Portal de Periódicos da CAPES

Crystal Structure of Aclacinomycin Methylesterase with Bound Product Analogues

2003; Elsevier BV; Volume: 278; Issue: 40 Linguagem: Inglês

10.1074/jbc.m304008200

1083-351X

Anna Jansson, Jarmo Niemi, Pekka Mäntsälä, G. Schneider,

Enzyme Structure and Function

Aclacinomycin methylesterase (RdmC) is one of the tailoring enzymes that modify the aklavinone skeleton in the biosynthesis of anthracyclines in Streptomyces species. The crystal structures of this enzyme from Streptomyces purpurascens in complex with the product analogues 10-decarboxymethylaclacinomycin T and 10-decarboxymethylaclacinomycin A were determined to nominal resolutions of 1.45 and 1.95 Å, respectively. RdmC is built up of two domains. The larger α/β domain shows the common α/β hydrolase fold, whereas the smaller domain is α-helical. The active site and substrate binding pocket are located at the interface between the two domains. Decarboxymethylaclacinomycin T and decarboxymethylaclacinomycin A bind close to the catalytic triad (Ser102-His276-Asp248) in a hydrophobic pocket, with the sugar moieties located at the surface of the enzyme. The binding of the ligands is dominated by hydrophobic interactions, and specificity appears to be controlled mainly by the shape of the binding pocket rather than through specific hydrogen bonds. Mechanistic key features consistent with the structure of complexes of RdmC with product analogues are Ser102 acting as nucleophile and transition state stabilization by an oxyanion hole formed by the backbone amides of residues Gly32 and Met103. Aclacinomycin methylesterase (RdmC) is one of the tailoring enzymes that modify the aklavinone skeleton in the biosynthesis of anthracyclines in Streptomyces species. The crystal structures of this enzyme from Streptomyces purpurascens in complex with the product analogues 10-decarboxymethylaclacinomycin T and 10-decarboxymethylaclacinomycin A were determined to nominal resolutions of 1.45 and 1.95 Å, respectively. RdmC is built up of two domains. The larger α/β domain shows the common α/β hydrolase fold, whereas the smaller domain is α-helical. The active site and substrate binding pocket are located at the interface between the two domains. Decarboxymethylaclacinomycin T and decarboxymethylaclacinomycin A bind close to the catalytic triad (Ser102-His276-Asp248) in a hydrophobic pocket, with the sugar moieties located at the surface of the enzyme. The binding of the ligands is dominated by hydrophobic interactions, and specificity appears to be controlled mainly by the shape of the binding pocket rather than through specific hydrogen bonds. Mechanistic key features consistent with the structure of complexes of RdmC with product analogues are Ser102 acting as nucleophile and transition state stabilization by an oxyanion hole formed by the backbone amides of residues Gly32 and Met103. Anthracyclines are aromatic polyketide antibiotics that are synthesized by Streptomyces species. They were recognized as early as 1939 for their antibacterial properties (1Krassilnikov N.A. Koreniako A.J. Mikrobiologiya. 1939; : 673-675Google Scholar) and have been used as important anti-tumor drugs in chemotherapy for more than three decades (2di Marco A. Gaetani M. Orezzi P. Scarpinato B.M. Silvestrini R. Soldati M. Dasdia T. Valentini L. Nature. 1964; 201: 706-707Crossref PubMed Scopus (169) Google Scholar). Their mode of action is not yet fully understood, but there is evidence that their cytotoxic activity is at least in part because of inhibition of topoisomerase II (3Booser D.J. Hortobagyi G.N. Drugs. 1994; 47: 223-258Crossref PubMed Scopus (199) Google Scholar). However, anthracyclines exhibit severe cardiotoxic side effects that limit their long-term use in the clinic (4Myers C.E. Mimnaugh G.C. Yeh G.C. Sinha B.K. Lown J.W. Anthracycline and Anthracenedione-based Anticancer Agents. Elsevier, Amsterdam1988: 527-569Google Scholar). Because of their complex composition and structure, these compounds are difficult to synthesize chemically. The search for novel derivatives with less toxic side effects has therefore been limited so far to the hybrid antibiotic approach that rests on gene transfer of a particular set of genes from one Streptomyces species to another, resulting in production of novel anthracycline metabolites (5Niemi J. Ylihonko K. Hakala J. Pärssinen R. Kopio A. Mäntsälä P. Microbiology. 1994; 140: 1351-1358Crossref PubMed Scopus (50) Google Scholar, 6Hopwood D.A. Malpartida F. Kieser H.M. Ikeda H. Duncan J. Fujii I. Rudd B.A. Floss H.G. Omura S. Nature. 