert Schnell 1 , Zhanliang Liu 2 , Pekka Mäntsälä 2 , Jarmo Niemi 2 and Gunter Schneider 1 ⁎ 1
Department of Medical Biochemistry and Biophysics, Karolinska Institutet, S-171 77 Stockholm, Sweden 2
Department of Biochemistry and Food Chemistry, University of Turku, FIN-20014 Turku, Finland Received 30 July 2009; received in revised form 28 August 2009; accepted 2 September 2009 Available online 8 September 2009
In the biosynthesis of several anthracyclines, aromatic polyketides produced by many Streptomyces species, the aglycone core is modified by a specific flavin adenine dinucleotide (FAD)- and NAD(P)H-dependent aklavinone-11-hydroxylase. Here, we report the crystal structure of a ternary complex of this enzyme from Streptomyces purpurascens, RdmE, with FAD and the substrate aklavinone. The enzyme is built up of three domains, a FAD-binding domain, a domain involved in substrate binding, and a Cterminal thioredoxin-like domain of unknown function. RdmE exhibits structural similarity to aromatic hydroxylases from the p-hydroxybenzoate hydroxylase family, but unlike most other related enzymes, RdmE is a monomer. The substrate is bound in a hydrophobic pocket in the interior of the enzyme, and access to this pocket is provided through a different route than for the isoalloxazine ring of FAD—the backside of the ligand binding cleft. The architecture of the substrate binding pocket and the observed enzyme–aklavinone interactions provide a structural explanation for the specificity of the enzyme for non-glycosylated substrates with C9-R stereochemistry. The isoalloxazine ring of the flavin cofactor is bound in the “out” conformation but can be modeled in the “in” conformation without invoking large conformational changes of the enzyme. This model places the flavin ring in a position suitable for catalysis, almost perpendicular to the tetracyclic ring system of the substrate and with a distance of the C4a carbon atom of the isoalloxazine ring to the C-11 carbon atom of the substrate of 4.8 Å. The structure suggested that a Tyr224-Arg373 pair might be involved in proton abstraction at the C-6 hydroxyl group, thereby increasing the nucleophilicity of the aromatic ring system and facilitating electrophilic attack by the perhydroxy-flavin intermediate. Replacement of Tyr224 by phenylalanine results in inactive enzyme, whereas mutants at position Arg373 retain catalytic activity close to wildtype level. These data establish an essential role of residue Tyr224 in catalysis, possibly in aligning the substrate in a position suitable for catalysis. © 2009 Elsevier Ltd. All rights reserved.
Edited by R. Huber
Keywords: flavoenzyme; anthracycline; polyketide biosynthesis; aromatic hydroxylase; merohedral twinning
*Corresponding author. E-mail address: [email protected]. Present address: Z. Liu, Department of Plant Pathology, Agricultural University of Hebei, Baoding 071001, China. Abbreviations used: FAD, flavin adenine dinucleotide; PHHY, phenol hydroxylase; pHBH, p-hydroxybenzoate hydroxylase; MHBH, 3-hydroxybenzoate hydroxylase; RebC, rebeccamycin biosynthetic enzyme; PgaE, angucycline biosynthetic enzyme; PDB, Protein Data Bank; TEV, tobacco etch virus. 0022-2836/$ - see front matter © 2009 Elsevier Ltd. All rights reserved.
Aklavinone-11-Hydroxylase Ternary Complex
Introduction Anthracyclines are aromatic polyketides that contain a common tetrahydrotetracene-5,12-quinone core. These metabolites are synthesized in vivo by complex enzymatic pathways through a process reminiscent of fatty acid biosynthesis.1–5 A type II, iterative polyketide synthase complex assembles the polyketide chain from acetyl- and malonyl-coenzyme A units, which is subsequently cyclized by specific aromatases and/or cyclases. In most of these pathways, chain elongation by polyketide synthase and subsequent β-keto-reduction and cyclization events result in the formation of aklavinone (see Fig. 1 for structure and atom numbering), a common precursor of many anthracyclines. This metabolic intermediate is further modified by a number of enzymes in so-called tailoring steps, which include hydroxylation, oxidation, dehydrogenation, methylation, and dehydration and which result in differences in the substitution patterns of the polyketide core, thus generating the remarkable chemical diversity of the anthracyclines.4,5 A modification common to all anthracyclines is glycosylation, generally at position C-7, as it is a prerequisite for biological activity. Many of the anthracyclines have medically important properties, and several are in clinical use, for instance, as anticancer drugs (doxorubicin, daunorubicin).1 The rdm gene cluster of Streptomyces purpurascens contains the genes that code for proteins of rhodomycin biosynthesis.6,7 These enzymes include RdmA, an aklanonic acid methyl ester cyclase that is a member of the cyclase family represented by SnoaL and AknH,8,9 and the tailoring enzymes RdmB, RdmC, and RdmE. RdmC is a methylesterase that catalyzes removal of the methoxygroup at the C-15 position of aklavinone,10,11 and RdmB is a novel S-adenosyl-methionine-dependent hydroxylase that converts 15-demethoxy-ɛ-rhodomycin glycoside to β-rhodomycin glycoside.10,12 RdmE encodes an aklavinone-11-hydroxylase, a NAD (P)H- and flavin adenine dinucleotide (FAD)-dependent monooxygenase.13 The enzyme catalyzes the incorporation of a hydroxyl group at position C-11 of aklavinone, resulting in ɛ-rhodomycinone (Fig. 1). This tailoring reaction is shared between biosynthetic pathways leading to different anthracyclines, for instance, daunorubicin in S. peucetius14,15 and rhodo-
967 mycin in S. purpurascens.13 Aklavinone hydroxylases act before the glycosylation step; that is, they cannot accept glycosylated substrates. RdmE is also specific for the C-9R configuration of anthracyclines and does not hydroxylate compounds with C-9S configuration.9 Aklavinone hydroxylases consist of a polypeptide chain of about 500–540 amino acids, with a molecular mass in the range of 52–55 kDa, and show weak to moderate sequence similarities (18–30% amino acid identity) to the p-hydroxybenzoate hydroxylase (pHBH) family of flavoenzymes.16 The first insights into the structural basis of catalysis in this family of aromatic hydroxylases were obtained from studies of pHBH.17–20 These studies were subsequently extended to include phenol hydroxylase (PHHY),21,22 3-hydroxybenzoate hydroxylase (MHBH),23 angucycline biosynthetic enzyme (PgaE)/CabE,24 and rebeccamycin biosynthetic enzyme (RebC).25,26 An intriguing feature common to these enzymes is the two conformations of the FAD cofactor that can be associated with the two half-reactions, reduction of the FAD cofactor by NAD(P)H and oxidation of the substrate using molecular oxygen. In a number of crystal structures of these enzymes, FAD is observed in the so-called “in” conformation (or closed conformation), in which the isoalloxazine ring lies within the active site and the substrate binding pocket is inaccessible from the bulk solvent.18,21,24,25 The reactions with molecular oxygen and the hydroxylation of the aromatic substrate are expected to occur while FAD is in this conformation. In order to be reduced by NAD(P)H, the isoalloxazine ring of FAD has to move from the active site towards the solvent. This conformation is termed the “out” conformation and has been observed in crystal structures of several members of the pHBH family.19–21,23,26 In this article, we report the determination of the three-dimensional structure of a ternary complex of RdmE with bound FAD and aklavinone by means of X-ray crystallography to a resolution of 2.5 Å. The crystal structure reveals the basis for the stereospecific recognition of aklavinone and for the inability for the aklavinone glycoside to act as substrate. It also identifies Tyr224 as an important residue in the interaction with the substrate and essential for catalysis, which is confirmed by site-directed mutagenesis. Structural comparison of RdmE with other
Fig. 1. Overall reaction catalyzed by aklavinone-11-hydroxylases. Numbering of the carbon atoms and labeling of the tetracyclic ring system of aklavinone are also shown.
968 aromatic hydroxylases, in particular with pHBH,17 PHHY,21,27,28 and RebC, 25,26 reveals that while overall features of structure and catalytic function appear to hold for all members of this family, there are intriguing differences among these enzymes with regard to substrate entry/product release, conformational changes during catalysis, and catalytic features.
Results and Discussion Structure determination The structure of RdmE was determined by molecular replacement from merohedrally twinned crystals using search models based on the three most related enzymes, all with sequence identities to RdmE of less than 30% [Protein Data Bank (PDB) codes 2r0c, 2qa1, and 2qa2]. The search models were truncated, retaining only conserved structural elements. Initial molecular replacement trials using different programs were unsuccessful; only after removal of the strong peak at 2/3, 1/3, 0 in the native Patterson map was a correct solution showing an excellent signal in the P63 space group obtained. The quality of the refined electron density maps and model and refinement statistics confirm that this solution is correct. The electron density map is continuous for the whole polypeptide chain, and with a few exceptions, all amino acids, from residue 1 to residue 535, are well defined in electron density. In particular, the electron density for the bound FAD and aklavinone is well defined and allows unambiguous placement of these molecules. Less welldefined electron density is found for residues 151– 155, and a few solvent-exposed side chains are too flexible to be observed. The overall quality of the model of RdmE is as expected for this resolution, and the majority (95%) of the residues are found in the most favored region of the Ramachandran plot (Table 1). Overall structure of the enzyme The monomer of RdmE consists of three domains (Fig. 2), the FAD-binding domain (amino acids 1–71, 99–198, and 289–393), the middle domain (amino acids 72–98 and 199–288), and the C-terminal domain (residues 394–535). The core of the molecule is formed by the FAD-binding domain, which is dominated by a Rossmann-type βαβ-fold comprising a mixed, six-stranded βsheet, flanked on one side by a small α-helix and a three-stranded antiparallel β-sheet (Fig. 2). The other side is lined by four α-helices, of which the longest helix (residues 345–373) spans the whole length of the domain. An important additional secondary-structure element is the three-stranded antiparallel β-sheet, which contributes to the formation of the FAD-binding groove close to the isoalloxazine ring.
Aklavinone-11-Hydroxylase Ternary Complex Table 1. Statistics of data collection and crystallographic refinement Data collection Resolution (Å) Beamline λ (Å) Number of observations Number of unique reflections Rsym (%) Completeness (%) I/σ(I) Multiplicity Wilson B-factor (Å2) Refinement Resolution interval (Å) Rwork (%) Rfree (%) Number of amino acids Number of atoms Protein FAD Aklavinone Water/sulfate B-factor (Å2) Protein FAD Aklavinone Water molecules rmsd from ideal geometry Bond length (Å) Bond angle (°) Refined twin fraction Ramachandran plot (%) Residues in most favored region Residues in allowed regions Residues in disallowed regions
50–2.5 (2.62–2.5) 7–11, MAX-laboratory 1.01590 351,856 (22,851) 64,506 (7778) 8.2 (23.5) 97.1 (80.8) 16.0 (4.7) 5.5 (2.9) 38.7 30.0–2.5 20.4 25.5 3 × 535 12,191 159 90 199/90 23.0 19.3 23.6 11.3 0.019 1.877 0.49 95.0 4.3 0.7
Values in parentheses are for the highest-resolution shell.