1985; 314: 642-644Crossref PubMed Scopus (265) Google Scholar). Alternative routes to non-natural anthracyclines could be based on knowledge of the three-dimensional structures of the enzymes involved in the biosynthesis of these metabolites. Mechanistic and structural information can be fed into the redesign of these enzymes, which may be used for the biosynthesis of anthracyclines with improved or novel toxicological activities. Most anthracyclines consist of a 7,8,9,10-tetrahydrotetracene-5,12-quinone skeleton, which is glycosylated at position C-7 or C-10 (Fig. 1A). Variations in the modifications of the aglycone skeleton or the type of carbohydrate attached lead to considerable chemical diversity, and several hundreds of these natural products are known to date (7Hutchinson C.R. Chem. Rev. 1997; 97: 2525-2535Crossref PubMed Scopus (119) Google Scholar). Biosynthesis starts with the polyketide pathway, where the carbon chain is built up by repeated Claisen condensations from acyl and malonate carbonyl units by the large polyketide synthase complex (8Hopwood D.A. Chem. Rev. 1997; 97: 2465-2497Crossref PubMed Scopus (621) Google Scholar). These condensation steps are followed by a final cyclization to give the aglycone polyketide skeleton (9Ye J. Dickens M.L. Plater R. Li Y. Lawrence J. Strohl W.R. J. Bacteriol. 1994; 176: 6270-6280Crossref PubMed Google Scholar). The cyclized product is then further modified by a series of tailoring enzymes leading to, for example, glycosylation, hydroxylation, methylester hydrolysis, and decarboxylation (10Connors N.C. Bartel P.L. Strohl W.R. J. Gen. Microbiol. 1990; 136: 1895-1898Crossref Scopus (20) Google Scholar, 11Niemi J. Wang Y. Airas K. Ylihonko K. Hakala J. Mantsala P. Biochim. Biophys. Acta. 1999; 1430: 57-64Crossref PubMed Scopus (17) Google Scholar, 12Dickens M.L. Priestley N.D. Strohl W.R. J. Bacteriol. 1997; 179: 2641-2650Crossref PubMed Google Scholar, 13Madduri K. Torti F. Colombo A.L. Hutchinson C.R. J. Bacteriol. 1993; 175: 3900-3904Crossref PubMed Google Scholar, 14Wang Y. Niemi J. Airas K. Ylihonko K. Hakala J. Mäntsälä P. Biochim. Biophys. Acta. 2000; 1480: 191-200Crossref PubMed Scopus (18) Google Scholar, 15Walczak R.J. Dickens M.L. Priestley N.D. Strohl W.R. J. Bacteriol. 1999; 181: 298-304Crossref PubMed Google Scholar). In Streptomyces purpurascens, the rdm operon contains several genes that code for tailoring enzymes of rhodomycin biosynthesis (5Niemi J. Ylihonko K. Hakala J. Pärssinen R. Kopio A. Mäntsälä P. Microbiology. 1994; 140: 1351-1358Crossref PubMed Scopus (50) Google Scholar, 16Niemi J. Mäntsälä P. J. Bacteriol. 1995; 177: 2942-2945Crossref PubMed Google Scholar). Enzymes encoded by this gene cluster include RdmA, an aklanonic acid methylester cyclase, and aklavinone-11-hydroxylase RdmE, which converts aklavinone to ϵ-rhodomycinone (11Niemi J. Wang Y. Airas K. Ylihonko K. Hakala J. Mantsala P. Biochim. Biophys. Acta. 1999; 1430: 57-64Crossref PubMed Scopus (17) Google Scholar). Genetic evidence suggests that rdmB and rdmC code for two enzymes that catalyze the steps leading from ϵ-rhodomycinone glycoside to β-rhodomycinone glycoside (16Niemi J. Mäntsälä P. J. Bacteriol. 1995; 177: 2942-2945Crossref PubMed Google Scholar). Both in vivo and in vitro experiments have shown that these two enzymes can also catalyze the conversion of aclacinomycin to 10-hydroxy-10-decarboxymethylaklavin (14Wang Y. Niemi J. Airas K. Ylihonko K. Hakala J. Mäntsälä P. Biochim. Biophys. Acta. 2000; 1480: 191-200Crossref PubMed Scopus (18) Google Scholar). Aclacinomycin methyl esterase (RdmC) 1The abbreviations used are: RdmC, aclacinomycin methylesterase; AcmT/A, aclacinomycin T/A; DcmaT/A 10-decarboxymethylaclacinomycin T/A; PEG, polyethylene glycol.1The abbreviations used are: RdmC, aclacinomycin methylesterase; AcmT/A, aclacinomycin T/A; DcmaT/A 10-decarboxymethylaclacinomycin T/A; PEG, polyethylene glycol. (14Wang Y. Niemi J. Airas K. Ylihonko K. Hakala J. Mäntsälä P. Biochim. Biophys. Acta. 2000; 1480: 191-200Crossref PubMed Scopus (18) Google Scholar) catalyzes removal of the methoxy group at the C-15 position of aclacinomycin T and aclacinomycin A (Fig. 1). The polypeptide chain contains 298 amino acids, and the enzyme is monomeric in solution (14Wang Y. Niemi J. Airas K. Ylihonko K. Hakala J. Mäntsälä P. Biochim. Biophys. Acta. 2000; 1480: 191-200Crossref PubMed Scopus (18) Google Scholar). RdmC shows about 50% sequence identity to related enzymes from other Streptomyces species, for instance DnrP from Streptomyces peucetius and DauP from Streptomyces sp strain C5 (17Madduri K. Hutchinson C.R. J. Bacteriol. 1995; 177: 3879-3884Crossref PubMed Google Scholar, 18Dickens M.L. Ye J. Strohl W.R. J. Bacteriol. 1995; 177: 536-543Crossref PubMed Google Scholar). These enzymes catalyze similar reactions, hydrolyzing 10-carbomethoxy-13-deoxycarminomycin to 10-carboxy-13-deoxycarminomycin (12Dickens M.L. Priestley N.D. Strohl W.R. J. Bacteriol. 