The middle domain is inserted into the FADbinding domain and folds into a seven-stranded antiparallel β-sheet, which is flanked by an α-helix. The active site is located at the interface between the FAD-binding domain and the middle domain. The latter forms the roof of the substrate binding pocket. The C-terminal domain displays a typical thioredoxin fold, as in other members of this enzyme family with the exception of pHBH, which lacks this domain. While the C-terminal domain participates in quaternary-structure formation in MHBH and PHHY, it is not part of the dimer interface in PgaE, CabE,24 and the Bayer–Villiger monooxygenase MtmOIV from the mithramycin biosynthetic pathway.29 Quaternary structure The asymmetric unit of the RdmE crystals contains three molecules. The interface areas between these molecules are in the range of 230– 320 Å2, typical in size of crystal contacts rather than protein–protein interfaces. Larger interfaces (approximately 920 Å2) are formed with molecules related by crystallographic symmetry, which lead to the formation of a trimer along the 6-fold screw axis. Although these values approach the range of protein–protein interfaces, we consider these interactions as crystal packing effects rather than
Aklavinone-11-Hydroxylase Ternary Complex
969
Fig. 2. Overall structure of RdmE. The FAD domain is shown in green, the middle domain in red, and the C-terminal domain in yellow. The bound ligands in the ternary complex, FAD (light blue) and aklavinone (dark blue), are shown as stick models.
displaying biologically relevant assemblies. Analytical gel-filtration data also suggest a monomeric structure of RdmE in solution (data not shown). Relation to the pHBH family Sequence comparisons suggested that RdmE is a member of the pHBH family of flavoenzymes. A search of the PDB with the RdmE coordinates using DALI30 returned a number of related structures, with the most similar being the flavin-dependent monooxygenases RebC,25,26 PHHY,21 and MHBH23 with Z-scores of 40.8, 36.9, and 35.5, respectively. Superposition of RdmE with the subunit of RebC (PDB code 2r0p) gave an rmsd value of 1.9 Å for 442 aligned residues. This value is only slightly larger than the rmsd values for the superposition of individual domains (1.6, 1.9, and 1.7 Å), showing that there are no larger differences in domain orientation between these two proteins. This finding also holds for other relatives of this enzyme family; that is, structural comparisons of RdmE with PHHY, MHBH, and PgaE return similar rmsd values and domain orientations. A structure-based sequence alignment of RdmE with other members of this sequence family of known structure, RebC, PHHY, PgaE, CabE, pHBH, and MBBH, results in pairwise sequence identities in the range 18–30%, with 37 invariant residues that are mostly involved in flavin binding or inter-domain interactions. The FAD binding site The cofactor is bound in the active site in the “out” conformation, with the ADP and ribityl moieties
embedded in the FAD-binding domain and the isoalloxazine ring located at the interface between the FAD-binding domain and the middle domain. FAD is tightly anchored to the enzyme by a significant number of direct or water-mediated hydrogen bonds to protein residues (Fig. 3). Most of the FAD-binding residues are well conserved in the pHBH family. Gln119, Asp308, and Asn322 are invariant (the latter residue binds FAD only in the “in” conformation), whereas Arg36 and Arg45 are often conservatively exchanged to lysine or glutamine, and Glu35 is replaced by aspartate in MHBH. This residue belongs to the dinucleotide binding motif that is conserved among flavin- and NAD(P) H-utilizing proteins31 and in the pHBH family.32 In the ternary complex of RdmE, the FAD-ring system is observed in the “out” conformation. The isoalloxazine ring of FAD is wedged between the side chain of Trp288 on one side and the polypeptide chain Arg45-Ala46, and in particular the aliphatic part of the side chain of Arg45, on the other side. There are no hydrogen bonds from the protein to the ring system that could stabilize this conformation further unlike in PHHY (subunit A) and MHBH, where a hydrogen bond between the N3 nitrogen atom of FAD to the side chain of a tyrosine residue was observed.21,23 This residue is replaced by a phenylalanine (Phe246) in RdmE, unable to engage in a similar interaction. There is, however, a hydrogen bond between the N3 nitrogen atom of the flavin ring and the oxygen atom of ring B of the substrate aklavinone, and it is presumably this interaction that stabilizes the flavin ring in the “out” conformation. A similar substrate–N3 flavin hydrogen bond has been found in the complex of
970
Aklavinone-11-Hydroxylase Ternary Complex
Fig. 3. Stereoview of the FAD binding site in RdmE. The carbon atoms of FAD are shown in green, and protein carbon atoms are shown in yellow. Hydrogen bonds are indicated by broken lines.
pHBH–2,4 dihydroxybenzoate, and also in this case, flavin was found in the “out” conformation.19,20 Structural basis of substrate recognition and specificity At the interface between the FAD- and substratebinding domains, in the vicinity of the isoalloxazine ring of FAD, there is a hydrophobic pocket that contains the bound substrate aklavinone (Fig. 4a). The location of the substrate binding pocket is similar to that of other members of the pHBH family, and the positions of the bound substrates overlap well. The tetracyclic ring system of aklavinone is bound in a rather narrow cleft, which is lined by predominantly hydrophobic amino acids (Fig. 4b). One side of the aklavinone ring system is packed against the aromatic side chain Trp222 and flanked by Phe79 and Phe246, and the other side packs against the wall of the active-site pocket defined by the polypeptide stretch Pro315-Thr316-Gly317-Gly318. The edge of ring D of the substrate is adjacent to another hydrophobic patch in the substrate pocket defined by the side chains of Met202, Thr233, Trp288, and Met290. The hydroxyl group of ring B of aklavinone forms a hydrogen bond to the side chain of Tyr224, and the C-12 carbonyl group of ring C is anchored to the main-chain nitrogen atom of Ala47 via a watermediated hydrogen-bond interaction. The same oxygen atom is hydrogen bonded to the N3 nitrogen atom of the isoalloxazine ring (see above). Ring A of aklavinone, which is observed in halfchair conformation, interacts via the C-7 hydroxyl group with the side chain of Arg373 through an indirect hydrogen bond involving a water molecule. The C-9 hydroxyl also interacts indirectly via a
bridging water molecule with the carbonyl oxygen atom of Gly318, and the ethyl substituent points into a hydrophobic pocket created by Phe79, Ile81, Met102, Trp114, and Met116 (Fig. 4b). These features seem to be responsible for the stereospecific recognition of the C9-R isomer of aklavinone by RdmE. Reversal of the stereochemistry at this position would result in a mismatch of the physicochemical properties of the substituents and their respective binding pockets; that is, hydrophilic groups of the substrate would point into hydrophobic patches on the enzyme and vice versa. This should disfavor binding of the S-isomer. Indeed, nogalaviketone, which differs from aklavinone by reversal of the stereochemistry at the C-9 carbon, carbonyl instead of hydroxyl group at C-7, and replacement of the ethyl by a methyl substituent, is not hydroxylated by RdmE, whereas its C9-R isomer auraviketone is a substrate.9 In contrast to other tailoring enzymes in anthracycline biosynthesis, RdmE does not accept aklavinone substrates glycosylated at position C-7. In the structure of the RdmE–aklavinone complex, there is a pocket adjacent to the C-7 hydroxyl group, which extends to the surface of the enzyme. However, modeling of bound aklavinone glycoside resulted in severe clashes of carbohydrate atoms with enzyme residues, for instance, the side chain of Ile81, Val95, Tyr224, or Arg373. Thus, this pocket is too small in size to accommodate the carbohydrate moiety of aklavinone glycosides, which leads to hydroxylation of only non-glycosylated polyketide substrates. Site-directed mutagenesis Residues Tyr224 and Arg373 are among the few polar amino acids in close vicinity of the substrate.
971
Aklavinone-11-Hydroxylase Ternary Complex
Fig. 4. Substrate binding site in RdmE. (a) Part of an omit electron density map, calculated with the algorithm of Vagin et al.,33 implemented in SFCHECK in the CCP4 program package,34 at the active site of RdmE, contoured at 1.2 σ. The refined position of bound substrate aklavinone is superposed. AKV denotes the bound substrate aklavinone. (b) Stereoview of the active site in the ternary complex of RdmE with FAD and aklavinone. Amino acids lining the wall of the substrate binding pocket are shown. Residues whose function has been probed by site-directed mutagenesis are indicated with carbon atoms in magenta. FAD is drawn with carbon atoms in green, and aklavinone is drawn with carbon atoms in blue.
We therefore changed these residues to a phenylalanine (Tyr224Phe) and alanine, glutamine, and methionine (Arg373Ala, Arg373Gln, and Arg373Met), respectively, by site-directed mutagenesis to probe their role in catalysis. Substitutions at the position Arg373 resulted in mutants with similar specific activities as the wild-type enzyme, whereas replacement of Tyr224 with a phenylalanine abolished catalytic activity (Table 2).
Table 2. Specific activities of wild-type and mutant RdmE Species Wild type Arg373Ala Arg373Gln Arg373Met Tyr224Phe
Vmax (nM/min mg)
Specific activity (%)
35.8 ± 1.2 36.2 ± 7.8 41.7 ± 3.5 19.2 ± 1.2 Not detectable
100 101 115 53 0
Substrate entry In PHHY and pHBH, access of the substrate to the active site is controlled by conformational changes that allow entry through a channel passing the re face of the flavin ring.21,35 In MHBH, the substrate binding pocket is rather open and would allow entry of the substrate by two different routes of access without the need for larger conformational changes.23 For RebC and RdmE, which use more bulky substrates, access to the active site via the flavin pocket is highly unlikely. For RebC, a path for substrate entry/product release has been proposed, which involves a disorder/order transition of a peptide stretch at the “backside” of the substrate pocket, opposite to the flavin ring.25 Folding of the disordered stretch into an α-helix after binding of substrate closes off the pocket and seals the ligand in its binding site. Interestingly, the largest differences between the structures of RdmE and RebC are
972 localized at this “backdoor” of the substrate pocket. Compared to RdmE, RebC has a 24-residue insertion that completely shields the substrate once the “melting” helix is folded. In RdmE, the polypeptide chain does not run across the backside completely, and there is an opening into the substrate pocket from this direction (Fig. 5). In the ternary complex, this opening is in part occluded by protein side chains, but small conformational changes at the side-chain level (Tyr224, Arg373, and Ile369) would allow passage of the aklavinone substrate to its binding pocket. We thus conclude that (i) while the general entry points for substrates into their binding site are conserved in RdmE and RebC, the precise mechanism in RebC appears more elaborate, and (ii) for both enzymes, it is different from the proposed channel adjacent to the flavin ring used by the smaller substrates phenol and p-hydroxybenzoate in PHHY and pHBH, respectively. Mechanistic implications Crystallographic, spectroscopic, and kinetic studies with pHBH, PHHY, and RebC concluded that catalysis in these hydroxylases can be summarized by the following overall steps (Fig. 6). In the absence of substrate, the enzymes most likely exist as binary complexes with oxidized FAD bound in two conformations in a dynamic equilibrium. The reduction of FAD by NAD(P)H in the absence of substrate is very slow, thus avoiding wasteful usage of reduction equivalents.25 Interaction of NAD(P)H with the ternary enzyme–FAD–substrate complex shifts the FAD to the “out” conformation where the isoalloxazine ring is in favorable position for hydride transfer.36 Release of NAD(P)H and shift of the reduced FAD back to the ”in” position comprise the next step, which is then followed by oxygen activation and hydroxylation of the substrate. The structure of the ternary complex of RdmE with bound substrate shows preferential binding of flavin in the “out” conformation also in the absence of NAD(P)H. This is in line with studies of PHHY where both the “in” and “out” conformations were observed in the presence of the substrate phenol and the absence of NAD(P)H21 and the structure of the substrate complex of MHBH,23 where flavin was observed in the “out” position. In these cases, this conformation is stabilized by hydrogen bonds to the isoalloxazine ring, either to enzyme residues (PHHY, MHBH) or to the substrate (RdmE). In these enzymes, FAD is thus poised for reaction with NAD(P)H, whereas in pHBH, deprotonation of the substrate and binding of NAD(P)H appear to trigger the shift from the “in” to the “out” conformation of the isoalloxazine ring.37 Reduction of FAD results in preferential binding of the flavin ring in the “in” conformation,20,26 associated with the enzymatic states during oxygen activation and hydroxylation reaction. For RdmE, we were not successful in obtaining an experimental structure with the flavin in the “in” conformation, and we therefore modeled this state based on the
Aklavinone-11-Hydroxylase Ternary Complex
structure described here and the isoalloxazine “in” conformation observed in the closest related enzyme, RebC.26 This conformation of FAD can be obtained straightforwardly by rotation around two bonds, C4′–C5′ and C1′–C2′. The conformation of the side chain of Trp288 was changed by a rotation around the Cα–Cβ bond in order to accommodate the isoalloxazine ring in the active site and to avoid unfavorable steric interactions. In this model, the isoalloxazine ring is almost perpendicular to the tetracyclic ring system of aklavinone, with the C4a atom of FAD about 4.8 Å away from the C-11 of the substrate (Fig. 7). The putative flavin 4a-hydroperoxide intermediate would place the hydroxyl group within 2.8 Å of C-11, the target of the electrophilic attack. A model of the hydroperoxide intermediate in the active site of RdmE also shows the characteristic hydrogen-bond interaction of the O-4 atom of the isoalloxazine ring with the main-chain nitrogen of residue Ala-47 that might promote formation of the hydroperoxide intermediate during catalysis as proposed for pHBH.38 The electrophilic attack of the hydroperoxide intermediate onto the aromatic ring system of the substrate could be further facilitated by an increase in electron density due to deprotonation of the substrate as shown for pHBH.37 In RdmE, the C-6 hydroxyl group of aklavinone forms a hydrogen bond to the side chain of Tyr224, which, in turn, is within hydrogen-bonding distance to the side chain of Arg373. We hypothesized that the close proximity of the positive charge of Arg373 and the tyrosine side chain might shift the pKa of Tyr224 such that it would be able to act as a catalytic base in the deprotonation of the substrate. The high catalytic activities of the mutants at position 373 (Table 2) suggest, however, that the arginine side chain is not required for catalysis, which makes this mechanistic scenario less likely. On the other hand, the mutagenesis data clearly indicate that Tyr224 plays an essential role in catalysis, possibly by orienting the substrate in a position suitable for catalysis.
Fig. 5. Surface representation of RdmE illustrating the route of access of the active site by the substrate. The bound aklavinone is shown as yellow stick model. The electrostatic potential is contoured from −25 to +25 kT.
Aklavinone-11-Hydroxylase Ternary Complex
Fig. 6. Proposed steps of the conversion of aklavinone to ɛ-rhodomycinone by RdmE.
973
974
Aklavinone-11-Hydroxylase Ternary Complex
Fig. 7. Comparison of the “out” conformation of FAD (light green) as observed in the crystal structure of the ternary complex of RdmE with the modeled “in” conformation (dark green). Modeling of the “in” state is based on the conformation of FAD as observed in subunit A of RebC (2rop).
Loss of the C4a hydroxyl group, that is, regeneration of the oxidized flavin ring, requires protonation of the hydroxyl in order to generate a better leaving group. There is no enzyme group in sufficient close vicinity to act as catalytic acid, but there are several well-defined water molecules in the proximity of the modeled position of the perhydroxy group that could participate in proton transfer.
Materials and Methods Protein production and preparation of substrate Recombinant RdmE was produced in Streptomyces lividans TK24 and purified with some modifications to the procedure described.13 The medium used was Tryptone Soya Broth (Oxoid); only anion exchange using a HiPrep 16/10 Q XL column [0–1 mol/l NaCl gradient in 50 mmol/l Tris–HCl, pH 7.5, 0.5 mM dithiothreitol (DTT), and 10% glycerol] and gel exclusion chromatography using a HiLoad 26/60 Superdex 200 column (50 mmol/ l Tris–HCl, pH 7.5, 0.5 mM DTT, and 10% glycerol) were performed in an Äkta FPLC system (Amersham Pharmacia). Protein samples were concentrated and stored at −20 °C in gel exclusion buffer. Screening for production of soluble RdmE in Escherichia coli was carried out using constructs comprising fulllength [residues 2–535; the N-terminal methionine is replaced by a tobacco etch virus (TEV) cleavage site] and truncated sequences. The latter were designed based on sequence comparison and predicted domain borders. Coding sequences for these constructs were cloned by ligation-independent cloning39 into the expression vector pNIC28-Bsa4 (GenBank Accession No. EF198106), containing a T7/lacO promoter, an N-terminal His6 tag, and a TEV protease site. The presence of the insert was confirmed by DNA sequencing. The expression constructs were tested for soluble protein production in small scale using the expression strain BL21-Gold (DE3) (Stratagene)
carrying the pRARE2 plasmid from the Rosetta2 strain (Novagen). This procedure resulted in the identification of one construct (pNIC28-RdmE), comprising an N-terminal His6 tag, the TEV cleavage site, and amino acids Asn2Arg535 of RdmE that produced soluble protein in E. coli. For large-scale purification, this construct was expressed in the E. coli strain BL21-Gold (DE3) pRARE2. The bacterial cells were grown at 37 °C in TB medium supplemented with 35 μg/ml kanamycin and 20 μg/ml chloramphenicol. At an OD600 (optical density at 600 nm) value of 0.5–0.7, gene expression was induced by addition of 0.5 mM IPTG and the temperature was lowered to 18 °C. The cells were harvested ca 15 h after induction, and the bacterial cell walls were disrupted by lysozyme treatment and sonication. The clarified lysate was loaded on Ni-NTA resin (Qiagen) and eluted from the column with an imidazole gradient. The protein sample was stored until further use at −80 °C in 50 mM Tris buffer, pH 7.5, and 50 mM NaCl containing 5% (v/v) glycerol and 1 mM TCEP. Aklavinone was prepared using the protocol described previously.13 Site-directed mutagenesis The coding sequences for the active-site mutants Tyr224Phe, Arg373Ala, Arg373Gln, andArg373Met were produced by PCR using the plasmid pNIC28-RdmE as template and mutation-specific primers in combination with the original amplification primers. The DNA fragments carrying the mutations were cloned in the pNIC28Bsa4 vector, and their sequences were verified by DNA sequencing. Production of the mutant proteins in E. coli and purification to electrophoretic homogeneity followed the same protocol as outlined above for wild-type RdmE. Activity assay Enzymatic activity was determined using the assay described previously.13 Due to overlap of the absorption spectra of NADPH and the substrate aklavinone, the reaction was followed spectroscopically by monitoring the
975
Aklavinone-11-Hydroxylase Ternary Complex appearance of the product at 500 nm. The reaction mixture (total volume, 0.1 ml) consisted of 40 μM aklavinone (dissolved in 5 μl methanol), 10 μM FAD, and 3 mM NADPH in 100 mM potassium phosphate buffer, pH 7.5. The reaction was initiated by addition of the enzyme to a concentration of 1.0 μM, and the reaction was monitored for 30 min. All activity measurements were carried out in triplicate. Crystallization The ternary complex of recombinant RdmE, produced in S. lividans, with FAD and substrate was prepared by the addition of FAD (final concentration, 0.8 mM) and aklavinone (final concentration, 1.74 mM) to a solution of RdmE (2.2 mg/ml) in 20 mM Tris buffer, pH 7.5, containing 1 mM DTT. After several hours of incubation on ice, the protein was concentrated to 10 mg/ml. The complex was crystallized by the vapor diffusion method. Orange–yellow crystals of RdmE appeared after several days of equilibration at 20 °C in drops composed of 2 μl protein solution and 2 μl reservoir solution (20 mM Tris, pH 7.6, containing 2.1 M ammonium sulfate). Data collection The crystals of RdmE were cryo-protected by quick soaking in mother liquor containing 25% ethylenglycol (v/v). A native data set was collected to 2.5 Å resolution at beamline 7/11, MAX-laboratory, Sweden. The data set was processed with the program MOSFLM and scaled with SCALA from the CCP4 package.34 Detailed data collection statistics are given in Table 1. The sigmoidal cumulative intensity distribution indicated crystal twinning, and additional twinning indicators were obtained using the twinning server†.40 Crystals of RdmE belong to the space group P63 with unit cell dimensions of a = b = 183.0 Å and c = 99.6 Å. The asymmetric unit contains three molecules of RdmE, resulting in a Matthews coefficient of 2.6 Da/Å3 and a solvent content of approximately 53%.
structure determination by molecular replacement with the program Phaser42 and the “core” search model ensemble described above. A clear solution with the positions of the three molecules in the asymmetric unit was obtained, and during refinement using Refmac,43 electron density for FAD and the bound substrate appeared at the expected positions, strong evidence that the molecules were correctly placed in the crystal lattice. Model building was carried out using the program Coot,44 interspersed with refinement using Refmac.43 For determination of Rfree, 5% of the reflections were omitted from refinement. Comparisons of refinement revealed that the switch back to the uncorrected data set (but including merohedral twinning operators) resulted in better maps and statistics than using the corrected data. Refinement using the latter stalled at an Rfree value of 36%. Presumably, the reason for success with molecular replacement only with the “corrected data” was due to removal of the translation peak in the Patterson map. Refinement was therefore continued with the uncorrected data set and the twin operator −h − k, k, −l, twinning fraction ∼0.5, and proceeded smoothly. Throughout refinement, tight to moderate NCS restraints were imposed on the three polypeptide chains. After Rfree had dropped below 30%, solvent molecules and the substrate were included in the model. The refinement converged at an R-factor/Rfree of 20.4/25.5%. The final model contains 3 × 535 amino acid residues, 3 FADs, 3 aklavinone molecules, 18 bound sulfate ions, and 199 water molecules. Details of the model statistics are given in Table 1. The quality of the protein models was analyzed with the program PROCHECK45 and Coot.44 Sequence comparisons were carried out with the programs BLAST46 and ClustalW,47 and figures were generated with the program PyMOL.48 Accession numbers The crystallographic data have been deposited in the PDB with the accession code 3IHG.
Structure determination
Acknowledgements Selenomethionine-substituted enzyme samples did not result in crystals useful for data collection, and in spite of numerous trials, no heavy-metal derivative was found. Attempts to solve the structure by molecular replacement, using the atomic coordinates of PHHY,22 PgaE, CabE,24 or RebC25 as search models, including a “core” model, failed. This “core” search model had been constructed by superposition of the closest members of the sequence family RebC (29% sequence identity to RdmE, PDB code 2r0c), PgaE (28% identity, PDB code 2qa1), and CabE (28% identity, PDB code 2qa2) and retaining only those parts that were identical in structure in all four enzymes. The side chains of conserved amino acids were retained; all other residues were truncated to alanine. The native Patterson map contains a strong pseudotranslation peak at 2/3, 1/3, 0 (30% of the origin peak), which, together with a somewhat streaky diffraction pattern, led us to suspect that RdmE crystals may suffer from lattice-translocation defects.41 The diffraction data were corrected for this effect using the procedure outlined by Wang et al.41 The corrected data were then used for † http://nihserver.mbi.ucla.edu/twinning/
We gratefully acknowledge access to synchrotron radiation at the European Molecular Biology Laboratory outstation at Deutsches Elektronen-Synchrotron, Hamburg; European Synchrotron Radiation Facility, France; and MAX-laboratory, Sweden. This work was supported by the Swedish Research Council (G.S.) and the Finnish Academy (grants 210576 and 121688).
Supplementary Data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/ j.jmb.2009.09.003
References 1. Staunton, J. & Weissman, K. J. (2001). Polyketide biosynthesis: a millennium review. Nat. Prod. Rep. 18, 380–416.