1997; 179: 2641-2650Crossref PubMed Google Scholar, 17Madduri K. Hutchinson C.R. J. Bacteriol. 1995; 177: 3879-3884Crossref PubMed Google Scholar). They all contain the same sequence motif G-X-S-X-G, typical for a serine hydrolase active site (18Dickens M.L. Ye J. Strohl W.R. J. Bacteriol. 1995; 177: 536-543Crossref PubMed Google Scholar). Here we report the crystal structures of RdmC with the bound product analogues 10-decarboxymethylaclacinomycin T (DcmaT) and 10-decarboxymethylaclacinomycin A (DcmaA) at molecular resolution. The enzyme contains the α/β hydrolase fold with a catalytic triad common to this enzyme family. Modeling of the substrate complex based on these structures suggests that the catalytic serine residue is perfectly located to act as nucleophile during hydrolysis and that the negative charge developing in the transition state and the tetrahedral intermediate could be stabilized by an oxyanion hole consisting of the main chain nitrogens of residues Gly32 and Met103. Protein Purification and Crystallization—RdmC from S. purpurascens was expressed in Streptomyces lividans and purified as previously described (11Niemi J. Wang Y. Airas K. Ylihonko K. Hakala J. Mantsala P. Biochim. Biophys. Acta. 1999; 1430: 57-64Crossref PubMed Scopus (17) Google Scholar). AcmT was produced from AcmA (Calbiochem) by partial hydrolysis as previously described (5Niemi J. Ylihonko K. Hakala J. Pärssinen R. Kopio A. Mäntsälä P. Microbiology. 1994; 140: 1351-1358Crossref PubMed Scopus (50) Google Scholar). Crystallization of the enzyme complexes was achieved by adding 1.2 mm AcmA or 1.2 mm AcmT, respectively, to 10 mg/ml protein solution in 50 mm TRIS-HCl buffer containing 1 mm dithiothreitol. The complex was crystallized by vapor diffusion in hanging drops with ammonium sulfate as precipitant at pH 7.5. Details of the crystallization protocol are described elsewhere (19Jansson A. Niemi J. Mäntsälä P. Schneider G. Acta Crystallogr. Sect. D Biol. Crystallogr. 2003; 59: 1637-1639Crossref PubMed Scopus (5) Google Scholar). An isomorphous platinum derivative was obtained by soaking the native crystals for 45 min in 5 mm K2PtNO3 dissolved in water. X-ray Data Collection and Processing—Native diffraction data for the AcmT and AcmA complexes were collected at the European Molecular Biology Laboratory (EMBL), beamline X11, Deutsches Elektronen Synchrotron (DESY), Hamburg to a resolution of 1.45 and 1.95 Å, respectively. The data were collected at cryogenic temperature using 20% PEG400 as a cryoprotectant and processed with DENZO and SCALEPACK (20Otwinowski Z. Sawyer L. Isaacs N. Bailey S. Proceedings of the CCP4 Study Weekend: Data Collection and Processing. Daresbury Laboratory, Warrington, UK1993: 56-62Google Scholar). The crystals were monoclinic, space group P21, with cell dimensions a = 38.2 Å, b = 84.7 Å, c = 44.3 Å, and β = 99.9°. The derivative data were collected at beamline B711, MAX-laboratory, Lund, Sweden to a resolution of 2.25 Å and processed with MOSFLM (21Leslie A. Joint CCP4 and ESF-EACBM Newsletter on Protein Crystallography, 26. Daresbury Laboratory, Warrington, UK1992Google Scholar) and SCALA (22Collaborative Computational Project 4Acta Crystallogr. Sect. D Biol. Crystallogr. 1994; 50: 760-763Crossref PubMed Scopus (19665) Google Scholar). Statistics of the data collection are given in Table I.Table IData collection statistics Values in parenthesis are for the highest resolution shell.AcmT complexAcmA complexK2PtNO3Resolution (Å)20.0-1.4530.0-1.9530.0-2.25Wavelength (Å)0.80940.81101.098No. of observations2347289594587630No. of unique reflections347331963613055R sym (%)aRsym=ΣhklΣi|Ii-〈I〉|/ΣhklΣi〈I〉 , where I i is the intensity measurements for a reflection and 〈I〉 is the mean value for this reflection.4.1 (27.3)10.6 (41.3)3.9 (7.9)Completeness (%)84.8 (63.5)99.9 (99.5)98.4 (95.2)I/σ (I)15.1 (2.0)15.0 (3.3)23.1 (14.6)a Rsym=ΣhklΣi|Ii-〈I〉|/ΣhklΣi〈I〉 , where I i is the intensity measurements for a reflection and 〈I〉 is the mean value for this reflection. Open table in a new tab Structure Determination and Model Building—For structure determination, single isomorphous replacement with anomalous scattering was used, based on the K2PtNO3 derivative (Table II). The heavy atom positions were determined using the program SOLVE (23Terwilliger T.C. Berendzen J. Acta Crystallogr. Sect. D Biol. Crystallogr. 1999; 55: 849-861Crossref PubMed Scopus (3216) Google Scholar). RESOLVE (24Terwilliger T. Acta Crystallogr. Sect. D Biol. Crystallogr. 2000; 56: 965-972Crossref PubMed Scopus (1628) Google Scholar) was used for density modifications to a resolution of 2.5 Å. Because of the high resolution of the native data set, the automatic model building program ARP/wARP (25Perrakis A. Morris R. Lamzin V.S. Nat. Struct. Biol. 1999; 6: 458-463Crossref PubMed Scopus (2561) Google Scholar) could be used to trace 260 of the 298 residues of the protein. Model building was performed with the program O (26Jones T.A. Zou J. Cowan S. Kjeldgaard M. Acta Crystallogr. Sect. A. 1991; 47: 110-119Crossref PubMed Scopus (12999) Google Scholar). Examination of the refined model showed that the platinum sites were all located adjacent to the imidazole side chains of histidine residues.Table IIPhasing statisticsDerivativeK2PtNO3Number of sites5R der (%)*aRder=Σ|FPH-FP|/Σ|FP| where F PH is the structure factor amplitude of the derivative crystal and F P that of the native.22.3R cullis*bRcullis=Σ||FPH±FP|-|FH(calc)|/Σ|FPH-FP| , where F PH and F P are defined as above and F H(calc) is the calculated heavy atom structure factor amplitude summed over centric reflections.0.53Phasing power*cPhasing power = F(H)/E, the root mean square heavy atom structure factor amplitude divided by the lack of closure error.3.17Figure of merit0.76a Rder=Σ|FPH-FP|/Σ|FP| where F PH is the structure factor amplitude of the derivative crystal and F P that of the native.b Rcullis=Σ||FPH±FP|-|FH(calc)|/Σ|FPH-FP| , where F PH and F P are defined as above and F H(calc) is the calculated heavy atom structure factor amplitude summed over centric reflections.c Phasing power = F(H)/E, the root mean square heavy atom structure factor amplitude divided by the lack of closure error. Open table in a new tab Refinement—Initial refinement was performed with CNS (27Brünger A.T. Adams P.D. Clore G.M. DeLano W.L. Gros P. Grosse-Kunstleve R.W. Jiang J.-S. Kuszewski J. Nilges M. Pannu N.S. Read R.J. Rice L.M. Simonson T. Warren G.L. Acta Crystallogr. Sect. D Biol. Crystallogr. 1998; 54: 905-921Crossref PubMed Scopus (16915) Google Scholar). 8% of the reflections were set aside for cross-validation by the use of R free (28Brünger A.T. Nature. 1992; 355: 472-475Crossref PubMed Scopus (3836) Google Scholar). The first steps consisted of rigid body refinement and simulated annealing procedures. Subsequent rounds of refinement included energy minimization and B-factor refinement, followed by manual adjustment of the model using O (26Jones T.A. Zou J. Cowan S. Kjeldgaard M. Acta Crystallogr. Sect. A. 1991; 47: 110-119Crossref PubMed Scopus (12999) Google Scholar). The bound ligand could be fitted straightforwardly into the electron density map, and water molecules were identified using PEAKMAX and WATPEAK (22Collaborative Computational Project 4Acta Crystallogr. Sect. D Biol. Crystallogr. 1994; 50: 760-763Crossref PubMed Scopus (19665) Google Scholar). The final cycles of refinement using REFMAC (29Murshudov G. Vagin A.A. Dodson E.J. Acta Crystallogr. Sect. D Biol. Crystallogr. 1997; 53: 240-253Crossref PubMed Scopus (13710) Google Scholar) resulted in a well-defined model with R work of 16.3% and R free of 18.7% (Table III). Individual, isotropic B-factors were used throughout the refinement, because anisotropic B-factors did not lead to a decrease in R free.Table IIIRefinement statisticsDcmaT complexDcmaA complexResolution (Å)1.451.95R work (%)aRwork=Σ||Fobs|-|Fcalc||/Σ|Fobs| ;|F obs| is the observed and|F calc| is the calculated structure factor amplitudes.16.315.7R free (%)bR free was calculated from 8% of measured reflections omitted from refinement.18.718.5Number of amino acids297297Number of atomsProtein22232223Ligands5370Water306297Ion55B-factor (Å2)Wilson plot1221ProteinMainchain11.315.1Sidechain12.816.7Water24.728.0Ligands19.432.8Ion24.328.8Root mean square deviation from ideal geometryBond length (Å)0.0080.008Bond angles (°)1.3311.294Ramachandran plot (%)Residues in most favored regions92.692.6Residues in additional allowed7.07.0Residues in generously allowed0.40.4a Rwork=Σ||Fobs|-|Fcalc||/Σ|Fobs| ;|F obs| is the observed and|F calc| is the calculated structure factor amplitudes.b R free was calculated from 8% of measured reflections omitted from refinement. Open table in a new tab The RdmC·AcmA complex crystallized isomorphously to the RdmC·AcmT complex, and the refined coordinates of the RdmC·AcmT complex were used for the calculation of initial 2 F o-F c and F o-F c difference electron density maps, without contributions of bound AcmT in the structure factor calculation. Refinement of the RdmC·AcmA complex followed the protocol outlined above except that REFMAC (29Murshudov G. Vagin A.A. Dodson E.J. Acta Crystallogr. Sect. D Biol. Crystallogr. 1997; 53: 240-253Crossref PubMed Scopus (13710) Google Scholar) was used throughout. The stereochemistry of the final models was analyzed with PROCHECK (30Laskowski R.A. MacArthur M.W. Moss D.S. Thornton J. J. Appl. Crystallogr. 