976 2. Schneider, G. (2005). Enzymes in the biosynthesis of aromatic polyketide antibiotics. Curr. Opin. Struct. Biol. 15, 629–636. 3. Fischbach, M. A. & Walsh, C. T. (2006). Assembly-line enzymology for polyketide and nonribosomal peptide antibiotics: logic, machinery, and mechanisms. Chem. Rev. 106, 3468–3496. 4. Hertweck, C., Luzhetskyy, A., Rebets, Y. & Bechthold, A. (2007). Type II polyketide synthases: gaining a deeper insight into enzymatic teamwork. Nat. Prod. Rep. 24, 162–190. 5. Metsä-Ketelä, M., Niemi, J., Mäntsälä, P. & Schneider, G. (2008). Anthracycline biosynthesis: genes, enzymes and mechanisms. Top. Curr. Chem. 282, 101–140. 6. Niemi, J. & Mäntsälä, P. (1995). Nucleotide sequences and expression of genes from Streptomyces purpurascens that cause the production of new anthracyclines in Streptomyces galilaeus. J. Bacteriol. 177, 2942–2945. 7. Niemi, J., Ylihonko, K., Hakala, Pärssinen, R., Kopio, A. & Mäntsälä, P. (1994). Hybrid anthracycline antibiotics: production of new anthracyclines by cloned genes from Streptomyces purpurascens in Streptomyces galilaeus. Microbiology, 140, 1351–1358. 8. Sultana, A., Kallio, K., Jansson, A., Wang, J. S., Niem, J., Mäntsälä, P. & Schneider, G. (2004). Structure of the polyketide cyclase SnoaL reveals a novel mechanism for enzymatic aldol condensation. EMBO J. 23, 1911–1921. 9. Kallio, P., Sultana, A., Niemi, J., Mäntsälä, P. & Schneider, G. (2006). Crystal structure of the polyketide cyclase AknH with bound substrate and product analogue: implications for catalytic mechanism and product stereoselectivity. J. Mol. Biol. 357, 210–220. 10. Wang, Y., Niemi, J., Airas, K., Ylihonko, K., Hakala, J. & Mäntsälä, P. (2000). Modifications of aclacinomycin T by aclacinomycin methyl esterase (RdmC) and aclacinomycin-10-hydroxylase (RdmB) from Streptomyces purpurascens. Biochim. Biophys. Acta, 1480, 191–200. 11. Jansson, A., Niemi, J., Mäntsälä, P. & Schneider, G. (2003). Crystal structure of aclacinomycin methylesterase with bound product analogues: implications for anthracycline recognition and mechanism. J. Biol. Chem. 278, 39006–39013. 12. Jansson, A., Koskiniemi, H., Erola, A., Wang, J., Mäntsälä, P., Schneider, G. & Niemi, J. (2005). Aclacinomycin-10-hydroxylase is a novel substrateassisted hydroxylase requiring S-adenosyl-L-methionine as cofactor. J. Biol. Chem. 280, 3636–3644. 13. Niemi, J., Wang, Y., Airas, K., Ylihonko, K., Hakala, J. & Mäntsälä, P. (1999). Characterization of aklavinone-11hydroxylase from Streptomyces purpurascens. Biochim. Biophys. Acta, 1430, 57–64. 14. Hong, Y.-S., Hwang, C. K., Hong, S.-K., Kim, Y. H. & Lee, J. J. (1994). Molecular cloning and characterization of the aklavinone 11-hydroxylase gene of Streptomyces peucetius subsp. caesius ATCC 27952. J. Bacteriol. 176, 7096–7101. 15. Filippini, S., Solinas, M. M., Breme, U., Schlüter, M. B., Gabellini, D., Biamonti, G. et al. (1995). Streptomyces peucetius daunorubicin biosynthesis gene, dnrF: sequence and heterologous expression. Microbiology, 141, 1007–1016. 16. Mattevi, A. (1998). The PHBH fold: not only flavoenzymes. Biophys. Chem. 70, 217–222. 17. Entsch, B., Cole, L. J. & Ballou, D. P. (2005). Protein dynamics and electrostatics in the function of phydroxybenzoate hydroxylase. Arch. Biochem. Biophys. 433, 297–311.
Aklavinone-11-Hydroxylase Ternary Complex 18. Schreuder, H. A., Prick, P. A., Wierenga, R. K., Vriend, G., Wilson, K. S., Hol, W. G. & Drenth, J. (1989). Crystal structure of the p-hydroxybenzoate hydroxylase–substrate complex refined at 1.9 Å resolution. Analysis of the enzyme–substrate and enzyme– product complexes. J. Mol. Biol. 208, 679–696. 19. Schreuder, H. A., Mattevi, A., Obmolova, G., Kalk, K. H., Hol, W. G., van der Bolt, F. J. & van Berkel, W. J. (1994). Crystal structures of wild-type phydroxybenzoate hydroxylase complexed with 4aminobenzoate, 2,4-dihydroxybenzoate, and 2-hydroxy-4-aminobenzoate and of the Tyr222Ala mutant complexes with 2-hydroxy-4-aminobenzoate. Evidence for a proton channel and a new binding mode of the flavin ring. Biochemistry, 33, 10161–10170. 20. Gatti, D. L., Palfey, B. A., Lah, M. S., Entsch, B., Massey, V., Ballou, D. P. & Ludwig, M. L. (1994). The mobile flavin of 4-OH benzoate hydroxylase. Science, 266, 110–114. 21. Enroth, C., Neujahr, H., Schneider, G. & Lindqvist, Y. (1998). The crystal structure of phenol hydroxylase in complex with FAD and phenol provides evidence for a concerted conformational change in the enzyme and its cofactor during catalysis. Structure, 6, 605–617. 22. Enroth, C. (2003). High-resolution structure of phenol hydroxylase and correction of sequence errors. Acta Crystallogr., Sect. D: Biol. Crystallogr. 59, 1597–1602. 23. Hiromoto, T., Fujiwara, S., Hosokawa, K. & Yamaguchi, H. (2006). Crystal structure of 3-hydroxybenxoate hydroxylase from Comamonas testosteroni has a large tunnel for substrate and oxygen access to the active site. J. Mol. Biol. 364, 878–896. 24. Koskiniemi, H., Metsä-Ketelä, M., Dobritzsch, D., Kallio, P., Korhonen, H., Mäntsälä, P. et al. (2007). Crystal structures of two aromatic hydroxylases involved in the early tailoring steps of angucycline biosynthesis. J. Mol. Biol. 372, 633–648. 25. Ryan, K. S., Howard-Jones, A. R., Hamill, M. J., Elliott, S. J., Walsh, C. T. & Drennan, C. L. (2007). Crystallographic trapping in the rebeccamycin biosynthetic enzyme RebC. Proc. Natl Acad. Sci. USA, 104, 15311–15316. 26. Ryan, K. S., Chakraborty, C., Howard-Jones, A. R., Walsh, C. T., Ballou, D. P. & Drennan, C. L. (2008). The FAD cofactor of RebC shifts to an IN conformation upon flavin reduction. Biochemistry, 47, 13506–13513. 27. Xu, D., Ballou, D. P. & Massey, V. (2001). Studies of the mechanism of phenol hydroxylase: mutants Tyr289Phe, Asp54Asn and Arg281Met. Biochemistry, 40, 12369–12378. 28. Xu, D., Enroth, C., Lindqvist, Y., Ballou, D. P. & Massey, V. (2002). Studies of the mechanism of phenol hydroxylase: effect of mutation of proline 364 to serine. Biochemistry, 41, 13627–13636. 29. Beam, M. P., Bosserman, M. A., Noinaj, N., Wehenkel, M. & Rohr, J. (2009). Crystal structure of Bayer– Villiger monooxygenase MtmOIV, the key enzyme of the mithramycin biosynthetic pathway. Biochemistry, 48, 4476–4487. 30. Holm, L. & Sander, C. (1994). Searching protein structure databases has come of age. Proteins, 19, 165–173. 31. Wierenga, R. K., Terpstra, P. & Hol, W. G. J. (1986). Prediction of the occurrence of the ADP-binding βαβfold in proteins, using an amino acid sequence fingerprint. J. Mol. Biol. 187, 101–107. 32. Eppink, M. H. M., Schreuder, H. A. & van Berkel, W. J. H. (1997). Identification of a novel conserved sequence motif in flavoprotein hydroxylases with a
Aklavinone-11-Hydroxylase Ternary Complex
33.
34.
35.
36.
37.
38.
39.
putative dual function in FAD/NAD(P)H binding. Protein Sci. 6, 2454–2458. Vagin, A. A., Richelle, J. & Wodak, S. J. (1999). SFCHECK: a unified set of procedure for evaluating the quality of macromolecular structure-factor data and their agreement with atomic model. Acta Crystallogr., Sect. D: Biol. Crystallogr. 55, 191–205. Collaborative Computational Project, Number 4. (1994). The CCP4 suite: programs for protein crystallography. Acta Crystallogr., Sect. D: Biol. Crystallogr. 50, 760–763. Wang, J., Ortiz-Maldonado, M., Entsch, B., Massey, V., Ballou, D. & Gatti, D. L. (2002). Protein and ligand dynamics in 4-hydroxybenzoate hydroxylase. Proc. Natl Acad. Sci. USA, 99, 608–613. Ortiz-Maldonado, M., Entsch, B. & Ballou, D. P. (2003). Conformational changes combined with charge–transfer interactions are essential for reduction in catalysis by p-hydroxybenzoate hydroxylase. Biochemistry, 42, 11234–11242. Palfey, B. A., Moran, G. R., Entsch, B., Ballou, D. P. & Massey, V. (1999). Substrate recognition by “password” in p-hydroxybenzoate hydroxylase. Biochemistry, 38, 1153–1158. Schreuder, H. A., Hol, W. G. & Drenth, J. (1988). Molecular modeling reveals the possible importance of a carbonyl oxygen binding pocket for the catalytic mechanism of p-hydroxybenzoate hydroxylase. J. Biol. Chem. 263, 3131–3136. Aslanidis, C. & de Jong, P. J. (1990). Ligationindependent cloning of PCR products (LIC-PCR). Nucleic Acids Res. 18, 6069–6074.
977 40. Yeates, T. O. (1997). Detecting and overcoming crystal twinning. Methods Enzymol. 276, 344–358. 41. Wang, J., Kamtekar, S., Berman, A. J. & Steitz, T. A. (2005). Correction of X-ray intensities from single crystals containing lattice-translocation defects. Acta Crystallogr., Sect. D: Biol. Crystallogr. 61, 67–74. 42. McCoy, A. J., Grosse-Kunstleve, R. W., Adams, P. D., Winn, M. D., Storoni, L. C. & Read, R. J. (2007). Phaser crystallographic software. J. Appl. Crystallogr. 40, 658–674. 43. Murshudov, G. N., Vagin, A. A. & Dodson, E. J. (1997). Refinement of macromolecular structures by the maximum-likelihood method. Acta Crystallogr., Sect. D: Biol. Crystallogr. 53, 240–255. 44. Emsley, P. & Cowtan, K. (2004). Coot: model-building tools for molecular graphics. Acta Crystallogr., Sect. D: Biol. Crystallogr. 60, 2126–2132. 45. Laskowski, R. A., MacArthur, M. W., Moss, D. S. & Thornton, J. M. (1993). PROCHECK: a program to check the stereochemical quality of protein structures. J. Appl. Crystallogr. 26, 283–291. 46. Altschul, S. F., Gish, W., Miller, W., Myers, E. W. & Lipman, D. J. (1990). Basic local alignment search tool. J. Mol. Biol. 215, 403–410. 47. Thompson, J. D., Higgins, D. G. & Gibson, T. J. (1994). CLUSTAL W: improving the sensitivity of progressive multiple sequence alignment through sequence weighting, position-specific gap penalties and weight matrix choice. Nucleic Acids Res. 22, 4673–4680. 48. DeLano, W. L. (2002). The PyMOL Molecular Graphics System, http://www.pymol.org.
J. Mol. Biol. (2009) 393, 966–977
Available online at www.sciencedirect.com
Structural Basis for Substrate Recognition and Specificity in Aklavinone-11-Hydroxylase from Rhodomycin Biosynthesis Ylva Lindqvist 1 , Hanna Koskiniemi 1 , Anna Jansson 1 , Tatyana Sandalova 1 , Robert Schnell 1 , Zhanliang Liu 2 , Pekka Mäntsälä 2 , Jarmo Niemi 2 and Gunter Schneider 1 ⁎ 1
Department of Medical Biochemistry and Biophysics, Karolinska Institutet, S-171 77 Stockholm, Sweden 2
Department of Biochemistry and Food Chemistry, University of Turku, FIN-20014 Turku, Finland Received 30 July 2009; received in revised form 28 August 2009; accepted 2 September 2009 Available online 8 September 2009
In the biosynthesis of several anthracyclines, aromatic polyketides produced by many Streptomyces species, the aglycone core is modified by a specific flavin adenine dinucleotide (FAD)- and NAD(P)H-dependent aklavinone-11-hydroxylase. Here, we report the crystal structure of a ternary complex of this enzyme from Streptomyces purpurascens, RdmE, with FAD and the substrate aklavinone. The enzyme is built up of three domains, a FAD-binding domain, a domain involved in substrate binding, and a Cterminal thioredoxin-like domain of unknown function. RdmE exhibits structural similarity to aromatic hydroxylases from the p-hydroxybenzoate hydroxylase family, but unlike most other related enzymes, RdmE is a monomer. The substrate is bound in a hydrophobic pocket in the interior of the enzyme, and access to this pocket is provided through a different route than for the isoalloxazine ring of FAD—the backside of the ligand binding cleft. The architecture of the substrate binding pocket and the observed enzyme–aklavinone interactions provide a structural explanation for the specificity of the enzyme for non-glycosylated substrates with C9-R stereochemistry. The isoalloxazine ring of the flavin cofactor is bound in the “out” conformation but can be modeled in the “in” conformation without invoking large conformational changes of the enzyme. This model places the flavin ring in a position suitable for catalysis, almost perpendicular to the tetracyclic ring system of the substrate and with a distance of the C4a carbon atom of the isoalloxazine ring to the C-11 carbon atom of the substrate of 4.8 Å. The structure suggested that a Tyr224-Arg373 pair might be involved in proton abstraction at the C-6 hydroxyl group, thereby increasing the nucleophilicity of the aromatic ring system and facilitating electrophilic attack by the perhydroxy-flavin intermediate. Replacement of Tyr224 by phenylalanine results in inactive enzyme, whereas mutants at position Arg373 retain catalytic activity close to wildtype level. These data establish an essential role of residue Tyr224 in catalysis, possibly in aligning the substrate in a position suitable for catalysis. © 2009 Elsevier Ltd. All rights reserved.
Edited by R. Huber
Keywords: flavoenzyme; anthracycline; polyketide biosynthesis; aromatic hydroxylase; merohedral twinning
*Corresponding author. E-mail address: [email protected]. Present address: Z. Liu, Department of Plant Pathology, Agricultural University of Hebei, Baoding 071001, China. Abbreviations used: FAD, flavin adenine dinucleotide; PHHY, phenol hydroxylase; pHBH, p-hydroxybenzoate hydroxylase; MHBH, 3-hydroxybenzoate hydroxylase; RebC, rebeccamycin biosynthetic enzyme; PgaE, angucycline biosynthetic enzyme; PDB, Protein Data Bank; TEV, tobacco etch virus. 0022-2836/$ - see front matter © 2009 Elsevier Ltd. All rights reserved.
Aklavinone-11-Hydroxylase Ternary Complex
Introduction Anthracyclines are aromatic polyketides that contain a common tetrahydrotetracene-5,12-quinone core. These metabolites are synthesized in vivo by complex enzymatic pathways through a process reminiscent of fatty acid biosynthesis.1–5 A type II, iterative polyketide synthase complex assembles the polyketide chain from acetyl- and malonyl-coenzyme A units, which is subsequently cyclized by specific aromatases and/or cyclases. In most of these pathways, chain elongation by polyketide synthase and subsequent β-keto-reduction and cyclization events result in the formation of aklavinone (see Fig. 1 for structure and atom numbering), a common precursor of many anthracyclines. This metabolic intermediate is further modified by a number of enzymes in so-called tailoring steps, which include hydroxylation, oxidation, dehydrogenation, methylation, and dehydration and which result in differences in the substitution patterns of the polyketide core, thus generating the remarkable chemical diversity of the anthracyclines.4,5 A modification common to all anthracyclines is glycosylation, generally at position C-7, as it is a prerequisite for biological activity. Many of the anthracyclines have medically important properties, and several are in clinical use, for instance, as anticancer drugs (doxorubicin, daunorubicin).1 The rdm gene cluster of Streptomyces purpurascens contains the genes that code for proteins of rhodomycin biosynthesis.6,7 These enzymes include RdmA, an aklanonic acid methyl ester cyclase that is a member of the cyclase family represented by SnoaL and AknH,8,9 and the tailoring enzymes RdmB, RdmC, and RdmE. RdmC is a methylesterase that catalyzes removal of the methoxygroup at the C-15 position of aklavinone,10,11 and RdmB is a novel S-adenosyl-methionine-dependent hydroxylase that converts 15-demethoxy-ɛ-rhodomycin glycoside to β-rhodomycin glycoside.10,12 RdmE encodes an aklavinone-11-hydroxylase, a NAD (P)H- and flavin adenine dinucleotide (FAD)-dependent monooxygenase.13 The enzyme catalyzes the incorporation of a hydroxyl group at position C-11 of aklavinone, resulting in ɛ-rhodomycinone (Fig. 1). This tailoring reaction is shared between biosynthetic pathways leading to different anthracyclines, for instance, daunorubicin in S. peucetius14,15 and rhodo-
967 mycin in S. purpurascens.13 Aklavinone hydroxylases act before the glycosylation step; that is, they cannot accept glycosylated substrates. RdmE is also specific for the C-9R configuration of anthracyclines and does not hydroxylate compounds with C-9S configuration.9 Aklavinone hydroxylases consist of a polypeptide chain of about 500–540 amino acids, with a molecular mass in the range of 52–55 kDa, and show weak to moderate sequence similarities (18–30% amino acid identity) to the p-hydroxybenzoate hydroxylase (pHBH) family of flavoenzymes.16 The first insights into the structural basis of catalysis in this family of aromatic hydroxylases were obtained from studies of pHBH.17–20 These studies were subsequently extended to include phenol hydroxylase (PHHY),21,22 3-hydroxybenzoate hydroxylase (MHBH),23 angucycline biosynthetic enzyme (PgaE)/CabE,24 and rebeccamycin biosynthetic enzyme (RebC).25,26 An intriguing feature common to these enzymes is the two conformations of the FAD cofactor that can be associated with the two half-reactions, reduction of the FAD cofactor by NAD(P)H and oxidation of the substrate using molecular oxygen. In a number of crystal structures of these enzymes, FAD is observed in the so-called “in” conformation (or closed conformation), in which the isoalloxazine ring lies within the active site and the substrate binding pocket is inaccessible from the bulk solvent.18,21,24,25 The reactions with molecular oxygen and the hydroxylation of the aromatic substrate are expected to occur while FAD is in this conformation. In order to be reduced by NAD(P)H, the isoalloxazine ring of FAD has to move from the active site towards the solvent. This conformation is termed the “out” conformation and has been observed in crystal structures of several members of the pHBH family.19–21,23,26 In this article, we report the determination of the three-dimensional structure of a ternary complex of RdmE with bound FAD and aklavinone by means of X-ray crystallography to a resolution of 2.5 Å. The crystal structure reveals the basis for the stereospecific recognition of aklavinone and for the inability for the aklavinone glycoside to act as substrate. It also identifies Tyr224 as an important residue in the interaction with the substrate and essential for catalysis, which is confirmed by site-directed mutagenesis. Structural comparison of RdmE with other
Fig. 1. Overall reaction catalyzed by aklavinone-11-hydroxylases. Numbering of the carbon atoms and labeling of the tetracyclic ring system of aklavinone are also shown.
968 aromatic hydroxylases, in particular with pHBH,17 PHHY,21,27,28 and RebC, 25,26 reveals that while overall features of structure and catalytic function appear to hold for all members of this family, there are intriguing differences among these enzymes with regard to substrate entry/product release, conformational changes during catalysis, and catalytic features.
Results and Discussion Structure determination The structure of RdmE was determined by molecular replacement from merohedrally twinned crystals using search models based on the three most related enzymes, all with sequence identities to RdmE of less than 30% [Protein Data Bank (PDB) codes 2r0c, 2qa1, and 2qa2]. The search models were truncated, retaining only conserved structural elements. Initial molecular replacement trials using different programs were unsuccessful; only after removal of the strong peak at 2/3, 1/3, 0 in the native Patterson map was a correct solution showing an excellent signal in the P63 space group obtained. The quality of the refined electron density maps and model and refinement statistics confirm that this solution is correct. The electron density map is continuous for the whole polypeptide chain, and with a few exceptions, all amino acids, from residue 1 to residue 535, are well defined in electron density. In particular, the electron density for the bound FAD and aklavinone is well defined and allows unambiguous placement of these molecules. Less welldefined electron density is found for residues 151– 155, and a few solvent-exposed side chains are too flexible to be observed. The overall quality of the model of RdmE is as expected for this resolution, and the majority (95%) of the residues are found in the most favored region of the Ramachandran plot (Table 1). Overall structure of the enzyme The monomer of RdmE consists of three domains (Fig. 2), the FAD-binding domain (amino acids 1–71, 99–198, and 289–393), the middle domain (amino acids 72–98 and 199–288), and the C-terminal domain (residues 394–535). The core of the molecule is formed by the FAD-binding domain, which is dominated by a Rossmann-type βαβ-fold comprising a mixed, six-stranded βsheet, flanked on one side by a small α-helix and a three-stranded antiparallel β-sheet (Fig. 2). The other side is lined by four α-helices, of which the longest helix (residues 345–373) spans the whole length of the domain. An important additional secondary-structure element is the three-stranded antiparallel β-sheet, which contributes to the formation of the FAD-binding groove close to the isoalloxazine ring.
Aklavinone-11-Hydroxylase Ternary Complex Table 1. Statistics of data collection and crystallographic refinement Data collection Resolution (Å) Beamline λ (Å) Number of observations Number of unique reflections Rsym (%) Completeness (%) I/σ(I) Multiplicity Wilson B-factor (Å2) Refinement Resolution interval (Å) Rwork (%) Rfree (%) Number of amino acids Number of atoms Protein FAD Aklavinone Water/sulfate B-factor (Å2) Protein FAD Aklavinone Water molecules rmsd from ideal geometry Bond length (Å) Bond angle (°) Refined twin fraction Ramachandran plot (%) Residues in most favored region Residues in allowed regions Residues in disallowed regions
50–2.5 (2.62–2.5) 7–11, MAX-laboratory 1.01590 351,856 (22,851) 64,506 (7778) 8.2 (23.5) 97.1 (80.8) 16.0 (4.7) 5.5 (2.9) 38.7 30.0–2.5 20.4 25.5 3 × 535 12,191 159 90 199/90 23.0 19.3 23.6 11.3 0.019 1.877 0.49 95.0 4.3 0.7
Values in parentheses are for the highest-resolution shell.