1993; 26: 282-291Crossref Google Scholar). The atomic coordinates and observed structure factor amplitudes have been deposited with the Protein Data Bank, accession codes 1q0z (RdmC·AcmA complex) and 1q0r (RdmC·AcmT complex). Structural Comparisons—BLAST (31Altschul S.F. Miller W. Myers E.W. Lipman D.J. J. Mol. Biol. 1990; 215: 403-410Crossref PubMed Scopus (68313) Google Scholar) was used to search for similar sequences, and sequence alignments were performed with ClustalW (www2.eb.ac.uk/clustalw/). Structure comparisons were done with TOP (32Lu G. J. Appl. Crystallogr. 2000; 33: 176-183Crossref Scopus (193) Google Scholar) and the lsq option in O (26Jones T.A. Zou J. Cowan S. Kjeldgaard M. Acta Crystallogr. Sect. A. 1991; 47: 110-119Crossref PubMed Scopus (12999) Google Scholar) using default parameters. Figures were made with O (26Jones T.A. Zou J. Cowan S. Kjeldgaard M. Acta Crystallogr. Sect. A. 1991; 47: 110-119Crossref PubMed Scopus (12999) Google Scholar), BOBSCRIPT (33Esnouf R.M. Journal of Mol. Graph. Model. 1997; 15: 133-138Crossref Scopus (1793) Google Scholar), and RENDER (34Meritt E.A. Murphy M.E.P. Acta Crystallogr. Sect. D Biol. Crystallogr. 1994; 50: 869-873Crossref PubMed Scopus (2855) Google Scholar). Quality of the Electron Density Map and the Model—RdmC was crystallized as a complex with AcmA or AcmT, and the three-dimensional structures of these complexes were determined to 1.45 and 1.95 Å, respectively, using single isomorphous replacement, including anomalous scattering based on a platinum derivative as described under "Experimental Procedures." The electron density is well defined for all 298 amino acid residues except for the N-terminal methionine, which is removed by the cloning host (11Niemi J. Wang Y. Airas K. Ylihonko K. Hakala J. Mantsala P. Biochim. Biophys. Acta. 1999; 1430: 57-64Crossref PubMed Scopus (17) Google Scholar), and one loop, comprising residues 166-169, where the side chains are not visible. This presumably reflects disorder of this loop, which is at the surface of the enzyme. The enzyme was crystallized in the presence of substrate, and the electron density maps clearly show a bound ligand in the active site (Fig. 2). However, there is no electron density for the carboxymethyl group at position C-10 of the product. The loss of the C-15 methyl group was expected, because the enzyme is active at the pH of crystallization. However, the product of the RdmC reaction, 15-demethoxyaclacinomycin, is unstable in water and slowly undergoes decarboxylation in solution (35Tanaka H. Yoshioka T. Yoshimoto A. Shimauchi Y. Ishikura T. Takeuchi T. Umezawa H. J. Antibiot. 1983; 36: 601-603Crossref PubMed Scopus (3) Google Scholar). The electron density is consistent with the bound product analogues 10-decarboxymethyl aclacinomycin DcmaA or DcmaT, respectively, and thus suggests that decarboxylation has occurred during the crystallization process. The final models of the protein consist of 297 amino acids, 406 water molecules, one sulfate ion, one PEG400 molecule, and a DcmaA or DcmaT molecule. They also contain three residues with double conformations and Pro254 in cis conformation. The models have an R work of 16.3% (DcmaT) and 15.7% (DcmaA) and an R free of 18.7% (DcmaT) and 18.5% (DcmaA). The overall structures of the polypeptide chain in the two complexes as well as interaction of the bound ligands with the enzyme are very similar. Unless otherwise noted, in the remainder of this report we discuss the structure of the RdmC·DcmaT complex because of its higher resolution and superior electron density maps. Fold—The RdmC monomer, with overall dimensions of ∼35 Å × 40 Å × 45 Å, is built up of two domains. The core of the larger domain (residues 2-132 and 223-298) contains an eight-stranded β-sheet (β1-8) flanked by six helices (A, B, C, D, E, and F) where helix D is a 310 helix (Fig. 3). The sheet is highly twisted and forms a half-barrel, with strand β2 antiparallel to the rest of the sheet and strand order 12435678. The first and the last strands are arranged almost perpendicular to each other. Helices B, C, D, and E are flanking the concave side, and helices A and F are found at the convex side of the β-sheet. A second, α-helical domain is formed by residues 133-222, which extend from β-strand 6 of the α/β domain and fold into five helices, D′1-D′5 (D′5 is a 310 helix). The chain enters the α/β domain again with helix D packing against the β-sheet (Fig. 3). This smaller domain forms a lid over the α/β sheet and participates in the formation of the substrate binding pocket. RdmC belongs to the α/β hydrolase superfamily (36Ollis D.L. Cheah E. Cygler M. Dijkstra B. Frolow F. Franken S.M. Harel M. Remington S.J. Silman I. Schrag J. Sussman J.L. Verschueren K.H.G. Goldman A. Protein Eng. 1992; 5: 197-211Crossref PubMed Scopus (1821) Google Scholar), and despite non-detectable or low sequence homology, structural comparisons show similarities to other members of this family of esterases, hydrolases, proteases, and lipases. A search of the Protein Data Base for structures related to RdmC using the program TOP (32Lu G. J. Appl. Crystallogr. 2000; 33: 176-183Crossref Scopus (193) Google Scholar) revealed that the enzymes structurally most similar to RdmC are haloperoxidases. These enzymes catalyze the halogenation of organic compounds in the presence of halide ions and peroxides such as H2O2 (37Hofmann B. Tolzer S. Pelletier I. Altenbuchner J. van Pee K.H. Hecht H.J. J. Mol. Biol. 1998; 279: 889-900Crossref PubMed Scopus (143) Google Scholar). Structures of bromoperoxidase from Streptomyces aureofaciens (38Hecht H.J. Sobek H. Haag T. Pfeifer O. van Pee K.H. Nat. Struct. Biol. 1994; 1: 532-537Crossref PubMed Scopus (113) Google Scholar) and chloroperoxidases from S. lividans, S. aureofaciens and Pseudomonas fluorescens (37Hofmann B. Tolzer S. Pelletier I. Altenbuchner J. van Pee K.H. Hecht H.J. J. Mol. Biol. 1998; 279: 889-900Crossref PubMed Scopus (143) Google Scholar) are known. They all show ∼22% sequence identity to RdmC, and their three-dimensional structures superimpose well for their α/β domains. Superposition of bromoperoxidase with RdmC gives 176 equivalent residues with a root mean square deviation of 1.73 Å; a similar value is obtained for the structural alignment of chloroperoxidase with RdmC (159 equivalent residues with a root mean square deviation of 1.63 Å). However, the haloperoxidases differ from RdmC in the structure of the helical domain in the number and orientation of the D′-helices. A search of the Protein Data Base with only the helical domain (residues 133-222) of RdmC did not result in any other similar structures except uteroglobin, which had been noted earlier (39Callebaut I. Poupon A. Bally R. Demaret J.P. Housset D. Delettre J. Hossenlopp P. Mornon J.P. Ann. N. Y. Acad. Sci. 2000; 923: 90-112Crossref PubMed Scopus (31) Google Scholar). Active Site and Substrate Binding—The active site and substrate binding pocket of RdmC is located at the interface between the two domains. Residues from helices D′1, D′2, D′5, β6, and αC form the predominantly hydrophobic substrate binding pocket, and the catalytic triad is located at the end of this deep pocket (Fig. 3A). Ser102 is located in a strand-turn-helix motif at the so-called nucleophilic elbow and is a central residue in the very sharp turn between strand five and helix C. The nucleophilic elbow is the most conserved structure within the α/β hydrolase fold and includes the conserved motif G-X-S-X-G (40Heikinheimo P. Goldman A. Jeffries C. Ollis D.L. Struct. Fold Des. 1999; 7: R141-R146Abstract Full Text Full Text PDF Scopus (219) Google Scholar). Asp248 is found in a loop between strand seven and helix E (in some enzymes of the α/β hydrolase family, it could also be found after strand six; Ref. 36Ollis D.L. Cheah E. Cygler M. Dijkstra B. Frolow F. Franken S.M. Harel M. Remington S.J. Silman I. Schrag J. Sussman J.L. Verschueren K.H.G. Goldman A. Protein Eng. 1992; 5: 197-211Crossref PubMed Scopus (1821) Google Scholar), and His276 is located between strand eight and helix F. The product analogue, DcmaT, is bound to the enzyme with the hydrophobic aklavinone part extending into this deep pocket and the carbohydrate moiety located at the entrance of the cavity (Fig. 4A). The high resolution of the electron density map allows a precise modeling of the bound product analogue. The electron density is consistent with planar conformations for rings B, C, and D and half-chair conformation of ring A. The configuration in ring A is 7S, 9R with O-7 and O-9 in axial positions and C-13 equatorial. The stereochemistry of the bound aglycone moiety is thus consistent with the x-ray structure of aklavinone (41Arora S.K. J. Antibiot. 1985; 38: 1788-1791Crossref PubMed Scopus (3) Google Scholar). The interactions of the aglycone moiety of DcmaT with enzyme residues are predominantly hydrophobic, which is not surprising given the aromatic, hydrophobic character of this molecule. Residues Met103, Thr106, Leu126, Phe134, Gly127, Gly128, Phe158, Leu224, Ile250, and Ala251 of the hydrophobic binding pocket are within van der Waals distance (3.9 Å) of the aglycone moiety (Fig. 4B). The side chain of Phe134 forms a stacking interaction with ring C of DcmaT. Of the six possible hydrogen bond donors and/or acceptors of the aglycone moiety, only one is actually involved in such an interaction. The oxygen atom O-4 of the aglycone skeleton forms a hydrogen bond to a bound water molecule, which in turn is within hydrogen bond distance to the