The middle domain is inserted into the FADbinding domain and folds into a seven-stranded antiparallel β-sheet, which is flanked by an α-helix. The active site is located at the interface between the FAD-binding domain and the middle domain. The latter forms the roof of the substrate binding pocket. The C-terminal domain displays a typical thioredoxin fold, as in other members of this enzyme family with the exception of pHBH, which lacks this domain. While the C-terminal domain participates in quaternary-structure formation in MHBH and PHHY, it is not part of the dimer interface in PgaE, CabE,24 and the Bayer–Villiger monooxygenase MtmOIV from the mithramycin biosynthetic pathway.29 Quaternary structure The asymmetric unit of the RdmE crystals contains three molecules. The interface areas between these molecules are in the range of 230– 320 Å2, typical in size of crystal contacts rather than protein–protein interfaces. Larger interfaces (approximately 920 Å2) are formed with molecules related by crystallographic symmetry, which lead to the formation of a trimer along the 6-fold screw axis. Although these values approach the range of protein–protein interfaces, we consider these interactions as crystal packing effects rather than
Aklavinone-11-Hydroxylase Ternary Complex
969
Fig. 2. Overall structure of RdmE. The FAD domain is shown in green, the middle domain in red, and the C-terminal domain in yellow. The bound ligands in the ternary complex, FAD (light blue) and aklavinone (dark blue), are shown as stick models.
displaying biologically relevant assemblies. Analytical gel-filtration data also suggest a monomeric structure of RdmE in solution (data not shown). Relation to the pHBH family Sequence comparisons suggested that RdmE is a member of the pHBH family of flavoenzymes. A search of the PDB with the RdmE coordinates using DALI30 returned a number of related structures, with the most similar being the flavin-dependent monooxygenases RebC,25,26 PHHY,21 and MHBH23 with Z-scores of 40.8, 36.9, and 35.5, respectively. Superposition of RdmE with the subunit of RebC (PDB code 2r0p) gave an rmsd value of 1.9 Å for 442 aligned residues. This value is only slightly larger than the rmsd values for the superposition of individual domains (1.6, 1.9, and 1.7 Å), showing that there are no larger differences in domain orientation between these two proteins. This finding also holds for other relatives of this enzyme family; that is, structural comparisons of RdmE with PHHY, MHBH, and PgaE return similar rmsd values and domain orientations. A structure-based sequence alignment of RdmE with other members of this sequence family of known structure, RebC, PHHY, PgaE, CabE, pHBH, and MBBH, results in pairwise sequence identities in the range 18–30%, with 37 invariant residues that are mostly involved in flavin binding or inter-domain interactions. The FAD binding site The cofactor is bound in the active site in the “out” conformation, with the ADP and ribityl moieties
embedded in the FAD-binding domain and the isoalloxazine ring located at the interface between the FAD-binding domain and the middle domain. FAD is tightly anchored to the enzyme by a significant number of direct or water-mediated hydrogen bonds to protein residues (Fig. 3). Most of the FAD-binding residues are well conserved in the pHBH family. Gln119, Asp308, and Asn322 are invariant (the latter residue binds FAD only in the “in” conformation), whereas Arg36 and Arg45 are often conservatively exchanged to lysine or glutamine, and Glu35 is replaced by aspartate in MHBH. This residue belongs to the dinucleotide binding motif that is conserved among flavin- and NAD(P) H-utilizing proteins31 and in the pHBH family.32 In the ternary complex of RdmE, the FAD-ring system is observed in the “out” conformation. The isoalloxazine ring of FAD is wedged between the side chain of Trp288 on one side and the polypeptide chain Arg45-Ala46, and in particular the aliphatic part of the side chain of Arg45, on the other side. There are no hydrogen bonds from the protein to the ring system that could stabilize this conformation further unlike in PHHY (subunit A) and MHBH, where a hydrogen bond between the N3 nitrogen atom of FAD to the side chain of a tyrosine residue was observed.21,23 This residue is replaced by a phenylalanine (Phe246) in RdmE, unable to engage in a similar interaction. There is, however, a hydrogen bond between the N3 nitrogen atom of the flavin ring and the oxygen atom of ring B of the substrate aklavinone, and it is presumably this interaction that stabilizes the flavin ring in the “out” conformation. A similar substrate–N3 flavin hydrogen bond has been found in the complex of
970
Aklavinone-11-Hydroxylase Ternary Complex
Fig. 3. Stereoview of the FAD binding site in RdmE. The carbon atoms of FAD are shown in green, and protein carbon atoms are shown in yellow. Hydrogen bonds are indicated by broken lines.
pHBH–2,4 dihydroxybenzoate, and also in this case, flavin was found in the “out” conformation.19,20 Structural basis of substrate recognition and specificity At the interface between the FAD- and substratebinding domains, in the vicinity of the isoalloxazine ring of FAD, there is a hydrophobic pocket that contains the bound substrate aklavinone (Fig. 4a). The location of the substrate binding pocket is similar to that of other members of the pHBH family, and the positions of the bound substrates overlap well. The tetracyclic ring system of aklavinone is bound in a rather narrow cleft, which is lined by predominantly hydrophobic amino acids (Fig. 4b). One side of the aklavinone ring system is packed against the aromatic side chain Trp222 and flanked by Phe79 and Phe246, and the other side packs against the wall of the active-site pocket defined by the polypeptide stretch Pro315-Thr316-Gly317-Gly318. The edge of ring D of the substrate is adjacent to another hydrophobic patch in the substrate pocket defined by the side chains of Met202, Thr233, Trp288, and Met290. The hydroxyl group of ring B of aklavinone forms a hydrogen bond to the side chain of Tyr224, and the C-12 carbonyl group of ring C is anchored to the main-chain nitrogen atom of Ala47 via a watermediated hydrogen-bond interaction. The same oxygen atom is hydrogen bonded to the N3 nitrogen atom of the isoalloxazine ring (see above). Ring A of aklavinone, which is observed in halfchair conformation, interacts via the C-7 hydroxyl group with the side chain of Arg373 through an indirect hydrogen bond involving a water molecule. The C-9 hydroxyl also interacts indirectly via a
bridging water molecule with the carbonyl oxygen atom of Gly318, and the ethyl substituent points into a hydrophobic pocket created by Phe79, Ile81, Met102, Trp114, and Met116 (Fig. 4b). These features seem to be responsible for the stereospecific recognition of the C9-R isomer of aklavinone by RdmE. Reversal of the stereochemistry at this position would result in a mismatch of the physicochemical properties of the substituents and their respective binding pockets; that is, hydrophilic groups of the substrate would point into hydrophobic patches on the enzyme and vice versa. This should disfavor binding of the S-isomer. Indeed, nogalaviketone, which differs from aklavinone by reversal of the stereochemistry at the C-9 carbon, carbonyl instead of hydroxyl group at C-7, and replacement of the ethyl by a methyl substituent, is not hydroxylated by RdmE, whereas its C9-R isomer auraviketone is a substrate.9 In contrast to other tailoring enzymes in anthracycline biosynthesis, RdmE does not accept aklavinone substrates glycosylated at position C-7. In the structure of the RdmE–aklavinone complex, there is a pocket adjacent to the C-7 hydroxyl group, which extends to the surface of the enzyme. However, modeling of bound aklavinone glycoside resulted in severe clashes of carbohydrate atoms with enzyme residues, for instance, the side chain of Ile81, Val95, Tyr224, or Arg373. Thus, this pocket is too small in size to accommodate the carbohydrate moiety of aklavinone glycosides, which leads to hydroxylation of only non-glycosylated polyketide substrates. Site-directed mutagenesis Residues Tyr224 and Arg373 are among the few polar amino acids in close vicinity of the substrate.
971
Aklavinone-11-Hydroxylase Ternary Complex
Fig. 4. Substrate binding site in RdmE. (a) Part of an omit electron density map, calculated with the algorithm of Vagin et al.,33 implemented in SFCHECK in the CCP4 program package,34 at the active site of RdmE, contoured at 1.2 σ. The refined position of bound substrate aklavinone is superposed. AKV denotes the bound substrate aklavinone. (b) Stereoview of the active site in the ternary complex of RdmE with FAD and aklavinone. Amino acids lining the wall of the substrate binding pocket are shown. Residues whose function has been probed by site-directed mutagenesis are indicated with carbon atoms in magenta. FAD is drawn with carbon atoms in green, and aklavinone is drawn with carbon atoms in blue.
We therefore changed these residues to a phenylalanine (Tyr224Phe) and alanine, glutamine, and methionine (Arg373Ala, Arg373Gln, and Arg373Met), respectively, by site-directed mutagenesis to probe their role in catalysis. Substitutions at the position Arg373 resulted in mutants with similar specific activities as the wild-type enzyme, whereas replacement of Tyr224 with a phenylalanine abolished catalytic activity (Table 2).
Table 2. Specific activities of wild-type and mutant RdmE Species Wild type Arg373Ala Arg373Gln Arg373Met Tyr224Phe
Vmax (nM/min mg)
Specific activity (%)
35.8 ± 1.2 36.2 ± 7.8 41.7 ± 3.5 19.2 ± 1.2 Not detectable
100 101 115 53 0
Substrate entry In PHHY and pHBH, access of the substrate to the active site is controlled by conformational changes that allow entry through a channel passing the re face of the flavin ring.21,35 In MHBH, the substrate binding pocket is rather open and would allow entry of the substrate by two different routes of access without the need for larger conformational changes.23 For RebC and RdmE, which use more bulky substrates, access to the active site via the flavin pocket is highly unlikely. For RebC, a path for substrate entry/product release has been proposed, which involves a disorder/order transition of a peptide stretch at the “backside” of the substrate pocket, opposite to the flavin ring.25 Folding of the disordered stretch into an α-helix after binding of substrate closes off the pocket and seals the ligand in its binding site. Interestingly, the largest differences between the structures of RdmE and RebC are
972 localized at this “backdoor” of the substrate pocket. Compared to RdmE, RebC has a 24-residue insertion that completely shields the substrate once the “melting” helix is folded. In RdmE, the polypeptide chain does not run across the backside completely, and there is an opening into the substrate pocket from this direction (Fig. 5). In the ternary complex, this opening is in part occluded by protein side chains, but small conformational changes at the side-chain level (Tyr224, Arg373, and Ile369) would allow passage of the aklavinone substrate to its binding pocket. We thus conclude that (i) while the general entry points for substrates into their binding site are conserved in RdmE and RebC, the precise mechanism in RebC appears more elaborate, and (ii) for both enzymes, it is different from the proposed channel adjacent to the flavin ring used by the smaller substrates phenol and p-hydroxybenzoate in PHHY and pHBH, respectively. Mechanistic implications Crystallographic, spectroscopic, and kinetic studies with pHBH, PHHY, and RebC concluded that catalysis in these hydroxylases can be summarized by the following overall steps (Fig. 6). In the absence of substrate, the enzymes most likely exist as binary complexes with oxidized FAD bound in two conformations in a dynamic equilibrium. The reduction of FAD by NAD(P)H in the absence of substrate is very slow, thus avoiding wasteful usage of reduction equivalents.25 Interaction of NAD(P)H with the ternary enzyme–FAD–substrate complex shifts the FAD to the “out” conformation where the isoalloxazine ring is in favorable position for hydride transfer.36 Release of NAD(P)H and shift of the reduced FAD back to the ”in” position comprise the next step, which is then followed by oxygen activation and hydroxylation of the substrate. The structure of the ternary complex of RdmE with bound substrate shows preferential binding of flavin in the “out” conformation also in the absence of NAD(P)H. This is in line with studies of PHHY where both the “in” and “out” conformations were observed in the presence of the substrate phenol and the absence of NAD(P)H21 and the structure of the substrate complex of MHBH,23 where flavin was observed in the “out” position. In these cases, this conformation is stabilized by hydrogen bonds to the isoalloxazine ring, either to enzyme residues (PHHY, MHBH) or to the substrate (RdmE). In these enzymes, FAD is thus poised for reaction with NAD(P)H, whereas in pHBH, deprotonation of the substrate and binding of NAD(P)H appear to trigger the shift from the “in” to the “out” conformation of the isoalloxazine ring.37 Reduction of FAD results in preferential binding of the flavin ring in the “in” conformation,20,26 associated with the enzymatic states during oxygen activation and hydroxylation reaction. For RdmE, we were not successful in obtaining an experimental structure with the flavin in the “in” conformation, and we therefore modeled this state based on the
Aklavinone-11-Hydroxylase Ternary Complex
structure described here and the isoalloxazine “in” conformation observed in the closest related enzyme, RebC.26 This conformation of FAD can be obtained straightforwardly by rotation around two bonds, C4′–C5′ and C1′–C2′. The conformation of the side chain of Trp288 was changed by a rotation around the Cα–Cβ bond in order to accommodate the isoalloxazine ring in the active site and to avoid unfavorable steric interactions. In this model, the isoalloxazine ring is almost perpendicular to the tetracyclic ring system of aklavinone, with the C4a atom of FAD about 4.8 Å away from the C-11 of the substrate (Fig. 7). The putative flavin 4a-hydroperoxide intermediate would place the hydroxyl group within 2.8 Å of C-11, the target of the electrophilic attack. A model of the hydroperoxide intermediate in the active site of RdmE also shows the characteristic hydrogen-bond interaction of the O-4 atom of the isoalloxazine ring with the main-chain nitrogen of residue Ala-47 that might promote formation of the hydroperoxide intermediate during catalysis as proposed for pHBH.38 The electrophilic attack of the hydroperoxide intermediate onto the aromatic ring system of the substrate could be further facilitated by an increase in electron density due to deprotonation of the substrate as shown for pHBH.37 In RdmE, the C-6 hydroxyl group of aklavinone forms a hydrogen bond to the side chain of Tyr224, which, in turn, is within hydrogen-bonding distance to the side chain of Arg373. We hypothesized that the close proximity of the positive charge of Arg373 and the tyrosine side chain might shift the pKa of Tyr224 such that it would be able to act as a catalytic base in the deprotonation of the substrate. The high catalytic activities of the mutants at position 373 (Table 2) suggest, however, that the arginine side chain is not required for catalysis, which makes this mechanistic scenario less likely. On the other hand, the mutagenesis data clearly indicate that Tyr224 plays an essential role in catalysis, possibly by orienting the substrate in a position suitable for catalysis.