main chain oxygen atom of Ile132. In addition to van der Waals interactions of residues Met103, His219, Tyr220, and Leu222 with the primary amino sugar, a hydrogen bond is formed between the nitrogen group (N3*) of the sugar moiety to the side chain oxygen atoms of Asp135. The presence of only one hydrogen bond between enzyme and the aglycone moiety suggests that hydrogen bonds are not the major determinant of substrate binding in RdmC. Substrate recognition therefore appears to be controlled predominantly by the shape of the binding pocket, defined by the lining hydrophobic amino acid residues. The importance of the shape and hydrophobicity of the binding pocket for substrate binding has recently been demonstrated for another tailoring enzyme in anthracycline biosynthesis, ActVA-Orf6, a monooxygenase structurally and functionally unrelated to RdmC (42Sciara G. Kendrew S.G. Miele A.E. Marsh N.G. Federice L. Malatesta F. Schimperna G. Savino C. Vallone B. EMBO J. 2003; 22: 205-213Crossref PubMed Scopus (138) Google Scholar). In the RdmC·DcmaA complex, the density for the aglycone and the first carbohydrate moiety is well defined in the electron density map. This part of DcmaA binds to RdmC in a manner indistinguishable from that of DcmaT. The additional carbohydrate moieties of DcmaA extend out of the binding pocket into the bulk solution (Fig. 4A) and form only one interaction with an enzyme residue, a hydrogen bond between the oxygen atom O-12 of the second sugar unit and the main chain oxygen atom of Tyr220. As a result, these two carbohydrate units are less well defined in the electron density map. This mode of binding, on the other hand, explains why both compounds are recognized as substrates by RdmC and why there is only modest discrimination between the mono- and triglycosylated anthracyclines (43Wang Y. Niemi J. Mäntsälä P. FEMS Microbiol. Lett. 2002; 208: 117-122Crossref PubMed Google Scholar). Polyethylene Glycol Binding—Strong difference density was found in a hydrophobic pocket adjacent to the active site. The shape and size of the electron density suggested a bound polyethylene glycol molecule consisting of 10 carbons and 6 oxygens (C10H20O6) (Fig. 2). PEG400 had been used in the crystallization liquor and was also used as cryoprotectant. The electron density suggests that a fraction of the bound polyethylene glycol molecules is covalently attached via the terminal aldehyde group to the catalytic Ser102, thereby mimicking the covalent acyl-enzyme complex. Refinement of such a model, however, shows that the fraction of the covalently linked PEG molecules is rather small. The polyethylene molecule forms hydrogen bonds to the side chain oxygens of Tyr198 and Glu202 and two water molecules. The PEG binding pocket is accessible from the outside solution via a small opening (Fig. 4A) lined mostly by hydrophobic residues such as Val212, Pro216, Gly63, Ala209, Thr62, Leu213, and Ile206. A similar pocket is found both in chloroperoxidase and bromoxidase, but in these enzymes the pocket is smaller in size. It has been suggested that this pocket may be used for transport of reactants/products such as peroxides, halides, and water molecules in bromoperoxidase (38Hecht H.J. Sobek H. Haag T. Pfeifer O. van Pee K.H. Nat. Struct. Biol. 1994; 1: 532-537Crossref PubMed Scopus (113) Google Scholar) and may be utilized in a similar manner in RdmC. Mechanistic Implications—RdmC is an esterase that removes the methoxy side chain at carbon 15 of AcmA or AcmT. In the structures of both complexes, obtained at pH 7.5, we observe the demethylated and decarboxylated product analogues, DcmaA and DcmaT, respectively. These structures allow the modeling of the AcmT substrate into the active site pocket, thus providing a picture of the enzyme-substrate interactions before the reaction takes place (Fig. 5). In this model, the side chain of the nucleophile, Ser102, is placed about 3 Å away from the C-15 carbon atom of the carboxymethyl group of AcmT. The side chain of His276 is within hydrogen bonding distance to the hydroxyl group of Ser102, suitable for proton abstraction and activation of the nucleophile. His276, in turn, forms hydrogen bonds to the carboxylate group of Asp248, which can stabilize the positive charge of the imidazolium ring during catalysis. Residues in the turns between strand 5 and helix C and strand 3 and helix A very likely participate in the stabilization of the negative charge building up at the carbonyl oxygen atom of the substrate during turnover. Backbone amides of residues Gly32 and Met103 point into a small cavity that harbors the carbonyl oxygen atom of the bound substrate. The oxygen atom is close to the main chain nitrogen of residue Met103 (4 Å) and forms a hydrogen bond to the main chain nitrogen of Gly32. The negative charge building up at this carbonyl oxygen atom of the substrate in the transition state could therefore be stabilized by this oxyanion hole, similar to other enzymes of this family (36Ollis D.L. Cheah E. Cygler M. Dijkstra B. Frolow F. Franken S.M. Harel M. Remington S.J. Silman I. Schrag J. Sussman J.L. Verschueren K.H.G. Goldman A. Protein Eng. 1992; 5: 197-211Crossref PubMed Scopus (1821) Google Scholar). The nitrogen atom (Nδ2) of His276 is also close to the oxygen atom of the methyl ester (∼3.5 Å) and is suitably positioned to protonate the leaving methoxy group. There are several water molecules in the vicinity of the catalytic serine residue that could act as nucleophiles in the hydrolysis of the acyl-enzyme intermediate. The three-dimensional structure thus supports a mechanism of ester hydrolysis by RdmC similar to that of related enzymes in the α/β hydrolase family. RdmC Is a Representative of a Class of Esterases in Anthracycline Biosynthesis—The highest sequence similarity of RdmC is to esterases from other Streptomyces species. DnrP from Streptomyces peucetius and DauP from Streptomyces sp. C5 exhibit sequence identities to RdmC of 51 and 53%, respectively (12Dickens M.L. Priestley N.D. Strohl W.R. J. Bacteriol. 1997; 179: 2641-2650Crossref PubMed Google Scholar, 17Madduri K. Hutchinson C.R. J. Bacteriol. 1995; 177: 3879-3884Crossref PubMed Google Scholar). DnrP and DauP have both been proposed to act as 10-carbomethoxy-13-deoxycarminomycin esterases on anthracycline substrates removing the carboxymethyl side chain at the C-10 position of, for example, the rhodomycin D aglycones (DauP) (18Dickens M.L. Ye J. Strohl W.R. J. Bacteriol. 1995; 177: 536-543Crossref PubMed Google Scholar). Rhodomycin D is one of the precursors in the production of daunorubicin and doxorubicin, both clinically important anthracycline chemotherapeutic agents. In the genome of Streptomyces coelicolor (44Bentley S.D. Chater K.F. Cerdeno-Tarraga A.M. Challis G.L. Thomson N.R. James K.D. Harris D.E. Quail M.A. Kieser H. Harper D. Bateman A. Brown S. Chandra G. Chen C.W. Collins M. Cronin A. Fraser A. Goble A. Hidalgo J. Hornsby T. Howarth S. Huang C.H. Kieser T. Larke L. Murphy L. Oliver K. O'Neil S. Rabbinowitsch E. Rajandream M.A. Rutherford K. Rutter S. Seeger K. Saunders D. Sharp S. Squares R. Squares S. Taylor K. Warren T. Wietzorrek A. Woodward J. Barrell B.G. Parkhill J. Hopwood D.A. Nature. 2002; 417: 141-147Crossref PubMed Scopus (2530) Google Scholar), we found a putative hydrolase gene with 40% sequence identity to RdmC. Sequence alignments show that the catalytic triad for these four esterases overlaps well (Fig. 6), including the G-X-S-X-G nucleophile motif. The only difference is seen in S. coelicolor, where the acid is glutamate instead of aspartate. Whereas in the putative hydrolase from S. coelicolor only a few of the substrate binding residues are conserved, DnrP and DauP contain most of the residues involved in binding the substrate in RdmC. Most of the substitutions that do occur are conservative, and the size and the hydrophobic character of the binding pocket is conserved (Fig. 6). The differences in the chemical structure of the substrates of these enzymes are small; the substrates of DauP/DnrP have a primary amino group (NH2) at the N3* position of the primary sugar, whereas AcmT contains a tertiary amino group (N(CH3)2). The natural substrates of both RdmC and DauP/DnrP are presumed to be hydroxylated at the C-11 position of the aglycone. The part of the substrate binding pocket that accommodates the C-11 hydroxyl group is very conserved in these enzymes and lacks polar residues capable of forming a specific hydrogen bond to the hydroxyl group. In conclusion, RdmC is a member of a group of esterases involved in the biosynthesis of anthracyclines in Streptomyces. The crystal structure determination of RdmC provides a structural framework for these enzymes and shows that they are members of the large superfamily of α/β hydrolases. The binding of the substrate is dominated by hydrophobic interactions, and specificity appears to be controlled by the shape of the binding pocket rather than through specific hydrogen bonds. Mechanistic key features suggested by the structure of complexes of RdmC with product analogues are the catalytic Ser-His-Asp triad and transition state stabilization by an oxyanion hole formed by the backbone amides of residues Gly32 and Met103. We gratefully acknowledge the MAX-laboratory, Lund and the European Molecular Biology Laboratory (EMBL) outstation, Deutsches Elektronen Synchrotron (DESY), Hamburg for provision of synchrotron radiation. We especially thank Christoffer Enroth, EMBL, for help using ARP/wARP.