Fig. 5. Surface representation of RdmE illustrating the route of access of the active site by the substrate. The bound aklavinone is shown as yellow stick model. The electrostatic potential is contoured from −25 to +25 kT.
Aklavinone-11-Hydroxylase Ternary Complex
Fig. 6. Proposed steps of the conversion of aklavinone to ɛ-rhodomycinone by RdmE.
973
974
Aklavinone-11-Hydroxylase Ternary Complex
Fig. 7. Comparison of the “out” conformation of FAD (light green) as observed in the crystal structure of the ternary complex of RdmE with the modeled “in” conformation (dark green). Modeling of the “in” state is based on the conformation of FAD as observed in subunit A of RebC (2rop).
Loss of the C4a hydroxyl group, that is, regeneration of the oxidized flavin ring, requires protonation of the hydroxyl in order to generate a better leaving group. There is no enzyme group in sufficient close vicinity to act as catalytic acid, but there are several well-defined water molecules in the proximity of the modeled position of the perhydroxy group that could participate in proton transfer.
Materials and Methods Protein production and preparation of substrate Recombinant RdmE was produced in Streptomyces lividans TK24 and purified with some modifications to the procedure described.13 The medium used was Tryptone Soya Broth (Oxoid); only anion exchange using a HiPrep 16/10 Q XL column [0–1 mol/l NaCl gradient in 50 mmol/l Tris–HCl, pH 7.5, 0.5 mM dithiothreitol (DTT), and 10% glycerol] and gel exclusion chromatography using a HiLoad 26/60 Superdex 200 column (50 mmol/ l Tris–HCl, pH 7.5, 0.5 mM DTT, and 10% glycerol) were performed in an Äkta FPLC system (Amersham Pharmacia). Protein samples were concentrated and stored at −20 °C in gel exclusion buffer. Screening for production of soluble RdmE in Escherichia coli was carried out using constructs comprising fulllength [residues 2–535; the N-terminal methionine is replaced by a tobacco etch virus (TEV) cleavage site] and truncated sequences. The latter were designed based on sequence comparison and predicted domain borders. Coding sequences for these constructs were cloned by ligation-independent cloning39 into the expression vector pNIC28-Bsa4 (GenBank Accession No. EF198106), containing a T7/lacO promoter, an N-terminal His6 tag, and a TEV protease site. The presence of the insert was confirmed by DNA sequencing. The expression constructs were tested for soluble protein production in small scale using the expression strain BL21-Gold (DE3) (Stratagene)
carrying the pRARE2 plasmid from the Rosetta2 strain (Novagen). This procedure resulted in the identification of one construct (pNIC28-RdmE), comprising an N-terminal His6 tag, the TEV cleavage site, and amino acids Asn2Arg535 of RdmE that produced soluble protein in E. coli. For large-scale purification, this construct was expressed in the E. coli strain BL21-Gold (DE3) pRARE2. The bacterial cells were grown at 37 °C in TB medium supplemented with 35 μg/ml kanamycin and 20 μg/ml chloramphenicol. At an OD600 (optical density at 600 nm) value of 0.5–0.7, gene expression was induced by addition of 0.5 mM IPTG and the temperature was lowered to 18 °C. The cells were harvested ca 15 h after induction, and the bacterial cell walls were disrupted by lysozyme treatment and sonication. The clarified lysate was loaded on Ni-NTA resin (Qiagen) and eluted from the column with an imidazole gradient. The protein sample was stored until further use at −80 °C in 50 mM Tris buffer, pH 7.5, and 50 mM NaCl containing 5% (v/v) glycerol and 1 mM TCEP. Aklavinone was prepared using the protocol described previously.13 Site-directed mutagenesis The coding sequences for the active-site mutants Tyr224Phe, Arg373Ala, Arg373Gln, andArg373Met were produced by PCR using the plasmid pNIC28-RdmE as template and mutation-specific primers in combination with the original amplification primers. The DNA fragments carrying the mutations were cloned in the pNIC28Bsa4 vector, and their sequences were verified by DNA sequencing. Production of the mutant proteins in E. coli and purification to electrophoretic homogeneity followed the same protocol as outlined above for wild-type RdmE. Activity assay Enzymatic activity was determined using the assay described previously.13 Due to overlap of the absorption spectra of NADPH and the substrate aklavinone, the reaction was followed spectroscopically by monitoring the
975
Aklavinone-11-Hydroxylase Ternary Complex appearance of the product at 500 nm. The reaction mixture (total volume, 0.1 ml) consisted of 40 μM aklavinone (dissolved in 5 μl methanol), 10 μM FAD, and 3 mM NADPH in 100 mM potassium phosphate buffer, pH 7.5. The reaction was initiated by addition of the enzyme to a concentration of 1.0 μM, and the reaction was monitored for 30 min. All activity measurements were carried out in triplicate. Crystallization The ternary complex of recombinant RdmE, produced in S. lividans, with FAD and substrate was prepared by the addition of FAD (final concentration, 0.8 mM) and aklavinone (final concentration, 1.74 mM) to a solution of RdmE (2.2 mg/ml) in 20 mM Tris buffer, pH 7.5, containing 1 mM DTT. After several hours of incubation on ice, the protein was concentrated to 10 mg/ml. The complex was crystallized by the vapor diffusion method. Orange–yellow crystals of RdmE appeared after several days of equilibration at 20 °C in drops composed of 2 μl protein solution and 2 μl reservoir solution (20 mM Tris, pH 7.6, containing 2.1 M ammonium sulfate). Data collection The crystals of RdmE were cryo-protected by quick soaking in mother liquor containing 25% ethylenglycol (v/v). A native data set was collected to 2.5 Å resolution at beamline 7/11, MAX-laboratory, Sweden. The data set was processed with the program MOSFLM and scaled with SCALA from the CCP4 package.34 Detailed data collection statistics are given in Table 1. The sigmoidal cumulative intensity distribution indicated crystal twinning, and additional twinning indicators were obtained using the twinning server†.40 Crystals of RdmE belong to the space group P63 with unit cell dimensions of a = b = 183.0 Å and c = 99.6 Å. The asymmetric unit contains three molecules of RdmE, resulting in a Matthews coefficient of 2.6 Da/Å3 and a solvent content of approximately 53%.
structure determination by molecular replacement with the program Phaser42 and the “core” search model ensemble described above. A clear solution with the positions of the three molecules in the asymmetric unit was obtained, and during refinement using Refmac,43 electron density for FAD and the bound substrate appeared at the expected positions, strong evidence that the molecules were correctly placed in the crystal lattice. Model building was carried out using the program Coot,44 interspersed with refinement using Refmac.43 For determination of Rfree, 5% of the reflections were omitted from refinement. Comparisons of refinement revealed that the switch back to the uncorrected data set (but including merohedral twinning operators) resulted in better maps and statistics than using the corrected data. Refinement using the latter stalled at an Rfree value of 36%. Presumably, the reason for success with molecular replacement only with the “corrected data” was due to removal of the translation peak in the Patterson map. Refinement was therefore continued with the uncorrected data set and the twin operator −h − k, k, −l, twinning fraction ∼0.5, and proceeded smoothly. Throughout refinement, tight to moderate NCS restraints were imposed on the three polypeptide chains. After Rfree had dropped below 30%, solvent molecules and the substrate were included in the model. The refinement converged at an R-factor/Rfree of 20.4/25.5%. The final model contains 3 × 535 amino acid residues, 3 FADs, 3 aklavinone molecules, 18 bound sulfate ions, and 199 water molecules. Details of the model statistics are given in Table 1. The quality of the protein models was analyzed with the program PROCHECK45 and Coot.44 Sequence comparisons were carried out with the programs BLAST46 and ClustalW,47 and figures were generated with the program PyMOL.48 Accession numbers The crystallographic data have been deposited in the PDB with the accession code 3IHG.
Structure determination
Acknowledgements Selenomethionine-substituted enzyme samples did not result in crystals useful for data collection, and in spite of numerous trials, no heavy-metal derivative was found. Attempts to solve the structure by molecular replacement, using the atomic coordinates of PHHY,22 PgaE, CabE,24 or RebC25 as search models, including a “core” model, failed. This “core” search model had been constructed by superposition of the closest members of the sequence family RebC (29% sequence identity to RdmE, PDB code 2r0c), PgaE (28% identity, PDB code 2qa1), and CabE (28% identity, PDB code 2qa2) and retaining only those parts that were identical in structure in all four enzymes. The side chains of conserved amino acids were retained; all other residues were truncated to alanine. The native Patterson map contains a strong pseudotranslation peak at 2/3, 1/3, 0 (30% of the origin peak), which, together with a somewhat streaky diffraction pattern, led us to suspect that RdmE crystals may suffer from lattice-translocation defects.41 The diffraction data were corrected for this effect using the procedure outlined by Wang et al.41 The corrected data were then used for † http://nihserver.mbi.ucla.edu/twinning/
We gratefully acknowledge access to synchrotron radiation at the European Molecular Biology Laboratory outstation at Deutsches Elektronen-Synchrotron, Hamburg; European Synchrotron Radiation Facility, France; and MAX-laboratory, Sweden. This work was supported by the Swedish Research Council (G.S.) and the Finnish Academy (grants 210576 and 121688).
Supplementary Data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/ j.jmb.2009.09.003
References 1. Staunton, J. & Weissman, K. J. (2001). Polyketide biosynthesis: a millennium review. Nat. Prod. Rep. 18, 380–416.
976 2. Schneider, G. (2005). Enzymes in the biosynthesis of aromatic polyketide antibiotics. Curr. Opin. Struct. Biol. 15, 629–636. 3. Fischbach, M. A. & Walsh, C. T. (2006). Assembly-line enzymology for polyketide and nonribosomal peptide antibiotics: logic, machinery, and mechanisms. Chem. Rev. 106, 3468–3496. 4. Hertweck, C., Luzhetskyy, A., Rebets, Y. & Bechthold, A. (2007). Type II polyketide synthases: gaining a deeper insight into enzymatic teamwork. Nat. Prod. Rep. 24, 162–190. 5. Metsä-Ketelä, M., Niemi, J., Mäntsälä, P. & Schneider, G. (2008). Anthracycline biosynthesis: genes, enzymes and mechanisms. Top. Curr. Chem. 282, 101–140. 6. Niemi, J. & Mäntsälä, P. (1995). Nucleotide sequences and expression of genes from Streptomyces purpurascens that cause the production of new anthracyclines in Streptomyces galilaeus. J. Bacteriol. 177, 2942–2945. 7. Niemi, J., Ylihonko, K., Hakala, Pärssinen, R., Kopio, A. & Mäntsälä, P. (1994). Hybrid anthracycline antibiotics: production of new anthracyclines by cloned genes from Streptomyces purpurascens in Streptomyces galilaeus. Microbiology, 140, 1351–1358. 8. Sultana, A., Kallio, K., Jansson, A., Wang, J. S., Niem, J., Mäntsälä, P. & Schneider, G. (2004). Structure of the polyketide cyclase SnoaL reveals a novel mechanism for enzymatic aldol condensation. EMBO J. 23, 1911–1921. 9. Kallio, P., Sultana, A., Niemi, J., Mäntsälä, P. & Schneider, G. (2006). Crystal structure of the polyketide cyclase AknH with bound substrate and product analogue: implications for catalytic mechanism and product stereoselectivity. J. Mol. Biol. 357, 210–220. 10. Wang, Y., Niemi, J., Airas, K., Ylihonko, K., Hakala, J. & Mäntsälä, P. (2000). Modifications of aclacinomycin T by aclacinomycin methyl esterase (RdmC) and aclacinomycin-10-hydroxylase (RdmB) from Streptomyces purpurascens. Biochim. Biophys. Acta, 1480, 191–200. 11. Jansson, A., Niemi, J., Mäntsälä, P. & Schneider, G. (2003). Crystal structure of aclacinomycin methylesterase with bound product analogues: implications for anthracycline recognition and mechanism. J. Biol. Chem. 278, 39006–39013. 12. Jansson, A., Koskiniemi, H., Erola, A., Wang, J., Mäntsälä, P., Schneider, G. & Niemi, J. (2005). Aclacinomycin-10-hydroxylase is a novel substrateassisted hydroxylase requiring S-adenosyl-L-methionine as cofactor. J. Biol. Chem. 280, 3636–3644. 13. Niemi, J., Wang, Y., Airas, K., Ylihonko, K., Hakala, J. & Mäntsälä, P. (1999). Characterization of aklavinone-11hydroxylase from Streptomyces purpurascens. Biochim. Biophys. Acta, 1430, 57–64. 14. Hong, Y.-S., Hwang, C. K., Hong, S.-K., Kim, Y. H. & Lee, J. J. (1994). Molecular cloning and characterization of the aklavinone 11-hydroxylase gene of Streptomyces peucetius subsp. caesius ATCC 27952. J. Bacteriol. 176, 7096–7101. 15. Filippini, S., Solinas, M. M., Breme, U., Schlüter, M. B., Gabellini, D., Biamonti, G. et al. (1995). Streptomyces peucetius daunorubicin biosynthesis gene, dnrF: sequence and heterologous expression. Microbiology, 141, 1007–1016. 16. Mattevi, A. (1998). The PHBH fold: not only flavoenzymes. Biophys. Chem. 70, 217–222. 17. Entsch, B., Cole, L. J. & Ballou, D. P. (2005). Protein dynamics and electrostatics in the function of phydroxybenzoate hydroxylase. Arch. Biochem. Biophys. 433, 297–311.
Aklavinone-11-Hydroxylase Ternary Complex 18. Schreuder, H. A., Prick, P. A., Wierenga, R. K., Vriend, G., Wilson, K. S., Hol, W. G. & Drenth, J. (1989). Crystal structure of the p-hydroxybenzoate hydroxylase–substrate complex refined at 1.9 Å resolution. Analysis of the enzyme–substrate and enzyme– product complexes. J. Mol. Biol. 208, 679–696. 19. Schreuder, H. A., Mattevi, A., Obmolova, G., Kalk, K. H., Hol, W. G., van der Bolt, F. J. & van Berkel, W. J. (1994). Crystal structures of wild-type phydroxybenzoate hydroxylase complexed with 4aminobenzoate, 2,4-dihydroxybenzoate, and 2-hydroxy-4-aminobenzoate and of the Tyr222Ala mutant complexes with 2-hydroxy-4-aminobenzoate. Evidence for a proton channel and a new binding mode of the flavin ring. Biochemistry, 33, 10161–10170. 20. Gatti, D. L., Palfey, B. A., Lah, M. S., Entsch, B., Massey, V., Ballou, D. P. & Ludwig, M. L. (1994). The mobile flavin of 4-OH benzoate hydroxylase. Science, 266, 110–114. 21. Enroth, C., Neujahr, H., Schneider, G. & Lindqvist, Y. (1998). The crystal structure of phenol hydroxylase in complex with FAD and phenol provides evidence for a concerted conformational change in the enzyme and its cofactor during catalysis. Structure, 6, 605–617. 22. Enroth, C. (2003). High-resolution structure of phenol hydroxylase and correction of sequence errors. Acta Crystallogr., Sect. D: Biol. Crystallogr. 59, 1597–1602. 23. Hiromoto, T., Fujiwara, S., Hosokawa, K. & Yamaguchi, H. (2006). Crystal structure of 3-hydroxybenxoate hydroxylase from Comamonas testosteroni has a large tunnel for substrate and oxygen access to the active site. J. Mol. Biol. 364, 878–896. 24. Koskiniemi, H., Metsä-Ketelä, M., Dobritzsch, D., Kallio, P., Korhonen, H., Mäntsälä, P. et al. (2007). Crystal structures of two aromatic hydroxylases involved in the early tailoring steps of angucycline biosynthesis. J. Mol. Biol. 372, 633–648. 25. Ryan, K. S., Howard-Jones, A. R., Hamill, M. J., Elliott, S. J., Walsh, C. T. & Drennan, C. L. (2007). Crystallographic trapping in the rebeccamycin biosynthetic enzyme RebC. Proc. Natl Acad. Sci. USA, 104, 15311–15316. 26. Ryan, K. S., Chakraborty, C., Howard-Jones, A. R., Walsh, C. T., Ballou, D. P. & Drennan, C. L. (2008). The FAD cofactor of RebC shifts to an IN conformation upon flavin reduction. Biochemistry, 47, 13506–13513. 27. Xu, D., Ballou, D. P. & Massey, V. (2001). Studies of the mechanism of phenol hydroxylase: mutants Tyr289Phe, Asp54Asn and Arg281Met. Biochemistry, 40, 12369–12378. 28. Xu, D., Enroth, C., Lindqvist, Y., Ballou, D. P. & Massey, V. (2002). Studies of the mechanism of phenol hydroxylase: effect of mutation of proline 364 to serine. Biochemistry, 41, 13627–13636. 29. Beam, M. P., Bosserman, M. A., Noinaj, N., Wehenkel, M. & Rohr, J. (2009). Crystal structure of Bayer– Villiger monooxygenase MtmOIV, the key enzyme of the mithramycin biosynthetic pathway. Biochemistry, 48, 4476–4487. 30. Holm, L. & Sander, C. (1994). Searching protein structure databases has come of age. Proteins, 19, 165–173. 31. Wierenga, R. K., Terpstra, P. & Hol, W. G. J. (1986). Prediction of the occurrence of the ADP-binding βαβfold in proteins, using an amino acid sequence fingerprint. J. Mol. Biol. 187, 101–107. 32. Eppink, M. H. M., Schreuder, H. A. & van Berkel, W. J. H. (1997). Identification of a novel conserved sequence motif in flavoprotein hydroxylases with a
Aklavinone-11-Hydroxylase Ternary Complex
33.
34.
35.
36.
37.
38.
39.
putative dual function in FAD/NAD(P)H binding. Protein Sci. 6, 2454–2458. Vagin, A. A., Richelle, J. & Wodak, S. J. (1999). SFCHECK: a unified set of procedure for evaluating the quality of macromolecular structure-factor data and their agreement with atomic model. Acta Crystallogr., Sect. D: Biol. Crystallogr. 55, 191–205. Collaborative Computational Project, Number 4. (1994). The CCP4 suite: programs for protein crystallography. Acta Crystallogr., Sect. D: Biol. Crystallogr. 50, 760–763. Wang, J., Ortiz-Maldonado, M., Entsch, B., Massey, V., Ballou, D. & Gatti, D. L. (2002). Protein and ligand dynamics in 4-hydroxybenzoate hydroxylase. Proc. Natl Acad. Sci. USA, 99, 608–613. Ortiz-Maldonado, M., Entsch, B. & Ballou, D. P. (2003). Conformational changes combined with charge–transfer interactions are essential for reduction in catalysis by p-hydroxybenzoate hydroxylase. Biochemistry, 42, 11234–11242. Palfey, B. A., Moran, G. R., Entsch, B., Ballou, D. P. & Massey, V. (1999). Substrate recognition by “password” in p-hydroxybenzoate hydroxylase. Biochemistry, 38, 1153–1158. Schreuder, H. A., Hol, W. G. & Drenth, J. (1988). Molecular modeling reveals the possible importance of a carbonyl oxygen binding pocket for the catalytic mechanism of p-hydroxybenzoate hydroxylase. J. Biol. Chem. 263, 3131–3136. Aslanidis, C. & de Jong, P. J. (1990). Ligationindependent cloning of PCR products (LIC-PCR). Nucleic Acids Res. 18, 6069–6074.
977 40. Yeates, T. O. (1997). Detecting and overcoming crystal twinning. Methods Enzymol. 276, 344–358. 41. Wang, J., Kamtekar, S., Berman, A. J. & Steitz, T. A. (2005). Correction of X-ray intensities from single crystals containing lattice-translocation defects. Acta Crystallogr., Sect. D: Biol. Crystallogr. 61, 67–74. 42. McCoy, A. J., Grosse-Kunstleve, R. W., Adams, P. D., Winn, M. D., Storoni, L. C. & Read, R. J. (2007). Phaser crystallographic software. J. Appl. Crystallogr. 40, 658–674. 43. Murshudov, G. N., Vagin, A. A. & Dodson, E. J. (1997). Refinement of macromolecular structures by the maximum-likelihood method. Acta Crystallogr., Sect. D: Biol. Crystallogr. 53, 240–255. 44. Emsley, P. & Cowtan, K. (2004). Coot: model-building tools for molecular graphics. Acta Crystallogr., Sect. D: Biol. Crystallogr. 60, 2126–2132. 45. Laskowski, R. A., MacArthur, M. W., Moss, D. S. & Thornton, J. M. (1993). PROCHECK: a program to check the stereochemical quality of protein structures. J. Appl. Crystallogr. 26, 283–291. 46. Altschul, S. F., Gish, W., Miller, W., Myers, E. W. & Lipman, D. J. (1990). Basic local alignment search tool. J. Mol. Biol. 215, 403–410. 47. Thompson, J. D., Higgins, D. G. & Gibson, T. J. (1994). CLUSTAL W: improving the sensitivity of progressive multiple sequence alignment through sequence weighting, position-specific gap penalties and weight matrix choice. Nucleic Acids Res. 22, 4673–4680. 48. DeLano, W. L. (2002). The PyMOL Molecular Graphics System, http://www.pymol.org.